Datasets:

griffin

/

ChemSum

like 11

Manganese complex-catalyzed oxidation and oxidative kinetic resolution of secondary alcohols by hydrogen peroxide

The highly efficient catalytic oxidation and oxidative kinetic resolution (OKR) of secondary alcohols has been achieved using a synthetic manganese catalyst with low loading and hydrogen peroxide as an environmentally benign oxidant in the presence of a small amount of sulfuric acid as an additive. The product yields were high (up to 93%) for alcohol oxidation and the enantioselectivity was excellent (>90% ee) for the OKR of secondary alcohols. Mechanistic studies revealed that alcohol oxidation occurs via hydrogen atom (H-atom) abstraction from an a-CH bond of the alcohol substrate and a twoelectron process by an electrophilic Mn-oxo species. Density functional theory calculations revealed the difference in reaction energy barriers for H-atom abstraction from the a-CH bonds of Rand Senantiomers by a chiral high-valent manganese-oxo complex, supporting the experimental result from the OKR of secondary alcohols.

manganese_complex-catalyzed_oxidation_and_oxidative_kinetic_resolution_of_secondary_alcohols_by_hydr

Introduction<!>Results and discussion<!>Conclusions<!>Materials<!>Determination of the relative rate constants (k rel )<!>Determination of the KIE value<!>Computational details<!>Conflicts of interest

<p>The selective oxidation of organic substrates using earthabundant transition metal catalysts (e.g. manganese and iron) and environmentally benign oxidants (e.g. molecular oxygen and hydrogen peroxide) is fundamentally important in enzymatic/biomimetic reactions and immensely useful in organic synthesis. 1,2 Therefore, tremendous efforts have been devoted to elucidating the biomimetic oxidation reactions and developing highly efficient, selective (asymmetric) oxidation reactions using earth-abundant transition metal catalysts and environmentally benign oxidants under mild conditions. [2][3][4][5][6][7] As a result, a great advance has been achieved recently in catalytic (asymmetric) epoxidation and hydroxylation reactions using synthetic iron and manganese catalysts and aqueous H 2 O 2 as an environmentally benign oxidant in the presence of carboxylic acid as an additive. [2][3][4][5][6] In these reactions, it has been proposed that high-valent metal-oxo intermediates are the active oxidants that affect the (asymmetric) epoxidation and hydroxylation reactions, and that the role of the carboxylic acid is to facilitate the heterolytic O-O bond cleavage of putative metal-hydroperoxo species to form high-valent metal-oxo intermediates. [3][4][5] Very recently, we reported that a manganese complex bearing a tetradentate N4 ligand is an efficient catalyst in the asymmetric epoxidation of olens by aqueous H 2 O 2 in the presence of a small amount of H 2 SO 4 , affording high product yields with excellent stereo-and enantioselectivities. 7 In the latter reaction, it was shown that carboxylic acid can be replaced by H 2 SO 4 for activating H 2 O 2 by manganese complexes, generating highvalent manganese-oxo species as active oxidants, 7 although the role of H 2 SO 4 remains elusive.</p><p>Another important research area in oxidation reactions is the oxidation of alcohols to aldehydes or ketones. 8,9 Recently, nonporphyrinic manganese complexes have been employed as catalysts in the development of efficient catalytic systems for alcohol oxidation reactions, especially in those using H 2 O 2 as an environmentally benign oxidant in the presence of carboxylic acid. 9 Mechanistic studies have been performed to elucidate the alcohol oxidation reactions using synthetic metal-oxo complexes. 10 In alcohol oxidation chemistry, the oxidative kinetic resolution (OKR) of racemic secondary alcohols has attracted much attention for developing efficient catalytic systems to obtain enantio-enriched alcohols, [11][12][13] since chiral secondary alcohols are valuable synthetic intermediates in the pharmaceutical and ne chemical industries. In the OKR of racemic secondary alcohols, chiral metal complexes (e.g. the Mn(III) salen complex) with articial oxidants (e.g. iodobenzene diacetate and sodium hypochlorite) have been used to produce chiral secondary alcohols. 12d,13 However, to the best of our knowledge, the OKR of secondary alcohols has never been explored using synthetic mononuclear manganese catalysts and aqueous H 2 O 2 as the terminal oxidant.</p><p>Herein, we report that manganese complexes bearing tetradentate N4 ligands, such as Mn(II)(P-MCP)(OTf) 2 (1) and Mn(II)(Dbp-MCP)(OTf) 2 (2) (Scheme 1A), 7 are highly efficient catalysts for the oxidation of alcohols by aqueous 30% H 2 O 2 in the presence of a catalytic amount of H 2 SO 4 (Scheme 1B). Furthermore, to the best of our knowledge, the present study reports the rst example of the use of a chiral manganese catalyst with low loading, aqueous H 2 O 2 as the terminal oxidant, and a catalytic amount of H 2 SO 4 as an additive in the OKR of racemic secondary alcohols (Scheme 1C). Density functional theory (DFT) calculations reveal that the energy barriers for the a-CH bond activation of chiral 1-phenylethanol (R-and S-enantiomers) by a high-valent manganese-oxo complex are signicantly different, explaining the experimental observation of high enantioselectivity in the OKR of racemic secondary alcohols.</p><!><p>Firstly, the reaction conditions for the catalytic oxidation of alcohols by manganese complexes and aqueous H 2 O 2 in the presence of H 2 SO 4 were optimized using 1-phenylethanol as a model substrate (see the Experimental section). Among the tested manganese catalysts (see Scheme 1A for the structures), Mn(II)(P-MCP)(OTf) 2 (1) and Mn(II)(Dbp-MCP)(OTf) 2 (2) exhibited high catalytic activity (Table 1, entries 1 and 2), whereas Mn(II)(MCP)(OTf) 2 (3) was a poor catalyst (Table 1, entry 3). In the absence of the manganese catalyst, the oxidation of 1-phenylethanol to acetophenone was not observed (Table 1, entry 4). Since 1 can be prepared easily and cost-effectively, 1 was used as the catalyst to determine the optimal reaction conditions by varying the amount of catalyst (Table 1, entries 5 and 6), H 2 O 2 (Table 1, entries 5 and 7) and H 2 SO 4 (Table 1, entries 7-10). In addition, as reported in the olen epoxidation reactions by nonporphyrinic Mn catalysts and H 2 O 2 , 7 other Brønsted acids, such as HClO 4 , H 3 PO 4 , HCl and CF 3 SO 3 H, turned out to be poor additives (Table 1, entries 11-14).</p><p>Aer optimizing the reaction conditions (Table 1, entry 9), we investigated the substrate scope for the oxidation of secondary alcohols by 1 and H 2 O 2 in the presence of H 2 SO 4 (Table 2). 1-Phenylethanol derivatives with electron-donating and -withdrawing substituents at the para-position of the phenyl group were oxidized to their corresponding ketones with good yields (e.g. >80%) (Table 2, le column). However, the product yields were moderate (e.g. $50%) for the oxidation of 1-phenylethanol derivatives with steric hindrance at the ortho-position of the phenyl group (Table 2, le column). Similarly, increasing the chain length and steric hindrance on the methyl side of the 1phenylethanol derivatives, such as 1-phenylbutan-1-ol, 2-methyl-1-phenylpropan-1-ol and 2,2-dimethyl-1-phenylpropan-1-ol, Scheme 1 (A) Schematic structures of manganese complexes bearing N4 ligands: Mn II (P-MCP)(OTf) 2 (1; P-MCP ¼ (1R,2R)-N,N 0 -dimethyl-N,N 0 -bis-(phenyl-2-pyridinylmethyl)cyclohexane-1,2-diamine and OTf Entry Catalyst (mol%) Additive (mol%) H 2 O 2 (equiv.) Yield (%) decreased the product yields (Table 2, middle column). On the other hand, a series of diphenylmethanol derivatives were converted to the desired products with yields of >80% (Table 2, middle column). Unactivated aliphatic secondary alcohols were also oxidized to their corresponding ketones with moderate to high yields (Table 2, right column).</p><p>We then investigated the OKR of secondary alcohols utilizing the manganese catalyst, H 2 O 2 oxidant and H 2 SO 4 additive system. Firstly, we optimized the reaction conditions by varying the amount of catalyst, oxidant and H 2 SO 4 , and the reaction temperature (see Fig. 1; also see Tables S1 and S2, ESI †). Obviously, the conversion of 1-phenylethanol would increase with an increasing amount of H 2 O 2 , as shown in Fig. 1. Due to the preference for oxidation of the S-enantiomer in the OKR of racemic secondary alcohols using the sulfuric acid-enabled manganese system, the ee value improved when increasing the number of equivalents of H 2 O 2 from 0 to 0.80, while no signicant change occurred when further increasing the number of equivalents of H 2 O 2 up to 1.0. Therefore, the oxidant amount was chosen to be 0.80 equiv. for the OKR of 1-phenylethanol, as a result of the excellent ee with lower conversion. Besides, 2 was chosen as the catalyst since 2 afforded a higher enantiomeric excess (ee) value (90%) than 1 (65%) in the oxidation of 1-phenylethanol (Table 3, entry 1 and footnote c). Under the optimized catalytic conditions, we obtained high ee values (>90% ee) irrespective of the substituents on the phenyl group of the benzylic alcohols (i.e. no effect from steric hindrance or the electronic nature of the substrates) (Table 3, entries 2-6). Importantly, 1-phenylpropan-1-ol derivatives, which were reported to be poor substrates in most manganese salen systems, 12d,13,14 also worked well in this sulfuric acidenabled manganese system, with ee values of >90% (Table 3, entries 7-12). In addition, increasing the steric hindrance and chain length in the 1-phenylpropan-1-ol derivatives did not affect the ee values either (Table 3, entries 13-15).</p><p>In order to gain mechanistic insight into the manganesecatalyzed alcohol oxidation reactions, we rstly investigated the effect of para-substituents on the reactivity of the benzyl alcohol by carrying out competitive alcohol oxidation of the benzyl alcohol against para-substituted benzyl alcohols (see the Experimental section). A good linear correlation was obtained when the k rel values were plotted against the Hammett parameters of the substituents (Fig. 2); the small but negative r value of À0.58 indicates that the active intermediate possesses electrophilic character, as reported in the oxidation of benzyl alcohol derivatives by synthetic metal-oxo complexes. 10 Secondly, when the intermolecular competitive oxidation of 1phenylethanol or its a-deuterated compound (1-deuterated 1phenylethanol) was carried out together with 1-(p-chlorophenyl) ethanol as a mediator, a kinetic isotope effect (KIE) value of 1.8 was obtained (see the Experimental section for the detailed method), suggesting that hydrogen atom (H-atom) abstraction from an a-CH bond may be the rate-determining step, as observed in other manganese complex-catalyzed alcohol oxidation reactions. 8b,c It is also notable that the KIE values determined in C-H bond activation reactions by synthetic metal-oxo complexes under stoichiometric conditions (e.g. KIE values of > 10) 10,15 are much higher than those obtained in metal complex-catalyzed oxidation reactions under catalytic conditions (e.g. KIE values of < 4). 8b,c It would be of interest to understand the reason for the difference in the KIE values obtained from the stoichiometric and catalytic reactions (e.g. the involvement of different metal-oxygen intermediates in the catalytic oxidation reactions). Thirdly, since cyclobutanol has oen been used as a substrate probe to distinguish one-electron and two-electron processes in alcohol oxidation reactions, 10a,c,16 we performed the oxidation of cyclobutanol and found that cyclobutanone was yielded exclusively and the ring-opened Table 2 Substrate scope for the oxidation of secondary alcohols a,b,c a Reaction conditions: a CH 3 CN (0.50 mL) solution containing 30% H 2 O 2 (1.2 equiv.) was added dropwise to a CH 3 CN (1.0 mL) solution containing the substrate (0.50 mmol), 1 (0.30 mol%) and H 2 SO 4 (0.30 mol%), using a syringe pump at 25 C for 1 h. b Yields were determined by GC. c Numbers in parentheses are product yields. d 1.0 mol% H 2 SO 4 was used. product, 4-hydroxylbutyraldehyde, was not detected (Scheme 2). Based on the mechanistic studies discussed above, we conclude that alcohol oxidation by an electrophilic manganese-oxo species is a two-electron process. This reactive manganese-oxo intermediate has been proposed previously in the asymmetric epoxidation of olens by the manganese catalyst (2) and H 2 O 2 in the presence of H 2 SO 4 . 7 Density functional theory (DFT) computations were then carried out to explore the enantioselectivity in the a-CH bond activation of the chiral 1-phenylethanols (both R-and S-enantiomers) by a chiral [(P-MCP)Mn(V)(O)(SO 4 )] + complex (I), which was proposed previously as the reactive intermediate in the reaction of 1 and H 2 O 2 in the presence of H 2 SO 4 . 7 The computational results reveal that H-atom abstraction from the a-CH bond of the S-enantiomer, with an energy barrier of 7.7/ 5.5 kcal mol À1 for the triplet/quintet spin state, is easier than that from the a-CH bond of the R-enantiomer, with an energy barrier of 10.6/6.8 kcal mol À1 for the triplet/quintet spin state (Table S3, ESI †). This reactivity trend was obtained by comparing the geometric character of these two transition states, in which TS S with r C-H ¼ 1.203 Å has a smaller elongation of the C-H bond, while TS R with r C-H ¼ 1.240 Å has a larger elongation of the C-H bond for the quintet ground state (Fig. S32, ESI †). In addition, the energy barrier difference for the two isomers might be due to the non-covalent anion-p interaction between the phenyl group of the substrate and the sulfuric acid anion ligand, which can stabilize the transition state, lowering the energy barrier. This kind of interaction only exists for TS S , with a distance of ca. 3.4 Å between the two groups. Thus, we may conclude that the S-enantiomer is an easier substrate than the R-enantiomer for oxidation by the high-valent Mn-oxo intermediate, regardless of the spin states; therefore the R-enantiomer remains in the reaction solution.</p><p>Based on the Arrhenius equation, an energy barrier difference of 1.3 kcal mol À1 will give a rate constant ratio (k (R) /k (S) ) of 0.112, which corresponds to a high ee value ($80%), as obtained from the experiments (vide supra). From inspection of the spin densities (Table S4, Fig. S32 and S33, ESI †), we can see that for the quintet spin state, the sulfuric acid group has a spin density of ca. 0.5 in the 5 I+ sub species and ca. 0.0 in the transition state. a Reaction conditions: a CH 3 CN (0.50 mL) solution containing 30% H 2 O 2 (0.70-0.90 equiv.) was added dropwise to a CH 3 CN (1.0 mL) solution containing the secondary alcohol (0.50 mmol), 2 (0.20 mol%) and H 2 SO 4 (1.0 mol%), using a syringe pump at 0 C for 1 h. b Conversion yields and ee values were determined by GC with a CP-Chirasil-Dex CB column. c When 1 was used as a catalyst under identical reaction conditions, the conversion yield and ee value were 66% and 65%, respectively. d Conversion yields were calculated from the isolated products and the ee values were determined by HPLC with an IA column. This indicates that the sulfuric acid group should be noninnocent to the reaction, and the non-innocence of the sulfuric acid group may make the reaction on the quintet spin state approachable. In addition, the low energy barrier indicates the high electrophilicity of the high-valent manganese-oxo species, which may originate from the existence of a low lying s* orbital that can accept electronic density from the C-H bond (Fig. S33, ESI †).</p><!><p>In summary, we have reported the rst example of sulfuric acidenabled chemoselective oxidation of secondary alcohols by manganese catalysts and hydrogen peroxide. 17 Secondary alcohols were oxidized to their corresponding ketones with good yields and an efficient OKR of racemic secondary alcohols was achieved with excellent enantioselectivities (>90% ee). Mechanistic studies revealed that the active manganese oxidant possesses electrophilic character, H-atom abstraction from an a-CH bond of the alcohol substrate is the rate-determining step, and alcohol oxidation occurs via a two-electron process. DFT calculations revealed that the difference in reaction energy barriers for H-atom abstraction from the a-CH bonds of the Rand S-enantiomers by a putative high-valent manganese-oxo intermediate is signicant (i.e. 1.3 kcal mol À1 ), affording the enantioselectivity in the OKR of racemic secondary alcohols. Future studies will focus on the improvement of the catalytic activity as well as the enantioselectivity in the OKR of secondary alcohols, using synthetic nonheme iron and related manganese catalysts and environmentally benign oxidants such as molecular oxygen and hydrogen peroxide.</p><!><p>All chemicals were purchased from Aldrich, Alfa Aesar and TCI, which were of the best available purity and were used without further purication unless otherwise indicated. Solvents were dried according to published procedures and distilled under argon prior to use. 18 PhCD(OH)CH 3 was prepared from acetophenone according to the literature method. 19 The ligands, MCP, P-MCP and Dbp-MCP, and their corresponding Mn II complexes, Mn II (MCP)(OTf) 2 , Mn II (P-MCP)(OTf) 2 and Mn II (Dbp-MCP)(OTf) 2 , were prepared according to the reported methods. Typical procedure for the oxidation of 1-phenylethanol All reactions were performed under an Ar atmosphere using a dried solvent and standard Schlenk techniques. The catalyst (0.30 mol%), 1-phenylethanol (0.50 mmol) as a substrate and H 2 SO 4 (0.30 mol%) were added into a Schlenk tube containing CH 3 CN (1.0 mL) at 25 C. Subsequently, a solution of CH 3 CN (0.50 mL) containing 30% H 2 O 2 (1.2 equiv.) was added dropwise using a syringe pump for 1 h. Then, the reaction solution was quenched with NaHCO 3 and Na 2 S 2 O 3 , and n-decane was added into the solution as an internal standard. The yields were determined by GC.</p><p>Typical procedure for the OKR of 1-phenylethanol All reactions were performed under an Ar atmosphere using a dried solvent and standard Schlenk techniques. The catalyst (0.20 mol%), racemic 1-phenylethanol (0.50 mmol) as a substrate and H 2 SO 4 (1.0 mol%) were added into a Schlenk tube containing CH 3 CN (1.0 mL) at 0 C. Subsequently, a solution of CH 3 CN (0.50 mL) containing 30% H 2 O 2 (0.80 equiv.) was added dropwise using a syringe pump for 1 h. Then, the reaction solution was quenched with NaHCO 3 and Na 2 S 2 O 3 , and ndecane was added into the solution as an internal standard. The yields were determined by GC (see Fig. 1 and Table 3; see also Tables S1 and S2, ESI †).</p><!><p>A solution of CH 3 CN (0.50 mL) containing 30% H 2 O 2 (0.40 equiv.) was added dropwise into a CH 3 CN solution (1.0 mL) containing a mixture of 1-phenylethanol (0.50 mmol) and parasubstituted 1-phenylethanol (0.50 mmol), the catalyst (1, 0.30 mol%) and H 2 SO 4 (0.30 mol%) using a syringe pump at 25 C for 1 h. Then, the reaction solution was quenched with NaHCO 3 and Na 2 S 2 O 3 , and n-decane was added into the solution as an internal standard. The yields were determined by GC.</p><p>The k rel values were calculated using eqn (1), 19a</p><p>where [R] i and [R] f are the initial and nal concentrations of para-substituted 1-phenylethanol, respectively, and [H] i and [H] f are the initial and nal concentrations of 1-phenylethanol, respectively.</p><!><p>A solution of CH 3 CN (0.50 mL) containing 30% H 2 O 2 (0.40 equiv.) was added dropwise into a CH 3 CN solution (1.0 mL) containing a mixture of substrate (0.50 mmol; 1-phenylethanol or 1-deuterated 1-phenylethanol) and 1-(4-chlorophenyl)ethanol (0.50 mmol), 1 (0.30 mol%) and H 2 SO 4 (0.30 mol%) using a syringe pump at 25 C for 1 h. Then, the reaction solution was quenched with NaHCO 3 and Na 2 S 2 O 3 , and n-decane was added into the solution as an internal standard. The yields were determined by GC. The KIE values were calculated using eqn (2)-( 4 D] f are the initial and nal concentrations of 1deuterated 1-phenylethanol, respectively.</p><!><p>Density functional theory calculations were performed using Gaussian 09 soware. 21 The high-valent [(P-MCP) Mn 5+ (O 2À )(SO 4 2À )] + species (I) was chosen as the reactive intermediate, and two enantiomers of the chiral 1-phenylethanol (S-and R-enantiomers) were used as the substrates. The spin-unrestricted B3LYP (UB3LYP) functional 22,23 was employed with two basis sets: (1) the LACVP basis set for Mn and the 6-31G* basis set for the rest of the atoms, denoted as B1, were used to optimize the minima and transition states; (2) the LACV3P basis set for Mn and the 6-311+G** basis set for the rest of the atoms, denoted as B2, were used to obtain the single point energy corrections. 24,25 The transition states and optimized minima were ascertained by vibrational frequency analysis with only one and zero imaginary frequencies, respectively. All calculations were performed in acetonitrile solvent using the self-consistent reaction eld (SCRF) in the conductor-like polarizable continuum model (CPCM).</p><!><p>There are no conicts to declare.</p>

Royal Society of Chemistry (RSC)

Development of a universal glycosyltransferase assay amenable to high-throughput formats

The development of a general 1-Zn(II) NDP sensor assay for rapid evaluation of GT activity is described. The 1-Zn(II) NDP sensor assay offers submicromolar sensitivity, compatibility with both purified enzymes and crude cell extracts, and exquisite selectivity for nucleoside diphosphates over the corresponding NDP-sugars. Thus, the 1-Zn(II) NDP sensor assay is anticipated to offer broad applicability in the context of GT engineering and characterization.

development_of_a_universal_glycosyltransferase_assay_amenable_to_high-throughput_formats

<!>Materials and methods<!>Preparation of ligand 1<!>TDP binding assay<!>Enzyme assays<!>Crude cell extract assays<!>OleD Kinetics<!>Results and discussion

<p>Complex carbohydrates are found in a wide range of biomolecules in cells, including polysaccharides, proteoglycans, glycolipids, glycoproteins, and antibodies. They play important roles in a number of biological processes such as cell growth, cell-cell interactions [1], immune response [2], inflammation [3], and viral and parasitic infections [4]. The attachment of carbohydrates to the biomolecules is catalyzed by glycosyltransferases (GTs) which transfer a monosaccharide unit from a nucleotide or lipid sugar donor to acceptor substrates in a regio- and stereospecific manner. Given the importance of carbohydrates in biology and medicine, the development of methods for glycan synthesis and modification remains a major focus of research [5-8].</p><p>While both chemical and enzymatic methods have been developed for glycan synthesis, enzymatic processes are often advantageous due to both their efficiency as well as their stringent regio- and stereochemical control [9, 10]. However, the lack of availability of suitable glycosyltransferases (GTs), and/or the requisite sugar nucleotide donors [11], for targeted glycosyl-bond formation often restricts the alternative application of enzymes. Thus, technologies to enable the generation of tailor-made GTs, either via rational design and/or directed evolution [10, 12], are anticipated to greatly augment the utility of GTs in this regard. Although there are recent examples in which GTs were successfully evolved to modulate their substrate specificity [13-15], in all cases the corresponding assays were developed for a specific acceptor. While other GT assays, including radiochemical assays, immunological assays, pH-based assays, or phosphatase-coupled assays exist [16, 17, 18], each has limits in the context of high throughput screening. In this study, we describe the development of a truly general fluorescence-based GT assay, based upon a xanthene-based Zn(II) complex nucleoside diphosphate chemosensor [19]. Given this 1-Zn(II) NDP sensor assay is highly sensitive, is compatible with both purified enzymes or crude extracts, and relies upon a sensor for the general leaving group of most Leloir-type GT-catalyzed reactions, the assay is anticipated to have broader applicability.</p><!><p>Unless otherwise specified, all chemicals and enzymes were reagent grade or better obtained from Sigma-Aldrich (St. Louis, MO, USA) or Fisher Scientific (Pittsburgh, PA, USA) and were used without further purification. Recombinant Streptomyces antibioticus wild-type oleandomycin glycosyltransferase (OleD) and corresponding mutants (OleD-ASP, OleD-AIP and OleD-TDP16) were produced and purified as described previously [14, 20, 21]. Absorbance readings were performed on a Beckman Coulter DU 800 spectrophotometer (Fullerton, CA, USA) and fluorescence was measured by a BMG Labtech FLUOstar Optima plate reader (microtiter plate scale, Durham, NC, USA). Mass spectrometric data were obtained on either a Waters (Milford, MA) LCT time-of-flight spectrometer for electrospray ionization (ESI) or a Varian ProMALDI (Palo Alto, CA) Fourier transform ion cyclotron resonance mass spectrometer (FTICR) equipped with a 7.0 T actively-shielded superconducting magnet and a Nd:YAG laser.</p><!><p>Ligand 1 was prepared as previously described without modifications starting from orcinol and ethyl orsellinate [19, 22, 23].</p><!><p>Complex 1-2Zn(II) stock solution was prepared by dissolving ligand 1 (2.5 mM) and ZnCl2 (6.3 mM, 2.5 equivalent) in 10 mM HCl. The 1-Zn(II) NDP sensor assay solution was prepared by adding the complex stock solution (10 μL) to the assay buffer (10 mL) containing 50% methanol in 25 mM HEPES (pH 7.4), 10 mM NaCl, 1 mM MgCl2. TDP or TDP-Glc at different concentrations (0.01, 0.02, 0.04, 0.08, 0.16, 0.31, 0.62, 1.25, 2.50, 5.00, 10.00 μM, final concentrations) was added to the assay solution (200 μL, final volume) in a 96 well plate and the fluorescence was measured at 520 nm with excitation at 485 nm. The dissociation constant was obtained by calculating the free TDP concentration at which ΔF/ΔFmax equals 0.5 (ΔF, fluorescence intensity change; ΔFmax, maximum fluorescence intensity change).</p><!><p>Representative GT (wtOleD or OleD variants, 1 μM final concentration) was added to the reaction buffer containing 10 mM Tris (pH 8.0), 1 mM TDP-Glc, 1 mM 4-MU and 1 mM MgCl2, and the mixture incubated at room temperature. For each GT activity determination, an aliquot of the GT reaction mixture (5 μL) was added to the 1-Zn(II) NDP sensor assay solution (195 μL) and the fluorescence was measured at 520 nm as described for the TDP binding assay.</p><p>The corresponding 4-MU fluorescence assay (where 4-MU glycosylation directly correlates to a reduction in 4-MU fluorescence) was conducted as previously described [14, 24]. Briefly, for this study the GT reaction mixture (10 μL) was added to 10 mM Tris (pH 8.0, 990 μL) and the fluorescence was measured at 460 nm with excitation at 355 nm.</p><!><p>Cells from OleD-expressing bacterial cell cultures (25 mL) were harvested by centrifugation (4000 rpm) and frozen at −80 °C. The frozen cell pellets were thawed on ice, resuspended in the lysis buffer (2 mL) containing 50 mM Tris (pH 8.0), lysozyme (1 mg/mL, 50 kU/mL) and benzonaze (125 U/mL, Novagen, San Diego, CA, USA, Cat# 70746-3), and incubated on ice for 1 hour. Removal of the cell debris by centrifugation (12,000 rpm) afforded crude cell extracts. OleD assays with crude cell extracts were carried out as described for the assays with purified enzymes by adding crude cell extracts (1 μL for 100 μL reaction, 1%) instead of purified enzymes to the reaction buffer. A Z factor for the assay containing TDP16 at 100 min was calculated by using the equation, Z = 1 − (3σs + 3σc)/|μs − μs| where σs and σc are denoted for the standard deviations of the sample signal and control signal, and μs and μs for the means of the sample signal and control signal [25].</p><!><p>Kinetics were performed with constant concentrations of OleD-TDP16 (1 μM) and TDP-Glc (1 mM) in 50 mM Tris-HCl (pH 8.0) containing 1 mM MgCl2 while varying 4-MU concentrations (0.05, 0.1, 0.2, 0.4, 0.8 and 1.6 mM). TDP production was assessed using the 1-Zn(II) NDP sensor assay at 20, 60, 180 seconds by fluorescence change at 520 nm. Initial reaction velocities, obtained as the slope of best fit to the initial linear portion of the reaction time course, were subsequently fit to the Michaelis-Menten equation.</p><!><p>Given nearly all LeLoir GT-catalyzed reactions produce NDP as a product, a sensitive NDP sensor would be advantageous for the development of a general GT assay strategy. Among the fluorescence-based NDP sensors that have been developed for biochemical applications [19, 26, 27], the xanthene-based Zn(II) complex [Fig. 1, 1-Zn(II)] offers both high sensitivity and selectivity for NDP over NDP-sugar (the requisite GT substrate). The complex contains two sites of 2,2′-dipicolylamine-Zn(II) and xanthene as a fluorescent sensing unit for nucleoside polyphosphates. This chemosensor selectively senses nucleoside di- or triphosphates with a large fluorescence enhancement (F/Fo > 15) and strong binding affinity (about 1 μM of apparent dissociation constant, K'd), whereas no detectable fluorescence change is induced by monophosphate species, NDP-sugars or various other anions. Therefore we expected the complex 1-Zn(II) could serve as enabling feature for the development of a general NDP sensor-based GT assay.</p><p>The xanthene-based ligand 1 was prepared as previously described starting from the commercially available compounds, orcinol and ethyl orsellinate [19, 22, 23]. To test the feasibility of 1-Zn(II) NDP sensor assay in the context of a GT assay, the well-studied macrolide-inactivating GT from S. antibioticus (OleD) was selected as a model system [14, 20, 21, 28]. As a first step, the binding affinity of 1-Zn(II) to TDP-Glc and TDP, the OleD substrate and product, respectively, was assessed in a 96-well plate format (Fig. 2). As anticipated, the large fluorescence increase at 520 nm directly correlated with an increase of [TDP] while increasing [TDP-Glc] had no effect. Thus, this standard analysis confirmed the complex to provide submicromolar sensitivity and cleanly distinguish between NDP and NDP-sugar [19], providing the selectivity (TDP K'd = 0.44 μM) and sensitivity required for a general GT assay.</p><p>Next, the 1-Zn(II) NDP sensor assay was applied to an in vitro GT assay containing purified enzymes. OleD-WT and three GT variants which display different proficiencies (ASP, AIP, TDP16 [14, 20, 21]) were employed as a representative GT series with TDP-Glc and 4-methylumbelliferone (4-MU) serving as the glycosyl donor and acceptor, respectively. Notably, the established order of 4-MU/TDP-Glc turnover across this series was TDP16 > ASP > AIP with no conversion expected using WT [14, 20, 21]. Consistent with this, an enzyme-dependent and time-dependent increase of fluorescence was observed which directly correlates to the variant efficiency of NDP production (and corresponding glucosyltransfer) among the series of reactions evaluated (Fig. 3). As expected, controls lacking enzyme, NDP-sugar or acceptor also lacked Δfluorescence. As further confirmation, the validated quenching of 4-MU fluorescence upon 4-MU 7-O-glucosylation measured in parallel at 460 nm with excitation at 355 nm (Fig. 3) [14], revealed an identical trend of catalyst proficiency to that determined by the 1-Zn(II) NDP sensor assay. For the most active variant TDP16, steady state kinetic parameters were also determined using the 1-Zn(II) NDP sensor assay in a 96-well plate format (Fig. S1). Saturation was observed by varying 4-MU at a fixed concentration of TDP-Glc (1 mM) to provide an apparent KM of 0.24 ± 0.011 mM and kcat of 12.3 ± 0.45 min−1 (kcat/KM = 51 mM−1 min−1) and these parameters are comparable to those previously determined via a discontinuous HPLC assay for TDP16 [21].</p><p>Finally, to assess the high throughput applicability of 1-Zn(II) NDP sensor assay, we examined crude extract compatiblity. Specifically, the plate-based 1-Zn(II) NDP sensor assay was applied to crude cell extracts from OleD-expressing cells (E. coli BL21). Each cell extract for the four OleD variants was added to the reaction mixture containing TDP-Glc and 4-MU, and the mixture was transferred to the assay solution in a 96-well plate and the fluorescence was measured at 520 nm with excitation at 485 nm (Fig. 4). Based upon this analysis, the observed crude extract reactivity trends were identical prior assessments using homogenous catalysts. Importantly, this study clearly demonstrates the 1-Zn(II) NDP sensor assay to be fully compatible with crude extract analyses as controls lacking expressed GT or less active GT displayed little to no detectable background signal (a Z factor of 0.82 was determined for the assay containing TDP16 at 100 min [25]). In addition, this study clearly demonstrates the ability of the 1-Zn(II) NDP sensor assay, even in a plate-based crude extract format, to distinguish among a range GT mutants which display differing proficiencies.</p><p>In conclusion, a general 1-Zn(II) NDP sensor assay has been developed for rapid evaluation of GT activity. The assay as described is sensitive, amenable to both purified enzymes and crude cell extracts, and, given the 1-Zn(II) NDP sensor selectivity for all five nucleoside diphosphates (K'd < 1 μM for ADP, TDP, UDP, GDP or CDP [19]) over the corresponding NDP-sugars (K'd > 20 μM), is anticipated to offer broad applicability.</p>

PubMed Author Manuscript

Description and Analysis of the First Spectrum of Iodine

An extensive survey of the spectra of iodine has led to a list of more than 900 lines emitted by neutral atoms in the region from 23070 A in the infrared to 1195 A in the extreme ultraviolet. Wavelengths between 12304 A and 2061 A were derived from measurements of spectrograms obtained with gratings of high dispersion. Wavelengths of lines outside these limits are the computed values for lines observed on photometric tracings of the infrared, inaccessible to photography, and in the ultraviolet with a vacuum-grating spectrograph. For many of the lines Zeeman patterns were obtained in a magnetic field of about 37,000 oersteds. With these data many of the lines have been classified as combinations between odd levels from the electron configurations 5s2 5p4 np and 5s2 5p4 nf, and even levels from the configurations 5s2 5p4 ns and 5s2 5p4 nd. Among these levels several sets have been recognized as forming Rydberg sequences that are in close agreement in placing the ground state 5p5 2P1½o of I i at 84,340 cm−1 below the ground state 5p4 3P2 of I ii. This gives 10.45 electron-volts for the ionization potential of the neutral iodine atom. A strong infrared line at 13148.8 A is explained as a magnetic dipole transition between the levels of the ground term 5p5 2P°.

description_and_analysis_of_the_first_spectrum_of_iodine

1. Introduction<!>2. Experimental Procedure<!>3. Results<!>4. Term Structure of I i<!>5. Series and Ionization Potential<!>6. The Continuous Spectra<!>7. Discussion<!>

<p>The spectra absorbed and emitted by iodine in its atomic and molecular states have been the object of many investigations. In volumes 5 and 8 of his Handbuch der Spectroscopie, Kayser lists 352 papers, which appeared up to 1933, dealing with various aspects of the spectral behavior of this heavy member of the halogen family. Since that date additional papers have appeared, of which some are cited below. But in spite of this abundant material, representing a vast amount of work, knowledge of the first spectrum of iodine, I i, emitted by neutral atoms, has remained scanty and fragmentary, largely owing to the fact that important parts of the spectrum lie in the not easily accessible ultraviolet and infrared regions. It is the purpose of this paper to present a new description of the first spectrum of iodine and an analysis of its term structure.</p><!><p>The investigations of the first spectrum of iodine at the Bureau were made at two different times under different experimental conditions. The first series of observations was made more than 30 years ago when chlorine and bromine [1]1 also were being investigated. The light source was a Geissler tube of Pyrex glass into which a small amount of dry iodine vapor could be admitted from time-to-time to replace that which was adsorbed on the walls of the tube or absorbed by its aluminum electrodes. The lamp was similar to that used in the experiments on chlorine and bromine, and was excited to luminescence in the same way, with an uncondensed discharge from the high-voltage side of a 40-kv transformer. The spectrograms were recorded on plates sensitized to the green, orange, red, and infrared regions of the spectrum by bathing ordinary photoplates in solutions of the photosensitizing dyes available at that time; namely, pinaverdol, pinacyanol, dicyanin, and the newly discovered kryptocyanin. The spectrographs carried concave gratings of 21-ft radius of curvature, ruled with 7,500 and 20,000 lines per in. and set up in Wadsworth mountings. Each exposure to the light source was made with one-half the length of the spectrograph slit covered with a colored-glass filter so that both the first-order spectrum and the overlapping second order could be obtained at the same time. Each plate was exposed also to light from the iron arc, in both the first and second orders, to obtain the necessary standard wavelengths for use in deriving the iodine wavelengths. Because the capillary of the discharge became discolored after a run of a few hours, it was necessary to make exposures of nearly 24-hr duration in order to photograph the lines of longest wavelength recorded on the plates.</p><p>Measurement of these spectrograms yielded a list of approximately 400 wavelengths, with estimated intensities, extending from 9732 A in the infrared to 3820 A in the ultraviolet. This list was seemingly the most extensive description of the first spectrum of iodine then available and was being prepared for publication when the paper, "The Arc Spectrum of Iodine," by Evans [2] appeared. A comparison of his list and ours showed that they were essentially identical. This fact and also the fact that his paper contains the first real results for a classification of the iodine spectrum outside of the Schumann region induced us to defer publication of our results until a substantial addition could be made to the description and analysis of the spectrum.</p><p>The second series of observations was made at various times during the period from 1953 to 1957. Improved apparatus and new experimental procedures have made it possible to advance the description of the iodine spectrum beyond the limits reached in the earlier work, and also to obtain Zeeman-effect observations that have led to a revision and extension of its term structure. The new light source was an electrodeless discharge tube of the type described by Corliss, Westfall, and Bozman [3]. It was excited to luminescence in a field of 2,450 Me from a microwave oscillator. The plates used to record the spectra were EK types 103a–O–UV, 103 a–O, I–F, I–N, I–Q, and I–Z, according to the region investigated, and, where required, were hypersensitized in an ammonia bath by the method recommended by Burka [4]. Four concave-grating spectrographs and a Hilger E 1 quartz-prism spectrograph were used to obtain the spectrograms. The spectrographs carrying the gratings with 7,500 and 15,000 lines per in. were used for the infrared and red regions where many new strong lines were found. For the shorter regions the grating with 30,000 lines per in. was used as well as the one with 15,000. For the extreme ultraviolet both the Hilger E 1 instrument and a 2-m glass grating ruled with 30,000 lines per in. and mounted in a vacuum chamber were used. All the spectrograms bore exposures to the iron arc or other sources of standard lines to be used in the determination of the iodine wavelengths.</p><p>For the Zeeman-effect observations the Weiss water-cooled magnet of the Bureau was used. With a current of 160 amp through the coils and a gap of 5 mm between the pole pieces, a field of approximately 37,000 oersteds was produced. The source between the pole pieces was also an electrodeless lamp of the type mentioned above, but of diameter 4 mm. A Wollaston prism of quartz placed between the light source and the projection lens of the spectrograph separated the two polarizations on the slit, with space between them for a no-field exposure. On plates appropriate for the regions under investigation resolved magnetic patterns were recorded for nearly all the strong lines of I i from 2062 A in the ultraviolet out to 11246 A in the infrared. Zeeman patterns were recorded also for some I ii lines of long wavelength.</p><!><p>The observational data and the deductions from atomic theory that are essential for the description of the spectrum of iodine and its term structure are embodied in the tables appended to this paper. In tables 1, 2, and 3 are listed, in the first three columns, the wavelengths of the lines of I i, their estimated intensities and characteristics, and their wave numbers in vacuum. The letters after the intensity numbers have the following significance: c=partially resolved hyperfine structure (hfs); d=double; w=widened line owing to unresolved hfs; h=hazy, diffuse; Z=Zeeman pattern given in table 4. The letters A and B indicate the type of shading displayed by unresolved patterns; thus: A=⩘|\; B=|\/|. The term combinations in the fourth column of the tables are based on g- and J-values derived from the Zeeman-effect patterns of table 4.</p><p>Although the intensities are estimates made according to the usual practice of spectroscopists, an attempt has been made to bring them into closer relationship with photometric standards than is possible in a compressed linear scale in which the numbers are roughly proportional to the logarithms of the actual intensities. In table 2 the intensities are the measured heights of the peaks of the lines above the noise level of the recorder tracings.</p><p>In tables 1 and 3, however, an attempt has been made to bring the estimated intensities into harmony with a photometric scale that reflects the enormous range in the strength of the lines. On this scale the faintest lines are arbitrarily assigned an intensity 1, and the strongest lines, that occur in the multiplet 6s4P—6p4P° are designated as 105 times as strong. This ratio is based on an accurate determination of the relative strengths of Rowland ghosts to their parent lines. Thus, for the strongest lines estimates were made of the intensities of their ghosts, and then multiplied by the corresponding factors to establish the intensities of the parent lines. Although the scale was thus established to represent true relative strengths of lines in short ranges of the spectrum, no attempt was made to standardize them over longer ranges of wavelengths. In particular the intensity scale used in the vacuum region is several orders of magnitude less than that of the visible and infrared. A few of the longest wavelengths listed in table 1 were measured also on the infrared recordings described above. The intensities derived for them from these observations are given in parentheses following the estimated photographic intensities, thereby giving a comparison of the relative sensitivity of the two modes of observation in a region near the limit of photographic detection.</p><p>The terms to be expected theoretically on the assumption that LS-coupling governs the orbital and spin angular momenta of the atom are given in table 5. Those actually found in this investigation are given in tables 6 and 7. These terms make it possible to calculate accurate wavelengths for lines in the infrared and extreme ultraviolet regions that lie beyond the reach of photographic recording with high-dispersion spectrographs. Such lines as have been observed by other investigators are given in tables 2 and 3 with wavelengths calculated from the terms of tables 6 and 7.</p><p>Lines listed in table 2 were observed by C. J. Humphreys [5] at the Naval Ordnance Laboratory, Corona, and by E. K. Plyler [6] at the Bureau. In this work they used electrodeless-discharge lamps, similar to those mentioned above in conjunction with their recording infrared spectrometers. These observations, which were made expressly for this investigation, verify all but two of the new infrared lines measured by Eshbach and Fisher [7] and add several lines not previously observed. All the ultraviolet lines in table 3 were measured on spectrograms obtained with the 2-m vacuum-grating spectrograph of the Bureau. These data not only confirm the descriptions of II given by Turner [8], La Croute [9], McLeod [10], and Hellerman [11], but increase by a factor of more than 4 the number of lines reported by these earlier observers. The light source was an electrodeless discharge of the type described above but modified to incorporate a LiF window sealed to the tube with an O-ring. The wavelengths listed in the first column are not the values derived directly from the measurements, but are values calculated from the energy levels given in tables 6 and 7. They were derived, therefore, indirectly from international secondary standard wavelengths, and are believed to be correct to less than 0.005 A. They are recommended for use as standards in the vacuum region.</p><p>A problem of prime importance for the analysis of the first spectrum of iodine is the evaluation of the separation of the levels of the ground state 5p5 2P°. The lines at 2061.6 and 1830.4 A, which are due to the transitions 5p5 2P0½o—6s 2P1½ and 5p5 2P1½o—6s 4P2½, may be used for this purpose. These lines have been measured several times by different observers, but the wavelengths reported for them are only approximately correct and are afflicted with the errors that are inherent in the reference lines against which they were measured. In the present work the mean wavelength of the longer line has been determined as 2061.633 A, from seven observations, made with high dispersion in the higher orders of the gratings, relative to international secondary standards in the iron arc spectrum.</p><p>The wavelength of the shorter line was determined as 1830.380 A from measurements relative to internal standards selected from the iodine spectrum itself. These lines are at 1876, 1844, 1799, 1702, and 1593 A, and appear with 1830 A on spectrograms made with the vacuum-grating spectrograph. We have determined accurate wavelengths for them, from measurements of lines of longer wavelength on high-dispersion spectrograms, by making use of the combination principle. Thus, we find the following level-separations, which are mean values of the wave-number differences between numerous pairs of well-measured lines: 6s 4P0½—6s 2P1½=4803.39cm−16s 4P1½—6s 2P1½=5726.93cm−16s 2P0½—6s 2P1½=7093.88cm−16s′ 2D1½—6s 2P1½=10262.33cm−1nd5.11½—6s 2P1½=14262.05cm−1</p><p>By adding each of these numbers to 48489.73 cm−1, the wavenumber of 2061 A, we obtain the wavenumbers, and thence the exact wavelengths, of the selected standards. These are given in table 3.</p><p>With the accurate values thus established for the two transitions given above we have: 1830.380A is5p5 2P0½o—6s 4P2½=54633.46cm−16s 4P2½—6s 2P1½=14659.42cm−11782.758A is5p5 2P1½o—6s 2P1½=56092.88¯cm−12061.6338A is5p5 2P0½o—6s 2P1½=48489.73cm−15p5 2P1½o—5p5 2P0½o=7603.15¯cm−1.</p><!><p>The spectrum to be expected theoretically for neutral iodine atoms, if LS-coupling governs their behavior under excitation, is that based on the terms given in table 5. In the unexcited state of the atom the electron configuration is 5s2 5p5, which yields an inverted 2P term of odd parity. The higher states of even and odd parity arise when excitation of the atom leads to the electron configurations listed in the first column of table 5. Ionization of the atom leaves it in one of the states represented by the 3P, 1D, and 1S terms of the basic electron configuration 5s2 5p4 of the ion I+. These three terms give rise, therefore, to the three families of terms that are likely to produce the strongest lines of the spectrum I i. A fourth family also is expected based on the addition of s, p, d, etc., electrons to the 5p6 configuration. The terms from these configurations will be doublets, and the lines arising from their combinations probably will be among the weaker lines of the spectrum.</p><p>The first real regularity in the spectrum of iodine was announced by Turner [12] who found the wavenumber interval of approximately 7,600 cm−1 recurring among several pairs of the lines observed by him in the Schumann region. Subsequently he correctly suggested that this difference represents the separation of the levels in the ground term 2P of the 5s2 5p5 electron configuration. From this starting point all further advances in the interpretation of the first spectrum of iodine have been made. Evans [2], Deb [13], and Murakawa [14], in their analyses, have recognized these lines as resulting from the 6s → 5p transition, but they are not in agreement on their designations of the individual levels of the 5p4 6s configuration. Inasmuch as the interpretations of I i offered by these investigators all rest on the 6s-levels it is, therefore, not surprising that there is disagreement among them.</p><p>The classifications of the lines of I i given in this paper likewise are based on the 6s-levels. With g- and J-values derived from well-resolved Zeeman patterns it is now possible to designate these levels with certainty and also some of the higher np-levels with which they combine. Increasing excitation of the iodine atoms brings into play still higher levels from the np, nd, ns, and nf electrons, but it is difficult to designate them with certainty because the g-values indicate a breakdown of the orbital and spin momenta to a coupling scheme between pure LS-coupling and jj-coupling, probably jl-coupling described by Racali [15].</p><p>Most of the levels of I i that are given in tables 6 and 7 result from the addition of ns, np, nd, and nf electrons to the lowest energy states 3P, 1D, and 1S of I ii. These were reported first by LaCroute [9] whose analysis, giving the relative positions of these terms, shows that the levels of 3P are separated by large differences in wave number. This fact has been an important guide in the analysis of I i, for nearly all the terms derived from 3P show similar characteristics. The eight levels of the 5s2 5p4 6s electron configuration all closely conform to the pattern of spacing exhibited by their parentage. This is confirmed by the g-values of these levels, which are nearly equal to the g's for LS-coupling, their sum being ∑gobs=11.985 as compared with ∑gLS=12.000.</p><p>Of the levels coming from configurations with np, nd, and nf electrons, only those derived from 3P levels can be designated with certainty. Here again the classifications rest on level intervals and g-values; but the g-values now show marked deviations from LS-g's, owing to configuration interaction, and there is a tendency for levels to form pairs. In the case of the 13 levels derived by adding a 6p-electron to the parent term 3P, g-sharing is such as to remove all resemblance to LS g-values, yet their sums are nearly equal, namely ∑gobs=17.847 and ∑gLS=18.000. Few if any levels from the parent terms, 1D and 1S, are definitely established.</p><!><p>Before the first attempts were made to unravel the spectrum of iodine, several investigators reported the results of their measurements of critical and ionizing potentials in iodine vapor. Thus, in the period from 1920 to 1924, values of the ionization potential from 10 to 10.5 ev were reported by Found [16], Mohler and Foote [17], Duffendack [18], and Mackay [19]. From these results it was evident that the separation  2P1½o− 3P2 between the ground states of I i and I ii is approximately 85,000 cm−1. Although the series announced by Deb [9] and by Price [20] give limits in close agreement with this value; yet, in the light of the present analysis, their series must be considered fictitious and the ionization potentials derived from them fortuitous.</p><p>The first physically real series of I i was given by Evans [2] in his analysis of the spectrum. Two of his levels in combination with the 6s-term accounted for two pairs of strong lines separated by the same difference in wave number. The limit and ionization potential derived from them are within the range of the experimental values cited above. This series has been confirmed and extended in the present work. Other series of three and more members have been found also, as set forth in table 8. Four of these series, with four and five members, representing the migration of the s, p, and d electrons, have been used to calculate the separation of the ground states of I i and I ii and the corresponding ionization potential. A Ritz formula R/(m+α+β/m2)2 has been evaluated for the variable terms of these series, and with the values given in the table for α and β, it is found to represent the series very closely. In fact the first solutions of this formula were found to fit series of three members closely enough to predict higher series members that were subsequently found. The individual determinations of the interval  2P1½o− 3P2 between the ground states of I i and I ii are in close agreement among themselves, and when an unweighted average of them is taken, yield a value of 84,340 cm−1. From this an ionization potential of 10.45 ev is derived for the work required to remove a p-electron from the configuration 5s2 5p5.</p><!><p>An outstanding feature of the spectra of iodine emitted by the electrodeless lamps used in this investigation is a succession of continua extending from 4800 A in the blue to 2900 A in the ultraviolet. These bands have been encountered by nearly all observers who have studied the spectra emitted by iodine molecules and atoms subjected to various modes of excitation. Perhaps the best description of them is that published by Curtis and Evans [21]. Our observations are in agreement with their view that these continua belong to two systems, the one consisting of a single band or of several broad overlapping bands between 4800 and 4035 A, the other consisting of a very broad strong band followed by several groups of much fainter and narrower bands. In each of these groups the individual bands are about 25 A in width, nearly equally spaced, and increase in intensity to a maximum near the center of the group and then decline. The mean separation of the bands is 210 cm−1, which is exactly the separation of the vibrational levels of the ground state of the I2 molecule as found by Kimura and Miyanashi [22] in their study of the ultraviolet absorption bands of I2. This fact casts doubt on the results of various investigators who have sought to ascribe the bands to atomic recombination processes. A description of these emission features as recorded on our spectrograms is given in table 9. Their presence is unwelcome in investigations of I i because they mask completely the fainter and widened lines and attenuate differences in intensity of measurable lines that appear on the continuous background.</p><!><p>The analysis of the first spectrum of iodine, presented in the preceding pages, shows clearly that in its excited states the atom has departed from LS-coupling and has reached a stage in intermediate coupling. Since the g-value of 5p5 2P0½o, of the ground state, is very closely that given by LS-coupling, we may assume that the g-value of the lowest member of the ground term, 5p5 2P1½o, is also that for LS-coupling. In the first excited states, however, arising from the electron configuration 5s2 5p4 6s, the departure from LS-coupling is marked. Although the g-values deviate only slightly from the LS-values, the levels of the 4P and 2P terms show the separations that are characteristic of their parent term, 3P, the ground state of I ii. This is in accord with the scheme for the p4 configuration in intermediate coupling as illustrated by Condon and Shortley [23].</p><p>In the configurations containing np-electrons, in which n ≧ 6, almost all resemblance to LS-coupling has disappeared from the term structure. The g-values for all the levels deviate strongly from the LS-values, except those for  4D3½o and  4D2½o for which the deviation is slight. The levels fall into a pair structure, of the kind prescribed by Racah [15] for jl-coupling, in which the level separations bear no resemblance whatsoever to those of LS-coupling. The g-values calculated for the np-levels with Racah's formula fit the observed g's more closely than do the g's of LS-coupling. A similar situation holds for levels from configurations with nd-electrons, and also with nf-electrons. The nf-levels fall very close to each other so that lines originating in them are separated by intervals less than those imposed by a magnetic field, say, of 35,000 oersteds. All of these lines for which magnetic patterns appear on our spectrograms of the Zeeman effect show the unsymmetrical structures, due to Paschen-Back interaction, similar to those described by Kiess and Shortley [24] for lines of oxygen and nitrogen.</p><p>These matters raise the question as to the appropriateness of the designations used for the energy levels of a heavy atom such as iodine. It is obvious that in a complex spectrum manifesting in its various states a transition from one coupling scheme to another, no single scheme of notation will be adequate or satisfactory. The only notation scheme that has achieved a status of widespread usage and permanence is the one devised for spectra built on LS-coupling. It is here emphasized that the use of it, in this paper, for levels that do not result from this coupling scheme, is for convenience in designating them, and not for attaching to them the quantum significance usually conveyed by the LS-symbols.</p><p>An inspection of table 1 will show that some lines have resisted all attempts to classify them. There is no doubt, however, that they belong to I i; but the necessary links to connect them to known or, as yet, unknown levels have not been found. One of the lines left unclassified after the bulk of the analysis had been completed is the relatively strong infrared line at 13149.19 A as measured by Eshback and Fisher [7]. This line, measured by us also on the recordings of the infrared spectrum of iodine by Plyler and by Humphreys, has a wave number practically identical with the separation of the levels in the 2P° ground term of the iodine atom. Accordingly it is designated in table 2 as the forbidden transition  2P0½o→ 2P1½o and a wavelength corresponding to the wave number 7603.15 cm−1 is calculated for it. A similar transition has been reported by Edlén [25] for the isoelectronic spectrum Xe ii. We have tried to photograph this line on EK–IZ plates so as to get a more accurate value of its wavelength, but the experiments were unsuccessful. Our belief that our designation of it is correct is substantiated by the discovery of similar transitions between the metastable levels of I ii, as reported by Martin and Corliss in their forthcoming paper on I ii.</p><p>In conclusion we acknowledge our indebtedness to several of our colleagues for data used in this investigation. Both C. J. Humphreys and E. K. Plyler made observations of the infrared spectrum of iodine beyond the reach of photography. Their data are presented in table 3. W. F. Meggers and R. Zalubas measured the magnetic patterns of numerous iodine lines during their investigations of spectra emitted by various electrodeless metal-halide lamps. Finally, W. C. Martin, Jr., made new measurements of iodine spectra in the extreme ultraviolet that have surpassed in extent and accuracy earlier descriptions of these spectra. It is a pleasure for us to thank each for his contribution to this paper.</p><!><p>Figures in brackets indicate the literature references at the end of this paper.</p><p>C.J. Humphreys, NAVORD Report 4571 (March 1956) NOLC Report 321.</p><p>E. K. Plyler (unpublished data).</p><p>Wavelengths and term combinations of I i</p><p>Infrared lines of I i</p><p>Observed wavelength.</p><p>Wavelengths of Ii in the ultraviolet</p><p>Wavelengths measured by C. H. Corliss and W. C. Martin.</p><p>Also I ii.</p><p>Masked by I ii and Lyα.</p><p>Zeeman effect of I i</p><p>Predicted terms of I i</p><p>Odd terms of I i</p><p>Even terms of I i</p><p>Series of I i</p><p>Continua in the spectrum of iodine</p>

PubMed Open Access

Identification of two novel RET kinase inhibitors through MCR-based drug discovery: Design, synthesis and evaluation

From an MCR fragment library, two novel chemical series have been developed as inhibitors of RET, which is a kinase involved in the pathology of medullary thyroid cancer (MTC). Structure activity relationship studies (SAR) identified two sub-micromolar tractable leads, 6g and 13g. 6g was confirmed to be a Type-II RET inhibitor. 13g and 6g inhibited RET in cells transformed by RET/C634. A RET DFG-out homology model was established and utilized to predict Type-II inhibitor binding modes.

identification_of_two_novel_ret_kinase_inhibitors_through_mcr-based_drug_discovery:_design,_synthesi

1. Introduction<!>2.1. MCR hit generation<!>2.2. RET inhibitor optimization<!>2.3. Cell-based studies<!>2.4. Homology model development<!>2.5. RET computational modeling<!>3. Conclusion<!>4.1. General experimental<!>4.2.1. 5-Phenylpyridin-2-amine<!>4.2.2. N-(tert-butyl)-6-phenylimidazo[1,2-a]pyridin-3-amine (1)<!>4.2.3. Ethyl 2-(4-bromophenyl)acetate (2)<!>4.2.4. Ethyl 2-(4-(6-aminopyridin-3-yl)phenyl)acetate (3)<!>4.2.5. Ethyl 2-(4-(imidazo[1,2-a]pyridin-6-yl)phenyl)acetate (4)<!>4.2.6. 2-(4-(Imidazo[1,2-a]pyridin-6-yl)phenyl)acetic acid (5)<!>4.2.7. N-(5-(tert-butyl)isoxazol-3-yl)-2-(4-(imidazo[1,2-a]pyridin-6-yl)phenyl)acetamide (6g)<!>4.2.8. 2-(4-(Imidazo[1,2-a]pyridin-6-yl)phenyl)-N-(3-(trifluoromethyl)phenyl)acetamide (6f)<!>4.2.9. 2-(4-(Imidazo[1,2-a]pyridin-6-yl)phenyl)-N-phenylacetamide (6e)<!>4.2.10. N-(3-chlorophenyl)-2-(4-(imidazo[1,2-a]pyridin-6-yl) phenyl)acetamide (6d)<!>4.2.11. N-(3-fluorophenyl)-2-(4-(imidazo[1,2-a]pyridin-6-yl) phenyl)acetamide (6c)<!>4.2.12. 2-(4-(Imidazo[1,2-a]pyridin-6-yl)phenyl)-N-(m-tolyl) acetamide (6b)<!>4.2.13. N-(3-bromophenyl)-2-(4-(imidazo[1,2-a]pyridin-6-yl) phenyl)acetamide (6a)<!>4.2.14. Methyl 2-(3-hydroxyphenyl)acetate (7)<!>4.2.15. Methyl 2-(3-((6-chloropyridazin-3-yl)oxy)phenyl)acetate (8)<!>4.2.16. Methyl 2-(3-((6-((diphenylmethylene)amino)pyridazin-3-yl)oxy)phenyl)acetate (9)<!>4.2.17. Methyl 2-(3-((6-aminopyridazin-3-yl)oxy)phenyl)acetate (10)<!>4.2.18. Methyl 2-(3-(imidazo[1,2-b]pyridazin-6-yloxy)phenyl) acetate (11)<!>4.2.19. 2-(3-(Imidazo[1,2-b]pyridazin-6-yloxy)phenyl)acetic acid (12)<!>4.2.20. N-(5-(tert-butyl)isoxazol-3-yl)-2-(3-(imidazo[1,2-b] pyridazin-6-yloxy)phenyl)acetamide (13g)<!>4.2.21. 2-(3-(Imidazo[1,2-b]pyridazin-6-yloxy)phenyl)-N-(3-(trifluoromethyl)phenyl)acetamide (13f)<!>4.2.22. 2-(3-(Imidazo[1,2-b]pyridazin-6-yloxy)phenyl)-N-phenylacetamide (13e)<!>4.2.23. N-(3-bromophenyl)-2-(3-(imidazo[1,2-b]pyridazin-6-yloxy)phenyl)acetamide (13d)<!>4.2.24. N-(3-chlorophenyl)-2-(3-(imidazo[1,2-b]pyridazin-6-yloxy) phenyl)acetamide (13c)<!>4.2.25. 2-(3-(Imidazo[1,2-b]pyridazin-6-yloxy)phenyl)-N-(m-tolyl) acetamide (13b)<!>4.2.26. N-(3-fluorophenyl)-2-(3-(imidazo[1,2-b]pyridazin-6-yloxy) phenyl)acetamide (13a)<!>4.3. RET biochemical inhibition assay<!>4.4. RET kinetic assay<!>4.5. RET homology model development<!>4.6. RET computational modeling<!>4.7. Cell cultures<!>4.8. Protein studies

<p>In 1985, the RET (Rearranged during transfection) gene was identified as a novel oncogene activated by DNA rearrangement. The isolated oncogene resulted from a recombination event between two unlinked human DNA segments, which occurs during the transfection process [1]. It was discovered that point mutations in RET lead to the development of medullary thyroid cancer (MTC) [2–9]. The identification of a small molecule inhibitor for the RET kinase represents an important strategy for the intervention of RET-driven cancers; however, the target has largely been neglected. Research manuscripts have been published on RET inhibition [10–16] and anticancer drugs vandetanib [17] and cabozantinib [18] were found to exhibit RET activity. However, both drugs were developed as vascular endothelial growth factor receptor 2 (VEGFR2) inhibitors and have VEGFR2 as the primary target. To improve efficacy and toxicity profiles, it is highly desirable to develop RET selective inhibitors for the treatment of MTC. We sought out to utilize MCR enabling chemistries to assist in the discovery of novel RET inhibitors.</p><p>Multicomponent reactions (MCRs) encompass a broad synthetic landscape where multiple inputs react to create a diversified final product [19,20]. The synthetic ease and facile analoging ability make MCR enabling chemistries an attractive tool for drug-discovery to rapidly identify novel, tractable leads. In the drug-discovery realm, MCR chemistry has been utilized to identify novel therapeutics for tuberculosis [21], viral infections [22], malaria [23], etc. Therefore, MCR enabling chemistries represent a highly validated means to generate biologically relevant chemotypes to treat human disease. Despite the large scope and application of kinase directed MCRs [24–27], no group has displayed the ability to generate RET kinase inhibitors utilizing MCR enabling chemistries. RET represents an unmet therapeutic target, and tailoring an MCR for lead development can help facilitate the discovery of novel inhibitor functionality.</p><p>In order to utilize MCR enabling chemistries to target the RET kinase we have generated a kinase inhibitor library utilizing a modified MCR protocol. The MCR is based on the Groebke–Blackburn–Bienaymé reaction [28], but glyoxylic acid is employed in order to create monosubstitution on the 5 membered ring of imidazopyridine (Scheme 1) [29,30]. The modified MCR permits the rapid generation of diverse, drug-like scaffolds that contain kinase inhibitor functionality. By using this chemistry, we have synthesized more than 100 imidazopyridine analogues. The screening of these analogues, followed by non-MCR hit optimization, produced two novel chemical scaffolds that inhibited the RET kinase at sub-micromolar concentrations. We report herein the progress and development of RET kinase inhibitors derived from an imidazopyridine MCR chemical library.</p><!><p>MCR kinase library screening was completed with a mid-throughput screening platform utilizing microfluidics technology established by Caliper® [31]. The assay contains a microfluidics chip that can separate phosphorylated peptide from unphosphorylated starting peptide. The starting peptide is FAM tagged, and can be tracked in each well using the EZ Reader® plate reader. With this screening technique, an MCR kinase library with 100 kinase fragments was screened and RET active 1 was identified (Fig. 1, A).</p><p>Based on the structure of compound 1, the warhead region (imidazopyridine) is predicted to bind to the RET hinge, while the phenyl ring system is predicted to engage a lipophilic pocket adjacent to the RET allosteric pocket. This is an atypical binding modality [32], permitting the fragment to bind the kinase 'backward' in comparison to the typical binding of ATP to the active site. Computational modeling showed the phenyl ring system was in close proximity to the allosteric pocket on RET and therefore adding allosteric functionality to compound 1 was hypothesized to increase potency (Fig. 1, B).</p><!><p>A synthetic protocol was developed to grow MCR active 1 into the RET allosteric pocket (Scheme 2). The solvent exposed tert-butyl moiety was replaced with a hydrogen to decrease lipophilicity at the solvent region. Using Fischer esterification, starting acid 2 was esterified to compound 3. Employing developed Suzuki coupling conditions, compound 4 was generated from a coupling reaction between an amino-pyridine boronic ester and compound 3. Compound 4 was cyclized to compound 5 using chloroacetaldehyde. After, compound 5 was saponified with base to produce 6 and subsequently coupled to various anilines using EDC to generate compounds 6a–6g.</p><p>As predicted, growing MCR active 1 into the allosteric region of RET drastically increased inhibitor potency as seen with compound 6g (RET IC50 = 0.21 ± 0.04 µM) (Table 1). Other compounds, 6e (RET IC50 = 18.8 ± 4.6 µM) and 6f (RET IC50 = 15.7 ± 2.9 µM), also displayed an increase in activity. The structure activity relationship (SAR) trend on RET is that bulky, lipophilic moieties at the meta position increase potency by engaging a conserved hydrophobic region. Compound 6f is >6 times more potent than 6b because the –CF3 moiety from compound 6f is more lipophilic and bulky than the –CH3 moiety on 6c. Interestingly, the SAR trend does not hold true for compound 6e, which contains a hydrogen atom at the meta position. Likely, the hydrophobicity of the unsubstituted aniline on 6e creates high-binding affinity at the hydrophobic region. From MCR active 1, a series of RET inhibitors were developed, which uncovered compound 6g, a RET lead inhibitor with sub-micromolar activity.</p><p>6g was strategically designed to engage the conserved hydrophobic region of RET (Fig. 2). Taking into account the increase in activity achieved with 6f, it was hypothesized that exchanging the –CF3 moiety with a tert-butyl group would have a favorable impact on activity (Fig. 2). Because of computational modeling results, 3-amino-5-tert-butylisoxazole was used as the amine input to generate compound 6g, which displayed ~75 times increase in activity when compared to 6f. It is hypothesized that the increase in activity is a direct result from better engagement in lipophilic contacts at the allosteric pocket (Fig. 2).</p><p>Compound 6g was subjected to an incubation assay to determine the kinetics of RET inhibition. In general, compounds that access the kinase allosteric pocket bind the DFG-out kinase conformation and incubation is typically required to achieve maximal inhibition [33]. With an increase in incubation time, 6g achieved greater RET inhibition (Fig. 3). A linear line was fit to the IC50 vs incubation time and the IC50 value for no incubation time (Y = 0) was determined to be 0.44 ± 0.03 µM. Therefore, through incubation, RET IC50 increases greater than 2 times and supports the hypothesis that 6g binds the DFG-out fold of RET. This suggests 6g is a Type-II kinase inhibitor that is not directly competitive with the binding of ATP.</p><p>Compounds based on the structure of 6g have very ridged, liner geometries and work was completed to identify a more flexible scaffold to define additional SARs on RET. An ether linker was hypothesized to be tolerated from a computational modeling study, and therefore a synthetic protocol was developed (Scheme 3).</p><p>Using Fischer esterification, compound 7 was generated from 7a. Compound 8 was synthetized employing a strong base to couple 7 with 3,6-dichloropyridazine. Buchwald conditions were utilized to convert compound 8 into compound 10, an amino-pyridazine. Cyclization of compound 10 with chloroacetaldehyde generated intermediate 11. Compound 11 was hydrolyzed with base to generate compound 12, which was subsequently coupled to various anilines using EDC to generate compounds 13a–g.</p><p>Through scaffold hopping from a novel MCR-based hit 1, compound 13g (RET IC50 =0.75 ± 0.03 µM) was identified, generating an additional RET lead inhibitor. The SAR on the scaffold based on compound 13g displayed a more direct trend than SAR from compound 6g (Table 1). On average, compounds generated from Scheme 3 displayed higher potency on RET, and bulkier groups at the meta position of the allosteric region produced compounds with higher potency. Similar to the case of compound 6g, compounds 13a–g engage a hydrophobic region in the allosteric pocket, which increases compound potency. Interestingly, despite having on average weaker RET inhibitors, Scheme 2 produced the most potent compound, 6g. Also, both 6g and 13g contain a t-butyl isoxazole structural moiety in the allosteric pocket and are at least 10-times more potent than any other compound with different substituents at the same region.</p><!><p>To further evaluate RET inhibition of 13g and 6g, the compounds were progressed into cell-based assays. The assay determined the amount of RET target inhibition by monitoring phosphorylation status of Y905 and Y1062 on the RET kinase domain (Fig. 4). Y1062 is responsible for activating PI3K/AKT and RAS/ERK pathways and is important to inhibit to block RET oncogene signaling. 13g was found active on RET in RAT1 cells transformed with a RET/C634R oncogene at an IC50 between 2.5 and 10 µM (Fig. 4, A). This value corresponds well to the determined biochemical IC50 of 0.75 ± 0.03 µM. 6g was found active on RET in RAT1 cells transformed with a RET/C634R oncogene at an IC50 between 0.25 and 0.50 µM (Fig. 4, B). Like 13g, the cell activity of 6g corresponds well to the determined biochemical IC50 of 0.21 ± 0.04 µM. Because 6g was shown to bind RET in the DFG-out fold (Fig. 3), there is not a large difference in biochemical vs cellular IC50s again suggesting the compound is not directly competitive with ATP. 6g represents a strong lead candidate that can be further developed into a RET advanced lead.</p><!><p>In order to better understand activity for compounds 6g and 13g a novel RET DFG-out homology model was generated utilizing Swiss-Model [34–36]. Crystal structure coordinates of a VEGFR-2 DFG-out structure was used as a template [37], and the RET amino acid sequence [38] was employed to build a RET DFG-out crystal structure. The resulting RET DFG-out model clearly displays the correct shift in the DFG-loop, which opens up the allosteric pocket of the kinase for inhibitors to access. There are two interchangeable folds of the RET kinase: DFG-in, the kinase is active and the allosteric pocket is closed; DFG-out, the kinase is not active and the allosteric pocket is open [33]. The developed DFG-out model served as a tool to evaluate binding modes of compounds 6g and 13g.</p><!><p>Compounds developed from the MCR-hit 1 were created to access the allosteric pocket of RET, which is only available in the DFG-out fold and a DFG-out homology model was developed and utilized to computationally model 6g and 13g (Figs. 5 and 6). Compound 6g is predicted to bind to the hinge of RET with imidazopyridine, making a hydrogen bond with A807 and pi–pi stacking with Y806. The amide of compound 6g sits on a 'bridge' held up by two hydrogen bonds with E775 and D892 (from the DFG motif), which allows the inhibitor to enter the allosteric pocket. This is a typical binding modality that is also observed with imatinib (PDB# 3K5V). Within the allosteric pocket, there is a hydrophobic region containing conserved valine, leucine, and isoleucine residues that the tert-butyl moiety engages (Fig. 5). Compound 13g is predicted to bind RET in a similar mode to compound 6g, where the imidazopyridine ring system makes a hydrogen bond with A807 and pi–pi stacks with Y806 at the RET hinge. At the 'bridge' region, the aryl-ring system from engages L758 through an ion-induced dipole interaction. Also at the 'bridge' region, E775 and D892 (from the DFG motif) hydrogen bond to 13g. This permits access into the allosteric pocket where the tert-butyl moiety can engage the hydrophobic region (Fig. 6). Because of similar biding-modes observed with compounds 6g and 13g, both compounds display somewhat similar potency. However, compound 6g was found ~3 times more potent than compound 13g in the RET biochemical assay. In cell-based assays, 6g was found ~10 times more potent than 13g.</p><!><p>MCR fragment-based drug discovery has been utilized to uncover two novel kinase inhibitors (6g and 13g) from active fragment 1, and established 'proof-of-concept' for MCR-directed RET hit to lead generation. The binding geometries of both compounds are unique for the RET kinase and present as valid starting points for drug-discovery campaigns. A RET DFG-out homology model was generated and utilized to model both lead inhibitors. The tert-butyl moiety was found essential for efficient potency, and engages a hydrophobic region in the allosteric pocket of RET. An incubation based binding study was completed that identified 6g binds RET slowly, typical of inhibitors that bind the DFG-out form of the kinase. Further, 13g and 6g inhibited RET in RAT1 cells transformed by RET/C634R. Both compounds are being developed into advanced leads through optimization of potency and activity, and the work will be published in due course.</p><!><p>All solvents were reagent grade or HPLC grade and all starting materials were obtained from commercial sources and used without further purification. Purity of final compounds was assessed using a Shimadzu ultra-high throughput LC/MS system (SIL-20A, LC-20AD, LC-MS 2020, Phenomenex® Onyx Monolithic C-18 Column) at variable wavelengths of 254 nM and 214 nM (Shimadzu PDA Detector, SPD-MN20A) and was >95%, unless otherwise noted. The HPLC mobile phase consisted of a water–acetonitrile gradient buffered with 0.1% formic acid. 1H NMR spectra were recorded at 400 MHz and 13C spectra were recorded at 100 MHz, both completed on a Varian 400 MHz instrument (Model# 4001S41ASP). Compound activity and kinetics were determined with the EZ Reader II plate reader (PerkinElmer®, Walthman, USA). Vandetanib was obtained from LC Laboratories® (Woburn, USA, Lot# BTB-105). All compounds were purified using silicagel (0.035–0.070 mm, 60 Å) flash chromatography, unless otherwise noted. Microwave assisted reactions were completed in sealed vessels using a Biotage Initiator microwave synthesizer.</p><!><p>5-Chloropyridin-2-amine (500 mg, 3.89 mmol), phenyl boronic acid (711 mg, 5.83 mmol), and Na2CO3 (11.67 mmol) were added to a mixture of 4:1 DMF/Water (10 mL). The resulting solution was stirred and degassed with argon for 5 min. P(Cy)3 (49.0 mg, 0.175 mmol) and Pd2(dba)3 (53.4 mg, 0.58 mmol) were then added and the reaction was heated under microwave irradiation for 30 min at 130 °C. The crude reaction was adsorbed onto silica and purified using flash chromatography utilizing a DCM/MeOH gradient to yield compound 5-phenylpyridin-2-amine (345 mg, 52.1%). 1H NMR(400 MHz, DMSO-d6) δ 8.23 (dd, J = 2.5, 0.6 Hz, 1H), 7.77 (dd, J = 8.0, 1.5 Hz, 1H), 7.69 (dd, J = 8.6, 2.6 Hz, 2H), 7.53 (d, J = 7.2 Hz, 2H), 7.38 (d, J = 7.2 Hz, 2H), 6.53 (d, J = 9.3 Hz, 1H), 6.10 (s, 2H). ESIMS m/z [M+H]+ 171.</p><!><p>Glyoxylic acid 50% solution in water (0.097 mL, 0.881 mmol), 5- phenylpyridin-2-amine (100 mg, 0.588 mmol), and acetic acid (5.08 µL, 0.088 mmol) were added to MeOH (1 mL) under stirring. After 15 min, 2-isocyano-2-methylpropane (0.070 mL, 0.617 mmol) was added to the reaction and was stirred overnight. After 12 h, the reaction was adsorbed onto silica and purified using flash chromatography using a DCM/MeOH gradient to yield compound 1 (90 mg, 57.7%). 1H NMR (400 MHz, DMSO-d6) δ 8.52 (dd, J = 1.8, 1.0 Hz, 1H), 7.99 (s, 1H), 7.75 (dd J = 8.0,1.5 Hz, 1H), 7.69 (dd J = 8.4, 1.2 Hz, 2H), 7.49 (t J = 8.0 Hz, 3H), 7.42 (dd J = 9.3,1.9 Hz, 1H), 7.15 (s, 1H), 4.61 (s, 1H), 1.17 (s, 9H). ESIMS m/z [M+H]+ 266. HPLC Purity >95% at 254 nM.</p><!><p>2-(4-bromophenyl)acetic acid (15 g, 69.8 mmol) was added to ethanol (125 mL). Then, sulfuric acid (0.186 mL, 3.49 mmol) was added to the reaction and the reaction was heated to 80 °C for 12 h. The reaction was confirmed complete based on TLC. Solid NaHCO3 was added to the reaction and then the ethanol was evaporated. The product was extracted with ether and washed 2× with water and 1× with brine. The organic layer was collected, dried with MgSO4, and condensed to generate compound 2 as a clear oil (16.113 g, 95%). 1H NMR (400 MHz, CDCl3) δ 7.45 (d, J = 8.4 Hz, 2H), 7.16 (d, J = 8.4 Hz, 2H), 4.15 (q, J = 7.1 Hz, 2H), 3.56 (s, 2H), 1.25 (t, J = 7.1 Hz, 3H). ESIMS m/z [M+H]+ 243.</p><!><p>Compound 2 (1.270 g, 5.23 mmol), 5-(4,4,5,5-tetramethyl-1,3,2- dioxaborolan-2-yl)pyridin-2-amine (1 g, 4.54 mmol), and Na2CO3 (1.445 g, 13.63 mmol) were added to a 20 mL microwave vessel. The degassed 4:1 DMF/water v/v (10 mL) was then added to the reaction and the reaction was degassed with argon for 5 min. P(Cy)3 (0.057 g, 0.204 mmol) and Pd2(dba)3 (0.062 g, 0.068 mmol) were then added to the reaction vessel followed by degassing with argon for an additional 5 min. The reaction vessel was sealed and heated under microwave irradiation for 30 min at 130°C. TLC confirmed all of 5-(4,4,5,5-tetramethyl-1,3,2-dioxaborolan-2-yl)pyridin-2-amine was consumed. The solvent from the reaction was evaporated and the reaction was adsorbed onto silica and purified with flash chromatography using a gradient from 100% DCM to 40% 4:1 DCM/ MeOH to yield compound 3 as yellow crystals (1.153 g, 99%). 1H NMR (400 MHz, DMSO-d6) δ 8.21 (d, J = 2.6 Hz, 1H), 7.66 (dd, J = 8.6, 2.6 Hz, 1H), 7.49 (d, J = 8.3 Hz, 2H), 7.26 (d, J = 8.3 Hz, 2H), 6.49 (d, J = 8.6 Hz, 1H), 6.02 (s, 2H), 4.06 (q,J = 7.1 Hz, 2H), 3.64 (s, 2H), 1.17 (t, J = 7.1 Hz, 3H). ESIMS m/z [M+H]+ 257.</p><!><p>Compound 3 (1.165 g, 4.55 mmol) was added into a reaction vessel equipped with a stir bar. Ethanol (40 mL) and Na2CO3 (1.205 g, 11.36 mmol) were then added to the reaction vessel and the reaction was heated to 100°C in an oil bath. After 5 min, chloroacetaldehyde 50% solution in water (6.39 mL, 45.5 mmol) was added dropwise and the reaction was heated to 100°C for 48 h. TLC confirmed that all of starting material was consumed, the solvent was condensed and the reaction was adsorbed onto silica. The reaction was purified with flash chromatography using a gradient from 100% DCM to 40% 4:1 DCM/MeOH to yield compound 4 (1.210 g, 95%). 1H NMR (400 MHz, CDCl3) δ 8.30 (s, 1H), 7.69 (d, J = 9.3 Hz, 1H), 7.66 (d, J = 1.2 Hz, 1H), 7.63 (s, 1H), 7.52 (d, J = 8.2 Hz, 2H), 7.43 (dd, J = 9.3, 1.8 Hz, 1H), 7.40 (d, J = 8.2 Hz, 2H), 4.18 (q, J = 7.2 Hz, 2H), 3.67 (s, 2H), 1.29 (d, J = 7.2 Hz, 3H). ESIMS m/z [M+H]+ 281.</p><!><p>Compound 4 (1.2 g, 4.28 mmol) was added to 1:1 THF/Water (40 mL) in a pressure reaction vessel. LiOH (1.798 g, 21.4 mmol) was then added and the reaction was heated to 100 °C for 5 h. TLC confirmed complete consumption of compound 4. All organic solvent was evaporated and the water solution was extracted with 5× DCM and all extracts were discarded. Then, the reaction was acidified with 3 M HCl to pH ~4. The acidified aqueous solution was extracted 10× with 4:1 DCM/IPA. All extracts were combined, dried, and condensed to yield compound 5 (455 mg, 42.1%). 1H NMR (400 MHz, DMSO-d6) δ 12.33 (s, 1H), 8.97 (s, 1H), 8.02 (s, 1H), 7.73–7.64 (m, 5H), 7.40 (d, J = 7.9 Hz, 2H), 3.65 (s, 1H). ESIMS m/z [M+H]+ 253.</p><!><p>Compound 5 (40 mg, 0.159 mmol), 5-(tert-butyl)isoxazol-3-amine (33.3 mg, 0.238 mmol), EDC (60.9 mg, 0.37 mmol), HOAt (21.56 mg, 0.159 mmol), and DIPEA (0.043 mL, 0.238 mmol) were all added to a reaction vessel. DMF (1 mL) was added and the reaction was stirred at room temperature overnight. After the reaction was complete, the organic layer was condensed and the crude reaction was transferred using DCM to a silica column and purified using DCM to 30% 4:1 DCM/MeOH. The desired compound 6g was isolated in moderate yield (20.4 mg, 34.4%).1H NMR (400 MHz, CDCl3) δ 9.83 (s, 1H), 8.29 (s, 1H), 7.70 (m, 2H), 7.64 (s, 1H), 7.53 (d, J = 7.6 Hz, 2H), 7.45 (d, J = 7.6 Hz, 2H), 7.40 (d, J = 9.4 Hz, 1H), 6.76 (s, 1H), 3.84 (s, 2H), 1.35 (s, 9H). 13C NMR (101 MHz, CDCl3) δ 181.78, 169.13, 157.97, 144.70, 136.53, 133.94, 133.72, 130.09, 127.44, 126.38, 125.10, 123.04, 117.76, 112.77, 93.42, 43.61, 33.05, 28.62. ESIMS m/z [M+H]+ 375. HPLC Purity 100%.</p><!><p>Compound 6f was synthesized according to the procedure outlined for 6g (17.5 mg, 27.9%). 1H NMR (400 MHz, DMSO-d6) δ 10.54 (s, 1H), 8.88 (s, 1H), 8.09 (s, 1H), 7.94 (s, 1H), 7.78 (d, J = 8.2 Hz, 1H), 7.62 (m, 4H), 7.56–7.48 (m, 2H), 7.44 (d, J = 7.9 Hz, 2H), 7.37 (d, J = 7.7 Hz, 1H), 3.72 (s, 2H).13C NMR (101 MHz, DMSO-d6) δ 170.06, 140.35, 130.44, 130.35, 126.91, 125.18, 124.79, 124.43, 123.05, 120.02 (q, J = 3.7 Hz), 117.33, 115.55 (q, J = 3.9 Hz), 114.07, 43.33. ESIMS m/z [M+H]+ 396. HPLC Purity 98%.</p><!><p>Compound 6e was synthesized according to the procedure outlined for 6g (20.4 mg, 39.3%). 1H NMR (400 MHz, DMSO-d6) δ 10.18 (s, 1H), 8.88 (s, 1H), 7.95 (s, 1H), 7.64 (d, J = 8.2 Hz, 2H), 7.59 7.60–7.58 (m, 4H), 7.57–7.52 (m,1H), 7.44 (d, J = 8.2 Hz, 2H), 7.28 (t, J = 7.9 Hz, 2H), 7.02 (t, J = 7.4 Hz, 1H), 3.68 (s, 2H). 13C NMR (101 MHz, DMSO-d6) δ 169.41, 139.63, 135.97, 135.44, 134.10, 130.29, 129.16, 126.88, 125.23, 124.81, 124.41, 123.67, 119.55, 117.33, 114.07, 43.37. ESIMS m/z [M+H]+ 328. HPLC Purity 100%.</p><!><p>Compound 6d was synthesized according to the procedure outlined for 6g (18.8 mg, 32.8%). 1H NMR (400 MHz, DMSO-d6) δ 10.38 (s, 1H), 8.88 (s, 1H), 7.95 (s, 1H), 7.81 (d, J = 1.7 Hz, 1H), 7.63 (m, 4H), 7.54 (d, J =9.4 Hz, 1H), 7.44 (m, 3H), 7.31 (t, J =8.1 Hz, 1H), 7.08 (dd, J =8.1, 2.1 Hz,1H), 3.69 (s, 2H).13C NMR (101 MHz, DMSO-d6) δ 169.84, 141.05, 135.60, 135.54, 134.11, 133.51, 130.89, 130.32, 126.90, 125.18, 124.79, 124.44, 123.40, 118.99, 117.91, 117.34, 114.05, 43.33. ESIMS m/z [M+H]+ 362. HPLC Purity 96%.</p><!><p>Compound 6c was synthesized according to the procedure outlined for 6g (15.7 mg, 28.7%). 1H NMR (400 MHz, DMSO-d6) δ 10.41 (s, 1H), 8.88 (s, 1H), 7.94 (s, 1H), 7.64 (d, J = 8.2 Hz, 2H), 7.59 (m, 3H), 7.54 (dd, J = 9.4, 1.6 Hz, 1H), 7.43 (d, J = 8.2 Hz, 2H), 7.35–7.27 (m, 2H), 6.88–6.82 (m, 1H), 3.69 (s, 2H). 13C NMR (101 MHz, DMSO-d6) δ 169.82, 162.56 (d, J = 241.3 Hz), 141.33 (d, J = 11.1 Hz), 135.63, 135.53, 134.10, 130.82 (d, J = 9.6 Hz), 130.33, 126.90, 125.19, 124.81, 124.43, 117.33, 115.26 (d, J = 2.7 Hz), 114.08, 110.13 (d, J = 21.1 Hz), 106.30 (d, J = 26.3 Hz), 43.33. ESIMS m/z [M+H]+ 346. HPLC Purity 100%.</p><!><p>Compound 6b was synthesized according to the procedure outlined for 6g (19.2 mg, 35.5%). 1H NMR (400 MHz, DMSO-d6) δ 10.11 (s, 1H), 8.88 (s, 1H), 7.94 (s, 1H), 7.65–7.59 (m, 4H), 7.54 (dd, J = 9.4, 1.7 Hz, 1H), 7.45–7.41 (m, 3H), 7.38 (d, J = 7.8 Hz, 1H), 7.15 (t, J = 7.8 Hz, 1H), 6.83 (d, J = 7.8 Hz, 1H), 3.66 (s, 2H), 2.24 (s, 3H). 13C NMR (101 MHz, DMSO-d6) δ 169.34, 139.56, 138.32, 136.02, 135.43, 134.10, 130.26, 129.00, 126.87, 125.22, 124.81, 124.41, 124.37, 120.09, 117.33, 116.74, 114.07, 43.40, 21.62. ESIMS m/z [M+H]+ 342. HPLC Purity 100%.</p><!><p>Compound 6a was synthesized according to the procedure outlined for 6g (17.1 mg, 26.5%). 1H NMR (400 MHz, DMSO-d6) δ 10.43 (s, 1H), 8.94 (s, 1H), 8.01 (s, 2H), 7.69 (m, 4H), 7.60 (d, J = 8.8 Hz, 1H), 7.55 (d, J = 7.9 Hz, 1H), 7.48 (d, J = 8.1 Hz, 2H), 7.31 (t, J = 7.9 Hz, 1H), 7.27 (d, J = 8.1 Hz, 1H), 3.75 (s, 2H). 13C NMR (101 MHz, DMSO-d6) δ 169.82, 141.19, 135.60, 135.53, 134.11, 131.20, 130.32, 126.90, 126.29, 125.18, 124.79, 124.44, 122.00, 121.85, 118.29, 117.34, 114.04, 43.33. ESIMS m/z [M+H]+ 406. HPLC Purity 100%.</p><!><p>2-(3-hydroxyphenyl)acetic acid (15 g, 99 mmol) was added to methanol (197 mL). Sulfuric acid (3.94 mL, 73.9 mmol) was added and the reaction was heated reflux for three days. The reaction was cooled to room temperature and solid Na2CO3 was added until the pH was neutral. Solvent was evaporated and the reaction was redissolved in ether and washed 3× with water and 2× with brine. The organic layer was collected, dried, and condensed to generate compound 7 as a dark oil (14.764 g, 90%). 1H NMR (400 MHz, CDCl3) δ 7.17 (t, J = 7.8 Hz, 1H), 6.82 (d, J = 7.8 Hz, 1H), 6.77 (t, J = 1.8 Hz, 1H), 6.73 (dd, J = 8.1, 2.5 Hz, 1H), 5.56 (s, 1H), 3.70 (s, 3H), 3.58 (s, 2H). ESIMS m/z [M+H]+ 167.</p><!><p>Compound 7 (4.07 mL, 30.1 mmol) was added to DMF in a reaction vessel cooled to 0 °C. NaH 60% dispersion (1.565 g, 39.1 mmol) was added to the reaction and the reaction was stirred at 0°C for 45 min. 3,6-Dichloropyridazine (6.01 g, 39.1 mmol) was added and the reaction was raised to 80°C using an oil bath and was heated for 48 h. After the reaction was complete as judged by TLC, the solvent was evaporated and the reaction was adsorbed onto silica. The reaction was purified using flash chromatography on a silica column using a gradient from hexanes to EtOAc (50%). The desired product 8 was isolated in moderate yield (3.02 g, 36%). 1H NMR (400 MHz, CDCl3) δ 7.49 (d, J = 9.1 Hz, 1H), 7.38 (t, J = 7.9 Hz, 1H), 7.21–7.14 (m, 3H), 7.11 (ddd, J = 8.2, 2.4, 0.9 Hz, 1H), 3.70 (s, 3H), 3.65 (s, 2H). ESIMS m/z [M+H]+ 279.</p><!><p>Compound 8 (2.768 g, 9.93 mmol), diphenylmethamine (1.980 g, 10.93 mmol), and potassium phosphate (10.53 g, 49.7 mmol) were added to dioxane (40 mL) and the reaction was degassed with argon for 10 min. Xantphos (0.287 g, 0.497 mmol) and Pd2(dba)3 (0.454 g, 0.497 mmol) were added to the reaction. The reaction was then sealed under positive argon pressure and heated under reflux for 12 h. After the overnight, the reaction was condensed and adsorbed onto silica. The reaction was purified with a silica column using flash chromatography with a hexane/EtOAc gradient to yield compound 9 (1.02 g, 24.5%). 1H NMR (400 MHz, CDCl3) δ 7.82 (d, J = 7.6 Hz, 2H), 7.52 (t, J = 7.3 Hz, 1H), 7.43 (t, J = 7.6 Hz, 2H), 7.31 (m, Hz, 4H), 7.19 (d, J = 6.5 Hz, 2H), 7.11–7.06 (m, 2H), 7.06–7.02 (m, 1H), 6.96 (d, J = 9.1 Hz, 1H), 6.88 (d, J = 9.1 Hz, 1H), 3.69 (s, 3H), 3.61 (s, 2H). ESIMS m/z [M+H]+ 424.</p><!><p>Compound 9 (1.02 g, 2.409 mmol) was dissolved in methanol (100 mL) and was cooled to 0 °C. 12 M HCl (2.007 mL, 24.09 mmol) was added to the reaction and stirred for 30 min at 0 °C. The reaction was removed from the ice bath and stirred at room temperature for 12 h. The reaction was purified with flash chromatography using a silica column and a gradient from DCM to 50% 4:1 DCM/MeOH. Compound 10 was isolated as a sticky-yellow oil (456 mg, 73%). 1H NMR (400 MHz, DMSO-d6) δ 7.30 (t, J = 7.8 Hz, 1H), 7.09 (d, J = 9.3 Hz, 1H), 7.02 (d, J = 7.8 Hz, 1H), 6.96–6.89 (m, 3H), 6.20 (s, 2H), 3.66 (s, 2H), 3.59 (s, 3H). ESIMS m/z [M+H]+ 260.</p><!><p>Compound 11 was synthesized according to the procedure outlined for compound 4 (290 mg, 98%). 1H NMR (400 MHz, CDCl3) δ 7.95 (d, J = 9.7 Hz, 1H), 7.71 (s, 1H), 7.64 (s, 1H), 7.40 (t, J = 7.8 Hz, 1H), 7.19 (d, J = 7.8 Hz, 1H), 7.17–7.12 (m, 2H), 6.90 (d, J = 9.7 Hz, 1H), 3.71 (s, 3H), 3.67 (s, 2H). ESIMS m/z [M+H]+ 284. HPLC Purity 70%.</p><!><p>Compound 12 was synthesized according to the procedure outlined for 5(451.4 mg, 95%).1H NMR (400 MHz, DMSO-d6) δ 12.41 (s,1H), 8.14 (d, J = 9.6 Hz, 1H), 8.04 (s, 1H), 7.63 (s, 1H), 7.42–7.34 (m, 1H), 7.16–7.15 (m, 3H), 7.08 (d, J = 9.6 Hz,1H), 3.61 (s, 2H). ESIMS m/ z [M+H]+ 270.</p><!><p>Compound 13g was synthesized according to the procedure outlined for 6g (14.7 mg, 33.7%).1H NMR (400 MHz, CDCl3) δ 9.70 (s, 1H), 7.93 (d, J = 9.6 Hz, 1H), 7.70 (s, 1H), 7.63 (s, 1H), 7.42 (t, J = 8.0 Hz), 7.30–7.24 (m, 2H), 7.16 (dd, J = 8.0, 2.2 Hz, 1H), 6.87 (d, J = 9.6 Hz, 1H), 6.72 (s, 1H), 3.81 (s, 2H), 1.31 (s, 9H). 13C NMR (101 MHz, CDCl3) δ 181.79, 168.69, 159.53, 157.90, 153.43, 135.79, 133.05, 130.26, 128.17, 126.50, 122.05, 120.16, 117.35, 111.47, 93.40, 43.83, 33.02, 28.57. ESIMS m/z [M+H]+ 392. HPLC Purity 100%.</p><!><p>Compound 13f was synthesized according to the procedure outlined for 6g (17.2 mg, 37.4%).1H NMR (400 MHz, CDCl3) δ 8.28 (s, 1H), 7.87 (d, J = 9.6 Hz, 1H), 7.79 (s, 1H), 7.71 (d, J = 7.9 Hz, 1H), 7.63–7.61 (m, 2H), 7.44 (t, J = 7.9 Hz, 1H), 7.39 (t, J = 7.9 Hz, 1H), 7.33 (d, J = 7.6 Hz, 1H), 7.23 (d, J = 7.6 Hz, 1H), 7.21–7.14 (m, 2H), 6.86 (d, J = 9.6 Hz, 1H), 3.77 (s, 2H). 13C NMR (101 MHz, Chloroform-d) δ 168.91, 159.59, 153.48, 138.38, 137.36, 136.18, 133.07, 131.29 (q, J = 32.4 Hz), 130.40, 129.50, 128.09, 126.52, 122.87, 121.98, 120.94, 120.20, 117.31, 116.75–116.26 (m), 111.69, 44.23. ESIMS m/z [M+H]+ 413. HPLC Purity 100%.</p><!><p>Compound 13e was synthesized according to the procedure outlined for 6g (14.4 mg, 37.5%).1H NMR (400 MHz, CDCl3) δ 7.87 (d, J = 9.6 Hz, 2H), 7.61 (d, J = 6.0 Hz, 2H), 7.47 (d, J = 8.0 Hz, 2H), 7.41 (d, J = 8.0 Hz, 1H), 7.28–7.19 (m, 4H), 7.15 (d, J = 8.0 Hz, 1H), 7.08 (t, J = 7.4 Hz, 1H), 6.85 (d, J = 9.6 Hz, 1H), 3.74 (s, 2H). 13C NMR (101 MHz, CDCl3) δ 168.60, 159.59, 153.47, 137.76, 137.40, 136.60, 133.06, 130.33, 128.95, 128.12, 126.56, 124.49, 122.00, 120.04, 119.89, 117.24, 111.59, 44.32. ESIMS m/z [M+H]+ 345. HPLC Purity 100%.</p><!><p>Compound 13d was synthesized according to the procedure outlined for 6g (19.4 mg, 41.1%).1H NMR (400 MHz, CDCl3) δ 8.03 (s, 1H), 7.88 (d, J = 9.6 Hz, 1H), 7.72 (s, 1H), 7.62 (s, 2H), 7.45–7.40 (m, 2H), 7.24–7.05 (m, 4H), 6.86 (d, J = 9.6 Hz,1H), 3.74 (s, 2H).13C NMR (101 MHz, CDCl3) δ 168.71, 159.59, 153.49, 139.07, 137.39, 136.24, 133.07, 130.54, 130.24, 128.14, 128.07, 127.41, 126.52, 122.86, 121.98, 120.18, 118.28, 117.28, 111.64, 44.46. ESIMS m/z [M+H]+ 423. HPLC Purity 100%.</p><!><p>Compound 13c was synthesized according to the procedure outlined in for 6g (16 mg, 37.9%). 1H NMR (400 MHz, CDCl3) δ 8.00 (s, 1H), 7.88 (d, J = 9.6 Hz, 1H), 7.62 (s, 2H), 7.58 (t, J = 1.8 Hz, 1H), 7.43 (t, J = 7.8 Hz, 1H), 7.34 (d, J = 8.0 Hz, 1H), 7.24–7.13 (m, 4H), 7.06 (d, J = 8.0 Hz, 2H), 6.86 (d, J = 9.6 Hz, 1H), 3.74 (s, 2H). 13C NMR (101 MHz, CDCl3) δ 168.70, 159.59, 153.50, 138.93, 137.39, 136.25, 134.58, 133.07, 130.55, 129.95, 128.11, 126.52, 124.48, 121.98, 120.19, 119.89, 117.78, 117.26, 111.63, 44.47. ESIMS m/z [M+H]+ 379. HPLC Purity 100%.</p><!><p>Compound 13b was synthesized according to the procedure outlined for 6g(20.4 mg, 51.1%). 1H NMR(400 MHz, CDCl3) δ 7.88 (d, J = 9.6 Hz, 1H), 7.82 (s, 1H), 7.61 (d, J = 5.5 Hz, 2H), 7.42 (t, J = 7.8 Hz, 1H), 7.33 (s, 1H), 7.25–7.19 (m, 3H), 7.17–7.13 (m, 2H), 6.90 (d, J = 7.5 Hz, 1H), 6.85 (d, J = 9.6 Hz, 1H), 3.73 (s, 2H), 2.28 (s, 3H). 13C NMR (101 MHz, CDCl3) δ 168.57, 159.60, 153.46, 138.89, 137.67, 137.40, 136.65, 133.06, 130.45, 128.76, 128.12, 126.56, 125.30, 122.01, 120.53, 120.02, 117.23, 116.96, 111.58, 44.54, 21.43. ESIMS m/z [M+H]+ 359. HPLC Purity 100%.</p><!><p>Compound 13a was synthesized according to the procedure outlined for 6g (21.0 mg, 52.0%). 1H NMR (400 MHz, CDCl3) δ 8.19 (s, 1H), 7.88 (d, J = 9.6 Hz, 1H), 7.66–7.60 (m, 2H), 7.48–7.40 (m, 2H), 7.24–7.11 (m, 5H), 6.86 (d, J = 9.6 Hz, 1H), 6.78 (t, J = 8.3 Hz, 1H), 3.74 (s, 2H). 13C NMR (101 MHz, Chloroform-d) δ 168.75, 162.89 (d, J = 244.9 Hz), 159.59, 153.46, 139.40 (d, J = 10.8 Hz), 137.39, 136.34, 133.04 (d, J = 9.7 Hz), 130.35, 130.00 (d, J = 8.6 Hz), 128.12, 126.52, 121.95, 120.11, 117.30, 114.99 (d, J = 2.8 Hz), 111.26, 111.20, 107.28 (dd, J = 26.7, 7.4 Hz), 44.27. ESIMS m/z [M+H]+ 363. HPLC Purity 100%.</p><!><p>Kinase activity was measured in a microfluidic assay that monitors the separation of a phosphorylated product from substrate. The assay was run using a 12-sipper chip on a Caliper EZ Reader II (PerkinElmer®, Walthman, USA) with separation buffer (100 mM HEPES, 10 mM EDTA, 0.015% Brij-35, 0.1% CR-3 [PerkinElmer®, Walthman, USA]). In 96-well polypropylene plates (Greiner, Frickenhausen, Germany) compound stocks (20 mM in DMSO) were diluted into kinase buffer (50 mM HEPES, 0.075% Brij-35, 0.1% Tween 20,2 mM DTT, 10 mM MgCl2, and 0.02% NaN3) in 12-point ½log dilutions (2 mM–6.32 nM). After, 1 µL was transferred into a 384-well polypropylene assay plate (Greiner, Frickenhausen, Germany). The RET enzyme (Invitrogen™, Grand Island, USA) was diluted in kinase buffer to a concentration of 2 nM and 5 µL of the enzyme mixture was transferred to the assay plate. The inhibitors/ RET enzyme were incubated for 60 min with minor shaking. A substrate mix was prepared containing ATP (Ambresco®, Solon, USA) and 5FAM tagged RET peptide (peptide #22, 5' FAM-EPLYWSFPA, PerkinElmer®, Walthman, USA) dissolved in kinase buffer, and 5 µL of the substrate mix was added to the assay plate. Running concentrations were as follows: ATP (9 µM), peptide (1.5 µM), compound 12-point ½log dilutions (0.2 mM–0.632 nM). For positive control, no inhibitor was added. For negative control, no enzyme was added. For running control, vandetanib (LC Laboratories®, Lot# BTB-105) was utilized. The plate was run until 10–20% conversion based on the positive control wells. The following separation conditions were utilized: upstream voltage −500 V; downstream voltage, −1900 V; chip pressure −0.8. Percent inhibition was measured for each well comparing starting peptide to phosphorylated product peaks relative to the baseline. Dose response curves, spanning the IC50 dose, were generated in GraphPad Prisim 6 and fit to an exponential one-phase decay line and IC50 values were obtained from the half-life value of the curve. IC50 values were generated in duplicate and error was calculated from the standard deviation between values.</p><!><p>The assay was completed identically to the RET Biochemical Inhibition Assay, but compounds were incubated with RET for 60, 20, and 3 min before addition of the substrate mix. The IC50 verses incubation time graph was fit to a liner equation (y = mx + b) in GraphPad Prisim 6 and extrapolated to determine 0 min incubation.</p><!><p>A VEGFR-2 DFG-out crystal structure [36] and the amino acid sequence of RET [37] was obtained. Using SwissModel Automatic Modelling Mode (swissmodel.expassy.org) [33–35], the RET amino acid sequence was employed to build a RET DFG-out homology model using the VEGFR2 DFG-out structure as a template. The resulting RET DFG-out homology model displayed the predicated shift in the DFG-out motif and harbored a newly accessible allosteric pocket.</p><!><p>Computational modeling studies were completed using Auto-Dock Vina [38], AutoDock Tools, and Discovery Studio 3.5. Using AutoDock Tools, the RET homology model was prepared as follows: 1) All hydrogens were added as 'Polar Only' 2) A grid box for the ATP binding site was created (center x = −25.881, center y = 9.55, center z = −10.927/size x = 16, size y = 44, size z = 18). Compounds to be computationally modeled were assigned appropriate rotatable bonds using AutoDock Tools. To computational model the compounds, AutoDock Vina [38] was employed. After the modeling study, the results were visualized and analyzed with Discovery Studio 3.5.</p><!><p>RAT1 fibroblasts expressing RET/C634R point mutant cells have been described previously and were grown in RPMI 1640 supplemented with 10% fetal bovine serum (FBS) (Invitrogen, Carlsbad, CA, USA) [40].</p><!><p>Immunoblotting experiments were performed according to standard procedures. Cells were harvested in lysis buffer (50 mM Hepes, pH 7.5, 150 mM NaCl, 10% glycerol, 1% Triton X-100, 1 mM EGTA, 1.5 mM MgCl2, 10 mM NaF, 10 mM sodium pyrophosphate, 1 mM Na3VO4, 10 µg of aprotinin/ml, 10 µg of leupeptin/ml) and clarified by centrifugation at 10,000× g. Protein concentration was estimated with a modified Bradford assay (Bio-Rad Laboratories, Berkeley, CA, USA). Antigens were revealed by an enhanced chemiluminescence detection kit (ECL, Amersham Pharmacia Biotech). Anti-RET is a polyclonal antibody raised against the tyrosine kinase protein fragment of human RET. Anti-pTyr905 and anti-pTyr1062 are phospho-specific affinity-purified polyclonal antibodies that recognize RET proteins phosphorylated at Y905 and Y1062, respectively. [17]</p>

PubMed Author Manuscript

Stereochemical investigations on the biosynthesis of achiral (Z)-γ-bisabolene in Cryptosporangium arvum

A newly identified bacterial (Z)-γ-bisabolene synthase was used for investigating the cyclisation mechanism of the sesquiterpene. Since the stereoinformation of both chiral putative intermediates, nerolidyl diphosphate (NPP) and the bisabolyl cation, is lost during formation of the achiral product, the intriguing question of their absolute configurations was addressed by incubating both enantiomers of NPP with the recombinant enzyme, which resolved in an exclusive cyclisation of (R)-NPP, while (S)-NPP that is non-natural to the (Z)-γ-bisabolene synthase was specifically converted into (E)-β-farnesene. A hypothetical enzyme mechanistic model that explains these observations is presented.

stereochemical_investigations_on_the_biosynthesis_of_achiral_(z)-γ-bisabolene_in_cryptosporangium_ar

<!>Introduction<!><!>Functional characterisation of a bacterial (Z)-γ-bisabolene synthase<!><!>Functional characterisation of a bacterial (Z)-γ-bisabolene synthase<!><!>The absolute configuration of the intermediates nerolidyl diphosphate and the bisabolyl cation<!><!>The absolute configuration of the intermediates nerolidyl diphosphate and the bisabolyl cation<!><!>Conclusion<!>Supporting Information<!>

<p>This article is part of the thematic issue "Reactive intermediates – carbocations".</p><!><p>Given the enormous impact of chirality within biomolecules for all forms of life, it is fascinating to see how nature is able to maintain and reproduce stereochemical information. This concept largely involves the introduction of stereocentres to achiral starting materials by the action of enzymes. While reactions fulfilling this category are still challenging within synthetic chemistry, and methods managing to reach this goal are desperately desired, in the enzymatic world with its completely chiral environment these transformations are ubiquitous, which diminishes the hard border between achiral and chiral. One intriguing example for this kind of reactivity is represented by terpene synthases (TSs), arguably building up the class of natural products with the highest density of stereochemical information, the terpenes. By providing a defined cavity including its molecular coating together with binding and activation of the diphosphate (OPP) moiety, these enzymes convert simple achiral oligoprenyl diphosphates into often complex, polycyclic hydrocarbons or alcohols with introduction of multiple stereocentres in just one enzymatic step [1–3]. With this approach, nature makes perfect use of the versatile chemistry of carbocations with its hydride or proton shifts and Wagner–Meerwein rearrangements leading to a large variety of possible structures. Among terpenoid natural products, achiral compounds are rarely found, but still present. In this group, there are acyclic compounds like the linear sesquiterpene (E)-β-farnesene (1, Figure 1), which is known as an alarm pheromone in aphids [4–5], but also monocyclic terpenes like α-humulene (2), a widely occurring sesquiterpene in many essential oils [6–7]. Whereas the stereochemical imprint of a TS on achiral products is not directly visible, there still can be a chiral cyclisation cascade behind these terpenes. This is true as well for examples featuring a mirror plane like the monoterpene 1,8-cineol (eucalyptol, 3), for which the absolute configuration of the intermediary terpinyl cation has been investigated using deuterium labelling, demonstrating different stereochemical courses in the plant Salvia officinalis [8–9] and in the bacterium Streptomyces clavuligerus [10]. Also the highly unusual methylated sesquiterpene sodorifen (4) possesses a mirror plane [11] making any labelling experiment hard to interpret and is nevertheless most likely biosynthesised through chiral intermediates [12]. For these cases, it is a particular challenge to uncover the stereochemical information hidden behind the achiral product structure. In this study, we addressed the chiral intermediates in the biosynthesis of the achiral sesquiterpene (Z)-γ-bisabolene (5) by a TS from a soil bacterium.</p><!><p>Structures of achiral terpenes: (E)-β-farnesene (1), α-humulene (2), 1,8-cineol (3) and sodorifen (4).</p><!><p>Within our efforts to characterise bacterial TSs with new functions and mechanisms, a TS (WP_035857999) from the soil actinomycete Cryptosporangium arvum DSM 44712 was cloned into the E. coli expression vector pYE-express [13] (Table S1, Supporting Information File 1), because of its phylogenetic distance to characterised TSs (Figure S1, Supporting Information File 1). The amino acid sequence of the enzyme features known conserved motifs both for binding [14] and activation [15] of the diphosphate moiety together with structurally important residues [16–17] (Figure S2, Supporting Information File 1). For in vitro activity testing, the enzyme was expressed in E. coli BL21(DE3), purified (Figure S3, Supporting Information File 1) and incubated with the common terpene precursors geranyl- (GPP, C10), farnesyl- (FPP, C15), geranylgeranyl- (GGPP, C20) and geranylfarnesyl (GFPP, C25) diphosphate. With hexane extraction and GC–MS analysis, only the incubation with FPP yielded a terpene product (Figure 2) that was isolated and identified by one- and two dimensional NMR spectroscopy (Table S2, Supporting Information File 1), EIMS databases and GC retention index as the known sesquiterpene (Z)-γ-bisabolene (5). Because the two olefinic carbon atoms of its quaternary double bond could not be unambiguously assigned from HMBC data, labelling experiments with (6-13C)- and (7-13C)FPP [18] were also conducted (Figure S4, Supporting Information File 1). These results characterise the TS from C. arvum as a (Z)-γ-bisabolene synthase (BbS).</p><!><p>A) Total ion chromatogram of a hexane extract from the incubation of FPP with BbS and B) EI mass spectrum of the main product identified as (Z)-γ-bisabolene (5).</p><!><p>The achiral, monocyclic sesquiterpene 5 is abundant in many essential oils and was reported from different sources such as the liverwort Dumortiera hirsuta [19]. Its (Z)-configured exocyclic double bond has also attracted the attention of synthetic chemistry for a diastereoselective total synthesis [20–22]. The wide occurrence of 5 is likely connected to the simple biosynthesis from FPP featuring the common bisabolyl cation (A) as an intermediate after 1,6-cyclisation (Scheme 1). For this cyclisation, a formal isomerisation of the (E)-configured double bond in FPP to the (Z)-configured double bond in A is needed. To address this problem, a 1,3-suprafacial transposition of OPP to nerolidyl diphosphate (NPP) is usually assumed [23]. This tertiary allylic diphosphate can undergo 1,6-cyclisation after a rotation around the C-2,C-3 single bond. Both NPP and A are chiral which raises the question of the active enantiomers in the BbS-catalysed reaction. This problem is challenging since the stereoinformation is destroyed in the final deprotonation step, which prevents any conclusion at the product stage, e.g., by use of enantioselectively labelled substrates [8–10]. If the nucleophilic attack of the C-6,C-7 double bond at the allylic system proceeds with an anti stereochemistry (anti-SN2' reaction), which is favoured for a concerted process and is also discussed for other cyclisation mechanisms [24–26], the four theoretically possible options for the BbS cyclisation mechanism are narrowed down to two possibilities: Either the reaction takes place via (R)-NPP resulting in (S)-A after ring closure (Scheme 1, path A) or via (S)-NPP, which would suggest involvement of (R)-A (Scheme 1, path B). This stereochemical link between NPP and A was also observed in a theoretical docking study with epi-isozizaene synthase suggesting (S)-NPP and (R)-A to be included in its cyclisation mechanism [27].</p><!><p>Cyclisation mechanism to 5 involving either the intermediates (R)-NPP and (S)-A (path A) or (S)-NPP and (R)-A (path B). Numbering of carbons in 5 reflect their origin in FPP.</p><!><p>To address this question experimentally, (R)- and (S)-NPP were synthesised following a known route for enantioselective preparation of nerolidol [28] by Sharpless epoxidation of farnesol in analogy to the reported synthesis of geranyllinaloyl diphosphates [29] (Scheme S1, Supporting Information File 1). Aiming for an easy and unambiguous interpretation of the incubation experiments, the synthesised nerolidol samples (showing moderate to good ee values as judged by Mosher ester analysis of the preceeding epoxides, Figure S5, Supporting Information File 1) were purified by preparative HPLC on a chiral stationary phase to >99% ee for both samples (Figure S6, Supporting Information File 1), before converting them into the NPPs. For comparison, also racemic NPP was synthesised by a Grignard reaction of geranylacetone with vinylmagnesium bromide. The two NPP samples featuring a well-defined stereocentre, and (rac)-NPP, were incubated with recombinant BbS, the experiments were extracted with hexane and analysed by GC–MS (Figure 3). A selective product formation was observed for the two enantiomers of NPP, which is surprising in the light of the fact that these reactive tertiary allylic diphosphates were often found to result in a complex mixture of terpene cyclase products and Mg2+-catalysed spontaneous hydrolysis products for other TSs [29–30]. While the reaction with (R)-NPP leads to BbS's native product 5, for (S)-NPP formation of the acyclic elimination product (E)-β-farnesene (1) was observed, which was identified by EI mass spectral library and GC retention index (I = 1460 (HP-5MS), lit: I = 1459 (HP-5MS) [31]). Incubation with (rac)-NPP resulted in a nearly 1:1 mixture of both products, showing that both enantiomers of NPP were converted with similar efficiency.</p><!><p>Total ion chromatograms of hexane extracts from incubation experiments with BbS and A) (R)-NPP, B) (S)-NPP and C) (rac)-NPP.</p><!><p>These results clearly rule out (S)-NPP, but rationalise (R)-NPP as an intermediate in the cyclisation mechanism of BbS, and are in favour of the (S)-bisabolyl cation (A) to be deprotonated to 5 within the cascade reaction (Scheme 1, path A) [24–26], although the stereochemical link between NPP and A could not be demonstrated experimentally in this study. The formation of (R)-NPP from FPP as a 1,3-syn-allylic rearrangement can be rationalised in a binding mode of FPP, in which OPP is located on a defined face of the C-2,C-3-double bond by the enzyme's active site (e.g., on top of it, Figure 4A). This migration of OPP to C-3 results in a reorganisation of the resulting structure to a cisoid conformation for the follow up 1,6-ring closure. To explain the astonishing selectivity between the two NPP enantiomers by BbS, different NPP conformations inside the chiral environment of the active site in BbS have to be assumed (Figure 4B + 4C). The architecture of the active site may stay the same in both cases, so a fixed OPP moiety with binding by the trinuclear Mg2+ cluster and a comparable folding of the isoprenoid chain in both cases is reasonable. Therefore, the two smallest substituents at the stereocentre formally change their places for the two enantiomers of NPP, representing a minor structural change of the substrate that can be tolerated in the active site. While the binding of (R)-NPP leads to a productive conformation that exhibits a close proximity between C-6 and C-1 for ring closure to (S)-A initiated by OPP abstraction, (S)-NPP cannot occupy this conformation for its different stereocentre. Instead, abstraction of the OPP moiety leads to an allylic cation with an unproductive conformation for further ring closure and is thus quenched by abstraction of a proton, presumably by participation of the diphosphate nearby, to give 1.</p><!><p>Hypothetical BbS active site comparable conformational folds of A) FPP, B) (R)- and C) (S)-NPP explaining the selective formation of 5 from (R)-NPP by ring closure via (S)-A and of 1 from (S)-NPP by hampering ring closure leading to elimination.</p><!><p>During the course of this work, a new TS from C. arvum was characterised as a (Z)-γ-bisabolene (5) synthase (BbS). Despite its monocyclic achiral structure, the biosynthesis of 5 proceeds via two chiral intermediates, NPP and the bisabolyl cation (A), whereas the absolute configuration of the first was addressed experimentally by the synthesis and in vitro incubation of both enantiomers of NPP. These experiments clearly showed the involvement of (R)-NPP in the BbS-catalysed reaction, whereas diphosphate was selectively eliminated from (S)-NPP by BbS to yield 1. The selectivity is understandable by the fixed, chiral active site architecture of BbS promoting ring closure only for (R)-NPP. In future studies, this experimental approach will not only provide insights into the stereochemical identity of intermediates in cases of achiral terpenes inhibiting any conclusion from the product structure as shown here, but will also deepen our knowledge of general NPP utilisation by sesquiterpene synthases. The chirality of this tertiary diphosphate is currently largely underinvestigated in the characterisation of TSs, even for cascades requiring its involvement.</p><!><p>Experimental details of culture conditions, gene cloning, protein purification, incubation experiments, isolation of 5 and HPLC purifications, the amino acid sequence of BbS, a phylogenetic tree with the location of BbS, SDS-PAGE analysis, listed NMR data of 5, labelling experiments for NMR assignment, synthetic procedures for the NPPs, Mosher ester analysis of epoxides, and chiral GC analysis of nerolidols.</p><!><p>Experimental part.</p>

PubMed Open Access

The "Decisive" Role for Secondary Coordination Sphere Nucleophiles on Hydrogen Atom Transfer (HAT) Reactions: Does it Exist and What is its Origin?

Although it has been reported that some radical reactions are possibly promoted by external ions, the origin of this phenomenon is unclear. In this work, several hydrogen atom transfer (HAT) reactions in the presence of anions were studied by density functional theory (DFT) calculations, electronic structure analysis and other methods, and it is concluded that both the electrostatic interaction and polarization of the transition state (TS) by the electric field generated by anions play a fundamental role in the TS stabilization effect, whereas the "charge shift bonding" that was previously presumed to be a major contributor is ruled out. Although the stabilization toward TSs in terms of electronic energy (and thus enthalpy) is significant, it should be noted that the effect is almost completely cancelled by entropy and solvation, and further cancelled by the formation of stable resting states. Thus there is still a long way for this effect to be used in actual catalysis.

the_"decisive"_role_for_secondary_coordination_sphere_nucleophiles_on_hydrogen_atom_transfer_(hat)_r

Introduction<!>Computational Methods<!>Results and Discussions<!>Conclusion<!>Supporting Information

<p>The "electrostatic catalysis" or "salt effect" is a long-standing and well-established concept. Early in 1990s, the catalytic effect of ions that seem inert at the first glance toward organic chemical transformations has been studied by Craig Wilcox [1][2][3] . Later on, the promotion of cobalt-carbon bond dissociation by a nearby charge was found in a biochemistry-related Vitamin B complex 4 . In the recent years, the catalytic effect of charged groups toward Diels-Alder reaction was studied by Michelle Coote 5,6 , and Kendall Houk 7 . The catalytic effect of charged groups is believed to have an electrostatic nature, proceeding through interaction between the dipolar moment of transition states (TSs) and the electric field generated by nearby charges, and thus is closely relevant to the external electric field effect in chemical reactions, which has been documented extensively in many cases [8][9][10][11][12][13] . It is noteworthy that hydrogen atom transfer (HAT) reactions have also been reported to be affected by metal ions and ligands [14][15][16][17] , which is believed to be a field-induced phenomenon (chargeinduced catalysis).</p><p>On the other hand, however, Thomas Cundari and coworkers reported that external anions provide "decisive" stabilization to the TS for the hydrogen atom transfer (HAT) reaction between methane and hydroperoxyl radical very recently 18 . The authors concluded that the interaction between anions and the HAT TS is due to "charge shift bonding 19,20 " originating from a 2-center-3electron interaction, which is a brand new explanation for the influence of external ions. Thus it is interesting how much role charge shift bonding plays in the reported reactions, and also in other examples that were previously believed to be field-originated. In this work, we conducted a more detailed investigation on the "salt effect" for HAT reactions, which will provide new understanding toward this long-standing concept.</p><!><p>The geometry optimization of all structures were performed with the Gaussian 16 program 21 , at M11/6-311+G(d,p) level [22][23][24][25][26] , if not specially mentioned. DLPNO-CCSD(T) calculations were carried out with the ORCA 4.2 program 27,28 , in combination with the aug-cc-pVTZ basis set [29][30][31] . All electronic structure analysis, including but not limited to bond critical point (BCP) properties, electron localization function (ELF), electron density Laplacian, were performed using the Multiwfn program 32 , based on the wavefunction obtained at M11/ma-def2-TZVPP level 33 . The GAMESS-US program 34 was employed for LMO-EDA calculations 35 . It is found that the GAMESS-US program gave wrong results with the M11 functional, and thus M06-2X/6-311+G(d,p) level 36 was selected to perform energy decomposition with the M11-optimized geometry.</p><p>The SMD implicit solvation model 37 was used for calculations with solvation effect, and the solvation free energies were obtained by G(M05-2X 38 /6-31G(d), with SMD(DMSO)) -G(M05-2X/6-31G(d), gas phase). The final Gibbs free energies were obtained by the sum of DLPNO-CCSD(T) single point energy, M11/6-311+G(d,p) correction to free energy, and solvation free energy (the last term only for calculations in DMSO). Particularly, the solvation free energies for Cl -, Brand proton are taken from experimental reports 39 , and those for F -, HOand HSwere derived from experimental pKa of HF, H2O and H2S in DMSO, which is 15.0 40 , 31.4 41 and 13.7 42 respectively.</p><!><p>Although the meta-GGA M06-L functional 43 was employed in Cundari's report, it was found from a benchmark study involving M06-L, B3LYP-D3BJ 44,45 , M06-2X, wB97xD 46 and M11 that the performances of density functionals parallel their Hartree-Fock (HF) components (Table S1), and thus the range-separated functional M11 with a large HF component was chosen to be the functional used for geometry optimization in this work. The TSs were located for the HAT reactions between methane and hydroperoxyl radical (Scheme 1a), in the presence of various anions X -. The energetics and optimized C-X bond lengths at DLPNO-CCSD(T)/aug-cc-pVTZ//M11/6-311+G(d,p) level are listed in Table 1. Note that at this stage only the TSs, but not preactivation complexes or any other resting states are discussed. The full Gibbs free energy surface will be discussed later. Scheme 1. The reactions studied in this work, and the definitions for some quantities discussed. 1 that all anions provide significant stabilization (i.e. negative E) to TS1 in terms of electronic energy, although largely cancelled by entropy. The second-row anions, fluoride and hydroxyl anion, are among the most stabilizing ones, whereas the HCOOwith delocalized negative charge exhibits much less stabilization. Interestingly, the heavier anions, Cl -, Brand HSprovide similar stabilization energy at ~8.5 kcal/mol, while the E value becomes surprisingly much more negative upon replacement of the hydrogen in HSwith the methyl group. Overall, it is concluded that the combination of anions to TS1 is exothermic in most cases, and next we are about to figure out the reason.</p><p>As suggested by Cundari's work, we firstly examined the existence of charge shift bonding, which originates from the interplay between two resonance structures shown in Figure 1a. It has been proposed in the early research on the charge shift bonding in silicon-halogen bonds that resonance structures close in energy might lead to larger resonance energy 47 , and thus stronger charge shift bonding. Herein the energy difference between resonance structure 1 and 2 could simply be characterized by the difference in the vertical electron affinities (EA shown in Scheme 1) for the anion and the "bared" TS1(X=none). In addition, the magnitude of the involvement of 2 is reflected by the spin population on X. It is expected that a EA close to zero, as well as large spin population on X, should indicate strong charge shift interaction. The spin population values are listed in Table 1, and it is seen that for most anions the spin population on Xis negligible, except for OH -, HSand MeS -, which also exhibit EAs closest to 0. Furthermore, the TS stabilization energy E correlates with both EA and spin population on X terribly. Thus it is questionable whether charge shift interaction could explain the observed energy change. (f) The ELF contour for F2 molecule, a molecule with typical charge shift bonding. (g) The RDG isosurface for TS1(X=HO -), in which weak noncovalent interaction is shown in blue (stronger ones) and green (weaker ones).</p><p>Traditionally the existence of charge shift bonding is characterized by the properties at bond critical point (BCP), such as the electron density, electron density Laplacian, and electron localization function (ELF). Typical charge shift bonds, such as that in F2 molecule, exhibit large electron density, positive Laplacian, and slightly accumulated ELF at BCP. The electron densities at the BCP located between the carbon atom and anion are recorded in Table 1, and all TSs exhibit negligible electron densities at BCPs. The contours for ELF and electron density Laplacian for TS1(X=HO -) clearly show that there is no accumulation of ELF, and only near-zero Laplacian in the C-X interatomic region (Figure 1d and e). Furthermore, the reduced density gradient (RDG) analysis 48 , a method that directly shows the region with noncovalent weak interaction and filters covalent interaction, directly shows that the C-X interaction lies in the region of weak interaction, even with OHas the anion. All these results indicate that there is negligible chemical bonding between C atom and anions. The ELF, Laplacian and RDG analysis results for TS1(X=MeS -) are shown in Figure S3, and there is no different conclusion. However, it still cannot be concluded that there is no charge shift interaction that contributes to the TS stabilization, although charge shift bonding has been ruled out. Next, however, we will focus on alternative ways to explain the observed energy change, and then reconsider the existence of charge shift interaction.</p><p>There are three possible types of interaction that may contribute to the TS stabilization besides charge shift bonding, namely hydrogen bonding, intrinsic dipole-anion interaction, and polarization effect due to the electric field generated by anion. It has been reported by Tian Lu et al that the electron density at the BCP between a hydrogen acceptor and the corresponding hydrogen atom is a good indicator of hydrogen bond strength 49 . Unfortunately, in all TSs studied there is no BCP between X and H atoms. Thus we turned to core-valence bifurcation (CVB) 50 , another well-known hydrogen bonding indicator. The CVB values for all anions except HSand OHindicate very weak hydrogen bonding (Table S2). Based on these observations, it is proposed that hydrogen bonding contributes only little to the total E.</p><p>In order to clarify how much role electric field plays in the TS stabilization, a uniform electric field model was employed (Figure 2). In this model, the bared TS1(X=none) is placed in a uniform electric field simulating the influence of external anions, and an "effective" field strength is defined by the electric field at the midpoint of TS1(X=none) along z-axis. TSs were re-optimized in the presence of varied external field (Figure 2b) and the stabilization energies of external field along zaxis at M11/6-311+G(d,p) level exhibit quadratic correlation with field strength. Interestingly, field along y-axis, which is perpendicular to the hydrogen atom transfer path (and also the intrinsic dipole moment of TS1), affords stabilization effect in similar magnitude, and thus it seems that the stabilization originates from polarization, but not intrinsic dipole interaction. The effective field strength resulted from each anion is calculated from the restrained electrostatic potential (RESP) atomic charge 51 on the electronegative atom, and the E predicted using the fitted relationship in Figure 2b is in very good consistence with the M11-calculated E in the presence of anions (since the relationship is fitted with M11 data, it is considerable that it is more comparable with Es at M11 level), with only X=MeSas an exception. Since the field model above is able to provide pretty good explanation for the observed E, the contribution of charge shift interaction is further precluded, and it is suggested that it is the "field effect" that plays the major role in the TS stabilization. Further investigation on the role of spin population delocalization is performed for the MeScase (Figure 2d). Upon re-optimization of TSs at elongated C-S distance, the spin population on S atom drops to zero rapidly, and the energy raises by about 2.5 kcal/mol. It is then concluded that the radical delocalization, or "charge shift interaction", contributes around 2.5 kcal/mol for the stabilization in TS1(X=MeS -), in consistence with the ~3 kcal/mol underestimation of E by electric field model. Since TS1(X=MeS -) is the one with largest spin population on X among all cases studied, it is concluded that charge shift bonding plays only a minor role in "salt effect" for TS1, and its contribution should be even less for anions other than MeS -.</p><p>On the other hand, however, the full Gibbs free energetics give a different scene of the salt effect, although it is established that external anions combine with TS1 significantly. All anions combine even more strongly with the OOH radical by hydrogen bonding, affording stable resting states, and increasing overall barrier, which is not mentioned in Cundari's work. The overall Gibbs free energy surface is shown in Figure 3. The energies for X-HOO complexes are raised upon solvation, but the overall barriers for all reactions still raise. Moreover, the solvation effect further cancels the TS-stabilization effect. As a result of all the effects above, the reaction is actually inhibited by external anions. Figure 3. The Gibbs free energy surface for reaction a in the presence of anions.</p><p>Although it has been revealed that anions cannot catalyze the HAT from methane to hydroperoxyl radical in the above work, the reaction studied above gives good insight into the physical picture of "salt effect". After establishment of the field-originated nature, we next moved to the HAT from methane to phenyl and methyl radical (reaction b and c in Scheme 1). The absence of strongly hydrogen bond donating hydroperoxyl radical is expected to avoid the formation of stable resting states. It is noticed that both TS2 and TS3 adopt different geometry: the anions are no longer collinear with the methane carbon center and the coming radical, but form a triangle shape to maximum the hydrogen bonding, which could be understood considering the largely reduced intrinsic dipole in these two cases. However, it is believed that there is no particularity for TS1, and further comments could be found in Supporting Information. Again, all anions provide strong stabilization in terms of electronic energies, while most of them are cancelled by entropy. For both TS2 and TS3, there are no longer spin densities on all anions, and the involvement of charge shift interaction is further precluded. BCPs appear in the interatomic region between anions and nearby hydrogen atoms, and thus the contribution of hydrogen bonding is easily estimated from Lu's work, indicating that HB contributes about 50% of E for these cases. The total E for TS3 can again be predicted by the uniform electric field model (while the effective field strength is hard to define for TS2). Since both TS2(X=none) and TS3(X=none) have negligible dipole moments, the field effect must come from either polarization or interaction with partly charged atoms, whereas the latter is also the origin of hydrogen bonding. In addition, energy decomposition methods further support the major role of electrostatic and polarization interaction (Supporting Information). The full Gibbs free energy surfaces for reaction b and c are shown in Figure 5. Despite the fact that the combining of anions with methyl or phenyl radical is much weaker than in reaction a, the overall barriers are still not lowered with only X=OHfor reaction b as an exception. Therefore, for the three reactions examined in this work, the catalytic effect of anions for HAT reactions previously reported by Cundari et al does not exist considering the full Gibbs free energy surface.</p><!><p>With computational methods, we studied the influence of external anions on the HAT reactions from methane to hydroperoxyl radical (reaction a), phenyl radical (reaction b) and methyl radical (reaction c). Although it was previously proposed by Cundari that it was charge shift bonding that made the major contribution to the stabilization effect toward HAT TSs, detailed electronic structure analysis shows that there is no chemical bonding between anions and methane carbon atom, and the spin delocalization is negligible for most anions. For the case with MeS -, which the anion bears the largest spin population among all reactions studied, the spin delocalization only contributes around 2.5 kcal/mol to the total energy. In contrast, a uniform electric field model gains quantitative success in interpreting the transition stabilization energy. By combining all these results, as well as energy decomposition study, it is believed that the electrostatic and polarization effect due to charged species play a major role.</p><p>On the other hand, although TSs are stabilized in terms of electric energy (and thus enthalpy), the effect cannot compensate the unfavorable entropy. As a result, the overall barrier, which is further influenced by stable resting states formed by substrates and anions, is not lowered or even raised in the presence of anions. In a summary, in contrast to the well-known existence of salt effect for polar cycloaddition or other reactions with polar TSs, our results are strongly against the presence of the Cundari-type "salt-effect" in HAT reactions. Strategical design, such as spatially constrained anions in order to compensate the entropy effect, is highly in need to further make use of the salt effect in the design of real "salt-catalyzed" HAT reaction in the future.</p><!><p>Benchmark data, CVB values, further discussion on TS geometry and the uniform electric field model, results for energy decomposition analysis, electronic structure analysis for TS1(X=MeS -), and all geometries involved in this work can be found in Supporting Information.</p>

ChemRxiv

The min-max test: an objective method for discriminating mass spectra

Deciding whether the mass spectra of seized drug evidence and a reference standard are measurements of two different compounds is a central challenge in forensic chemistry. Normally, an analyst will compute a mass spectral similarity score between spectra from the sample and reference and make a judgment using both the score and their visual interpretation of the spectra. This approach is inherently subjective and not ideal when rapid assessment of several samples is necessary. Making decisions using only the score and a threshold value greatly improves analysis throughput and removes analyst-to-analyst subjectivity, but selecting an appropriate threshold is itself a non-trivial task. In this manuscript, we describe and evaluate the min-max test -a simple and objective method for classifying mass spectra that leverages replicate measurements from each sample to remove analyst subjectivity. We demonstrate that the min-max test has an intuitive interpretation for decision-making, and its performance exceeds thresholding with similarity scores even when the best performing threshold for a fixed dataset is prescribed. Determining whether the underlying framework of the min-max test can incorporate retention indices for objectively deciding whether spectra are measurements of the same compound is on-going work.

the_min-max_test:_an_objective_method_for_discriminating_mass_spectra

Introduction<!>The min-max test<!>Evaluation methodology<!>Numerical Results<!>Discussion<!>Conclusions<!>Acknowledgements<!>Author Contributions<!>Competing Interests<!>Disclaimer

<p>"Is it a fentanyl?" Drug analysts are routinely tasked with answering questions of this nature when presented with seized drug evidence. The most commonly employed analytical technique towards confirmatory tasks in seized drug analysis is gas chromatography mass spectrometry 1 ; gas chromatography is used to physically and temporally separate the case sample into constituent components, and electron ionization (EI) mass spectrometry is used to propose the molecular structure (identity) of each component.</p><p>A mass spectrum can be thought of as a roughly reproducible a representation of a compound's structural information. In some cases, a mass spectrum includes near complete structural information, allowing it to be directly interpreted and the compound to be uniquely identified (see † To whom correspondence should be addressed: arun.moorthy@nist.gov a We introduce the phrase roughly reproducible to indicate that we expect replicate measurements to be self-similar, but we never expect replicate measurements to be identical.</p><p>introductory examples in references [2][3][4] among other textbooks). In most cases, however, mass spectra contain incomplete and non-unique structural information that render interpretation impractical without comparison to reference spectra. And while many mass spectral libraries 5,6 and interactive software tools 7,8 are available to assist in the interpretation process, the burden of decision-making still lies with the analyst. 9 In an application area like seized drug analysis, where decisions must be made rapidly and analyst subjectivity can be of significant consequence, having an explainable numerical approach for deciding whether mass spectra are measurements of different compounds is an obvious need.</p><p>We use the term similarity score to represent any numerical index that estimates the similarity between a pair of mass spectra. Although inconsistent in terminology and notation, several similarity scores have been explored in the literature. [10][11][12][13][14][15][16][17] The most well-known in mass spectrometry is the dot product, 11 or more commonly known as the cosine similarity in other pattern recognition applications. The cosine similarity between mass spectra will evaluate to a real number between 0 and 1 arbitrary units (au), inclusive, with 0 au indicating that the spectra share no common peaks, and the value 1 au indicating that the spectra are identical. A refinement of cosine similarity is the identity match factor, 17 based on the composite score described in the seminal paper by Stein and Scott. 11 The identity match factor considers the ratio of relative intensities of adjacent peaks when estimating similarity, thus capturing subtle information about isotopic patterns that is not necessarily reflected in cosine similarity. In most software implementations, identity match factors have been scaled to 100 (e.g., AMDIS 18 ) or 999 (e.g., NIST MS Search 19 ) and are reported as integers; all similarity scores discussed in this manuscript remain unscaled values between 0 and 1 au.</p><p>While not explicitly recommended b , we can use a threshold similarity score for deciding whether two mass spectra are measurements of different compounds -a task we refer to as negative confirmation in this manuscript. For example, if the identity match factor between the mass spectrum of a case sample and the spectrum of a reference standard is 0.3 au, the case sample is unlikely to be the same compound as the reference. Formally, we can think of this process in terms of binary classification and define the similarity score test as</p><p>where 𝑀 is a similarity score between two spectra, 𝜏 ! is a threshold similarity score, and class prediction 0 implies that the spectra are measurements of different compounds. Class prediction 1 implies that the spectra are measurements of the same compound, but we know that confirming whether two samples are the same compound (positive confirmation) with mass spectral comparisons alone is problematic as discussed later in the manuscript. The challenge with implementing the similarity score test for negative confirmation is that there is no obvious choice for a threshold value, especially not one that is universal across all classes of drugs and all varieties of similarity scores; equivalent similarity scores could be computed in very different situations (Figure 1). Algorithms and numerical approaches that produce counter-intuitive results and require analysts to make critical decisions are unappealing in forensics applications. 20 The min-max test was initially formulated during our recent work developing targeted gas chromatography-mass spectrometry methods for identifying synthetic cannabinoids 21 and cathinones. 22 We needed to determine if we could discriminate several pairs of closely eluting compounds based on their mass spectra. We wanted to avoid the ambiguity of threshold setting with a similarity score test and the subjectivity of visual comparison with manual interpretation. The min-max test was able to meet these requirements by leveraging replicate mass spectra to characterize spectral self-similarity within a sample and effectively remove subjectivity from the analysis. In this manuscript, we present the method more completely, with updated notation and refinements to reflect what we learned from our initial experimentation. We discuss how the minmax test, by construction, has an intuitive interpretation for deciding whether spectra are measures of different compounds (negative confirmation), and demonstrate how it out-performs the similarity score test for general classification using EI mass spectra of assorted drugs.</p><!><p>At the core of the min-max test are replicate measurements. By computing similarity scores between replicate mass spectra of individual compounds, we have context for decisions about pairs of compounds. Assume we have two samples to compare: the first is an unidentified compound isolated from seized evidence, and the second is a standard reference compound. Let 𝑺 𝟏𝟏 and 𝑺 𝟐𝟐 be sets of intra-sample similarity scores computed between two or more replicate mass spectra of samples 1 and 2, respectively. Let 𝑺 𝟏𝟐 be the set of inter-sample similarity scores computed between the spectra of the two samples. Using these sets of values, we can formulate a spectral comparison index (𝛿 $ ) that follows the general form</p><p>where 𝑓(⋅), 𝑔(⋅), and ℎ(⋅) are functions chosen to reduce the input sets of data into single representative values. In the min-max test, we decide whether two samples are different compounds by comparing the most conservative estimate of intra-sample spectral similarity to the most generous estimate of inter-sample similarity. Formally, we compute the min-max index (𝛿 !! ) as</p><p>where the functions min(⋅) and max (⋅) denotes the minimum and maximum values contained in the specified sets, respectively. For ease of reading, we will drop the subscript from 𝛿 !! and refer to the min-max index with the symbol 𝛿 as no other spectral comparison indices are computed in this manuscript.</p><p>As presented in Eq. ( 2) and with score sets 𝑺 𝟏𝟏 , 𝑺 𝟐𝟐 , and 𝑺 𝟏𝟐 constructed using similarity scores that evaluate between 0 and 1 au, the evaluated 𝛿 will be a real number between -1 and 1 au with practical values ranging between -0.1 and 0.9 au. The most intuitive employment of the min-max index is to assess whether 𝛿 > 0 au and infer that the compared sets of spectra are measures of different compounds; the larger the value of 𝛿, the more certain we are of the claim. If 𝛿 ≤ 0 au, there is at least some overlap in the observed intra and inter-sample spectral similarity and so we cannot confidently claim the samples are different compounds.</p><p>A transformation such as 𝛿 % = 1 − max(0, 𝛿) allows us to compare min-max indices more readily to similarity scores. The transformed min-max index evaluates between 0 and 1 au with uncertainty of a negative confirmation increasing with index values due to increased spectral similarity, as is the case with similarity scores. We can define the min-max test as</p><p>where 𝜏 &% is a min-max index threshold value, and prediction classes 0 and 1 implying that the spectra are measures of different compounds and the same compound, respectively. An intuitive employment of the min-max test in this formulation is to set 𝜏 & ! = 1 au. For the remainder of this manuscript, all discussion of min-max indices will reference these transformed values.</p><!><p>To demonstrate and evaluate the min-max test, we collated previously published and newly measured mass spectra into a single collection. The collection contained 10 replicate measurements of 144 illicit drug standards (comprised of synthetic cannabinoids, cathinones and opioids), totaling 1440 mass spectra labeled with names and molecular formulae. With these mass spectra, we computed two min-max datasets, the first with min-max indices computed using cosine similarity as the representative similarity score, and the second using identity match factors. These datasets were constructed by computing the min-max indices between all possible pairs of compounds using 3 randomly selected replicate spectra, repeating the experiment 100 times. In cases where the same compounds were being compared, we ensured the selected replicates were not overlapping. The resulting datasets contained 2073600 min-max indices each, approximately 0.7 % of which were computed between the same compound and the rest computed between different compounds. We also generated two similarity score datasets to evaluate the similarity score test for comparison. Each of these datasets were constructed by computing all possible nontrivial cosine similarity scores and identity match factors between the 1440 spectra, thus each dataset contained 2072160 total scores. Approximately 0.6 % of the similarity scores in the datasets were computed between spectra of the same compound and the rest between different compounds. The raw mass spectra, computed datasets, and source code and scripts used to analyze results are available for review. 23 Performance measures: For convenience, we use 𝑝 to denote an index and 𝜏 ' to denote a threshold value associated with 𝑝. A positive prediction is when 𝑝 ≥ 𝜏 ' , and a negative prediction is when 𝑝 < 𝜏 ' . For a set of indices that can be mapped to binary classifications, the number of True Positives (TP) is the count of positive predictions associated with positive classification (i.e., the compared spectra were replicates of the same compound). True Negatives (TN) are the negative predictions associated with negative classification (i.e., the compared spectra were measurements of different compound). False Positives (FP) are positive predictions that should have been associated with a negative classification, and False Negative (FN) are the negative predictions that should have been associated with positive classifications. Several standard performance measures can be derived from these quantities. In this manuscript, we considered accuracy, true positive rate (TPR) or recall, specificity, precision, and false positive rate (FPR) as described by Fawcett 24 and summarized in Table 1. Determining optimal thresholds: There are several options for determining optimal threshold values. One approach is to simply select the threshold value that optimizes the objective function (e.g., maximizes test accuracy) for the entire available dataset. While this method is likely to identify a unique threshold for each objective function, we gain no insights about how the identified threshold will perform with new data. A second approach is to use several subsets of the data to determine a range of threshold values, removing some data dependency and shedding some light on what an ideal threshold might look like for completely new data. In this manuscript, we obtained optimal thresholds using complete datasets and an iterative subset selection approach. For iteratives subsets, we randomly selected 10000 indices, ensuring that exactly 60 were values from replicate mass spectra, and repeated the process 1000 times. We determined threshold indices that maximized (1) accuracy, and (2) the difference between recall (TPR) and FPR.</p><!><p>To give a general overview of how similarity scores and min-max indices are distributed across our datasets, we generated box and whisker plots with indices distinguished as estimates of either different compounds or the same compound (Figure 2 recall. The observed accuracy of the similarity score test using the identity match factors dataset (Fig 2b) and a threshold of 0.9 au is approaching 100 %, but with a recall of only 50 %. With the two min-max datasets (Figs 2c and d), the effect of similarity score selection is more subtle, and the indices computed between replicate measurents of the same compound hover close to the 1 au decision-making threshold we intuitively expected. The accuracy of the min-max test with a threshold of 1 au is approaching 100 % with recall values also just under 100 %, using either dataset (Figs 2c and d). A comprehensive performance assessment of both tests using both similarity scoring selections and several fixed threshold values is provided in the Supplementary Information.</p><p>Figures 2a and b also confirm that confident negative confirmations can be made using just the similarity score test. For example, there was not a single false negative (incorrect negative confirmation) using a threshold of 0.7 au with either of the similarity score datasets in this manuscript. However, using such a low threshold is a very inefficient filter. We define the gray area as the range of similarity scores that will falsely characterize true negatives as false positives using the similarity score test. The lowest observed cosine similarity score for a true positive in the cosine similarity dataset is 0.88 au. We refer to this value as the lower gray area threshold. There were 370 pairs of different compounds where at least one of the computed cosine similarity scores was greater that then the lower gray zone threshold, and so would be falsely classified as the same compound if the lower gray zone threshold was used as the threshold in a similarity score test. Of these pairs, 100 % would be correctly classified as true negatives using the min-max test with a 1 au threshold. Similar results were observed with the gray area false positives from the similarity score test being correctly classified as true negatives using the min-max test when identity match factors were used to estimate spectral similarity (see Table S5 in Supplemental Information).</p><p>The optimal decision-making thresholds for both maximizing the accuracy and maximizing the difference between recall and false positive rates using all combinations of tests and similarity scores are summarized in Table 2. The optimal threshold value for using the min-max test is always around 1 au, regardless of similarity estimate and objective function. There is at least one instance where the threshold that maximized the difference between recall and false positive rate was 0.975 au. Because we used randomly selected subsets of data, we cannot easily disaggregate the exact conditions that led to that threshold value. The optimal threshold values for the similarity score tests varied more substantially with choice of similarity estimate and objective, but the tests performed well. That said, the objective values at the optimal threshold with the similarity score test were always less than the objective values using the min-max test with intuitive threshold 1 au.</p><!><p>To begin, we must acknowledge that our seemingly large datasets, with over 2 million data points each, represent a small fraction of the potential chemical space one might explore using EI mass spectrometry -we only considered mass spectra of 144 synthetic cannabinoids, cathinones and opioids. Additionally, our manuscript only evaluated the effect of two different similarity score choices. That said, the numerical results presented in the previous section and as supplemental information support our notion that the min-max test can be an excellent method for objective negative confirmation.</p><p>The original motivation for developing the min-max test was to overcome the known limitation of the similarity score test -a good decision-making threshold is difficult to select. While using cosine similarities and identity match factors to estimate spectral similarity is mature in its application, the rough reproducibility of mass spectra and non-linearity of the computations underlying these similarity estimates make them difficult to interpret without visual appraisal, hence limiting their objectivity. In our study, we were able to identify a range of optimal thresholds to maximize similarity score test accuracy and the difference between recall and false positive rate using both cosine similarities and identity match factors. These performance results give us a new data-driven basis from which to select decision-making thresholds for similarity score tests goingforward, yet these values lack the type of intuitive grounding necessary to be completely satisfying. And as with any quantity derived from data, there is concern that the dataset used to determine optimal thresholds was inadequate or inappropriate for the next application of the similarity score test.</p><p>Selecting a threshold value for the min-max test is intuitive. A min-max index of 1 au means that the minimum self-similarity observed within the sets of replicate spectra is equal to the maximum similarity observed between the spectra from either set, implying that the sets of spectra are not discernible. The performance results for the min-max test using the full dataset and the threshold of 1 au were excellent, with accuracy, recall and specificity all at least 99.8 %, regardless of similarity score imposed on the method. Using a slightly more conservative threshold value of 0.977 au, we have slightly worse accuracy and specificity, but observe 100 % recall.</p><p>The obvious limitation of the min-max test is that it requires replicate mass spectra of each of the compared samples. In an application like seized drug analysis, this requirement is likely not an issue as there is often enough seized evidence and reference standard to take several replicate measurements. However, it is not impossible that a lack of material or analysis time make taking replicate measurements impractical. One useful strategy in these scenarios is to run the similarity score test first, followed by the min-max test if the result falls within the gray area as demonstrated in the numerical results. other more subtle limitation of the min-max test is that poor quality spectra (e.g., spectra containing contaminants, or measured to inconsistent mass limits) will produce unreliable results. The effectiveness of the min-max test stems from our ability to quantify the conceptual notion of self-similarity. If a poor replicate is included in the test, our understanding of self-similarity is misrepresented in the min-max index, and the utility of the min-max test is essentially void. Alternative spectral comparison indices can be formulated using the general framework in Eq. ( 1) that are less susceptible to failing when provided a single incorrect measurement (e.g., choosing function 𝑔(⋅) to select the median value contained in the set). However, these alternate function choices may lack the simple and intuitive justifications that make the min-max test so satisfying. The similarity score test is also susceptible to failure if provided with poor quality mass spectra; this limitation is far more self-evident and is presently defenseless.</p><p>Our manuscript focused on the utility of the min-max test for negative confirmation tasks. Positive confirmation with mass spectrometry is appreciably more difficult. The performance metric of interest for positive confirmation tasks would be precision; the fraction of positive predictions that are true positives. With our datasets and a threshold of 1 au, the precision of the min-max test precision was greater than 85 % (using either cosine similarity or identity match factors) with nearly 100 % recall. This means that up to 15 % of positive predictions were false positives. Given that our dataset consists of several pairs of isomeric drugs, with near identical mass spectra, this high false positive rate is not surprising. A well-known strategy for enhancing the value of similarity scores is to combine them with measures from orthogonal technologies such as retention indices obtained from gas chromatography. [25][26][27] Exploring how we can effectively combine retention indices within on our framework of spectral comparison indices using replicate measurements for the purposes of objective positive confirmation is on-going work in our lab.</p><!><p>Forensic chemists are routinely required to make consequential decisions as quickly as possible. One example of a major decision is confirming that seized evidence does not contain an illegal drug (negative confirmation). The traditional method employed for this task is gas chromatography mass spectrometry followed by mass spectral interpretation. While effective, this approach is inherently subjective. An alternative is to compute numerical estimates of mass spectral similarity and use these values to make decisions. Unfortunately, setting decision making thresholds is a non-trivial challenge, especially across various classes of drugs. In this manuscript, we introduced the min-max test as an alternative method for negative confirmation using mass spectra. We discussed how selecting a threshold for the min-max test is intuitive and demonstrated that the method outperforms automated decision-making using spectral similarity estimates and threshold values, even when the threshold has been optimally chosen for a fixed dataset.</p><p>In addition to being effective, algorithms and software tools must be objective and explainable to be of any practical use in a forensic science setting. We constructed the min-max test with these two considerations in mind. We believe the min-max test will be an indispensable tool for forensic chemists performing negative confirmation tasks using mass spectra, and that this manuscript provides a template for further developing objective and explainable methods in the forensic sciences.</p><!><p>The authors would like to thank Drs. Kearsley, Mallard, Stein and Wallace of the National Institute of Standards and Technology for sharing insights that were instrumental in developing the min-max test. The authors would also like to thank Amber Burns of the Maryland State Police Forensic Sciences Division for sharing practical perspectives on analytical challenges faced by forensic laboratories.</p><!><p>A.S.M. and E.S. conceived the methods and designed the research plan. E.S. conducted laboratory experiments. A.S.M. performed computational analyses. A.S.M. and E.S. co-wrote the manuscript.</p><!><p>The authors declare no competing interests.</p><!><p>Official contribution of the National Institute of Standards and Technology (NIST); not subject to copyright in the United States. Certain commercial products are identified in order to adequately specify the procedure; this does not imply endorsement or recommendation by NIST, nor does it imply that such products are necessarily the best available for the purpose.</p>

ChemRxiv

Fully-automated radiosynthesis of the amyloid tracer [11C] PiB via direct [11C]CO2 fixation-reduction

BackgroundThe β-amyloid radiotracer [11C] PiB is extensively used for the Positron Emission Tomography (PET) diagnosis of Alzheimer’s Disease and related dementias. For clinical use, [11C] PiB is produced using the 11C-methylation method ([11C] Methyl iodide or [11C] methyl triflate as 11C-methylation agents), which represents the most employed 11C-labelling strategy for the synthesis of 11C-radiopharmaceuticals. Recently, the use of direct [11C]CO2 fixation for the syntheses of 11C-tracers has gained interest in the radiochemical community due to its importance in terms of radiochemical versatility and for permitting the direct employment of the cyclotron-produced precursor [11C]CO2.This paper presents an optimised alternative one-pot methodology of [11C]CO2 fixation-reduction for the rapid synthesis of [11C] PiB using an automated commercial platform and its quality control.Results[11C] PiB was obtained from a (25.9 ± 13.2)% (Average ± Variation Coefficient, n = 3) (end of synthesis, decay corrected) radiochemical yield from trapped [11C]CO2 after 1 min of labelling time using PhSiH3 / TBAF as the fixation-reduction system in Diglyme at 150 °C. The radiochemical purity was higher than 95% in all cases, and the molar activity was (61.4 ± 1.6) GBq/μmol. The radiochemical yield and activity (EOS) of formulated [11C] PiB from cyclotron-produced [11C]CO2 was (14.8 ± 12.1)%, decay corrected) and 9.88 GBq (± 6.0%), respectively. These are higher values compared to that of the 11C-methylation method with [11C]CH3OTf (~ 8.3%).ConclusionsThe viability of the system PhSiH3 / TBAF to efficiently promote the radiosynthesis of [11C] PiB via direct [11C]CO2 fixation-reduction has been demonstrated. [11C] PiB was obtained through a fully automated radiosynthesis with a satisfactory yield, purity and molar activity. According to the results, the one-pot methodology employed could reliably yield sufficiently high tracer amounts for preclinical and clinical use.Electronic supplementary materialThe online version of this article (10.1186/s41181-019-0065-4) contains supplementary material, which is available to authorized users.

fully-automated_radiosynthesis_of_the_amyloid_tracer_[11c]_pib_via_direct_[11c]co2_fixation-reductio

Introduction<!><!>Introduction<!><!>Introduction<!>Chemicals and materials<!>Instruments<!>Preparation of the module<!>Flushing of the target and lines<!>Preparation of the reagent solution<!>Optimisation of N-[11C-methyl]-4-toluidine<!>Automated radiosynthesis of [11C]PiB<!>Physicochemical quality control<!><!>Optimisation of N-[11C-methyl]-4-toluidine<!><!>Optimisation of N-[11C-methyl]-4-toluidine<!><!>Radiosynthesis of [11C]PiB<!><!>Conclusion<!>

<p>The compound 2-(4′-N-[11C]methylaminophenyl)-6-hydroxybenzothiazole, also known as [11C]6-OH-BTA-1 or [11C] Pittsburg Compound B ([11C]PiB), has long been recognised as a potent PET radiotracer for beta-amyloid (Aβ) plaque imaging in the brains of patients with Alzheimer's Disease (AD) and other forms of dementia (Engler et al. 2008; Herholz et al. 2007; Klunk et al. 2004; Nordberg 2004, 2008; Rabinovici and Jagust 2009). [11C] PiB still remains the gold standard for amyloid imaging in AD diagnosis due to its high affinity for Aβ plaques (Kd = 1.4 nM) (Mathis et al. 2003), fast uptake and low non-specific binding.</p><!><p>Different radiosyntheses of [11C] PiB published in the literature and their parameters</p><!><p>During the last decade, additional optimised radiosyntheses of [11C] PiB have been published in the literature. (Philippe et al. 2011; Coliva et al. 2015; Boudjemeline et al. 2017) Although these radiosyntheses proved to be reliable and easily adaptable to a GMP compliant production for patients, all of them use the secondary precursors [11C] methyl iodide or triflate. The preparation of these secondary precursors from cyclotron-produced [11C]CO2 is always time and activity consuming. (Långström et al. 1999). The overall RCY of any [11C] PiB radiosynthesis, and subsequently the total activity available for patients, is directly affected by these losses of radioactivity during the preparation of [11C]CH3I or [11C]CH3OTf. In this sense, any approach capable of eliminating the steps of synthesis of the 11C-methylating agents would be advantageous for a better overall performance in terms of the RCY and the available activity for PET scans.</p><!><p>Radiosynthesis of [11C] PIB through amine [11C] methylation by the direct use of cyclotron-produced [11C]CO2</p><!><p>With an ageing population, there is an increasing demand for [11C] PiB for clinical applications, and therefore the possibility of applying a methodology of [11C]CO2 fixation-reduction for the routine preparation of [11C] PiB to increase its radiochemical yield and thus perform more PET studies for each batch of radiopharmaceuticals was devised. In this context, attention was focused on the strategy recently developed by Liu and co-workers (Liu et al. 2016) for the reductive functionalisation of amines with CO2. In their work, the authors explored the utility of the formylation and methylation of amines with CO2 using hydrosilanes / TBAF as the catalytic system.</p><p>This approach employed mild reaction conditions because it is a metal-free protocol, and the use of the unstable N-heterocyclic carbene was avoided.. The radiochemical translation of this methodology was proposed for the radiosynthesis of amines based on radiopharmaceutical interest in particular [11C] PiB (Scheme 2).</p><p>The aim of this work was to assess the applicability of the PhSiH3 / TBAF mediated reductive incorporation of [11C]CO2 for the radiosynthesis of the β-amyloid tracer [11C]PiB. A fully automated, GMP adaptable, fast radiosynthesis of [11C] PiB by means of [11C]CO2 fixation-reduction was developed employing the commercially available platform GE TRACERlab® FX C Pro and its quality control.</p><!><p>All chemicals and reagents used in this work were commercially available products and were used without further purification. Anhydrous solvents MeCN (99.8%), DMF (99.8%), DMSO (99.9%) and bis (2-methoxyethyl) ether (Dyglime, 99.5%) were purchased from Sigma-Aldrich and stored and handled under inert atmosphere. Tetrabutylammonium fluoride (TBAF) 1.0 M in THF and PhSiH3 (97%) were acquired from Aldrich, stored in a desiccator and handled in an inert atmosphere (N2). 6-OH-BTA-0 was obtained from Siquimia. 4-Toluidine was purchased from Fluka. 6-OH-BTA-0, 6-OH-BTA-1 and 6-(MeO)-BTA-0 were used as analytical standards and were purchased from ABX (GmbH). N-methyl-4-toluidine, N-formyl-4-toluidine and 2-(4′-N-[11C]formamidophenyl)-6-hydroxybenzothiazole (PiB N-formyl derivative) were synthesised using standard organic chemistry procedures described in the literature (Vogel 1956; Shekhar et al. 2009). Absolute Ethanol (99.8%) was purchased from Merck. Saline and Water for Injection (sterile, USP grade) were acquired from Farmaco Uruguayo. Sep-Pak® C18 light cartridges were purchased from Waters and were preconditioned with ethanol (5 mL) followed by water for injection (10 mL) and air (5 mL). Millex® GV sterile filters (0.22 μm, PVDF, 33 mm) were purchased from Millipore. The semipreparative HPLC column used for [11C] PiB purification was a Luna® 5 μm C18(2) 100 Å 250 × 10 mm column (Phenomenex). The analytical HPLC column used for both model amine (4-toluidine) 11C-methylation and [11C] PiB radiosynthesis was a Nucleodur 100–5 C18-ec 250 × 4.6 mm column (Macherey-Nagel). The analytical GC column was a DB-WAX that was 30-m in length, 0.53-mm in diameter and 1.00-mm in film thickness (Agilent).</p><!><p>[11C]CO2 was produced by the 14N(p,α)11C nuclear reaction in a PETtrace™ 800 16.5 MeV cyclotron (GE Healthcare). A high-performance target was used for [11C]CO2 production. The target content was a mixture of N2 and 1.0% O2 (Praxair). To assess the best labelling conditions, ~ 18.5 GBq of [11C]CO2 (50 μA, 3 min) were used, and ~ 185 GBq (70 μA, 35 min) were used for complete radiosynthesis (labelling, purification and formulation).</p><p>Radiosyntheses were carried out using a TRACERlab® FX C PRO module (GE Healthcare) (see Additional file 1: Figure S1). All valves of the TRACERLab® module were controlled according to the pre-programmed time intervals (time lists) to transfer the reagents from one part to another part of the instruments. Helium pressure was used to transfer the reagents. The transfer of the radioactivity was traced and recorded with an inbuilt radioactivity detector. A by-pass of the iodination loop was made to redirect the purified [11C]CO2 towards the reaction vessel (reactor). A pre-injection vial (10 mL) was installed before the injection loop, controlled by valve 10 and pressurised with helium from valve 19. The purpose of this vial was to collect the reactor content and the portions of rinse solvent (acetone) added from Vial 3 before the loading of the HPLC loop.</p><p>HPLC analyses were performed with a Shimadzu UFLC equipped with UV and a gamma detector (Lablogic Flow RAM HPLC detector). The GC analyses of ethanol, residual reagents and residual solvents were carried out using a Shimadzu GC-2010 Plus equipped with an FID detector. The gamma spectrometry was performed using a 1023-channel Ortec multichannel analyser with a 1″ × 1″ NaI (Tl) crystal. The activity measurements were performed using a Capintec CRC 25 ionisation chamber.</p><!><p>Prior to any radiosynthetic procedure, the molecular sieves (MS, 4 Å, 60–80 mesh) column of the TRACERlab™ FX C Pro was heated to 350 °C under helium flow (40 mL/min) for 15 min and then cooled to room temperature under a helium atmosphere. Simultaneously, the reactor was rinsed with acetone, flushed with helium and further dried under a vacuum to 100 °C for 30 min. The reactor was then cooled to 25 °C and was kept in a positive helium atmosphere (> 300 kPa) until its use in the radiosynthetic experiment.</p><!><p>To improve Am, a protocol described by our group (Savio et al. 2012) was followed. In short, the content of the target was delivered to the molecular sieves column at 350 °C under flowing helium (40 mL/min) to decrease the amount of unlabelled CO2 and to send it to waste ('cold flush'). Immediately before beginning the irradiation for the radiosynthesis, the target was bombarded at 70 μA for 5 min, and its content was directly sent to waste ('hot flush').</p><!><p>PhSiH3 was added slowly over a solution of the amine (4-toluidine or 6-OH-BTA-0) in the desired solvent (0.5 mL) in an inert atmosphere. The solution was vigorously mixed using a vortex agitator, and then TBAF 1.0 M in THF was carefully added over the mixture. The evolution of gas and changes in colour were usually observed during this step. The resulting solution was mixed using a vortex agitator, taken by a syringe, loaded into the reactor of the TRACERlab™ module and sparged with helium (40 mL/min) for 5 min, ideally no more than 5 min before the end of bombardment (EOB).</p><!><p>The cyclotron produced [11C]CO2 (EOB Activity: A0) was sent to the module and trapped in the MS column at room temperature for further purification. The delivery duration was approximately 3 min. The MS column was then heated to 350 °C to desorb purified [11C]CO2, which was transferred under a helium stream (15 mL/min) to the reactor where the amine solution was placed. Trapping was performed at room temperature. The trapped [11C]CO2 activity was monitored and registered. Once trapping was complete (a maximum activity AT is reached), the reactor was sealed, and the solution was heated to the chosen temperature. For evaluating the losses of [11C]CO2 during the heating step, 'start of labelling activity' (ASOL) was registered once the temperature reached the desired value. Likewise, 'end of labelling activity' (AEOL) was registered once the labelling time was finished. The solution was cooled to approximately 70 °C and was diluted with 0.5 mL of the same solvent used in the radiolabelling step. The solution was collected in a vial, its activity (AVIAL) was measured and radio HPLC analyses were performed to determine the relative radiochemical proportion of the expected species.</p><!><p>The same protocol described was conducted using 6-OH-BTA-0 as the precursor amine. Once the labelling step was finished, the solution was cooled and sent to the pre-injection vial. The reactor was then rinsed with acetone (1 mL) from Vial 3, combined with the reaction crude and injected into the HPLC. The separation of [11C] PiB was achieved using MeCN:H2O (50,50 v/v) at a flow rate of 4 mL/min (tR: 8.5–9.5 min). The fraction containing [11C] PiB was collected over 50 mL of water for injection and then passed through a Sep-Pak® C18 light cartridge. The excess HPLC solvent was washed with water for injection (10 mL). [11C] PiB was eluted from the SPE cartridge with 0.9 mL of absolute ethanol and collected over 5 mL of preloaded saline. In addition, 4 mL of saline were used to rinse the SPE cartridge. The solution of formulated [11C] PiB was filtered through a 0.22 μm sterilising filter. The total time of radiosynthesis was about 32 min (since EOB) or 25 min (since the end of [11C]CO2 trapping).</p><!><p>Radiochemical purity (RCP) was determined using analytical radio-HPLC. An isocratic condition with a CH3COONH4 / CH3COOH buffer 0.1 M, pH = 5.0 and MeCN (40:60 v/v) at a flow rate of 1.2 mL/min was used for 4-toluidine. For [11C] PiB, an isocratic condition with H2O and MeCN (50:50 v/v) at a flow rate of 1.2 mL / min was used. RCP was calculated considering the area of the peak corresponding to the desired analyte in relation to the sum of the areas of all peaks. The identity of radioactive products was confirmed by co-elution with the non-radioactive standard compounds. UV detection was 270 nm for 4-toluidine and 340 nm for PiB. Am was calculated considering the activity of [11C] PiB EOS x RCP in relation to the molar amount of PiB in the sample.</p><p>Residual solvents and reagents (such as acetone, acetonitrile, Diglyme and PhSiH3) and ethanol were analysed by gas chromatography (GC) in accordance with USP general chapter < 467>. The temperature programme for GC runs was a gradient of 40 °C hold for 2 min, 1 °C/min to 44 °C, 20 °C/min to 200 °C and 200 °C hold for 2 min (total time of 15 min) with helium (11.3 mL/min) as the carrier gas.</p><p>The appearance of the solution was checked by visual inspection. The pH level was determined using a calibrated pH-meter. Radionuclidic purity was assessed by recording the corresponding gamma spectrum, and radionuclidic identity was assessed by measuring the physical half-life.</p><!><p>Direct 11C-methylation of 1 with [11C]CO2</p><p>[a] Experiments were performed using approximately 18.5 GBq of [11C]CO2 as a starting activity from the cyclotron; [b] Radiochemical yield (%, dc) was estimated as the actual final activity of [11C]2 or [11C]3 in relation to trapped [11C]CO2 (AT) (decay corrected); [c] TBAF (0.05 mmol) [d] without TBAF [e] blank experiments in DMSO</p><!><p>The reactivity of the system was evaluated through the radiochemical yield (RCY) of N-[11C-methyl]-4-toluidine ([11C]2). The formation of [11C]2 was observed in practically all conditions where the capture of [11C]CO2 was considerable. Based on the work of Liu and co-workers (Liu et al. 2016), the experimental conditions began with using MeCN as a solvent. Thus, in MeCN at 80 °C, a radiochemical yield of 14% was obtained for [11C]2 (entry 1, Table 1). An increase in the amount of the amine 1 precursor led to an improvement in the radiochemical yields of [11C]2 (entries 2 and 3). Similar radiolabelling results were obtained when both the temperature and the reaction time were increased (entries 4 and 5). Nonetheless, under these conditions, the RCY observed was lower than 45% in all cases (entries 1–7, Table 1).</p><!><p>The evolution of the radioactivity and labelling temperature within the reaction vessel. AT: maximum [11C]CO2 radioactivity trapped in the reactor; ASOL: Radioactivity at the Start of Labelling; AEOL: Radioactivity at the End of Labelling</p><!><p>Finally, a significantly lower radiochemical yield was observed when DMF was tested as a reaction solvent compared to DMSO (entry 15 vs entry 9, Table 1). In some cases, the presence of N-[11C]-4-tolylformamide ([11C]3) was detected. This is in agreement with the proposed reaction mechanism in which a N-11C-formamide would be a precursor for the N-11C-methyl amine (Additional file 1: Scheme S1). Thus, when DMF was used as a solvent at 100 °C, ([11C]3) was obtained with an RCY of 26% (entry 15, Table 1), whereas when using MeCN and DMSO as solvents at 80 and 50 °C, [11C]3 was formed in 8 and 14% of RCY, respectively (entries 1 and 8, Table 1).</p><p>Another important parameter of the [11C]CO2 fixation methodology is the trapping efficiency of the reagent solution. This was evaluated as the relationship between maximum trapped [11C]CO2 activity (AT) and [11C]CO2 EOB theoretical activity (A0), and it was decay corrected. The trapping of [11C]CO2 was dramatically affected by the amount of TBAF in the solution. When no TBAF was present, the trapping efficiency of [11C]CO2 was less than 5% (entries 12 and 16, Table 1). This is consistent with the mechanism in which TBAF was necessary to form the adduct [PhSiH3F]− that would have been responsible for solubilising the [11C]CO2 in the form of [11C]HCOO− (See Additional file 1: Scheme S1); however, increasing the amount of TBAF to 0.05 mmol did not appear to be favourable for a more efficient trapping (60%, entry 6, Table 1). All the solvents used showed good performance for trapping [11C]CO2 at room temperature when 0.01 mmol of TBAF was used independently from the amine concentration. Nevertheless, when no amine was added (blank runs in DMSO), a small decrease in the trapping efficiency was observed (entries 13 and 14, Table 1).</p><p>No losses of radioactivity due to the volatilisation of [11C]CO2 or other radioactive derivatives were observed during the radiolabelling step, which indicated that the 11C species were quantitatively solubilised into the solution (Fig. 1). The losses of radioactivity were evaluated based on the relationship between AEOL/ASOL and were decay corrected. In all cases, this relationship was approximately 100%.</p><!><p>Optimisation of reaction conditions for radiolabelling [11C] PIB from direct [11C]CO2 incorporation</p><p>[a] Experiments were performed using approximately 18.5 GBq of [11C]CO2 as a starting activity from the cyclotron; [b] 'cold' and 'hot' flushes of targets and lines; [c] solution sparged with He (5 min, 40 mL/min); [d] without [11C]CO2, [e] without [11C]CO2 and without PhSiH3 / TBAF, [f] n = 3</p><!><p>These results could be correlated with a higher degree of reaction between increasing concentrations of the precursor and environmental CO2 to obtain PIB (considering an excess of environmental CO2 that dilutes [11C]CO2 during the experiments). In this context, a further flushing of the lines and the target lowered the concentration of unlabelled PiB, but the total concentration was still high for the purposes (entry 8, Table 2). The elevated values in the concentrations of PiB led to the assumption that other phenomena could be added to the incorporation of environmental CO2, and in this sense, a contribution of a methyl group from the DMSO used as a solvent was suspected. Indeed, as Jiang and co-workers (Jiang et al. 2014) reported, DMSO can be used as a methylating agent for amines in the presence of HCOOH at 150 °C.</p><p>In this system, the presence of unlabelled formoxysilane (PhSiH2OCHO) could lead to the activation of DMSO in the form of a methylmethylenesulphonium cation, which could act as the methylating agent depicted by the authors (as shown in Additional file 1: Scheme S2). To demonstrate the contribution of the methyl group from DMSO, two [11C]CO2 blank experiments were carried out (entries 9 and 10, Table 2). As expected, a higher concentration of PiB was observed in the presence of DMSO and PhSiH3/TBAF compared to that obtained when no PhSiH3 / TBAF were added to the DMSO solution.</p><p>In view of these assumptions, it is proposed that the use of an alternative solvent with a high boiling point might be favourable for improving the Am of [11C]PiB. Indeed, the work of Liger and co-workers employed Diglyme as a reaction solvent for the direct [11C] methylation of amines from [11C]CO2 (Liger et al. 2015). When using Diglyme in a [11C]CO2 blank experiment carried out in the presence of PhSiH3 / TBAF, the concentration of PiB decreased considerably with respect to the corresponding experiment with DMSO (entry 11, Table 2).</p><p>Similar results were obtained in the presence of [11C]CO2 in Diglyme at 150 °C for 2.5 min, decreasing the concentration of PiB to 105 μg/mL (entry 12). The RCY and RCP of the [11C] PiB achieved under these conditions were 57% and 64%, respectively. Further increasing the labelling time to 5 min allowed for obtaining a higher RCY and RCP for [11C] PiB, and the highest Am achieved (entry 13, Table 2). Furthermore, the [11C]CO2 trapping efficiency was 83% for this condition.</p><p>It is important to highlight that the corresponding products of N-formylation and O-methylation, 2-(4′-N-[11C]formamidophenyl)-6-hydroxybenzothiazole and 2-(4′-aminophenyl)-6-O [11C] metoxybenzothiazole, respectively, were not observed under the conditions assayed (data not shown).</p><!><p>Optimised conditions for the complete radiosynthesis of [11C] PiB, starting from 185 GBq of [11C]CO2 (N = 3 each experiment)</p><p>Analytical radio-HPLC for formulated [11C]PiB: upper: UV (340 nm); lower: gamma</p><p>Physicochemical quality control parameters for three consecutive batches of [11C] PiB (labelling time 1 min)</p><p>Comparison of the performance of the methodology for the radiosynthesis of [11C] PiB using [11C]CH3OTf and the approach presented in this work employing [11C]CO2 fixation-reduction with PhSiH3 / TBAF. Values are expressed as Average ± Variation Coefficient (%)</p><!><p>It has been demonstrated that the direct [11C]CO2 fixation-reduction for the radiosynthesis of [11C] PiB can be achieved using the PhSiH3 / TBAF system. To obtain knowledge related to the radiochemical nature of the methodology and thus to produce [11C] PiB with good and reproducible RCY and Am, the influence of physical and radiochemical parameters was investigated. Higher radiochemical yields and activities (EOS) of formulated [11C] PiB from cyclotron-produced [11C]CO2 were obtained compared to that of the 11C-methylation method using PhSiH3 / TBAF as a fixation-reduction system in Diglyme at 150 °C for 1 min.</p><p>Based on these results, a rapid one-pot methodology for the radiosynthesis of [11C] PiB by means of the direct use of the primary precursor [11C]CO2 was developed by employing an automated commercial platform along with a physicochemical quality control proposed for its analysis. In this context, the study indicates the advantages of the unique published work in the application of a direct [11C]CO2 fixation-reduction methodology for the radiochemical productions of [11C] PiB, especially in terms of reaction conditions (carbene- and metal-free), Am and overall radiosynthesis time.</p><!><p>Semipreparative HPLC chromatogram for [11C]PiB. (DOCX 735 kb)</p><p>2-(4′-N-aminophenyl)-6-hydroxybenzothiazole</p><p>Alzheimer's Disease</p><p>Molar activity</p><p>decay corrected</p><p>Bis (2-methoxyethyl) ether</p><p>End of bombardment</p><p>End of synthesis</p><p>Gas chromatography</p><p>Good manufacture practice</p><p>High-performance liquid chromatography</p><p>no decay corrected</p><p>Positron emission tomography</p><p>Phenylsilane</p><p>Pittsburg compound B (2-(4′-N-methylaminophenyl)-6-hydroxybenzothiazole)</p><p>Quality control</p><p>Radiochemical purity</p><p>Radiochemical yield</p><p>Tetrabutylammonium fluoride</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

A Sephin1-insensitive tripartite holophosphatase dephosphorylates translation initiation factor 2α

The integrated stress response (ISR) is regulated by kinases that phosphorylate the α subunit of translation initiation factor 2 and phosphatases that dephosphorylate it. Genetic and biochemical observations indicate that the eIF2αP-directed holophosphatase, a therapeutic target in diseases of protein misfolding, is comprised of a regulatory subunit, PPP1R15, and a catalytic subunit, protein phosphatase 1 (PP1). In mammals, there are two isoforms of the regulatory subunit, PPP1R15A and PPP1R15B, with overlapping roles in the essential function of eIF2αP dephosphorylation. However, conflicting reports have appeared regarding the requirement for an additional co-factor, G-actin, in enabling substrate-specific dephosphorylation by PPP1R15-containing PP1 holoenzymes. An additional concern relates to the sensitivity of the holoenzyme to the [(o-chlorobenzylidene)amino]guanidines Sephin1 or guanabenz, putative small-molecule proteostasis modulators. It has been suggested that the source and method of purification of the PP1 catalytic subunit and the presence or absence of an N-terminal repeat–containing region in the PPP1R15A regulatory subunit might influence the requirement for G-actin and sensitivity of the holoenzyme to inhibitors. We found that eIF2αP dephosphorylation by PP1 was moderately stimulated by repeat-containing PPP1R15A in an unphysiological low ionic strength buffer, whereas stimulation imparted by the co-presence of PPP1R15A and G-actin was observed under a broad range of conditions, low and physiological ionic strength, regardless of whether the PPP1R15A regulatory subunit had or lacked the N-terminal repeat–containing region and whether it was paired with native PP1 purified from rabbit muscle or recombinant PP1 purified from bacteria. Furthermore, none of the PPP1R15A-containing holophosphatases tested were inhibited by Sephin1 or guanabenz.

a_sephin1-insensitive_tripartite_holophosphatase_dephosphorylates_translation_initiation_factor_2α

Introduction<!><!>Introduction<!>Both native PP1 and bacterially expressed PP1α require the presence of G-actin to promote PPP1R15A-regulated eIF2αP dephosphorylation<!><!>Both native PP1 and bacterially expressed PP1α require the presence of G-actin to promote PPP1R15A-regulated eIF2αP dephosphorylation<!>Two-fold stimulation of eIF2αP dephosphorylation by repeat-containing PPP1R15A in an unphysiological low ionic strength buffer<!><!>Two-fold stimulation of eIF2αP dephosphorylation by repeat-containing PPP1R15A in an unphysiological low ionic strength buffer<!>Lengthy incubation of the enzymatic reactions does not uncover PPP1R15A's ability to promote G-actin–independent eIF2αP dephosphorylation at physiological salt concentrations<!><!>Substrate recruitment by the repeat-containing PPP1R15A325–512 region plays a secondary role in the kinetics of eIF2αP dephosphorylation, and its disruption is unlikely to account for sensitivity to Sephin1<!><!>Substrate recruitment by the repeat-containing PPP1R15A325–512 region plays a secondary role in the kinetics of eIF2αP dephosphorylation, and its disruption is unlikely to account for sensitivity to Sephin1<!><!>Substrate recruitment by the repeat-containing PPP1R15A325–512 region plays a secondary role in the kinetics of eIF2αP dephosphorylation, and its disruption is unlikely to account for sensitivity to Sephin1<!>Discussion<!><!>Discussion<!>Protein expression and purification<!>In vitro dephosphorylation reactions<!>Author contributions<!>

<p>The integrated stress response (ISR)6 is a signal transduction pathway that couples diverse stressful conditions to the activation of a rectifying translational and transcriptional program that is implicated in biological processes ranging from memory formation to immunity and metabolism (reviewed in Ref. 1). The mammalian ISR and its yeast counterpart (the general control response) are initiated by phosphorylation of the α subunit of translation initiation factor 2 (eIF2α) on serine 51 (2, 3), and its activity is terminated by eIF2αP dephosphorylation.</p><p>Two related regulatory proteins, PPP1R15A/GADD34 and PPP1R15B/CReP, encoded in mammals by PPP1R15A and PPP1R15B, direct the unspecific protein phosphatase 1 (PP1) to promote eIF2αP dephosphorylation (4–7). PPP1R15A or PPP1R15B form a complex with PP1 via a conserved region of ∼70 amino acids (PPP1R15A residues 555–624) located at their C termini (5, 8–11) (Fig. 1A). This conserved C-terminal region of either PPP1R15 regulatory subunit is sufficient to promote eIF2αP dephosphorylation and to inactivate the ISR (4, 5, 10, 11). Indeed, herpesviruses have exploited this activity and encode a small protein homologous to the C terminus of PPP1R15 to reverse eIF2α phosphorylation, undoing a defensive strategy of infected cells (12).</p><!><p>G-actin stimulates PPP1R15A-dependent eIF2αP dephosphorylation by either PP1N or PP1α. A, cartoon representation of human PPP1R15A protein (1–674) and the different constructs used in this study (sequence provided in Table S1). Key residues used for truncated versions of the proteins in this study are annotated. The ER localization domain and the proline, glutamate, serine, and threonine-rich (PEST) repeats are highlighted, as are the PP1 and G-actin binding sites in the conserved C-terminal region. The MBP solubility tag is also represented in the cartoons of the constructs. B, top panel, Coomassie-stained PhosTag SDS-PAGE containing resolved samples of dephosphorylation reactions (30 min at 30 °C) in which 2 μm eIF2αP was dephosphorylated by PP1N purified from rabbit skeletal muscle in the presence or absence of PPP1R15A325–636-MBP (50 nm) and/or G-actin (400 nm).The position of the various protein species is indicated. eIF2αP and eIF2α0 refer to the phosphorylated and nonphosphorylated form of the bacterially expressed N-terminal domain (residues 1–185) of eIF2α, respectively. Note that both G-actin and PP1N preparation gave rise to two bands: a major full-length species and minor degradation product in the case of G-actin and a PP1and tropomyosin band in the case of PP1N (see also Fig. S1). Shown is a representative experiment of two independent repetitions performed. Center panel, plot of the rate of eIF2αP dephosphorylation as a function of the concentration of PP1N from lanes 1–12 of the experiment above. Bottom panel, plot of the velocity of each enzyme relative to the mean of velocity of PP1 alone calculated from all the informative reactions in the two repeats of this experiment. Statistical significance was derived from Mann-Whitney test (ns, nonsignificant, p > 0.05; ***, p ≤ 0.001). C, as in B but using bacterially expressed PP1α as the catalytic subunit (96, 48, 24, or 12 nm), MBP-PPP1R15A325–636 (50 nm), and G-actin (400 nm). The assays were performed during 20 min at 30 °C. Shown is a representative experiment of two independent repetitions performed.</p><!><p>Despite genetic evidence pointing to the sufficiency of the conserved C-terminal portion of PPP1R15 in reversing the eIF2αP-dependent ISR in vivo (4, 5, 10), complexes formed in vitro between PPP1R15 regulatory subunit fragments and PP1 have not been observed to accelerate eIF2αP dephosphorylation. Dephosphorylation of eIF2αP is no faster by a complex of PPP1R15A–PP1 (or PPP1R15B–PP1) than by PP1 alone, showing that, when added as single components, PPP1R15A/B do not influence kcat or Km of PP1 toward the substrate eIF2αP (10). However, addition of G-actin to the binary complex of PPP1R15 and PP1 selectively accelerates eIF2αP dephosphorylation. G-actin binds directly to the conserved C terminus of PPP1R15 alongside PP1 to form a ternary complex, whose affinity (Kd∼10−8 m) matches the EC50 of G-actin's stimulatory effect (10, 13). The in vivo relevance of G-actin for eIF2αP dephosphorylation is attested to by the finding that actin sequestration in fibers (as F-actin) enfeebles eIF2αP dephosphorylation, implying a role for factors that affect the actin cytoskeleton in ISR regulation (14).</p><p>The ability to dephosphorylate eIF2αP is an essential function in developing mammals (15). Nonetheless, inactivation of the PPP1R15A gene, which decelerates eIF2αP dephosphorylation and prolongs the ISR, is protective in certain cellular and animal models of diseases associated with enhanced unfolded protein stress (16–19). This has generated interest in targeting the PPP1R15A-containing holophosphatase for inhibition by small molecules (reviewed in Ref. 20), an endeavor that requires detailed knowledge of the enzymatic mode of action.</p><p>A recent report challenged the need for G-actin as a co-factor in PPP1R15A-mediated eIF2αP dephosphorylation (21). Instead, it suggested that a binary complex assembled from PP1α and a fragment of PPP1R15A (PPP1R15A325–636), encompassing both the C-terminal PP1-binding region and the N-terminal repeat–containing extension, dephosphorylates eIF2αP faster than PP1 alone (21). Importantly, dephosphorylation of eIF2αP by this active binary complex was reported to be selectively inhibited in vitro by guanabenz and Sephin1, two structurally related small molecules reputed to function in vivo as proteostasis modifiers (22, 23). The new study contradicts previous observations that neither a PPP1R15A–PP1 binary complex nor a PPP1R15A–PP1–G-actin ternary complex were susceptible to inhibition by guanabenz or Sephin1 (9, 13).</p><p>Here we address three important questions raised by these discrepant reports. Does the isotype of the PP1 catalytic subunit or its source (recombinant versus native) influence the requirement for G-actin by the eIF2αP-directed holophosphatase? What role does the N-terminal repeat–containing region of PPP1R15A play in eIF2αP dephosphorylation by the holophosphatase? Do these factors influence the sensitivity of eIF2αP dephosphorylation to guanabenz and Sephin1?</p><!><p>PP1 produced in Escherichia coli may differ in its enzymatic activity from PP1 purified from animal tissues, both in its substrate specificity and in its sensitivity to regulatory subunits (reviewed in Ref. 24). To determine whether the G-actin dependence of PP1–PPP1R15A–mediated eIF2αP dephosphorylation is a peculiarity of the bacterially expressed PP1γ isoform used previously (10, 13), we purified the native catalytic subunit of PP1 from rabbit skeletal muscle (PP1N), following an established protocol (25), and compared the two PP1 preparations. Native PP1 (PP1N) is a mixture of PP1α, PP1β, and PP1γ isoforms and gave rise to two prominent bands on SDS-PAGE (Fig. S1A, left panel). The mass spectra of tryptic peptides derived from the PP1N sample were analyzed by Maxquant with iBAQ (intensity-based absolute quant) to identify the major contaminating species (tropomyosin), and to estimate the relative contribution of PP1 and contaminants to the protein preparation. This enabled a comparison of the catalytic subunit content of PP1N preparation with the bacterially expressed PP1γ, which served as a reference.</p><p>The N-terminal portion of PPP1R15A, which includes the membrane association region (26), compromises expression in bacteria and recovery of a functional protein (27). Therefore, we used a PPP1R15A325–636 fragment lacking this region, which is soluble when expressed in E. coli. Fig. 1B shows that addition of either PPP1R15A325–636-MBP (lanes 5–8) or G-actin alone (lanes 13 and 14) did not stimulate eIF2αP dephosphorylation by nanomolar concentrations of PP1N. However, addition of both G-actin and PPP1R15A325–636-MBP (Fig. 1B, lanes 9–12) stimulated dephosphorylation by 5-fold, similar to the increase observed with bacterially expressed PP1γ (Fig. S1B) (10).</p><p>PP1 purified from rabbit muscle is a mixture of α, β, and γ isoforms, whereas it has been reported that the PP1α isoform possesses in vivo selectivity for PPP1R15A (6). Therefore, we prepared bacterially expressed PP1α by a method that promotes its native-like state (28). To control for effects the location of the tag might have on activity, we also generated an N-terminally MBP-tagged PPP1R15A325–636 (MBP-PPP1R15A325–636; Fig. 1A and Table S1). The holophosphatase comprised of PP1α and MBP-PPP1R15A325–636 also exhibited a stringent requirement for G-actin (Fig. 1C).</p><p>A concentration-dependent stimulatory effect of PPP1R15A on eIF2αP dephosphorylation by the three component holoenzyme (PP1, PP1R15A, and G-actin) was observed with constructs tagged at either their N or C termini and with either native or bacterially expressed PP1 (Fig. 2, A and B). The difference in EC50 values obtained for PPP1R15A325–636-MPB with PP1N (58 nm) or MBP-PPP1R15A325–636 with PP1α (6 nm) may reflect the effect of the position of the MBP tag, the contaminating tropomyosin (in PP1N), or both. Importantly, the data agreed with similar experiments in which PPP1R15A325–636 and bacterially expressed PP1γ were used, with an EC50 of 10 nm (see Fig. 8A in Ref. 13).</p><!><p>The source of the catalytic subunit does not affect the kinetics of PPP1R15A and G-actin–mediated stimulation of eIF2αP dephosphorylation. A, top panel, Coomassie-stained PhosTag SDS-PAGE of dephosphorylation reactions (30 min at 30 °C) in which 2 μm eIF2αP was dephosphorylated by PP1N (20 nm) in the presence of G-actin (400 nm) and increasing concentrations of PPP1R15A325–636-MBP (0–100 nm). Shown is a representative experiment of three independent experiments performed. Bottom panel, plot of the rate of dephosphorylation of eIF2αP as a function of PPP1R15A325–636-MBP concentration from the three experiments performed. The EC50 was calculated using the "[Agonist] versus response − variable slope (four parameters)" function in GraphPad Prism v7. The gray lines represent the 95% confidence interval of the fitting. Shown are values obtained for EC50 and information of goodness of the fit (r2). B, as in A but using bacterially expressed PP1α (24 nm) and increasing concentrations of MBP-PPP1R15A325–636 (0–100 nm) in reactions performed over 20 min at 30 °C. Shown is a representative experiment of three independent experiments performed. C, as in A but with fixed concentrations of PP1N (20 nm) and PPP1R15A325–636-MBP (50 nm) and varying the concentrations of G-actin (1–2000 nm). Shown is a representative experiment of three independent experiments performed.</p><!><p>G-actin also exerted a saturable concentration-dependent stimulatory effect on the activity of a three-component holophosphatase constituted with native PP1N (Fig. 2C). The EC50 for G-actin with PP1N (30 nm) was similar to that observed previously using bacterially expressed PP1γ, with an EC50 of 13 nm (see Fig. 2C in Ref. 13). Hence, despite variations in the estimated EC50 values for PPP1R15A or G-actin, the combinations of catalytic and regulatory subunits tested showed consistent PPP1R15A and G-actin concentration-dependent enzymatic activity. These experiments, conducted in a buffer of physiological ionic strength over a physiological protein concentration range (nanomolar catalytic subunit and micromolar substrate) and over a timescale aimed to minimize the effect of substrate depletion on enzyme kinetics, indicate that neither the source of PP1 nor the position of the tag in PPP1R15A are likely to account for the reported G-actin–independent ability of PPP1R15A to stimulate eIF2αP dephosphorylation.</p><!><p>To explore the discrepant findings on the G-actin independent stimulatory activity of MBP-PPP1R15A325–636, we sought to reproduce the experiments reported in Ref. 21 as closely as possible. We received from the Bertolotti laboratory their expression plasmid. The encoded protein, referred to here as MBP∼PPP1R15A325–636 (Fig. 1A), differs from the one used above (MBP-PPP1R15A325–636) by the absence of three residues in the linker separating the MBP from PPP1R15A and 11 residues in the linker separating PPP1R15A from the C-terminal polyhistidine tag (Table S1). The MBP∼PPP1R15A325–636 fusion protein was produced in E. coli and purified as described previously (21), and dephosphorylation reactions were carried out in a salt-free, low ionic strength buffer designed to mimic as closely as possible the one used in that study (50 mm Tris (pH 7.4), 1.5 mm EGTA, and 2 mm MnCl2, with the notable exception of 0.5 mm TCEP, added here to prevent inactivation of the catalytic subunit by oxidation).</p><p>A 2-fold stimulation of eIF2αP dephosphorylation by MBP∼PPP1R15A325–636 was apparent in reactions conducted at low salt concentration (15 mm) but lost at more physiological concentrations (100 mm), whereas the 5-fold stimulatory effect of G-actin was observed at both low and physiological salt concentration (Fig. 3A). The stimulatory effect of MBP∼PPP1R15A325–636 at low salt concentration depended on the N-terminal repeat–containing region of PPP1R15A (Fig. 3B), as reported previously (21), and was not observed with a nonspecific dephosphorylation substrate (Fig. S2A).</p><!><p>PPP1R15A325–636 accelerates eIF2αP dephosphorylation by PP1α in a low ionic strength buffer. A, top panel, Coomassie-stained PhosTag SDS-PAGE containing resolved samples of dephosphorylation reactions (30 min at 30 °C) in which 2 μm eIF2αP was dephosphorylated by PP1α (25 or 100 nm) in the presence or absence of MBP∼PPP1R15A325–636 (1 μm) with or without G-actin (400 nm) in low (15 mm NaCl) or physiological (100 mm NaCl) ionic strength buffer. Shown is a representative experiment of three independent repetitions performed. Bottom panel, plot of the percentage of eIF2αP dephosphorylation under the different conditions from the experiment above and the two other repeats performed. Statistical significance was derived from paired two-tailed t test (ns, nonsignificant, p > 0.05; *, p ≤ 0.05; **, p ≤ 0.01). B, as in A but using PPP1R15A533–624-MBP (200 nm). Shown is a representative experiment of three independent repetitions performed.</p><!><p>Although modest (2-fold) and confined to nonphysiological, low ionic strength conditions, this stimulatory effect was also reproducibly observed with the MBP-PPP1R15A325–636 and PPP1R15A325–636-MBP proteins used in Figs. 1 and 2 (Fig. S2B), negating a role for the linkers or the position of the tag in this activity. Notably, in both unphysiological low ionic strength buffer (in which PPP1R15A alone has a stimulatory effect) and under physiological conditions, the presence of G-actin dominates the kinetics of eIF2αP dephosphorylation.</p><!><p>Upon inhibition of the phosphorylating kinase, the eIF2αP signal decays with a t½ of <10 min (with no change in the total eIF2α content) in both cultured mouse fibroblasts (see Fig. 6 in Ref. 14) and Chinese hamster ovary cells (see Fig. 10 in Ref. 13). Despite the rapid in vivo kinetics of the dephosphorylation reaction, the experiments pointing to G-actin–independent eIF2αP dephosphorylation were conducted with long incubations of 16 h at 30 °C (21). In the absence of other components, PP1α is markedly unstable at 30 °C, losing about half of its activity by 1 h and all detectable activity by 3 h (Fig. S3, A and B). Thus, a stabilizing effect of a PP1 binding co-factor might have accounted for the apparent G-actin–independent stimulatory effect of MBP-PPP1R15A325–636 on PP1α-mediated eIF2αP dephosphorylation at physiological salt concentrations. However, over a range of PP1 concentrations (0.2–200 nm), the presence of MBP-PPP1R15A325–636 failed to stimulate eIF2αP dephosphorylation, regardless of whether PP1N (Fig. 4A) or PP1α (Fig. 4B) was used as the catalytic subunit.</p><!><p>At physiological ionic strength and in the absence of G-actin, PPP1R15A is unable to stimulate dephosphorylation of eIF2αP (despite extended incubation of 16 h). A, top panel, Coomassie-stained PhosTag SDS-PAGE containing dephosphorylation reactions (16 h at 30 °C) in which 2 μm eIF2αP was dephosphorylated by the indicated concentration of PP1N in the presence or absence of PPP1R15A325–636-MBP (50 nm). Quantification of the percentage of dephosphorylation (%dP) is shown below the image. Shown is a representative experiment of two independent repetitions performed. Bottom panel, plot of the rate of dephosphorylation of eIF2αP as a function of PP1N concentration. Data were obtained by quantification of bands of images shown above and the other repeat performed. B, as in A but using PP1α as the source of the catalytic subunit and MBP-PPP1R15A325–636 (50 nm) as the regulatory subunit. Shown is a representative experiment of two independent repetitions performed.</p><!><p>PPP1R15A interacts directly with eIF2α, both in cells (9) and in vitro (21). This interaction maps to the repeat-containing region of PPP1R15A, residues 325–512, N-terminal to PPP1R15A's PP1-binding domain (Fig. 1A) and was proposed to play an important role in the catalytic cycle of PPP1R15A-containing holoenzymes (21). However, in the presence of G-actin, PPP1R15A325–636-MBP and PPP1R15A533–624-MBP (Fig. 1A and Table S1) stimulated eIF2αP dephosphorylation similarly when paired either with PP1N (compare our Figs. 2B and 5A) or with PP1γ (compare Figs. 8A and 2B in Ref. 13). These findings suggest that the conserved C-terminal PPP1R15 fragment that binds PP1 and G-actin simultaneously is sufficient to promote eIF2αP dephosphorylation and to dominate its kinetics in vitro and call into question the importance of the N-terminal repeats in PPP1R15A to the fundamentals of the holoenzyme's catalytic cycle.</p><!><p>The C-terminal portion of PPP1R15A is sufficient to promote eIF2αP dephosphorylation. A, top panel, Coomassie-stained PhosTag SDS-PAGE containing resolved samples from dephosphorylation reactions (30 min at 30 °C) in which 2 μm eIF2αP was dephosphorylated by PP1N (20 nm) in the presence of G-actin (400 nm) and increasing concentrations of PPP1R15A533–624-MBP (0–100 nm). Shown is a representative experiment of three independent repetitions performed. Bottom panel, plot of the rate of dephosphorylation of eIF2αP as a function of PPP1R15A533–624-MBP concentration from the three experiments performed. The EC50 was calculated using the "[Agonist] versus response − variable slope (four parameters)" function in GraphPad Prism v7. The gray lines represent the 95% confidence interval of the fitting. Shown are values obtained for EC50 and information of goodness of the fit (r2). B, as in A but using PP1α (24 nm) in the presence of MBP-PPP1R15A325–636 (24 nm), G-actin (400 nm), and increasing concentrations of MBP-PPP1R15A325–512 as a competitor (0–8 μm). The assays were performed during 20 min at 30 °C. Lane 8, loaded with only MBP-PPP1R15A325–512 shows the absence of a species co-migrating with eIF2α0 (which might otherwise obscure an inhibitory effect on dephosphorylation). Lanes 9 and 10 control for the dependence of enzymatic activity on PPP1R15A and G-actin in this experiment. Quantification of the percentage of dephosphorylation (%dP) is shown below the image. Shown is a representative experiment of two independent repetitions performed.</p><!><p>We considered that an important contributory role for substrate engagement by the PPP1R15A325–533 repeat-containing fragment to the catalytic cycle of the holophosphatase might have been masked by compensatory features that diverge between the different regulatory subunit constructs, fortuitously equalizing their activity. To address this possibility, we measured the ability of MBP-PPP1R15A325–512 containing the repeats but lacking the C-terminal PP1 binding region (Fig. 1A and Table S1) to compete with MBP-PPP1R15A325–636–mediated (G-actin–dependent) eIF2αP dephosphorylation using PP1α as the catalytic subunit. Minimal inhibition of the dephosphorylation reaction was observed at competitor concentrations of up to 8 μm (Fig. 5B), which is a >300-fold excess over the MBP-PPP1R15A325–636 regulatory subunit (present in the reaction at 24 nm) and a concentration of 18-fold above the reported Kd of the interaction between MBP-PPP1R15A325–512 and eIF2αP (21).</p><p>These data suggest that substrate recruitment by the N-terminal extension of PPP1R15A plays a secondary role in the kinetics of the dephosphorylation reaction in vitro and that the reported role of Sephin1 and guanabenz in disrupting that interaction is unlikely to make an important contribution to their pharmacological activity. Consistent with these conclusions, we found that, under physiological salt conditions where eIF2αP dephosphorylation depends on the concentration of PP1α, MBP-PPP1R15A325–636, and G-actin, we were unable to observe an inhibitory effect of either Sephin1 (Fig. 6A and Fig. S4A, lanes 8–11) or guanabenz (Fig. 6B and Fig. S4B, lanes 8–11) at a concentration of up to 100 μm, which exceeds by 100-fold the concentration required for a proteostatic effect in cultured cells (1 μm; see Fig. 1F in Ref. 23). Similarly, no effect of the compounds was observed on the PP1–PPP1R15A holophosphatase activity under low-salt conditions (Fig. S4, A, lanes 1–7, and B, lanes 1–6).</p><!><p>Neither Sephin1 nor GBZ interfere with eIF2αP dephosphorylation. A, Coomassie-stained PhosTag SDS-PAGE containing resolved samples from dephosphorylation reactions (20 min, 30 °C) in which 2 μm eIF2αP was dephosphorylated by PP1α (24 nm) in the presence or absence of MBP-PPP1R15A325–636 (60 nm) and/or G-actin (400 nm). The components were preincubated as specified with either Sephin1 (100 μm), tautomycin (80 nm), or DMSO (vehicle) for 15 min at room temperature before being added to the reaction. The bottom panel shows a long exposure of the relevant section of the image above corresponding to the phosphorylated and nonphosphorylated forms of eIF2α. Quantification of the percentage of dephosphorylation (%dP) is shown below the image. Shown is a representative experiment of three independent experiments performed. B, as in A but with guanabenz (GBZ). Shown is a representative experiment of two independent experiments performed.</p><!><p>Complete inhibition of PPP1R15A-mediated eIF2αP-dephosphorylation by Sephin 1 was reported in an assay conducted over 16 h at 30 °C in a low ionic strength buffer (21). We wished to test whether the reported Sephin1 inhibition might be unmasked by this long incubation (in which the enzyme is undergoing inactivation; Fig. S3B). Using identical MBP∼PPP1R15A325–636 and PP1α constructs, in an identical low ionic strength buffer and following overnight incubation at 30 °C, we observed a 2-fold stimulation of eIF2αP dephosphorylation by MBP∼PPP1R15A325–636 (similar to that noted in shorter reactions; Fig. 3). However, even under these conditions, designed to mimic as closely as possible those used in Ref. 21, the presence of 100 μm Sephin1 was devoid of an inhibitory effect on substrate dephosphorylation (Fig. S4C).</p><!><p>The new experiments presented here cover a range of conditions with realistic concentrations and time regimes. Incorporation of multiple time points and titrations of reaction components enabled a comparison of enzyme kinetics that accounts for the effect of substrate depletion. Our observations were made with four different PPP1R15A preparations, three different PP1 preparations, and both buffer conditions used previously in our laboratory and those used in Ref. 21, all of which consistently show the requirement for G-actin as an additional co-factor in enabling PPP1R15A to stimulate eIF2αP dephosphorylation in vitro. Therefore, the results presented here are in keeping with previous observations that G-actin has an essential role in promoting eIF2αP dephosphorylation both in vitro and in vivo (10, 13, 14).</p><p>The PP1 apo-enzyme is salt-sensitive and inhibited by buffers of physiological ionic strength (29). By contrast, PP1 holoenzymes retain their regulated enzymatic activity at physiological ionic strength (30). These considerations call into question the significance of the 2-fold stimulation of eIF2αP dephosphorylation by PPP1R15A325–636 observed in buffer of low ionic strength. Our experiments also cast doubt on the importance of the physical interaction between the repeat-containing region of PPP1R15A (residues 325–512) and eIF2αP in the substrate-specific dephosphorylation reaction carried out with physiological ionic strength. PPP1R15 regulatory subunits are found throughout the animal kingdom, but only their C-terminal ∼70 residues are conserved (11). This C-terminal fragment contains all the information needed to promote eIF2αP dephosphorylation, as exemplified by its selective hijacking by herpesviruses (12) and by experimentally targeted expression in cells (see Fig. 1C in Ref. 10). In complex with G-actin, the conserved C-terminal fragment of the PPP1R15s is also able to direct PP1 to selectively dephosphorylate eIF2αP in vitro (Figs. 2 and 5A here and Refs. 10, 13).</p><p>The prominent stimulatory role of G-actin on eIF2αP dephosphorylation, observed both in vivo and in vitro, should not obscure the possibility that binary complex formation with PPP1R15 might also favor eIF2αP dephosphorylation independently of G-actin joining the complex. Regulatory subunit binding restricts access to PP1 (24, 31), favoring the phosphorylation of one class of substrates over another. Mere exclusion of some substrates from access to the catalytic subunit might accelerate eIF2αP dephosphorylation when levels of PPP1R15A levels are sufficiently elevated in cells, even though in vitro (and in the absence of competing substrates), the PPP1R15A-PP1 binary complex is not a faster eIF2αP phosphatase than PP1 alone (provided the experiments are conducted at physiological salt concentrations). As neither Sephin1 nor guanabenz affect the stability of the PPP1R15A–PP1 complex (13), it is unlikely that they achieve any measure of inhibition by weakening PPP1R15A's ability to compete with other regulatory subunits for limiting amounts of catalytic subunit. These considerations lead us to propose a dual role for PPP1R15A in cells: diverting limiting amounts of PP1 away from other substrates toward eIF2αP and, in conjunction with G-actin as an essential co-activator, stimulating the intrinsic rate of dephosphorylation by the holoenzyme thus formed. Actin, too, has a dual role in stimulating eIF2αP dephosphorylation: by stabilizing the PPP1R15-PP1 complex (14), G-actin favors the exclusion of other regulatory subunits while stimulating enzyme kinetics selectively toward eIF2αP (Fig. 7).</p><!><p>Model depicting PPP1R15A's role in regulating eIF2αP dephosphorylation. A, in the absence of PPP1R15A, the cellular pool of the catalytic subunit (PP1) is preferentially bound by a variety of regulatory subunits (R1, R2, and R3), which direct its phosphatase activity toward their specific substrates (S1, S2, and S3), excluding eIF2αP. In the Substrate conversion section, see the preferential dephosphorylation of substrates S1, S2, and S3 (solid arrows) compared with eIF2α (dotted arrows). B, rising levels of PPP1R15A recruit PP1 away from other regulatory subunits, redirecting its phosphatase activity toward eIF2αP by excluding other substrates. In the Substrate conversion section, observe the inverted preferential dephosphorylation of substrates compared with A. C, when present, G-actin joins the PPP1R15A–PP1 holophosphatase, increasing its intrinsic eIF2αP-directed catalytic activity. In the Substrate conversion section, see the increased arrow thickness for eIF2αP dephosphorylation compared with B.</p><!><p>Here we present no argument against an important function for the divergent N-terminal extensions of PPP1R15 regulatory subunits. This role may play out in terms of subcellular localization (26) or protein stability (32) and might be influenced by a physical interaction with the substrate (9, 21). However, our findings argue that the physical interaction noted previously between PPP1R15A residues 325–512 and eIF2αP (21) is unlikely to play an important role in formation of the enzyme–substrate complex required for catalysis under physiological conditions, and, hence, its reported disruption by guanabenz or Sephin1 is unlikely to underscore an inhibitory effect on eIF2αP dephosphorylation.</p><p>Most importantly perhaps, the findings presented here argue that the inability of previous efforts to uncover a role for guanabenz or Sephin1 in inhibiting eIF2αP dephosphorylation in vitro (9, 13) was unlikely to have arisen from choice of catalytic subunit, from features of the PPP1R15A regulatory subunit, or the buffer conditions used. Rather, the findings reported here, made in vitro, reinforce observations that Sephin1 and guanabenz have no measurable effect on the rate of eIF2αP dephosphorylation in cells (13). The recent description of PPP1R15A/GADD34-independent cellular effects of guanabenz (33) and our observations that Sephin1-induced changes in gene expression were noted both in cells lacking PPP1R15A and in cells with nonphosphorylatable eIF2α (13) suggest the need to reconsider the role of these two compounds as eIF2αP dephosphorylation inhibitors.</p><!><p>The plasmids used to express protein in E. coli and the sequence of the encoded proteins are listed in Tables S1 and S2.</p><p>PPP1R15A325–636-MBP and PPP1R15A533–624-MBP were produced as described previously (13). Briefly, proteins were expressed in E. coli BL21 (New England Biolabs, catalog no. C3013) as N-terminally tagged GSH S-transferase fusion proteins and purified by tandem affinity chromatography, bound to a GSH-Sepharose 4B resin and eluted with GSH, followed by an overnight cleavage with tobacco etch virus protease (to remove the glutathione S-transferase tag), binding to amylose beads, and elution in maltose-containing buffer.</p><p>MBP-PPP1R15A325–636 and MBP-PPP1R15A325–512 were constructed in the C-terminally hexahistidine tag–containing pMAL-c5x-His plasmid (New England Biolabs, catalog no. N8114). Transformed E. coli BL21 (New England Biolabs, catalog no. C3013) were selected on lysogeny broth (LB) agar plates supplemented with 100 μg/ml ampicillin. A single colony was picked to grow overnight in 5 ml of starter culture that served to inoculate 2 liters of lysogeny broth (LB) (all supplemented with 100 μg/ml ampicillin), which was kept at 37 °C. At A600 = 0.6–0.8, protein expression was induced using 1 mm isopropyl β-d-thiogalactopyranoside at 18 °C for 20 h. Bacteria were pelleted and resuspended in ice-cold His6 lysis buffer containing 50 mm Tris (pH 7.4), 500 mm NaCl, 1 mm MgCl2, 1 mm tris(2-carboxyethyl)phosphine (TCEP), 100 μm phenylmethylsulfonyl fluoride, 20 trypsin inhibitory units per liter aprotinin, 2 μm leupeptin, 2 μg/ml pepstatin, 20 mm imidazole, and 10% glycerol. Bacterial suspensions were lysed using an Emulsi-Flex-C3 homogenizer (Avestin, Inc., Ottawa, ON, Canada) and clarified in a JA-25.50 rotor (Beckman Coulter) at 33,000 × g for 30 min at 4 °C. Pre-equilibrated nickel-nitrilotriacetic acid beads (Qiagen, catalog no. 30230) were incubated with the samples for 2 h at 4 °C. Proteins were eluted in 2 ml of imidazole elution buffer (50 mm Tris (pH 8), 100 mm NaCl, 500 mm imidazole, and 10% glycerol) and incubated with amylose beads (New England Biolabs, catalog no. E8021S) pre-equilibrated with lysis buffer (His6 lysis buffer without imidazole) for 2 h at 4 °C. The amylose beads were batch-washed using 25 bed volumes of lysis buffer, and proteins were eluted with amylose elution buffer (lysis buffer + 10 mm maltose). MBP-R15A325–512 purification required an additional buffer exchange step (into lysis buffer) using Centripure P1 desalting columns (EMP Biotech, catalog no. CP-0110) to eliminate maltose (which appeared to interfere with the dephosphorylation reactions when present at high concentrations).</p><p>MBP∼PPP1R15A325–636 (a gift from the Bertolotti laboratory) was expressed and purified as described previously (21) with minor modifications. The isopropyl β-d-thiogalactopyranoside–induced culture was maintained for 16 h at 18 °C, and 0.5 mm TCEP was included in all buffers, throughout the purification procedure, and in the final dialysis buffer (50 mm Tris (pH 7.4), 200 mm NaCl, and 0.5 mm TCEP).</p><p>For eIF2αP, the N-terminal fragment of human eIF2α (1–185, with three solubilizing mutations) was purified from bacteria and phosphorylated in vitro using the kinase domain of PKR-like ER kinase (PERK), as described previously (10). G-actin was purified from rabbit muscle according to Ref. 34 as modified in Ref. 10. PP1γ (7–300) was purified according to Ref. 13. PP1α (7–330) was purified from BL21 E. coli according to Refs. 28, 35. PP1N was purified from rabbit muscle according to Ref. 25.</p><!><p>Unless otherwise stated, dephosphorylation reactions were performed at a final volume of 20 μl by assembling 5 μl of 4× solution of each component: PP1, PPP1R15A, G-actin, and eIF2αP (or their respective buffers). A 10× assay buffer (500 mm Tris (pH 7.4), 1 m NaCl, 1 mm EDTA, 0.1% Triton X-100, and 10 mm MgCl2) was diluted 1:10, supplemented with 1 mm DTT, and used to create working solutions of PP1, PPP1R15A, and eIF2αP at the desired concentrations. G-actin working solutions were created using G buffer (2 mm Tris-HCl (pH 8), 0.2 mm ATP, 0.5 mm DTT, and 0.1 mm CaCl2). Holoenzyme components (PP1, PPP1R15A, and G-actin) were combined first, and substrate (eIF2αP) was added last to initiate the reactions, which were conducted under shaking at 500 rpm and at 30 °C for the specified time. The final buffer composition was 36 mm Tris (pH 7.4), 76 mm NaCl, 74 μm EDTA, 0.007% Triton X-100, 0.7 mm MgCl2, 25 μm CaCl2, 0.05 mm ATP, 0.8 mm DTT, 0.5 μm Latrunculin B, 0.4–3 μm MnCl2, 0.5% glycerol, and 50 μm TCEP in the experiments performed for Figs. 1, 2, and 4–6 and Figs. S1, S3, and S4, A, lanes 8–11, and B, lanes 7–10.</p><p>Dephosphorylation reactions designed to reproduce the observations in Ref. 21 were performed in the assay buffer described therein (50 mm Tris-HCl (pH 7.4), 1.5 mm EGTA (pH 8.0), and 2 mm MnCl2), with the modification that 0.5 mm TCEP was added to disfavor oxidative inactivation of the enzyme. The NaCl content of the final reaction was constrained by the contribution of the protein solutions added to each reaction. To maintain parity between reactions performed with and without PPP1R15A, an equal volume of the PPP1R15A buffer was added to reactions lacking the protein. The final salt concentration in the various reactions is noted in the figure legends.</p><p>The stability test of PP1α (Fig. S3) was performed by preparing a fresh 240 nm solution of PP1α in the assay buffer described above. Separate aliquots were preincubated either at 30 °C or on ice for the specified times (30 min to 7 h, see schematic in Fig. S3A). At termination of the preincubation, 5 μl of these preincubated solutions were added to 20 μl of dephosphorylation reactions as described above.</p><p>Dephosphorylation reactions to test the activity of Sephin1 or guanabenz (Fig. 6 and Fig. S4) included 15-min preincubation of the enzymatic components at room temperature (before the addition of substrate) with either Sephin1 (Enamine, catalog no. EN300-195090), guanabenz (Sigma-Aldrich, catalog no. D6270), tautomycin (Calbiochem, catalog no. 5805551), or an equal volume of DMSO (vehicle).</p><p>Reactions were terminated by addition of 10 μl of 3× Laemmli buffer supplemented with 100 mm DTT and heating the samples for 5 min at 70 °C. A third (10 μl) of the final volume was resolved in 12.5% PhosTag SDS gels (Wako, catalog no. NARD AAL-107) at 200 V for 1 h. Gels were stained with Coomassie Instant Blue and imaged on an Odyssey imager (LI-COR, Lincoln, NE).</p><p>ImageJ was used to quantify eIF2αP dephosphorylation, as reflected by the intensity of the fluorescence arising from the Coomassie stain of the eIF2αP and eIF2α0 bands resolved by the PhosTag SDS-PAGE gels and captured as a TIF file on the Odyssey imager. GraphPad Prism v8 was used to fit the plot and perform statistical analysis. Table S3 lists the number of times each experiment was performed.</p><!><p>A. C.-C. and D. R. conceptualization; A. C.-C. data curation; A. C.-C. formal analysis; A. C.-C. investigation; A. C.-C., Z. C., M. S. C., W. P., M. B., and D. R. writing-review and editing; D. R. supervision; D. R. funding acquisition; D. R. writing-original draft; D. R. project administration. A. C.-C. conceived the study, co-designed and conducted the experiments, interpreted the results, created the figures, and co-wrote the paper. Z. C. co-designed the experiments, assisted with the preparation of PP1 from rabbit muscle, interpreted the results, and edited the manuscript. M. S. C. expressed and purified PP1α from E. coli, interpreted the results, and edited the manuscript. W. P. oversaw the expression and purification of PP1α from E. coli, interpreted the results, and edited the manuscript. M. B. co-designed the experiments, oversaw the purification of PP1 from rabbit muscle, interpreted the results, and edited the manuscript. D. R. conceived the study, co-designed the experiments, interpreted the results, and co-wrote the paper.</p><!><p>This work was supported by a Wellcome Trust principal research fellowship (Wellcome 200848/Z/16/Z) to D. R. and a Wellcome Trust strategic award to the Cambridge Institute for Medical Research (Wellcome 100140). The authors declare that they have no conflicts of interest with the contents of this article. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.</p><p>This article contains Figs. S1–S4 and Tables S1–S3.</p><p>integrated stress response</p><p>eukaryotic initiation factor</p><p>maltose-binding protein</p><p>endoplasmic reticulum</p><p>tris(2-carboxyethyl)phosphine.</p>

PubMed Open Access

Mechanistic role of cytochrome P450 (CYP)1B1 in oxygen-mediated toxicity in pulmonary cells: a novel target for prevention of hyperoxic lung injury

Supplemental oxygen, which is routinely administered to preterm infants with pulmonary insufficiency, contributes to bronchopulmonary dysplasia (BPD) in these infants. Hyperoxia also contributes to the development of acute lung injury (ALI) and acute respiratory distress syndrome (ARDS) in adults. The mechanisms of oxygen-mediated pulmonary toxicity are not completely understood. Recent studies have suggested an important role for cytochrome P450 (CYP)1A1/1A2 in the protection against hyperoxic lung injury. The role of CYP1B1 in oxygen-mediated pulmonary toxicity has not been studied. In this investigation, we tested the hypothesis that CYP1B1 plays a mechanistic role in oxygen toxicity in pulmonary cells in vitro. In human bronchial epithelial cell line BEAS-2B, hyperoxic treatment for 1\xe2\x80\x933 days led to decreased cell viability by about 50\xe2\x80\x9380%. Hyperoxic cytotoxicity was accompanied by an increase in levels of reactive oxygen species (ROS) by up to 110%, and an increase of TUNEL-positive cells by up to 4.8-fold. Western blot analysis showed hyperoxia to significantly down-regulated CYP1B1 protein level. Also, there was a decrease of CYP1B1 mRNA by up to 38% and Cyp1b1 promoter activity by up to 65%. On the other hand, CYP1B1 siRNA appeared to rescue the cell viability under hyperoxia stress, and overexpression of CYP1B1 significantly attenuated hyperoxic cytotoxicity after 48 h of incubation. In immortalized lung endothelial cells derived from Cyp1b1-null and wild-type mice, hyperoxia increased caspase 3/7 activities in a time-dependent manner, but endothelial cells lacking the Cyp1b1 gene showed significantly decreased caspase 3/7 activities after 48 and 72 h of incubation, implying that CYP1B1 might promote apoptosis in wild type lung endothelial cells under hyperoxic stress. In conclusion, our results support the hypothesis that CYP1B1 plays a mechanistic role in pulmonary oxygen toxicity, and CYP1B1-mediated apoptosis could be one of the mechanisms of oxygen toxicity. Thus, CYP1B1 could be a novel target for preventative and/or therapeutic interventions against BPD in infants and ALI/ARDS in adults.

mechanistic_role_of_cytochrome_p450_(cyp)1b1_in_oxygen-mediated_toxicity_in_pulmonary_cells:_a_novel

1. Introduction<!>2.1. Cell culture<!>2.2. Cell proliferation assays<!>2.3. Intracellular ROS<!>2.4. TUNEL fluorescein assay<!>2.5. Western blot analysis of CYP1B1 apoprotein<!>2.6. qPCR<!>2.7. Cyp1b1 promoter assay<!>2.8. CYP1B1 siRNA knockdown<!>2.9. ApoTox-Glo Triplex Assay for cytotoxicity and caspase 3/7 activities<!>3.1 Cytotoxicity of hyperoxia to BEAS-2B cells<!>3.2 Hyperoxia downregulated CYP1B1 in BEAS-2B cells<!>3.3 CYP1B1 siRNA protected cells from hyperoxic cytotoxicity<!>3.4 CYP1B1 overexpression promoted hyperoxic cytotoxicity<!>3.5 Knockout of Cyp1b1 gene in lung endothelial cells alleviated hyperoxic toxicity<!>DISCUSSION<!>

<p>Supplemental oxygen, which is frequently administered to premature infants with pulmonary insufficiency, is one of the major risk factors for the development of bronchopulmonary dysplasia (BPD) in premature infants [1]. Hyperoxia also contributes to the development of acute lung injury (ALI) and acute respiratory distress syndrome (ARDS) in adults. Infants with BPD often are re-hospitalized before 1 year of age, and have worse neurological outcomes with significantly higher rates of cerebral palsy, spastic dyplegia, and lower cognitive and language scores [2]. BPD has a multifactorial etiology, including genetic predisposition, prematurity, mechanical ventilation, oxygen exposure, and inflammation [3].</p><p>Supplemental oxygen, which is often a life-saving therapy for premature neonates and ALI/ARDS patients, may lead to hyperoxia, which in turn is a major risk factor for the development of lung injury that is associated with increased pulmonary permeability, increased inflammatory cell count, and injuries to endothelial and epithelial cells [4]. Multiple studies have shown an involvement of CYP enzymes CYP1A (i.e. CYP1A1 and CYP1A2) in oxygen toxicity [5–8]. CYP is a family of heme-containing proteins that are involved in the metabolism of numerous endogenous substrates and xenobiotics [9]. Induction of CYP1A by β-naphtoflavone (BNF) or 3-methylcholanthrene (MC) prior to hyperoxia protects mice and rats against the toxic effects of oxygen exposure [10, 11]. Meanwhile, pre-treatment of rats with a CYP1A inhibitor, aminobenzotriazole, followed by exposure to 95% O2 leads to severe inflammation, pleural effusions, and severe lung injury [12]. During the initial 48 h of hyperoxia exposure, pulmonary and hepatic CYP1A upregulation is observed [12, 13]. Between 48–60 h, the animals develop severe respiratory distress, accompanied with downregulation of CYP1A [6, 12]. Aryl hydrocarbon receptor (AhR) is a key regulatory transcription factor of many CYP proteins and other important developmental genes, including CYP1A1/2 [14, 15]. Experiments with AhR null mice indicated that the protective effect of CYP1A1/2 on hyperoxic toxicity is dependent on the AhR in the lung [16].</p><p>CYP1B1 is the newest member of CYP1 family, described in 1994, after being cloned from a 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD)-treated human keratinocyte cell line [17], and shares only 40% homology with CYP1A1 and CYP1A2 [17]. CYP1B1 activates PAH in human lung [18] and inactivates benzo[a]pyrene (BP) in mouse aortic smooth muscle cells [19]. CYP1B1 is endogenously expressed in the lung and other tissues, but not hepatocytes [20]. BNF induces CYP1B1 mRNA, but to a lesser extent than that of CYP1A1 [21, 22]. Thakur et al. [23] have reported that maternal treatment of BP, followed by 7-days of postnatal hyperoxia exposure leads to upregulation of CYP1B1 [23].</p><p>CYP1B1 has a conserved DNA sequence and appears in the early stages in ontogenesis [24]. It may play a role in evolution and development. In this investigation, we tested the hypothesis that CYP1B1, which is also regulated by the AHR, plays a mechanistic role in oxygen toxicity in pulmonary cells in vitro.</p><!><p>BEAS-2B and H358 cells (ATCC, Manassas, VA) were maintained in RPMI 1640 or high-glucose DMEM containing 10% fetal bovine serum (FBS), 100 U/ml penicillin, and 100 µg/ml streptomycin. The cells were maintained in room air (RA) or exposed to hyperoxia (95% O2 and 5% CO2), as described earlier [25].</p><p>Lung endothelial cells were isolated from wild-type and Cyp1b1-null mice as previously described [26, 27]. The EC growth medium was DMEM containing 10% FBS, 2 mM L-glutamine, 2 mM sodium pyruvate, 20 mM HEPES, 1% nonessential amino acids, 100 µg/ml streptomycin, 100 U/ml penicillin, freshly added heparin at 55 U/ml (Sigma), 100 µg/ml endothelial growth supplement (Sigma), and murine recombinant interferon-γ (R&D Systems, Minneapolis, MN) at 44 U/ml. Cells were incubated at 33°C with 5% CO2 and progressively passaged to larger plates. Cells were normally maintained in 60-mm dishes coated with 1% gelatin prepared in phosphate-buffered saline (PBS) [27].</p><p>To establish the CYP1B1-overexpressed cell line, total RNA was isolated from MC-treated BEAS-2B cells using RNeasy Mini Kit (Qiagen, Germantown, MD), and was subjected to reverse transcription (Bio-Rad, Hercules, CA). RT-PCR was performed using 5'-ATGCTAGCGCCGCCACCATGGGCACCAGCCTCAG-3' and 5'-TAGGTACCCTTATTGGCAAGTTTCCTTGG-3' as the cloning primers. The open reading frame of human CYP1B1 cDNA was subcloned into pcDNA3.1 between the NheI/KpnI sites. The insert sequence of pCD-CYP1B1 was verified by DNA sequencing. To obtain the stable overexpressed cells, pCD-CYP1B1 or pcDNA3.1 was transfected into H358 cells using SuperFect (Qiagen), and maintained in 500 µg/ml geneticin (Life Technologies). Clones that overexpressed CYP1B1 were screened by CYP1B1 antibody (Santa Cruz Biotechnology, Santa Cruz, CA) using immunofluorescence staining, and verified by real-time RT-PCR (data not shown).</p><!><p>The conventional trypan blue exclusion assay was performed using 0.4% trypan blue as previously described [28]. 3-(4,5-dimethylthiazolyl-2)-2,5-diphenyltetrazolium bromide) (MTT) assay was performed according to the manufacturer's instruction (ATCC) as previously described [15].</p><!><p>Intracellular ROS level was quantified according to the manufacturer's instruction (Life Technologies) as previously described [15].</p><!><p>TUNEL assay was performed using Click-iT TUNEL Alexa-Fluor Imaging Assay Kit according to the manufacturer's instruction (Life Technologies). Briefly, cells in 96-well black-walled plates (BD Biosciences) were fixed with 4% paraformaldehyde, permeabilized with 0.25% Triton X-100, incubated with 50 µl/well of TdT reaction buffer at RT for 10 min, incubated with 50 µl/well of TdT reaction cocktail at 37°C for 60 min, washed twice with 3% BSA in PBS, and incubated with 50 µl/well of the Click-iT reaction cocktail at RT for 30 min. After the final wash with PBS containing 3% BSA, fluorescence was measured with excitation at 495 nm and emission at 519 nm.</p><!><p>Cell lysates (30 µg protein per well) were resolved in SDS-PAGE and detected with CYP1B1 antibody (Santa Cruz Biotechnology) or β-actin (Actb) antibody (Sigma-Aldrich) by Western blot analyses, as described previously [13, 28].</p><!><p>Total RNA was extracted from the cell lysates using RNeasy Mini Kit (Qiagen). Message RNA was quantified with SuperScript III Two-Step qRT-PCR Kit (Life Technologies). The primer sequences were 5'-CACTGCCAACACCTCTGTCTT-3' and 5'-CAAGGAGCTCCATGGACTCT-3' for CYP1B1 [29], 5'-AAACGCATTAACTGGCGAAC-3' and 5'-GAACTCCAGGAGAACTGCAAA-3' for human ornithine decarboxylase antizyme OAZ1 [29]. The SYBR Green (Bio-Rad) qPCR condition was 95°C fo r 5 s (denaturation), 65°C for 10 s (annealing), and 72°C for 20 s (extension) for CYP1 B1, and was 95°C for 5 s, 60°C for 10 s, and 72°C for 20 s for OAZ1. The PCR specificity was confirmed with the melting curves. The PCR efficiency was 95.7% for CYP1B1 and 96.5% for OAZ1.</p><!><p>The dual-luciferase assay (Promega, Madison, WI) was performed in BEAS-2B cells double transfected with 2 plasmid constructs using Qiagen SuperFect. The firefly luciferase construct p1.1 contained the rat Cyp1b1 5'-UTR and 1.0 Kb of the proximal 5'-flanking sequence. The renilla luciferase construct was pRL-TK (Promega). Cyp1b1 promoter activities were determined by the dual-luciferase assay, which entailed normalizing the firefly luciferase activities against those of renilla luciferase.</p><!><p>Cells were transfected with either ON-TARGETplus hCYP1B1 siRNA or the non-targeting control (Dharmacon, Chicago, IL) using Lipofectamine 2000 (Life Technologies). The siRNA effect was examined by qPCR as described in Materials and methods.</p><!><p>Cytotoxicity and caspase 3/7 activities were determined using ApoTox-Glo™ Triplex Assay (Promega) according to the manufacturer's instruction. Briefly, cells in 96-well black-walled plates were incubated with 20 µl of Cytotoxicity Reagent at 37°C for 30 min. Cytot oxicity was determined by fluorescence (excitation at 485 nm and emission at 520 nm). The cells were then incubated with 100 µl of Caspase-Glo 3/7 Reagent at RT for 30 min. The caspase 3/7 activities were determined by bioluminescence.</p><!><p>Hyperoxia impairs lung development in premature babies, as well as in newborn mice and other animals [30, 31]. In vitro experiments consistently demonstrated hyperoxic toxicities to pulmonary cell lines such as H358 (unpublished data), H441, and A549 [25]. In BEAS-2B cells, trypan blue exclusion assay showed that the number of live cells increased by about 60% per day under RA conditions (RA) (Figure 1A). Hyperoxia (95% O2 plus 5% CO2) [28] showed no effect on cell proliferation during the first 24 h, but exhibited 44 and 81% inhibition at 48 and 72 h, respectively, based on cell numbers (Figure 1A). The MTT cell proliferation assay measures the activity of NAD(P)H-dependent oxidoreductases which represents the metabolic rate of entire cell population, live and dead, in each well. Hyperoxia decreased the A570nm in the MTT assay of BEAS-2B cells by 14%, 24%, and 51% at 24, 48, and 72 h, respectively (Figure 1B).</p><p>According to the literature, hyperoxic cytotoxicity is associated with increased production of ROS [32]. We measured the effect of hyperoxia on intracellular ROS in BEAS-2B cells using CM-H2DCFDA as the probe. ROS converts the fluorescent probe into 5-(and 6-)chloromethyl-2′,7′-dichlorofluorescin (CM-DCF). As anticipated, we found that hyperoxia increased the CM-DCF fluorescence or intracellular ROS by 26% at 48 h and 110% at 72 h (Figure 1C).</p><p>Since hyperoxia caused cell death (Figure 1A), we performed TUNEL apoptosis assay, a method based on terminal deoxynucleotidyl transferase (TdT)-associated incorporation of dUTPs at the 3'-OH groups of fragmented DNA. Hyperoxia increased dUTP incorporation in the BEAS-2B cells by 1.5-, 2.7-, and 4.8-fold at 24, 48, and 72 h, respectively (Figure 1D), indicating the involvement of apoptosis in the ROS-associated hyperoxic cytotoxicity.</p><!><p>Previous reports indicate that hyperoxia induces CYP1A1 in the lung or cultured pulmonary cells [5, 25]. When BEAS-2B pulmonary cells were exposed to hyperoxia, CYP1B1 apoprotein was significantly downregulated at 24 and 48 h in Western blot analysis (Figure 2A). qPCR indicated that hyperoxia decreased CYP1B1 mRNA level by 38%, 21%, and 19% at 24, 48, and 72 h, respectively (Figure 2B). The reference gene OAZ1 was not affected by hyperoxia, consistent with our previous publication on the effect of hyperoxia in H441 cells [30]. CYP1A1 mRNA was induced by 94% by a 24 h hyperoxic treatment (not shown), consistent with previous observations [25].</p><p>In order to investigate the mechanisms of downregulation of Cyp1b1a 1.1 Kbp of rat Cyp1b1 promoter sequence was subcloned into pGL3 luciferase reporter system (see Materials and methods). Dual luciferase assay of the p1.1/pRL double transfected BEAS-2B cells indicated that Cyp1b1 promoter activity was down-regulated by hyperoxia by 45–65% at 24, 48, and 72 h (Figure 2C).</p><!><p>BEAS-2B cells were transiently transfected with either non-specific control siRNA or CYP1B1 siRNA. The qPCR analysis indicated that CYP1B1 siRNA significantly knocked down CYP1B1 mRNA in both room air (21.0%) and hyperoxia (85.5%) conditions (Figure 3A). Cells that were transfected with CYP1B1 siRNA showed elevated the cell viability under hyperoxia conditions by 44% (Figure 3B), suggesting that CYP1B1 siRNA rescued cells from oxygen toxicity, albeit this was not statistically significant.</p><!><p>CYP1B1-H358 cells were generated by stably transfecting CYP1B1 cDNA into H358 cells using a pCDNA vector. The control cells (pCD-H358 cells) were created by transfecting the H358 cells with the empty vector pCDNA3.1(+). The cells were maintained in room air or subjected to hyperoxia for 24, 48, and 72 h. MTT assay measures the total NAD(P)H-dependent oxidoreductase activity in each well. But the two stable cell lines exhibit a different growth rate, resulting in different amount of cells in each well. Therefore, in Figure 3C, we converted the original MTT A570nm readings into cell densities, using a standard curve of MTT A570nm readings versus cell count. We found that CYP1B1-H358 was more susceptible to hyperoxic toxicity (Figure 3C). A 48 h hyperoxic treatment caused a greater decrease in cell viability of CYP1B1-H358 cells, as compared to pCD-H358 cells, and this difference was much more pronounced at 72 h (Figure 3C).</p><!><p>Lung endothelial cell line prepared from wild-type and Cyp1b1-null mice were subjected to RA and hyperoxic condition. Hyperoxic toxicity to cell viability was observed in both cell line (Figure 4A). Although the Cyp1b1−/− cells grew faster than the Cyp1b1+/+ (WT) cells (Figure 4A), the hyperoxia-induced increase of caspase 3/7 activities was lower in the Cyp1b1−/− cells (Figure 4B). The differences became statistically significant at 72 h (Figure 4B), suggesting Cyp1b1 expression promoted apoptosis under hyperoxic stress.</p><!><p>Lung injury due to prolonged hyperoxia is characterized by increased ROS production [32]. ROS could initiate lung damage by oxidative damage to proteins, lipids, and DNA, which could in turn lead to enhanced expression of pro-inflammatory genes [33]. ROS plays a central role in the subsequent extensive inflammatory response, destruction of the alveolo-capillary barrier, impaired gas exchange, and pulmonary edema [34].</p><p>In this study, we focused on the mechanistic role of CYP1B1 in oxygen-mediated pulmonary toxicity using human pulmonary cell lines. Using BEAS-2B cells, we found that the onset of hyperoxic cytotoxicity was most noticeable at 48 h. No significant cell death (Figure 1A), ROS production (Figure 1C), or dUTP incorporation-associated apoptosis (Figure 1D) was observed after 24 h exposure of 95% oxygen. This observation was consistent with our previous observations that CYP1A1 was induced by hyperoxia within 24 h, and this may have led to protection of the cells from hyperoxic toxicity in that time frame [25]. At later time points, CYP1A1 expression declined and there was more cytotoxicity, suggesting that CYP1A1 decrease may have contributed to the toxicity mediated by oxygen [25].</p><p>The mechanisms by which hyperoxia caused downregulation of CYP1B1 at the protein (Figure. 2A) and mRNA levels (Figure. 2B) is not completely understood, but probably entailed attenuation of CYP1B1 gene at the transcriptional level based on our transient transfection experiments using the luciferase reporter gene (Figure. 2C). However, that fact that CYP1B1 mRNA repression by hyperoxia was modest relative to that of CYP1B1 protein could be explained by a combination of transcriptional and post-transcriptional/translational mechanisms. It is possible that ROS produced under hyperoxic conditions may have decreased the stability of CYP1B1 protein [35].</p><p>The downregulation of CYP1B1 might be a protective mechanism because we found that CYP1B1 promoted hyperoxic cytotoxicity. This idea is supported by our observation that overexpression of CYP1B1 in the cells led to increased cytotoxicity caused by hyperoxic incubation for 48 h or longer (Figure 3C). Furthermore, knock-down of CYP1B1 mRNA rescued the cells from toxicity after 48 h hyperoxia (Figure. 3A&B), although this was statistically significant. Conditional deletion of CYP1B1 in endothelial cells significantly decreased hyperoxia-mediated apoptosis (Figure 4B). The fact that endothelial cells lacking CYP1B1 showed lesser apoptosis and toxicity than wild type cells supports the hypothesis that CYP1B1 in pulmonary endothelial cells in vivo contributes to the pro-oxidant effects under hyperoxic conditions. Our recent studies showing decreased hyperoxic lung injury in mice lacking the gene for Cyp1b1 (Veith et al., 2016, unpublished results) lends further credence to this hypothesis.</p><p>Tang et al. reported that Cyp1b1 deficiency in retinal endothelial cells showed increased oxidative stress and protection against abnormal angiogenesis, and this was due increased production of thrombospondin-2 (TSP-2), an inhibitor of angiogenesis [36].</p><p>In summary, the results of this investigation support the hypothesis that CYP1B1 plays a mechanistic role in pulmonary oxygen toxicity, and CYP1B1-mediated apoptosis could be one of the mechanisms of oxygen toxicity. CYP1B1 could thus be one of the targets for preventative and/or therapeutic interventions against BPD in infants and ALI/ARDS in adults.</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

Cbr1 is a Dph3 reductase required for the tRNA wobble uridine modification

Diphthamide and the tRNA wobble uridine modifications both require Dph3 (DiPHthamide biosynthesis 3) protein as an electron donor for the iron-sulfur clusters in their biosynthetic enzymes. Here, using a proteomic approach, we identified Saccharomyces cerevisiae cytochrome B5 reductase (Cbr1) as a NADH-dependent reductase for Dph3. The NADH- and Cbr1-dependent reduction of Dph3 may provide a regulatory linkage between cellular metabolic state and protein translation.

cbr1_is_a_dph3_reductase_required_for_the_trna_wobble_uridine_modification

<!>Yeast Strains<!>Sample preparation of Dph3 SILAC study<!>Nano LC/MS/MS and data Analysis of Dph3 SILAC interactome sample<!>Cloning, expression and purification of Cbr1, Mcr1, Pga3, and Ncp1<!>Ultraviolet\xe2\x80\x93visible spectroscopy of Dph3 reduction reactions<!>Anaerobic reconstitution of the first step of yeast diphthamide biosynthesis<!>Purification of eEF2 and its in vitro ADP-ribosylation by DT<!>DT and \xce\xb3-toxin sensitivity assays<!>Cloning, expression and purification of \xce\xb3-toxin<!>Bulk tRNA isolation and in vitro \xce\xb3-toxin treatment

<p>Dph3 (DiPHthamide biosynthesis 3), also known as Kti11 (Kluveromyces lactis Toxin Insensitive 11), is required for two highly conserved modifications in eukaryotes: diphthamide, a unique protein post-translational modification on eukaryotic elongation factor 2 (eEF2), and the tRNA wobble uridine modification, 5-carboxymethyl-2-thiouridine (mcm5s2U). These two distinct modifications were both suggested to be important for translation fidelity.1,2</p><p>Studies of the diphthamide biosynthesis pathway have elucidated the molecular function of Dph3. Formation of diphthamide in eukaryotes takes four steps (Supplementary Results, Supplementary Fig. 1a), requiring seven proteins (Dph1-Dph7).3-5 The first step requires an unconventional radical SAM enzyme, the Dph1-Dph2 heterodimer. Dph1-Dph2 contains [4Fe-4S] clusters and relies on Dph3 as an electron donor to keep the [4Fe-4S] clusters in the active and reduced state.6,7 The functions of Dph5, Dph6, and Dph7 in the subsequent steps of the biosynthesis pathway are well characterized4,5,8 while the role of Dph4 is still unknown.</p><p>In eukaryotes, approximately 25% of cytoplasmic tRNAs have the wobble uridine modified to 5-methoxycarbonylmethyluridine (mcm5U), 5-carbamoylmethyl-2'-O-methyluridine (ncm5U) or mcm5s2U. Synthesis of a common intermediate, 5-carboxymethyluridine (cm5U), requires the eukaryotic elongator complex consisting of six subunits (Elp1-Elp6) and seven other associated proteins (Kti11-Kti14, Sit4, Sap185 and Sap190) in eukaryotes (Supplementary Fig. 1b).9,10 Elp3, a radical SAM enzyme, is the catalytic subunit for the first step. Recombinant archaeal Elp3 from Methanocaldococcus infernus catalyzes the formation of cm5U in the presence of SAM and sodium dithionite in vitro via a radical mechanism.11 Given that Dph3 (also known as Kti11) is an electron donor for the radical SAM enzyme Dph1-Dph2 in diphthamide biosynthesis, it is believed that Dph3 also acts as an electron carrier for Elp3 in the tRNA modification reaction.12 Thus, Dph3 connects the two modifications that are important for translation elongation. However, the physiological reductase(s) that ultimately provides electrons to Dph3 is not known.</p><p>To identify candidate reductase(s) for Dph3, we performed a protein interactome study on Dph3 in Saccharomyces cerevisiae (Supplementary Fig. 2a) using stable isotope labeling by amino acids in cell culture (SILAC).13 We generated a yeast BY4741 strain expressing endogenous level of FLAG-tagged Dph3 by inserting a C-terminal triple flag tag on the endogenous dph3 gene and cultured this strain in heavy media. A BY4741 strain expressing untagged Dph3 was cultured in light media. The cell lysates were immunoprecipitated with anti-FLAG resins. The eluted fractions from both immunopreicipitation were mixed, precipitated, digested with trypsin, and analyzed by mass spectrometry to identify proteins with high heavy to light (H/L) ratios, which would be potential interacting partners of Dph3. Among the list of proteins with high H/L ratios (Supplementary Fig. 2b), we found known interactors of Dph3, such as Dph1-Dph2 and the elongator complex subunits Elp1-Elp3.14 Our SILAC also identified Kti13. This is consistent with the recent findings reporting that Kti13 forms a heterodimer with Dph3.15,16 The identification of known Dph3 interacting partners suggested that our SILAC results were reliable. We suspected that the reductase is likely a flavin-containing protein because electron transfer in cells from the common two-electron donors (NADH or NADPH) to Fe-containing one-electron acceptors typically require flavin cofactors, which are capable of both two-electron and one-electron transfers. To our delight, we found an NADH dependent flavoprotein, cytochrome b5 reductase (Cbr1), among the proteins with high H/L ratio (Supplementary Fig. 2b). Cbr1 is a transmembrane protein embedded in the endoplasmic reticulum (ER) membrane and mitochondrion outer membrane with the catalytic domain residing on the cytosolic side (Fig. 1a).17 Therefore, it is plausible that Cbr1 reduces the cytosolic Dph3.</p><p>To test if Cbr1 could reduce Dph3 in vitro, we cloned, expressed and purified the recombinant Cbr1. We first monitored the reduction of Dph3 by detecting the 488 nm absorption of oxidized Dph3 as previously described.7 Upon addition of NADH to initiate the reaction, Dph3 was rapidly reduced by Cbr1 (Fig. 1b). In contrast, addition of NADPH did not lead to reduction of Dph3, suggesting that Cbr1 is an NADH-specific enzyme (Supplementary Fig. 3). To confirm Cbr1's role as a Dph3 reductase, we tested if this reduction system can be used to reduce Dph1-Dph2 and reconstitute the first step of diphthamide biosynthesis in vitro. Using 14C-SAM, the substrate eEF2 was radioactively labeled in the presence of Dph1-Dph2, Dph3, Cbr1 and NADH (Fig. 1c, and Supplementary Fig. 7, lane3). The Cbr1/NADH/Dph3 reduction system was able to reduce Dph1-Dph2 similarly to dithionite, a chemical reductant for Fe-S clusters (Fig. 1c, and Supplementary Fig. 7, lane 1). These results demonstrated that Cbr1 can reduce Dph3 using NADH in vitro.</p><p>To confirm that Cbr1 is the reductase of Dph3 in vivo, we examined the formation of diphthamide in CBR1 knockout (cbr1Δ) yeast. Diphthamide is targeted by the diphtheria toxin (DT) which inactivates eEF2 and kills the cells.18 Thus, we used the established DT sensitivity assay19 to test if cbr1Δ strain abolished diphthamide formation. Interestingly, the cbr1Δ strain conferred no resistance to DT (Supplementary Fig. 4), suggesting that formation of diphthamide is not affected. One possibility is that multiple proteins could serve as reductases for Dph3. Thus, we used the Basic Local Alignment Search Tool (BLAST) to search for other potential reductases in Saccharomyces cerevisiae using the protein sequence of Cbr1. Three proteins with highly similar sequence to Cbr1 were found: Mitochondrial cytochrome b5 reductase (Mcr1), Plasma membrane-associated coenzyme Q6 reductase (Pga3) and Altered inheritance of mitochondria protein 33 (Aim33). Since both Cbr1 and the NADPH dependent Cytochrome P450 reductase (Ncp1) were reported to provide electrons for Cytochromes P450 involved in ergosterol biosynthesis,20 we also tested Ncp1 for Dph3 reduction. We cloned, overexpressed and purified the recombinant Mcr1, Pga3 and Ncp1 (over-expression of the putative protein Aim33 in E.coli or yeast did not yield any protein) and tested their Dph3 reduction activity by monitoring the 488 nm absorption. Interestingly, we found that both Mcr1 and Ncp1 reduced Dph3 at a slower rate compared to that of Cbr1 under similar reaction conditions (Supplementary Fig. 5a and 5c). Pga3 displayed no Dph3 reduction activity (Supplementary Fig. 5b). We found that both the Cbr1/NADH/Dph3 or Ncp1/NADPH/Dph3 reduction systems were able to reduce Dph1-Dph2 and support the first step of diphthamide in vitro (Supplementary Fig. 5d and Supplementary Fig. 8). These results suggested that Mcr1 and Ncp1 could also be reductases for Dph3.</p><p>We next examined diphthamide formation in multiple reductase deletion strains by DT sensitivity assay. We failed to generate a cbr1Δmcr1Δncp1Δ strain as NCP1 was found to be an essential gene by the Saccharomyces Genome Deletion Project (http://wwwsequence.stanford.edu/group/yeast_deletion_project/Essential_ORFs.txt). Surprisingly, the cbr1Δmcr1Δ strain still did not confer any resistance to DT (Supplementary Fig. 4). The lack of obvious phenotype in diphthamide biosynthesis for the double deletion strain prompted us to investigate the tRNA modifications in these reductase deletion strains. We reasoned that the synthesis of the much more abundant wobble uridine modifications (about 25% of the tRNA population)12 is likely to depend more heavily on the efficient reduction of Dph3 compared to the formation of the irreversible diphthamide modification on the eEF2 with a slow protein turnover rate.21 The mcm5s2U modified tRNAs are targeted by the Kluyveromyces lactis killer toxin which cleaves the modified tRNAs, leading to cell-cycle arrest. Therefore, we tested if the reductase(s) deletion strains conferred resistance to inducible expression of γ-toxin, the killer toxin catalytic subunit, using a reported assay.22,23 As expected, BY4741 wild type and dph2Δ strain which contained the mcm5s2U tRNA modification were unable to grow when the expression of γ-toxin was induced (Fig. 2a). The dph3Δ or elp3Δ strain lacking the mcm5s2U tRNA modification survived under such conditions. Remarkably, we found that CBR1 deletion alone conferred some resistance to γ-toxin, suggesting that formation of tRNA modifications was partially impaired. Furthermore, while mcr1Δ strain did not confer any resistance, cbr1Δmcr1Δ strain exhibited greater resistance to γ-toxin. To confirm that the mcm5s2U formation is impaired in cbr1Δ strain, we isolated total tRNAs and treated the tRNAs with γ-toxin to probe for mcm5s2U. Using northern blot, we found that the glu-tRNA from cbr1Δ strain had significantly lower cleaved glu-tRNA fragment compared to that from BY4741 WT strain (Fig. 2b, and Supplementary Fig. 6). Taken together, these results support a role for Cbr1 as the major reductase of Dph3 under normal physiological conditions.</p><p>Interestingly, a small fraction of mcm5s2U is still formed in the cbr1Δmcr1Δ strain (Fig. 2b, and Supplementary Fig. 6), indicating some residual reduction of Dph3 by other reductase, possibly by Ncp1. We hypothesized that this residual reduction of Dph3 is sufficient to support the diphthamide modification, which needs much fewer electrons compared to the more abundant tRNA wobble uridine modifications. To test this, we checked if diphthamide formation in the cbr1Δmcr1Δ cells would be affected when eEF2 was over-expressed. While the BY4741 WT cells with over-expressed eEF2 do not confer any resistance to DT, cbr1Δ cells with over-expressed eEF2 are partially resistant to DT, and CBR1 and MCR1 double deletion render the cells almost full resistance to DT (Fig. 2c). Furthermore, purified eEF2 from cbr1Δ strain over-expressing eEF2 showed a significant decrease in diphthamide modification level (Fig. 2d, and Supplementary Fig. 9). These results suggested that Cbr1 is required for efficient reduction of Dph3 to support the increased electron demand in diphthamide biosynthesis when eEF2 is over-expressed.</p><p>In summary, we identified Cbr1 as the major physiological reductase of Dph3, the electron carrier for the radical SAM enzymes required for two distinct modifications that are important for translation fidelity (Fig. 2e). Because Cbr1 is a NADH dependent reductase, it is plausible that the overall amount of reduced Dph3 is regulated by the redox state of the cell via the NAD+/NADH ratio. Moreover, the abundant tRNA wobble uridine modifications are dependent on the availability of reduced Dph3. Thus, the Cbr1 and NADH-dependent reduction of Dph3 may provide a regulatory linkage between the metabolic state of the cells and protein translation. In bacteria, it is known that flavodoxin serves as the reductase for radical SAM enzymes.24 However, in eukaryotes, the identity of the physiological reduction system for radical SAM enzymes is largely unknown. The Cbr1/Dph3 system is the first physiological reduction system identified for radical SAM enzymes in eukaryotes. This finding may be useful for the identification of other eukaryotic reduction systems for radical SAM enzymes.</p><!><p>All strains used in this study are listed in Supplementary Table 1. The HL1352Y strain expressing endogenous FLAG-tagged Dph3 was generated using PCR-based tagging method as previously described.25 Briefly, the PCR fragment amplified from the plasmid pFA6a-6xGLY-3xFLAG-HIS3MX6 (Addgene plasmid 20753) with primers ZL210 and ZL211 (Supplementary Table 2) was transformed into BY4741 strain and plated on synthetic complete agar plates with histidine dropout for selection. The DPH2, DPH3, CBR1 and MCR1 single deletion strains were obtained from Open Biosystems (GE Dharmacon). The cbr1Δmcr1Δ strain was generated from cbr1Δ strain by Longtine PCR-based method as previously described.26 The PCR fragment amplified from the plasmid pFA6a-NATMX6 with primers ZL244 and ZL245 was transformed into the cbr1Δ strain. Transformed cells were grown in YPD media for three hours to recover and plated on YPD plates with 100 μg/mL nourseothricin for selection. The cbr1Δmcr1Δaim33Δ strain was generated by transforming cbr1Δmcr1Δ strain with PCR fragment amplified from pFA6a-His3MX6 (Addgene plasmid 41596) using primers ZL242 and Zl243 and selecting with synthetic complete histidine drop out plates. The HL1352Y strain with endogenous FLAG-tagged Dph3 was verified by anti-Flag western blot. Single deletion strains obtained from Open Biosystems were verified by PCR method using strain associated barcode primers. Deletion of MCR1 or AIM33 by Longtine fragments were confirmed by PCR method using 5' UTR and 3' UTR primers (Supplementary Table 2).</p><!><p>Saccharomyces cerevisiae BY4741 strain was cultured in synthetic complete media (200 mL) with light lysine and arginine with an initial A600 of 0.02 untill the A600 reached approximately 0.5. The HL1352Y strain expressing Flag-tagged Dph3 was first cultured in synthetic complete media with heavy lysine (Sigma 608041) and arginine (Sigma 608033) for about eight generation. The overnight culture was then used to inoculate 200 mL heavy synthetic complete media with an initial A600 of 0.02 untill the A600 reached approximately 0.5. Cells were harvested by centrifugation at 4,000 g and lysed with 600 μL of glass beads (OPS Diagnostics BAWG400-200-04) and 1 mL lysis buffer containing 50 mM Tris pH 8.0, 0.2% NP-40, 150 mM sodium chloride, 5 mM EDTA and 1 mM phenylmethanesulfonyl fluoride. Cells were lysed by 5 intervals of vortexing for 2 minutes with 2 minutes cooling on ice between intervals. Total cell lysates were cleared by centrifuging for 10 min at 13,000 g and 4 °C. The supernatants containing 2.5 mg of proteins were incubated with 25 μL anti-flag resins (Sigma A2220) for 4 hours at 4°C. The resins were washed with 1 mL of lysis buffer five times and eluted with 90 μL of elution buffer (50 mM Tris pH 8.0 and 1% SDS) and heated at 95 °C for 5 minutes. The eluted fractions were reduced by DTT (10 mM) for 30 minutes at room temperature and then alkylated by iodoacetamide (40 mM) for 30 minutes at room temperature. The heavy (HL1352Y) and light (BY4741) eluates from equal amounts of beads loaded with equivalent amounts of total lysate (2.5 mg) from the two cultures were mixed. Proteins were precipitated by adding 600 μL of precipitation buffer containing 50% Acetone, 49.9% ethanol and 0.1% acetic acid and cooling on ice for 30 minutes. The protein pellet was washed with 400 μL of ice-cold precipitation buffer and air-dried. The resultant pellet was resuspended in 50 μL of resolubilization buffer containing 8 M urea and 50 mM Tris pH 8.0, and then diluted with 400 μL 50 mM Tris pH 8.0 and digested by 1 μg trypsin overnight at 37 °C.</p><!><p>The SILAC tryptic digests were reconstituted in 50 μL of 0.5% formic acid (FA) estimated at 0.1 μg/μL for nanoLC-ESI-MS/MS analysis, which was carried out on an Orbitrap Elite mass spectrometer (Thermo-Fisher Scientific, San Jose, CA) equipped with a "CorConneX" nano ion source device (CorSolutions LLC, Ithaca, NY). The Orbitrap was interfaced with a Dionex UltiMate3000RSLCnano system (Thermo, Sunnyvale, CA). Each SILAC peptide sample (5 μL) was injected under "User Defined Program" onto a PepMap C18 trap column-nano Viper (5 μm, 100 μm × 2 cm, Thermo) at 20 μL/min flow rate and then separated on a PepMap C18 RP nano column (3 μm, 75 μm × 25 cm, Thermo) which was installed in the nano device with a 10-μm spray emitter (NewObjective, Woburn, MA). The peptides were eluted with a 120 minutes gradient of 5% to 38% acetonitrile (ACN) in 0.1% formic acid at a flow rate of 300 nL/min, followed by a 7-min ramping to 95% ACN-0.1% FA and a 8-min hold at 95% ACN-0.1% FA. The column was re-equilibrated with 2% ACN-0.1% FA for 25 minutes prior to the next run. The Orbitrap Elite was operated in positive ion mode with nano spray voltage set at 1.6 kV and source temperature at 250 °C with external calibration for FT mass analyzer being performed. The instrument was operated in parallel data-dependent acquisition (DDA) under FT-IT mode using FT mass analyzer for one MS survey scan from m/z 375 to 1800 with a resolving power of 120,000 (FWHM at m/z 400) followed by MS/MS scans on top 20 most intensive peaks with multiple charged ions above a threshold ion count of 10,000 in FT mass analyzer. Dynamic exclusion parameters were set at repeat count 1 with a 30 s repeat duration, an exclusion list size of 500, 60 s exclusion duration with ±10 ppm exclusion mass width. Collision induced dissociation (CID) parameters were set at the following values: isolation width 2.0 m/z, normalized collision energy at 35 %, activation Q at 0.25, and activation time 10 ms. All data were acquired under Xcalibur 2.2 operation software (Thermo). All MS and MS/MS raw spectra were processed and searched using Sequest HT software within the Proteome Discoverer 1.4.1.14 (PD 1.4, Thermo Scientific). The Saccharomyces cerevisiae RefSeq sequence database (5,847 entries, downloaded on 5/17/2015 from NCBInr) was used for database searches. The database search was performed under a search workflow with the "Precursor Ions Quantifier" node for SILAC 2plex (Arg10, Lys8) quantitation. The default settings for protein identification in Sequest node were: two mis-cleavages for full trypsin with fixed carbamidomethyl modification of cysteine, variable modifications of 10.008 Da on Arginine and 8.014 Da on lysine, N-terminal acetylation, methionine oxidation and deamidation of asparagine and glutamine residues. The peptide mass tolerance and fragment mass tolerance values were 15 ppm and 0.8 Da, respectively. Only high confidence peptides defined by Sequest HT with a 1% FDR by Percolator were considered for the peptide identification. The mass precision for expected standard deviation of the detected mass used to create extracted ion chromatograms was set to 3 ppm. The SILAC 2-plex quantification method within PD 1.4 was used to calculate the heavy/light ratios of all identified proteins. Only unique peptides were used for quantification. The final protein list was filtered by proteins with at least two peptides identified. 500 was set as the maximum H/L ratio to make it mathematically meaningful for peptides essentially not present in the light sample. Table in Supplementary Figure 2b lists protein with H/L ratio greater than 10.</p><!><p>Genomic DNA was extracted from Saccharomyces cerevisiae BY4741 strain using Pierce Yeast DNA Extraction Kit. The DNA sequence encoding for Cbr1 with an N-terminal truncation of 27 amino acids was amplified by PCR with primers ZL224 and ZL222 (Supplementary Table 2) from the genomic DNA. The amplified gene fragment was inserted into the pET28a vector and transformed into the Escherichia coli expression strain BL21 pRARE2. A single colony was used to inoculate an overnight starter culture, which was used to inoculate 2 liters of lysogeny broth (LB) containing 50 μg/mL kanamycin and 20 μg/mL chloramphenicol. Cells were grown at 37°C to A600 of approximately 0.6 and cooled to 16°C. Expression was induced with 0.5 mM isopropyl β-D-1-thiogalactopyranoside (IPTG) and 50 μM riboflavin, and grown overnight at 16°C. Cells were harvested by centrifugation and lysed using the EmulsiFlex-C3 cell disruptor (Avestin, Inc., Canada). The protein was purified on BioLogic DuoFlow 10 System (Bio-Rad, Hercules, CA). The purification was performed on a HisTrap HP column (GE Healthcare, Piscataway, NJ) with a linear gradient from 30 mM imidazole to 500 mM imidazole in 30 minutes. The yellow protein fractions were collected and dialyzed against 25 mM Tris-HCl buffer (pH 8.0) containing 150 mM NaCl. Protein concentration was determined by standard Bradford assay.</p><p>The DNA sequence encoding for Pga3 with an N-terminal truncation of 61 amino acids was amplified by PCR with primers ZL234 and ZL235 (Supplementary Table 2) from the genomic DNA. The DNA sequence encoding Ncp1 with an N-terminal truncation of 33 amino acids was amplified by PCR with primers ZL316 and ZL318 from the genomic DNA. Yeast complementary DNA was used as template for amplification of the mcr1 gene. Yeast RNA was purified using the TRIzol® reagent (Thermo Fisher Scientific). DNA was removed from the purified RNA using DNase I (New England Biolabs). The RNA was extracted with phenol/chloroform to inactivate DNase I. Total cDNA was synthesized from the purified RNA using SuperScript® III Reverse Transcriptase (Thermo Fisher Scientific). The DNA sequence encoding for Mcr1 with an N-terminal truncation of 31 amino acids was amplified by PCR with primers ZL240 and ZL241 (Supplementary Table 2) from the synthesized cDNA. Subsequent cloning and expression of recombinant Pga3, Ncp1, and Mcr1 were similar to that of Cbr1.</p><!><p>Recombinant yeast Dph3 was purified as previously described.7 The reaction was monitored on a Cary 50 Bio UV-Vis spectrophotometer (Varian) at 488 nm. Cbr1 (0.5 μM) was mixed with 100 μM of Dph3 in a cuvette. The reaction was initiated by addition of NADH or NADPH at a final concentration of 0.2 mM or as indicated. Reduction of Dph3 by Mcr1 or Pga3 was monitored under the same conditions as that of Cbr1. Figure 1b and Supplementary Figure 5a-5c show a representative image from three experimental repeats.</p><!><p>The Dph3 protein, Dph1-Dph2 complex and eEF2 proteins from BY4741 dph2 deletion strain were expressed and purified as previously described.7 The reaction mixture was assembled in the anaerobic chamber under strictly anaerobic conditions. Aerobically purified Dph3, Cbr1 and eEF2 were degassed by a schlenk line. The reconstitution reactions were set up in an anaerobic chamber. Dph1-Dph2 (5 μM), Dph3 (10 μM), eEF2 (2 μM), Cbr1 or Mcr1 or Ncp1 (5 μM), and NADH or NADPH (200 μM) were added to a buffer containing 150 mM NaCl and 200 mM Tris-HCl at pH 7.4. Reactions without NADH or without reductases were also carried out as negative controls. The reaction vials were sealed before being taken out of the anaerobic chamber. 14C-SAM (ARC 0343-50, 18 μM) was injected into each reaction vial to initiate the reaction. The reaction mixtures were mixed by brief vortexing and incubated at 30 °C for 60 minutes. The reactions were stopped by adding protein loading dye and subsequently heating at 95 °C for 5 minutes, and then resolved by 12% SDS–polyacrylamide gel electrophoresis. The dried gel was exposed to a Phosphor Imaging screen and scanned using a Typhoon FLA 7000 (GE Healthcare Life Sciences). Figure 1c and Supplementary Figure 7 show a representative image from three experimental repeats. Supplementary Figure 5d and Supplementary Figure 9 show a representative image from two experimental repeats.</p><!><p>Cells transformed with p423Met25-eEF2 (allowing over-expression of eEF2 with a 8 His C-terminal tag) were cultured in synthetic complete media with histidine dropout at 30 °C with an initial A600 of 0.02 until the A600 reached approximately 1.0. Cells harvested were lysed and eEF2 was purified as previously described.5 Purified eEF2 (1 μM) and Rh-NAD (25 μM) were incubated with DT (1μM) at 30 °C in 50 mM NaCl, 30 mM dithiothreitol (DTT), 2 mM ethylenediaminetetraacetic acid (EDTA), and 25 mM Tris-HCl at pH 8.0 for 15 minutes as previously described.4 Figure 2d shows a representative image from two experimental repeats.</p><!><p>For DT sensitivity assays, cells were transformed with plasmid pHL1015, which allows for galactose-inducible, glucose-repressible expression of diphtheria toxin as previously described.4 Transformed cells were cultured in synthetic complete media with uracil dropout at 30 °C overnight, adjusted to A600 of 0.2 with autoclaved water, and then diluted serially in 4-fold increments. Aliquots of each dilution were spotted on glucose-containing or galactose-containing agar plates using a replica plater. Plates were incubated at 30 °C. Cell growth was recorded 2-3 days after plating. DT sensitivity assay for cells co-transformed with p423Met25-eEF2 and pHL1015 were performed similarly and plated onto galactose-containing or glucose-containing synthetic complete with histidine and uracil dropout agar plates. For γ-toxin sensitivity assays, cells were transformed with plasmid pLF16 which allows for galactose-inducible, glucose-repressible expression of γ-toxin as previously described.23 Cultures were grown in synthetic complete media with leucine dropout at 30°C. Plating of cells on agar plates was performed similar to that of DT assays. Figure 2a, 2c and Supplementary Figure 4 are representative of three biological triplicates.</p><!><p>The DNA sequence encoding for γ-toxin with an N-terminal truncation of 19 amino acids was amplified by PCR with primers ZL436 and ZL437 (Supplementary Table 2) from pLF16. The amplified gene fragment was inserted into the pET28a vector and transformed into the Escherichia coli expression strain BL21 pRARE2. A single colony was used to inoculate an overnight starter culture, which was used to inoculate 2 liters of LB containing 50 μg/mL kanamycin and 20 μg/mL chloramphenicol. Cells were grown at 37°C to A600 of approximately 0.6 and cooled to 16°C. Expression was induced with 0.5 mM IPTG and grown overnight at 16°C. Subsequent protein purification was performed similar to that of Cbr1.</p><!><p>Yeast cells were cultured in 2 liters YPD from initial A600 of 0.02 till A600 reached approximately 0.5. Cells were harvested and bulk tRNA was purified as previously described.27 The total tRNAs (5μg) were incubated withγ-toxin (5 μM) in 10 mM NaCl, 10 mM MgCl2, 1 mM DTT and 10 mM Tris-HCl at pH 7.4 for 2 hours at 30°C. The time course of γ-toxin treatment in Figure S5 was performed under the same reaction conditions with specified incubation time. The samples were separated on 12% polyacrylamide, 8M urea gels, and transferred to nylon membrane (GE Health, rpn119b) for northern blot. The oligonucleotide used to detect 5' of glu-tRNA was ordered from IDT (/5AmMC6/GTGATAGCCGTTACACTATATCGGA) and conjugated to Alexa Fluor® 680 (ThermoFisher, A37567). Northern blots were visualized by Odyssey® CLx imaging system (LI-COR). Figure 2b shows a representative image from two experimental repeats.</p>

PubMed Author Manuscript

Reversible dispersion and release of carbon nanotubes <i>via</i> cooperative clamping interactions with hydrogen-bonded nanorings

Due to their outstanding electronic and mechanical properties, single-walled carbon nanotubes (SWCNTs) are promising nanomaterials for the future generation of optoelectronic devices and composites. However, their scarce solubility limits their application in many technologies that demand solution-processing of high-purity SWCNT samples. Although some non-covalent functionalization approaches have demonstrated their utility in extracting SWCNTs into different media, many of them produce short-lived dispersions or ultimately suffer from contamination by the dispersing agent. Here, we introduce an unprecedented strategy that relies on a cooperative clamping process. When mixing (6,5)SWCNTs with a dinucleoside monomer that is able to self-assemble in nanorings via Watson-Crick base-pairing, a synergistic relationship is established. On one hand, the H-bonded rings are able to associate intimately with SWCNTs by embracing the tube sidewalls, which allows for an efficient SWCNT debundling and for the production of long-lasting SWCNT dispersions of high optical quality along a broad concentration range. On the other, nanoring stability is enhanced in the presence of SWCNTs, which are suitable guests for the ring cavity and contribute to the establishment of multiple cooperative noncovalent interactions. The inhibition of these reversible interactions, by just adding, for instance, a competing solvent for hydrogen-bonding, proved to be a simple and effective method to recover the pristine nanomaterial with no trace of the dispersing agent.

reversible_dispersion_and_release_of_carbon_nanotubes_<i>via</i>_cooperative_clamping_interactions_w

Introduction<!>Initial experiments and sample preparation<!>Spectroscopic and thermogravimetric measurements<!>AFM, TEM and DFT analysis<!>Analysis of the synergistic interactions<!>Recovery of the pristine SWCNTs<!>Conclusions

<p>Carbon nanotubes (CNTs) are nanomaterials with impressive electronic and mechanical properties and are considered as strong candidates for the next generation of transistors, photovoltaics, and (bio)chemical sensors. [1][2][3] However, they are insoluble in most solvents and infusible at any temperature, due to strong bundling caused by numerous van der Waals interactions, which hampers many of their applications. 4,5 Covalent and noncovalent surface modication with molecules, leading to homogeneous CNT dispersions in various solvents or bulk materials, is now common practice to facilitate their processing and to attach specic functions. [6][7][8][9] The covalent approach grants more stable dispersions, but the modication of the p-conjugated CNT surface with graed molecules has a detrimental impact on their (semi)conducting and mechanical properties. That is the reason why noncovalent strategies, achieved by promoting strong interactions between CNT sidewalls and p-conjugated or apolar moieties in adequate solvents, are oen considered more suitable. [9][10][11] However, the weak and dynamic nature of supramolecular interactions frequently produces less durable dispersions of individual CNTs because bundling, which ultimately leads to CNT reprecipitation, remains as a strongly competing supramolecular process in solution. Bundling can be reduced by increasing the number of surface-binding functions in the dispersing agent and, importantly, by inducing wrapping interactions around single CNT cylinders (interaction mode II in Fig. 1a below). A large number of composites with p-conjugated polymers, 12,13 DNA, [14][15][16] synthetic peptides, [17][18][19][20] or, in general, oligomers featuring multiple aromatic units [21][22][23][24] have been tested to stabilize CNTs selectively in solution by specic helical wrapping conformations around the tube.</p><p>A main drawback that originates from such a robust marriage is that polymer desorption via washing processes is oen difficult to achieve, and CNTs puried or processed in this way may remain contaminated with the dispersing agent. 25,26 In order to address this issue, recent strategies have been reported that focus mainly on: (1) switching between tightly and loosely bound polymer conformations, [27][28][29] which can be triggered thermally, photochemically 27 or by a change in solvent 28 or pH, 29 and on (2) inducing depolymerization processes, [30][31][32][33][34] which can also be made reversible by introducing supramolecular [32][33][34] or dynamic covalent bonds 35 along the polymer main chain.</p><p>Here, we introduce a novel approach that is instead based on clamping discrete self-assembled nanorings around the tube cross-section to efficiently and reversibly produce durable dispersions of individual single-walled CNTs (SWCNTs) in organic solvents. The encapsulation of SWCNTs within well-dened macrocycles, 19,[36][37][38][39][40][41] forming rotaxane-type ensembles, has been recently explored using covalent cyclization reactions. The strategy followed by some of us relied on two extended aromatic binding sites to promote the supramolecular association of U-shaped molecules to SWCNT, followed by "clipping" through ring-closing metathesis to produce the mechanically interlocked species. [36][37][38][39][40][41] Depending on the chemical nature of the recognition motif, we have also observed the formation of oligomers that wrap around the SWCNTs. 38 The approach followed herein, in contrast, prots from a dynamic, strongly cooperative noncovalent macrocyclization process.</p><p>Our design focuses on a dinucleoside monomer (GC1; Fig. 1b) that has been rationally synthesized to comply with these objectives (see the ESI † for synthetic details). GC1 structure consists of complementary guanosine (G) and cytidine (C) DNA bases connected by a rigid, linear central block. We demonstrated recently that related molecules self-assembled into cyclic tetramer species 42 through Watson-Crick G-C Hbonding interactions 43 exhibiting record chelate cooperativities, [44][45][46] which allow the nanorings to be formed quantitatively in nonpolar solvents within a wide concentration range. [47][48][49][50] The ribose groups in GC1 feature multiple long alkyl chains that, upon cyclic assembly, would point in all directions towards the exterior medium, which should benet individual CNT debundling and thus afford high solubility to the nal composites. The monomer-CNT interaction strength is an additional key factor in our design. A p-conjugated dialkoxyarene central block with modest affinity for the CNT surface has been installed between the lipophilic bases. The idea is that stable dispersions would only be obtained if the rings are able to embrace the tube, so that its sidewalls can interact with several monomers in the nanoring cavity (mode IV in Fig. 1), and not through external binding with sections of the rings (mode III) or with individual monomers or linear oligomers (modes I/II). We also explore herein the special synergy of the GC1-SWCNT marriage, which simultaneously brings an enhanced solubility to the nanotubes (when compared to a related molecule that cannot cyclize: CC1) and an increased stability to the macrocycles. This synergy originates from the cooperative action of G-C H-bonding and monomer-nanotube van der Waals interactions, so we reasoned that the disturbance of any of these distinct noncovalent interactions would cause collapse of the supramolecular ensemble, thus facilitating the recovery of the extracted CNTs.</p><!><p>Prior to their combination with CNTs, we conrmed that this novel GC1 monomer displayed a similar self-assembly process to the one already reported by us with closely related dinucleosides (see the ESI † for further details). 47,50 NMR and optical spectroscopy experiments indicated that GC1 tetrameric rings are indeed formed close to quantitatively in apolar chlorinated solvents within the 10 À1 to 5 Â 10 À4 M concentration range. These cyclic assemblies are characterized by red-shied and low intensity emission maxima at ca. 505 nm, and by the presence of a characteristic negative Cotton effect, with maxima at 340 and 387 and a minimum at 428 nm. At lower concentrations, they dissociate gradually into monomeric species, which display distinct emission maxima at 421 and 445 nm and null CD signals (vide infra). The addition of polar cosolvents that can compete for H-bonding, like DMSO or DMF, also results in monomer dissociation. The 1 H NMR spectra of GC1 recorded by modifying the CDCl 3 : DMF-D 7 volume ratio (Fig. S1 †) revealed a strong all-or-nothing behavior: no signicant participation of any other H-bonded oligomer but the tetrameric macrocycle is detected in solution. This is in agreement with the formation of stable ring species with remarkably high chelate cooperativities, as determined in our previous work. An initial theoretical study using DFT calculations (see below and the ESI † for further details) served to properly design the interacting ring-tube system and experimental measurements. We decided to use (6,5)-enriched SWCNTs with a mean diameter of 0.7 nm and very narrow polydispersity (we observe (6,5) SWCNTs exclusively in photoluminescence experiments; see Fig. S2 †) because they should t adequately within the GC1 nanoring cavity. Thus, a dispersion of (6,5)SWCNTs in CHCl 3 (0.2 mg mL À1 ) was produced rst by ultrasonication followed by centrifugation, in order to remove large SWCNT bundles and any other carbonaceous impurities. These non-stabilized dispersions are rather short-lived and signicant precipitation of (6,5)SWCNTs was clearly observed aer a few hours (see central image in Fig. 2a). Immediately aer centrifugation, GC1 (or CC1) solutions in CHCl 3 (different concentrations were tested, from 5.0 Â 10 À4 to 10 À6 M) were added to the supernatant (6,5)SWCNT dispersion and the mixtures were stirred at room temperature. The (6,5)SWCNT-GC1 suspensions produced (right image in Fig. 2a) were kept under dark and checked at different periods of time. A few spectroscopic changes were observed within the rst hours aer mixing, but aer ca. 12 hours spectral properties remained constant for weeks and are reproducible.</p><!><p>The (6,5)SWCNT-GC1/CC1 composites were analyzed by absorption, CD, Raman, and emission spectroscopies and by TGA, and compared with pristine (6,5)SWCNTs and GC1/CC1 in the same conditions (Fig. 2). The absorption spectra (Fig. 2b) showed the typical features of both components in the mixture and only minor deviations in absorption maxima were noted. The typical scattering phenomena, evidenced by a baseline rise, was notorious in the original (6,5)SWCNTs dispersions and in mixtures with CC1 or with low GC1 content. However, as the GC1/(6,5)SWCNTs ratio is increased, scattering is concomitantly reduced until it is no longer perceptible in the absorption spectrum, which underlines the high dispersing power of this agent. As a matter of fact, the GC1-(6,5)SWCNTs mixtures produced are clear suspensions (Fig. 2a) that show no evidence for nanotube precipitation along several weeks.</p><p>CD can be considered as one of the most powerful techniques for stereochemical analysis: it is sensitive to the absolute conguration as well as to conformational features, which are oen completely obscured in ordinary absorption spectra. 51 Regarding this, the CD spectral shape of GC1-(6,5)SWCNTs composites does not exhibit important differences with GC1 in the same conditions (Fig. 2c). This is a strong indication that the cyclic tetramer structure is preserved in the presence of (6,5) SWCNT, and no dissociation or structural change in the Hbonded assembly was noted. In order to obtain additional proof from a theoretical perspective, we calculated the CD spectrum for the optimized structure of cGC1 4 (see Fig. 3g below), which is displayed as a dotted red line in Fig. 2c. Our calculations revealed a strong dependency of the CD spectra with the symmetry of the macrocycle. Only when the C 4 axis was maintained in the cGC1 4 ring structure, we obtain a good match between the experimental and the theoretical results (with the exception of the vibronic structure of the positive Cotton effect). If the C 4 symmetry axis is lost, due for instance to other binding modes of GC1 to the CNT sidewalls, the CD spectrum is signicantly perturbed. As an example, we show in Fig. S3 † the structure and calculated spectrum for a cGC1 4 structure, which is close in energy, but has lost the C 4 symmetry. We can therefore conclude that this C 4 symmetry is conserved upon the formation of (6,5)SWCNT-cGC1 4 conjugates, which is a strong indication that the SWCNT is inside the cavity of cGC1 4 ring.</p><p>Another important proof that demonstrates GC1-(6,5) SWCNTs interactions came from comparing the emission spectra (Fig. 2d) with and without CNTs. As a matter of fact GC1 uorescence is considerably quenched in the presence of (6,5) SWCNTs. This is very characteristic in noncovalent SWCNT assemblies 22 and is presumably caused by an energy transfer process between the dinucleoside p-conjugated system and the SWCNT when they are in close contact.</p><p>On the other hand, TGA studies showed that macrocycle loading increases with GC1 concentration, to yield 17% and 28% loading at 10 À6 and 10 À4 M GC1 initial concentration, respectively. Finally, no signicant (6,5)SWCNT electronic perturbance was noted in Raman experiments when GC1 or CC1 were added, which is a good indication of the formation of noncovalently bound composites.</p><!><p>Exploration of the GC1-SWCNT suspensions under atomic force microscopy (AFM) was consistent with a picture in which the SWCNTs are encapsulated within GC1 macrocycles. AFM images were obtained upon spin-coating the SWCNT suspensions on mica, and show individualized SWCNTs decorated with objects of about 2 nm in height (Fig. 3a and b), in full accord with the DFT-modelled size of the GC1 tetramer (Fig. 3h). We also explored our samples under transmission electron microscopy (TEM), where we observe mostly rebundled SWCNTs, most likely due to sample preparation issues. However, wherever individualized SWCNTs were located, they showed heavily functionalized sidewalls (Fig. 3c and d). Unfortunately, and in contrast to covalently linked macrocycles, 41 attempts at obtaining higher resolution TEM images were precluded by the instability of the H-bonded organic macrocycle in the conventional transmission electron microscope at 200 kV. High resolution images were obtained in an aberration corrected microscope at 60 kV (Fig. 3e and f). We observe structures of around 2.0 nm surrounding the SWCNTs that are consistent with the picture provided by spectroscopy, but even under low voltage (see also Fig. S5 †), [52][53][54] the macrocycles were observed to quickly reorganize and eventually decompose under the electron beam irradiation, probably due to radiolysis damage associated to hydrogen.</p><p>Further theoretical studies were performed in order to gain a deeper insight into the structure and interactions in the (6,5) SWCNT-cGC1 4 composites. As explained above, the optimized geometry of cGC1 4 belongs to the C 4 point group of symmetry and is depicted in Fig. 3g. The inner cavity has a mean diameter of 2.1 nm and the central aromatic moieties are bent at an angle of 30 , so that planarity is not conserved, allowing the carbon nanotube to maintain a stronger interaction with the ring. In order to have a second proof of the arrangement between the two moieties, we additionally calculated the intensity of the interaction between SWCNTs and cGC1 4 ring with close p-p contacts in the range of 3.2-3.5 Å. Among the different possible congurations (see Fig. S4 †), we observed that the (6,5)SWCNT and the cGC1 4 tetramer interact strongly (E int ¼ À22.7 kcal mol À1 ) when the macrocycles are clamping the nanotube and the macrocycle tilts slightly with respect to the tube axis, in order to maximize non-covalent interactions (interaction mode IV in Fig. 1). In contrast, all external binding modes investigated (interaction mode III in Fig. 1), result in positive (i.e. disfavorable) interaction energies ranging between +0.7 and +42.0 kcal mol À1 . The geometry of the lowest energy conguration is shown in Fig. 3h.</p><!><p>Concentration-dependent experiments, where we recorded absorption, CD and emission spectra, provided a deeper understanding of the mutual interaction between CNTs and Hbonded nanorings. Two parallel dilution measurements were conducted from GC1 CHCl 3 solutions: one in the absence (Fig. 4a and c) and the other one in the presence of (6,5)SWCNTs (Fig. 4b and d). The spectral changes with concentration were monitored by CD from 4.0 Â 10 À4 M (blue lines) down to 3.0 Â 10 À6 M (CD) or down to 2.0 Â 10 À7 M (emission) (red lines). The trends obtained from both techniques are compared in Fig. 4i.</p><p>A rst experimental observation that came from comparison of both dilution experiments is that the presence of (6,5) SWCNTs preserved the characteristic cyclic tetramer features along a wider concentration range. This is unambiguously observed in the evolution of the CD spectra as a function of concentration. While GC1 CD features readily disappear below ca. 10 À5 M (Fig. 4a), indicating cyclic tetramer dissociation, [47][48][49][50] they remain distinct down to 3 Â 10 À6 M in the presence of (6,5) SWCNTs (Fig. 4b). The degree of cyclotetramerization, that is, the molar fraction of GC1 molecules associated as cyclic tetramers (horizontal equilibria in Fig. 1a), can be calculated from each dilution experiment by integrating CD intensity. The comparison of both trends, represented with solid blue and green circles in Fig. 4i, suggests that GC1 nanorings are signicantly stabilized when mixed with (6,5)SWCNTs. Fitting these data to cyclotetramerization processes (blue and green lines in Fig. 4i) afforded cyclotetramerization constants (K T ) in the order of K T ¼ 3.2 Â 10 14 M À3 for GC1 and K T ¼ 4.0 Â 10 16 M À3 for GC1 + (6,5)SWCNTs. That is, chelate cooperativity greatly benets from the presence of SWCNTs and nanoring stability is considerably increased.</p><p>We then turned our attention to the evolution of the uorescence features in these dilutions experiments. The emission spectra were analyzed in two different ways, attending to: (1) their relative intensity, which can be correlated to the fraction of CNT-bound molecules (vertical equilibrium in Fig. 1a), or (2) the spectral shape and emission wavelength, which reports again on the degree of cyclotetramerization (horizontal equilibria in Fig. 1a). In the rst case we compared GC1 emission intensity in the 400-650 nm range in the absence (I GC ) or presence (I GC-CNT ) of (6,5)SWCNTs at each concentration (see Fig. S6A †). The degree of emission quenching, dened as 1 À I GC-CNT /I GC , was then calculated and plotted as red squares in Fig. 4i. Since energy transfer from the photoexcited dinucleoside molecules to the CNT is close to quantitative when they are closely interacting, as determined from the fully quenched emission of the samples at high concentration, these curves actually reveal the molar fraction of GC1 that is bound to (6,5)SWCNTs at each concentration. Our results show that GC1-(6,5)SWCNTs association is virtually quantitative at concentrations above 10 À4 M and then falls in the 10 À4 to 10 À6 M range. This nonlinear trend suggests a cooperative GC1-(6,5)SWCNT interaction that seems to be coupled to the cyclotetramerization equilibrium. In order to compare our GC monomer with a related dispersing agent that does not associate in cyclic systems, we carried out another set of parallel dilution experiments with CC1 in the absence and presence of (6,5)SWCNTs. The degree of emission quenching was likewise calculated at each concentration and the results are shown as orange squares in Fig. 4i (see also Fig. S6B †). The CC1-(6,5)SWCNTs association is clearly weaker and no longer detected below 10 À5 M, a concentration where ca. half of the GC1 molecules are still bound to CNTs.</p><p>In the second case, the degree of cyclotetramerization was estimated by analysing the shape of each normalized uorescence spectra during the dilution measurements shown in Fig. 4c and d, as we described in our previous work (see also the ESI †). 47,50 As commented above, when the monomeric species, showing emission maxima at 421 and 445 nm in CHCl 3 (see red spectra in Fig. 4c and d), associate in cyclic tetramers, a red-shi to >500 nm and a reduction in emission intensity is noted. The results obtained at each concentration are also included in Fig. 4i as solid green and blue triangles for the GC1 samples without and with CNTs, respectively. The cyclic tetramer association trends calculated by CD and uorescence spectroscopy (green circles and triangles) display a quite decent correlation for GC1. That is not the case when (6,5)SWCNTs are present (please compare blue circles and triangles), and this is because CD and emission spectroscopy are not reporting on the same GC1 population. Whereas the measured CD spectra is representative of all GC1 molecules in solution, the emission spectra primarily provide information of the fraction of molecules that are not bound to the CNTs, since the emission of the GC1 molecules that are bound is strongly quenched. As a consequence, the shape evolution of the emission spectra when the CNTs are present exhibits again a strong coupling between the horizontal cyclotetramerization and vertical GC1-(6,5)SWCNT association equilibria in Fig. 1a. At concentrations above 10 À4 M most GC1 molecules are bound to the CNTs and the residual emission recorded is extremely weak in intensity and representative of the cyclic tetramer in shape and emission wavelength (see blue spectrum in Fig. 4d as an example). However, when GC1-(6,5)SWCNT association is no longer quantitative in the 10 À4 to 10 À6 M range, the emission spectra becomes rapidly dominated by the fraction of GC1 molecules that are not bound to CNTs, which have stronger monomer-like features (red spectrum in Fig. 4d) because their actual concentration is lower. This is reected in a very sharp transition around 10 À4 M in the blue-triangle trend in Fig. 4i. In other words, in the 10 À4 to 10 À6 M range, the shape of the emission spectrum is more shied to the monomer features in the presence of CNTs, since the actual concentration of emissive GC1 molecules (i.e. not bound to CNTs) is lower than in the absence of CNTs. Only when the GC1-(6,5)SWCNT interactions become no longer important, below 10 À6 M, samples with and without CNTs display similar emission intensity and shape (see also Fig. S6A †).</p><p>We believe the graph in Fig. 4i provides a rather faithful description of the self-assembly of this two-component mixture in solution as a function of concentration. The results gathered from these dilution experiments in CHCl 3 clearly indicate that each species benets synergistically from the presence of the other, as it is schematically represented in Fig. 5. On one hand, the nanorings are more stable in the presence of nanotubes, a gain that is represented by the blue area in Fig. 4i. On the other, the nanotubes can host a higher number of dispersing agent molecules and thus enjoy enhanced solubility along a broader concentration range when the monomer cyclizes, a gain that is represented by the red area in Fig. 4i when comparing GC1 and CC1. In short, H-bonding between complementary bases and van der Waals dispersion forces between monomer and the p-conjugated CNT sidewalls are noncovalent interactions that work here cooperatively to build strongly-associated GC1-(6,5)SWCNTs composites. It is interesting to note that smaller nanorings, where the nucleobases are directly connected by a triple bond (see GC3 in the ESI †), whose cavities cannot host (6,5)SWCNTs, do not exhibit the same synergistic effects as GC1, but rather behave as a regular dispersing agent, like CC1. On the other hand, GC1 cannot solubilize as efficiently multi-walled nanotubes of larger diameter. Both observations are in agreement with the clamping interaction mode IV, shown in Fig. 1a and 5, being the most likely and abundant in CHCl 3 solutions (Fig. 1a). We should nonetheless consider that, when the nanorings are assembled around the tube, local concentration effects may occur, so that a rearrangement into linear polymers may take place when several rings coincide locally. However, these ring-to-chain rearrangements should lead to a reduction in CD intensity at high concentrations that we did not observe, so we can discard this is a relevant situation. As we reduce concentration, these effects are even less likely to occur, because the rings are favoured entropically.</p><!><p>At this point we reasoned that the whole assembly could be demolished by addressing just one of the supramolecular interactions, which could be a simple and straightforward strategy for pristine (6,5)SWCNTs recovery. For instance, an increase in solvent polarity should disrupt H-bonding interactions without strongly affecting monomer-(6,5)SWCNTs van der Waals interactions. In order to prove this, we performed now the same dilution experiments from GC1 samples with and without CNTs by gradual addition of DMF instead of CHCl 3 , and recorded again nanoring dissociation by CD (Fig. 4e and f; open blue and green circles, respectively, in Fig. 4i), and the degree of emission quenching by uorescence spectroscopy (Fig. 4g and h; open red squares in Fig. 4i).</p><p>In line with the previous observations, the cycles are more stable and the CD features resist a higher amount of DMF when (6,5)SWCNTs are present (please compare open blue and green circles). Together with the disappearance of the CD signals, GC1 monomer emission was recovered in a narrower concentration range when the parent GC1-(6,5)SWCNTs dispersion was diluted with DMF instead of CHCl 3 (please compare solid and open red squares), which suggests much weaker interactions with the CNTs in this polar solvent. Interestingly, several minutes aer DMF addition a clear precipitate emerged from the original dispersion that, aer washing, showed no residual GC1 spectroscopic features. Despite the strong association and remarkable endurance of the GC1-(6,5)SWCNTs solutions in CHCl 3 , which can last for several weeks maintaining the original optical quality, CNT rebundling and precipitation occurred rapidly in DMF, likely due to the lower efficiency of the dissociated monomer as a dispersing agent (see Fig. 5). It is also interesting to note that cGC1 4 dissociation can also be induced by increasing the temperature in CHCl 3 or CHCl 2 CHCl 2 solutions. Aer being subjected to a heating-cooling cycle the spectra of the initial and nal samples differed considerably and clear precipitation of the CNTs was noted. However, the results are highly dependent on the concentration and not as reproducible and efficient as the addition of DMF to release and recover the pristine (6,5)SWCNTs.</p><!><p>In conclusion, we have explored herein an unprecedented approach to solubilize SWCNTs in apolar solvents that relies on dynamic macrocycle clamping around the nanotube sidewalls, which allows for efficient SWCNT debundling, and on cooperative noncovalent interactions, which supplies the required reversibility to simply and effectively recover the pristine material. The combination of theoretical DFT-based (6,5)SWCNTs (bottom). At high concentrations in CHCl 3 , GC1 ring assemblies are formed quantitatively and establish strong interactions with the CNTs. Upon dilution with CHCl 3 the cyclic tetramers gradually dissociate, but such dissociation occurs to a lower extent when CNTs are present, due to the stronger (6,5) SWCNT-cGC1 4 clamping interactions. On the contrary, upon dilution with DMF, the cGC1 4 rings are fully dissociated at relatively high concentrations, which eventually produces CNT precipitation due to the weaker interaction of GC1 monomers with the (6,5)SWCNTs. In this last situation, a simple filtration and washing protocol allows to separate efficiently the (6,5)SWCNTs from the GC1 monomer.</p><p>methodologies, spectroscopic techniques, as well as AFM and TEM microscopies, provide solid evidence for a preferred association mode where the H-bonded nanorings are embracing the tube (mode IV in Fig. 1). Furthermore, a comparison between dilution experiments performed on GC1 in the absence or presence of (6,5)SWCNTs provided a deep insight into the mutual benets offered by the combination of these two species. On one hand, nanoring stability is unambiguously enhanced in the presence of the CNTs, likely due to the interaction of more than one GC1 monomer with the tube sidewalls. On the other, the GC1 molecule exhibited an extraordinary solubilizing power, in comparison with a regular dispersing agent that cannot cyclize, like CC1. When comparing these two monomers, we observe dramatic differences in terms of the quality of the dispersions produced (that only show very low scattering in the case of GC1; see Fig. 2b), their durability (the dispersions produced from CC1 show a CNT precipitate aer a few days, while those from GC1 were seen to resist for many weeks), and the resistance of the monomer-SWCNT interaction to dilution (as determined from the experiments shown in Fig. 4). We believe that these remarkable differences, which are translated in the generation of clear, long-lasting (6,5) SWCNTs dispersions along a broad concentration range, can only be explained if the dispersing agent is able to surround individual nanotubes, as in mode IV in Fig. 1.</p><p>Future efforts will be directed to study if our self-assembled nanorings can selectively extract CNTs as a function of diameter and chirality.</p>

Royal Society of Chemistry (RSC)

Smartphone-Based Mobile Detection Platform for Molecular Diagnostics and Spatiotemporal Disease Mapping

Rapid and quantitative molecular diagnostics in the field, at home, and at remote clinics is essential for evidence-based disease management, control, and prevention. Conventional molecular diagnostics requires extensive sample preparation, relatively sophisticated instruments, and trained personnel, restricting its use to centralized laboratories. To overcome these limitations, we designed a simple, inexpensive, hand-held, smartphone-based mobile detection platform, dubbed \xe2\x80\x9csmart-connected cup\xe2\x80\x9d (SCC), for rapid, connected, and quantitative molecular diagnostics. Our platform combines bioluminescent assay in real-time and loop-mediated isothermal amplification (BART-LAMP) technology with smartphone-based detection, eliminating the need for an excitation source and optical filters that are essential in fluorescent-based detection. The incubation heating for the isothermal amplification is provided, electricity-free, with an exothermic chemical reaction, and incubation temperature is regulated with a phase change material. A custom Android App was developed for bioluminescent signal monitoring and analysis, target quantification, data sharing, and spatiotemporal mapping of disease. SCC\xe2\x80\x99s utility is demonstrated by quantitative detection of Zika virus (ZIKV) in urine and saliva and HIV in blood within 45 min. We demonstrate SCC\xe2\x80\x99s connectivity for disease spatiotemporal mapping with a custom-designed website. Such a smart- and connected-diagnostic system does not require any lab facilities and is suitable for use at home, in the field, in the clinic, and particularly in resource-limited settings in the context of Internet of Medical Things (IoMT).

smartphone-based_mobile_detection_platform_for_molecular_diagnostics_and_spatiotemporal_disease_mapp

<!>RT-LAMP Primer Design<!>Virus Samples<!>Samples Spiked with Viruses<!>Benchtop RT-LAMP Amplification<!>Operation of Multifunctional Isothermal Amplification Reactor (MIAR) Chip<!>Operation of Smart-Connected Cup (SCC)<!>Smart-Connected Cup Platform<!>Android-Based Smartphone App<!>Excitation/Filter-Free BART-LAMP Assay<!>Effects of BART Assay Composition on Light Emission Intensity<!>Quantitative Detection of ZIKV in Urine and Saliva<!>HIV Detection in Blood<!>Smart and Connected Disease Surveillance<!>CONCLUSIONS<!>

<p>Rapid and quantitative molecular diagnostics in the field, at home, and at remote clinics is essential for evidence-based disease management, control, and prevention.1–3 Nucleic acid amplification tests (NAATs) based on enzymatic amplification, such as polymerase chain reaction (PCR), are the "gold standard" for disease diagnostics due to their high sensitivity and specificity and quantification capability.4–6 However, current PCR-based NAATs require extensive sample preparation to extract, isolate, and purify target nucleic acids from heterogeneous samples and instrumentation for precise thermal cycling, neither of which is practical at the point of care (POC). There is a need for simple, rapid, inexpensive, and easy-to-use NAATs that can be used outside the clinical laboratory.</p><p>Recent advances in isothermal nucleic acid amplification assays and microfluidics enable portable detection systems.7–10 However, many of these systems11,12 are bulky and expensive and often lack connectivity. Most currently available systems are appropriate for centralized clinics, not for point of care diagnostics. Since smartphones are ubiquitous, in particular in developing countries that lack land-based telecommunications,13,14 it is sensible to take advantage of smartphones to replace dedicated instruments for inexpensive and reliable signal acquisition and analysis; transmission of test results to the patient's file and doctor's office; communication of test Global Positioning System (GPS) location, time, and deidentified results to public health officials for spatiotemporal mapping, epidemiological surveillance, resource allocation, and policy decisions. The use of smartphones in NAATs not only reduces test cost but also enhances capabilities beyond what is available with existing instruments. Most recently, a few smartphone-based platforms15–18 have been reported for nucleic acid amplification detection. However, all these platforms rely on fluorescence detection with optical components (i.e., fluorescence excitation light source and optical filters) that add cost and complexity. Although they use smartphone cameras to record signals, most of them lack smartphone-based image processing and quantitative detection and do not take advantage of connectivity.</p><p>Here, we describe a simple, inexpensive, hand-held, smartphone-based mobile detection platform (dubbed "smart-connected cup", SCC) for rapid, connected, and quantitative molecular diagnostics (Figures 1a,b and S1a). Our platform combines a bioluminescent assay (BART, bioluminescent assay in real-time)19 with loop-mediated isothermal amplification (LAMP), in which luciferin, fueled by polymerase byproducts, produces bioluminescence light. In contrast to fluorescent reporters,15–18 BART does not require an excitation light source and optical filters and is not susceptible to background emission. Incubation heating for the isothermal amplification is generated by an exothermic chemical reaction ("thermal battery"), and incubation temperature is regulated with a phase-change material, enabling electricity-free nucleic acid amplification.18 Our smartphone App monitors luciferin emission in real-time, quantifies emission intensity, and determines target concentration. Quantitative test results are displayed on the smartphone screen and, when desired, wirelessly transmitted to a remote server to be recorded in the patient's files and made available to the patient's doctor and, together with GPS coordinates, to public health officials. To demonstrate our platform's capabilities, we selected applications of current interest: molecular diagnostics of Zika virus (ZIKV) and HIV infection. We also illustrate our platform's capability of mapping disease spread with a custom-designed website.</p><!><p>The RT-LAMP primer set for ZIKV was designed as we have previously described.20 Briefly, we aligned and analyzed complete genome sequences of various ZIKV strains to identify conserved sequences among ZIKV strains. These sequences were compared to other flaviviruses' sequences with DNAMAN software. A 600-nt sequence in the envelope protein coding region was selected to serve as our template due to its high homology among ZIKV and high divergence from all other flaviviruses examined. The RT-LAMP primer set was designed with the PrimerExplorer V4 software (Eiken Chemical Co. Ltd.). A BLAST search of the GenBank nucleotide database was carried out for the selected primers' sequences to verify specificity. The RT-LAMP sequences were synthesized by a commercial vendor (IDT, Coralville, IA) and documented in Table S1 together with the concentrations used in our reaction mix.</p><!><p>ZIKV American strain (mex 2–81), DENV-2 NGC, and CHIKV (181/clone25) were obtained from the World Reference Center for Emerging Viruses and Arboviruses (WRCEVA, Dr. Robert Tesh, Director) and propagated in the mosquito cell line C6/36 cells. The viral concentration (PFU/mL) was determined with a plaque assay on Vero cells. Inactivated HIV-1 virus was purchased from Thermofisher Scientific (AcroMetrix HIV-1 High Control, Benicia, CA).</p><!><p>We collected urine, saliva, and whole blood samples from healthy, consenting, adult volunteers. The ZIKV was spiked in the urine samples and saliva samples at various concentrations. For safety, the ZIKV was first inactivated by mixing the samples laden with ZIKV with binding/lysis buffer (AVL, QIAamp Viral RNA Mini Kit) in a BSL2, followed with addition of ethanol to the lysate according to the manufacturer's protocol.21 Similarly, inactivated HIV virus was spiked in plasma separated from whole blood with our POC plasma separator.22</p><!><p>Zika viral RNA was extracted with Qiagen Viral RNA mini kit (QIAGEN, Valencia, CA), following the manufacturer's protocol. In addition to the primers and template, the RT-LAMP reaction mix (25 μL) included: 1× OptiGene Isothermal Master Mix ISO-100 (OptiGene, U.K.), 3 U of AMV reverse transcriptase (Invitrogen, Carlsbad, CA), and 0.5 μL of EvaGreen fluorescent dye (Biotium, Hayward, CA). Amplification was carried out and monitored with Peltier Thermal Cycler PTC-200 (Bio-Rad DNA Engine, Hercules, CA) at 63°C. Fluorescence emission intensity data were collected once every minute for 60 min.</p><!><p>We used our custom-made microfluidic chips with four independent multifunctional, isothermal amplification reactors (MIAR) (Figure 1c). For each test, lysate was filtered through the nucleic acid isolation membrane of one of the amplification reactors. The nucleic acids bound to the membrane. Subsequent to the sample introduction, 150 μL of Qiagen wash buffer 1 (AW1) was injected into the reactor to remove amplification inhibitors. Then, the silica membrane was washed with 150 μL of Qiagen wash buffer 2 (AW2), followed by air-drying for 30 s. Next, 25 μL of bioluminescent assay in real-time and loop-mediated isothermal amplification (BART-LAMP) master mix, which includes 1× OptiGene Isothermal Master Mix ISO-100nd (OptiGene, U.K.), 3 U of AMV reverse transcriptase (Promega, USA), 2.5 μL of BART reporter (Lot: 1434201; ERBA Molecular, UK), and LAMP primers (Table S1) were injected into each reactor. The inlet and outlet ports were then sealed with transparent tape.</p><!><p>Commercially available Mg–Fe alloy pouch (chemical heater) of the type used in Meal, Ready to Eat (MRE, Innotech Products Ltd., USA) was first placed in the drawer (Figures 1a and S1a). When 7.5 mL of tap water was added to the drawer, the magnesium oxidized rapidly and the reaction produced heat. After approximately 10 min, once the heat sink's temperature exceeded 60°C, the MIAR chip was inserted into the SCC and the smartphone was initiated to record in real-time the bioluminescence emission from the reactors. The phone camera acquired an image once every min for 60 min. The images obtained with the smartphone camera were analyzed, and the average bioluminescence intensity signal for each reactor was extracted and depicted as a function of time. The detection results can be sent to the cloud.</p><!><p>To enable on-site, rapid, connected, molecular detection, we designed, prototyped, and tested a simple, inexpensive, compact, smart-connected cup (SCC; Figures 1a,b and S1a and Video S1). The SCC consists of (i) a Thermos cup body with vacuum insulation, (ii) a smartphone (i.e., Samsung Galaxy S6) with custom written App, and (iii) a 3D-printed holder (i.e., chip holder, cup lid, and a smartphone adaptor). The chip holder was designed to accommodate our MIAR chip (Figure 1c) and interface our chip with a heat sink. To facilitate real-time bioluminescence monitoring, a detection window was formed in the center of the cup lid. The smartphone adapter was part of the cup lid. Figure 1b is a photograph of our SCC equipped with a smartphone. The smartphone with our App is capable of real-time imaging/recording of the bioluminescence signal (Figure 1d), processing images, quantifying nucleic acid concentration, and reporting/transmitting test results.</p><p>A phase-change material (PCM) with a melting temperature of 68°C (PureTemp 68, Entropy Solutions Inc., Plymouth) regulates the MIAR's incubation temperature at a level appropriate for isothermal amplification, typically about 63°C. The heat generated by the exothermic reaction is transferred through the PCM and aluminum heat sink to the MIAR chip. Excess heat is consumed as a latent heat without any temperature increase beyond the desired value, independent of ambient conditions.</p><!><p>We developed a custom Android App that controls camera exposure time and image acquisition rate, processes smartphone camera images of the reactors, and reports/transmits test results (Figure 2). Detailed information about this App is given in Figure S2. Our App was developed with Eclipse Integrated Development Environment (IDE) in Android Developer Tools and Java. Our App has three primary functions: (i) real-time bioluminescence image capture, (ii) image processing/analysis, and (iii) test result reporting and transmission.</p><p>The main menu of the App (Figure 2b (i)) provides operating instructions, solicits information from the user, and allows the user to alter default parameters such as camera exposure time, total test time, and image capture frequency. The user is also prompted to select the detection regions of the reactor array displayed in the preview window (Figure 2b (ii)). The App takes images at regular intervals (i.e., once a minute) of the light emission from the user-defined regions and constructs arrays of bioluminescence intensities as functions of incubation time for all reactors (Figure 2b (ii)). To ensure consistent conditions, the camera focus and exposure time are locked during the entire process. To shorten the detection distance, a 0.67× magnetically mounted wide lens (Amazon) is used. All images are saved in Joint Photographic Experts Group (JPEG) format with a resolution of 96 dots per inch (dpi) and 8 bits per RGB channel (24 bits in total). Intensity values are calculated by summing the individual RGB values of each pixel. The intensity values are then graphed in real-time; threshold times are calculated, and the data is saved for future analysis.</p><p>For target quantification, one can operate two of the reactors as calibration reactors with known numbers of template copies to establish a linear relationship between the threshold time and the log of the number of template copies (Figure 2b (iii)). Alternatively, a calibration formula can be stored in the App. At the conclusion of the test, the calibration formula can be used to determine the number of templates in each test.</p><p>To enable spatiotemporal disease mapping, GPS coordinates of each test location are recorded. The test results, location, and time stamp can then be communicated to a secure server by our smartphone App (Figure 2a). The App can also provide counseling and links to support resources.</p><!><p>The MIAR chip (Figure 1c) that was used in our experiments contained four independent multifunctional isothermal amplification reactors. Each reactor includes a porous silica nucleic acid (NA) isolation membrane (Figure 3a) for nucleic acid capture, concentration, and purification from raw samples (<5 min), eliminating the need for centrifugation and spin columns that are commonly used in centralized laboratories without sacrificing performance.23 More importantly, our flow-through NA-isolation membrane decouples sample volume from the reactor's volume, enabling the use of relatively large sample volumes (several hundreds microliters) for high sensitivity while maintaining small reaction volume and small reagent consumption. We fabricated the chip with PMMA (poly-(methyl methacrylate)) via computer numeric control (CNC) machining.23</p><p>Real-time monitoring of nucleic acid amplification is commonly used for the quantification of target nucleic acids in a sample. Typically, real-time quantitative detection records fluorescence emission intensity of an intercalating dye or fluorescent beacons to monitor synthesis of polynucleotides during enzymatic amplification.24–27 However, fluorescence readout necessitates an excitation light source and optical filters to separate excitation and emission spectra. Since many materials exhibit autofluorescence, fluorescence detection is also susceptible to unwanted background emission. All these increase cost and complexity. Since the BART reporter's emission is induced by the amplification reaction itself, one can directly monitor the amplification process with a smartphone camera (Figures 1d, 3b, and S3) without a need for excitation and optical filters, enabling a simple, inexpensive, and portable smartphone-based molecular detection platform.</p><p>Since the emission intensity detected by the smartphone camera increases with increasing camera's exposure time (Figure 3c), our App overrides the automated camera shutter exposure time and maintains shutter exposure time fixed for all images, independent of light intensity. At the start of the monitoring process, we imaged the background to establish a baseline. This background was then subtracted from all subsequent images.</p><p>To select appropriate exposure time, we evaluated the recorded light intensity as a function of camera exposure time (Figure 3d). As the camera exposure time increases, so does the recorded intensity. However, too long camera exposure times lead to increased background noise. As a reasonable compromise, we use camera exposure time of 40 s in all our experiments.</p><!><p>BART reporter includes adenosine-5′-O-phosphosulfate (APS), ATP sulfurylase, firefly luciferase, and luciferin (Figure S3). We examined the effect of BART reporter concentration on light emission by varying its volume in the LAMP reaction mix from 0 to 4 μL (Figure 4). Figure 4a depicts the emission intensity as a function of time for various BART reporter concentrations. In the absence of the reporter, the bioluminescence intensity remains very low throughout the entire incubation time. In the presence of the BART reporter, the emission intensity is initially low but increases sharply as the rate of DNA synthesis increases. The emission intensity peaks and then, as the polymerase reaction slows, it decays slowly with a long tail. As the BART reporter concentration increases so does the emission intensity peak height (Figure 4b). In the range of reporter concentrations considered here, we did not observe any inhibitory effects of BART reporters. Importantly, BART reporter concentration does not significantly affect the threshold time, which we define as the time interval from the beginning of incubation until the signal reaches half its peak height (Figure 4c). In all the experiments reported in the remainder of this paper, we used 2.5 μL of BART reporter solution.</p><p>Earlier workers19 reported that BART emission decays rapidly after peaking, while in our experiments, we observe a slow decay (Figure 4a). The sharp peak is attributed19 to the competing effects of pyrophosphate (PPi). During the exponential phase of amplification, PPi produced by polymerase rapidly converts into ATP that fuels bioluminescence emission. Subsequently, excess free PPi inhibits luciferase, reducing light intensity after the peak. Previous researchers19 used LAMP reaction mix without pyrophosphatases (PPases). In contrast, we use OptiGene LAMP28 assay in which PPases is added to improve amplification efficiency.29,30 We hypothesize that PPases' interaction with PPi drives conversion of luciferin into Oxiluciferin to produce light and is responsible for the light emission after the peak. To examine the effect of PPases on real-time BART-LAMP amplification curves, we compared Eiken LAMP buffer31 in the absence and presence of PPases. Eiken LAMP without PPases produced a high initial background signal and a unique sharp peak (Figure S4a) consistent with previous reports19 of amplification curves obtained in the absence of PPases. When we added 1U PPase (New England Biolabs, Inc.) to the Eiken LAMP buffer, we observed a reduction in the background signal and broadening of the bioluminescence peak (Figure S4b) similar to the one observed with the OptiGene LAMP buffer (Figure S4c) that contains PPases. In this work, we use OptiGene LAMP buffer that includes PPases because it produces low initial background.</p><!><p>ZIKV infection is of growing concern since it is implicated in brain defects in newborns, including microcephaly, and in severe neurological and autoimmune complications.32–34 A few studies suggest that ZIKV detection in urine is more sensitive than in blood and has a longer window of detection.35,36 Since ZIKV is coendemic with and shares similar initial symptoms as infections caused by other arboviruses, such as dengue viruses (DENV), yellow fever virus (YFV), and chikungunya virus (CHIKV),37,38 diagnosis is needed for disease management.</p><p>To demonstrate the compatibility of our SCC platform with bioluminescent detection for point-of-care molecular diagnostics, we used ZIKV as the model analyte and carried out viral RNA extraction, amplification, and detection on chip (Figures 1c and 3a). We redesigned, for improved efficiency, a set of RT-LAMP primers,20 targeting the highly conserved envelope protein coding region (Figure S5). To address for possible mutations and maintain high amplification efficiency, we use degenerate primers. The primers' sequences are given in Table S1. The primers were evaluated for possible cross-reactivity with other pathogens' nucleic acids en silico and with DENV and CHIKV nucleic acids (Figures 5a and S5a). Our primers successfully distinguished ZIKV from DENV and CHIKV.</p><p>Video S2 shows real-time bioluminescence monitoring of four multifunctional amplification reactors with 500, 50, 5, and 0 (negative control) PFU ZIKV (American strain mex 2–81) in urine samples. The higher the target concentration, the earlier the amplification reactor lights up. In the negative control reactor, the bioluminescence intensity remained nearly constant and very low throughout the entire incubation period. Figure 5b shows the bioluminescence emission intensity as a function of time when the sample contains 500, 50, 5, and 0 PFU of ZIKV in urine samples. We define the threshold time (Tt) as the time that elapses from the start of the enzymatic reaction until the bioluminescence intensity reaches half its peak height. Figure 5c depicts the threshold time Tt (min) as a function of ZIKV concentration (C) on a semilog plot. Our experiments indicate that our SCC is suitable for nucleic acid amplification and that the use of threshold time measurement provides reliable target quantification with a sensitivity of 5 PFU per urine sample.</p><p>Saliva is considered a suitable and convenient sample type for ZIKV.39,40 Here, we carried out sequence experiments to detect ZIKV in saliva (Figure 5d,e). We achieved a similar performance to the one reported with urine samples (Figure 5b,c). Our SCC's results are also comparable with the ones obtained with a benchtop instrument (Figure S5b). Our studies indicate that ZIKV in urine and saliva can be detected with a simple POC device with performance comparable to benchtop equipment.</p><!><p>Individuals undergoing HIV therapy require periodic viral load monitoring to (i) assess evolution of mutations that may require alternations in therapy and (ii) ensure compliance with the drug regimen.41 To demonstrate our SCC platform's capability to process diverse human samples and targets, we tested the device's suitability for HIV detection (Figure 5f). The test was carried out with plasma spiked with HIV virions. Plasma was separated from whole blood with our POC plasma separator22 and then inserted into our chip.</p><!><p>Interconnectivity is a hallmark of contemporary technology. Sensors, actuators, and embedded electronics enable real-time, networked data collection and analysis, as well as distributed control, connectivity, customization or personalization, and adaptability, forming an Internet of Things. In the area of healthcare, an Internet of Medical Things (IoMT),42–45 i.e., interconnected, networked medical devices, will offer new paradigms of diagnosis, personalized therapy, epidemiology, and healthcare services in a timely, sustainable, individualized, and cost-effective mode of delivery. However, current IoMT devices are limited to detecting or monitoring physiological data such as heart rate, blood pressure, body temperature, ECG, and blood glucose.42–45 IoMT POC molecular diagnostic devices that monitor diseases have many potential benefits.</p><p>To demonstrate the connectivity of our SCC as an IoMT POC molecular diagnostic device and its utilization for spatial disease mapping, we designed a website to map GPS locations of the tests in a designated region (Figure 6). The website can show the detection information, such as location of tests and test results. When desired, other patients' information such as age, gender, ethnic group, and occupation can be included for spatial epidemiology studies based on big data analysis.</p><!><p>We designed, fabricated, and tested a simple, inexpensive, handheld, smartphone-based mobile detection platform for molecular diagnostics of infectious diseases in diverse biological samples such as urine, saliva, and blood. Our disposable microfluidic chip houses multifunctional reactors, each equipped with a flow-through nucleic acid immobilization membrane to capture, concentrate, and purify nucleic acids from raw samples. Our flow-through configuration decouples sample volume from reaction volume and enables processing of large sample volumes for high sensitivity with little added complexity over that of rapid tests operating with raw samples. Our method also calls for the storing of lyophilized reaction mix, encapsulated in paraffin, in the amplification reactor for just-in-time release and hydration once the reactor reaches its operating temperature. This storage method enables long shelf life without refrigeration and streamlines flow control. We demonstrated this storage strategy elsewhere46 but have not used it here due to cost considerations of lyophilizing small quantities of reaction mixes. Our SCC utilizes exothermic chemical reaction as a heat source to incubate the isothermal amplification and a phase-change material to regulate incubation temperature and maintain it independent of ambient conditions. In other words, the SCC operates without any need for electrical power and with a very simple, inexpensive means to control temperature. Our SCC also provides an interface for a smartphone.</p><p>In contrast to the fluorescence reporters, we use here luciferin reporters fueled by byproducts of polymerase to emit light. Bioluminescent reporters have several important advantages over the more commonly used fluorescent reporters. First, bioluminescent reporters do not require excitation. Second, microfluidic components may self-fluoresce and produce background emission, while background emission is absent with bioluminescent reporters. Third, fluorescent reporters require the use of optical filters to separate between excitation and emission spectra while no filters are needed when bioluminescent reporters are used, which reduces cost and complexity.</p><p>We have developed a custom smartphone App that can instruct the user in carrying out the various steps needed for the test. Once the diagnostic chip is inserted into the SCC, our custom App adjusts the camera's shutter to provide a sufficiently long exposure time to acquire detectable bioluminescent emission and dictates camera frame rate. The smartphone camera then monitors the bioluminescent emission during polymerase amplification. The App digitizes and stores emission intensity as a function of time and calculates threshold time. The software can compare the test's threshold time with a calibration table and report positive or negative test and number of virions on the smartphone's screen.</p><p>Here, we demonstrated our SCC's performance by quantitative detection of zika virus in urine and saliva and HIV in blood within 45 min. In comparison, current emergency approved tests utilize expensive and complex thermal cyclers (i.e., PCR machines); require trained personnel; require sample shipment to a centralized laboratory, risking sample degradation; may take a few days from test to results, assuming reliable logistic networks. The current technology deprives health care providers from essential information to make timely evidence-based decisions. With our SCC, the test can be performed by the patients themselves or next to the patient by minimally trained personnel, making sophisticated molecular diagnostics accessible at home and in resource-limited settings. The SCC can, of course, be adapted to detect other targets than ZIKV and HIV and is amenable to multiplexing.47</p><p>Additionally, we have demonstrated SCC's connectivity, enabling spatiotemporal disease mapping on a custom-designed website. Our SCC serves as an IoMT device for molecular diagnostics that can transmit test results to the doctor's office and communicate test information (i.e., GPS location, time, and deidentified results) to public health officials, providing critical data to policy makers and epidemiologists.</p><!><p>Video S1: An animation illustrating the various components of the smart cup (MPG)</p><p>Video S2: Real time smartphone-based detection of BART-LAMP emission from a multifunctional amplification chip (MP4)</p>

PubMed Author Manuscript

Linear Atomic Cluster Expansion Force Fields for Organic Molecules: beyond RMSE

We demonstrate that fast and accurate linear force fields can be built for molecules using the Atomic Cluster Expansion (ACE) framework. The ACE models parametrize the Potential Energy Surface in terms of body ordered symmetric polynomials making the functional form reminiscent of traditional molecular mechanics force fields. We show that the 4 or 5body ACE force fields improve on the accuracy of the empirical force fields by up to a factor of 10, reaching the accuracy typical of recently proposed machine learning based approaches. We not only show state of the art accuracy and speed on the widely used MD17 and ISO17 benchmark datasets, but also go beyond RMSE by comparing a number of ML and empirical force fields to ACE on more important tasks such as normal mode prediction, high temperature molecular dynamics, dihedral torsional profile prediction and even bond breaking. We also demonstrate the smoothness, transferability and extrapolation capabilities of ACE on a new challenging benchmark dataset comprising a potential energy surface of a flexible drug-like molecule.

linear_atomic_cluster_expansion_force_fields_for_organic_molecules:_beyond_rmse

Introduction<!>Atomic Cluster Expansion basis functions<!>Choice of radial basis<!>Basis Selection<!>Parametrization of the linear ACE potentials<!>MD17<!>Learning curves<!>Normal mode analysis<!>Extrapolation in temperature<!>Extrapolation far from the training set<!>Fitting multiple molecules<!>Flexible molecule test: 3BPA<!>Preparation of the data set<!>Comparison of force fields models<!>Conclusions

<p>The efficient simulation of the dynamics of molecules and materials based on first principles electronic structure theory is a long standing challenge in computational chemistry and materials science. There is a trade-off between the accuracy of describing the Born-Oppenheimer potential energy surface (PES) 1 and the length and time scales that are accessible in practice. A convenient way to measure this trade-off is by considering the total number of simulated atoms, which can be a result of either generating a few configurations consisting of many atoms, or many configurations (e.g. a long molecular dynamics trajectory) each consisting of fewer atoms. Explicit electronic structure simulations are extremely accurate and systematically improvable. They can treat on the order of a million simulated atoms in total using either cubic scaling methods and molecular dynamics, or linear scaling algorithms on larger systems. Alternatively, in order to simulate many orders of magnitude more atoms, the PES can be parametrized in terms of the nuclear coordinates only. In this way, the electrons do not have to be treated explicitly, which simplifies the simulations con-siderably. These methods can routinely model a trillion (10 12 ) or more simulated atoms.</p><p>When parametrizing the PES, it is natural to decompose the total energy of the system into body ordered contributions, which can then be resummed into local atomic (or site) energies. The site energy of atom i is written as</p><p>where indices j, k run over all neighbors of atom i (either unrestricted, or within a cutoff distance r cut ), z i denotes the chemical element of atom i and r ij = r j − r i the relative atomic positions.</p><p>The traditional approach to the parametrization of the body ordered terms for molecular systems is to use physically motivated simple functional forms with few parameters, leading to "empirical force-fields". These models typically require a pre-determined topology, meaning that the parameters describing the interactions of a certain atom depend on its neighbors in the bonding graph that is specified before the simulation and is not allowed to change. [2][3][4][5] The potential energy is then written as a sum of body-ordered bonded and non-bonded terms, for example:</p><p>where r, θ and φ describe the intramolecular bond lengths, angles and dihedral angles in the molecule, and E non−bonded contains a Lennard-Jones (LJ) term accounting for van der Waals and short-range repulsive interactions and a Coulomb term to describe the long-range electrostatics. The bonded terms can be made equivalent to the body order in eq (1) by rewriting the sum over atom-tuples into sums over sites. The advantage of the simple functional form of the bonded terms is very fast evalua-tion and ease of fitting due to the small number of free parameters. 2,[6][7][8] On the other hand, this simplicity limits the achievable accuracy 9 and requires significant modification to incorporate reactivity. 10 Note that while in the most widely used force fields, the non-bonded interactions are two-body, this is not the case for polarizable force fields, such as Amoeba. 11 Moreover, the direct evaluation of terms beyond 3-body contributions is computationally expensive, in general growing exponentially with the body order, which severely limits the possibility of systematically improving force fields by adding higher body order terms.</p><p>Over the past ten years a new approach has emerged, employing machine learning (ML) methods to parametrize the PES. Instead of the body order expansion, the site energy is approximated by a neural network or a Gaussian process regressor (GPR) both of which are extremely flexible functional forms, proven to be universal approximators. 25 Due to this flexibility there is no need to specify topology or atom types beyond the identity of the chemical element, and much higher model accuracy can be achieved given an appropriate (typically rather large) training set. On the other hand, this flexibility comes also at a cost: there is no guarantee that the behavior of these ML models remains chemically sensible in regions of configuration space where there is not enough training data. Spurious local minima or even wildly wrong atomization energies are par for the course. 26 The most prominent examples of ML models are Atom Centred Symmetry Function based feed forward neural networks introduced by Behler and Parinello 27 that also includes the family of ANI force fields 17,28 and DeepMD, 18 the atomic neighborhood density based GPR models like Gaussian Approximation Potentials (GAP) 14,29 and FCHL, 15 the gradient domain kernel based sGDML, 16 and message passing graph neural network based Schnet, 23 Physnet, 24 DTNN, 30 and DimeNet 21 and most recently the covariant or equivariant neural network based Cormorant 22 and PaiNN. 19 There is also a third family of methods, which expands the PES as a linear combination of body-ordered symmetric polynomial ba-Table 1: Comparison of different force field fitting approaches. Molecular mechanics (e.g. AMBER, 2 CHARMM 12 and OPLS 13 ), machine learning: Kernels (GAP, 14 FCHL 15 and sGDML 16 ), Neural Networks (ANI, 17 DeepMD, 18 PaiNN, 19 GMsNN, 20 DimeNet, 21 31,32 (Permutationally Invariant Polynomials (PIPs)), which approximated the PES of small molecules to extremely high accuracy, albeit with exponential scaling in the number of atoms. Introducing finite distance cutoffs reduces this scaling to linear, and the resulting atomic body-ordered permutationally invariant polynomials (aPIPs) have been shown to achieve high accuracy and better extrapolation compared to the above nonlinear machine learning based approaches in both molecular and materials systems. 26,33 The main limitation of the aPIPs approach is that the evaluation time of the site energy increases quickly with body order, making it essentially impossible to go above body-order 5 (certainly when the five atoms are of the same element). More recently, the Atomic Cluster Expansion (ACE) 34,35 (and the earlier Moment Tensor Potentials 36 ) are formulations of symmetric polynomial approximations that remove the steep scaling of the evaluation of the site energy with the number of neighbors independently of body order, resulting in highly efficient interatomic potentials for materials. 37 Table 1 compares the main features of the classical force fields, machine learning based potentials and the linear Atomic Cluster Expansion force fields. In one sense, the linear ACE constitutes a middle ground between the other two: it retains the chemically natural body order, but lifts the limitations of fixed topology and inflexible functional form embodied in eq (2).</p><p>The purpose of the present paper is to demonstrate the performance of linear ACE force fields for small organic molecules. After briefly reviewing the general ACE framework and outlining the necessary choices that go into fitting our linear models, we start with the MD17 38 and ISO17 23 benchmark data sets. We are particularly interested in going beyond the RMSE (or MAE) of energies and forces (the typical target of the loss function in the fit), because practically useful force fields have other desirable properties too: chemically sensible extrapolation, good description of vibrational modes, and accuracy on trajectories self-generated with the force field, just to name a few. The insufficient nature of mean error metrics has been pointed out before. [39][40][41] In addition to the above data sets, we also demonstrate the use of ACE on a slightly larger, significantly more flexible molecule that is more representative of the needs of medicinal chemistry applications.</p><p>The programme of tests as we outlined is designed to explore the capabilities and properties of different approaches to making force fields. We emphasize here that we are not making or testing force fields that are in and of themselves generally useful to others. That is a significant undertaking and it is to be attempted once we better understand these capabilities and properties, and are able to select which approach has the best prospects. Therefore, in addition to quoting literature results for recently published ML schemes, we refit a number of them, where the necessary software is available (sGDML, ANI and GAP in particular), so that we can show their performance on our tests. We also refit a classical empirical force field (eq (2)) to exactly the same training data to more rigorously quantify the anticipated accuracy gains of the ML and ACE approaches.</p><!><p>The atomic cluster expansion (ACE) model 34,35 keeps the body ordering of terms defined in eq (1), but reduces the evaluation cost by eliminating the explicit summation over atomtuples. This is accomplished by projecting the atomic neighbor density onto isometry invariant basis functions. This idea, detailed below, is referred to as the "density trick", and was introduced originally to construct the power spectrum (also known as SOAP) and bispectrum descriptors 14,42 (which are in fact equivalent to the 3-and 4-body terms in ACE, respectively, so in a sense the ACE invariants can be considered a generalization of these to arbitrary body order). We start by defining the neighborhood density of atom i as</p><p>where ρ z i denotes the density of atoms of element z in the neighborhood of atom i. This density is projected onto a set of 1-particle basis functions, which we choose to be a product of a radial basis and real spherical harmonics:</p><p>Here the "1-particle" refers to the single sum over neighbors, with the central atom i serving as the center of the expansion. There is considerable flexibility in the choice of the radial basis; the specifics for this work are documented at the end of this subsection. We then define the atomic base as the projection of the neighborhood density onto the 1-particle basis functions</p><p>where the index z i refers to the chemical element of atom i. For notational convenience, we collect the rest of the 1-particle basis indices into a multi-index,</p><p>From the atomic base A z i v , we obtain permutation-invariant basis functions, which we will call the "A-basis", by forming the products,</p><p>The product containing ν factors gives a basis function that is the sum of terms each of which depends on the coordinates of at most ν neighbors, and we refer to it either as a ν-correlation or as a (ν +1)-body basis function (the extra +1 comes from the central atom i). A graphical illustration of this construction is shown in fig 1 for the special case where the two factors are the same. For many (different) factors, taking products of the atomic base (left side of fig 1) takes a lot less time to evaluate than the explicit sum of all possible products (right side of fig 1). This is the key step that we referred to as the density trick.</p><p>The A-basis is not rotationally invariant. We therefore construct a fully permutation and isometry-invariant overcomplete set of functions, which we call the B-basis (technically not a basis but a spanning set), by averaging the Abasis over the three dimensional rotation group,</p><p>Figure 1: Construction of high body order invariant basis functions. A graphical illustration showing how higher body-order basis functions can be constructed as products of the projected neighborhood density. The evaluation cost of the basis functions scales linearly with the number of neighbors rather than exponentially by doing the density projection first and than taking the products to obtain higher order basis functions. The figure (and expression) also makes explicit the occurrence of self-interaction terms in the ACE basis. They are automatically corrected through the inclusion of lower-order correlations in the basis.</p><p>O(3),</p><p>where the matrix of Clebsch-Gordan coupling coefficients C vv is extremely sparse. Many of the resulting basis functions will be linearly dependent (or even zero), but it is relatively straightforward to remove these dependencies in a pre-processing step, to arrive at an actual basis set. We refer to Dusson et al. 35 for the details of the procedure outlined up to this point. The B-basis in eq (8) is complete in the sense that any function of the neighboring atoms that is invariant to permutations and rotations can be expanded as a linear combination of the basis functions. We therefore write the site energy of ACE as</p><p>The above equation makes it clear that the model is linear in its free parameters, the c coefficients. The B-basis functions are polynomials of the atomic coordinates, and in order to show that the explicit body ordering has been retained, we can switch back to using the Abasis (with the product explicitly written out),</p><p>where the c can be obtained as linear combina-tions of the c coefficients appearing in eq (10), using the transformation defined in eq (9). Now the body-ordering is readily identified. Each term corresponds precisely to a sum of ν-correlations, i.e. (ν + 1)-body terms as in the traditional body-order expansion, eq (1). In practice, we use a recursive scheme 35 that leads to an evaluation cost that is O(1) per basis function, independent of body-order. The number of basis functions does grow with body order, at a rate that has an exponent ν.</p><p>The construction outlined so far yields infinitely many polynomials B z i v , which can be characterized by their correlation-order ν, and their (modified) polynomial degree D = ν t n t + w Y l t , where n t and l t come from the multi-index v t and the weight w Y is used to trade-off the radial and angular resolution of the basis set. When it comes to defining a model in practice the expansion is truncated both in the body-order and in the maximum polynomial degree at each body-order.</p><!><p>In the models in this paper we will not use much of the flexibility of the ACE framework, and simply take R</p><p>r → x(r) is a one dimensional radial transformation, f cut is a cutoff or envelope function and p n are orthogonal polynomials. For the radial transform we take</p><p>which amplifies the effect of neighbors closer to the central atom. For the cutoff function we specify both inner and outer cutoffs, r in < r out , and define</p><p>The polynomials p n are then defined recursively by specifying that p 0 (x) = 1, p 1 (x) = x, and the orthogonality requirement</p><p>where we have used the inverse of the radial transform, x → r(x). Eq (15) implies that the radial basis R n and not the polynomials p n are orthonormal in x-coordinates.</p><p>The introduction of an inner cut-off is necessary to prevent wildly oscillating behaviour in high energy regions of configuration space where pairs of atoms are very close to one another and little or no training data is available. Alternatively, one could introduce such training data, but that would unnecessarily complicate the construction of training data sets and this inner cutoff mechanism is sufficient. To ensure short range repulsion we augment the large multi-body ACE basis by a small auxiliary basis set, consisting only of low-polynomial-degree pair interaction (two-body) functions. The construction is exactly the same as before, but we change the cut-off function to</p><!><p>Before we can parametrize the ACE force field we need to select a specific finite basis set chosen from the complete ACE basis constructed in the previous section. There are three approximation parameters: the cutoff radius (r cut = r out ), the maximum correlation order ν max , and the maximum polynomial degrees D max ν corresponding to order ν basis functions. We have already specified the cut-off radius in the definition of the radial basis in eq (12). The basis is then chosen as (a linearly independent subset of) all possible basis functions B iv with correlation order at most ν max and polynomial degree at most D max ν . In all models for molecules with three or fewer distinct elements we take ν max = 4, which corresponds to a general 5-body potential. In models for molecules with four or more distinct elements we reduce this to ν max = 3 (4-body potential). The weight w Y specifies the rela-tive importance of the radial and angular basis components; here we choose w Y = 2. The maximum polynomial degrees D max ν can be adjusted to balance the size of the basis set against fit accuracy and evaluation time; the precise parameters we choose for each molecule are given in Table S1. The basis truncation we specified here is just one, rather simple, way to obtain a finite basis. There may very well be more sophisticated methods to choose an optimal subset of the complete basis.</p><!><p>We define the total energy of a linear ACE model with parameters c corresponding to a spatial configuration of atoms (denoted by X, e.g. a molecule in a particular configuration) as the sum of the site energies</p><p>where E i is a site energy defined in eq (10). Optimal parameters are obtained by minimizing the loss function</p><p>where the E QM and F QM are energies and forces, respectively, in the training data, obtained from electronic structure calculations. The sum is taken over all configurations in the training set, and w E X , w F X are weights specifying the relative importance of energies and forces. Since the model energy and force are both linear in the free parameters, the loss can be written in a linear least squares form,</p><p>where the vector t contains the QM energy and force observations, and the design matrix Ψ contains the values and gradients of the basis evaluated at the training geometries. Ψ has a number of rows equal to the total number of observations (energies and force components) in the training set, and a number of columns equal to the total number of basis functions.</p><p>The least squares problem has to be regularized, especially when the basis contains high degree polynomials. 33 One option is to apply Tychonov regularization, where the loss function is modified as</p><p>This is widely used to regularize linear regression, often by taking Γ as just the identity matrix, or alternatively in the case of kernel ridge regression (and Gaussian process regression) as the square root of the kernel matrix. 43 In the present case, we use a diagonal Γ with entries corresponding to a rough estimate for the p-th derivative of the basis functions,</p><p>where n t and l t are part of the elements of the multi-index vector v (cf. eq (6)). This scales down high degree basis functions, encouraging a smooth potential, which is crucial for extrapolation, and is loosely analogous to the smooth Gaussian prior of GPR. The actual solutions are then found using the standard iterative LSQR solver, 44 for the details see the SI.</p><p>In the other approach we used for solving the least squares problem the same Γ matrix is introduced, but without a Tychonov term,</p><p>and the solution is found using the rank revealing QR factorisation 45 (RRQR), in which we perform a QR factorization of the scaled design matrix ΨΓ −1 , and truncate the small singular values below some tolerance parameter λ.</p><p>For more details of the exact implementation see Refs. 26,45. We found that when the linear system is not underdetermined, RRQR gave somewhat better solutions than LSQR. All parameters of the optimization (w E X , w F X , p, λ) are given in the SI.</p><p>The last modelling choice that needs to be made is the 1-body term, that is the energies of the isolated atoms of each element in our model. One can use the energy of the isolated atoms evaluated with the reference electronic structure method, which ensures the correct behavior of the model in the dissociation limit. In other words, that the force field is modelling the binding energy of the atoms. An alternative approach, often used in the ML fitting of molecular energies, is to take the average energy of the training set, divided by the number of atoms in the molecule, and assign the result to each element. In this case, the fitted model has zero mean energy. This usually improves the fit accuracy slightly, by reducing the variance of the function that we need to fit, in case the data spans a narrow energy range around its average, e.g. because it came from samples of moderate temperature molecular dynamics.</p><p>A third option is to not use any reference potential energies for the fit, but only forces. Once the coefficients are determined, the potential can be shifted by a constant energy chosen to minimize the training set energy error. In the current work, we evaluated all three strategies for ACE and found that using the isolated atom energies for the 1-body term gives slightly higher RMS errors, but leads to far superior extrapolation. The other two strategies (using the average energy for the 1-body term, and fitting only to forces) result in similar somewhat lower test set errors, but inferior physical extrapolation properties.</p><p>As mentioned in the introduction, we view tests on data sets such as MD17 and ISO17 as proxies: the models thus created are not useful for any scientific purpose. The promise of ML force fields is greatest when the intention is to describe a very wide variety of compounds and conformations, perhaps including chemical reactions. With this in mind, the most natural choice for the 1-body term is to choose it to match the energy of the isolated atom in vacuum. This choice is independent of any particular data set, and the apparent advantages of the other choices in terms of lower errors are expected to diminish in the limit of a large and wide ranging data set.</p><!><p>The original MD17 benchmark data set consists of configurations of 10 small organic molecules in vacuum sampled from density functional theory (DFT) molecular dynamics simulations at 500 K. 38 It has recently been recognized, that some of the calculations in the original data set did not properly converge, in particular, many of the forces are noisy. A subset of the full data set was recomputed with very tight SCF convergence settings and is called the rMD17 (revised MD17) data set. 46 We have used this new version of the data set and the five train-test splits as reported in Ref. 46. These revised training sets consist of 1,000 configurations to avoid the problem of correlated training and test sets: when more than 1,000 configurations are used from the full published trajectory, some of the test set configurations will necessarily fall between two neighboring training set data points that are separated by a much smaller time difference than the decorrelation time of the trajectory, resulting in an underestimation of the generalization error. 46 Table 2 shows the Mean Absolute Error (MAE) of the different force field models trained on 1,000 configurations. For comparison, in Tables 2 and 3, we include a wide selection of models from the various classes of force field fitting approaches that we discussed in the Introduction (Table 1). They include ML approaches such as feed forward neural networks (ANI, GMsNN, GMsNN), Gaussian Process regression models (sGDML, FCHL, GAP) and graph neural network based models (DimeNet, Schnet, Physnet, Cormorant and PaiNN). The models on the left of Table 2 were trained by us, (except for FCHL) using the exact traintest splits of rMD17, whereas the models on the right of the solid vertical line are from the literature and were trained on the original MD17 data set using different train-test splits. The precise details of the fitting procedures and parameters for each of the models can be found in the SI.</p><p>Of the descriptor based models, sGDML, 17 This was crucial for achiev-ing the errors shown. When the weights were initialized randomly, the errors are higher by factor of 2 (Table S2). The GAP model, using SOAP features to describe the atomic geometry (which are similar to ANI's features), achieves similar errors to the ANI model with pre-training. The fact that ANI is only competitive with GAP if it is pre-trained can be rationalized by the relative sample efficiency of kernel models compared to neural networks. The FCHL kernel models also use 2-and 3-body correlations as features, but they have been more carefully optimized for molecular systems and hence are able to achieve very low errors. 15 The classical force field (FF) refers to a reparametrization of the GAFF functional form 2,47 using the ForceBalance program 6,47 and the rMD17 training set. This model gives at least an order of magnitude higher errors compared to the ML force fields. This is not a huge surprise, but is nevertheless a quantitative characterization of the limitations of the fixed functional form for a situation in which the empirical force fields are designed to do well.</p><p>For completeness, in Table 3 we show the MAEs of the neural network models reported in the literature that were trained on 50,000 structures from the original MD17 trajectories. The test set errors of these models are probably underestimating the true generalization error, because the large training set contains configurations that are correlated with the test set, as discussed above. 46 It is still interesting to note that the Cormorant equivariant neural network 22 achieves very low energy errors compared to PaiNN, even though it was trained on energy labels only, but the force errors for this model were not reported. On the other hand, the PhysNet 24 graph neural network achieves remarkably low force errors compared to the other models. But similarly to the other equivariant graph neural network models, this comes at the expense of having close to 3 times larger energy errors compared to ACE and FCHL.</p><!><p>The first property to consider beyond the raw energy and force errors is the learning curve, showing how a model's performance improves with additional training data. For kernel models such as FCHL and sGDML, the "kernel basis" grows precisely together with the training data, which is why these methods are universal approximators. Subject to the radial cutoff, the infinite set of Atomic Cluster Expansion basis functions forms a complete basis for invariant functions, so in principle they can also be used to approximate the potential energy surface to arbitrary accuracy. 35 In this case however, the size of the training set and the size of the basis are decoupled. One advantage is that the evaluation cost is independent of training set size, but we have to choose a finite basis set to work with by selecting a maximum body order and the truncation of the 1-particle basis. In order to motivate our choice, we show in fig 2 the force accuracy of ACE as a function of basis set size and the corresponding evaluation time, trained on 1,000 azobenzene configurations (the largest molecule in MD17).</p><p>The timings were obtained using a 2.3 GHz Intel Xeon Gold 5218 CPU. For context, we show the accuracy and evaluation time of the other ML models we trained, each called in their native environment: ACE in julia, GAP via the fortran executable, and sGDML and ANI directly from their respective Python packages. (Note that in the case of ANI considerable speed up could be achieved using a GPU when multiple molecules are evaluated simultaneously, see fig S1, though our single molecule results are in agreement with the timings reported in the original ANI paper 17 ). The solid part of the ACE curve corresponds to 4-body potentials (ν = 3) and we varied only the polynomial degrees, whereas for the last point (dashed), we increased the body order to 5, because the 4-body part of the curve showed saturating accuracy. Increasing the body order further is likely to bring the error down even more, however, the cost of evaluation would also grow unacceptably if all basis functions for the given body and polynomial degree are retained. In the future, effective sparsification strategies need to be developed that would allow the inclusion of some high body order basis functions without the concomitant very large increase of the overall basis set size. For the purposes of the present paper, for each molecule in MD17 we selected a basis set size such that the evaluation cost was roughly comparable with the other ML models. (Note however that in a real ML force field application, one might very well choose a much smaller basis, e.g. 10K, to take advantage of the sub-millisecond evaluation times.)</p><p>In fig 3 we show the learning curves for linear ACE and sGDML (the best models we trained from Table 2) and compare to the literature results of FCHL. 46 The low body order linear ACE is equal or better than the other manybody kernel models in the low data limit, but with additional training data the kernel models overtake ACE in several cases. The latter also saturates, showing the limitations of the relatively low body order model. The learning curves for the forces are given in fig S2, and show a broadly similar trend, with less pronounced saturation for ACE. In the case of ACE the number of basis functions is shown in parentheses. The classical force field has a timing of about 1 µs, which would not fit on this scale. For the ANI model we show both the CPU and GPU timings.</p><!><p>The normal modes and their corresponding vibrational frequencies characterize the potential energy surface near equilibrium. This is interesting in the context of the MD17 models because their training set contains geometries sampled at 500 K which means they are, in general, far from the equilibrium geometry. The ability of the models to describe the minima of the PES, even if it is not in the training set, is particularly important when considering larger systems with potentially many local minima, where finding all the different local minima at the target level of theory can be infeasible.</p><p>To test how well the different models infer the normal modes we took the DFT optimized geometry of each of the 10 molecules and rerelaxed them with the force field models. At the force field minima we carried out a vibrational analysis to find the normal modes and their corresponding vibrational frequencies.</p><p>Fig 4 shows the errors in the predicted normal mode vibrational frequencies for each of the 10 MD17 molecules. The ACE model achieves the lowest error for all 10 molecules, surprisingly even for those for which sGDML has lower errors based on the 500 K MD test set of Table 2. For example, for toluene sGDML has both lower energy and force errors, but at the same time the ACE model has significantly lower errors in predicting the vibrational frequencies, achieving a MAE of 1.0 cm -1 compared to sGDML with an error of 1.4 cm -1 . Observing the individual molecules in Fig 4 it is notable that the ACE model has the lowest fluctuation in the errors of the normal modes, achieving nearly uniform accuracy across the entire spectrum. The case of benzene also shows the limitations of characterizing the models by the force MAE alone. The linear ACE model has only slightly lower force MAE than sGDML (0.5 meV/ Å compared to 0.8 meV/ Å) but the normal mode frequency prediction is more than 3 times more accurate: 0.2 cm -1 compared to 0.7 cm -1 . The linear ACE model has very low errors for all normal modes, whereas sGDML has much higher errors for the high frequency modes.</p><p>Similarly, in the case of aspirin, even though the ANI model has lower MAE on the test set both for energies and forces than the GAP model, its vibrational frequency error is significantly larger than those of GAP (8.3 cm -1 compared to 6.4 cm -1 ). We also compared the models to the accuracy of a classical force field. The normal mode frequency errors of the empirical FF are about 10 times higher than the errors of the ML force fields. These errors do not fit on the scale of</p><!><p>When building a new force field for a molecule, beyond high accuracy, we also need robustness, by which we mean that there should not be areas of accessible configuration space where the model predictions are unphysical or nonsensical. Sometimes called "holes" in the potential energy surface, these can be remedied by regularization 33 or by iterative fitting 39 and additional data. 48 In the context of the MD17 benchmark, with its fixed training set, we test the robustness of the models we fitted by run- Where a point is missing, the model hit a hole in the potential and the MD run was terminated. This happened most often with the GAP model, indicating that this potential was the least regular. The linear ACE and ANI models can also be prone to hitting holes in the potential at the highest temperatures. Of all models sGDML was the most stable, it al-ways kept the molecule intact even at 950 K for the duration of the simulations. Such extreme stability is not necessarily chemically realistic (see the next section on extrapolation to bond breaking).</p><p>Looking at the increase in errors with temperature for the different models we can see that the linear ACE often keeps the errors low with a small slope whereas the other models show a clearer increase as the temperature increases. This can be best observed for ethanol, malonaldehyde and uracil. It is notable that the model that works best at lower temperatures (in the training regime) also works best at higher temperatures confirming that the models are able to smoothly extrapolate away from the training data. Furthermore, we can see a good agreement of the test set force MAE in Table 2 with the force MAEs estimated from the models' own trajectories. This hints that the models explore similar regions of the configuration space as the original ab initio trajectories.</p><!><p>To test the extrapolation properties of the different models further we looked at two tests probing the torsional profile of azobenzene We carried out these tests with several different versions of the linear ACE models differing in the definition of their 1-body terms, because we expect this choice to make a significant difference in how chemically reasonable the fitted models are far from the training set. We denote the ACE models fitted using force data only by ACE F. This has the lowest force error on the test set (comparison shown in Table S3). For the other two ACE models, energies were also included in the training. They differ in the 1-body term only, the model using average per-atom training set energy is denoted as ACE AVG, whereas the model using the isolated atom energies as the 1-body term is de-noted ACE E0. The third option is the natural choice, as this ensures that if all atoms are separated from each other the predicted energy will correctly correspond to the sum of the isolated atom energies. Fig 6(a) shows the torsional energy profile of the azobenzene molecule. The ACE E0 model with the isolated atom 1-body term is able to extrapolate furthest, somewhat overestimating the energy, while the ANI and sGDML models also extrapolate smoothly, but slightly underestimate the energy. The linear ACE model with the average energy 1-body term and the GAP model fail to extrapolate and predict a completely nonphysical drop in energy for smaller values of the dihedral angle. 2. geometry of ethanol. The only force field that shows qualitative agreement with DFT is the ACE E0 model. (Note that we do not expect any of the fitted models to quantitatively reproduce the DFT energy profile, even when the isolated H atom is described correctly by design, because the C 2 H 5 O • radical is not.) We attribute this success to the explicit body ordered nature of the linear ACE model, including using the isolated atom as the 1-body term, and careful regularization -as was the case in a similar test for other polynomial models. 26 Fig 6 (c) shows a detailed comparison of the different ACE models together with their test set MAE value. This shows that having the lowest possible test set error does not coincide with the most physically reasonable model, and using stronger regularization can lead to much smoother extrapolation. The more strongly regularized ACE models with relatively higher force error are still significantly more accurate than sGDML, ANI, GAP or the classical force field.</p><p>Interestingly, having the isolated atom as the 1-body term is not sufficient for good extrap-olation. This is shown by the two different GAP models in fig 6(b), which show essentially no difference to the extrapolation, presumably due to the very poor description of the radical. GAP is not an explicitly body ordered model.</p><!><p>Apart from sGDML, whose descriptor is tied to a given molecule with fixed topology, the models under consideration can all be fitted to multiple molecules simultaneously. Therefore, having evaluated their capacity to approximate individual potential energy surfaces one by one, it is interesting to see how the they cope with describing all of the rMD17 data set pooled together.</p><p>Table 4 shows the energy and force errors for the combined fit with linear ACE, GAP, ANI and the empirical force field. GAP and ANI errors only go up by around 30%, reflecting the fact that these are very flexible functional forms. The ANI model (which is pre-trained by starting from ANI-2x neural network weights) is now distinctly better than GAP. The empirical force field error increases by even less. In this case that is due to the use of atom-types, which help to separate the energy contribution of different functional groups. The increase in the error is largest for ACE, about a factor of two, although for most molecules it is still the combined ACE model that has the lowest error amongst these models.</p><p>In addition, we also show the performance of the original unmodified ANI-2x model (its energies and forces were tested against values recomputed with exactly the same electronic structure method and parameters that were used in its fitting 49 ). Its energies and forces are better than those of the empirical force fields by factors of around 2-3 and 5, respectively. (The exception is azobenene, for which its energies are worse). The difference between ANI-2x and the re-trained ANI is about a factor of 2-4 for energies (the average over all the molecules is at the high end) and a factor of two for forces.</p><p>The other commonly used benchmark data set for machine learning based molecular force fields that contains multiple molecules is ISO17. 23 The full data set contains 5000-step ab initio molecular dynamics simulation trajectories of 129 molecules, all with the same chemical formula C 7 H 10 O 2 . The standard task is to train a force field using a randomly selected 4000 configurations of 103 molecules (so about 400K configurations altogether, although these are highly correlated) and evaluate it on the remaining 1000 structures of the trajectory ("known molecules") and on the full trajectories of the "unknown molecules". We note that when all 400K training configurations are used, the conformations of "known molecules" that are usually reported as a test set are very close to the training set, at most 1 or 2 MD steps away on the trajectory from the actual training set, so the error measured on these is essentially the same as the training error.</p><p>We trained a linear ACE model on only a total of 5,000 configurations and a GAP model on only a total of 10,000 configurations sampled uniformly from the training set and evaluated them on both the known and unknown molecules. The results in Table 5 show that the linear ACE model performs significantly better than GAP, achieving errors in the same ballpark as the other methods for the unknown molecules, but using orders of magnitudes less training data. In particular, the ACE model matches the energy error of the state of the art GM-sNN 20 on the unknown molecules, demonstrating its excellent extrapolation capabilities. For all the neural network models, the error on known molecules is quite a bit lower than that for the unknown molecules, which we consider to be a sign of overfitting. For ACE and GAP, the error is still lower but by a much smaller factor, helped by the explicit regularization. Tellingly, the most similar ratio is for GM-sNN, which is a shallow neural network.</p><!><p>Finally, noting that all the MD17 molecules are rather rigid, our last test is to assess the capabilities of the different force field models on a more challenging system that has relevance for medicinal chemistry applications. We created a new benchmark data set for the flexible drug-like molecule 3-(benzyloxy)pyridin-2amine (3BPA). 50 Though smaller than typical drug-like molecules, with a molecular weight of 200, this molecule has three consecutive rotatable bonds, as shown in fig 7. This leads to a complex dihedral potential energy surface with many local minima, which can be challenging to approximate using classical or ML force fields. 51</p><!><p>To prepare a suitable training data set we started by creating a grid of the three dihedral angles (α, β and γ) removing only the configurations with atom overlap. From each of the configurations corresponding to the grid points, we started short (0.5 ps) MD simulations using the ANI-1x force field. 28 This time scale is sufficient to perturb the structures towards lower potential energies, but is not enough to significantly equilibrate them. In this way we obtained a set of 7000 configurations as shown in the left panel of Fig 7 . From the distribution of dihedral angles, five different densely populated pockets were identified in the space of the three dihedral angles. One random configuration was selected from each of the 5 pockets and a long 25 ps MD simulation was performed at three different temperatures (300 K, 600 K, 1200 K) using the Langevin thermostat and 1 fs timestep. We sampled 460 configurations from each of the trajectories starting after a delay of 2 ps. In this way the final data set of 2300 configurations was obtained. The configurations were re-evaluated using ORCA 52 at the DFT level of theory using the ωB97X exchange correlation functional 53 and the 6-31G(d) basis set.</p><p>(These settings are similar to that used in the creation of the ANI-1x data set 49 ). From the total data set we created two training sets, one using 500 randomly selected geometries from the 300 K set, and another one, labelled "mixed-T", selecting 133 random configurations from each of the trajectories at the three temperatures. The rest of the data in each case makes up the three test sets, each corresponding to a different temperature. The right hand panels of Fig 7 show the distribution of dihedral angles in the test sets. At 300 K the separate pockets of the configuration space are sampled mostly individually, whereas at 1200 K the distribution widens significantly, and the sampling connects the pockets across multiple barriers with ease.</p><!><p>We trained linear ACE, sGDML, ANI and GAP force fields, and re-parametrized the bonded terms of a classical force field (FF), using the 300 K and the mixed-T training sets. Table 6 shows the energy and force RMSEs of the different models alongside the general purpose ANI-2x force field errors on the same configurations. Just as before, the weights of the re-trained ANI model were initialized form the ANI-2x weights, giving it a considerable advantage over the other models, especially because the DFT functional and basis set that we use are the same as that of the underlying DFT method of the ANI-2x model.</p><p>For the case of training on the 300 K configurations the linear ACE and sGDML models are able to achieve very low errors when tested at the same temperature, but the ACE model shows significantly better extrapolation properties to the configurations sampled at higher temperatures. The model extrapolating most accurately to 1200 K is the re-trained ANI force field, but the linear ACE is not far behind, especially considering how poor the extrapolation of the other models are. Just as for the smaller molecules, the fitted empirical force field shows much higher errors, about a factor of 2-4 for energies and a factor of 4 for forces compared with the ANI-2x force field. Only at 1200 K does ANI-2x become competitive with the ACE trained at 300 K.</p><p>Training on the mixed-T training set leads to a significant drop in the errors at the higher temperature test sets for all ML models, but not for the empirical force field. The linear ACE model achieves the lowest error in every case, showing approximately 40% decrease in the error for the high temperature test set. The other ML models improve also, by even bigger factors (because their extrapolation power was less). The gains over the general ANI-2x force field, nearly a factor of two in energies for all three test sets, show the potential scope for parametrizing such custom force fields in medicinal chemistry applications. The errors in the empirical force field are mostly unchanged, quantifying the limitations of the fixed functional form when describing the anharmonic high energy parts of the potential energy surface.</p><p>To look beyond the energy and force RMSE, we performed a constrained geometry optimization using the different force field models and DFT to map out the dihedral potential energy surface of the molecule. The complex energy landscape is visualized in S8. The energy landscape of the empirical force field has most of the features of the DFT landscape and is even correctly predicting the position of the lowest energy minimum in the β = 120 • plane. Some of the potential energies on this plane are clearly too high however. On the other hand the landscape of the GAP model is quite irregular, some of the most basic features are either missing or blurred together. The ANI landscape is also quite irregular, somewhat less than GAP, and some of the high energy peaks are too high and too broad. This is an example where the fixed functional form of the classical force field gives better extrapolation behavior to parts of the configuration space where there is little training data. The RMSE results clearly do not give a full, perhaps not even a very useful distinction between these models. The ACE and sGDML models reproduce the landscape much more closely (and indeed these are the models with the lowest RMSE as well). Some differences include the sGDML getting the position of the lowest energy minimum wrong and ACE having too high a peak at α = 230 • , γ = 150 • .</p><!><p>In this paper we have demonstrated how the Atomic Cluster Expansion framework can be used as linear models of molecular force fields. We showed that body ordered linear models built using the ACE basis are competitive with the state of the art short range ML models on a variety of standard tests. Furthermore we carried out a number of "beyond RMSE" tests to compare the ML approaches, and to study the smoothness and extrapolation properties of the fitted force fields: vibrational frequencies, forcefield driven molecular dynamics and extrapolation to bond-breaking.</p><p>We also introduced a data set on a flexible drug-like molecule, with the idea that testing the performance on it is more predictive of the quality of the model for medicinal chemistry applications. The linear ACE model was signifi-cantly smoother than other transferable models and was able to extrapolate to higher potential energy regions than all other models.</p><p>We showed that the ACE framework allows us to build accurate force fields with very low evaluation cost. Together with competing approaches that are in the recent literature and in our comparison tables, the prospects are good for being able to carry out large scale simulations of systems ranging from biomolecular applications to other complex molecular systems such as polymers with electronic structure accuracy in the near future. A number of bottlenecks remain for ACE, which include the steep increase in the number of basis functions as new chemical elements are added to the model. This can be tackled via sparsification strategies, which is the focus of our future work. Furthermore the inclusion of long range electrostatics and charge transfer are essential for the simulation of biomolecular systems and an integration of these into the ACE framework is also underway. Currently ACE is implemented in the Julia language, but can readily be called from Python via the Atomic Simulation Environment (ASE). The fitted models can also be evaluated via LAMMPS. models, a comparison of different 1-body energy ACE models and energy landscapes for 3BPA with β = 150 • and β = 180 • are shown in the Supporting Information. The 3BPA dataset is available as a .zip file free of charge via the Internet at http://pubs.acs.org</p>

ChemRxiv

Pd/Xiang-Phos-catalyzed enantioselective intermolecular carboheterofunctionalization under mild conditions

A mild and practical Pd/Xiang-Phos-catalyzed enantioselective intermolecular carboheterofunctionalization reaction of 2,3-dihydrofurans is developed, leading to various optically active fused furoindolines and tetrahydrofurobenzofurans. The key to this transformation is employing two newly modified N-Me-Xiang-Phos ligands ((S, R S )-N-Me-X4/X5) as chiral ligands under mild conditions. Moreover, this synthetic methodology can be efficiently applied to a variety of complex polysubstituted heterocycles with high chemo-, regio-, and enantio-selectivities via introducing diverse substituents on furan rings, which were hard to access by other routes.

pd/xiang-phos-catalyzed_enantioselective_intermolecular_carboheterofunctionalization_under_mild_cond

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

<p>Benzofused heterocycles are ubiquitous moieties in natural products, pharmaceuticals, dyes and herbicides, in which furoindolines and tetrahydrofurobenzofurans are prevalent as key core structures (Fig. 1). 1 These derivatives have shown signicant anticancer, antimalarial and antimicrobial activities, as well as antioxidant properties, for instance, Makomotindoline and Aspidophylline A have shown a distinct effect on mammalian cells. 2,3 The rst enantioselective total synthesis of Aspidophylline A was described by Garg via a reductive interrupted Fischer indolization. 4 You and co-workers developed a copper-catalyzed intermolecular dearomative cascade reaction of indoles, which also provided a powerful synthetic method for the construction of furoindolines. 5 E. J. Corey achieved a short, asymmetric total synthesis of Aatoxin B 2 via an aromative cascade reaction. 6 Despite these seminal reports, hetero-annulation of alkenes developed by Catellani and Larock has become a classic and useful strategy for the construction of various heterocycles from readily available starting materials. 7 Although various methods have been developed to construct these two skeletons, it still remains a considerable challenge to extend the substrate scope of asymmetric variants, particularly those that enable access to poly-substituted benzofused heterocycles.</p><p>Over the past two decades, palladium-catalyzed carboheterofunctionalization of alkenes has been proved to be a reliable and efficient method for the synthesis of a variety of poly-cyclic heterocycles. 8 The majority of these reactions proceeded through a crucial hetero-palladation of alkenes with aryl halides along with N-or O-nucleophiles. 9,10 However, the development of an enantioselective version, especially under mild conditions, poses a considerable challenge due to the lack of any suitable robust chiral catalyst. Recently, Mazet and coworkers reported the rst asymmetric Pd-catalyzed syn-carboe-therication and syn-carboamination of 2,3-dihydrofurans (2,3dhfs) at 110 C by utilizing two different chiral ligands (Scheme 1a). 11 Inspired by the good performance of our chiral sulnamide phosphine (Sadphos) ligands in the asymmetric construction of C-C and C-X bonds, 12 we wondered whether Sadphos could realize the highly enantioselective carboether-ication and carboamination of 2,3-dhfs under mild conditions and also address the low enantioselectivity issue of the carboamination reaction. Herein, we report a highly chemo-, regio-, and enantioselective palladium-catalyzed carboheterofunctionalization of 2,3-dhfs employing two newly modied Xiang-Phos ligands as chiral ligands, which can give direct access to enantioenriched poly-substituted functionalized furoindolines and tetrahydrofurobenzofurans in moderate to high yields with high enantio-selectivities at a reduced reaction temperature (Scheme 1b).</p><!><p>With the use of our developed chiral sulnamide phosphine ligands as the chiral ligands, [13][14][15][16] the carboamination reaction of 2-bromoaniline derivative 1a and 2,3-dhf 2a was investigated. It was found that Ming-Phos M1, PC-Phos PC1, Xu-Phos Xu1 and Xiang-Phos X1 did not efficiently deliver the desired product. As observed in our previous work, the N-H bond in ligands could greatly affect the reactivity as well as enantioselectivity in some cases. 16 Several representative N-Me sulnamide phosphine ligands lacking the hydrogen-bonding site were further investigated. We were pleased to nd that the desired product 3aa could be obtained in 81% yield with a 48% ee value in the presence of (S, R S )-N-Me-X1 and CH 3 ONa, albeit with a small amount of the Heck byproduct 4aa. Other chiral ligands such as Ming-Phos N-Me-M1 and Xu-Phos N-Me-Xu2 showed less efficiency comparatively, leading to a lower yield and enantioselectivity along with a poor regioselectivity (Fig. 2).</p><p>Under the conditions of Xiang-Phos (S, R S )-N-Me-X1 utilized as the chiral ligand, NaOPh appeared to be the optimal base, affording 3aa in 78% ee albeit with a 2 : 1 regioselectivity ratio (r.r.) (Table 1, entries 1-5). Solvent screening showed that 1,2-DCE gave a better yield (81%) and r.r. (9 : 1) with 87% ee (Table 1, entries 6-9). The result obtained employing other chiral N-Me-Xiang-Phos ligands indicated that the introduction of steric hindrance on the phenyl backbone and enhancement of the electron-donating character were benecial for the catalytic enantio-selectivity and regioselectivity (Table 1, entries 10-14). Employing the newly modied N-Me-X5 as the chiral ligand, a series of Pd precursors were then screened, showing that a vemembered cyclic palladium precatalyst was competent for the carboamination cyclization (Table 1, entries 15-19). Comparably better outcomes were obtained under mild conditions by lowering the temperature to 20 C (Table 1, entries 20-23). Inspired by previous ndings that the addition of a trace amount of water may help to increase the reactivity and the stereoselectivity, 17 the water effect was studied, and indeed, we found that the addition of 2 equivalents of water to the system led to a signicantly improved reactivity and reproducible enantioselectivity. In terms of the reactivity and enantio-and regio-selectivity, the reaction conditions illustrated in entry 22 were utilized in the following substrate scope investigations (please see Table S1 in the ESI for details †).</p><p>Various substituted N-(2-bromophenyl)-p-tolylsulfonamide derivatives 1a-r were subsequently employed as coupling partners in the enantioselective intermolecular carboamination of 2,3-dhf (Scheme 2). Remarkably, a wide range of 2-Br-anilines bearing electronically diverse substituents at C4 and C5 such as halogens, -Me, -OMe, -CF 3 , -OCF 3 , and -CO 2 Me reacted smoothly and furnished the corresponding furoindolines 3aa-3la in good yields (up to 97%) and ee's (up to 96%). Further substrate scope investigations demonstrated that the electronic properties of substituents at C3 and C6 did affect the yields and enantioselectivities, and produced products 3na and 3oa in lower yields comparatively. Notably, the disubstituents on phenyl rings were also applicable in this cyclization reaction, affording 3pa in 84% yield with 89% ee, as well as 3qa containing a heterocycle in 87% yield and 95% ee. When N-(2bromophenyl)-benzenesulfonamide was explored as an alternative to 1a, to our delight, the furoindoline product 3ra was formed in a nearly quantitative yield (92%) with high enantioselectivity (95%). We also replaced the protective groups on the nitrogen atom with Ms and Ns, but only trace products could be detected by NMR. Other nitrogen protecting groups on aniline, such as Boc, Cbz and Bz, were not tolerated, and in these cases the desired product was not observed. To our delight, a gram-scale reaction was conducted to further demonstrate the potential synthetic utility of this methodology, delivering 1.2 g of 3aa in 77% yield and 94% ee with 2.5 mol% palladium catalyst at 20 C for 6 days. The absolute conguration of this series of products was conrmed by the X-ray diffraction analysis of 3aa. 18 Aer a quick survey of the construction of tetrahydrofurobenzofuran (Table 2, please see Table S2 in the ESI for details †), the optimal reaction conditions were identied (Table 2, entry 7). A series of substituted 2-bromophenol derivatives 5a-h were subsequently employed as coupling partners in the enantio-selective intermolecular carboetherication of 2,3-dhf with the use of N-Me-X4 under mild conditions (Scheme 3). Of note, an exciting enantioselective cyclization was realized when substrates containing diverse substituents at C4, C5 and C6 with different electronic properties (-F, -Me, and -OMe) participated smoothly, delivering products in 90% to 99% ee's. It is a pity that only a trace amount of product was observed when substituents were present at C3 of the phenyl ring.</p><p>To delve into the construction of poly-substituted fused furoindolines and tetrahydrofurobenzofurans, a variety of dihydrofuran derivatives 2b-2e, which could be readily prepared by a classic Heck reaction, were subjected to carbofunctionalization. In these two cases, variation of the electronic parameters of the phenyl groups on dhf rings had a slight inuence on the yields and enantiocontrol of the carbohetero-functionalizations (Scheme 4, 3ab/6ab, 3ac/6ac). Considering the O-and Ncontaining heterocycle substituents on dhf rings, the phenyl groups could be swapped for the benzofuran and quinoline substituents (3ad/6ad, 6ae), still maintaining the efficiency of the transformations. 5-Methyl-2,3-dhf was next examined to investigate the formation of an all-carbon quaternary stereocenter. The carboetherication reaction took place smoothly a Unless otherwise specied, all reactions were carried out with 1a (0.2 mmol), 2a (0.8 mmol, 4 eq.), a [Pd] source (0.01 mmol, 5 mol%), N-Me-Xiangphos (0.024 mmol, 12 mol%), base (0.8 mmol, 4 eq.), and H 2 O (7.2 mL, 2 eq.) in a solvent (1 mL, 0.2 M). when increasing the loading of the palladium precatalyst and chiral ligand at a higher temperature (6af). However, only the corresponding debromination product was detected in the carboamination reaction system. The absolute conguration of the poly-substituted carboamination products was conrmed by X-ray diffraction analysis of 3ac, 18 while the absolute conguration of poly-substituted carbo-etherication products was assigned by comparing the rotational value and 1 H, 1 H-NOESY-NMR spectrum (please see the ESI for details †) of 6ab between our work and Mazet's work. 11 The substituted aromatic ring on the tetrahydrofuran ring was in the (S)-conguration, and is a diastereomer of the corresponding product in Mazet's work.</p><p>Based on Mazet's studies, as well as our observations on Pd/ Sadphos catalytic systems, the chirality-induction models of carbo-amination and -etherication were proposed according to the absolute conguration of products 3aa and 6aa, as shown in Scheme 5. We supposed that the reaction was initiated by a classic oxidative addition, which would be followed by ligand exchange, deprotonation and coordination of 2,3-dhf. The key step of asymmetric hetero-palladation was hypothesized to occur to ultimately construct optically active benzofused heterocycles with high regio-and enantio-selectivities. a Unless otherwise specied, all reactions were carried out with 5a (0.2 mmol), 2a (1 mmol, 5 eq.), a [Pd] source (0.005 mmol, 2.5 mol%), N-Me-Xiang-Phos (0.01 mmol, 5 mol%), base (0.4 mmol, 2 eq.), and H 2 O (3.6 mL, 1 eq.) in a solvent (1 mL, 0.2 M). b Yield of isolated product. c Determined by chiral HPLC. d Pd 2 (dba) 3 was added to 5 mol%, and L3 was added to 10 mol%.</p><p>Scheme 2 Scope of carboamination of 2,3-dhf with 2-bromoanilines.</p><!><p>In summary, we have demonstrated an efficient Pd-catalyzed enantioselective intermolecular carboheterofunctionalization of 2,3-dihydrofurans for the synthesis of poly-substituted benzofused heterocycles. The new N-Me-Xiang-Phos X4/X5 ligands are responsible for the high reactivity and enantioselectivity. This strategy could be conducted under mild conditions and easily extended to a wide range of chiral fused furoindolines and tetrahydrofurobenzofurans with high chemo-, regio-, and enantio-selectivities, which made the method extremely attractive. In addition, a gram-scale reaction of the representative product 3aa was investigated to further demonstrate the potential synthetic applications of this method. Further applications of Sadphos in other transition-metal-catalyzed reactions are underway in our group and will be reported in due course.</p><!><p>There are no conicts to declare.</p>

Royal Society of Chemistry (RSC)

Considerations of Protein Subpockets in Fragment-Based Drug Design

While the fragment-based drug design approach continues to gain importance, gaps in the tools and methods available in the identification and accurate utilization of protein subpockets have limited the scope. The importance of these features of small molecule\xe2\x80\x93protein recognition is highlighted with several examples. A generalized solution for the identification of subpockets and corresponding chemical fragments remains elusive, but there are numerous advancements in methods that can be used in combination to address subpockets. Finally, additional examples of approaches that consider the relative importance of small-molecule co-dependence of protein conformations are highlighted to emphasize an increased significance of subpockets, especially at protein interfaces.

considerations_of_protein_subpockets_in_fragment-based_drug_design

<!>Concept of shared subpockets<!>Fragment diversity and protein interactions<!>Identifying Subpockets<!>Using protein topology to search for binding sites<!>Ligand interaction approaches to compare binding sites<!>Shared Subpockets in ATP-Binding Sites and Relevance to Ligand Selectivity<!>Consideration of Protein Interfaces and Cooperative Interactions<!>Summary<!>Declaration of Interest<!>

<p>Fragment-based drug design (FBDD) is an important strategy in both industry and academia for the discovery of novel ligands and aids the progression toward lead compounds (1). It is based on the idea that through the use of low molecular weight chemical fragments, which typically only bind weakly to their intended target, higher affinity lead ligands can be obtained by combining or 'growing' these small compounds into larger drug-like molecules. Due to the high levels of diversity between biological targets, incorporating FBDD as a high-throughput screening tool can have significant advantages over traditional higher molecular weight chemical libraries (2). The approach considers factors such as compound availability, ease of synthesis, large chemical space, and limits on steric 'bulki-ness,' which may otherwise preclude many higher molecular weight ligands from recognizing non-covalent enthalpically driven affinity factors (e.g. hydrogen bonding, etc.) at a target-binding site. Indeed, FBDD can prove to be robust for rational fragment identification in the absence of 3D structural data (3). However, FBDD still faces several challenges, such as the general lack of accountability for ligand specificity or selectivity (4,5), and the fact that key interactions and geometry of an original fragment hit may need to be changed when incorporated into a lead compound (6). Furthermore, the role of ligand-dependent receptor conformations has been largely untested.</p><p>A number of reviews have discussed recent advances in fragment-based drug design and how these tools can be used to improve the lead design process (4,7–14). Yet, relatively few evaluations have appeared which address the importance of small localized environments in a protein-binding site, and how microenvironments ultimately drive ligand binding and observed effects such as drug non-specificity. The field of drug discovery has long struggled with the accurate prediction of a drug's cross-pharmaco-logic profile (15–18) and side-effects. However, recent advances offer avenues toward understanding the significance of protein 'subpockets'—the physical, chemical, and geometric properties surrounding an individual residue. Traditional methods of computational analysis to find binding sites have sought to identify the similarity between proteins based on inherent sequence conservation or overall structural similarity. However, the localized chemical environments to be sampled by chemical fragment screens are potentially vast. Further development of generalized methods for the discovery of similar protein-/ligand-binding sites and predicting the interaction profile of molecular fragments remains of high interest (19).</p><p>Recently, multiple methods have been developed that compare proteins based on localized environments within binding sites, or the chemical environment around protein-bound ligands derived from PDB database crystal structures. These methods take one of two general approaches: (i) proteins are compared and binding sites organized based on their structural or chemical similarities, or (ii) data from known ligand–protein interactions are used to identify potentially similar sites in other proteins that could bind equivalent chemical fragments. This review aims to provide an overview of the background and emerging computational approaches that can define, and take into account, the significance of protein subpockets in the design of new chemical probes and pharmaceuticals. Many of these same tools offer methods which can also enhance understanding of drug activity. The relevance of new methodological insights into fragment-based drug discovery and the implications for lead development when considering factors such as non-specificity and side-effects are also considered.</p><!><p>In many modern cases of structure-based drug discovery, structural information for a target in question drives the ligand development process (20,21). The RCSB protein data bank (PDB) (22) currently contains more than 89 000 structures solved by X-ray crystallography, with more than 10 000 others solved through other means (NMR, electron microscopy, etc.). This knowledge database, which continues to expand, is a vital tool for understanding the general structures of target proteins and the topology of target ligand-binding sites. Among 'sibling' members of a protein subfamily, binding sites may have a moderate-to-high level of overall similarity, such as the ATP-binding sites in the large family of protein kinases (23). Although, even in cases where two arbitrarily selected proteins have overall dissimilar ligand-binding sites, they may still bind the same chemical fragments if they share similar or equivalent topo-logical features at the subpocket level (see Figure 1 for an example). This explains why two binding sites with substantially different sequences can bind identical chemical fragments, such as the case with the trifluoromethyl- and sulfonyl-binding subpockets between isoenzymes of the carbonic anhydrase family and cyclooxygenase-2 (24). Convergent evolution in nature is commonly observed to reveal similar enzyme active sites between proteins (25). Likewise, divergent evolution can result in shared ligand-binding sites among proteins that have highly dissimilar overall structures (26). Between proteins even of distant origin, similar ligand-binding motifs containing conserved consecutive residues can be found, indicating that it is quite efficient for nature to reuse localized features to bind similar molecules, even when the functional relationships may not be clear.</p><p>The principle of protein subpockets requires definition at the level of individual amino acids within a protein. Each residue is surrounded in three dimensions by other amino acid residues, water molecules, and/or metal ions. The resultant microenvironments create the recognition features for a particular chemical fragment, albeit with likely moderate ranges of affinities. These localized environments created by the protein generally define what are termed 'subpockets' within a binding cavity. Even within the limits of 20 different standard amino acids and the amide backbone, a substantial number of distinct subpockets with unique 'fingerprints' are possible. These subpockets differ by the relative proximity of specific amino acids to each other and distances between chemical functional groups. The amount of chemical space covered by chemical fragments is vast, but not all of them comprise preferable features for protein binding (27). Therefore, any given chemical fragment is not guaranteed to bind to a protein subpocket across the entire proteome.</p><!><p>For a chemical fragment to bind with favorable free energy, it must have the appropriate overall shape, proper spacing of its chemical functional groups, and generally compatible hydrophobic features to establish meaningful lifetimes within a subpocket. How these characteristics overlap with the physiochemical properties associated with the characteristics of drug-like molecules warrants consideration. In principle, chemical fragments can represent high diversity in shape and chemical features, but only a subset of compounds that could be classified as 'fragments' are useful for the purposes of screening against biological targets. A study by Zuegg and Cooper (28) analyzed more than eight million unique compounds from more than 100 chemical vendors and found that only 400 000 of these passed the fragment-like filter for the analysis of drug- and lead-likeness based upon 'the rule of three' (molecular weight < 300 Da, ClogP < 3, the number of hydrogen bond donors and acceptors < 3, and the number of rotatable bonds < 3) (29). A large number of fragments are represented in this subgroup, but there is an argument for ignoring a majority of commercially available compounds and enhancing practicality in a fragment-based screening platform. Of course, rules often have exceptions, and there is evidence that simply adhering to 'the rule of three' for the design of a fragment library may preclude compounds that would otherwise show up as hits (30).</p><p>To further facilitate the fragment library design process, there are a number of studies that have examined the binding preferences of proteins for chemical fragments (Table 1). By analyzing fragments from lead-like compounds in the PDB (22), Chan et al. (31) found that particular types of fragments are more likely to interact with specific amino acid side chains (Asp, Glu, Arg and His) and engage in hydrogen bonding interactions. Separately, through the use of a developed algorithm, LigFrag-RPM, Wang et al. (32) mapped the interaction profiles of 315 unique fragments—derived from 71 798 different PDB ligands—against 20 naturally occurring amino acids, also identifying the preferences of fragment types for particular amino acids. This map could be used to determine whether a given fragment is in a favorable or unfavorable environment, and potentially guide the lead chemistry process.</p><p>While this may serve as a quick, efficient way to direct the design of ligand topology, the orientations of individual amino acids have to be taken into account. Even when comparing the same binding site between two crystal structures of the same protein, the microenvironments in that site may differ significantly based on residue orientations, and this can substantially bias docking studies. As an example, Cox et al. (33) noted that although co-crystal structures had been obtained of the G-protein-coupled receptor (GPCR), CXC-motif chemokine receptor 4 (CXCR4), bound to the anti-HIV small-molecule IT1t and peptide CVX15 (34), docking of IT1t back into the CXCR4: IT1t crystal structure resulted in large RMSD values compared to experimentally solved structures. Similarly, docking other small-molecule antagonists into the crystal structure gave poses that lacked critical interactions identified by site-directed mutagenesis (35,36). It was observed that the CXCR4:CVX15 (CVX15: cyclic peptide antagonist) crystal structure produced a result that showed agreement between the computational binding pose of another small-molecule CXCR4 inhibitor, AMD11070, and independent site-directed mutagenesis data. When analyzed in more detail, the orientations of residues Gln200 and His203 in the interior of the binding pocket differed substantially between the two co-crystal structures. Although the same amino acids are present at the binding site, changes in side chain orientation can substantially affect predicted binding modes and docking scores in virtual-compound screens, and may result in an unacceptably large false-negative rate, particularly when rigid protein docking approaches are used. Therefore, the most accurate representations of protein subpockets must take into account multiple possible rotameric states of amino acid side chains if binding site microenvironments are to be generalized as chemical spaces represented by single 'fingerprints'. This presents a potential problem for many approaches to binding site comparison that make use of pre-existing ligand-bound data, as the amino acids that make up the microenvironment surrounding a ligand fragment are already conformationally biased in a given crystal structure.</p><p>While particular amino acids are seen to favor certain chemotypes, there exist some types of compounds that show relatively limited specificity and that can negatively affect fragment-based drug screening efforts. These molecules, being related to pan-assay interference compounds, are often widely promiscuous in the number and types of interactions with proteins and serve as artifacts, yielding false signals across many types of assays (5,37–41). There are varied reasons for the promiscuous nature of these compounds (42,43), but they represent a subclass of indirect observations of the potential 'fuzziness' in protein microenvironments. In essence, these compounds own a fragment chemical space that can recognize more general protein features that have recurring characteristics which contribute to the high level of promiscuity. When considered in this context, protein subpockets cannot always be rigidly defined by a size and shape boundary as some chemical functionalities are seen to interact with less well-defined features.</p><!><p>In reality, defining what constitutes a 'subpocket' for fragment screening is not always a clear task. A significant number of topological features on a protein could potentially be classified as being part of a subpocket if one were to focus on the environment of a single amino acid. A common approach for defining pockets is to focus on the proximity of protein atoms or residues to a bound ligand. There have been many studies dealing with the analysis of binding sites, ranging from sequence comparisons, to overall binding site structural similarity and pharmacophore ('fingerprint') searching (8,44–54). Until recently, the majority of the work in the field has looked at the relationship between binding sites in the context of an entire defined pocket. Traditional computational methods have attempted to reduce the representation of protein surfaces in an effort to maximize the speed with which a search query can be performed [for examples, see Jordan et al. (55) and Xie & Bourne (56) ]. It is likely that higher degrees of accuracy in predicting the similarity of binding sites can be obtained by taking into account more in-depth features such as hydrophobicity, aromaticity, and hydrogen-bonding capability at the level of single amino acids. As a result, attention in the field has turned to studying local areas of protein surfaces, with implications for traditional receptor-binding sites as well as protein–protein interaction interfaces. Table 2 highlights many of the currently available programs, methods, or databases used to screen binding sites, and while there is diversity between each of these, most can be generalized into one of two groups: protein topology- and ligand interaction-driven approaches.</p><!><p>A significant number of methods currently exist for comparing protein-binding sites (46). Some of these detect locally conserved residue patterns to define similar topo-logical features between proteins and their probability of containing equivalent interaction sites. These approaches can function independently of known ligand-binding data, as the protein structure drives the query. A web server such as ProBiS (83) is useful for detecting either global or local similarities between proteins and identifying structurally conserved binding sites on proteins of interest. This tool takes into account geometric as well as physicochem-ical properties for aligning the local structure of two proteins; a user can decide to compare the entire surface of a protein, or only a selected surface patch, against a database of more than 35,000 non-redundant structures. The program COFACTOR (64) is able to identify functional sites on a protein as well as predict its biological function by comparing the local and global 3D structure of a protein against the large BioLiP database (95). Similarly, the MAAAGINE web server (76) can be used to define an arrangement of 3-8 amino acid residues that are converted into a search pattern to query the PDB and identify structural motifs with equivalent localized environments among the arrangement of amino acids within a defined tolerance distance (default of 1.5 A). These examples demonstrate larger-scale (whole-protein) methods of comparison that identify conserved features of binding site topologies between proteins.</p><p>While these strategies can prove to be quite effective in certain circumstances, some of these 'globalized' methods may not be able to recognize distant similarities, particularly in cases where the dynamics of a protein allow a ligand to bind at sites that may be very structurally diverse. Recent tools have been developed to approach protein comparisons and binding site identification by taking a 'subpocket-focused' approach. Based on the 3D coordinates of a protein, DoGSiteScorer (66-68) is able to detect potential pockets on the surface and split each of them into subpockets. The program uses properties such as volume, depth, surface, ellipsoid main axes, and site-lining atoms and residues, as well as functional groups present to calculate the predicted pockets. This tool scores each of these pockets for their potential druggability, yielding accuracies of 88% when trained and tested against a dataset of 1069 different structures (67).</p><p>Among some of the other methods (81,96-98) used to compare binding sites are CavBase (59-63) and PrISE (55). CavBase derives 3D descriptors that characterize the surface properties of a binding cavity. The individual amino acids lining a cavity are analyzed to form descriptors of the localized chemical environment. Dummy atoms ('pseu-docenters') are placed on the surface, representing the overall chemical property expressed by the surrounding exposed atoms in that area. Therefore, cavities described by a series of pseudocenters are compared with a database [Relibase (99) ] of cavities from PDB protein structures. Similarity is determined through the matching of 3D property descriptors (pseudocenters) between the different sites to provide a 'pharmacophore-esque' search that takes advantage of localized environments within a protein-binding cavity. Additionally, a new evaluation formalism for entries in Cavbase has recently been reported that improves efficiency of large-scale mining for similar protein-binding pockets (100). Separately, PrISE is able to predict interface residues of protein-protein interactions by defining structural elements consisting of a central amino acid residue and its surrounding residues on the protein surface. PrISE deconstructs the surface of a query protein into its structural elements and compares those to similar elements in a database that contains elucidated structural elements in PDB format. Elements are labeled as interface or non-interface based on the characteristics of the central residue. The labels are then used to predict whether the central residue of each structural element of the query protein is an interface residue. Ultimately, although different, CavBase and PrISE are able to use the chemical environment around a specific amino acid residue to predict the similarity between sites of interaction on different proteins.</p><p>For analyzing similarities between binding sites in terms of subcavities, the PocketFEATURE algorithm (80) takes a modified approach by analyzing 'microenvironments' to assess overall similarity between proteins for prediction of shared ligands. The method does not rely on the sequence or relative pattern of amino acids in a binding site. Instead, Liu & Altman define a 'microenvironment' as a local, spherical region within a protein structure that may encompass amino acid residues that are discontinuous in sequence and structure. A set of 80 physiochemical properties (101) is calculated over six concentric spherical shells, centered on a predefined functional center, with the total radius of the 'microenvironment' being 7.5 A. Two sites are compared using an adjusted Tanimoto coefficient based on the presence of similar properties. Advantageously, this algorithm relies weakly on geometric requirements, instead using biophysical and biochemical measurements to characterize a subregion of a binding pocket. This approach allows for dynamic conformational changes in both the ligand at a binding site and the pocket itself. In this way, the algorithm may most closely align itself with a true subpocket-type search when compared to the aforementioned computational methods.</p><!><p>As opposed to protein structure-driven approaches, ligand-driven approaches utilize known interaction data of ligands (small molecules and chemical fragments alike) with proteins to develop binding models that can be used to compare binding sites. Traditionally, many of these methods analyze the environment surrounding a bound ligand in a protein's binding site, encoding physiochemical and/or geometric features to form a general pharmacophore. These descriptors can then be screened against databases of known structures to detect similar sites. A program such as G-LoSA (74) uses this general strategy to predict or design a ligand for a target protein given known interaction data between ligands and predicted similar binding sites. The region surrounding a bound ligand is compared to localized regions in other proteins, and those containing a large set of aligned residues are identified. Similarly, GIRAF (72,73) utilizes a database of known ligand–protein structures to create an index of the geometric features of the surrounding atomic environment. Similar ligand-binding sites can then be identified and aligned with the query structure, independent of sequence homology, or protein fold. PROLIX (84) uses fingerprints of ligand–protein interaction patterns to rapidly mine large crystal structure databases for similar patterns. The PoSSuM database (82) enables the rapid exploration of similar binding sites between proteins based on physiochemical and geometric similarities, sorted using the neighbor search algorithm SketchSort (102). Each of these methods is useful for finding similar binding sites to a query protein in a small amount of time based on already-known interaction data, but these generally take whole-site approaches as opposed to screening small microenvironments of the binding site itself.</p><p>As with protein topology-driven approaches for binding site analysis, recent methods utilizing ligand-bound data are advancing the analysis of subpockets. The web tool Patch-Surfer (78) represents a binding pocket as a set of small, localized surface patches. Each of these patches is further characterized using 3D Zernike descriptors (103), enabling the identification of corresponding regions in pockets on other proteins, even if the overall pocket shapes are different. This method is able to compare a queried pocket to known ligand-binding pockets and predicts binding ligands for the query. In its most recent version, Patch-Surfer was tested against a large dataset of more than 6000 non-redundant pockets, with 2707 different ligands, where it displayed better predictive performance than many other currently available methods to predict protein pockets. Another program, CrystalDock (65), is able to take a ligand-receptor complex, break the ligand into its constituent molecular parts, use the microenvironments surrounding the ligand fragments, and perform a geometric comparison to identify similar microenvironments in ligand-bound PDB structures. This information can be used to predict chemical fragments that would bind a site of interest. FragFEA-TURE (71) has a similar function in that it also compares the structural environments within a target protein to those in databases to find statistically preferred fragments at a binding site. KRIPO (48) is able to use microenvironment pharmacophore fingerprints to identify similar binding sites between proteins, and potential bioisosteric replacements for queried molecular fragments. Combining a subcavity comparison search with pharmacophoric analysis, Wallach and Lilien developed a method (94) to cluster similar binding site subcavities to predict patterns of binding between proteins that do not share any structural similarity with known systems.</p><p>Taking a modified approach to ligand-driven subpocket analysis, Kalliokoski et al. (90) developed SubCav, a tool for comparing and aligning protein subpockets. In this method, a modified version of the 3D pharmacophore fingerprint descriptor F-SPE-FP-PH3 (69) was used to define pharmacophoric features of all protein atoms within 4.5 A of a bound ligand. These atoms were described according to eight different chemical features. Fingerprints composed of 7680 elements incorporating pharmacophoric features and the three dimensional triangular distances between them were generated for a grouping of subpockets within a binding site. Normalized Tanimoto scores were then used to assess the similarity between two fingerprints. Subpockets were aligned using the methodology described by Kabsch (104) implemented in BioPython (105). An advantage of this method is that it can identify similar fragment-binding sites independent of protein structural or sequence similarity. This focus on local equivalent microenvironments enables more comprehensive predictions of the pharmacological selectivity for on- and off-target binding events for fragments in ligands.</p><!><p>A current challenge for small-molecule lead development is represented by target site similarities within the super-family of human protein kinases. Major efforts for drug discovery targeting these kinases have focused on their ATP-binding sites. Although these proteins have evolved distinct cellular functions and hence substrates, there exists a significant level of similarity at the ATP-binding sites. One of the challenges in developing inhibitors to these proteins is the optimization of selectivity for one or more related kinases (106). A study by Anastassiadis et al. (107) showed that a large proportion of commercially available kinase inhibitors display significant levels of cross-reactivity, with some having the ability to inhibit the catalytic activity by more than 50% in nearly 30% of all tested kinases. The clinical implications for these observations are not always clear, but offers a platform for understanding the role of subpockets in rendering predictable inhibitor kinase cross-reactivity.</p><p>Within kinase ATP-binding sites, there exists a conserved phenylalanine residue of the Asp-Phe-Gly (DFG) motif that is buried in a hydrophobic pocket, which is positioned in a groove between the two 'lobes' of the kinase. This motif is targeted by type I kinase inhibitors (108,109), being generally more promiscuous than type II and III inhibitors, which typically take advantage of an allosteric binding site that becomes available upon a structural change of the DFG motif (108–110). The idea of a protein subpocket finds significant application here given the large similarities in the binding pockets for type I inhibitors. An example can be seen with the JNK3 human kinase, where the residues Glu147, Met149, and Val196 form a microenvironment (Figure 2C) that recognizes shared chemical features between a natural ligand, adenosine monophosphate (AMP) (111), and a dihydroanthrapyrazole-based inhibitor (112), though their structures, overall, are quite different.</p><p>These binding site subpockets are used advantageously when designing mimics of natural ligands as inhibitors, but the similarities in these microenvironments between kinases can also give rise to promiscuous chemical landscapes. Staurosporine is a classical ATP-competitive inhibitor that is notoriously promiscuous (113), with there already being 48 crystal structures available in the PDB that show it bound to a variety of human kinases. With roughly 32% sequence similarity, the TAO2 and serine– threonine kinase 16 (STK16) kinases both bind stau-rosporine with good affinity. The sequence of residues at the ligand-binding site differs between the two proteins, but TAO2 (114) and STK16 (115) both fold in a way that causes chemically similar amino acids to overlap, forming shared subpockets that result in a favorable binding site for the competitive inhibitor (Figure 2A).</p><p>There are additional proteins in this family that display high levels of promiscuity as well. The RET tyrosine kinase has been observed to be significantly inhibited by a variety of competitive inhibitors (107). While many of these ligands have structurally distinct features and vary in size, one conserved binding orientation (116) makes use of an overall hydrophobic subpocket surrounding Val804, as can be seen in Figure 2B. Even though the inhibitors shown in this example are not vastly dissimilar, they do demonstrate that when designing ligands, consideration of the fact that some proteins recognize more generalized chemical features. This insight is especially important when considering kinases as biological targets, but importantly the example is instructive to apply to other protein families as well. Furthermore, documented ligand-kinase associations offer many cases to benchmark tests of methodologies to identify and utilize fragment subpockets for the desired selectivity or cross-reactivity.</p><!><p>In contrast to the types of ligand interaction sites considered so far, protein-protein interaction (PPI) interfaces give fewer clues on a residue-by-residue basis for what is required for a specific binding event. Protein surfaces most often are relatively flat, and the interface at which two proteins engage typically makes up a large surface area (1000-2000 Å2) with multiple contact points (117). Drug discovery efforts, particularly in the pharmaceutical industry, traditionally avoided exploring PPIs as drug targets, considering them 'undruggable' until recent challenges of this classification (118). Within the last decade, research into targeting PPIs as a therapeutic strategy for multiple diseases has continued to accelerate. Previous work has shown that not all of the residues at a PPI interface are critical for the interaction. In the best studied cases to date, there exist regions of 'hot spots' that confer most of the binding energy (119). These 'hot spots' are adept at binding protein or peptide-esque molecules and have been suitable targets for peptides and a number of mimetic compounds (120). However, the practical utility of the hot spot theory requires additional consideration with respect to the design of specific protein–ligand interactions.</p><p>Previous studies have concluded that amino acids such as Ile, Leu, Met, Phe, Trp, Tyr, and Val appear more frequently at PPI interfaces relative to other amino acids, and their average distribution throughout the genome (121,122). Some residues at interface hot spots form 'anchor sites,' where they serve as critical recognition features that drive the binding process (123). In this way, these sites are analogous to the previously discussed concept of a subpocket in that the localized environment and relative orientation of particular amino acids in anchoring sites are important for the recognition of specific chemo-types. The feasibility of targeting anchor sites was demonstrated when developing small-molecule PPI inhibitors (122). The surface of the E3 ubiquitin–protein ligase MDM2 forms a natural binding groove that the Phe19, Trp23, and Leu25 residues of p53 favorably bind (Figure 3A). Even when visualizing the unbound form of MDM2, the pocket where Trp23 of p53 ultimately binds is visible, suggesting that this site is important for the recognition and binding of p53. Phe19 and Leu25, upon binding, appear to induce the formation of hydrophobic pockets on MDM2 that result in favorable binding environments for the two residues. In this case, the recognition features of a subpocket are not necessarily prearranged for a meaningful binding event to occur. Rather, a molecular fragment may induce the formation of a new subpocket with optimal chemical environ ment for that fragment to bind, when it otherwise would not if considered a static system.</p><p>In a similar fashion, proliferating cell nuclear antigen (PCNA) contains a region on its surface that recognizes specific chemical features. PCNA acts as a homotrimeric scaffold protein that binds many different proteins associated with DNA replication and damage repair (124). Many of these proteins contain a conserved eight-member sequence known as the PCNA-interacting protein (PIP box) motif (125). While four of the eight amino acids in this sequence are highly variable, the remaining four residues represent conserved chemical functionalities that are critical for proteins to bind at the PIP box-binding site on PCNA (Figure 3B). The conserved aromatic residues (tyrosine or phenylalanine) in the seventh and eighth positions of the PIP box are part of a short 310 helix that binds in a highly flexible (126) hydrophobic pocket on the surface of PCNA. This site generally recognizes hydrophobic molecular fragments exemplified by the discovery of a small molecule, T2AA, that binds in that pocket and inhibits the interaction of PCNA with other PIP box-containing proteins (127). Upon comparison between the structures of PCNA bound to p21 (PDB ID: 1AXC) (128) and T3 (an analog of T2AA; PDB ID: 3VKX) (127), a subpocket formed by Ile128, Tyr133, Tyr250, Pro234, and Val236 binds an aromatic fragment of both T3 and p21's PIP box tyrosine (Figure 3C). Furthermore, the inhibitor's core iodine atoms induce a structural widening of the binding pocket, exposing a larger hydrophobic surface, resulting in a stably bound planar orientation of the molecule. The ligand dependence of the subpocket within the PIP box-binding region is another example of how the plasticity of protein interfaces must be taken into consideration.</p><p>Because drug-like compounds are often composed of several segmental fragments, any one substructure of a molecule could have affinity for a subpocket fingerprint shared between two or more proteins. Of course, the observable binding affinity of the drug molecule depends on all of its interacting components with a protein; a single interaction within a microenvironment may not be nearly sufficient enough to stabilize a specific bound conformation to elicit a pharmacological effect. However, protein systems do exist that have evolved to accept a variety of chemical structures, whether through flexible binding sites, or weak requirements for a binding event to produce a desired effect. The large family of GPCRs that make up the mammalian olfactory system is one example. Many individual members of this group become activated upon interaction with a plethora of different odorants (129–132) (Figure 3D), implicating the existence of 'fuzzy subcavities' making up interaction sites. Receptor OR1G1, and others like it, thus exhibits flexible recognition of general chemo-types, rather than a specific orientation of atoms.</p><p>The afrementioned examples demonstrate that PPI interfaces containing subpockets are analogous to those found in traditional small ligand-binding sites. However, these characteristics indicate the transient aspects of these sub-pockets and articulate both opportunities and challenges for discovery of selective ligands with useful pharmacological effects. Many cases highlight limitations of structure-based virtual screening, and improved methods to achieve overall accuracy continue to emerge (133–138). Computational approaches to address the issues of ligand and receptor dynamics has been an area of continued advancement (10,133,134,136,139,140). The distinctions are likely that induced-fit pathways and conformational selection play more significant roles in the formation of subpockets and the resulting affinity of a molecule composed of multiple substructures (10,141–143). These mechanistic principles have recently been exemplified in the formation of transient binding sites for protein interface inhibitors where evidence for both local conformational selection and induced-fit pathways have been evaluated (141).</p><p>An understanding of how subpockets are influenced by other microenvironments in proximity is a general characteristic for consideration in FBDD. When a fragment of a small molecule binds in a given subpocket, the binding event can induce structural changes that modulate the features of surrounding regions. This means that in a fragment-based screen, suitable binding in a particular chemical microenvironment may not occur in the absence of nearby transient binding events of other fragments, resulting in false negatives. A recent case study addressed how current FBDD methods are limited in the detection of fragments that exhibit cooperative binding (144). However, the deployment of these methods to provide improved accuracy in fragment-based screens is less than complete.</p><p>Several computational approaches to meet these limitations are emerging and show significant promise for expanding successes. Mahasenan & Li were able to develop an approach that incorporates multiple protein conformers via ligand-directed modeling (145,146). Substantial improvements for sampling and free energy scores in docking studies can be achieved when taking into account multiple fragment ligands simultaneously (147,148). Novel inhibitors of STAT3 and the IL-6/GP130 protein interface were discovered by considering multiple subpockets, and simultaneous ligand screening (149–152). Type II kinase inhibitors were also developed for the MELK kinase by taking into account an induced conformational shift of the protein (153). These examples implicate the importance for ligand-dependent subpockets when using FBDD approaches; they may represent a general process that can allow for cooperative binding into subpockets. The full extent to which this impacts current drug discovery efforts and computational screening methods remains to be fully established, but opens avenues for improvement in the utilization of computational and experimental approaches to discover multiple fragment sites at protein interfaces.</p><!><p>There are an increasing number of computational tools available to complement experimental fragment-based screens and ligand design. Advantages to these approaches are that they enable a potential general solution to the challenges of optimal fragment selectivity and linkages, but these goals remain to be fulfilled. The prediction of protein structure microenvironments and dynamics requires consideration, and methodological advances are emerging. As previously highlighted, the similarities of protein microenvironments between evolutionarily related or convergent proteins could argue for increased chances of off-target effects. To the extent that this may be a concern at protein interfaces still remains largely unknown (118,119). However, the potentially unique combinations of transient subpockets that arise upon the binding of a ligand may ultimately result in higher degrees of selectivity.</p><p>The occurrence of subpockets in unrelated proteins is generally recognized as a potential contribution to drug effects; however, an account of these features when selecting specific fragments in drug development has been traditionally based upon empirical deduction. Experimental and computational approaches that can account for potential cooperative interactions are subjects of considerable importance for structure-based drug design. It is increasingly feasible using computational approaches to consider a classification process to stratify both risk and benefit of new molecular candidates. These tools can also be applied to molecular fragments and their associated libraries. As opposed to promiscuous scaffolds, fragments with high propensity for subpocket interactions can be argued to represent important tools for lead identification when considered in an appropriate knowledge framework. Whether fragment libraries can be qualified on a pro-teome-wide scale remains a future challenge, but the combinations of experimental and computational methods offers increasing promise.</p><!><p>The authors declare no competing interests in the content of this paper. Support is recognized for a Purdue Research Foundation Fellowship (MDB), the Department of Defense Breast Cancer Research Program for Award W81XWH-10-0105 (VJD), and the Purdue Center for Cancer Research via an NIH NCI grant (P30CA023168).</p><!><p>MEDIT-SA [Internet]. Available from: http://medit-pharma.-com/ [cited October 1 2014].</p><p>fragment-based drug design</p><p>G-protein-coupled receptor</p><p>proliferating cell nuclear antigen</p><p>RCSB protein data bank</p><p>PCNA-interacting protein sequence motif</p><p>root-mean-square deviation</p><p>The conformation of the adenine portion of ACP is conserved between two structurally and sequentially diverse proteins, demonstrating a shared subpocket at the binding site (HSP90 N-terminal domain, yellow, PDB: 3t10; chemotaxis protein CheA, blue, PDB: 1i5a).</p><p>Conserved subpockets in protein kinases contribute to their inherent promiscuity. (A) Staurosporine (shown as partially transparent) is known to have inhibitory activity against a plethora of protein kinases from different families (107). Conserved subpockets between the TAO2 MAP3-level kinase (green loop) and serine-threonine kinase 16 (STK16) (blue loop) allow staurosporine to bind with comparable affinities to both proteins (TAO2, PDB ID: 2GCD; STK16, PDB ID: 2BUJ). (B) The promiscuous RET tyrosine kinase (orange surface) contains a binding site subpocket that allows for the binding of overall structurally distinct inhibitors. The red highlighted portion of each inhibitor overlaps in the subpocket, shown with the red circle on the right (green, PDB ID: 2IVU; blue, PDB ID: 2IVV). (C) Within the binding pocket of JNK3 MAP kinase, residues Glu147, Met149, and Val196 form a microenvironment that binds comparable chemical features between a natural ligand, adenosine monophosphate (AMP) (PDB ID: 4KKE), and inhibitors such as the dihydroanthrapyrazole-based antagonist shown on the right (PDB ID: 1PMV).</p><p>Protein interface hotspots contain inducible microenvironments that bind conserved fragments between molecule types. (A) Three residues of p53 (F19, W23 and L25; green sticks) become buried in the surface of MDM2 (blue surface), inducing the formation of hydrophobic subcavities. Residue W23, in particular, likely acts as an anchoring residue, substantially contributing to the binding affinity (PDB ID: 1YCR). (B) Four key residue positions (green sticks) in the highly conserved PIP box sequence are essential for the binding of PIP box-containing proteins (p21 shown as green loop) to PCNA (orange surface). The fifth, seventh, and eighth residues in the sequence bind at a surface pocket on PCNA made up of several hydrophobic microenvironments (PDB ID: 1AXC). (C) Within the hydrophobic PIP box-binding site on PCNA (orange surface), the subpocket defined by the relative orientation of I128, Y133, Y250, P234, and V236 to one another has affinity for an aromatic ring moiety, with a tyrosine residue (green sticks) of p21's PIP box sequence (green loop) anchored in the same location as the tyrosine-analogous fragment of the small-molecule inhibitor T3 (127) (PDB ID: 3VKX, p21 peptide from 1AXC). (D) Some receptors exist that themselves accept a variety of structurally diverse substrates, the best example being the large family of GPCR olfactory receptor proteins. OR1G1, a member of this family, becomes activated upon exposure to numerous diversified odorants (130).</p><p>Examples of chemical fragments that select amino acids interact with (31,32).</p><p>A selected set of programs, databases or general methods for assessing protein similarity and analyzing binding sites for the purposes of predicting new fragments for drug discovery/lead development.</p>

PubMed Author Manuscript

OX4 is an NADPH-dependent Dehydrogenase Catalyzing an Extended Michael Addition Reaction to Form the Six-membered Ring in the Antifungal HSAF

The polycyclic tetramate macrolactam HSAF is an antifungal natural product isolated from Lysobacter enzymogenes. HSAF and analogs have a distinct chemical structure and new mode of antifungal action. The mechanism by which the 5/5/6 tricycle of HSAF is formed from the polyene precursor is not totally clear. Here, we used purified OX4, a homologous enzyme of alcohol dehydrogenase/Zn-binding proteins, to show the enzymatic mechanism for the six-membered ring formation. The results from the deuterium isotope incorporation demonstrated that OX4 selectively transfers the pro-R hydride of NADPH to C21 and one proton from water to C10 of 3-deOH alteramide C (1), resulting in 3-deOH HSAF (2) through a reductive cyclization of the polyene precursor by a mechanism consistent with an extended 1,6-Michael addition reaction. The regioselective incorporation of the NADPH hydride into C21 of 1 is also stereoselective, leading to the 21S configuration of 2. This work represents the first characterization of the activity and selectivity of the enzyme for the six-membered ring formation in a group of distinct antifungal polycyclic tetramate macrolactams.

ox4_is_an_nadph-dependent_dehydrogenase_catalyzing_an_extended_michael_addition_reaction_to_form_the

<p>The antifungal natural product HSAF (Heat-Stable Antifungal Factor) is a polycyclic tetramate macrolactam (PoTeM) that has been isolated from diverse sources.1 The structure and mode of action of HSAF and analogs are distinct from the existing antifungal drugs and fungicides.2–4 PoTeMs share the same basic scaffold but differ by the cyclic systems, which can vary from 5/5, 5/5/6, 5/4/6, 5/6/5, to 5/5/5/8, carrying different stereochemistry.5–8 HSAF's precursors, 3-deOH alteramide C (1) and 3-deOH HSAF (2), contain a 5/5 and 5/5/6 cyclic system respectively (Figure 1). Our previous studies showed that four redox enzymes (OX1–4) are responsible for the cyclic system of HSAF and analogues.5–7 However, mechanistic details for the tricycle formation has not been fully elucidated. In this work, we performed deuterium isotope incorporation and NMR analyses of the labeled products resulted from the OX4-catalyzed reaction, which forms the six-membered ring in 2 from 1 (Figure 1).</p><p>We previously showed that 1 is the precursor of 2, and OX4, a homolog of alcohol dehydrogenase/Zn-binding protein, is responsible for this conversion.7 The conversion of 1 into 2 is a two-electron reduction whereby the two double bonds at C21-C22 and C10-C11 are replaced by a new single bond between C11 and C22, and isomerization of the trans double bond at C8-C9 to cis. The overall reaction is a reductive cyclization with hydrogen incorporation at C10 and C21. In theory, the hydride could attack either C21 (then C10 would be protonated) or C10 (then C21 would be protonated). To find out, we expressed the OX4 gene in E. coli and purified OX4 (Figure S1). The enzyme was incubated with 1, in the presence of NADPH, (R)-[4-2H]NADPH, or (S)-[4-2H]NADPH, and the isotope incorporation in products were analyzed by LC-HRMS (Figure 1). (R)-[4-2H]NADPH was generated from D-[1-2H]glucose by the glucose dehydrogenases (GDH) from Thermoplasma acidophilum ATCC 25905, and (S)-[4-2H]NADPH was generated from D-[1-2H]glucose by the GDH from Bacillus megaterium DSM 2894 (Figure S1).9–11 When OX4 was incubated with 1 in the presence of (S)-[4-2H]NADPH in H2O, the LC-HRMS gave m/z 497.2966 for [M + H]+ of 2, showing that no deuterium was incorporated in the product and OX4 does not use the pro-S hydride of NADPH for the reduction reaction (Figure 1). When OX4 was incubated with 1 in the presence of (R)-[4-2H]NADPH in H2O, the LC-HRMS gave m/z 498.3038 for [M + H]+ for 2a. The increase of one mass unit showed that OX4 specifically uses the pro-R hydride of NADPH for the reduction reaction. When OX4 was incubated with 1 in the presence of NADPH in heavy water (D2O), the LC-HRMS gave m/z 498.2988 for [M + H]+ of 2b. The one mass unit increase in the product supported that water is the proton source for the reduction reaction. Finally, when OX4 was incubated with 1 in the presence of (R)-[4-2H]NADPH in heavy water (D2O), the LC-HRMS gave m/z 499.3170 for [M + H]+ of 2c, demonstrating that the reductive cyclization involves the incorporation of the pro-R hydride of NADPH and one proton from the medium.</p><p>To further test the selectivity of OX4, we carried out the similar set of reactions using 3-deOH Alteramide B (3) as substrate (Figure 1 and Figure S3). Our previous result showed that OX4 could reduce the diene at C21-C22 and C23-C24 of 3, to produce Alteramide D (4) which contains a C22-C23 double bond.7 However, it was not clear which hydride of NADPH was used in the reduction. As shown in Figure S3, the result clearly demonstrated that OX4 incorporates the pro-R hydride of NADPH and one proton from water into 4 in this non-cyclized, reduction reaction.</p><p>Next, we investigated the regioselectivity and stereoselectivity of OX4 toward the hydride and the proton, through NMR analyses of the deuterium-labeled products. To obtain a sufficient amount of the substrate 1, we set out a large scale of solid culture (8 liter LB) of the mutant strain C3△OX147 and obtained 28 mg 1 after a series of purification steps (Supporting Information). We then performed a 40 mL enzymatic preparation of 2a, using OX4, 1 and (R)-[4-2H]NADPH generated in situ from D-[1-2H]glucose by TaGDH. After purification, we obtained approximately 5 mg of the deuterated product 2a from the enzymatic reaction. 2a was analyzed by 1H-NMR, 1H-1H COSY and NOESY (Table S1, Figure S4–S7). Comparison of 1H-1H COSY spectra of 2 and 2a clearly showed that the correlations of H-21b/H-20 and H-21a/H-21b in 2 was not observed in 2a, while the correlations of H-21a/H-20 and H-10a/H-10b were observed in both 2 and 2a (Figure 2). These observations confirmed the deuterium incorporation at C21 in 2a. Thus, OX4 regioselectively transfers the pro-R hydride of NADPH to C21 and the proton of water to C10 during the six-membered ring formation.</p><p>A further comparison of the 1H-NMR spectra showed that the H-8/H-9 coupling constant of 2a (J ~ 11.5) and 1 (J ~ 15.0) was consistent with the 8E configuration in 1 and 8Z configuration in 2a (Figure S4). The result confirmed the trans to cis isomerization of the C8-C9 double bond during the conversion from 1 to 2a. It also supports the reductive cyclization by OX4 goes via an extended Michael addition reaction, in which C8-C9 can transiently exist as a single bond in the intermediate that undergoes a tautomerization to afford 2 (or 2a, 2b, 2c) (Figure 3). Finally, a careful analysis of the NOESY data revealed a correlation between H-21a and H-19, which makes 21S configuration for C21 of 2a (Figure S7)</p><p>Together, our data revealed the regioselectivity and stereoselectivity of the OX4-catalyzed reductive cyclization that forms the six-membered ring of HSAF (Figure 3). The results support a mechanism consistent with a 1,6-Michael addition reaction that not only forms a new ring, but also leads to the unusual isomerization at the C8-C9 double bond, which is a challenging transformation considering the relatively rigid, "closed" macrolactam system of the PoTeM. It should be pointed out that D2O could exchange with an acidic group in the enzyme, which then protonates C10, although we proposed D2O directly protonate C10.</p><p>Zhang et al. first demonstrated the enzymatic mechanism for the IkaC-catalyzed inner five-membered ring in ikarugamycin.9 IkaC and OX4 share a 62.3% sequence identity (Figure S2). Their data showed that IkaC selectively transferred the deuterated pro-R hydride from (R)-[4-2H]NADPH to C15, resulting in a 15R configuration in ikarugamycin. This group also investigated the PtmC-catalyzed six-membered ring formation in pactamides.12 LC-MS data showed that the enzyme was selective for the pro-R hydride of NADPH. However, the regioselectivity and stereoselectivity for the hydride transfer in the PtmC-catalyzed six-membered ring formation were not determined. Here, our data show that OX4 is also selective for the pro-R hydride of NADPH. We further demonstrate that OX4 selectively transfers the pro-R hydride to C21, resulting in a 21S configuration in HSAF during the six-membered ring formation. This work represents the first characterization of both the regioselectivity and stereoselectivity of the enzyme for the six-membered ring formation in a group of distinct antifungal PoTeMs.</p>

PubMed Author Manuscript

Quantifying Carbohydrate-Active Enzyme Activity with Glycoprotein Substrates Using Electrospray Ionization Mass Spectrometry and Center-of-Mass Monitoring

Carbohydrate-active enzymes (CAZymes) play critical roles in diverse physiological and pathophysiological processes and are important for a wide range of biotechnology applications. Kinetic measurements offer insight into the activity and substrate specificity of CAZymes, information that is of fundamental interest and supports diverse applications. However, robust and versatile kinetic assays for monitoring the kinetics of intact glycoprotein and glycolipid substrates are lacking. Here, we introduce a simple but quantitative electrospray ionization mass spectrometry (ESI-MS) method for measuring the kinetics of CAZyme reactions involving glycoprotein substrates. The assay, referred to as center-of-mass (CoM) monitoring (CoMMon), relies on continuous (real-time) monitoring of the CoM of an ensemble of glycoprotein substrates and their corresponding CAZyme products. Notably, there is no requirement for calibration curves, internal standards, labeling, or mass spectrum deconvolution. To demonstrate the reliability of CoMMon, we applied the method to the neuraminidase-catalyzed cleavage of N-acetylneuraminic acid (Neu5Ac) residues from a series of glycoproteins of varying molecular weights and degrees of glycosylation. Reaction progress curves and initial rates determined with CoMMon are in good agreement (initial rates within \xe2\x89\xa45%) with results obtained, simultaneously, using an isotopically labeled Neu5Ac internal standard, which enabled the time-dependent concentration of released Neu5Ac to be precisely measured. To illustrate the applicability of CoMMon to glycosyltransferase reactions, the assay was used to measure the kinetics of sialylation of a series of asialo-glycoproteins by a human sialyltransferase. Finally, we show how combining CoMMon and the competitive universal proxy receptor assay enables the relative reactivity of glycoprotein substrates to be quantitatively established.

quantifying_carbohydrate-active_enzyme_activity_with_glycoprotein_substrates_using_electrospray_ioni

INTRODUCTION<!>Proteins, CAZymes, Glycans, and Other Reagents.<!>CUPRA Substrates.<!>Mass Spectrometry.<!>Implementation of CoMMon.<!>Calculation of CoMt.<!>CUPRA-ZYME.<!>IS Approach.<!>Initial Rates and Relative Initial Rates.<!>HILIC Analysis of N-Glycans.<!>Validation of CoMMon Using NEUC Desialylation of Glycoproteins.<!>Glycoprotein Specificity of NEUC.<!>Application of CoMMon to Large Glycoprotein Substrates.<!>Application of CoMMon to ST6Gal1.<!>CONCLUSIONS

<p>Carbohydrate-active enzymes (CAZymes), a large group of enzymes involved in the synthesis, degradation, and modification of carbohydrates (glycans), play critical roles in diverse physiological and pathophysiological processes.1,2 For example, glycosyltransferases (GTs) and glycosyl hydrolases (glycosidases, GHs), which catalyze the synthesis and cleavage of glycosidic bonds, respectively, regulate protein glycosylation, a post-translational modification that affects protein quality control, cell signaling, and host–pathogen interactions.3,4 CAZymes are also important for a wide range of biotechnology applications, such as protein glycoengineering, which seeks to control the glycosylation of biotherapeutics, vaccines, and diagnostics, as well as in glycomaterials.5,6 Measurements of CAZyme kinetics provide insight into enzyme activity and substrate specificity. Such information is of fundamental interest and supports diverse applications. However, versatile and robust kinetic assays suitable for natural CAZyme substrates, in particular, glycoproteins and glycolipids, are limited. Consequently, the structure–activity relationships of many CAZymes are poorly understood.</p><p>CAZyme activity and substrate specificity have traditionally been assessed using radiochemical, spectroscopic (spectrophotometric- and fluorescence-based), and separation techniques (e.g., reversed-phase, ion-exchange, or thin-layer chromatography).7–10 For example, radiolabeled CMP-14CNeu5Ac and fluorophore-labeled CMP-Neu5Ac-Bodipy have been widely used for the characterization of sialyltransferase (ST) activity.11,12 Although very sensitive, the radiolabeling technique usually requires a separation step and is limited to special safety certified labs. Spectroscopic techniques, such as fluorescence and luminescence, also offer high sensitivity but are not readily amenable to monitoring multiple substrates simultaneously. Moreover, the introduction of fluorophores (or chromophores) into the substrate can alter, in some cases substantially, CAZyme activity.13 Separation techniques (e.g., high-performance liquid chromatography) coupled to mass spectrometry (MS) detection represent label-free approaches. However, such methods generally require quenching of the reaction and, as such, do not allow for continuous monitoring.14,15</p><p>Over the past two decades, the use of electrospray ionization mass spectrometry (ESI-MS) to monitor CAZyme (and other classes of enzymes) reactions has grown significantly. The direct ESI-MS approach has the benefit of being label-free, allows for continuous measurements, and can simultaneously monitor multiple species (e.g., substrate(s), product(s), and intermediates).16–20 However, because the ESI-MS response factors for many CAZyme substrates and their corresponding products differ, correlating the measured substrate and product ion abundances with solution concentrations is challenging and generally requires the use of a calibration curve or an internal standard (IS), usually an isotopically labeled substrate or product.21–24 Appropriate IS for many CAZyme reactions are not available commercially and are difficult to synthesize; when available for purchase, they tend to be expensive. Recently, the competitive universal proxy receptor assay (CUPRA-ZYME), a quantitative ESI-MS method for measurement of CAZyme kinetics, was introduced.25 The assay relies on oligosaccharide substrates labeled with an affinity tag that binds to an appropriate protein receptor called a universal proxy protein, UniPproxy. Because both the substrate and corresponding product bind to the proxy protein, their concentrations can be determined directly from the relative abundances of the bound UniPproxy ions measured by ESI-MS, with no requirement for calibration curves or IS. Moreover, the assay is amenable to monitoring simultaneously multiple substrates (provided that they have distinct molecular weights (MWs)), thereby allowing the relative reactivity of substrates to be quantified. However, this method requires labeled substrates and is not amenable to native glycoproteins or glycolipids, which can provide important insight into CAZyme substrate specificities.</p><p>Despite important advances in direct ESI-MS-based methods for measuring CAZyme reactions involving oligosaccharide substrates, continuous, quantitative monitoring of intact glycoprotein and glycolipid substrates remains problematic. Glycoprotein substrates, in particular, represent a formidable challenge due to their macroheterogeneity (presence or absence of glycans at a specific glycosylation site) and microheterogeneity (different glycan structures at a particular glycosylation site). Glycosylation can occur at multiple sites on the protein, with a range of glycan structures possible at a given site, leading to a complex mixture of structures, many with distinct MWs, which are difficult and, often, impossible to fully resolve, even with the most advanced mass analyzers.26,27 Here, we describe a simple, versatile, and quantitative ESI-MS approach to measure the kinetics of CAZyme reactions involving intact glycoprotein substrates. The assay, referred to as center-of-mass (CoM) monitoring (CoMMon), relies on continuous (real-time) ESI-MS monitoring of the CoM of the ensemble of glycoprotein substrates and their corresponding CAZyme products (Figure 1). Importantly, there is no requirement for calibration curves, IS, labeling, or mass spectrum deconvolution. We validated the assay using a GH reaction. We compared reaction progress curves and initial rates for neuraminidase-catalyzed desialylation of a series of mammalian glycoproteins measured with CoMMon and, simultaneously, using an isotopically labeled IS. To illustrate the applicability of CoMMon to GT reactions, we used the assay to measure human ST-catalyzed sialylation of asialo-glycoproteins. The feasibility of applying CoMMon to large (MW > 500 kDa) and highly heterogeneous glycoprotein substrates was also shown. Finally, we highlight how CoMMon, combined with CUPRA-ZYME, enables the relative reactivity of glycoprotein substrates to be quantitatively established, thereby providing a unique opportunity to reliably study structure—reactivity relationships for CAZyme-catalyzed reactions involving glycoprotein substrates.</p><!><p>Details for the proteins, CAZymes, glycans, and other reagents used in this study and sample preparation are given in the Supporting Information.</p><!><p>CUPRA substrates (CS) employing two different affinity tags (sulfonamide or biotin) were used to perform CUPRA-ZYME.25 Representative structures of the CS (CS3SLS, CS3SLB, and CSLNnTS) are shown in Figure S1. Human carbonic anhydrase (hCA) and monostreptavidin (mSA) were used as UniPproxy for the CS containing sulfonamide and biotin affinity tags, respectively. The synthetic methods used to prepare the CS are described elsewhere;28 a brief description is given in the Supporting Information.</p><!><p>ESI-MS measurements were carried out in positive ion mode using a Q Exactive Orbitrap and ultra high mass range (UHMR) Q Exactive Orbitrap mass spectrometers (Thermo Fisher Scientific, Bremen, Germany), each equipped with a nanoflow ESI (nanoESI) source. CAZyme reactions, which were carried out in the nanoESI tip, were monitored continuously by ESI-MS starting at 3 min reaction time (the minimum time to prepare and load the sample into the tip). Details of the instrumental and experimental procedures used are given in the Supporting Information.</p><!><p>The CoMMon approach to measuring the progress of CAZyme reactions involving an ensemble (mixture of glycoforms) of glycoprotein substrates is based on changes in the time-dependent CoM (CoMt) of the substrates. Specifically, the time-dependent fractional abundance of the substrates (Ft) can be expressed in terms of CoMt as shown in eq 1: (1) Ft=|(CoM0−CoMt)|/I(CoM0−CoM∞)∣ where the CoM0 and CoM∞ represent the initial and final values of CoMt, respectively. The value of CoM0 is established prior to addition of CAZyme; in principle, CoM∞ represents the theoretical end point. However, as discussed in more detail below, in some cases, the CAZyme reaction does not proceed to completion on the timescale of the measurements. In such instances, CoM∞ represents the apparent (measured) end point of the reaction.</p><!><p>In cases where all MW-distinct (corresponding to unique glycan composition) glycoprotein species are fully resolved in the mass spectrum, the CoMt of a glycoprotein sample can be calculated from the arithmetic mean of the MWs of the individual species using eq 2: (2) CoMt=∑ifi,tMWi,t where i corresponds to a specific glycan composition (at reaction time t) with MWi,t and fractional abundance fi,t (which is calculated from the total, charge state (n) normalized relative abundances (Abi) of the corresponding gas-phase ions): (3) fi,t=∑nAbi(MWi,t)n∑i∑nAbi(MWi,t)n </p><p>Uniform ESI-MS response factors for all glycoforms are an implicit assumption made in the calculation of fi,t (and CoMt). Based on the results of ESI-MS protein–ligand binding measurements, this assumption is reasonable in cases where glycoprotein substrates undergo relatively small changes (≤5%) in MW.29,30</p><p>Typically, it is not possible to fully resolve all or even most MW-distinct glycoprotein species, making it challenging to accurately calculate Abi from ESI mass spectra. This limitation can be overcome by calculating CoMt from the total signal corresponding to all glycoprotein species (i.e., substrates, intermediates, and products). If the charge state distributions of all glycoprotein species are similar, CoMt can be approximated as the weighted average mass-to-charge ratio (m/z) of all detectable glycoprotein signal: (4) CoMt=At∑xIntx,t where At, which is calculated from eq 5, represents the m/z-weighted signal intensity within the selected m/z range; x represents the specific m/z value, and Intx,t represents the intensity of the signal at position (m/z)x: (5) At=∑xIntx,t(m/z)x </p><p>However, as the charge state distributions of glycoproteins often change over the course of the CAZyme reaction, CoMt is more reliably calculated as the average of CoMt values for individual glycoprotein charge states (CoMt,n): (6) CoMt,n=At,n∑xIntx,t,n where At,n represents the m/z-weighted signal intensity within the m/z range corresponding to charge state n: (7) At,n=∑xIntx,t,n(m/z)x,n Intx,t,n represents the peak intensity at (m/z)x,n, and x represents the specific m/z value, within the range, at charge state n. After determination of CoMt,n for each charge state, the final CoMt can be found using eq 8: (8) CoMt=∑nCoMt,nh where h is the number of total charge state envelops being considered.</p><p>A modified version of our previously reported SWARM software was used to extract CoM values from the time-resolved mass spectra.31 The original SWARM software was designed for the analysis of mass spectra acquired using mass analyzers for which the intensity of ion signals is independent of the charge state.31 In this work, as described in the Supporting Information, we modified SWARM to enable the analysis of mass spectra acquired using mass analyzers for which ion intensity is proportional to its charge state and to automatically calculate CoMt from time-resolved mass spectra (https://github.com/pkitov/CUPRA-SWARM). An overview of the data analysis procedures is given in Figure S2.</p><!><p>Where indicated, CUPRA-ZYME was performed simultaneously with CoMMon. Complete details of CUPRA-ZYME and its implementation can be found elsewhere.25 A brief overview of the implementation assay and data analysis is given in the Supporting Information.</p><!><p>Where specified, CoMMon was performed in the presence of an IS. Details of the implementation of the IS method and data analysis are given in the Supporting Information.</p><!><p>Details on the procedures used to determine initial rates (ν0) and relative initial rates (νrel) are given in the Supporting Information.</p><!><p>N-Glycans, released from glycoprotein substrates, were labeled (with 2-AB) and analyzed by hydrophilic interaction-ultrahigh-performance liquid chromatography (HILIC) with a Thermo Scientific Vanquish UHPLC system coupled with fluorescent (Thermo Scientific, Waltham, MA, USA) and ESI-MS detectors (Thermo Q Exactive Orbitrap). A description of the sample preparation methods, experimental and instrumental conditions, and data analysis is provided in the Supporting Information.</p><!><p>To test the reliability of CoMMon for monitoring the progress of CAZyme-catalyzed reactions, we first applied the method to measure the kinetics of NEUC-catalyzed cleavage of Neu5Ac from five glycoprotein substrates—prostate specific antigen (PSA), α1-antitrypsin (α1AT), bovine fetuin (BF), α1-acid glycoprotein (AGP), and haptoglobin (Hp). NEUC (NanI subtype secreted by Clostridium perfringens) is a nonspecific exo-neuraminidase that cleaves α2–3, α2–6, and α2–8-linked Neu5A, albeit with a preference for α2–3-linked Neu5Ac.32,33 Detailed glycosylation information for the five glycoproteins is given in the Supporting Information.</p><p>Shown in Figure S3 are representative mass spectra acquired for the five glycoproteins, as well as their corresponding asialo forms. The CoM of a given glycoprotein and its asialo-form represents the starting point (CoM0) and end point (CoM∞), respectively, for the NEUC-catalyzed reactions. Putative glycan compositions of the detected glycoforms, assigned based on possible combinations of N-glycans (PSA, α1AT, AGP, and Hp) or N- and O-glycans (BF), are given in the Glycoform Identification file. The Neu5Ac content of each glycoprotein sample is summarized in Figure S4a in the form of heat maps. According to the heat maps, Hp has the highest degree of sialylation (an average of ~19 Neu5Ac); AGP and BF exhibit a similar degree of sialylation (~13) followed by α1AT (~7) and PSA (~2).34,35 Similar average degrees of sialylation were obtained by considering the differences in the measured CoM0 and CoM∞ values (Figure S4b and S4c).</p><p>The NEUC-catalyzed cleavage of Neu5Ac from each of the five glycoprotein substrates was continuously monitored by ESI-MS performed on aqueous ammonium acetate (200 mM, pH 6.7, and 25 °C) of NEUC and PSA, α1AT, AGP, BF, or Hp at reaction times ranging from 3 to 240 min (α1AT, BF, and AGP) or 180 min (PSA and Hp). A known concentration of 13C3-Neu5Ac, which served as an IS to monitor the concentration of Neu5Ac released from the glycoproteins, was added to each solution. Representative ESI mass spectra acquired for α1AT at reaction times of 5 and 150 min are shown in Figure 2a. Mass spectra for the other glycoprotein substrates are shown in Figure S5. Over the course of the reaction, a noticeable shift to lower m/z was observed for each glycoprotein, consistent with the loss of Neu5Ac residues. Also detected are signals corresponding to protonated and sodiated ions of Neu5Ac and 13C3-Neu5Ac. The relative abundances of the Neu5Ac ions increased with reaction time, consistent with the increasing concentration of free Neu5Ac in solution.</p><p>Progress curves were constructed from the time-resolved ESI mass spectra acquired for the five glycoproteins using CoMMon (Figure 2b). Curves corresponding to the time-dependent concentration of free Neu5Ac were also calculated using the IS method, which is, arguably, the most reliable approach to measure reaction progress using ESI-MS (Figure 2b). It can be seen that, for all substrates, there is excellent agreement between progress curves measured using CoMMon and with the IS, with a normalized root-mean-square deviation of <5% (Table S1). Moreover, the initial rates obtained from the CoMMon- and IS-derived progress curves (Figure 2c) agree within 5% (Table S1). Together, these results indicate that CAZyme kinetic data for glycoprotein substrates extracted from time-resolved ESI-MS mass spectra using the CoMMon method are of comparable quality to data obtained using an IS. This new method, therefore, alleviates the requirement for ISs, which are not always available and can be difficult to synthesize.</p><!><p>To quantitatively compare the reactivity of the five glycoproteins with NEUC, we simultaneously performed CoMMon and CUPRA-ZYME in aqueous ammonium acetate (200 mM, pH 6.7, and 25 °C), using CS3SLB as a reference substrate to account for differences in NEUC activity. Shown in Figure 3a are representative ESI mass spectra for α1AT. Illustrative ESI mass spectra for the other glycoproteins and the corresponding progress curves are shown in Figures S6 and S7.</p><p>Inspection of the normalized progress curves (Figure 3b) reveals a number of interesting features. First, all five glycoproteins exhibit higher initial rates than CS3SLB, and the trend in relative (to CS3SLB) initial rates (Hp > BF > AGP > α1AT > PSA) qualitatively tracks with the average number of Neu5Ac residues present on each glycoprotein (Figures 3c,d, and S8 and Table S2), although BF is an outlier. NEUC is known to have a sialic acid binding module, in addition to a catalytic domain, which recruits substrates.36 The presence of the binding module likely accounts for the observed dependence of rate on the number of Neu5Ac.36 The reason for the higher relative initial rate observed for BF (compared to AGP) is not known but may reflect the presence of sialylated O-glycans (dominated with 2–3 linked Neu5Ac), which are possibly more preferred substrates than the sialylated N-glycans.37 Interestingly, the slope of a linear fit of the initial rate data for each glycoprotein substrate versus the corresponding number of Neu5Ac is only ~0.2, suggesting that, generally, the Neu5Ac residues are partially shielded from NEUC. At present, the origin of this shielding effect is not known and requires further investigation. Possibly, it is due to steric effects associated with the local protein environment, which limit accessibility of the Neu5Ac residues to NEUC, vide infra.</p><p>Second, the overall appearance of the progress curves differs considerably among the five glycoproteins. However, all the substrates exhibit kinetics (Figure 3e) that are reasonably described by a double exponential function (eq S5), albeit with significantly different rate constants (k1 and k2) and fractional abundances (f1 and f2, Table S3). Double exponential kinetics were previously observed for PSA treated with human neuraminidase 3 (NEU3).25 One possible explanation for the observed kinetics is the different reactivity of α2–3- and α2–6-linked terminal Neu5Ac. However, analogous measurements performed on PSA prepared with all α2–3- or all α2–6-linked Neu5Ac also revealed double exponential kinetics (Figure S9 and Table S4). These results rule out Neu5Ac linkage as the origin of this phenomenon and, instead, suggest that the double exponential kinetics observed for PSA arise from the different reactivity of Neu5Ac associated with the α1–3 and α1–6 branches of the biantennary N-glycans. It is also possible that such an explanation applies to triantennary and tetra-antennary N-glycans, wherein the Neu5Ac associated with the α1–3 and α1–6 branches has distinct kinetics. However, it must be stressed that the suggestion that branch specificity is responsible for the double exponential kinetics is speculative and requires follow-up investigations (by CoMMon) of additional glycoprotein substrates with known N-glycan compositions. It will also be important to test other neuraminidases to establish whether this is a general phenomenon.</p><!><p>To demonstrate the ability of CoMMon to measure CAZyme kinetics for large glycoprotein substrates, for which individual species (glycan compositions) cannot be resolved by ESI-MS, we applied the method to the NEUC-catalyzed desialylation of alpha-2-macroglobulin (α2M). Human α2M, an acute-phase protein that acts as the proteinase inhibitor, exists predominantly as a tetramer (MW ~720 kDa) at neutral pH; each α2M monomer possesses 8 N-glycosylation sites consisting predominantly of biantennary and triantennary complex type N-glycans.38 Representative ESI mass spectra for α2M and asialo-α2M are shown in Figure S10; mass spectra acquired at reaction times of 5 and 150 min are shown in Figure 4a. Notably, the optimal instrumental parameters required to detect the α2M tetramer were not ideal for the detection of free Neu5Ac and the associated IS ions. As a result, it was not possible (with the current instrumentation) to apply CoMMon to the α2M tetramer and, simultaneously, monitor the free Neu5Ac released. Instead, CUPRA-ZYME, using CS3SLS, was performed together with CoMMon to assess relative reactivity.</p><p>Comparison of the resulting progress curves and initial rates for tetrameric α2M and CS3SLS (Figure 4c) reveals that, despite possessing approximately 65 reactive Neu5Ac residues (calculated from the difference in CoM0 and CoM∞ values), α2M is a slightly poorer substrate than CS3SLS. This finding suggests significant shielding of the Neu5Ac residues in tetrameric α2M, which is qualitatively consistent with the results obtained for the other glycoproteins tested. Indeed, inspection of the available crystal structure of deglycosylated human α2M39 (PDB: 4ACQ) shows that all but one (N869) of the N-glycosylation sites are located in close proximity of subunit–subunit interfaces (Figure S11), which could limit their accessibility to NEUC. To test whether the slow relative kinetics of α2M result from protein shielding of the Neu5Ac residues, we reduced the α2M sample by treatment with DTT (to cleave the disulfide bonds) and then incubated with IAA (to bind with the free sulfhydryl to prevent the reformation of disulfide bonds) with the goal of making the N-glycans more accessible to NEUC. ESI-MS analysis of the resulting sample revealed the presence of both monomeric and dimeric α2M (Figure S12); ESI mass spectra acquired for the corresponding asialo-glycoforms are also shown. Representative ESI mass spectra of the monomeric and dimeric α2M measured after 5 and 150 min treatment with NEUC are shown in Figure 4b, and the corresponding progress curves are shown in Figure 4d. Notably, inspection of progress curves reveals that monomeric α2M and dimeric α2M are more reactive than tetrameric α2M, with relative initial rates that are ~3.5 and ~1.6 times higher, respectively (Figure 4e). The greater reactivity of the monomer and dimer provides support for the hypothesis of widespread Neu5Ac shielding in tetrameric α2M.</p><!><p>To demonstrate the applicability of CoMMon for monitoring the kinetics of GT reactions, the transfer of Neu5Ac from the CMP-Neu5Ac donor to asialo-PSA, asialo-α1AT, asialo-BF, and asialo-AGP by a soluble recombinant form of human ST6Gal1 was investigated. ST6Gal1, one of two human STs responsible for the formation of α2–6 Neu5Ac linkages to Gal, transfers Neu5Ac to Galβ1–4GlcNAc-R acceptors with high specificity.40 To quantitatively assess the specificity of ST6Gal1 for the four asialo-glycoproteins, CUPRA-ZYME, employing CSLNnTB, was performed simultaneously. Representative ESI mass spectra acquired for the four asialo-glycoproteins at three different reaction times (3, 30, and 90 min) are shown in Figures S13 and S14. Normalized (to CSLNnTB) progress curves for each of the asialo-glycoproteins are shown in Figure 5a; relative initial rates are plotted in Figure 5b, and detailed fitting results are shown in Table S5.</p><p>Inspection of the CoMMon-derived progress curves reveals a number of significant features. First, in contrast to what was observed for desialylation by NEUC, there is no correlation between the initial rate and the number of acceptor sites in the four asialo-glycoproteins (Figure 5c). Asialo-BF and asialo-AGP, which have a similar number of N-glycan acceptor sites, exhibit relative initial rates ~2.1 and ~1.4 times higher than those of CSLNnTB, respectively. The higher reactivity agrees with the results of a previous study of human ST6Gal1 using CMP-14CNeu5Ac as the donor.41,42 The higher relative initial rate measured for asialo-PSA (~1.7 times higher than that of CSLNnTB) also agrees with the findings of our previous study.25 In contrast, the initial rate for asialo-α1AT is ~0.75 times that of CSLNnTB, despite possessing an average of eight acceptor sites.</p><p>Second, the apparent kinetics for ST6Gal1-catalyzed sialylation of the four glycoproteins are reasonably well described by double exponential functions (Figure 5d). Although such a behavior was previously reported for PSA, it has not been shown to be a general feature of ST6Gal1-catalyzed reactions.25 Given that the fractional abundances (f1 and f2) of the two reactive forms of PSA are similar (~0.5, Table S6), the double exponential kinetics can be explained by the different reactivity (k1/k2 ≈ 55, Table S6) of Neu5Ac acceptor sites on the biantennary N-glycan, which is the dominant glycan found on the PSA sample used in this work (Figures 5e and S15). Indeed, analysis of the released PSA N-glycans at early reaction times shows that Neu5Ac is present only on the Gal residue associated with the α1–3 branch (Figure S15). A similar behavior was found for asialo-α1AT, which also contains predominantly (~85%) biantennary N-glycans (Figure 5e). Like with PSA, the fractional abundances are similar (~0.5, Table S6), but the rate constants differ significantly (k1/k2 ≈ 13, Table S6). Moreover, Neu5Ac is found only on the Gal residue associated with the α1–3 branch of the biantennary N-glycans at early reaction times (Figure S16). This clear preference for the α1–3 branch of biantennary N-glycans of asialo forms of PSA and α1AT is consistent with observations reported for the sialylation of the Fc fragment of immunoglobulin G by ST6Gal1.43,44 It is notable, however, that, even though both PSA and α1AT contain predominantly biantennary N-glycans, asialo-α1AT exhibits a much slower reactivity compared with asialo-PSA. These results suggest that, in addition to the inherent differences in the reactivity of α1–3 and α1–6 branches of the biantennary N-glycans, other factors (perhaps local protein environment) affect ST6Gal1 kinetics.</p><p>The kinetics measured for the other two glycoproteins, asialo-BF (which contains predominantly triantennary N-glycans, Figures 5e and S17) and asialo-AGP (triantennary and tetra-antennary N-glycans, Figures 5e and S18), exhibit a significantly larger (compared to asialo-PSA and asialo-α1AT) fast component (f1 ~0.7), suggesting that terminal Gal residues associated with these more branched N-glycans are better substrates. An analysis of N-glycan released from asialo-AGP and asialo-BF reveals that terminal Gal residues are more abundant on the antennae associated with the α1–3 branch (Galβ1–4GlcNAcβ1–2Manα1–3 or Galβ1–4GlcNAcβ1–4Manα1–3, Table S7) than on those on the α1–6 branch. Together, these findings suggest that the antennae associated with the α1–3 branch of triantennary or tetra-antennary N-glycans are better ST6Gal1 substrates than those associated with the α1–6 branch. This conclusion agrees with previous observations made for sialylation of asialo-transferrin and asialo-AGP by bovine ST6Gal1.45 It is also qualitatively consistent with conclusions drawn from the results of a nuclear magnetic resonance spectroscopy study of ST6Gal1 and glycopeptides possessing triantennary and tetra-antennary N-glycans, wherein the Galβ1–4GlcNAcβ1–6Manα1–6 branch was found to be a poorer substrate compared to the Galβ1–4GlcNAcβ1–2Manα1–3, Galβ1–4GlcNAcβ1–4Manα1–3, and Galβ1–4GlcNAcβ1–2Manα1–6 branches.46</p><!><p>This work introduces CoMMon, a simple, versatile, and quantitative ESI-MS method for measuring the kinetics of CAZyme reactions involving heterogeneous, intact glycoprotein substrates. The assay relies on the continuous monitoring of the CoM of the ensemble of glycoprotein substrates and their corresponding CAZyme products using ESI-MS. The reaction progress curve is generated from the time-dependent change in CoM. Importantly, there is no requirement for calibration curves, IS, labeling, or mass spectrum deconvolution.</p><p>To test reliability, the assay was used to monitor the NEUC-catalyzed cleavage of Neu5Ac residues from a series of glycoproteins of varying MWs and degrees of glycosylation. The measured progress curves and initial rates determined with CoMMon are in good agreement (within ≤5%) with results obtained, simultaneously, using an isotopically labeled IS. Interestingly, all the glycoprotein substrates tested exhibit progress curves that are reasonably well described by double exponential functions. This finding suggests distinct kinetic reactivity for Neu5Ac associated with the underlying α1–3 and the α1–6 branches of the N-glycans. Future efforts will be directed toward further testing this hypothesis and assessing whether it is a common feature of neuraminidase-catalyzed reactions.</p><p>A powerful feature of CoMMon is that it can be performed simultaneously with CUPRA-ZYME, thereby enabling the relative reactivity of glycoprotein substrates to be quantitatively established. Application of the combined CoMMon/CUPRA-ZYME approach to the NEUC-catalyzed cleavage of Neu5Ac residues from a series of glycoproteins revealed that the initial rates, normalized to that of a common CUPRA substrate, increase with the increasing Neu5Ac content, as expected. However, the increase in rate per Neu5Ac residue (~0.2×) suggests that the Neu5Ac residues of the glycoproteins are partially shielded from NEUC. Measurements performed using other neuraminidases are now needed to establish the extent to which shielding is enzyme dependent.</p><p>To illustrate versatility, the combined CoMMon/CUPRA-ZYME approach was used to monitor the kinetics of Neu5Ac transfer to a series of asialo-glycoproteins by human ST6Gal1. Notably, no correlation between relative initial rates and the number of acceptor sites was observed. Moreover, the initial rates of the asialo-glycoproteins are found to be similar to the initial rate measured for the CUPRA substrate (which has a single acceptor site), suggesting that ST6Gal1 has limited accessibility to the N-glycan acceptor sites. Similar to what was observed for the NEUC-catalyzed reactions, the progress curves measured for ST6Gal1 are reasonably described by double exponential functions. The results of N-glycan analysis showed that ST6Gal1 has a preference for the Galβ1–4GlcNAcα1–2Manα1–3 branch of biantennary N-glycans and suggests a preference for the Galβ1–4GlcNAcβ1–2Manα1–3 and Galβ1–4GlcNAcβ1–4Manα1–3 branches of triantennary and tetra-antennary N-glycans. Efforts to elucidate the relative kinetics of acceptor sites in triantennary and tetra-antennary N-glycans are ongoing.</p><p>Finally, it is important to emphasize that CoMMon is not limited to CAZymes and glycoprotein substrates. The method is general and, in principle, can be applied to many different enzyme classes to monitor reaction kinetics for heterogeneous substrates that are amenable to direct ESI-MS detection. Examples include protein post-translational modifications, such as kinase-catalyzed phosphorylation,47,48 and enzymatic degradation of polymers and of phospholipids in model membranes.49,50 Importantly, the CoMMon method removes the need for expensive homogeneous substrates to study kinetics and provides a reliable method suitable for natural substrates, including macromolecules. Furthermore, CoMMon is not restricted to enzyme reactions and can be used to measure the kinetics of chemical reactions involving one or more heterogeneous reactants, for the relative quantification of proteins and complexes in heterogeneous samples (e.g., determination of drug-to-antibody ratios and cargo loading of viral particles)51,52 and lipid binding to membrane proteins.53</p>

PubMed Author Manuscript

Recent advances in transition-metal-catalyzed incorporation of fluorine-containing groups

Fluorine chemistry plays an increasingly important role in pharmaceutical, agricultural, and materials industries. The incorporation of fluorine-containing groups into organic molecules can improve their chemical and physical properties, which attracts continuous interest in organic synthesis. Among various reported methods, transition-metal-catalyzed fluorination/fluoroalkylation has emerged as a powerful method for the construction of these compounds. This review attempts to describe the major advances in the transition-metal-catalyzed incorporation of fluorine, trifluoromethyl, difluoromethyl, trifluoromethylthio, and trifluoromethoxy groups reported between 2011 and 2019.

recent_advances_in_transition-metal-catalyzed_incorporation_of_fluorine-containing_groups

Introduction<!>Fluorination<!><!>Fluorination<!>Palladium catalysis<!><!>Palladium catalysis<!><!>Palladium catalysis<!><!>Palladium catalysis<!><!>Palladium catalysis<!><!>Palladium catalysis<!><!>Palladium catalysis<!><!>Palladium catalysis<!><!>Palladium catalysis<!><!>Palladium catalysis<!><!>Palladium catalysis<!><!>Palladium catalysis<!><!>Palladium catalysis<!><!>Palladium catalysis<!><!>Palladium catalysis<!><!>Palladium catalysis<!><!>Palladium catalysis<!><!>Palladium catalysis<!><!>Palladium catalysis<!><!>Palladium catalysis<!><!>Palladium catalysis<!><!>Palladium catalysis<!><!>Palladium catalysis<!><!>Copper catalysis<!><!>Copper catalysis<!><!>Copper catalysis<!><!>Copper catalysis<!><!>Copper catalysis<!><!>Copper catalysis<!><!>Copper catalysis<!><!>Copper catalysis<!><!>Copper catalysis<!><!>Copper catalysis<!><!>Copper catalysis<!><!>Copper catalysis<!><!>Other catalysts<!><!>Other catalysts<!><!>Other catalysts<!><!>Other catalysts<!><!>Other catalysts<!><!>Other catalysts<!><!>Other catalysts<!><!>Other catalysts<!><!>Other catalysts<!><!>Other catalysts<!><!>Other catalysts<!><!>Other catalysts<!><!>Other catalysts<!><!>Other catalysts<!><!>Other catalysts<!><!>Other catalysts<!><!>Other catalysts<!><!>Trifluoromethylation<!><!>C(sp3)–CF3 bond formation<!><!>C(sp3)–CF3 bond formation<!><!>C(sp3)–CF3 bond formation<!><!>C(sp3)–CF3 bond formation<!><!>C(sp3)–CF3 bond formation<!><!>C(sp3)–CF3 bond formation<!><!>C(sp3)–CF3 bond formation<!><!>C(sp3)–CF3 bond formation<!><!>C(sp2)–CF3 bond formation<!><!>C(sp2)–CF3 bond formation<!><!>C(sp2)–CF3 bond formation<!><!>C(sp2)–CF3 bond formation<!><!>C(sp2)–CF3 bond formation<!><!>C(sp2)–CF3 bond formation<!><!>C(sp2)–CF3 bond formation<!><!>C(sp2)–CF3 bond formation<!><!>C(sp2)–CF3 bond formation<!><!>C(sp2)–CF3 bond formation<!><!>C(sp2)–CF3 bond formation<!><!>C(sp2)–CF3 bond formation<!><!>C(sp2)–CF3 bond formation<!><!>C(sp2)–CF3 bond formation<!><!>C(sp2)–CF3 bond formation<!><!>C(sp2)–CF3 bond formation<!><!>C(sp2)–CF3 bond formation<!><!>C(sp2)–CF3 bond formation<!><!>C(sp2)–CF3 bond formation<!><!>C(sp2)–CF3 bond formation<!><!>C(sp2)–CF3 bond formation<!><!>C(sp2)–CF3 bond formation<!><!>C(sp2)–CF3 bond formation<!><!>C(sp2)–CF3 bond formation<!><!>C(sp2)–CF3 bond formation<!><!>C(sp2)–CF3 bond formation<!><!>C(sp2)–CF3 bond formation<!><!>C(sp2)–CF3 bond formation<!><!>C(sp2)–CF3 bond formation<!><!>C(sp2)–CF3 bond formation<!><!>C(sp2)–CF3 bond formation<!><!>C(sp2)–CF3 bond formation<!><!>C(sp2)–CF3 bond formation<!><!>C(sp2)–CF3 bond formation<!><!>C(sp2)–CF3 bond formation<!><!>C(sp2)–CF3 bond formation<!><!>C(sp)–CF3 bond formation<!><!>C(sp)–CF3 bond formation<!><!>C(sp)–CF3 bond formation<!><!>C(sp)–CF3 bond formation<!><!>C(sp)–CF3 bond formation<!>Difluoromethylation<!><!>Difluoromethylation<!><!>Difluoromethylation<!><!>Difluoromethylation<!><!>Difluoromethylation<!><!>Difluoromethylation<!><!>Difluoromethylation<!><!>Difluoromethylation<!><!>Difluoromethylation<!><!>Difluoromethylation<!><!>Difluoromethylation<!><!>Difluoromethylation<!><!>Difluoromethylation<!><!>Difluoromethylation<!><!>Trifluoromethylthiolation<!><!>Trifluoromethylthiolation<!><!>Trifluoromethylthiolation<!><!>Trifluoromethylthiolation<!><!>Trifluoromethylthiolation<!><!>Trifluoromethylthiolation<!><!>Trifluoromethylthiolation<!><!>Trifluoromethylthiolation<!><!>Trifluoromethylthiolation<!><!>Trifluoromethoxylation<!><!>Trifluoromethoxylation<!><!>Trifluoromethoxylation<!><!>Conclusion

<p>Compared with other halogens (Cl, Br, I), fluorine (F) has completely different physical and chemical properties, such as a unique electronic structure, strongest electronegativity, and small atomic radius similar to that of hydrogen atoms. Due to these unique properties, the introduction of fluorine into a molecule can cause dramatic changes, such as the acidity or basicity of neighboring groups, dipole moment, and properties such as lipophilicity, metabolic stability, and bioavailability [1]. Consequently, carbon–fluorine bonds have become an integral part of pharmaceutical [2–3], agricultural [4], materials industries [5], and tracers for positron emission tomography [6]. According to statistics, about 35% of agrochemicals and 20% of pharmaceuticals contain fluorine [7].</p><p>Although the content of fluorine in the Earth's crust is relatively abundant (13th most abundant element), scientists have identified only 21 kinds of fluorine-containing natural molecules [8–9]. Therefore, it is highly desirable to introduce a fluorine-containing substituent into a molecule artificially. However, traditional fluorination methods to these building blocks, such as Friedel–Crafts-type electrophilic halogenation [10–11], Sandmeyer-type reactions of diazonium salts [12], and halogenations of preformed organometallic reagents [13], commonly involve multiple steps, harsh reaction conditions, and the use of stoichiometric amounts and/or toxic reagents [14]. Also, low functional group tolerance, being limited to activated arenes, the production of metal salts as stoichiometric byproducts, and poor levels of regioselectivity would always be observed, limiting the progress of fluorine chemistry to some extent. In this regard, the use of various transition metals to catalyze the synthesis of organic fluorides has become a mature field, and the application of these methodologies has allowed decreasing the need of pre-functionalized substrates, less consumption of reaction time and costs, and enabled to produce enantioenriched target compounds [15–20]. Furthermore, transition metals have the unique advantage of possessing multiple mechanistic features, which translates into the ability to apply new substrate classes and provide hitherto novel and inaccessible structures. Therefore, transition-metal-catalyzed fluorination/fluoroalkylation reactions represent an important and hot topic in fluorine chemistry. In addition, among the various metals developed, palladium is the most commonly employed transition metal, followed by copper owing to its high efficiency and cheapness. Meanwhile, other transition metals, such as Fe, Ni, Rh, Ag, Co, etc., have received considerable attention and are widely applied due to their respective characteristics.</p><p>Over the past few years, several reviews on fluorination/fluoroalkylation have disclosed. Kamlet [17] mainly discussed progresses in catalyzed fluorination and trifluoromethylation before 2011, and Besset [21] focused on the direct introduction of fluorinated groups into alkenes and alkynes. Then, Toste [1] covered advances in catalytic enantioselective fluorination, mono‑, di‑, and trifluoromethylation, and trifluoromethylthiolation reactions. Recently, Zhang [14] offered a brief summary of the recent achievements in the ever-growing field of green fluoroalkylation. However, until now, no comprehensive survey of the literature has been reported on this topic. In this review, we highlight the recent progress of transition-metal-catalyzed fluorination and trifluoromethylation reported between 2011 and 2019. Meanwhile, we also present the incorporation of difluoromethyl, trifluoromethylthiol and trifluoromethoxy groups. Some sections of this review are structured around the synthesis of alkyl-, aryl- and vinyl- as well as alkynyl organofluorides. Notably, the current review covers mainly two types of transition-metal-catalyzed reactions: 1) cross-couplings with a fluorinated organometallic species or a halogenated fluorinated species and 2) the direct introduction of fluorinated moieties into nonfunctionalized substrates with a fluorinated reagent. We hope that this review will provide a comprehensive overview of this topic and attract significant attention.</p><!><p>For many years, specialists in the field of fluorine chemistry have been actively studying ways to introduce fluorine into organic molecules by aid of transition-metal catalysis. Depending on the transfer form of fluorine, there are three general strategies for constructing C–F bonds: nucleophilic, electrophilic and radical fluorination (Scheme 1) [22].</p><!><p>The main three strategies of fluorination: nucleophilic, electrophilic and radical fluorination.</p><!><p>In nucleophilic fluorination reactions, the fluoride anion (F−) or a derivative thereof, such as tetrafluoroborate (BF4−), is the fluorine source and behaves as a nucleophile. The electrophile, such as an alkyl chain or an aryl ring with halides or sulfonates, reacts with the fluoride source (Scheme 1a). On the other hand, in the electrophilic fluorination, the nucleophile may be a carbon anion (e.g., Grignard reagent), a compound with electron-rich unsaturated bonds (arene, alkene, or alkyne), or a substrate having a nucleophilic and labile bond (e.g., C−Si, C−Sn, and C−B), while the electrophile is the fluorination reagent (Scheme 1b). As shown in Scheme 1d, many nucleophilic and electrophilic fluorination reagents have been developed by chemists. In the radical fluorination, C–F bonds are produced by carbon-based radicals (generated in situ by various methods) with "atomic fluorine" sources, such as XeF2, hypofluorite, or molecular fluorine (Scheme 1c). Notably, transition metals are not biased to one reaction class, and the same metal may be successfully applied to all three kinds of fluorination.</p><p>Several reviews of fluorination have been published within the past few years, Buchwald [23], Weng [24], Gouverneur [25], Reiser [26], and etc. [22,27–33] discussed progresses of fluorination, such as Weng who focused on the recent advances in the transition-metal-assisted synthesis of alkyl fluorides, and Buchwald introduced the discovery and development of Pd(0)/Pd(II)-catalyzed aromatic fluorination reactions. Herein, we focus on the developments towards the construction of C(sp3)–F and C(sp2)–F bonds with different catalysts, such as palladium, copper, silver, iron, nickel, ruthenium, cobalt, etc.</p><!><p>Palladium is a member of the nickel triad in the periodic table, and palladium complexes exist in three oxidation states, Pd(0), Pd(II), and Pd(IV). Straightforward interconversion between different oxidation states, tolerance to various guiding groups, easy electroplating of C–H bonds, and the compatibility of many Pd(II) catalysts with oxidants make them act as ideal catalysts for C–H activations [34]. Over the last decade, a number of Pd-catalyzed methods have been developed to synthesize aryl fluorides [23,32].</p><p>Allylic fluorination: In 2010, Doyle and co-worker [35] developed a strategy for C–F bond formation of readily available cyclic allylic chlorides and AgF using a Pd(0) catalyst in combination with Trost's bisphosphine ligand at room temperature (Scheme 2a). They also proved that the allylic fluorination was achieved by an SN2-type attack of fluoride on an electrophilic Pd(II)-allyl intermediate. One year later, the same author extended this method to a highly regio- and enantioselective fluorination of acyclic allylic chlorides. Compared to the previous process, this reaction used a different chiral bisphosphine ligand resulting in larger bite angles and afforded the products in good yields (Scheme 2b) [36].</p><!><p>Doyle's Pd-catalyzed fluorination of allylic chlorides.</p><!><p>A palladium-catalyzed method for the formation of allylic C–F bonds from allyl p-nitrobenzoate using TBAF(t-BuOH)4 as the fluoride source was explored by Gouverneur et al. in the same year (Scheme 3) [37]. The 2- and 3-arylpropenyl fluorides can be quickly synthesized under mild conditions in moderate to good yields.</p><!><p>Allylic fluorination of 2- and 3-substituted propenyl esters.</p><!><p>In 2012, a Pd(0)-catalyzed allylic fluorination of allylic phosphorothioate esters with AgF was accomplished by Wu's group (Scheme 4) [38]. The formation of fluorinated products with an overall retention of the stereochemical configuration suggests a mechanism wherein a palladium-π-allyl intermediate undergoes a rapid π-σ-π isomerization.</p><!><p>Regioselective allylic fluorination of cinnamyl phosphorothioate esters.</p><!><p>In 2013, the first example of an allylic C–H fluorination reaction of simple alkenes with Et3N·3HF as a nucleophilic fluoride source was reported by Doyle and co-worker (Scheme 5) [39]. Herein, the authors utilized a Pd/Cr cocatalytic system to generate the allylic fluorides with high regioselectivity (branched > linear).</p><!><p>Palladium-catalyzed aliphatic C–H fluorination reported by Doyle.</p><!><p>Alkyl fluorination of acidic carbonyl compounds and other compounds: In 2012, the group of Sodeoka [40] reported the first example of an enantioselective monofluorination of α-keto esters catalyzed by Pd-μ-hydroxo complexes with cyclopentyl methyl ether (CPME) as the best solvent (Scheme 6). Also, they achieved the diastereoselective reduction of the remaining keto group with lithium tri(sec-butyl)borohydride (ʟ-Selectride). The syn-β-fluoro-α-hydroxy esters were obtained finally in good yields with excellent enantioselectivities (83–95% ee).</p><!><p>Pd-catalyzed enantioselective fluorination of α-ketoesters followed by stereoselective reduction to give β-fluoro-α-hydroxy esters.</p><!><p>There are two examples of a Pd-catalyzed fluorination of oxindoles. In 2012, Shi and co-workers [41] described the enantioselective asymmetric fluorination of oxindoles with an axially chiral C2-symmetric N-heterocyclic carbene (NHC) palladium complex as a catalyst (Scheme 7a). The corresponding products were obtained in excellent yields but low to moderate enantioselectivities. Meanwhile, Wu and co-workers [42] developed a similar system using a BINAP-derived palladium complex to perform the similar reaction with 4,4'-diF-NFSI as the fluorinating agent in higher enantioselectivities (Scheme 7b).</p><!><p>Pd-catalyzed C(sp3)–H fluorination of oxindoles.</p><!><p>In 2012 the group of Sanford [43] achieved the palladium-catalyzed C–H fluorination of 8-methylquinoline derivatives using AgF as the nucleophilic fluoride source and PhI(OPiv)2 as a hypervalent iodine oxidant (Scheme 8). Very recently, they [44] optimized this transformation and achieved the benzylic C–H radiofluorination with no-carrier-added Ag[18F]F. This method was applied to the radiolabeling of diversely substituted 8-methylquinoline derivatives. Notably, in this process, a new method was developed for generating Ag[18F]F by using a sep-pak cartridge.</p><!><p>C–H fluorination of 8-methylquinoline derivatives with F− reagents.</p><!><p>In 2012, van Leeuwen and co-workers [45] described the synthesis of new enantiopure wide-bite-angle diphosphanes and their application in the asymmetric fluorination of α-cyanoacetates with a palladium catalyst (Scheme 9). Under these conditions, the fluorination of ethyl 2-cyano-2-phenylacetate afforded the product with highest enantiomeric excess (93%).</p><!><p>Fluorination of α-cyano acetates reported by van Leeuwen.</p><!><p>In 2013, Kim's group [46] described an enantioselective electrophilic fluorination of α-chloro-β-keto phosphonates with up to 95% ee (Scheme 10). Notably, this reaction used an air and moisture-stable chiral palladium complex as the catalyst, which worked well at low catalyst loading (as low as 0.5 mol %).</p><!><p>The catalytic enantioselective electrophilic C–H fluorination of α-chloro-β-keto phosphonates.</p><!><p>In 2015, Shi et al. [47] introduced a Pd(II)/Pd(IV)-catalyzed fluorination of β-methylene C(sp3)–H bonds of α-amino acid derivatives (Scheme 11a). This process was carried out under the strongly binding bidentate 2-(pyridine-2-yl)isopropylamine (PIP) auxiliary. A range of substrates containing both aliphatic and benzylic C(sp3)–H bonds was finally converted to the corresponding fluorinated products with excellent diastereoselectivities. Based on the PIP auxiliary developed by Shi, Ge's group [48] developed a similar direct, highly site- and diastereoselective fluorination of aliphatic amides (Scheme 11b). Although the roles of Fe(OAc)2 and Ag2CO3 were unclear, their addition significantly improved the reaction yield. A catalytic cycle of these β-fluorination reactions is proposed in Scheme 11. Initially, coordination of the amide with the palladium species followed by a base-promoted ligand-exchange process yields the palladium complex A. Subsequently, cyclometallation of the palladium complex A occurs to produce the intermediate B through the C–H bond-activation process. Oxidative addition of the intermediate B with Selectfluor affords the palladium(IV) species C, followed by reductive elimination and ligand dissociation to give the final product.</p><!><p>Fluorination of unactivated C(sp3)–H bonds directed by the bidentate PIP auxiliary.</p><!><p>Similar to these publications in strategy and products, in the same year, Xu's group [49] presented the palladium-catalyzed direct fluorination of unactivated C(sp3)–H bonds at the β-position of carboxylic acids with NFSI (Scheme 12). To achieve this transformation, an 8-aminoquinoline-derived auxiliary was developed as an effective directing group for the activation of the C–H bonds. In this transformation the presence of Ag2O and pivalic acid was found to be crucial for the successful synthesis of β-fluorinated carboxylic acids.</p><!><p>Fluorination of C(sp3)–H bonds at the β-position of carboxylic acids.</p><!><p>Recently, the first example of a Pd-catalyzed protocol for the general enantioselective electrophilic C(sp3)–H fluorination of benzaldehyde substrates was reported by Yu and co-workers (Scheme 13) [50]. Enantioenriched benzyl fluorides were obtained by aid of a chiral α-amino amide transient directing group (TDG). Notably, the condensation of this bulky amino amide with the aldehyde led to control of the stereochemistry of the C–H insertion step, promoting the C–F over C–O bond formation via an inner-sphere pathway.</p><!><p>Enantioselective benzylic C–H fluorination with a chiral transient directing group.</p><!><p>Fluorination of arenes, aryl bromides, -alcohols, -triflates, and -boronic acid derivatives: In 2013, Larhed and co-workers [51] established a one-pot, two-step fluorination of aryl alcohols via aryl nonafluorobutylsulfonates. This transformation employed Pd2(dba)3/t-BuBrettPhos and CsF to convert aryl alcohols to aryl fluorides at 180 °C under microwave conditions (Scheme 14). The proposed catalytic cycle of this aryl fluorination is also shown. Only reductive elimination was investigated by Larhed, because this reaction step is crucial for product formation and a successful outcome of the reaction.</p><!><p>Microwave-heated Pd-catalyzed fluorination of aryl alcohols.</p><!><p>In the same year, the Ritter group [52] reported a Pd-catalyzed fluorination of arylboronic acid derivatives via a Pd(II)/Pd(III) cycle (Scheme 15). A single-electron-transfer (SET) mechanism involving a well-defined Pd(III) intermediate has been proposed. First, a bis(terpyridyl)Pd(II) complex B is oxidized by Selectfluor with turnover-limiting to obtain Pd(III) C and a Selectfluor radical cation. Then, a transfer of a F· radical from the Selectfluor radical cation to an aryl trifluoroborate occurs, forming the C−F bond and producing a delocalized radical. Finally, SET from the radical to C regenerates palladium species B, and affords a delocalized cation which converts to the aryl fluoride with loss of BF3. Notably, the addition of NaF increases the yield of aryl fluoride by reacting with the generated BF3.</p><!><p>Fluorination of aryl potassium trifluoroborates.</p><!><p>In 2013, Buchwald et al. [53] introduced an improved catalyst system involving CsF and the stable Pd(0) species [(1,5-cyclooctadiene)(L1·Pd)2] (L1 = AdBrettPhos), which is a precatalyst for the fluorination of aryl triflates and heteroaryl triflates (Scheme 16a). Furthermore, aryl fluorides were provided in good to excellent yields with easy to separate byproducts. A year later, the same catalyst was employed for the nucleophilic fluorination of aryl bromides and iodides with AgF and KF [54]. Meanwhile, with a slight modification of the phosphine ligand, Buchwald developed a similar Pd(0) precatalyst [L2Pd]2(cod), which was used to fluorinate nitrogen-containing heteroaryl bromides (Scheme 16b).</p><!><p>C(sp2)–F bond formation using precatalyst [L·Pd]2(cod).</p><!><p>In 2015, Buchwald and co-workers [55] explored a novel ligand for the Pd-catalyzed fluorination of (hetero)aryl triflates and bromides. The desired aryl fluorides were obtained with higher than 100:1 selectivity (Scheme 17).</p><!><p>Pd-catalyzed fluorination of (hetero)aryl triflates and bromides.</p><!><p>More recently, Yamamoto and co-workers [56] described a palladium-catalyzed general method for aromatic C–H fluorination with mild electrophilic fluorinating reagents at room temperature (Scheme 18). Notably, in this process, a reactive transition metal fluoride electrophile B is catalytically formed from A with Selectfluor or NFSI instead of an organometallic intermediate as usual. Then, the activated Pd(IV)–F electrophile B would be capable of electrophilic fluorination of weakly nucleophilic arenes. This unusual mechanism of catalysis may provide a new idea to the catalysis of C–H functionalization reactions.</p><!><p>The Pd-catalyzed C–H fluorination of arenes with Selectfluor/NFSI.</p><!><p>Aryl C–H fluorination with various directing groups: With Pd(OTf)2(MeCN)4 and N-methyl-2-pyrrolidinone (NMP) used as the catalyst system, in 2011 the Yu group [57] described the ortho-fluorination of benzoic acid substrates with a directing group, an electron-deficient removable acidic amide (Scheme 19). With this method, both mono- and difluorinated benzoic acid derivatives can be selectively obtained in high yields.</p><!><p>Pd(II)-catalyzed ortho-monofluorination protocol for benzoic acids.</p><!><p>In 2014, Pu and co-workers [58] devised the regioselective Pd(PPh3)4-catalyzed electrophilic ortho-fluorination of 2-arylbenzothiazoles with NFSI and ʟ-proline as the crucial promoter and the benzothiazoles as the directing groups (Scheme 20). This strategy plays an important role in the pharmaceutical and agrochemical industries.</p><!><p>Pd-catalyzed C(sp2)–H bond fluorination of 2-arylbenzothiazoles.</p><!><p>Meanwhile, Xu's group [59] used O-methyl oxime as the directing group for the Pd-catalyzed ortho-fluorination of aromatic and olefinic C(sp2)–H bonds (Scheme 21a). It is worth noting that a cheap and nontoxic nitrate was added as a highly efficient promoter in the presence of NFSI and Pd2(dba)3. In addition, the authors proposed a reaction mechanism that involves a Pd(II)/Pd(IV) catalytic cycle (Scheme 21b). At the early stage of this process, an in situ-generated cationic [Pd(NO3)]+ species facilitates the C–H bond activation to give intermediate A. The Pd(II)(1a)2 complex B is formed via further C–H bond activation of another molecule 1a by the cyclopalladation(II) intermediate A. Then, intermediate B undergoes oxidative addition by NFSI to give the highly reactive species F–Pd(IV)1a)2-N(SO2Ph)2 (C), which produces the product 2a and reductive elimination intermediate 1a-Pd(II)-N(SO2Ph)2 (D). Finally, intermediate A regenerates from intermediate D by aid of the catalytic amount of HNO3 released during the C–H activation step.</p><!><p>Nitrate-promoted fluorination of aromatic and olefinic C(sp2)–H bonds and proposed mechanism.</p><!><p>In 2015, Zhao et al. [60] discovered a Pd(II)-catalyzed ortho-selective C–H fluorination of oxalyl amide-protected benzylamines (Scheme 22). The yields were up to 95% with NFSI as the [F+] source and tert-amyl alcohol as the solvent.</p><!><p>Fluorination of oxalyl amide-protected benzylamine derivatives.</p><!><p>In 2018, the Sorensen group [61] described a direct Pd-catalyzed ortho-C–H fluorination of benzaldehydes. Notably, these transformations were achieved with orthanilic acids as new transient directing groups (TDGs) in DCE in the presence of air (Scheme 23). This approach employed 1-fluoro-2,4,6-trimethylpyridinium salts as a bystanding F+ oxidant or an electrophilic fluorinating reagent. A broad substrate scope and high functional group compatibility were observed.</p><!><p>C–H fluorination of benzaldehydes with orthanilic acids as transient directing group.</p><!><p>In addition to the methods discussed above, there are some other methods for the aromatic C–H fluorination using electrophilic fluorination reagents with various other directing groups [60,62–66]. Additionally, a diverse range of N-heterocycles, amides and motifs commonly encountered in medicinal chemistry were used as handles to direct C–H fluorination for the synthesis of pharmaceutical drugs (Scheme 24) [25].</p><!><p>Pd(II)-catalyzed aryl C–H fluorination with various directing groups.</p><!><p>Despite the success of Pd-catalyzed fluorinations, the more widespread use of these technologies has been partially hampered by the high cost and toxicity associated with Pd, in addition to the difficulty encountered when attempting to remove this metal from product mixtures. Therefore, cupper as low-cost, earth-abundant and readily available transition metal has emerged as a prevalent catalyst in a huge number of organic transformations. Similar to palladium complexes, copper complexes generally exist in four oxidation states, Cu(0), Cu(I), Cu(II), and Cu(III) and various fluorination reactions could be developed by different catalytic mechanisms.</p><p>Fluorination of inert C–H bonds, alkyl bromides and -triflates: In a 2012 study, Lectka's group [67] disclosed the catalytic fluorination of a series of aliphatic, benzylic, and allylic substrates with moderate yields. In this case, the authors employed a multicomponent catalytic system, involving Selectfluor, the radical precursor N-hydroxyphthalimide (NHPI), an anionic phase-transfer catalyst (KB(C6F5)4), and a Cu(I)-bisimine complex, to give the corresponding monofluorinated product (Scheme 25).</p><!><p>Cu-catalyzed aliphatic, allylic, and benzylic fluorination.</p><!><p>One year later, Weng and co-workers [68] synthesized and characterized a new copper(I) fluoride complex ligated by a phenanthroline derivative. This complex was applied to the SN2 fluorination of primary and secondary alkyl bromides, producing the corresponding alkyl fluorides in 40–90% yield (Scheme 26).</p><!><p>Cu-catalyzed SN2 fluorination of primary and secondary alkyl bromides.</p><!><p>In 2014, the group of Lalic [69] developed a mild fluorination of alkyl triflates with potassium fluoride catalyzed by a phase-transfer copper catalyst (Scheme 27). Notably, with 10 mol % of (IPr)CuOTf, full conversion can be accomplished in 10 minutes at 45 °C.</p><!><p>Copper-catalyzed fluorination of alkyl triflates.</p><!><p>Allylic fluorination: In 2013, there is an example of a copper-catalyzed fluorination of internal allylic bromides (Scheme 28). In Liu's study, this approach was achieved using Et3N·3HF as the fluorine source with a high catalyst loading (20–30 mol %) affording the products in 45–92% yield [70]. The heteroatom-containing functional group (R1) is necessary for good reactivity and regioselectivity.</p><!><p>Cu-catalyzed fluorination of allylic bromides and chlorides.</p><!><p>α-Fluorination of acidic carbonyl compounds: In 2011, Shibatomi and co-workers [71] described the one-pot asymmetric gem-chlorofluorination of active methylene compounds by using a copper(II) complex with a chiral spiro 2-pyridyl monooxazoline ligand (SPYMOX). The corresponding α-chloro-α-fluoro-β-keto esters were isolated with up to 92% ee (Scheme 29a). This approach could be extended to asymmetric gem-chlorofluorination of β-ketophosphonates. Two years later, the same authors [72] demonstrated the highly enantioselective fluorination of α-alkyl-β-keto esters and α-alkylmalonates using the same catalyst system (Scheme 29b). Moreover, various cyclic and acyclic substrates were successfully fluorinated with high enantioselectivities.</p><!><p>Synthetic strategy for the fluorination of active methylene compounds.</p><!><p>In 2013, the Kesavan group [73] reported the use of tartrate-derived bidentate bisoxazoline-Cu(II) complexes for the enantioselective fluorination of aliphatic cyclic and acyclic β-ketoesters with up to 98% yields (Scheme 30). In this method, (S,S)-Nap-(R,R)-Box as the most suitable diastereomeric ligand forms a 5-membered chelate with copper.</p><!><p>Fluorination of β-ketoesters using a tartrate-derived bidentate bisoxazoline-Cu(II) complex.</p><!><p>In the same year, an efficient and highly enantioselective fluorination of β-ketoesters catalyzed by diphenylamine-linked bis(thiazoline)-Cu(OTf)2 complexes was reported by Du and co-worker (Scheme 31a) [74]. Che and co-workers [75] achieved a similar α-fluorination of β-ketoesters and N-Boc-oxindoles (Scheme 31b). Compared with Du's method, Che employed both AgClO4 and chiral iron(III)-salan complexes as the catalyst.</p><!><p>Highly enantioselective fluorination of β-ketoesters and N-Boc-oxindoles.</p><!><p>In 2016, the group of Nishikata [76] described a copper-catalyzed site-selective fluorination of α-bromocarbonyl compounds using a copper/CsF catalyst system (Scheme 32). Tertiary alkyl fluorides could be generated by this fluorination through the assistance of an amide group. From the results, the catalytic cycle of this reaction includes: 1) copper salt induced generation of the alkyl radical species B from substrate A and 2) fluorination of the alkyl radical species B with CuF2, which is in situ-generated from the reaction of CuXBr and CsF with the aid of an amide group, gives the desired product and recyclable CuF.</p><!><p>Amide group-assisted site-selective fluorination of α-bromocarbonyl compounds.</p><!><p>Csp2–H bond formation catalyzed by Cu catalysts: In 2013, Sanford and co-workers [77] developed a simple and practical process for the nucleophilic fluorination of arylpotassium trifluoroborates. The reaction proceeds in CH3CN at 60 °C in the presence of Cu(OTf)2 as the catalyst and KF as the fluoride source (Scheme 33). A possible mechanism for this transformation is proposed in Scheme 33 below. Notably, Cu acts as both a mediator and an oxidizer in this reaction.</p><!><p>Cu-mediated aryl fluorination reported by Sanford [77].</p><!><p>In the same year, Daugulis et al. [78] presented a Cu-catalyzed selective fluorination of benzoic acid derivatives and benzylamine derivatives assisted by an aminoquinoline auxiliary. With a CuI catalyst, AgF as fluoride source, NMO as oxidant, and DMF as solvent, they achieved the selective mono- or difluorination in high yields (Scheme 34). Notably, pyridine as an additive could prevent the decomposition of an amide substrate in a long-time reaction.</p><!><p>Mono- or difluorination reactions of benzoic acid derivatives.</p><!><p>Meanwhile, the group of Sanford [79] presented the nucleophilic fluorination of diaryliodonium salts with KF through a Cu(I/III) catalytic cycle mechanism. This procedure preferentially fluorinates the smaller aromatic ligand on iodine(III). Also, the addition of Cu(OTf)2 and 18-crown-6 promoted the fluorination effectively. Finally, excellent yields, fast rate, high selectivity, and a broad substrate scope were observed by the authors (Scheme 35). The proposed mechanism is as follows: ligand exchange of the active Cu(I) catalyst A, which is generated via either reduction by the solvent or disproportionation of the precatalyst Cu(II)(OTf)2, provides Cu(I)-F (B). Then, oxidation of Cu(I)-F (B) by the diaryliodonium reagent forms Cu(III)–aryl intermediate C. Subsequently, a reductive elimination of intermediate C provides a putative π-complex D, which then releases the desired aryl–F product and regenerates the CuI catalyst A.</p><!><p>Cu-catalyzed fluorination of diaryliodonium salts with KF.</p><!><p>Subsequently, the Cu-catalyzed fluorination of 2-pyridylaryl bromides was achieved by Liu and co-workers [80] through a Cu(I/III) catalytic cycle as well (Scheme 36). This method is based on the aid of an important pyridyl directing group and the final aryl C–F bond is formed after the reductive elimination of ArCu(III)–F species.</p><!><p>Copper(I)-catalyzed cross-coupling of 2-pyridylaryl bromides.</p><!><p>Other transition metals, including Co, Ni, Fe, Ag, Ir, Mn, etc., have received more and more attention.</p><p>Aliphatic and benzylic C–H fluorination and decarboxylative fluorination: In 2012, a silver-catalyzed radical decarboxylative fluorination of aliphatic carboxylic acids in aqueous solution was provided by Li and co-workers (Scheme 37) [81]. The corresponding alkyl fluorides were produced in 47–95% yield under mild conditions. Additionally, the authors proposed a mechanism involving a Ag(III)-mediated SET followed by a fluorine transfer.</p><!><p>AgNO3-catalyzed decarboxylative fluorination of aliphatic carboxylic acids.</p><!><p>Subsequently, the group of Groves [82] developed two manganese catalysts for the fluorination of C(sp3)–H bonds (Scheme 38). On the one hand, they employed a manganese porphyrin to catalyze the oxidative aliphatic C–H fluorination with iodosylbenzene (PhIO) as a stoichiometric oxidant. A variety of substrates, including simple hydrocarbons, substituted cyclic molecules, terpenoids, and steroid derivatives, were selectively fluorinated at some otherwise inaccessible sites, however, in low to moderate yields. On the other hand, the same group [83] developed Mn(salen)Cl as a catalyst for the direct C–H fluorination at benzylic positions with a nucleophilic fluorine source. Notably, Groves adapted the method for the 18F-radiofluorination of benzylic and aliphatic C–H bonds using no-carrier-added [18F]-fluoride with Mn(salen)OTs [84].</p><!><p>The Mn-catalyzed aliphatic and benzylic C–H fluorination.</p><!><p>In 2013, Lectka's group [85–86] reported an iron-catalyzed C(sp3)–H fluorination of benzylic substrates with or without an electron-withdrawing group (EWG) in the presence of Selectfluor (Scheme 39). Notably, an EWG beta to the benzylic position is efficient for an excellent selectivity of the benzylic fluorination.</p><!><p>Iron(II)-promoted C–H fluorination of benzylic substrates.</p><!><p>Moreover, Gouverneur and co-workers [87] established the decarboxylative fluorination of α,α-difluoro- and α-fluoroarylacetic acids with a wide functional group compatibility in the presence of AgNO3 as catalyst in good yields (Scheme 40). Further, this approach was efficiently applied to the preparation of [18F]-labelled tri- and difluoromethylarenes using [18F]Selectfluor bis(triflate).</p><!><p>Ag-catalyzed fluorodecarboxylation of carboxylic acids.</p><!><p>In 2014, Chen and co-workers [88] described a selective direct C(sp3)–H fluorination catalyzed by a commercially available vanadium(III) oxide with Selectfluor in good yields (Scheme 41). It is noteworthy that the catalyst and the byproduct H-TEDA could be removed easily by filtration.</p><!><p>Vanadium-catalyzed C(sp3)–H fluorination.</p><!><p>A simple AgNO3-catalyzed synthesis of alkyl fluorides through radical deboronofluorination of alkyl boronates and boronic acids in acidic aqueous solution was also developed by Li and co-workers in 2014 [89]. This method features good yields and a wide functional group compatibility (Scheme 42).</p><!><p>AgNO3-catalyzed radical deboronofluorination of alkylboronates and boronic acids.</p><!><p>Recently, the group of Van Humbeck [90] reported a selective and mild method for the C–H fluorination of azaheterocycles with Selectfluor at room temperature. In this case, a new radical mechanism was proposed that electron transfer from the heterocyclic substrate to Selectfluor eventually generates a benzylic radical, leading to the desired C–F bond formation. The excellent selectivity of the desired fluorinated product was obtained without additives. In addition, a catalytic amount of iron(III) complex [FeCl4][FeCl2(dmf)3] was found to improve the yields in some cases (Scheme 43).</p><!><p>Selective heterobenzylic C–H fluorination with Selectfluor reported by Van Humbeck.</p><!><p>With an Fe(II)-catalyzed orchestrated redox process, an alkoxyl radical-guided strategy for the site-selective fluorination of unactivated methylene and methine C–H bonds was published by Liu and co-workers in 2018 (Scheme 44) [91]. The fluorination of various primary, secondary, and tertiary hydroperoxides was achieved in moderate to excellent yields, with the hydroperoxide functional group acting as a precursor of an alkoxy radical to control site-selective carbon-centered radical formation.</p><!><p>Fe(II)-catalyzed site-selective fluorination guided by an alkoxyl radical.</p><!><p>Allylic fluorination: In 2011, the group of Nguyen [92] developed the nucleophilic fluorination of allylic trichloroacetimidates, as shown in Scheme 45a. Cyclooctadiene iridium chloride dimer, [IrCl(COD)]2, was an effective catalyst to promote this fluorination with Et3N·3HF, forming allylic fluorides in moderate to good yields. This facile method shows a good regioselectivity to gain the branched isomer within 1 h. Later in 2017, they described a similar method for the asymmetric fluorination of racemic allylic trichloroacetimidates utilizing a chiral bicyclo[3.3.0]octadiene-ligated iridium complex (Scheme 45b) [93]. This reaction proceeded under mild conditions with an extremely broad substrate scope, as well as excellent branched-to-linear ratios and enantioselectivities.</p><!><p>Fluorination of allylic trichloroacetimidates reported by Nguyen et al.</p><!><p>In 2013, Gouverneur and co-workers [94] demonstrated the regio and stereocontrolled fluorination of allylic carbonates with [Ir(COD)Cl]2 as the catalyst and TBAF(t-BuOH)4 as the fluoride source to produce branched and linear allylic fluorides (Scheme 46). Remarkably, this was the first example to afford (Z)-allyl fluorides (Z:E ratio > 20:1).</p><!><p>Iridium-catalyzed fluorination of allylic carbonates with TBAF(t-BuOH)4.</p><!><p>In 2015, Nguyen et al. [95] explored the asymmetric fluorination of racemic, secondary allylic trichloroacetimidates with Et3N·3HF using a chiral-diene-ligated Ir complex (Scheme 47). This process proceeded under mild conditions with excellent enantioselectivity and yields, a broad substrate scope, as well as a wide range of functional group compatibility. Notably, this strategy overcomes the challenges associated with the formation of secondary allylic fluorides bearing α-linear substituents, providing complete regio and stereocontrolled acrylic allylic fluorides.</p><!><p>Iridium-catalyzed asymmetric fluorination of allylic trichloroacetimidates.</p><!><p>Fluorination of acidic carbonyl compounds: In 2010, Itoh and co-workers [96] demonstrated the asymmetric fluorination of cyclic and acyclic β-ketoesters by using a catalytic amount of Co(acac)2 with (R,R)-Jacobsen's salen ligand (Scheme 48). The α-fluorinated products were thus obtained with good enantioselectivity.</p><!><p>Cobalt-catalyzed α-fluorination of β-ketoesters.</p><!><p>In the same year, Kim's group [97] accomplished an efficient enantioselective electrophilic α-fluorination of various α-chloro-β-ketoesters catalyzed by chiral nickel complexes with good enantioselectivity (up to 99% ee). Notably, the chiral nickel-diamine complexes are air and moisture-stable (Scheme 49).</p><!><p>Nickel-catalyzed α-fluorination of various α-chloro-β-ketoesters.</p><!><p>In 2011, two nickel-catalyzed protocols for the enantioselective α-fluorination of β-ketoesters were reported separately. In van Leeuwen's reaction, SPANamine derivatives were synthesized and applied as chiral ligands in the asymmetric α-fluorination of β-ketoesters (Scheme 50a) [98]. Meanwhile, to achieve this transformation, Gade and co-workers [99] developed a new class of chiral tridentate N-donor pincer ligands, bis(oxazolinylmethylidene)isoindolines. They obtained the desired products under mild conditions with excellent enantioselectivities (up to >99% ee) and good yields (Scheme 50b).</p><!><p>Ni(II)-catalyzed enantioselective fluorination of oxindoles and β-ketoesters.</p><!><p>Also, Feng et al. [100] developed a new method for the highly enantioselective fluorination of N–H-free 3-substituted oxindoles catalyzed by a Sc(III)/N,N'-dioxide complex. A series of 3-aryl- and 3-alkyl-3-fluoro-2-oxindoles were obtained in excellent yields and enantioselectivities (89–99% ee) with NFSI under basic conditions (Scheme 51).</p><!><p>Scandium(III)-catalyzed asymmetric C–H fluorination of unprotected 3-substituted oxindoles.</p><!><p>In 2016, a mild, amide-directed fluorination of benzylic, allylic, and unactivated C–H bonds was described by the Cook group [101]. By the use of the iron(II) triflate (Fe(OTf)2) as catalyst, the desired fluorides were finally obtained through a F-transfer of a short-lived radical intermediate (N-fluoro-2-methylbenzamides) in up to 93% yield (Scheme 52).</p><!><p>Iron-catalyzed directed C–H fluorination.</p><!><p>Csp2–H bond formation catalyzed by Ag catalysts: In 2010, the Ritter group [102] firstly reported a Ag-catalyzed fluorination of arylstannane derivatives with the electrophilic fluorination reagent F-TEDA-PF6 (Scheme 53). Also, the reaction was applied to late-stage fluorination of small molecules. However, this method uses toxic arylstannanes as starting materials and requires an additional synthetic step from the triflate or halide to the stannanes.</p><!><p>Electrophilic silver-catalyzed Ar–F bond-forming reaction from arylstannanes.</p><!><p>Transition-metal-catalyzed trifluoromethylation reactions have made great progress in the joint efforts of organic fluorination scientists and metalorganic chemists over the past decade. Introducing trifluoromethyl groups into organic molecules can significantly alter their properties, such as their metabolic stability, lipophilicity, and the ability to penetrate the blood–brain barrier. Similar to fluorination, trifluoromethylation can also be achieved by three reaction types: nucleophilic, electrophilic and radical trifluoromethylation.</p><p>In recent years, many novel trifluoromethylation reagents, such as cationic, anionic and radical CF3 sources have been discovered and offer manifold choices to effect electrophilic, nucleophilic and radical trifluoromethylation [103] (Figure 1). The selection of the trifluoromethylation reagent has become the main factor in the optimization of these reactions. With a suitable trifluoromethylation reagent, a wide range of substrates are directly converted to the desired trifluoromethylated products. Several reviews [104–110] have been published on this subject, while this part mainly discusses trifluoromethylation reactions catalyzed by metals. However, there are only a few methods available for the C(sp3)–CF3 bond formation and this transformation still needs further examination.</p><!><p>Nucleophilic, electrophilic and radical CF3 sources.</p><!><p>Copper catalysis: In 2011, two Cu(I)-catalyzed allylic trifluoromethylation reactions of terminal olefins have been developed independently by the groups of Buchwald [111] and Wang [112] (Scheme 54). Under similar mild conditions using Togni's reagent II, the desired allyl–CF3 products were obtained and the methods well tolerated a variety of functional groups (e.g., esters, epoxides, amides, alcohols, or aldehydes). Moreover, the thermodynamically favored E-olefin was generated with high stereoselectivity in good yields.</p><!><p>Cu(I)-catalyzed allylic trifluoromethylation of unactivated terminal olefins.</p><!><p>In 2012, two different groups [113–114] individually reported the direct trifluoromethylation of allylsilanes under very similar conditions. These processes furnished various branched cyclic and acyclic allylic CF3 products using copper as the catalyst (Scheme 55).</p><!><p>Direct copper-catalyzed trifluoromethylation of allylsilanes.</p><!><p>Subsequently, an enantioselective trifluoromethylation of cyclic β-ketoesters with commercially available trifluoromethylating reagents was reported by Gade and co-workers using a Cu-boxmi catalyst [115]. Under mild conditions, both five and six-membered ring β-ketoesters were converted to the corresponding products in high yields and enantioselectivities (Scheme 56).</p><!><p>Cupper-catalyzed enantioselective trifluoromethylation of five and six-membered ring β-ketoesters.</p><!><p>In 2018, the first example for the copper-catalyzed stereospecific trifluoromethylation of secondary propargyl sulfonates was described by the group of Zhang [116]. The resulting chiral trifluoromethylated alkynes were acquired with high regioselectivity and stereospecificity (ees up to >99%). Furthermore, this reaction showed a broad substrate scope, as well as excellent functional-group compatibility (Scheme 57). A possible mechanism was proposed: firstly, trifluoromethylcopper complex A, generated from CuCN with TMSCF3, undergoes oxidative addition with a secondary propargyl sulfonate to give a configuration-inversed propargyl-Cu(III) species B. Then, the reductive elimination of B affords the final product with overall inversion of the configuration.</p><!><p>Cu-catalyzed highly stereoselective trifluoromethylation of secondary propargyl sulfonates.</p><!><p>Recently, Li and co-workers [117] explored a simple and facile method to access δ-trifluoromethylated carboxamides and sulfonamides through a copper-catalyzed 1,5-hydrogen atom transfer (Scheme 58).</p><!><p>Remote C(sp3)–H trifluoromethylation of carboxamides and sulfonamides.</p><!><p>Other catalysts: In 2013, Gouverneur and co-workers [118] described a photoredox-based catalytic approach to afford enantioenriched branched allylic CF3 products from allylsilanes using [Ru(bpy)3]Cl2 (Scheme 59). Herein, the silyl group in the substrate plays an important role in controlling the regioselectivity of the trifluoromethylation reaction.</p><!><p>Trifluoromethylation of allylsilanes with photoredox catalysis.</p><!><p>Later in 2017, Li's group [119] described a practical protocol for the decarboxylative trifluoromethylation of various primary and secondary aliphatic carboxylic acids. With AgNO3 as a catalyst, (bpy)Cu(CF3)3 (bpy = 2,2'-bipyridine) as a CF3 source and K2S2O8 as an oxidant, aliphatic carboxylic acids were converted to the corresponding trifluoromethylated products in good yields (Scheme 60). Also, mechanistic studies, a radical clock experiment, revealed the intermediacy of −Cu(CF3)3Me, which undergoes reductive elimination and subsequent oxidation to give the active species Cu(CF3)2. Meanwhile, aliphatic carboxylic acids give the corresponding alkyl radicals via Ag(II)-mediated oxidative decarboxylation. Then, Cu(CF3)2 provides a CF3 group to alkyl radicals to obtain the final product.</p><!><p>Ag-catalyzed decarboxylative trifluoromethylation of aliphatic carboxylic acids in aqueous CH3CN.</p><!><p>Very recently, MacMillan et al. [120] discovered an efficient approach to the decarboxylative trifluoromethylation of aliphatic carboxylic acids via the combination of photoredox and copper catalysis (Scheme 61). The method tolerates a myriad of primary, secondary and tertiary carboxylic acids and provides the corresponding CF3 analogue in good to excellent yields. Details of the proposed dual copper–photoredox cycle are shown in Scheme 61. The Ir(III) photocatalyst Ir[dF(CF3)ppy]2(4,4'-dCF3bpy)PF6 (1) undergoes photoexcitation with visible light to form the highly oxidizing excited state ·Ir(III) 2. Then, SET from copper carboxylate 4, derived from carboxylic acid 3 with the Cu(II) catalyst to ·Ir(III) 2 provides Cu(III) carboxylate 5, or in the dissociated form, a carboxyl radical and Cu(II) complex 6, along with reduced Ir(II) photocatalyst 7. The resulting carboxyl radical extrudes CO2 and sequentially recombines to generate Cu(III) species 9. At this stage, SET from 7 to 9 closes the photoredox catalytic cycle and produces an alkylcopper(II) species 10. Under the addition of Togni's reagent I (11), species 10 affords the final alkyl−CF3 product and complex 13, which is used for ligand exchange with 3.</p><!><p>Decarboxylative trifluoromethylation of aliphatic carboxylic acids via combined photoredox and copper catalysis.</p><!><p>Palladium-catalyzed trifluoromethylation of aryl and vinyl compounds: In 2010, Watson and co-workers [121] developed the first Pd-catalyzed trifluoromethylation of aryl/heterocyclic chlorides with the CF3 source TESCF3 (TES, triethylsilyl), which proceeded following a classical Pd(0)/Pd(II) catalytic cycle (Scheme 62). Also, the reaction tolerates a variety of functional groups, such as esters, amides, ethers, nitriles, etc., and therefore provides a new way for late-stage modifications.</p><!><p>Palladium-catalyzed Ar–CF3 bond-forming reaction.</p><!><p>In the same year, Yu's group [122] reported a Pd(II)-catalyzed C–H trifluoromethylation of arenes with an electrophilic trifluoromethylation reagent using diverse heterocyclic directing groups. Notably, the presence of trifluoroacetic acid (TFA) is crucial for the Ar–CF3 bond formation and Cu(OAc)2 can increase the catalytic turnover (Scheme 63). Based on three different modes of the ArPd(II) species reaction with nucleophiles, electrophiles and highly oxidizing reagents, three possible reaction pathways (A, B and C, respectively) are envisaged, that can follow the C–H activation event to the trifluoromethylated products, as described in Scheme 63. In this case, the specific catalytic mechanism remains to be studied.</p><!><p>Palladium-catalyzed trifluoromethylation of arenes with diverse heterocyclic directing groups.</p><!><p>In 2011, the group of Liu [123] accomplished a Pd(II)-catalyzed oxidative trifluoromethylation of indoles with TMSCF3 and PhI(OAc)2 at room temperature (Scheme 64). Through reductive elimination from the (Ar)Pd(IV)-CF3 intermediate, the aryl C–CF3 bond is generated. Notably, the bidentate nitrogen-containing ligand is crucial to the achievement of this process.</p><!><p>Pd-catalyzed trifluoromethylation of indoles as reported by Liu.</p><!><p>In the same year, Buchwald et al. [124] discovered a palladium-catalyzed trifluoromethylation of vinyl triflates and nonaflates (Scheme 65). A variety of trifluoromethylated cyclohexenes were obtained using a catalyst system, which was composed of Pd(dba)2 or [(allyl)PdCl]2 and the monodentate biaryl phosphine ligand t-BuXPhos. Also, TMSCF3 and KF were more suitable to the trifluoromethylation of triflate electrophiles, while the use of TESCF3 and RbF gave better results for nonaflate electrophiles.</p><!><p>Pd-catalyzed trifluoromethylation of vinyl triflates and vinyl nonaflates.</p><!><p>Subsequently, the Yu [15,125] and Shi group [126] independently reported the palladium-catalyzed ortho-trifluoromethylation of an aromatic C–H bond with Umemoto's trifluoromethylation reagent. Notably, Cu(II) salts were crucial for forming the aryl–CF3 bonds. In Yu's study, benzamides and benzylamines were well trifluoromethylated via a Pd(II)/Pd(IV) catalytic cycle with the addition of TFA (and Ag2O) (Scheme 66a). With an acetamido group as a directing group, Shi developed an efficient method to access ortho-CF3 acetanilides and anilines (Scheme 66b).</p><!><p>Pd(II)-catalyzed ortho-trifluoromethylation of aromatic C–H bonds.</p><!><p>Recently, Wang and co-workers [127] reported a visible-light-induced Pd-catalyzed ortho-trifluoromethylation of acetanilides. Without the need of an external photocatalyst and additive, various N-substituted anilides and acetanilides were obtained efficiently at room temperature in air. The strategy features good yields, broad functional group tolerance and high regioselectivity (Scheme 67).</p><!><p>Visible-light-induced Pd(OAc)2-catalyzed ortho-trifluoromethylation of acetanilides with CF3SO2Na.</p><!><p>Copper-catalyzed trifluoromethylation of aryl- and alkenylboronic acids: In 2011, Liu and Shen [128] developed a CuI-catalyzed method for the trifluoromethylation of aryl- and alkenylboronic acids with Togni's reagent (Scheme 68). A range of different substrates gave the corresponding trifluoromethylated (hetero)arenes in good to excellent yields.</p><!><p>CuI-catalyzed trifluoromethylation of aryl- and alkenylboronic acids.</p><!><p>Also, in 2012, Beller and co-workers [129] described a copper-catalyzed trifluoromethylation of aryl- and vinylboronic acids with the generation of CF3-radicals at room temperature. The mild reaction conditions allowed a wide variety of functional groups to be tolerated, though a large quantity of TBHP was required (Scheme 69). Notably, the authors proposed two mechanistic pathways for this trifluoromethylation reaction. The difference between path A and path B is that the sequence of CF3 radicals and aryl- and vinylboronic acids is reversed. In addition, the CF3 radical is generated from the reaction of TBHP with NaSO2CF3.</p><!><p>Cu-catalyzed trifluoromethylation of aryl- and vinylboronic acids.</p><!><p>Copper-catalyzed trifluoromethylation of alkenes: The method described by Hu [130] was applied to the trifluoromethylation of a wide range of α,β-unsaturated carboxylic acids through CuF2-catalyzed decarboxylative fluoroalkylation with high yields and excellent E/Z ratio (Scheme 70).</p><!><p>Copper-catalyzed trifluoromethylation of α,β-unsaturated carboxylic acids.</p><!><p>Additionally, a copper(I)-catalyzed trifluoromethylation of alkenes was disclosed by Sodeoka and co-workers in 2012 [131]. The reaction was carried out with Togni's reagent as the CF3 source and TsOH as a Brønsted acid in CH2Cl2 at 40 °C (Scheme 71). Notably, trifluoromethylstyrenes were formed through further transformations of the oxytrifluoromethylated products with high efficiency.</p><!><p>Formation of C(sp2)–CF3 bond catalyzed by copper(I) complex.</p><!><p>In the same year, Loh's group [132] used the same copper catalyst and Togni's reagent to achieve the trifluoromethylation of enamides in good yields at room temperature (Scheme 72a). Meanwhile, this reaction exhibited excellent stereoselectivity towards the E-isomer. One year later, the same group [133] extended this approach to the directing-group-assisted copper-catalyzed trifluoromethylation of electron-deficient alkenes (Scheme 72b). Moreover, α-aryl and α-alkyl-substituted acrylate derivatives could be used as substrates to form the C(sp2)–CF3 bond with a complete Z-selectivity. A radical species participated in the reaction's catalytic cycle.</p><!><p>Loh's Cu(I)-catalyzed trifluoromethylation of enamides and electron-deficient alkenes.</p><!><p>In 2013, the group of Liu [134] described a copper-catalyzed decarboxylative trifluoromethylation of α,β-unsaturated carboxylic acids with CF3SO2Na. This method was applied to a wide range of α,β-unsaturated carboxylic acids. Meanwhile, a similar radical process for the difluoromethylation of aryl-substituted acrylic acids was also achieved by Liu and co-workers. The HCF2-substituted E-alkenes were finally obtained with iron catalysis and zinc difluoromethanesulfinate ((CF2HSO2)2Zn, Baran reagent). Also, the authors proved that the formation of the Cvinyl–CF3/Cvinyl–CF2H bonds followed a radical addition–elimination process (Scheme 73).</p><!><p>Copper and iron-catalyzed decarboxylative tri- and difluoromethylation.</p><!><p>Subsequently, Bouyssi and co-workers [135–137] used Togni's reagent to conduct the trifluoromethylation of (hetero)aromatic aldehydes or corresponding N,N-dialkylhydrazones with CuCl as the catalyst at room temperature (Scheme 74). These reactions showed a broad substrate scope and good functional group compatibility with up to 99% yield.</p><!><p>Cu-catalyzed trifluoromethylation of hydrazones developed by Bouyssi.</p><!><p>In 2013, a simple and effective copper-catalyzed approach for the construction of Cvinyl–CF3 bonds without using pre-functionalized substrates was reported by Xiao et al. (Scheme 75) [138]. The process proceeded smoothly to give the trifluoromethylated alkenes in good to excellent yields via a radical mechanism.</p><!><p>Cu(I)-catalyzed trifluoromethylation of terminal alkenes.</p><!><p>Additionally, Duan and co-workers [139] discovered a copper/silver-catalyzed decarboxylative trifluoromethylation of α,β-unsaturated carboxylic acids with CF3SO2Na. This reaction proceeded efficiently for a wide range of alkyl and aryl-substituted α,β-unsaturated carboxylic acids derivatives with excellent E/Z selectivity (Scheme 76). It's worth mentioning that the addition of Ag2CO3 additives was crucial for promoting the decarboxylation of α,β-unsaturated carboxylic acids.</p><!><p>Cu/Ag-catalyzed decarboxylative trifluoromethylation of cinnamic acids.</p><!><p>In 2014, a Cu(I/II)-catalyzed α-trifluoromethylation of α,β-unsaturated carbonyl compounds were unfolded by the Bi group (Scheme 77) [140]. The reaction was applied to a broad range of carbonyl compounds, including enones, α,β-unsaturated esters, thioesters, and amides. Notably, the authors obtained products with stable E-configuration through a SET process.</p><!><p>Copper-catalyzed direct alkenyl C–H trifluoromethylation.</p><!><p>In 2017, Loh and co-workers [141] introduced a Cu(I/II)-catalyzed Cvinyl–H trifluoromethylation of a variety of styrene derivatives. This process was achieved by using 1-methylimidazole (NMI) as ligand and tetrabutylammonium iodide (TBAI) as an additive (Scheme 78). Mechanistic studies revealed that this reaction probably proceeds through a radical pathway.</p><!><p>Copper(I/II)-catalyzed direct trifluoromethylation of styrene derivatives.</p><!><p>Copper-catalyzed trifluoromethylation of arenes and heteroarenes: In 2013, Xi et al. [142] reported a CuCl-catalyzed direct trifluoromethylation of sp2 C–H bonds with Togni reagent (Scheme 79). Also, phenyl, thiophene, and pyridine derivatives achieved regioselectively trifluoromethylation with N-pivalamide as a directing group. The authors also proposed a possible radical pathway for this reaction. The final trifluoromethylated compounds were generated from pivalamido arenes and heteroarenes with the CF3 radical through a Cu(I/II) catalytic cycle.</p><!><p>Regioselective trifluoromethylation of pivalamido arenes and heteroarenes.</p><!><p>In 2013, the Szabó [143] and Wang group [144] described the copper-mediated C–H trifluoromethylation of quinones with Togni's reagent. Szabó utilized a stoichiometric amount of CuCN combined with catalytic bis(pinacolato)diboron, whereas Wang applied a stoichiometric amount of CuI. Notably, both groups proved a mechanism involving the formation of a CF3 radical with copper(I) acting as a one-electron reducing agent (Scheme 80).</p><!><p>Synthesis of trifluoromethylquinones in the presence of copper(I).</p><!><p>With catalytic cupric acetate and TBHP, the group of Tang [145] developed a green strategy for the trifluoromethylation of imidazoheterocycles with a recyclable mixed medium of 1-butyl-3-methylimidazolium tetrafluoroborate ([Bmim]BF4) and water (Scheme 81). By following this method, diverse trifluoromethylated imidazoheterocycles were obtained in up to 80% yield. The method features a green and recyclable solvent, mild reaction conditions (room temperature) and excellent functional group tolerance. In this instance, the copper catalyst may only promote the generation of the tert-butoxyl radical from TBHP. The oxidation of the intermediate A with t-BuOOH produces a carbocation B, followed by an oxidative dehydrogenation process to afford the target product.</p><!><p>Oxidative trifluoromethylation of imidazoheterocycles in ionic liquid/water.</p><!><p>Also in 2015, Li and co-workers [146] developed a mild and fast Cu(I/II)-catalyzed trifluoromethylation procedure to obtain 3-trifluoromethylcoumarins. The reaction was carried out with a CuCl/CF3SO2Na/TBHP system under continuous-flow conditions, affording the corresponding products with wide substrate tolerance in moderate to good yields (Scheme 82).</p><!><p>A mild and fast continuous-flow trifluoromethylation of coumarins using a CuI/CF3SO2Na/TBHP system.</p><!><p>After one year, the group of Cai [147] presented a Cu(II)-catalyzed 8-amido chelation-induced regioselective C5-trifluoromethylation of quinolines (Scheme 83a). With CuBr2 as a catalyst and azobisisobutyronitrile (AIBN) as an oxidant, a wide range of functional groups were well tolerated to provide the products in moderate to excellent yields. Simultaneously, Zhang and co-workers [148] described a similar, milder regioselective C–H trifluoromethylation of 8-aminoquinolines by using a chitosan-based heterogeneous copper catalyst (CS@Cu(OAc)2, CS = chitosan) (Scheme 83b).</p><!><p>Copper-catalyzed oxidative trifluoromethylation of various 8-aminoquinolines.</p><!><p>Recently, a picolinamide (PA)-directed method for the Cu-catalyzed trifluoromethylation of anilines was described by the group of Zhang [149]. The trifluoromethyl group was installed at the ortho position of the substrate, yielding 2-(trifluoromethyl)aniline derivatives in moderate to good yields (Scheme 84). Notably, the directing group could be recovered in excellent yields and this approach provided a new way for the efficient synthesis of floctafenine via a single-electron-transfer mechanism.</p><!><p>PA-directed copper-catalyzed trifluoromethylation of anilines.</p><!><p>Vinyl C–CF3 bond formation using Fe, Ir, Ru, and Ag catalysts: In 2012, Buchwald and co-workers [150] unfolded an iron(II)-catalyzed trifluoromethylation of potassium vinyltrifluoroborates at room temperature (Scheme 85). With this approach, 2-arylvinyl substrates, in particular, furnished the products in good yields and excellent E/Z ratios (E/Z > 95.5%).</p><!><p>Trifluoromethylation of potassium vinyltrifluoroborates catalyzed by Fe(II).</p><!><p>Also, Cho and co-workers [151] reported a direct method for an alkenyl trifluoromethylation employing a Ru photocatalyst. The method used CF3I as a CF3 radical source and 1,8-diazabicyclo[5.4.0]undec-7-ene (DBU) as the base (Scheme 86). Under these mild reaction conditions, the trifluoromethylation of a wide range of alkenes shows high functional-group tolerance with a low catalyst loading. Moreover, compared with other alkenes, this process works especially well for terminal alkenes.</p><!><p>Alkenyl trifluoromethylation catalyzed by Ru(phen)3Cl2 as photocatalyst.</p><!><p>In 2013, Akita's group [152] developed a radical-mediated trifluoromethylation of vinyltrifluoroborates promoted by the photoredox catalyst [Ru(bpy)3](PF6)2 under visible light irradiation (Scheme 87a). The trifluoromethylated alkenes were obtained in up to 95% yield. One year later, the same group [153] presented a procedure for trifluoromethylation of multisubstituted alkenes with a different CF3 source, Umemoto's reagent (Scheme 87b). Additionally, this reaction could be extended to double trifluoromethylation.</p><!><p>Ru-catalyzed trifluoromethylation of alkenes by Akita's group.</p><!><p>In 2014, a visible-light-induced decarboxylative trifluoromethylation of α,β-unsaturated carboxylic acids by using [Ir(ppy)3] as a photoredox catalyst was explored by Zhu and co-workers (Scheme 88) [154]. Notably, this procedure employed only 1 mol % catalyst loading to achieve an excellent reactivity and E/Z stereoselectivity at room temperature.</p><!><p>Ir-catalyzed Cvinyl–CF3 bond formation of α,β-unsaturated carboxylic acids.</p><!><p>In 2016, Duan and co-workers [155] disclosed a Ag(I)-catalyzed denitration/trifluoromethylation of β-nitrostyrenes with CF3SO2Na, which employed a large excess of di-tert-butyl peroxide (DTBP) as the oxidant and tetrabutylammonium iodide (TBAI) as phase-transfer catalyst (Scheme 89). Notably, only (E)-isomers of the products were obtained in moderate to high yields.</p><!><p>Ag(I)-catalyzed denitrative trifluoromethylation of β-nitrostyrenes.</p><!><p>Various transition-metal-catalyzed direct C–H bond trifluoromethylation of arenes and heteroarenes: In 2011, the group of MacMillan [156] reported a simple approach for the direct trifluoromethylation of unactivated arenes and heteroarenes through a radical-mediated mechanism (Scheme 90). Under exposure to 26 W fluorescent light, this process proceeded well in the presence of triflyl chloride and different photocatalysts depending on the substrate's nature, i.e., Ru(phen)3Cl2 for 5-membered heterocycles, Ir(Fppy)3 for 6-membered arenes and heterocycles. It is worth mentioning, that triflyl chloride provides a cheap and easy to handle CF3 source.</p><!><p>Photocatalyzed direct trifluoromethylation of aryl and heteroaryl C–H bonds.</p><!><p>A mild and simple electrophilic trifluoromethylation of various aromatic and heteroaromatic compounds was disclosed by the Togni group [157] in 2012. The authors used methyltrioxorhenium (MTO) as the catalyst (Scheme 91). Notably, the direct aromatic trifluoromethylation tolerates a broad substrate scope, however, is limited to electron-rich substrates.</p><!><p>Rhenium (MTO)-catalyzed direct trifluoromethylation of aromatic substrates.</p><!><p>In 2014, Ma et al. [158] developed the first visible-light-promoted radical trifluoromethylation of unprotected anilines. With [Ir(ppy)3] and Togni's reagent, the method afforded various fluorine-containing molecules and heterocyclic compounds at room temperature (Scheme 92).</p><!><p>Trifluoromethylation of unprotected anilines under [Ir(ppy)3] catalyst.</p><!><p>In 2015, the group of Hajra [159] described a method for the direct trifluoromethylation of imidazopyridines and other imidazoheterocycles. The CF3SO2Na/t-BuOOH/Ag system enables accomplishing the reaction at room temperature under ambient air (Scheme 93).</p><!><p>Oxidative trifluoromethylation of imidazopyridines and imidazoheterocycles.</p><!><p>A direct trifluoromethylation of (hetero)arenes in the presence of only 0.1 mol % [Ru(bpy)3]Cl2 as catalyst was reported by the Stephenson group in 2016 [160]. Notably, the authors utilized pyridine N-oxide derivatives in concert with trifluoroacetic anhydride to facilitate this process (Scheme 94). Moreover, the method has been successfully extended on a kilogram scale.</p><!><p>Ruthenium-catalyzed trifluoromethylation of (hetero)arenes with trifluoroacetic anhydride.</p><!><p>One year later Mizuno's group [161] introduced a direct C–H trifluoromethylation of (hetero)arenes with O2 as the terminal oxidant in the presence of catalytic amounts of phosphovanadomolybdic acids (Scheme 95). The reaction tolerated diverse (hetero)arenes to afford the corresponding trifluoromethylated products via a radical pathway in 26–92% yields.</p><!><p>Phosphovanadomolybdic acid-catalyzed direct C–H trifluoromethylation.</p><!><p>In 2017, Zhang and co-workers [162] were the first who reported a nickel(II)-catalyzed and picolinamide-assisted site-selective C–H bond trifluoromethylation of arylamines in water (Scheme 96a). This strategy displays several advantages: 1) inexpensive nickel catalyst, 2) recyclable catalyst, 3) aqueous phase reaction, and 4) high site selectivity. Only one year later, the group of Xia optimized this approach and established a convenient, oxidant-free protocol for the ortho-trifluoromethylation of arylamine under ultraviolet irradiation (Scheme 96b) [163].</p><!><p>Picolinamide-assisted ortho-trifluoromethylation of arylamines.</p><!><p>In 2018, Wu and co-workers [164] introduced a one-step strategy for the synthesis of trifluoromethylated free anilines using Togni's reagent using a nickel-catalyzed C–H trifluoromethylation. Moreover, free anilines with a variety of functional groups were trifluoromethylated under the mild reaction conditions in up to 90% yield (Scheme 97).</p><!><p>A nickel-catalyzed C–H trifluoromethylation of free anilines.</p><!><p>In 2010, Qing's group [165] reported the first example of a copper-mediated trifluoromethylation of terminal alkynes. Notably, the reaction was carried out with nucleophilic (trifluoromethyl)trimethylsilane (Me3SiCF3) as a CF3 source under air atmosphere (Scheme 98a). Moreover, this protocol was compatible with various terminal alkynes, such as aromatic and aliphatic alkynes, affording the trifluoromethylated alkynes in 47–91% yields. Subsequently, Qing [166] developed an efficient catalytic trifluoromethylation by adding terminal alkynes and Me3SiCF3 slowly with a syringe pump. Two years later, the same group [167] presented an improved Cu-mediated oxidative trifluoromethylation of aryl and heteroaryl terminal alkynes. In the latter case, the trifluoromethylation proceeded at room temperature by using Ag2CO3 as an oxidant with a significantly lower amount of TMSCF3 (Scheme 98b).</p><!><p>Cu-mediated trifluoromethylation of terminal alkynes reported by Qing.</p><!><p>In 2012, Huang et al. [168] reported a process for trifluoromethylation of terminal alkynes with Togni's reagent in DCM at room temperature (Scheme 99a). The trifluoromethylated acetylenes were obtained with up to 98% yield via a Cu(I/III) catalytic cycle with CF3+. As an extension of their work, this group [169] developed the trifluoromethylation of alkynyltrifluoroborates to form trifluoromethylated acetylenes under similar conditions without the addition of bases (Scheme 99b).</p><!><p>Huang's C(sp)–H trifluoromethylation using Togni's reagent.</p><!><p>In the same year, the groups of Guo [170] and Xiao [171] also developed a copper(I)-catalyzed trifluoromethylation of terminal alkynes with Umemoto's reagent as an electrophilic CF3 source (Scheme 100a). Compared with the reaction conditions reported by Guo, Xiao's method was carried out at higher temperature, using similar copper(I) catalysts, but with different ligands (Scheme 100b).</p><!><p>Cu-catalyzed methods for trifluoromethylation with Umemoto's reagent.</p><!><p>In 2014, the trifluoromethylation of aromatic alkynes through visible-light photoredox catalysis was described by Cho and co-workers [172]. With fac-[Ir(ppy)3] as photocatalyst and KOt-Bu as a base, the reaction was achieved under blue LED irradiation in moderate yields (Scheme 101). However, this approach was not suitable for aliphatic alkynes.</p><!><p>The synthesis of alkynyl-CF3 compounds in the presence of fac-[Ir(ppy)3] under visible-light irradiation.</p><!><p>Compared with the trifluoromethylation and fluorination mentioned above, the methodological research on difluoromethylation, trifluoromethylthiolation and trifluoromethoxylation of organic molecules are quite rare and scattered. Here we summarize the new developments within recent years.</p><!><p>The introduction of a difluoromethylene (CF2) group into organic molecules can significantly improve their metabolic stability and oral bioavailability [3]. Therefore, the difluoroalkylation has become a powerful strategy for regulating the biological activity of organic molecules. It is noteworthy that transition-metal-catalyzed difluoroalkylation is an effective route to obtain these valuable difluoroalkylated backbones. There are four modes of catalytic difluoroalkylation, including nucleophilic difluoroalkylation, electrophilic difluoroalkylation, radical difluoroalkylation, and metal-difluorocarbene coupling (MeDiC) [173]. Finally, a wide range of difluoroalkylated (hetero)arenes [(Het)Ar-CF2R, R = PO(OEt)2, CO2Et, CONR1R2, COR1, (Het)Ar, alkenyl, alkynyl, alkyl, H] and alkenes were obtained with excellent functional group tolerance.</p><p>In 2012, the Reutrakul group [174] firstly reported a Pd-catalyzed Heck-type reaction of [(bromodifluoromethyl)sulfonyl]benzene with styrene derivatives (Scheme 102). Notably, the reaction shows a broad substrate scope, including a variety of styrene derivatives, vinyl ethers, vinyl sulfides, and a few heteroaromatic substrates.</p><!><p>Pd-catalyzed Heck reaction reported by Reutrakul.</p><!><p>In the same year, Yu and co-workers [175] developed an iridium-catalyzed direct C–H functionalization of enamides and ene-carbamates with BrCF2CO2Et under visible-light photoredox conditions (Scheme 103). This method shows excellent yields and a wide substrate scope.</p><!><p>Difluoromethylation of enamides and ene-carbamates.</p><!><p>Moreover, Hu et al. [130] established a copper-catalyzed (phenylsulfonyl)-difluoromethylation of α,β-unsaturated carboxylic acids with excellent E/Z selectivity (Scheme 104). Notably, the Lewis acid (CuF2·2H2O) was used to enhance the electrophilicity of the Togni's reagent and to promote the decarboxylation of the carboxylic acids. The authors proposed that, under these conditions, the Togni's reagent may undergo a Cu-catalyzed bond cleavage to produce the highly electrophilic iodonium salt A, which then coordinates to the carboxylic acid functionality to generate the intermediate B. The latter then undergoes – through an intramolecular reaction – decarboxylation and reductive elimination to afford the species E and the desired product. Finally, species E reacts with HF regenerating the catalyst.</p><!><p>Difluoromethylation of α,β-unsaturated carboxylic acids.</p><!><p>In 2013, Pannecoucke and co-workers [176] developed a copper-catalyzed regioselective difluoroacetylation of dihydropyrans and glycals on the C-2 position. Notably, the corresponding products were obtained through a Cu(I/III) catalytic cycle without the involvement of radicals (Scheme 105a). Hence, in 2014, the same group extended this method to the olefinic difluoroacetylation of enamides [177]. In this reaction, they obtained the β-difluoroester-substituted enamides under operationally simple and mild conditions. Also, the method has a broad substrate scope, including cyclic and acyclic enamides (Scheme 105b).</p><!><p>Copper-catalyzed direct C(sp2)–H difluoroacetylation reported by Pannecoucke and co-workers.</p><!><p>In 2016, a Pd-catalyzed direct difluoroalkylation of aldehyde hydrazones with functionalized difluoromethyl bromides was described by Monteiro's group (Scheme 106a) [178]. The bromodifluoromethylated compounds are effective reagents for the difluoromethylation of aldehyde-derived hydrazones to the corresponding difluoromethyl ketone hydrazones. However, this strategy relies on the use of an expensive palladium/ligand catalyst system that makes it less attractive and practical. Subsequently, the same group [179] found that CuCl could also effectively catalyze the difluoromethylation of hydrazones. This method provided an efficient, cost-effective protocol for the multigram-scale preparation of functionalized difluoromethylketone hydrazines (Scheme 106b).</p><!><p>Difluoroalkylation of aldehyde-derived hydrazones with functionalized difluoromethyl bromides.</p><!><p>Compared with Monteiro's approaches, Zhu and co-workers [180] were the first who developed a visible-light-induced direct C–H-bond difluoroalkylation of aldehyde-derived hydrazones (Scheme 107a). Importantly, this unprecedented photoredox protocol is enabled by a novel aminyl radical/polar crossover mechanism. Meanwhile, a first gold-catalyzed photoredox difluoroalkylation of aromatic aldehyde hydrazones under sunlight was reported by Hashmi's group (Scheme 107b) [181]. Both methods smoothly work at room temperature affording the products with modest to excellent yields.</p><!><p>Photoredox-catalyzed C–H difluoroalkylation of aldehyde-derived hydrazones.</p><!><p>One year later, Ackermann and co-workers [182] presented a ruthenium(II)-catalyzed meta-selective C–H difluoromethylation with the cooperation of phosphine and carboxylate. This protocol is compatible with a variety of functional groups, such as pyridyl, pyrimidyl, pyrazolyl, and even purinyl assistance (Scheme 108).</p><!><p>Synergistic ruthenium(II)-catalyzed C–H difluoromethylation reported by Ackermann.</p><!><p>A visible-light photocatalytic decarboxylation strategy for the synthesis of difluoromethylated styrenes with fac-Ir(ppy)3 and BrCF2CO2Et was developed by Noël et al. in 2017 [183]. Herein, meta and para-substituted cinnamic acids afforded the expected E-isomers, while ortho-substituted cinnamic acids selectively provided the less stable Z-product. Notably, the conversion of the Z-isomer into the E-isomer was achieved by controlling the reaction time accurately. Furthermore, arylpropiolic acids could also be decarboxylative difluoromethylated by this method (Scheme 109).</p><!><p>Visible-light photocatalytic decarboxylation of α,β-unsaturated carboxylic acids.</p><!><p>Meanwhile, the group of Dilman [184] developed a method for the synthesis of difluorinated ketones with gem-difluorinated organozinc reagents (Scheme 110). Firstly, acyl chlorides reacted with potassium dithiocarbamate to generate S-acyl dithiocarbamates. Subsequently, the so-obtained dithiocarbamates were coupled with organozinc to produce the desired difluorinated ketones.</p><!><p>Synthesis of difluorinated ketones via S-alkyl dithiocarbamates obtained from acyl chlorides and potassium dithiocarbamate.</p><!><p>Additionally, Poisson and co-workers [185] developed a simple and efficient way to access various aryl and heteroaryl difluoromethylated phosphonates under mild conditions. The reaction proceeds smoothly with CuCF2PO(OEt)2 and a palladium catalyst in MeCN (Scheme 111). This transformation enabled the functionalization of various less reactive substrates, such as phenol, boronate, ketones, nitriles, esters, etc.</p><!><p>Synthesis of aryl and heteroaryl difluoromethylated phosphonates.</p><!><p>Notably, the above-mentioned copper-catalyzed highly stereoselective trifluoromethylation reaction of secondary propargyl sulfonates developed by Zhang [116] could also be extended to stereospecific propargylic difluoroalkylation (Scheme 112). In this reaction trimethylsilyldifluoroamide (TMSCF2CONEt2) is chosen as the difluoroalkylating reagent and proceeds under mild reaction conditions with high regioselectivity and stereospecificity (ee up to 99%).</p><!><p>Difluoroalkylation of secondary propargyl sulfonates using Cu as the catalyst.</p><!><p>In 2018, Zhao et al. [186–187] disclosed a ruthenium(II)-enabled para-selective C–H difluoromethylation of ketoxime ethers, anilides, indolines and tetrahydroquinolines (Scheme 113). The protocol is compatible with various functional groups, furnishing the para-difluoromethylated products in moderate to good yields. Moreover, chelation-assisted cycloruthenation plays a key role in the selective activation of para-CAr–H bonds.</p><!><p>Ru(II)-mediated para-selective difluoromethylation of anilides and their derivatives.</p><!><p>Subsequently, the Zhang group [188] disclosed an iron-catalyzed cross-coupling of a wide range of arylmagnesium and difluoroalkyl bromides with modified N,N,N',N'-tetramethyl-ethane-1,2-diamine (TMEDA) as a ligand. Notably, this bulky diamine is critical to improve the catalytic efficiency and suppress the side reaction of defluorination (Scheme 114).</p><!><p>Bulky diamine ligand promoted cross-coupling of difluoroalkyl bromides.</p><!><p>Recently, a synthetic method for difluoroacetylated quinoxalin-2(1H)-one derivatives was presented by the same group [189]. The direct difluoroacetylation of diverse quinoxalinones with a wide range of functional groups proceeded regioselectively at the C-3 position with ethyl bromodifluoroacetate under copper catalysis (Scheme 115).</p><!><p>Copper-catalyzed C3–H difluoroacetylation of quinoxalinones.</p><!><p>In the past few years, many methods for the direct introduction of trifluoromethylthio groups into organic compounds have been reported. Depending on the type of the trifluoromethylthiolating reagent used in the chemical conversion, the methods can also be classified into three classes: radical, nucleophilic and electrophilic trifluoromethylthiolation. Some reviews in this area focused on several aspects such as the syntheses of aromatic and heterocyclic perfluoroalkyl sulfides [190], direct trifluoromethylthiolation reactions [191–194], sulfur-based fluorination and fluoroalkylation reagents [195], trifluoromethylthio cation-donating ability (Tt+DA) [196], or synthetic methods leading to compounds containing CF3–S units [197]. Herein, based on the transition-metal catalysis, recent research advances in these types of synthetic methods are described in this section of this review.</p><p>In 2013, Shibata's group [198] developed an electrophilic trifluoromethansulfonyl hypervalent iodonium ylide for the trifluoromethylthiolation of enamines, indoles and β-ketoesters catalyzed by copper(I) chloride (Scheme 116). The desired CF3S-substituted products were formed with good yields in short times at room temperature.</p><!><p>Copper(I) chloride-catalyzed trifluoromethylthiolation of enamines, indoles and β-ketoesters.</p><!><p>In 2014, Gade et al. [199] applied a copper-boxmi complex as highly enantioselective catalyst to effect electrophilic trifluoromethylthiolations (Scheme 117). A number of α-SCF3-substituted β-ketoesters were obtained in good yields with high enantiomeric excess (ee) under mild conditions.</p><!><p>Copper-boxmi-catalyzed asymmetric trifluoromethylthiolation of β-ketoesters.</p><!><p>The group of Rueping [200] employed N-(trifluoromethylthio)phthalimide as an electrophilic source of F3CS+ for the direct trifluoromethylthiolation of boronic acids and alkynes under copper catalysis in 2014 (Scheme 118). Based on the mild conditions, this approach features high functional group tolerance and a broad substrate scope.</p><!><p>Direct Cu-catalyzed trifluoromethylthiolation of boronic acids and alkynes.</p><!><p>In the same year, a powerful protocol for the direct synthesis of α-trifluoromethylthio-substituted ketones was reported by Weng and co-workers [201]. Notably, the trifluoromethylthiolation reactions of primary and secondary α-bromoketones worked well with CF3SiMe3 and elemental sulfur as precursors (Scheme 119). Furthermore, this strategy shows a broad substrate scope and tolerates a variety of functional groups.</p><!><p>Cu-catalyzed synthesis of α-trifluoromethylthio-substituted ketones.</p><!><p>In 2016, a variety of enamines, indoles, β-keto esters, pyrroles, and anilines were trifluoromethylthiolated efficiently by Shibata's group [202] in the presence of diazotriflone and copper catalysis through an electrophilic-type reaction (Scheme 120).</p><!><p>Trifluoromethylthiolation reactions promoted by diazotriflone and copper.</p><!><p>In 2016, Glorius et al. [203] introduced the synthesis of vinyl-SCF3 compounds using N-(trifluoromethylthio)phthalimide as SCF3 source under blue LEDs irradiation. Notably, a variety of alkenes could be converted to vinyl-SCF3 compounds with the cooperation of an [Ir] photocatalyst and an ammonium bromide salt (Scheme 121). The formed adduct could be triggered via an oxidative quenching cycle and delivered a SCF3 radical. Under similar conditions, they also demonstrated a tandem photoinduced trifluoromethylthiolation/semi-pinacol-type rearrangement of ketones.</p><!><p>Halide activation of N-(trifluoromethylthio)phthalimide.</p><!><p>Meanwhile, the same group [204] developed a visible-light-mediated trifluoromethylthiolation of alkyl carboxylic acids with [Ir] and phtalimide-SCF3 as the trifluoromethylthiolating reagent (Scheme 122a). Moreover, tertiary, secondary, and primary alkyl carboxylic acids afforded the desired products in good to excellent yields. Notably, the use of an external sacrificial hydrogen atom donor, mesitylene or methyl (3-methyl)benzoate, avoided the bistrifluoromethylthiolation reaction in this process. Subsequently, the group [205] demonstrated that the Phth-SCF3 reagent could also be used for the direct trifluoromethylthiolation of C–H bonds under similar mild conditions (Scheme 122b). Also, a wide range of aliphatic substrates were converted to their trifluoromethylthiolated analogues in very good yields.</p><!><p>The visible light-promoted trifluoromethylthiolation reported by Glorius.</p><!><p>Additionally, the group of Goossen [206] disclosed a simple and practical strategy for the conversion of α-diazo esters to the corresponding trifluoromethylthiolated esters with a Me4NSCF3 salt. In the presence of copper thiocyanate, this transformation afforded the products with up to 98% yields at room temperature and was applied to a wide range of easily available α-diazo esters (Scheme 123).</p><!><p>Synthesis of α-trifluoromethylthioesters via Goossen's approach.</p><!><p>Recently, the formation of arenes-SCF3 was shown by the Jacobi von Wangelin group [207], the Zhao group [208] as well as the Tlili group [193] (Scheme 124). All three methodologies have been developed based on a [Ru]-based photocatalyst under LED irradiation. The group of Jacobi von Wangelin used the bis(trifluoromethyl) disulfide (CF3SSCF3) as the source of trifluoromethyl sulfide, while the other two groups employed shelf-stable reagents C and D for trifluoromethylthiolation, arenesulfonate-SCF3.</p><!><p>Photoinduced trifluoromethylthiolation of diazonium salts.</p><!><p>The introduction of a trifluoromethoxy (OCF3) group into a molecule can improve its metabolic stability and membrane permeability. Some strategies for the synthesis of trifluoromethoxylated compounds have been reviewed: Poisson [209] and Billard [210] discussed the recent advances toward the synthesis of OCF3-containing molecules; Hopkinson [211] depicted a radical revolution for trifluoromethoxylation; recently, Ngai [212] summarized some photoredox-based approaches to form tri- and difluoromethoxylated compounds. Despite the great interest in this functional group, only a few transition-metal-catalyzed methods have been developed for the synthesis of trifluoromethoxylated compounds over the past decade. This may be due to the fact that C–OCF3 bond formation reactions have many limitations, including reversible decomposition of the trifluoromethoxide anion in solution above room temperature to afford carbonic difluoride and fluoride [213–214], as well as β-fluoride elimination from transition-metal-trifluoromethoxide complexes [215–216].</p><p>In 2011, the first report of a transition-metal-mediated Caryl–OCF3 bond formation was described by the Ritter group (Scheme 125) [217]. Aryl trifluoromethyl ethers could be accessed through a silver-mediated cross-coupling of trifluoromethoxide with arylstannanes and arylboronic acids.</p><!><p>Ag-mediated trifluoromethoxylation of aryl stannanes and arylboronic acids.</p><!><p>In 2018, Ngai and co-workers [218] reported a direct C–H trifluoromethoxylation of (hetero)arenes under visible light irradiation. This approach proceeded at room temperature by employing the redoxactive catalyst Ru(bpy)3(PF6)2 (Scheme 126). Mechanism studies suggest a SET from the excited photoredox catalyst to 1 resulting in exclusive liberation of the OCF3 radical. The reaction of the trifluoromethoxyl radical with (hetero)arenes provides trifluoromethoxylated cyclohexadienyl radicals that undergo oxidation and deprotonation to generate the desired products.</p><!><p>Catalytic (hetero)aryl C–H trifluoromethoxylation under visible light.</p><!><p>Recently, Ngai's group [219] as well as the Togni group [220] independently developed cationic N–OCF3 reagents. Both reagents are reducible with an excited ruthenium-based photocatalyst. Herein, Togni synthesized a series of N-trifluoromethoxypyridinium reagents (Scheme 127a), while Ngai prepared a series of 1-CF3O-benzotriazole reagents (Scheme 127b). These reagents exhibit the best results for the direct C–H trifluoromethoxylation of (hetero)arenes. Notably, both groups used the same photocatalyst ([Ru(bpy)3](PF6)2) under blue LED irradiation. Of note, the cationic N–OCF3 reagents presented an impressive substrate scope tolerance including halides (I, Br, Cl and F), nitriles, ketones, amides, acids and phosphonates.</p><!><p>Photoinduced C–H-bond trifluromethoxylation of (hetero)arenes.</p><!><p>The development of methods for the transition-metal-catalyzed incorporation of fluorine-containing groups into target molecules is an active area of chemical research. In this review, we summarized the major advances in the field of transition-metal-catalyzed fluorination and fluoroalkylation reactions over the past few years. A variety of methods shows significant advantages in view of atom-economy, reaction diversity, selectivity and functional-group compatibility. The suitable catalytic systems and the newly developed reagents play a critical role in these reactions. Further, different directing groups also contribute greatly to the success of these reactions. Notably, copper shows a wider range of applications involving the catalysis of fluorination/fluoroalkylation reactions of various alkyl-, aryl- and vinyl- as well as alkynyl substrates, while palladium-catalyzed reactions in combination with suitable ligands show improved selectivity in some reactions. Also, because of milder and cleaner reaction conditions, photocatalysts have received extensive attention and have greatly applied recently. Despite diverse methods for the transition-metal-catalyzed fluorination/fluoroalkylation have been reported, chemists still face many challenges that need to be overcome. More economical, green, selective, general and practical strategies remain sought after. A suitable catalytic system and an effective fluorination reagent are two important aspects for achieving this goal. Overall, we hope that this review will provide further insight into this field and inspire chemists to develop new fluorination/fluoroalkylation reactions. Meanwhile, we believe that the methodologies mentioned in this review will contribute to future advances in the late-stage functionalization of molecules.</p>

PubMed Open Access

CHEMICAL CUES THAT GUIDE FEMALE REPRODUCTION IN DROSOPHILA MELANOGASTER

Chemicals released into the environment by food, predators and conspecifics play critical roles in Drosophila reproduction. Females and males live in an environment full of smells, whose molecules communicate to them the availability of food, potential mates, competitors or predators. Volatile chemicals derived from fruit, yeast growing on the fruit, and flies already present on the fruit attract Drosophila, concentrating flies at food sites, where they will also mate. Species-specific cuticular hydrocarbons displayed on female Drosophila as they mature are sensed by males and act as pheromones to stimulate mating by conspecific males and inhibit heterospecific mating. The pheromonal profile of a female is also responsive to her nutritional environment, providing an honest signal of her fertility potential. After mating, cuticular and semen hydrocarbons transferred by the male change the female\xe2\x80\x99s chemical profile. These molecules make the female less attractive to other males, thus protecting her mate\xe2\x80\x99s sperm investment. Females have evolved the capacity to counteract this inhibition by ejecting the semen hydrocarbon (along with the rest of the remaining ejaculate) a few hours after mating. Although this ejection can temporarily restore the female\xe2\x80\x99s attractiveness, shortly thereafter another male pheromone, a seminal peptide, decreases the female\xe2\x80\x99s propensity to re-mate, thus continuing to protect the male\xe2\x80\x99s investment. Females use olfaction and taste sensing to select optimal egg-laying sites, integrating cues for the availability of food for her offspring, and the presence of other flies and of harmful species. We argue that taking into account evolutionary considerations such as sexual conflict, and the ecological conditions in which flies live, is helpful in understanding the role of highly species-specific pheromones and blends thereof, as well as an individual\xe2\x80\x99s response to the chemical cues in its environment.

chemical_cues_that_guide_female_reproduction_in_drosophila_melanogaster

INTRODUCTION<!>1. FEMALE-DERIVED PHEROMONES THAT MEDIATE VIRGINS\xe2\x80\x99 ATTRACTIVENESS<!>1a. Intrinsic cues (age, sex, hormones) regulate the CH profile of a virgin female<!>1b. Extrinsic cues (diet, daily cues, environment) also regulate the CH profile of a virgin female<!>2a. Cuticular Hydrocarbons<!>2b. Molecules in seminal fluid<!>3. EVOLUTIONARY CONSIDERATIONS REGARDING PHEROMONES<!>3a. Pheromones display high species-specificity<!>3b. Pheromones often act in blends<!>3c. Pheromones may act as signals or as switches (on/off or rheostats)<!>3d. Sexual conflict is important to consider when evaluating pheromones\xe2\x80\x99 mechanisms<!>4. INFLUENCE OF THE ECOLOGICAL CONTEXT ON FEMALE REPRODUCTIVE BEHAVIORS<!>4a. Role of food in female post-copulatory sexual receptivity<!>4b. Role of chemical cues in egg-laying site choice<!>4b1. Chemical cues to find an appropriate egg-laying site<!>4b1. Local chemical cues that prompt egg-laying<!>CONCLUSIONS

<p>Sexual reproduction requires the interaction of a male and a female in order for their gametes to meet and combine resulting in the production of a zygote. The female typically exerts pre-copulatory selection on males based on a male's display of traits that are ideally honest indicators of his fecundity and his possession of alleles that confer adaptation to the environment. The female and her chosen mate then cooperate in the production of offspring. This simple picture omits significant, but less known mechanisms that impact the evolution and regulation of reproductive behaviors. First, females may "tune" their sexual receptivity to ecological conditions, and not solely to male quality. This is because the production of energy-intensive eggs by females is constrained by the available nutritional resources. The ecological community also matters, as the presence of predators and competing species can impact female reproductive decisions, such as how to recognize males from her own species, where to lay eggs or raise offspring (in taxa with parental care). Second, reproduction is a social phenomenon beyond the reproducing pair: finding a mate is easier in a group than when alone, risk of predation is reduced when in a group of conspecifics, and resource use by the offspring is facilitated by the presence of others. Thus, to maximize their reproductive output, females must not only judge male quality, but also integrate ecological conditions such as nutritional resources, presence of competing species, and social environment.</p><p>While reproduction is a cooperative process between males and females and between individual group members, it should also be viewed through the lens of conflict over fitness gains and competition for limited resources. The need for interaction between two sexes for reproduction can create a situation of sexual conflict between males and females, who have been selected to maximize their own fitness even if it generates a cost to the fitness of the other partner (Arnqvist and Rowe 2005; Moore and Pizzari 2005; Parker 2006; Chapman 2006). An example of such conflict is the evolution of traumatic copulatory appendages in male beetles, that damage the female reproductive tract but protect the male's sperm investment by making his mate less likely to mate with other males (Rönn et al. 2007). An additional consequence of considering reproduction as a social phenomenon is that reproductive phenotypes of an individual not only interact with those of the other sex, but also with those of group members (Moore and Pizzari 2005). The genotype of each individual also influences the social environment of other group members, including those of the opposite sex, creating new selection pressures; this phenomenon is termed indirect genetic effects (IGE;Wolf et al. 1998). As a result, the social environment can evolve more rapidly than the physical environment, leading to rapid changes in selection pressures. This perspective brings new predictions about the complexity of reproductive interactions: phenotypes that evolve under antagonistic interactions between males and females can enter a cycle of adaptation-counteradaptation (Moore and Pizzari 2005), and phenotypic changes in a group member create a new social environment for all other group members (Wolf et al. 1998). Sexual conflict and IGEs are predicted to lead to unusual levels of elaboration, plasticity and diversification in communication systems and their underlying sensing mechanisms (West-Eberhard 2014). Research on reproduction must thus expect complexity in the ecological, evolutionary and sensory mechanisms that have led to observed reproductive strategies and phenotypes.</p><p>Here we hope to illustrate the influence of ecological conditions and social context in modulating reproduction by reviewing current knowledge of the mechanisms that modulate the reproductive behavior of female Drosophila melanogaster fruit flies, in response to its environment. Most previous studies of reproductive behavior in Drosophila have been performed under reduced ecological complexity in the laboratory and have focused on males. However, this situation is rapidly changing with a recent increase in studies that consider the ecology of fruit flies when formulating hypotheses about environmental factors that affect behavior, and with a greater focus on females. Such experiments reveal that females are equipped with complex chemical senses that detect chemical substances emitted by their social and ecological milieus that change their behavior and physiology. Those substances, collectively called semiochemicals, include pheromones, which are chemical substance displayed by an individual that affect the behavior or physiology of others, as well as chemicals produced by food sources that act as cues of nutrient availability. Here, we review recent advances regarding the influence of semiochemicals on Drosophila melanogaster female reproduction. The genetics of this model organism, and the focus of many labs on its biology, has allowed dissection of how it senses chemical cues and how those affect behavior. We hope that this review will inform the development and testing of new hypotheses by neurogeneticists, behavioral geneticists, and chemical ecologists and will stimulate further integration between these two fields.</p><p>To provide the context for our review, we first outline Drosophila melanogaster's life cycle in an ecological context. Although these flies are primarily associated with fruits, they are not, strictly-speaking, phytophagous insects. They feed upon the microbial community (bacteria and fungi) responsible for fruit decomposition, as well as upon the decomposed fruit itself. Drosophila thus live in an intensely smelly environment (Mansourian and Stensmyr 2015), full of chemical cues that modulate reproduction. Flies are attracted to ripe fruits by the volatile fermentation products produced by yeasts breeding on the fruit, and are sexually stimulated by the smell of yeast (Becher et al. 2010; Gorter et al. 2016; Grosjean et al. 2011; Palanca et al. 2013; Scheidler et al. 2015). Utilization of yeast smell to locate a source of this fungus and to mate in its vicinity is likely due to the strong dependency of Drosophila reproductive success on yeast. For example, Drosophila larvae will not develop in absence of key nutrients, such as sterols, that they normally obtain from yeast (Baumberger 1917; Carvalho et al. 2010). Female flies also require a rich diet to produce eggs (Bownes & Blair 1986) needing proteins, lipids, and sugars that can be supplied by yeast (Bownes, Scott & Shirras 1988; Carvalho-Santos and Ribeiro 2017). This strong dependency on yeast appears to drive much of Drosophila reproductive strategy and life-cycle: from egg-laying on fermenting substrates, to aggregation in great numbers on fermenting fruits where the social and sexual interactions that lead to mating and the start of a new cycle. This attraction to yeast also exposes Drosophila to competitors, such as microorganisms that compete with yeast for fruit resources, and parasitoids that prey on Drosophila larvae. Figure 1 outlines the steps of Drosophila melanogaster female reproduction and the web of ecological interactions and the chemicals that influence it, as will be discussed in this review.</p><!><p>Virgin Drosophila melanogaster females display over 50 different types of hydrocarbon and fatty acid molecules on their cuticle (Dweck et al. 2015b; Everaerts et al. 2010; Lin et al. 2016; Yew et al. 2009; Yew and Chung 2015). These molecules, collectively called cuticular hydrocarbons (CHs), act as cues of sex and species identity as well as of age, microbiome status, nutritional state, and social context. The surface of a virgin female is thus a readout of her genetic identity as well as of her experience and context (Billeter and Levine 2013; Everaerts et al. 2010; Farine et al. 2012). We begin this review by exploring the function of these chemicals as pheromones in the reproduction of virgin females.</p><!><p>Drosophila melanogaster adult females emerge from their pupal case unready for reproduction. Their ovaries do not contain mature eggs (Bodenstein 1947), their cuticle is not yet hardened, and the females are typically not sexually receptive (Manning 1967)(Figure 1a). Their CHs have not yet acquired an adult quality, suggesting that CHs act as cues of females' reproductive readiness. Immature male and female flies display long chain CHs with multiple double bonds and methyl branches; there is no sexual dimorphism in CH expression at this stage (Arienti et al. 2010; Pechine et al. 1988). Oenocytes, which are subcuticular abdominal cells that produce CHs (Figure 1a), are still developing during this early adult stage (Chiang et al. 2016; Makki et al. 2014; Wicker-Thomas et al. 2015). Females' development is regulated by hormones that coordinate maturation of different parts of the female reproductive system. This is shown by the fact that genetic disruption of the Juvenile Hormone (JH) pathway or ablation of the corpora allata, the glands that produce JH, delay the maturation of both adult female CH production and oogenesis (Bilen et al. 2013; Bodenstein 1947; Soller et al. 1999). The steroid hormone ecdysone might also be part of this maturation process, as genetic disruptions in the ecdysone pathway lead to aberrant oenocyte maturation and survival (Chiang et al. 2016) and ecdysone levels change with age, correlating with CH maturation (Arienti et al. 2010; Wicker and Jallon 1995). Bursicon, a peptide hormone known for its effect on hardening and tanning of the cuticle shortly after adult emergence and on wing eversion (Honegger et al. 2008; Peabody and White 2013), increases CH production, suggesting that it plays a role in regulating the time course of CH maturation and/or accumulation on the cuticle (Flaven-Pouchon et al. 2016). CH maturation might be important for female choice, as males can coerce females to mate within the first 30 minutes after eclosion (Markow 2000; Seeley and Dukas 2011); these young females' inability to fly and their poor locomotion make them a prime target for males (Figure 1a). In this situation, the males do not display courtship. They just jump onto the females and mate, engendering a cost to the female because she cannot choose her mate. Such a lack of choice is consistent with the observation that these females will remate at higher frequencies when they become mature (Seeley and Dukas 2011). Maturation of the CH profile correlates with, and presumably induces, increased male courtship (Bilen et al. 2013). This increase in courtship by males can presumably give females more opportunities to evaluate the males and exert mate choice. That female CHs increase male courtship and delay mating is consistent with the observation that mature adult females lacking CH are more attractive to males than are females with a fully developed CH profile (Billeter et al. 2009).</p><p>Females become reproductively mature two days post-eclosion. At this time their ovaries are mature, (Bodenstein 1947) and the females have higher mating receptivity (Manning 1967). These females also have mature oenocytes that produce ~50 different CHs, most of these shared with males, except for long chain dienes including (7Z,11Z)-Heptacosadiene (7,11-HD) (Everaerts et al. 2010;Figure 1b; Figure 2). Female-specific production of dienes makes these molecules attractive candidate sex pheromones, but this role is not yet completely established. Mature females lacking CHs, including 7,11-HD, due to genetic ablation of the oenocytes, are preferred over control females as mating partners by males from some strains but not by others (Billeter et al. 2009; Pischedda et al. 2014). This means that for some strains, CHs can lower attractiveness, whereas in others CHs increase it, illustrating the importance of CHs as pheromones that regulate intra-specific courtship. The recent discovery of fatty acid-derived pheromones, such as methyl laurate (Figure 2), that are found on the cuticle of flies but are not produced by oenocytes, might shed light on how CH tune female attractiveness (Dweck et al. 2015b; Lin et al. 2016). These fatty acid-derived pheromones are found on both males and females of most Drosophila species and seem to function as general sex pheromones that stimulate male courtship. Methyl laurate, as well as a few other fatty acid-derived pheromones, is detected by olfactory receptor Or47b; males unable to sense this compound reduce courtship towards females (Dweck et al. 2015b; Lin et al. 2016). These fatty acid-derivatives likely act as general attractive sex pheromones in many Drosophila species, whereas sex- and species-specific CH made by the oenocytes likely mitigate the effects of these attractive pheromones both intra- and inter-specifically (Billeter et al. 2009; Savarit et al. 1999; Figure 1b). Indeed, the most striking and consistent phenotype of D. melanogaster females lacking CHs is that they are intensely courted by males of other species (Billeter et al. 2009; Savarit et al. 1999), an effect that can be countered by applying small amounts of synthetic 7,11-HD to CH-less females (Billeter et al. 2009). This shows that in addition to its pheromonal effects within D. melanogaster, female-specific 7,11-HD deters courtship by heterospecific males (Figure 1b).</p><p>A close look at the CH displayed by different Drosophila species reveals that males' CH are generally qualitatively invariant, but that there are large inter-species differences in CH displayed by females. For example, 7,11-HD is a female-characteristic CH only in D. melanogaster. Females of other species, like D. erecta, make the long chain diene (9Z, 23Z)-Tritriacontadiene, and females of species like D. simulans make monoenes, such as (Z)-7-Tricosene (7-T)(Jallon and David 1987; Figure 2). The prominent female CH from each species appears to be the pheromone that blocks inter-species mating; for instance 7-T appears to be the pheromone that normally blocks D. melanogaster males from courting D. simulans females, as D. melanogaster males mutant for the putative 7-T receptors Gr32a court D. simulans females (Fan et al. 2013), and D. melanogaster females lacking 7,11-HD are courted by males from other Drosophila species (Billeter et al. 2009). Differences in female CH expression arise from regulatory changes in the expression of CH-metabolic enzymes. 7,11-HD is synthesized in D. melanogaster female oenocytes through the action of a series of enzymes including desaturase1 (desat1) (Marcillac et al. 2005), desaturaseF (Chertemps et al. 2006), and elongaseF (Chertemps et al. 2007). The enzymatic pathway that controls 7,11-HD synthesis is expressed differently between species. For example, D. melanogaster desaturaseF, a gene necessary for 7,11-HD synthesis, is expressed specifically in females. However across several Drosophila species, the promoter of the desaturaseF gene shows repeated loss and gain of the sex-specific Doublesex-binding element, resulting in the gene's expression in both males and females of some species (Shirangi et al. 2009). There is also variation in the amount of two heptacosadiene (HD) isomers between D. melanogaster populations: 7,11-HD is predominant in most laboratory strains, whereas 5,9-HD levels are relatively high in Caribbean and sub-Saharan African strains (Ferveur et al. 1996). The 7,11-HD:5,9-HD ratio is controlled by two closely related genes, desat1 and desat2, (Coyne et al. 1994; Dallerac et al. 2000). A deletion in the promoter of desat2 has been correlated with high 7,11-HD and low 5,9-HD, whereby a more-active desat2 promoter results in high levels of 5,9-HD (Takahashi et al. 2001). The functional consequence of this difference is unclear. Females with 7,11-HD generally mate faster than do females with 5,9-HD, perhaps because the former elicits more courtship by males (Ferveur et al. 1996). Taken together, these data indicate that Drosophila female CHs function to block inter-species mating, by blocking courtship from heterospecifics, as well as having a role in promoting intraspecific mating (Figure 1b).</p><p>The high molecular weights of female dienes such as 7,11-HD (Figure 2), make them unlikely to be volatile pheromones (Antony and Jallon 1982). Rather, they remain on the female's cuticle and are detected by males through contact taste receptors. The involvement of the male's taste system in detecting female sex-specific CHs has been clearly documented. Males sense female CH via gustatory neurons, in the males' first pair of legs, that express the ion channel pickpocket 23 (ppk23) (Lu et al. 2012; Thistle et al. 2012). Although ppk23 is necessary for sensing 7,11-HD, it is unclear whether this ion channel is the receptor for this sex pheromone. It is a predicted sodium ion transporter, so it may act as an effector or amplifier of signal transduction downstream of chemoreceptors in taste neurons dedicated to pheromone sensing (Thistle et al. 2012). Although taste-sensing is clearly critical for detecting 7,11-HD, earlier work suggested the existence of volatile female compounds (Tompkins et al. 1980; Tompkins and Hall 1981), whose chemical identity remained undefined until recently. In 2017 Lebreton et al. discovered that 7,11-HD is oxidized quickly when in contact with air, generating a volatile degradation product (Z)-4-undecenal (Figure 2). This latter molecule is sensed by odorant receptor Or69a and can elicit upwind flight from both male and female D. melanogaster, but not from D. simulans. Thus, female D. melanogaster make a species-specific volatile pheromone that attracts conspecific males (Lebreton et al. 2017). It remains to be seen whether this pheromone also elicits courtship, or whether that is the role of the intact 7,11-HD. Resolution of this question will shed light on the relative role of olfactory versus gustatory detection of sex pheromones.</p><!><p>D. melanogaster males mate preferentially with more fecund females (Arbuthnott et al. 2017), suggesting that females display cues of their fecundity. Although visual cues such as size are probably important in mate-assessment (Laturney and Billeter 2014), it appears that pheromones also play a role (Arbuthnott et al. 2017). Indeed, the pheromones displayed by a female are responsive to external cues, providing information about the female's ecological context and thus her potential suitability as a mate. Three examples of this relate to her nutritional status, her circadian status and social environment, and the microbial environment in which the female is situated.</p><p>A central nutrient-sensing pathway, insulin signaling, affects CH production and female attractiveness (Kuo et al. 2012). Insulin signaling also regulates absorption of nutrients by the fat body. This tissue produces yolk proteins that will be taken up by the developing oocyte; yolk protein production is thus a major limiting step in oocyte maturation and hence fecundity (Fedina et al. 2017). Interestingly, under conditions of low insulin, where resources in the fat body may be released to fuel egg production, the females' oenocytes reduce production of the CHs 7-T and (Z)-5-Tricosene (5-T)(Fedina et al. 2017). This is intriguing because these two CHs, which are normally produced at higher levels in males than in females, had previously been associated with reducing male-male courtship, suggesting that they may play a role in decreasing the attractiveness of any fly to a romantically-inclined male (Billeter et al. 2009; Ferveur and Sureau 1996; Lacaille et al. 2007; Wang et al. 2011). Thus, a reduction in the quantity of these courtship-inhibitory pheromones displayed on the female cuticle should lead to increased attractiveness of that female (Figure 1b). This correlation between the effect of insulin on oocyte production and on the level of inhibitory CH suggests that males could use a female's CH profile as a proxy to sense her nutrient levels and, indirectly, her fecundity. Information about the nutritional state of a female, and hence her ability to produce eggs, may thus be cued by the same pheromone that increases her attractiveness.</p><p>CH production in mature adult females is also influenced by the circadian clock system that synchronizes the fly's behavior and physiology to day/night length and to the activity of other flies (Levine 2002; Hall 2003). Flies' CH profile changes throughout the day under the influence of central brain clock neurons that entrain to light conditions (Krupp et al. 2008). These neurons send information to the oenocytes through the neuropeptide Pigment Dispersing Factor to affect expression of genes that regulate CH synthesis (Krupp et al. 2008; Krupp et al. 2013). It seems likely that flies use circadian fluctuations in pheromone profiles to detect information about the phase of activity of other flies, since sensing the air from flies exposed to normal light:dark conditions can entrain the circadian clock of flies living in complete darkness (Levine 2002). The phase information provided by the CH profile could help flies synchronize their activity with one another. Strikingly, the circadian system also regulates CH production in the context of another external cue, the genotype of group members. The genotypic composition of males in a group affects the expression of CH synthetic enzymes in the oenocytes of other males, and thus CH production. This too is regulated by the circadian-gene system, which presumably detects environmental changes in social context in addition to its known function in adapting to changes in abiotic factors, such as photoperiod (Bloch et al. 2013; Kent et al. 2008; Krupp et al. 2008; Krupp et al. 2013). That pheromones displayed by one individual are affected by the genotype of others is an illustration of IGEs (Kent et al. 2008). These changes in male pheromone profiles might help females detect genetic diversity in nearby males. Indeed females mate more frequently when surrounded by genetically diverse males, whose CH production is modulated by one another's presence (Billeter et al. 2012; Kent et al. 2008; Krupp et al. 2008;). The genetic composition of a group thus influences male pheromones, which in turn can influence a female's mating decision. This will ultimately affect the genetic composition of the next generation, and thus the evolutionary process.</p><p>The pervasiveness of the ecological context in determining a fly's pheromonal profile is probably even broader. Bacteria living on the food consumed by Drosophila can become part of the fly's microbiome (Sharon et al. 2010), and through as yet unknown mechanisms, can affect their CH profiles. Some studies have proposed that this can lead to assortative mating between males and females raised on the same diet and infected by similar bacteria (Sharon et al. 2010). However, a new study presents data that challenge this interpretation (Leftwich et al. 2017). Pathogenic bacteria can also increase the production of Drosophila aggregation pheromones. These pheromones attract flies to the area, and flies that were uninfected can then become infected (Keesey et al. 2017).</p><p>Thus, the chemical profile of a virgin female fly reflects not only her sex, maturity, and species, but also her nutritional status, social context and experience (including mating status, see later section). We hope that future work will highlight more of the mechanism through which a fly's biography can be displayed on her cuticle and how other individuals can use this information to instruct their social interactions with that individual.</p><!><p>Male D. melanogaster also produce and display cuticular hydrocarbons. In particular, their oenocytes make 7-T (which females also display on their cuticles, including in response to nutrient cues (see above) but in much smaller quantities). When flies mate, the 7-T from the male rubs off onto the female, where it remains for at least two days and probably longer (Everaerts et al. 2010; Scott 1986; Yew et al. 2009). This additional 7-T on the female decreases her attractiveness immediately after mating to other males, which detect the molecule with gustatory receptor Gr32a on their forelegs (Laturney and Billeter 2016; Miyamoto and Amrein 2008; Wang et al. 2011) as well as Gr66a (Lacaille et al. 2007). Several other male CH are transferred to females during mating, but they have not yet been shown to have an effect on the female's attractiveness (Everaerts et al. 2010; Laturney and Billeter 2016).</p><!><p>In addition to transferring CH to females via cuticular contact, male D. melanogaster also make a fatty acid pheromone, cis Vaccenyl Acetate (cVA) that they transfer to females during mating by a different means (Butterworth 1969). cVA is made by a male reproductive tissue, the ejaculatory bulb (Brieger and Butterworth 1970; Guiraudie-Capraz et al. 2007), and is transferred to female in the seminal fluid (Figure 2). cVA, sensed by males' olfactory receptor Or67d, reduces male courtship (Ejima et al. 2007; Ha and Smith 2006; Kurtovic et al. 2007; Ronderos and Smith 2010; van der Goes van Naters and Carlson 2007). cVA might be also sensed by gustatory receptors in males' legs (Thistle et al. 2012). Finally, the pheromone (3R,11Z,19Z)-3-acetoxy-11,19-octacosadien-1-ol (CH503), is also transferred to females during mating; like cVA, it is made in the ejaculatory bulb and reduces male courtship (Ng et al. 2015; Yew et al. 2009); males sense this pheromone via gustatory receptor Gr68a (Shankar et al. 2015; Yew et al. 2009;). Thus, females receive several hydrocarbon and fatty acid-derivative pheromones from their mates via cuticular transfer and via seminal fluid (Figure 1c), and these pheromones affect the females' attractiveness (Figure 1d). We will see in section 3b that 7-T and cVA work synergistically, in a blend, to decrease the female's attractiveness to other males.</p><p>Most pheromones are relatively small molecules, including volatile molecules that are released into the environment or CHs that are displayed on the surface of an animal (e.g. see above; Yew and Chung 2015). However, a class of much larger and non-volatile molecules that are present in the semen of D. melanogaster males falls under the definition of pheromone as well. These molecules are transferred to the reproductive tract of females during mating (Figure 1C) and influence the female's behavior and physiology (Avila et al. 2011; Hopkins et al. 2017; Perry et al. 2013; Figure 1d). D. melanogaster males produce hundreds of Seminal Fluid Proteins and peptides (SFPs) whose effects transition females from virgin to reproductive states. Behavioral changes induced by these SFPs in females include diminished sexual receptivity (Avila et al. 2011; Chapman et al. 2003; Chen et al. 1988; Liu and Kubli 2003), increased protein feeding (Carvalho et al. 2006; Ribeiro and Dickson 2010), decreased sleep (Dove et al. 2017; Garbe et al. 2016; Isaac et al. 2010), and increased aggression (Bath et al. 2017). Physiological changes include increased egg production and ovulation (Chapman et al. 2003; Heifetz et al. 2000; Liu and Kubli 2003; Rubinstein and Wolfner 2013), changes in gut size and in the rate of food transit through the gut (Apger-McGlaughon and Wolfner 2013; Cognigni et al. 2011; Hudry et al. 2016; Reiff et al. 2015), changes in vesicle release and in neuromodulator levels in the reproductive tract (Heifetz and Wolfner 2004; Heifetz et al. 2014), and conformational changes of internal organs such as the oviducts (Rubinstein and Wolfner 2013) and uterus (Adams and Wolfner 2007; Avila and Wolfner 2009), likely due to muscle contraction/relaxations. During mating, several SFPs enter the circulation of the female by crossing the wall of the very distal part of her reproductive tract (Lung and Wolfner 1999). Recent studies have shown that the male's intromittent organ punctures the intima of the female reproductive tract, potentially providing a direct route for SFPs to enter the circulation (Kamimura 2007; Mattei et al. 2015; Mattei et al. 2017). From the circulation SFPs can reach all tissues of the female, including her neuroendocrine systems. Although the following paragraphs focus on D. melanogaster SFPs, seminal molecules with "pheromonal" effects need not be peptides or proteins. For example, in certain mosquitoes, the sesquiterpenoid juvenile hormone (JH; Aedes; Borovsky et al. 1994; Feinsod and Spielman 1980; Klowden and Chambers 1991; Shapiro et al. 1986) or the steroid hormone ecdysone (Anopheles) induce post-mating changes in behavior or physiology (Gabrieli et al. 2014).</p><p>Perhaps the best known of the D. melanogaster seminal pheromones is the "Sex Peptide" (SP), a 36-amino acid long (Figure 2) peptide that is produced in the male's reproductive accessory glands, and transferred to the female during mating. Different regions of SP exert different effects on the female. SP's N-terminal region increases JH production by the female's corpus allatum, as shown both in tissues incubated with SP (where JH BIII production is increased by SP (Moshitzky et al. 1996)), and in vivo (Bontonou et al. 2015; Soller et al. 1999). JH, in turn, is needed for the increased oogenesis that occurs after mating. SP's C-terminal region stimulate a decrease in female sexual receptivity starting a few hours after mating (Chapman et al. 2003; Liu and Kubli 2003; Peng et al. 2005). The C-terminal region of SP also stimulates oogenesis (Peng et al. 2005), excretory characteristics (Apger-McGlaughon and Wolfner 2013), and sperm release from storage (Avila et al. 2010). It is possible that this region is also sufficient to cause post-mating changes in siesta sleep (Dove et al. 2017; Garbe et al. 2016; Isaac et al. 2010), feeding (Carvalho et al. 2006; Ribeiro and Dickson 2010), and aggression (Bath et al. 2017) but these have not been tested directly. The C-terminal portion of SP acts through a G-protein coupled receptor, the Sex Peptide Receptor (SPR), which is found in the nervous system and elsewhere in the female including in her sperm storage organs (Yapici et al. 2008). Evidence has been presented for the existence of a second SP receptor whose molecular identity is as yet unknown (Haussmann et al. 2013). Genetic studies have defined neurons through which SP acts, via SPR, to affect female sexual receptivity behavior and egg-production; it is possible that it also acts through these neurons to induce the other behaviors (Feng et al. 2014; Hasemeyer et al. 2009; Hussain et al. 2016; Ottiger et al. 2000; Rezával et al. 2012; Rezával et al. 2014; Walker et al. 2015; Yang et al. 2009). SP's effects on sperm release from storage requires SPR in both the nervous system and the sperm storage organs (Avila et al. 2010; Avila et al. 2015b).</p><p>As with all other seminal proteins tested (Ravi Ram et al. 2005; Monsma et al. 1990), SP remains in the female's circulation for less than a day; it is degraded by proteases it encounters there (Pilpel et al. 2008), however, SP's effects on females' s receptivity and egg-laying, for instance, persist for much longer: 10–14 days. Peng et al. showed that the effects of SP are maintained because the peptide binds to sperm (via its N-terminal region), apparently protecting it from degradation once SP bound sperm are in storage (Peng et al. 2005). A cascade of seminal proteins is needed to bind the SP to sperm (Findlay et al. 2014; Ram and Wolfner 2009). SP's C-terminus is cleaved from sperm by a trypsin-like activity (Peng et al. 2005), releasing it to bind to its receptor SPR and to induce the behavioral and some long-term physiological changes.</p><p>Other seminal proteins also influence the female's physiology, thus also qualifying them as pheromones. For example, the SFP prohormone ovulin enters the female and stimulates octopaminergic signaling on her reproductive tract musculature (Rubinstein and Wolfner 2013), relaxing her oviduct and stimulating the ovulation of oocytes. Another seminal protein, Acp36DE, affects the conformation of the female's uterus (Avila and Wolfner 2009), presumably by regulating contraction of the muscles encircling it; this is suggested to facilitate the movement of sperm towards storage sites.</p><p>In summary, males transfer a complex mixture of pheromones to females during mating. These molecules are made in the oenocytes and in the reproductive tract, and can belong to completely different chemical classes (Figure 2). Despite these differences, these pheromones have in common two remarkable features: they are not transferred via the air but via contact or via semen, and they act on females to induce a transition from a virgin to a reproducing state. They effect this state change by both acting on the physiology and behavior of the female and by indicating to others that the female has mated by changing her pheromonal profile.</p><!><p>A female's sex pheromones play a critical role in broadcasting her age, species, mating-appropriateness and condition, and male pheromones that are transferred to her modulate her behavior, physiology, and attractiveness. It is interesting to consider these roles in light of the species-specificity required for successful reproduction, as well as the dynamic of conflicts in reproductive strategies of males and females. Several important areas reflect, or are influenced by, the roles of pheromones in this evolutionary context.</p><!><p>Female CH pheromones are often highly species-specific (Jallon and David 1987), attracting only conspecific males to the female for mating. Likewise, the primary amino acid sequences of an unusually high number of seminal proteins have evolved rapidly (Swanson 2004; Haerty et al. 2007), giving rise to species-specific sequences. Seminal proteins that show signs of positive selection at the amino acid level, or other evidence of positive selection, include ovulin (Aguadé et al. 1992), Acp36DE (Begun et al. 2000), and SP (Cirera and Aguadé 1997). Ovulin orthologs are difficult to find outside the melanogaster group (Mueller et al. 2005), and it is not yet known if those potential orthologs – some different enough not to cross-react with anti-melanogaster-ovulin - are functional. SP orthologs are found throughout Drosophila, though distant species' are very different in sequence and are not functional in D. melanogaster (Tsuda et al. 2015; Tsuda and Aigaki 2016). Thus, both ovulin and SP appear to be novel genes, that have evolved rapidly. Interestingly the SPR is highly conserved and found in many insect orders such as mosquitoes, Lepidoptera and beetles (Yapici et al. 2008). The ancestral ligands for this receptor are believed to be the well conserved myoinhibitory peptides, whose functions include regulating muscle contraction and fluid balance (Kim et al. 2010; Yamanaka et al. 2010). It appears that SP arose and coopted the receptor for myoinhibitory peptides, and then was selected for in certain Drosophila lineages as it conferred advantages to the male in preventing his mate from remating and in inducing his mate to undertake physiological and behavioral changes that can increase her progeny production. While it is likely that the rapid evolution of seminal proteins arose due to conflicts between the reproductive interests of males and females (e.g. Sirot et al. 2014), one could imagine potential contributions from a Muller-Dobzhansky system, in which protein partners that co-evolve within a lineage become incompatible with their counterparts in a different lineage (see Maheshwari and Barbash 2011 for review). Moreover, the rapid evolution of seminal proteins may also have been favored as part of a fail-safe mechanism: if pre-mating mechanisms like pheromones (and other cues) failed to maintain species isolation, the species-specificity of seminal proteins could help keep an interspecies mating unsuccessful.</p><!><p>While precise chemical nature of a pheromone can confine its effect to a single targeted species, further specificity can be gained if a successful signal involves a combination of molecules, in particular concentrations, e.g. a blend. This idea is well documented in the sex pheromones of Lepidoptera (Renou 2014), but is also is important in CH of D. melanogaster and, likely, in SFP blends in this species.</p><p>For example, as noted above, D. melanogaster males rub the CH 7-T (Figure 1c; Figure 2) onto females during mating and also transfer the volatile pheromone cVA to females in semen. These two pheromones act as a blend to reduce female attractiveness (thus acting as fast-acting mate-guarding pheromones that help the male protect his sperm investment; see below). The 7-T/cVA blend of mate guarding pheromones makes females smell partially like males. Virgin females do not produce cVA and only exhibit small amounts of 7-T, while males produce these two molecules in high quantities (Everaerts et al. 2010). That 7-T and cVA act as a blend to reduce male courtship is evident in that males devoid of CH, including 7-T, but who still produce cVA, are attractive to other males (Billeter et al. 2009). cVA is thus not sufficient to reduce male-male courtship. Rather, it requires 7-T as a partner for its mate-guarding effect, as perfuming males lacking CH with 7-T blocked courtship from other males (Billeter et al. 2009; Wang et al. 2011). In keeping with the idea that males "masculinize" the pheromone profiles of females during mating, perfuming virgin females with only 7-T or cVA at their natural dose does not decrease in females' attractiveness. It is only when both pheromones are present that they, together, cause decreased female attractiveness (Laturney and Billeter 2016)- the blend is uniquely effective.</p><p>By marking their mates with cVA and 7-T, males make these females smell like males (Scott 1986). Such a trick, in conjunction with mated females having lower sexual receptivity than virgins (Chen et al. 1988; Manning 1967), explains why males have not evolved to ignore this mate-guarding pheromone. Even though ignoring that pheromonal combination could lead them to sire most of the offspring from previously-mated females, ignoring it would also lead them to court males, a strategy with low fitness benefits. Interestingly, in addition to donating anti-aphrodisiac pheromones (7-T, cVA) to mated females, males also reduce females' expression of attractive sex pheromones such as 7,11-HD. Specifically, the seminal pheromone SP down-regulates females' hydrocarbon (CH) production (Bontonou et al. 2015). Thus, males lower female attractiveness by marking them with male pheromones and by altering females' production of sex pheromones. That male hydrocarbon pheromones make females unattractive immediately after mating, may buy time for SP to act via the receptors and neural pathways noted above to decrease female receptivity.</p><p>Blends may also be important in the optimal action of SFP pheromones. Male D. melanogaster appear to be capable of adjusting the amounts, and relative levels, of SFPs in response to social cues (related to perceived sperm competition risk. In the presence of rivals, males transfer more seminal proteins (ovulin and SP) and sperm (Garbaczewska et al. 2012; Perry et al. 2013; Sirot et al. 2011; Wigby et al. 2009; Wigby et al. 2016), thus transferring ejaculates of different composition in response to sperm competition risk (Parker and Pizzari 2010). In addition, when males are mated with previously mated females, they transfer more SP relative to ovulin than when they mate with previously unmated females (Sirot et al. 2011). Thus, males can adjust the blend of seminal proteins transferred when faced with a mated female, increasing the relative amount of a paternity-protecting molecule (SP) relative to the amount of a molecule (ovulin) that had already acted and on whose prior activity they can piggy-back. The idea of a different blend for different circumstances recalls the different pheromone blends used by different species for mate-attraction.</p><!><p>Pheromones change the characteristics, and in some cases the behavior and physiology, of the recipient. In some cases, they do so directly, acting as signals at the top of a signal transduction pathway –e.g. as ligands for odorant (or taste) receptors whose activated state is then transduced to the sensory system (Gomez-Diaz and Benton 2013). For example, the hydrocarbon and semen pheromones discussed above that are transferred by male D. melanogaster change a mated female's smell and taste to more male-like, resulting in her decreased attractiveness to other males, without any direct action by the female – the compounds simply act as ligands for the male's receptors.</p><p>In other cases, pheromones switch the recipient to a new state by hijacking pre-existing hormonal or sensory pathways in the recipient. This is particularly the case for the SFP pheromones. Here too however, as with the ligands just mentioned, the SFPs appear to act at the tops of pathways – in this case as switches that regulate the activity of a pre-existing pathway in the female. Sometimes SFPs can act as traditional on/off switches. For example, ovipositor extrusion is only seen in mated females; this behavior is turned "on" (from an "off" state) by SP (Chen et al. 1988). In other cases, SFPs act more like rheostat-switches, turning up (or down) pre-existing pathways and processes in mated females. Examples are that ovulin increases octopaminergic signaling (and thus ovulation) in mated females (Rubinstein & Wolfner, 2013), and that SP increases oogenesis and decreases receptivity by mated females (Chapman et al. 2003; Chen et al. 1988; Liu & Kubli, 2003). In all of these cases, the process seen post-mating (ovulation, oogenesis, occasional low receptivity) occurs to some low level in virgin females, but the SFP acts to increase its level, often greatly.</p><!><p>Thus far, we have discussed mechanisms of pheromone perception as important in initial proximate interactions between females and males, and have noted that male CH pheromones and SFPs also can affect the likelihood of a mated female mating with another male. This latter effect arises because polyandry is costly to males in species like D. melanogaster, due to last male sperm precedence (Manier et al. 2010; Parker and Pizzari 2010; Schnakenberg et al. 2012). This phenomenon, in which the last male to mate sires the majority (often 80–90%) of offspring, means that a male's reproductive success will be disadvantaged if his mate mates again (Parker and Pizzari 2010). Sexual conflict theory predicts that Drosophila males should evolve mechanisms that either reduce their partners' ability to re-mate ("mate guarding") and/or increase the chances of the female using his sperm over that of another male. As noted above, hydrocarbon pheromones perform mate-guarding functions, as does SP's inhibition of female receptivity. Other SFPs affect sperm competition outcomes (Chapman et al. 2000; Clark et al. 1995; Fiumera et al. 2005; Harshman and Prout 1994; Reinhart et al. 2014; Zhang et al. 2013), consistent with the effects on sperm use but are beyond the scope of this review. However, polyandry can be beneficial for females (Arnqvist and Nilsson 2000; Jennions and Petrie 2000;), for example by allowing them to acquire sperm from different males, with advantages in producing the most-fit progeny. This sets up a tension between the strategies/needs of females and males, leading to sexual conflict and an arms race. Simply put, males transfer pheromones that act to their advantage, and females acquire resistance to those pheromones, to their advantage. Males can then evolve new pheromones or other mechanisms to overcome female resistance.</p><p>Such conflicts have been suggested to have driven some aspects of the CH pheromone phenomenon in D. melanogaster (Laturney and Billeter 2016), and at least some of the rapid sequence evolution of critical D. melanogaster seminal proteins (Sirot et al. 2015). Focusing on the former, the transfer of 7-T and cVA by males reduces the female's chance of remating, which can reduce her potential to accrue fitness benefits. In such a situation, females that remove or deactivate those pheromones, thus restoring her attractiveness and/or remating propensity, will accrue fitness advantages. Recently it was shown that females exhibit a behavior that can mitigate the effect of mate guarding hydrocarbon pheromones and SFPs that affect her remating receptivity. A few hours after mating, females eject the surplus male ejaculate located in their uterus (bursa), including its cVA and possibly a large fraction of SFPs (Figure 1e; Dumenil et al. 2016; Laturney and Billeter 2016; Lee et al. 2015; Manier et al. 2010). This ejection removes ~90% of cVA, altering the blend with 7-T that had led to optimal mate–guarding, and thus restoring some female attractiveness (Laturney and Billeter 2016). Therefore, male manipulation of female attractiveness can be disrupted via ejaculate ejection, a typical post-mating behavior.</p><p>The cVA that males provide to females is ejected with the mating plug that formed at the entrance of her reproductive tract. This gelatinous plug is a coagulation of proteins from male reproductive glands (such as accessory glands and ejaculatory bulb), and female proteins (at least in mosquitoes (Rogers et al. 2009), and fills the bursa of the mated female. It is thought to help retain ejaculate, including pheromones like cVA in the female's bursa and may also assist sperm movement into storage. Prevention of D. melanogaster mating plug coagulation by knockdown of a critical mating plug protein from the male's ejaculatory duct (pEBme)(Avila et al. 2015a) results in premature loss of ejaculate from the female, minimizing sperm storage and affecting post-mating responses that depended on stored sperm. Moreover, at least one mating plug protein affects female remating behavior before mating plug ejection (Bretman et al. 2010), in addition to the effects of cVA. Indeed nearly half of the recently ejected females re-mated during a 30-minute observation period, whereas all the non-ejected females abstained (Figure 1e,f; Laturney and Billeter 2016). This finding suggests a close temporal relationship between ejection and remating: females that are faster to eject may also be faster to re-mate. The timing of ejaculate ejection is plastic. It is also socially modulated: females that were held in groups ejected the ejaculate 1 hour earlier than females that had mated in single-pairs and were then isolated after copulation (Laturney and Billeter 2016). As females also mate faster and more often when in social contexts that contain more flies and with more genetic diversity (Billeter et al. 2012; Gorter et al. 2016; Krupp et al. 2008; Laturney and Billeter 2016), females may be able to modulate timing of ejection to influence attractiveness in order to maximize reproduction. In contexts that are favorable for their reproduction, such as when genetically diverse males are present, females may shorten their ejection latency in order to attract potential mates and increase genetic diversity of offspring. However, if remating is not likely or beneficial, such as when the female is isolated or is with inbred males, females may lengthen their ejection latency in order to reduce unwanted sexual harassment or to ensure full usage of their already obtained ejaculate.</p><p>The continued evolutionary interplay between female and male reproductive strategies and molecules may explain why pheromones such as SFPs often act as switches (on/off or rheostats, as discussed above). Physiology and behavior are regulated by complex molecular pathways that often involve the action of pleiotropic molecules and conserved machinery (for example, ovulation in both Drosophila and locust require octopamine signaling in females; Monastirioti et al. 1996; Monastirioti 2003; Lange 2009). In this context, it may be simpler to evolve a new switch to turn on/off (or up/down) a pathway (e.g. ovulin in D. melanogaster) if the old switch has become ineffective due to counter-evolution, rather than to evolve a new or modified internal component of the intricate and pleiotropic pathway.</p><!><p>So far in this review, we have focused on chemicals that guide sexual interactions between flies. Ecological conditions, such as food availability and the presence of harmful organisms, can have dramatic effects on female reproductive behaviors. Interestingly those conditions mainly affect mated and not virgin females. They are sensed through chemical cues produced by the food sources, predators and pathogens. The female's reactions to those semiochemicals function to ensure offspring survival and might thus be considered a primitive form of maternal care.</p><!><p>Food availability is paramount to female reproduction because it provides the energy for both the production of eggs and the survival of offspring during development. As we have seen above, male seminal pheromones transferred to females during mating stimulate egg production, which puts high nutritional demands on females. These nutritional needs are accompanied by a change in female physiology, including a change in intestinal morphology and function (Apger-McGlaughon and Wolfner 2013; Cognigni et al. 2011; Lemaitre and Miguel-Aliaga 2013; Reiff et al. 2015) as well as a change in the female's diet. Mating shifts the dietary preference of female flies from sugar-rich food to yeast, which is a prime source of proteins (Carvalho et al. 2006; Ribeiro and Dickson 2010; Vargas et al. 2010). This shift in dietary preference towards yeast is adaptive, as the protein content of yeast fuels egg-production and yeast itself is necessary as a food source for offspring development and survival (Baumberger 1917; Terashima and Bownes 2004). Interestingly, these changes in food preference are not controlled directly by the demands of egg-production, as mated females lacking oocytes still consume more yeast (Barnes et al. 2007; Ribeiro and Dickson 2010). Instead this dietary change is triggered by SP. This male pheromone can shift a female's nutritional preferences towards a diet that benefits offspring production and survival (Barnes et al. 2008; Carvalho et al. 2006; Ribeiro and Dickson 2010; Vargas et al. 2010).</p><p>Yeasts produce several volatile fermentation products (Becher et al. 2012) that act as chemical cues that stimulate sexual activity, coupling a cue about the presence of an important nutritional resource for reproduction with sexual arousal. This simple coupling is seen for D. melanogaster males, whose courtship intensity increases in the presence of odors from food (Grosjean et al. 2011). Interestingly, the food odor-cue that triggers increased male courtship is sensed by olfactory neurons whose second order neuronal projections converge with those of sex pheromone sensing neurons, making the smell of certain foods akin to pheromones in both function and sensory mechanism (Grosjean et al. 2011), in contrast to the situation with other (general) odors. The mating propensity of females also changes when there is yeast in their diet (Fricke et al. 2010; Harshman et al. 1988): presence of yeast increases the sexual receptivity of mated (but not virgin) females (Figure 1f; Gorter et al. 2016). However the coupling of food (yeast) odors to sex is not as simple for D. melanogaster females as it is for males: the smell of yeast is not sufficient to increase mated females' likelihood of remating (Gorter et al. 2016). Rather, the combination of yeast odors, in particular acetic acid (Figure 2), plus the presence of the yeasts' amino acids is required to trigger an increase in females' mating receptivity. The olfactory sensing of yeast relevant to sexual behavior is mediated by the ionotropic olfactory receptor family in both males and females, in particular Ir84a (Grosjean et al. 2011) in males and Ir75a in females (Gorter et al. 2016), but the identity of the receptor for yeast amino acids remains unclear. Mated females must integrate two signals, one received through olfactory means, the other probably through gustatory, to increase sexual receptivity in the presence of yeast. This integration allows mated females to simultaneously measure the consumed yeast nutritional resources required for egg production and the presence of environmental yeast that will be required for offspring growth. Such integration insures against the inability of the peripheral nervous system to discriminate substances that smell or taste like yeast but cannot be metabolized to fuel egg production. Finally, why does yeast affect the sexual receptivity of mated but not virgin females? Multiple mating is costly to females (Chapman et al. 1995; Fowler and Partridge 1989; Wigby and Chapman 2005), so mated females would maximize fitness by only mating to acquire more sperm or ovulation-boosting male seminal peptides when there are enough yeast resources for sustained egg production and for offspring development. In contrast, it is sufficiently important that a virgin female acquire sperm that she would not benefit from restricting her mating only to the most optimal, yeast-laden, environments.</p><!><p>Female flies also rely on their sense of smell and taste to identify sites on which to lay their eggs. Selection of an egg-laying site occurs in two phases. First, females are attracted to a prospective oviposition site. Second, once there, females utilize different, more local, cues for the final decision to deposit eggs.</p><!><p>The first step in oviposition is locating an appropriate site (Figure 1g). Mated D. melanogaster females are attracted to decaying fruits (Becher et al. 2012; Reed 1938). Yet their "fruit fly" name is somewhat misleading, because they are actually attracted by the combination of fruit and the yeasts that grow on the fruits. Among volatile cues that attract flies are yeast fermentation products, and yeast alone is sufficient to attract flies (Becher et al. 2012). The ecological relevance and strength of this attraction is exemplified by the lily, Arum palaestinum, whose volatiles mimic yeast fermentation products. The lily volatiles attract flies, which then act as pollinators (Stökl et al. 2010). Several fly odorant receptors (Ors and Irs) detect the smell of yeast (Silbering et al. 2011; Stökl et al. 2010). This complex long-distance detection of, and attraction to, yeast volatiles is consistent with the ecological importance of yeast for flies. But the yeast volatiles alone are not the only chemical cues that attract D. melanogaster. The flies are also attracted to some fruit-specific products, such as antioxidants. These fruit antioxidants are hypothesized to help the immune defense of larvae that will eventually hatch on these sites and will have to defend against microorganisms that co-occur with beneficial food yeasts in the flies' habitat (Dweck et al. 2015a). Interestingly, the antioxidants are not sensed directly by the flies. Rather, a yeast metabolite of the antioxidants, the volatile ethylphenol (Figure 2), attracts the flies (Dweck et al. 2015a). Thus, flies are attracted to the combination of yeast and fruit, harking back to the idea that semiochemicals often work as blends, or in this case convey information as a blend, instructing much of a female fly's egg-laying decision.</p><p>The chemical cues that are sensed by females are not always attractants. Females can also evaluate chemical cues to avoid sites that are potentially dangerous to their eggs or offspring. For example, D. melanogaster females express a sensitive and selective olfactory channel for geosmin (Figure 2), a volatile chemical produced by harmful microorganisms. Detection of geosmin leads to strong repulsion and to avoidance of potential egg-laying (Figure 1g,h; Becher et al. 2010; Becher et al. 2012; Stensmyr et al. 2012). The fly olfactory system therefore is tuned to sensing microorganisms that affect their progeny. Whether these cues are attractant or repellent seems determined by whether they come from microorganisms that are beneficial (mostly yeast) or harmful (e.g. pathogens) (Figure 1g).</p><p>Sensing chemical cues associated with nutritional vs. pathogenic features is only one aspect of attraction to an egg-laying site. Other flies can also help. Flies that have already located a food source deposit aggregation pheromones that attract more flies. Drosophila females benefit from aggregation through communal egg-laying, which increases offspring survival through cooperation between larvae in fending off fungal growth on the food permitting better resource exploitation (Stamps et al. 2012; Trienens et al. 2017; Wertheim et al. 2002b). The aggregation pheromones that attract females to oviposition sites are complex. The first aggregation pheromone discovered in Drosophila (Bartelt et al. 1985) was cVA (Figure 2; Wertheim et al. 2002a; Wertheim et al. 2006). This pheromone was discussed earlier as contributing to decreased mating-attractiveness of mated females when co-detected with 7-T. When co-occurring with the smell of food, cVA functions to attract flies to aggregate (Bartelt et al. 1985; Billeter and Levine 2015; Lebreton et al. 2012; Schlief and Wilson 2007; Xu et al. 2005;), illustrating once again that being in a blend can confer different functions to the same pheromone. cVA is deposited by mated females on egg-laying sites during the process of ejaculate ejection discussed earlier (Figure 1e; Dumenil et al. 2016; Laturney and Billeter 2016). Several other aggregation pheromones that have been subsequently reported include 9-Tricosene (Figure 2), which is deposited by males when sensing the presence of food and attracts females to lay eggs nearby (Lin et al. 2015). Mated females also deposit a series of CH on yeast sources, promoting egg-laying by other mated females (Dumenil et al. 2016). These compounds are deposited through frass (fly excreta), which contains CH, some cVA, as well as more recently-discovered pheromones such as methyl laurate (Figure 2; Keesey et al. 2016). Aggregation pheromones deposited by D. melanogaster are not species-specific, so various Drosophila species are regularly found occupying the same oviposition sites. Thus, the benefit of breeding together is likely not restricted to the species level but may benefit closely related species (Symonds and Wertheim 2005; Wertheim 2005). The process of attracting flies to an egg-deposition site has some negative consequences: it can be hijacked by parasitoids. cVA also attracts parasitoid wasps to oviposition sites (Wertheim et al. 2003). These wasps inject their eggs in Drosophila larvae, where the eggs then and develop further. In nature, up to 80% of Drosophila larvae are parasitized by parasitoid wasps, including Leptopilina boulardi and L. heterotoma (Fleury et al. 2004).</p><!><p>Once a female has reached a suitable egg-laying substrate, she searches for a site on which to lay her eggs. This searching behavior is triggered by the presence of an ovulated egg in her reproductive tract (Gou et al. 2014). The female walks around, probing the substrate with her legs, proboscis and ovipositor, all of which contain sensory receptors (Yang et al. 2008). When sampling potential oviposition sites, females integrate input from smell and taste to weigh two competing options: egg-laying attraction vs. positional repulsion (Joseph et al. 2009). For instance, females select substrates containing acetic acid (Figure 2) for egg-laying, showing egg-laying attraction, but do not stay on such sites once they have laid the egg, showing positional repulsion. The egg-laying preference for acetic acid is primarily relayed through gustatory neurons, while positional aversion is relayed through the olfactory system (Joseph et al. 2009). Analogous contradictory behaviors in females' responses to a chemical cue are also observed in response to lobeline (Figure 2), an alkaloid that is naturally produced by plants in the genus Lobelia (Krochmal et al. 1972), and serves as a feeding repellent for several insect species (Detzel and Wink 1993; Wink and Schneider 1990). The presence of lobeline repels Drosophila females from long-stays on a substrate, but attracts these females to lay eggs there (Joseph and Heberlein 2012). The positional repulsion and egg-laying attraction of lobeline is regulated by gustatory neurons: neurons on the tarsi of the female's first pair of legs stimulate positional aversion (Chen and Amrein 2017; Joseph and Heberlein 2012), whereas neurons in the internal mouthparts lining the pharynx receive inputs that stimulate egg-laying attraction (Joseph and Heberlein 2012).</p><p>The chemical cues that guide the precise selection of an egg-laying site by females are not directly connected to the cues that attract females to oviposition sites from a long-distance. This is shown by the observation that several chemicals that promote oviposition do not attract flies in olfactory tests. This dichotomy between the mechanisms that determine general site attraction and egg-laying site choice is illustrated by oviposition site choice in the presence of parasitoid wasps. The adult fly's olfactory system is tuned to the Leptopilina odor iridomyrmecin, which is a major component of the female wasp sex pheromone (Figure 2; Ebrahim et al. 2015). Flies show no long-range repulsion to the smell of wasps, but female flies will avoid laying eggs when they smell the parasitoid (Figure 1h; Ebrahim et al. 2015; Lefèvre et al. 2012). As Drosophila parasitoid wasps do not attack adult flies but only the larvae that hatch from the fly's eggs, it makes sense that female flies are not repelled by the smell of parasitoids, but merely avoid laying eggs in their presence. The same logic applies to the observation that females prefer to lay egg on citrus fruits (Figure 1i), which produce limonene. Even though limonene does not normally attract flies, its smell repels parasitoid wasps, potentially explaining why flies prefer to oviposit at sites containing limonene (Dweck et al. 2013). Finally, flies avoid laying eggs near feces of carnivorous animals, perhaps because the feces contain harmful bacteria that produce phenol – a volatile detected by and repellent to flies (Mansourian et al. 2016). Egg-laying site preference is thus an chemically-driven behavioral strategy that ensures that offspring are reared in safer environments.</p><!><p>Chemicals are important agents of communication. In D. melanogaster, pheromones include cuticular and other hydrocarbon molecules and seminal proteins from males. These suites of molecules can telegraph the species, maturity, mating-status, and other reproductive characteristics of females to potential mates (of their own or other species). Given the importance of species-specificity in mating on the one hand, and the divergent optimal reproductive strategies of males and females on the other, pheromones show remarkable species specificity at the structural level and in the blends in which they function. Their activity as signals or switches, rather than as downstream effectors of biochemical pathways, is evidence of the broad effect that these molecules have on the fly's biology and evolution. Chemicals that contain information about food substrates and the presence of pathogens or predators, are also sensed by females (and males) including in the selection of optimal sites for egg-laying and progeny development. The valence (attractiveness or repulsiveness) of these latter chemical cues to mated females is connected to whether the species that produce them is beneficial to the female or her progeny, in which case it attracts females, or harmful, in what case it repels them. The complexity of chemical signaling is evident even in the cues made by, or affecting, females – on which we focused this article. The male likely encounters a similar complexity. There is much room for future research to uncover the complexity and expanse of the sensory mechanisms that underlie reproductive behaviors and the details of the chemical mechanisms that couple reproductive success to a fly's intrinsic state, and to ecological and social contexts.</p>

PubMed Author Manuscript

The role of hypervalent iodine(<scp>iii</scp>) reagents in promoting alkoxylation of unactivated C(sp³)–H bonds catalyzed by palladium(<scp>ii</scp>) complexes

Although Pd(OAc) 2 -catalysed alkoxylation of the C(sp 3 )-H bonds mediated by hypervalent iodine(III) reagents (ArIX 2 ) has been developed by several prominent researchers, there is no clear mechanism yet for such crucial transformations. In this study, we shed light on this important issue with the aid of the density functional theory (DFT) calculations for alkoxylation of butyramide derivatives. We found that the previously proposed mechanism in the literature is not consistent with the experimental observations and thus cannot be operating. The calculations allowed us to discover an unprecedented mechanism composed of four main steps as follows: (i) activation of the C(sp 3 )-H bond, (ii) oxidative addition, (iii) reductive elimination and (iv) regeneration of the active catalyst. After completion of step (i) via the CMD mechanism, the oxidative addition commences with an X ligand transfer from the iodine(III) reagent (ArIX 2 ) to Pd(II) to form a square pyramidal complex in which an iodonium occupies the apical position.Interestingly, a simple isomerization of the resultant five-coordinate complex triggers the Pd(II) oxidation. Accordingly, the movement of the ligand trans to the Pd-C(sp 3 ) bond to the apical position promotes the electron transfer from Pd(II) to iodine(III), resulting in the reduction of iodine(III) concomitant with the ejection of the second X ligand as a free anion. The ensuing Pd(IV) complex then undergoes the C-O reductive elimination by nucleophilic attack of the solvent (alcohol) on the sp 3 carbon via an outersphere S N 2 mechanism assisted by the X À anion. Noteworthy, starting from the five coordinate complex, the oxidative addition and reductive elimination processes occur with a very low activation barrier (DG ‡ 0-6 kcal mol À1 ). The strong coordination of the alkoxylated product to the Pd(II) centre causes the regeneration of the active catalyst, i.e. step (iv), to be considerably endergonic, leading to subsequent catalytic cycles to proceed with a much higher activation barrier than the first cycle. We also found that although, in most cases, the alkoxylation reactions proceed via a Pd(II)-Pd(IV)-Pd(II) catalytic cycle, the other alternative in which the oxidation state of the Pd(II) centre remains unchanged during the catalysis could be operative, depending on the nature of the organic substrate.Scheme 1 Palladium-catalyzed C(sp 3 )-H alkoxylation using the cyclic iodine(III) reagent 2 (BI-OMe) developed by Rao et al.

the_role_of_hypervalent_iodine(<scp>iii</scp>)_reagents_in_promoting_alkoxylation_of_unactivated_c(s

Introduction<!>CMD mechanism<!>Reductive elimination<!>Regeneration of active catalyst 13<!>Catalytic cycle proposed by the DFT calculations<!>Impact of the substituent on the reaction mechanism<!>Assessing the generality of our proposed mechanism<!>Conclusion<!>Computational details

<p>Hypervalent iodine compounds have been widely used in organic synthesis as oxidants and reaction promoters over recent decades. 1 Although these compounds themselves have the capacity to mediate many organic reactions, the addition of a catalyst is usually a prerequisite for certain transformations to occur. 2 Among these catalysts, palladium complexes have been demonstrated to have great potential to promote numerous iodine(III)-mediated processes including the C-H functionalization and coupling reactions. 3 In this context, Rao et al. developed a method for alkoxylation of unactivated methylene and methyl C(sp 3 )-H bonds to prepare alkyl ethers using hypervalent iodine(III) reagents in the presence of a Pd(II) catalyst. 4 According to the developed method, they showed that butyramide derivative 1 is alkoxylated in the presence of an alcohol as the solvent and the alkoxylation reagent, methoxybenziodoxole (BI-OMe, 2) as the oxidant, and Pd(OAc) 2 as the catalyst (Scheme 1). This strategy for alkoxylation might receive specic attention due to its signicance in modication of anti-inammatory drugs such as ibuprofen and naproxen. It is worth mentioning that the importance of such a transformation has also prompted others to use similar methodologies for installing alkoxy groups onto other organic molecules using hypervalent iodine(III) reagents under Pd(II) catalysis even, in some cases, before that developed by Rao et al. 5,6 In an attempt to explore the reaction mechanism, Rao et al. conducted an isotope labeling experiment in deuterated methanol (Scheme 2). On that basis, the only observed product was 4, implying that the methoxy group installed on the product must originate from the solvent and not from the oxidant. Interestingly, when different alcohols were used as the solvent, only product 5 was practically observed and when the reaction was run in 1,2dichloroethane (DCE), the carboxylated product 6 was attained. All these ndings suggest that the methoxy group of the oxidant must act as a spectator ligand and thus is never involved in the reductive elimination step in the catalytic cycle (vide infra).</p><p>Based on the preliminary results briey discussed above, Rao et al. proposed the catalytic cycle outlined in Scheme 3. Accordingly, the reaction was surmised to commence with the coordination of substrate 1 to Pd(OAc) 2 , followed by OAc ligands-assisted N-H and C(sp 3 )-H activation processes via the concerted metallation-deprotonation (CMD) mechanism 7 to yield cyclopalladated intermediate 7. Subsequently, the oxidation of Pd(II) to Pd(IV) by the cyclic iodine(III) reagent 2 generates intermediate 8 from which a ligand exchange with the alcohol occurs to afford 9. The resultant intermediate nally forms product 3 by undergoing the C-OR reductive elimination followed by a substitution reaction. It is worth noting that this plausible mechanism has also been summarized in several recent reviews, 3e,8 and proposed by several other researchers for interpreting alkoxylation reactions mediated by iodine(III) and catalyzed by palladium(II). 4,5,6e,f,9a,b,f,h-j,m,n The above-simplied description prompted us to investigate the detailed mechanism of the title reaction by applying density functional theory (DFT) with the aim of addressing the following questions: (i) how the Pd(II) is oxidized to the Pd(IV) by BI-OMe? (ii) Are 8 and 9 key intermediates on the catalytic cycle? If so, why does not the reductive elimination occur from 9 to give 11? If not, what are the key intermediates? Why is the carboxylate group, and not the methoxy, installed on the organic molecule when the solvent is not the alcohol (Scheme 2)? Is the formation of a Pd(IV) intermediate inevitable in the catalytic process? Does the C-O reductive elimination take place via an inner-sphere mechanism? Why does the reaction require a high temperature to occur? Through answering these questions, we hope to enhance understanding of the fundamental processes involved in numerous Pd(II)-catalyzed I(III)-mediated alkoxylation reactions. 4,5,6e-g,9 Results and discussion</p><!><p>As described in the Introduction, the reaction is proposed to be initiated by activation of the N-H and C-H bonds of substrate 1 by Pd(OAc) 2 via the CMD mechanism. In analogy with the previous DFT studies, 10 trimeric Pd 3 (OAc) 6 is considered as the precatalyst for the palladium acetate; the trimeric form is computed to be 15.6 kcal mol À1 more stable than the monomeric Pd(OAc) 2 (Fig. 1a). The breakdown of Pd 3 (OAc) 6 into square planar complex 13 is computed to be exergonic by about À1.8 kcal mol À1 (Fig. 1b).</p><p>The N-H deprotonation of the coordinated substrate by one of the acetate ligands gives complex 14 in which acetic acid occupies a vacant coordination site of Pd(II). This process is predicted to be extremely fast with DG ‡ ¼ 5.0 kcal mol À1 . The acetic acid in this complex (14) binds weakly to the metal center and thus it can be easily replaced to give complex 15 with a Pd-H-C agostic interaction (Fig. 1c). The resultant complex then involves a second deprotonation process by the other acetate ligand to yield 16. We found that this key step is nearly thermoneutral and proceeds with an overall activation barrier of 19.7 kcal mol À1 . The replacement of the weakly bonded acetic acid in 16 by BI-OMe gives 17 as an active species on the catalytic cycle (vide infra). process (Scheme 3). On that basis, the reductive elimination from 9 is expected to produce both products 3 and 11, which disagrees with the experimental ndings. Similar key intermediates were also suggested by some other researchers for analogous alkoxylation reactions catalyzed by Pd(II) complexes. 4,5,6e,f,9a,b,f,h-j,m,n To conrm the proposed mechanism is inoperative, we calculated the energy of intermediates 8 and 9a, and then evaluated the C-O reductive elimination from these two intermediates (Fig. 2). Several trends are apparent from this DFT investigation. First, based on the proposed mechanism (Scheme 3), the ligand exchange between 8 and the alcohol (solvent) is a prerequisite for the reductive elimination to produce the desired product. Our calculations show that the ligand exchange process is endergonic by 6.4 kcal mol À1 where the alcohol is MeOH, implying that the Pd center prefers to remain coordinated with the carboxylate ligand. Second, transition structures TS 1 and TS 2 are signicantly lower in energy than TS 3 . It follows that the reductive elimination should preferentially occur from 8 and not 9a. Third, transition structure TS 2 is by 4.5 kcal mol À1 lower in energy than TS 1 indicating that 8 is more prone to be involved in the carbon-carboxylate coupling and not the carbon-alkoxy coupling.</p><p>It is inferred from the above results that the mechanism proposed in the literature is not capable of accounting for the experimental data. This inconsistency spurred us to seek for some other alternatives. The following discloses a novel mechanism by which one can easily interpret the results of the Pd(II)catalyzed I(III)-mediated alkoxylation reactions developed by Rao et al. and others. 4,5,9 Oxidative addition Our calculations show that the oxidative addition step starts with the formation of adduct 17 in which reagent BI-OMe binds to the palladium center through its methoxy group (Fig. 3a). Next, the OMe ligand is transferred from the iodine to the palladium via transition structure TS 17-18 to give square pyramidal complex 18. The resultant ve coordinate complex is formally an iodonium salt in which the Pd d z 2 orbital interacts with an empty orbital on the iodine(III) having three-center-fourelectron (3c-4e) character (Fig. 3b); the spatial distribution for the LUMO of 18 (Fig. 3b) indicates the antibonding orbital relating to such an interaction. Interestingly, we found that this complex is extremely reactive toward a redox process through undergoing a simple isomerization. The strong trans-inuencing feature of the alkyl group causes the quinoline moiety of the tridentate ligand in 17 to coordinate relatively weakly to the Pd center and thus be highly susceptible to rearrangement. In this situation, the quinoline readily moves from the basal to the apical position via trigonal bipyramidal transition structure TS 18-19 lying only 3.1 kcal mol À1 above 18 to give another square pyramidal structure (19) in that the alkyl group occupies the apical position. This ligand movement turns on a repulsive interaction between the nitrogen lone pair and the lled d z 2 orbital, destabilizing the d z 2 orbital, resulting in two electrons being transferred from the Pd(II) to the iodine(III) center, furnishing 19. 11 As depicted in Fig. 3b, such an electron transfer turns off the Pd-N repulsive interaction and formally changes the oxidation state of the palladium center from +2 to +4. The spatial distribution for the LUMO of 19 (Fig. 3b reduced to the iodine(I) in 19. Due to an increase in population of the iodine p z orbital, the I-O a bond distance becomes longer upon going from 17 to 19 (Fig. 3a) and nally is cleaved in 19 supported by Wiberg bond index (WBI) analysis showing an almost zero-bond order (0.016) between the I and O a atoms. We found an excellent correlation between the I-O a bond distance and the population of the iodine p z orbital with R 2 ¼ 0.92 (Fig. 3d). 12 These computational results are clearly consistent with the oxidation of Pd(II) by the iodine(III) reagent through the unprecedented mechanism outlined in Fig. 3a.</p><!><p>Once the oxidative addition has taken place, the ve-coordinate Pd(IV) complex 19 with a pendant carboxylate group is formed. The sixth coordination site of this Pd(IV) complex can be lled by this pendant carboxylate group to furnish complex 20 (Fig. 4). However, owing to the strong trans-inuencing character of the alkyl moiety, the carboxylate binds weakly to the Pd(IV) center in 20, indicating a comparable stability for these two intermediates (19 and 20). Now, we turn our attention to investigation of the C-O reductive elimination from these two Pd(IV) complexes. Based on our calculations, this step can proceed via at least four different variants, as shown in Fig. 4. Accordingly, pathways A and B involve the inner-sphere reductive elimination from complexes 19 and 20 by crossing concerted transition structures TS 4 and TS 5 , respectively. These two pathways install the methoxy group of the oxidant on the nal product. In pathway C, the C-O reductive elimination occurs through an outersphere S N 2 mechanism involving the nucleophilic addition of the pendant carboxylate to the sp 3 carbon bonded to the palladium by passing through transition structure TS 6 . This pathway leads to formation of the carboxylated product. In pathway D, a methanol pre-activated by the pendant carboxylate nucleophilically attacks the sp 3 carbon to yield 22 via transition structure TS 21-22 . In this pathway, the solvent serves as the alkoxylation reagent, and addition of the methanol to the sp 3 carbon is accompanied by its deprotonation by the pendant carboxylate group. As can be seen from Fig. 4, TS 21-22 is much lower in energy than other transition structures, implying that pathway D is kinetically favored over pathways A, B, and C. However, when the solvent is not the alcohol, pathway D is turned off and thus, in this case, the reaction should preferentially proceed via pathway C, conrmed by the fact that TS 6 energetically lies considerably below TS 4 and TS 5 .</p><p>The energy prole given in Fig. 4 can answer several questions asked in the Introduction. For example, on that basis, one can explain why the OMe group on the oxidant does not appear in the product, why the solvent (alcohol) serves as the alkoxylation reagent and why when the solvent is not an alcohol, the carboxylate group is installed on the product.</p><!><p>Aer completion of the C-O reductive elimination via pathway D, intermediate 22 is formed. The dissociation of the carboxylic acid from the resulting intermediate via trigonal bipyramidal transition structure TS 22-23 yields 23 in which the methoxylated organic molecule acts as a tridentate ligand (Fig. 5). This complex subsequently participates in ligand exchange with one of the acetic acids produced from the CMD mechanism (Fig. 1) and generates the more stable intermediate 26 by passing through two transition structures TS 24-25 and TS 25-26 . The higher stability of 26 than 23 implies that the Pd(II) center prefers to coordinate to the acetate rather than the methoxy ligand. The addition of the second acetic acid to 26 via transition structure TS 26-27 gives 27 from which 28 is formed by a proton transfer from the coordinated acetic acid to the nitrogen atom of the methoxylated molecule. The overall activation barrier for formation of 28 from 22 is computed to be less than 15 kcal mol À1 . Finally, displacement of the coordinated product by substrate 1 leads to regeneration of active catalyst 13. From this point on, the second catalytic cycle begins. Since the active catalyst 13 lies 10.3 kcal mol À1 higher in energy than 26, the overall activation barrier to the C(sp 3 )-H activation in the second catalytic cycle increases to 30.0 kcal mol À1 (Fig. 5). Indeed, the ability of the N-H deprotonated product to form a tridentate complex causes this species to bind more strongly to the palladium(II) center than substrate 1. This feature retards the alkoxylation reaction by decreasing the activity of the catalyst, resulting in the process requiring a high temperature for completion.</p><!><p>The catalytic cycle shown in Scheme 4 summarizes our calculation results related to the mechanism of the title reaction. The reaction is initiated by coordination of substrate 1 to the Pd complex followed by deprotonation of the N-H and C(sp 3 )-H bonds by the OAc ligands to give cyclopalladated intermediate 16. 13 Subsequently, the substitution of BI-OMe for HOAc affords 17. The OMe ligand in the resultant intermediate ( 17) is then transferred from iodine(III) to palladium(II) to give iodonium 18 stabilized by an anionic palladium(II) complex. Aerward, intermediate 18 undergoes isomerization by moving the quinoline moiety from the basal to the apical position, triggering a redox process by promoting two electrons from Pd(II) to iodine(III), leading to formation of Pd(IV) complex 19 with a pendant carboxylate group. This isomerization not only promotes the Pd(II) oxidation but also sets the stage ready for the reductive elimination by formation of a ve-coordinate complex in which the reacting alkyl moiety occupies a position trans to the empty site. 14 The addition of a methanol (solvent) to the ensuing complex then affords 21. In this intermediate, the methanol is stabilized by a hydrogen bonding interaction with the pendant carboxylate group. Next, the C-O reductive elimination takes place by nucleophilic attack of the methanol on the sp 3 carbon via an outer-sphere S N 2 mechanism to yield 22. Our calculations predict that starting from iodonium salt 18, the oxidative addition and the reductive elimination steps are extremely fast with DG ‡ < 3.5 kcal mol À1 (Fig. 3 and 4). Later, the more stable complex 26 is furnished by following a series of chemical steps. The high stability of this species causes the regeneration of the active catalyst 13 to be considerably endergonic ($10 kcal mol À1 ), resulting in the overall activation barrier to the subsequent catalytic cycle increasing to 30.0 kcal mol À1 (Fig. 5). This nding clearly explains why the alkoxylation reaction developed by Rao et al. requires a high temperature for completion (Scheme 1).</p><!><p>In a separate study, Rao et al. used a similar method for preparation of acetals through double alkoxylation of the C(sp 3 )-H bonds of butyramide derivative 29 using BI-OMe as the oxidant and Pd(OAc) 2 as the catalyst (Scheme 5). 9b Based on the preliminary results, the authors proposed that the product of the rst alkoxylation is the substrate for the second one.</p><p>What interest us here is to explore how the electronic feature of the R 0 substituent on a substrate affects the alkoxylation mechanism. Fig. 6 compares the energy proles for alkoxylation of the substrates with R 0 ¼ H, Me, and OMe. Several points emerge from this comparison. First, the overall activation energy of the C-H activation step increases in the order R 0 ¼ H < Scheme 4 Catalytic cycle proposed by the DFT calculations for palladium-catalyzed C(sp 3 )-H alkoxylation using BI-OMe. Me < OMe. It is inferred from this result that the greater the electron donor property of the R group, the higher the activation barrier of the C-H activation. Indeed, a substituent with strong electron donating ability decreases the acidity of the hydrogen being abstracted, leading to the C-H activation to become more energy demanding. Second, the stability of three coordinate Fig. 6 Calculated energy profiles for palladium-catalyzed alkoxylation of the C(sp 3 )-H bond of butyramide derivatives with different R 0 substituent. Free energies (potential energies) are given in kcal mol À1 . complex 7 (7_R 0 ) is determined by the nature of the R 0 group; an R 0 group with strong electron donating ability increases the intrinsic stability of this coordinatively unsaturated species. It also nds that the relative energy of the transition structure TS 17-18 (TS 17-18_R 0 ) depends on the intrinsic stability of this three coordinate species and decreases in the order R ¼ H > Me > OMe. Third, although the formation of a Pd(IV) complex is unavoidable for the substrates with R 0 ¼ H and Me, this is not the case for the substrate with R 0 ¼ OMe. The IRC calculation shows that transition structure TS 17-18_OMe directly connects 17 -OMe to 19 0 _OMe. It follows that intermediates 18_OMe and 19_OMe are not local minimum and thus no Pd(IV) intermediate is formed in this case. Indeed, if we assume that 19_OMe is generated during the reaction, it is highly unstable and rapidly undergoes a redox process. This is because of the strong pdonor property of the OMe substituent, forcing the Pd IV -C(sp 3 ) s-bond in 19_OMe to be completely polarized toward the palladium center, giving zwitterion complex 19 0 _OMe in which the Pd center bears an oxidation state of +2. In this case, the C-O coupling process from this zwitterion complex does not involve the reductive elimination step, and instead, it takes place via nucleophilic addition of the carboxylate-activated MeOH to the oxonium ion. The DFT calculations show that the addition of the methanol to the oxonium ion is extremely fast and proceeds without involvement of an intermediate. As a result, it is evident from our calculations that a change in the R substituent leads to an abrupt alteration in the reaction mechanism. Scheme 6 shows the modied catalytic cycle for alkoxylation of the substrate with R ¼ OMe.</p><!><p>To evaluate whether the mechanism proposed in this study is applicable to interpret other alkoxylation reactions catalysed by Pd(II) complexes, we investigated the mechanism of the reaction shown in Scheme 7 developed by Shi. 5 In this reaction, iodobenzenediacetate (PIDA) is used as the oxidant. The energy prole given in Fig. 7</p><!><p>In this work, we performed DFT calculations to elucidate the mechanism of the alkoxylation of the C(sp 3 )-H bonds using hypervalent iodine(III) reagents (ArIX 2 ) catalysed by Pd(OAc) 2 . An unprecedented mechanism for this transformation is revealed by clarifying the oxidative addition step. The calculations indicate that this key step begins with the transfer of an X ligand from ArIX 2 to a Pd-alkyl intermediate. The ligand transfer forms a square pyramidal Pd(II) complex in which iodonium [ArIX] + occupies the apical position and is stabilized by interaction with the Pd d z 2 orbital. The resultant complex then undergoes an isomerization by moving the ligand trans to the Pd-alkyl bond to the apical position. This isomerization considerably destabilizes the d z 2 orbital, promoting the electron transfer from Pd(II) to I(III), resulting in the reduction of I(III) concomitant with the extrusion of the second X ligand as a free anion. The released anion then assists the alcohol (solvent) to nucleophilically attack the Pd(IV)-alkyl bond via an S N 2 mechanism to form a new C-O bond. The information reported in this study is important in enhancing our understanding of the fundamental processes that underpin many catalytic reactions mediated by iodine(III) reagents and catalyzed by Pd(II) complexes and could assist scientists to design new catalytic reactions.</p><!><p>Gaussian 16 (ref. 15) was used to fully optimize all the structures reported in this paper at the M06 level of theory. 16 For all the calculations, solvent effects were considered using the SMD solvation model 17 with methanol as the solvent. The SDD basis set 18 with effective core potential (ECP) was chosen to describe iodine and palladium. The 6-31G(d) basis set 19 was used for other atoms. This basis set combination will be referred to as BS1. Frequency calculations were carried out at the same level of theory as those for the structural optimization. Transition structures were located using the Berny algorithm. Intrinsic reaction coordinate (IRC) calculations were used to conrm the connectivity between transition structures and minima. 20 To further rene the energies obtained from the SMD/M06/BS1 calculations, we carried out single-point energy calculations in methanol using the M06 functional method for all of the structures with a larger basis set (BS2). BS2 utilizes the def2-TZVP basis set 21 on all atoms with an effective core potential including scalar relativistic effect for palladium and iodine. Tight convergence criterion was also employed to increase the accuracy of the calculations. In this work, the free energy for each species in solution was calculated using the following formula:</p><p>where DG 1 atm/1 M ¼ 1.89 kcal mol À1 is the free-energy change for compression of 1 mol of an ideal gas from 1 atm to the 1 M solution phase standard state.</p><p>An additional correction to Gibbs free energies was made to consider methanol concentration where a MeOH is directly involved in transformations. In such a case, the free energy of MeOH is described as follows:</p><p>G(MeOH) ¼ E(BS2) + G(BS1) À E(BS1) + DG 1 atm/1 M + RT ln(24.72)</p><p>where the last term corresponds to the free energy required to change the standard state of MeOH from 24.72 M to 1 M. 22 The orbital population analysis and determination of the WBI bond orders were carried out by the NBO7 program. 23</p>

Royal Society of Chemistry (RSC)

Relative Quantitation of Neuropeptides at Multiple Developmental Stages of the American Lobster Using N,N-Dimethyl Leucine Isobaric Tandem Mass Tags

Neuromodulators and neurotransmitters play important roles in neural network development. The quantitative changes of these signaling molecules often reflect their regulatory roles in physiological processes. Currently, several commercial tags (e.g., iTRAQ and TMT) have been widely used in proteomics. With reduced cost and higher labeling efficiency, we employed a set of custom-developed N,N-dimethyl leucine (DiLeu) 4-plex isobaric tandem mass tags as an attractive alternative for the relative quantitation of neuropeptides in brain tissue of American lobster Homarus americanus at multiple developmental stages. A general workflow for isobaric labeling of neuropeptides followed by LC-MS/MS analysis has been developed, including optimized sample handling procedures. Overall, we were able to quantify 18 trace-amount neuropeptides from 6 different families using a single adult brain as a control. The quantitation results indicated that the expressions of different neuropeptide families had significant changes over distinct developmental stages. Additionally, our data revealed intriguing elevated expression of neuropeptides in the early juvenile development stage. The methodology presented here advanced the workflow of DiLeu as an alternative labeling approach and the application of DiLeu-based quantitative peptidomics, which can be extended to areas beyond neuroscience.

relative_quantitation_of_neuropeptides_at_multiple_developmental_stages_of_the_american_lobster_usin

INTRODUCTION<!>DiLeu Labeling as a Quantitative Methodology<!>General Workflow<!>DiLeu Labeling of Multiple Developmental Stages of Lobster Brain for Relative Quantitation of Neuropeptides<!>Validation of DiLeu Labeling and MS Peptidomics Data Using Previous Literature<!>Neuropeptide Trends over Development and Function<!>CONCLUSIONS<!>Animals and Dissection<!>Brain Size Measurement<!>Tissue Extraction<!>Labeling of Peptide Standards and Neuropeptide Extracts<!>MALDI-Fourier Transform-Ion Cyclotron Resonance (FT-ICR) MS<!>Reversed Phase NanoLC-ESI-MS/MS<!>Data Analysis and Quantitation

<p>Decapod crustaceans, including the American lobster Homarus americanus, serve as a classic model for neurobiology studies due to their simple and well-characterized neurological systems.1–3 The stomatogastric nervous system (STNS) and central nervous system (CNS) are two major neurological systems in the American lobster that contain several important neural organs. Many behaviors such as swallowing, chewing, circadian rhythms, and feeding are controlled by these two systems.4,5 In particular, the brain plays very important role in neuromodulation. For example, sensory input to the brain of decapod crustaceans is provided by compound eyes mediating vision or an array of specialized sensilla located at antennules mediating all other sensory modalities.6 The brain is also connected with the STNS by projection through the commissural ganglia (CoG) or directly into the stomatogastric ganglion (STG) to regulate the motor patterns in the stomach.7,8 Considering this critical location and interweaving nerve connections, the identification of signaling molecules in this highly complex neuronal structure needs to be addressed first, which could provide molecular clues on the relationship between certain neurotransmitters/hormones and behaviors.</p><p>However, the dynamic changes of these signaling molecules throughout development have not been well studied. Neurogenesis can occur at any time in any developmental stage.9 In H. americanus, it takes about 7 years and over 20 molts for a lobster to become an adult. During this process, chemosensory receptors are continuously added in the antennules, and the new sensory axons grow into the primary chemosensory processing areas in the brain.10 Another study showed that the brain size as well as deutocerebral organization changes as the development proceeds.11 Therefore, a developmental study would be important to elucidate the changes of neurotransmitters complements that are occurring in the brain, in particular neuropeptides, which are known to be heavily involved in proper development.12 This information can offer clues on the regulation of neuropeptides and other neurotransmitters on growth and development.</p><p>Initially, immunocytochemistry and physiological studies have been employed to characterize the complement and acquisition of neuropeptides and their receptors.8,13–20 A study performed on fruit fly Drosophila compared the neuropeptide complements and distributions in the CNS of larvae and adult and showed distinctions between these two developmental stages.13 There are also reports on the CNS of the crustacean and other arthropods that describe the immunochemical reactivity of some neuropeptides like FMRFamide.14–17 Specifically, in H. americanus, it has been shown that the synthesis of neuropeptides in the STNS of the American lobster occurs sequentially, and some neuropeptides are not present until larval stages.8,18–20 Researchers now ask about the global neuropeptidomic changes in the brain, specifically if they are the same at each stage of development. If they are the same, we can ask if several neuropeptide families show distinct abundance differences and finally if these abundances can be compared to the brain size at each developmental stage.</p><p>To answer the above quantitative questions, mass spectrometry (MS) has been commonly utilized. MS was proven to be a sensitive and accurate tool for characterizing neuropeptides and other neurotransmitters, especially in crustaceans.1,2,21,22 Many of the previous immunocytochemical results have been confirmed and expanded by MS since the latter can be more specific to neuropeptide isoforms.23–28 A combination of online liquid chromatography (LC) tandem MS (MS/MS) and matrix-assisted laser desorption/ionization (MALDI)-MS methodology has been utilized enabling hundreds of neuropeptides to be characterized from brain and other organs of adult American lobster, mainly due to the fast speed, high sensitivity, and other capabilities of MS technology.5,29</p><p>A common approach for relative quantitation is through mass-difference based labeling to measure relative peak areas of MS spectra.5,30–33 This approach detects peptide levels in two different samples by labeling the peptides with either light or heavy stable isotopes, mixing them equally, and analysis by MS. Nevertheless, this classically binary (2-plex) set of reagents suffers from the limitations of quantitation of only 2–3 samples at a time and the increase of complexity of the spectra, although several recent instrument developments have made multiplexing more accessible.34–37 With the advent of isobaric tags for relative and absolute quantitation (iTRAQ) and tandem mass tag (TMT) reagents, the comparisons of more than 2–3 samples could be done simultaneously at the MS/MS level.35,37 In this method, relative quantitation can be performed by comparing the intensities of reporter ions produced by collision-induced dissociation (CID), although this means quantitation can only be done on molecules selected for MS/MS. Our group has developed a custom isobaric tag set entitled N,N-dimethyl leucine (DiLeu), which allows for the comparison of up to 12 samples.34,38 This labeling approach produces low mass immonium a1 ions (m/z 115–118) at the MS/MS level and has proven to be comparable with iTRAQ for quantitation but with much lower experimental cost.34,38</p><p>Herein, we utilized 4-plex DiLeu as isobaric tags to optimize sample handling and reaction conditions for neuropeptidomic studies. The optimized methods (Scheme 1) were applied to four different developmental stages of the American lobster brains to compare the neuropeptide contents and abundances in brain tissues using electrospray ionization-quadrupole time-of-flight (ESI-QTOF) MS. This manuscript represents the first application of DiLeu reagents for large-scale crustacean neuro-peptide expression analysis, offering an alternative strategy for multiplexed quantitative peptidomics.</p><!><p>The basic principles and synthesis of the DiLeu isobaric tagging reagent have been introduced elsewhere.34,38 Briefly, each tag contains a reporter, a balance, and amine reactive group. The final group is a triazine ester, which targets the N-terminus and exposed lysine groups of peptides (Supporting Information Scheme 1). Successful labeling results in a mass shift of 145.1 Da per amine group, which is the addition of the balance and reporter groups. As shown in Figure 1a and b, almost complete labeling with a DiLeu tag is observed for a standard peptide (m/z 1060.57 RPPGFSPFR). When four differentially labeled samples are combined and subjected to ESI-QTOF MS/MS analysis, distinctive fragmentation patterns can be easily obtained, including intense a1 ions (m/z 115.1, 116.1, 117.1, and 118.1) serving as quantitative reporters for the four distinct samples. The relative heights of these peaks correspond with the proportions of the labeled peptides in each sample, allowing relative quantitation between those four samples (Figure 1c and d).</p><p>Standard samples were tested to establish the robustness and reproducibility of DiLeu quantitation on neuropeptides for this study. Figure 2a shows the reporter ions for labeling four aliquots of a neuropeptide standard SGKWSNLRGAWamide (m/z 1260.66) that were mixed in equal amounts. Quantitative 1:1:1:1 confirms the reliability of DiLeu for neuropeptide study. A 10:5:2:1 mixing ratio was also investigated, proving that DiLeu is capable of accurate quantitation of neuropeptides of a wide dynamic range (Figure 2b).</p><!><p>The isobaric labeling workflow to analyze the protein/peptides in plasma and cells are well documented.35,37,39,40 However, to our knowledge, there is no literature detailing the isobaric labeling steps for neuropeptides. This is likely due to the fact that the detection of neuropeptides, which are inherently present in low abundance, is often masked by other highly concentrated molecules, including salts, lipids, and proteins, in complex tissue samples.41 It should be noted that a similarly structured tag, isotopic DiLeu (iDiLeu), has been used to label neuropeptides, although only clean, standard solutions were used.42 Here, in order to analyze developmental changes, a simple and reliable workflow was developed to analyze neuropeptides using DiLeu isobaric tagging strategies (Scheme 1). Briefly, animals of four developmental stages were dissected and the brains were pooled respectively for tissue extraction. The homogenate was centrifuged, and the supernatant for each stage was dried down. After reconstitution in aqueous 0.1% formic acid, a desalting step was needed before labeling reaction to allow efficient and reproducible coupling to the labeling reagents. The labeled samples were then combined and ready for LC separation and MS analysis.</p><p>As mentioned before, sample cleanup is necessary before labeling to improve the reaction efficiency. In order to improve this aspect of our study, we optimized the desalting process of C18 ZipTips from the manufacturer's protocol and compared it to a magnetic beads clean up method. First, the dried down samples were reconstituted in maximum tip volume (10 μL) to reduce the chance of a possible blockage. Second, the reconstitution was followed by a short, high speed centrifugation to remove any lipids left which could also block the Ziptip. Finally, the volume of the elution step was increased 100 μL instead of ∼3 μL to ensure maximum peptide elution and recovery. Overall, the optimized Ziptip method out-performed the magnetic beads (Figure 3a–d) and improved the reproducibility (Figure 5). In a recent paper, Ziptips and magnetic beads were also compared for human serum peptide profiling, and it was determined that Ziptips were a better desalting method in spectral quality, reproducibility, and time consumption.43</p><p>In label-free and isotopic labeling of neuropeptides, acidified methanol and other acidified organic solvents have been widely used in extracting neuropeptides.41 Nonetheless, caution needs to be given while using this solvent with DiLeu since the labeling reaction that follows extraction requires basic conditions.38 Previously dimethylformamide (DMF) has been utilized to extract neuropeptides before DiLeu labeling due to their compatibility.34,38,44 For this study, the extraction capability of acidified methanol (Figure 3a, b) and DMF (Figure 3c, d) was compared, and it was determined that DMF was better based on the number and intensity of the neuropeptide peaks, especially in conjunction with the optimized Ziptip desalting method.</p><p>To confirm our results, DiLeu tags were employed to label brain extracts obtained from the above extraction and purification methods to directly compare all the possible combinations. The labels corresponding to each of the conditions are shown in Figure 3e. Interestingly, the m/z 117.1 peak is higher than m/z 116.1 (Figure 3f), even though the spectral features (Figure 3c) corresponding to m/z 117.1 was lower. This further confirmed the compatibility of labeling with DMF over acidified methanol as an extraction solvent. Overall, m/z 118.1, corresponding to DMF extraction with Ziptip desalting, which was the preferred method individually, showed the highest intensity. The final workflow for DiLeu analysis of neuropeptide is shown in Scheme 1.</p><!><p>It is known that neuropeptides are acquired sequentially in the STG of the lobster, with some neuropeptides not appearing until larval stages.20 It would be interesting to explore whether similar trends occur in the brain. To assess neuropeptide expression during development, animals of embryo, larvae, juvenile, and adult stages were utilized.</p><p>Tandem MS spectra of labeled peptide m/z 1084. 63 + 290.2 (HI/LASLYKPR) and FMRF peptide m/z 1337.69 + 145.1 (FSHDRNFLRFamide) are highlighted in Figure 4. Previously, DiLeu labeling was shown to improve MS/MS fragmentation in de novo sequencing, allowing us to obtain highly confident identifications even with low resolution instrumentation.38 In de novo sequencing of m/z 1084.63, we noticed there was a fragment mass matching lysine, suggesting two labels, 145.1*2, were added to the original peptide. Due to the isobaric masses of leucine and isoleucine, the exact amino acid for the second residue has not yet been confirmed. For peptide m/z 1337.69, it was determined to belong to the FMRFamide family. One common feature of this family is that there is an almost complete y ion series observed. This may be because of the presence of highly basic arginine residue at the C-terminus, which would suppress the formation of b ions. With the aid of DiLeu tags, reliable fragmentation patterns were obtained and used for high confidence neuropeptide identifications. Overall, there were 18 neuropeptides from 6 different families identified and quantified (Supporting Information Table 1). Low level of incomplete labeling was observed for SIFamide peptide VYRKPPFNGSI-Famide (m/z 1423.78) and HI/LASLYKPR (m/z 1084.62) in which lysine residue competed with N terminal residue for labeling. The overall labeling efficiency was ∼90% and only fully labeled peptide sequences were used for quantitation.</p><p>As representative data, Figure 5a and c illustrate the reporter ions of labeled FMRFamide peptide GDRNFLRFamide (m/z 1023.55) and tachykinin peptide APSGFLGM(O)Ramide (m/z 950.49), showing the significant abundance differences among multiple stages as well as distinctive patterns among the selected peptides. Reversed labeling experiments have also been performed in order to eliminate any bias generated by labeling procedure (Figure 5b and d). Considering the huge differences in tissue size among multiple developmental stages and possibly wide dynamic range in neuropeptide abundance, it should be noted that we use more animals in early stages for comparative analyses. The total neuropeptide abundances from pooled animals have been converted to neuropeptide abundance in each individual animal, and the ratio of peak intensity between animal of each stage to adult stage is shown in Figure 6 and detailed in Supporting Information Table 1.</p><p>Several trends have been identified including: (1) Neuro-peptides are gradually acquired in the maturation process. All neuropeptides exhibit higher abundances as the lobster matures to later stages and the differences between developmental stages are significant (p value < 0.01, Figure 6 and Supporting Information Table 1). This is consistent with the continuous addition of receptors into the antennules11 and the increase in the number of neurons from previous immuno staining experiments.;13,45 (2) The abundances for different neuro-peptide families are distinct, with individual familial isoforms sharing similarities. This might suggest that the peptides in one family were expressed and processed simultaneously from the neuropeptide gene, and their release and actions also occur concurrently in the development and maturation of a neural network. However, two exceptions are noted, including tachykinin APSGFLGMRG (m/z 992.5) and RFamide QDL-DHVFLRFamide (m/z 1288.68). These two neuropeptides show higher abundances in embryonic and larvae 2 stages as compared to other neuropeptide family isoforms, although these differences are less obvious in juvenile stage. In general, neuropeptides originate from large, inactive preprohormones which undergo a series of enzymatic processing steps and proteolytic cleavages and modifications, with the last step being amidation at the C-terminus of glycine-ended sequences.46 If the glycine-ended form is converted to the amidated form as the development proceeds, that may contribute to the sharp increase in m/z 934.49 in later stages, which is the amidation product of m/z 992.5. Unfortunately, we were unable to quantify m/z 1271.68 due to detection limit of the instrument. (3) Different neuropeptide families exhibit distinct differences in abundances as a function of development. For example, tachykinin peptides exhibit about 1000-fold differences in abundance when transitioning from the embryonic stage to adult stage. Other peptides such as m/z 1084.63 display only about 100 times difference in relative abundance, even though the size difference of the lobster between these two stages is about 500 times. This might be due to the different roles of distinct neuropeptide families.</p><!><p>To better understand the relationship of neuropeptide amount with the brain, we further calculated the peptide concentration by considering the size of the brain. Notably, the neuropeptides are not evenly distributed in the brain.47,48 The conversion here is just an approximation to facilitate neurobiologists' understanding of relationship of total neuropeptide abundance as a function of lobster developmental stage. The calculated results for four selected peptides are illustrated in Figure 7. It was noted that the neuropeptide concentrations of these four representative neuropeptides were consistently highest in the juvenile seventh stage. Previous studies that measured the area of olfactory lobe (OL) and accessory lobe (AL) indicated that, in early juvenile stage, the volume of olfactory OL and AL took up the largest portion of the brain (Supporting Information Figure S1).11 By comparing the trend observed in our study, which describes the concentration change of neuropeptides, to the one measuring the volume of chemosensory processing areas (OL and AL), we observed very similar pattern. This suggested that the OL and AL regions may contain neurons involved in the synthesis and release of neuropeptides. It has been known that increase in number and diameter of the neural processes account for the growth of OL and AL, and an increase in the bulk of cortical glial tissue might contribute to the increase in bulk of the brain.11 Therefore, at around seventh juvenile stages, more new sensory axons are projected into glomeruli, which may suggest that this period is a fast growing or function acquiring period. More in-depth investigation is needed to further explore this speculation.</p><!><p>The trend for m/z 934.49 is shown in blue trace (Figure 7). Tachykinin related peptides (TRPs) are a family of neuropeptides which usually exist at high amounts in the adult lobster. The neuromodulation of tackykinin in the CNS may depend on the specific circuit and transmitters with tachykinin colocalized with TRPs.49,50 They are also involved in the modulation of photoreceptor sensitivity, olfactory sensory processing, and higher motor control.51,52 Our study shows that expression of this family is in very low concentration in the embryonic stage, but increases dramatically during the maturation process to the adult stage (Figure 6). Similarly, in a study of TRPs expression in different developmental stages of fruit fly Drosophila, the number of TRP-expressing neuronal cell bodies in the brain and ventral nerve cord increases during larval development and significantly increases after later pupal development to adult.45 In two other reports comparing the amplitude of neuropeptides in the STNS, the abundance of neuropeptides in this family showed significant increase from Larvae 1 to Juvenile 7 stages, which is consistent with our observation that tackykinin family was acquired quickly in the brain after the embryonic hatch.23,53 Further physiological experiments are needed to establish the precise role of this peptide family in development.</p><p>Val1-SIFamide also exhibits significant differential expression between early and later developmental stages. In contrast to tachykinin, the peak abundance of Val1-SIFamide of juvenile stage is close to that of an adult (Figure 6), indicating that this neuropeptide level changes dramatically between larvae and juvenile stage. This peptide has been identified in the STNS and the CNS of the American lobster with high abundance previously.27,29,54 In the STNS, the amplitude of this neuro-peptide family exhibits relatively stable expression levels throughout developmental stages, which differentiates the expression in the brain. It can activate several components of the pyloric motor pattern and the peptide exerts neuro-modulatory function on the STG via local release instead of hormonal regulation.54 The highly conserved sequences among several crustacean species and high abundance of this peptide suggest important function of this peptide.3 Previous studies have shown that SIFamide neuropeptides are involved in modulating sexual behavior, aggression, as well as visual, tactile, and olfactory stimuli.55 Mass spectral imaging data revealed that tachykinin and SIFamide were expressed in the OL and AL areas,10 which further confirms that the synthesis and release of neuropeptides might be related to OL and AL.</p><p>FMRFamide, orcokinin, and orcokinin related peptides are three large peptide families in the lobster brain. They can also be detected in many other organs of the American lobster.48 These three families display similar patterns in Figure 6, and the representative trend of orcokinin at m/z 1502.69 is shown in green trace in Figure 7. Previous reports indicated that FMRFamide and orcokinins appeared in the STNS as early as midembryonic stage (E50).8,56 A majority of the neuropeptides reported here can be found in the predicted preprohormones from a gene.29,57 In our study, we observed that neuropeptides in these three families share similar intensity patterns, and there was about a 250 times increase in the amount from embryonic to adult stage (Figure 6). The similar trends of individual isoforms within the same family might indicate that peptides are expressed and processed concurrently. The concentration calculation (described above) revealed a slight decrease from embryonic stage to larval second stage. This decrease might be caused by the morphological change in the brain, as we noticed that the expression of neuropeptides of these three families was concentrated on regions other than OL and AL (48 and unpublished data). However, consistent with other neuro-modulators, these peptides still exhibited the highest abundance in juvenile seventh stage (Figure 7).</p><p>The peptide at m/z 1084.63 is highly abundant in early stages but shows a remarkably lower abundance compared to other neuropeptides at later developmental stages (Figure 4). In this quantitative study (Figure 6), the intensity of this peptide in the juvenile stage is very similar to that of the adult, suggesting an early acquisition of this peptide during development. In Figure 7, this peptide exhibits its highest concentration in juvenile seventh stage. To our knowledge, there is no physiological study on this peptide yet, and it would be interesting to determine its potential relationship with development.</p><!><p>In summary, we demonstrated successful application of the recently developed DiLeu labeling tags to a neuropeptide quantitation study. The DiLeu labeling technique was optimized for neuropeptide detection in brain tissue, which lays the foundation for future application of this quantitative methodology for other tissues and other -omics studies. Relative quantitation of neuropeptides across multiple developmental stages was investigated to quantify 18 neuropeptides, showing that isoforms from the same family displayed similar trends of increase as lobster matures. A majority of neuropeptides displayed the highest concentration in juvenile seventh stage, suggesting potential important morphological and functional developments at this stage. Overall, our study systematically made use of the custom 4-plex DiLeu isobaric labeling technique for MS quantitation to study the neuropeptide changes during development, which will enable better characterization of the roles of neuropeptides play in maturation process.</p><!><p>Early stage lobsters were obtained from New England Aquarium (Boston, MA), and adult lobsters were ordered from Maine Lobster Direct Web site (https://www.mainelobsterdirect. com/). Ages of the lobster were estimated using an eye index scale for embryos58 or carapace length in the lobster growth chart for hatched animals. Embryos used in this study were in a stage of 85%–100% embryonic development (E85–E100). The dissection of lobsters was described previously.8,23,59 Four developmental stages were investigated, including adult, juvenile seventh, larvae second, and embryo.</p><!><p>In order to calculate the concentrations of neuropeptides across the whole brain, the sizes of brains of different developmental stages were measured under a microscope. The brains were assumed as cuboid for easy measurement and calculation for volume. Five animals of each stage were measured, and no significant differences were observed among individuals. The averaged values were compared with the previous studies on lobster brain volume as a function of developmental stage11 and the differences observed were less than 10%. Therefore, we used the averaged values from our measurements as brain sizes, which were 0.035 mm3 for the brain of embryo, 0.11 mm3 for larvae second stage, 0.85 mm3 for juvenile seventh stage, and 20 mm3 for adult.</p><!><p>Lobster brains were dissected out from animals in chilled saline and then transferred to 20 μL of cold DMF. Brains at each stage (1 brain for adult, 4 for juvenile seventh, 90 for larvae second, and 90 for embryo in each replicate) were pooled after dissection and homogenized using 60 μL of DMF in a 0.1 mL tissue grinder (Wheaton Inc., Millville, NJ). The homogenate was transferred to 0.6 mL microtube and centrifuged at 16,100g for 10 min in an Eppendorf 5415 D microcentrifuge (Brinkmann Instruments Inc., Westbury, NY). The resulting supernatant was saved, and the pellet was re-extracted twice using the same method. Supernatants from three extractions were combined and then dried down in a Savant SC 110 SpeedVac concentrator (Thermo Electron Corporation, West Palm Beach, FL). Subsequently, the brain extract was resuspended in 10 μL of aqueous 0.1% formic acid (v/v) and then desalted by ZipTipC18 Pipette Tip (Millipore Corporation, Billerica, MA) or magnetic Dynabeads (Invitrogen, Carlsbad, CA). Tissue extractions and desalting were performed for three different replicates. In order to reduce errors generated by extraction, in the next two replicates experiment, we used different number of animals to compensate for differences in brain size at various developmental stages. We used 0.5 brain for adult, 2 for juvenile seventh, 45 for larvae second, and 45 for embryo extracts for pooling. The ratio of animals throughout development was kept the same for easy comparison.</p><!><p>Four DiLeu labels were activated into their triazine ester form. Half a milligram of DiLeu in 25 μL of DMF was combined with 0.93 mg of DMTMM and 0.37 μL NMM. Mixing was performed at room temperature for 1 h and used immediately. The general labeling strategy is shown in Supporting Information Scheme 1. The standards or real samples were then labeled with 10 μL of one of the four DiLeu mass tags with the presence of 5 μL ethanol. Samples were then labeled at room temperature for 1.5 h and then quenched for 30 min by adding 100 μL of water. Quenched solutions were dried under Speedvac. Finally, the sample was resuspended in 20 μL of aqueous 0.1% formic acid (v/v).</p><!><p>Varian/IonSpec FTMS (Lake Forest, CA) was equipped with a 7.0 T actively shielded superconducting magnet. The FTMS instrument consisted of an external high-pressure MALDI source. A 355 nm Nd:YAG laser (Laser Science, Franklin, MA) was used to create ions followed by accumulation in the external hexapole storage trap before being transferred through a quadrupole ion guide to the ICR cell. All data were collected in positive ion mode. The ions were excited prior to detection with a radio frequency sweep beginning at 7050 ms with a width of 4 ms and amplitude of 150 V base to peak. Detection was performed in broadband mode from m/z 108.00 to 2500.00.</p><!><p>Labeled samples were combined and then analyzed using Waters nanoAcquity UPLC system online coupled to Waters Micromass QTOF mass spectrometer (Waters Corp., Milford, MA). An aliquot of 6 μL of solution was injected and trapped onto a C18 trap column (Zorbax 300SB-C18 Nano trapping column, Agilent Technologies, Santa Clara, CA) for 10 min, and eluted onto a homemade C18 column (75 μm × 150 mm, 3 μm, 100 Å) using a linear gradient (0.3 μL/min) from 5% buffer B [0.1% formic acid in acetonitrile (Fisher Scientific, Pittsburgh, PA)] to 50% buffer B over 60 min. Buffer A was 0.1% formic acid. The nanoflow ESI source conditions were set as follows: capillary voltage 3200 V, sample cone voltage 35 V, extraction cone voltage 15 V, source temperature 120 °C, and cone gas (N2) 10 l/h. MS survey scan range was from m/z 400–1800, and the MS/MS scan was from m/z 50–1800.</p><!><p>The MS/MS de novo sequencing was performed by manual sequencing and automatic sequencing by Waters Masslynx peptide sequencing software (PepSeq) (Waters Corp., Milford, MA). DiLeu labeled neuropeptides were then identified with PepSeq N-terminal labeling, lysine labeling, C-terminal amidation, and methionine oxidation selected for modifications. The precursor error tolerance was set to be <100 ppm. Identified, labeled peptide spectra were quantified by comparing the intensities of the MS/MS reporter ions 115.1, 116.1, 117.1, and 118.1. Quantitative values were calculated by dividing each reporter ion peak height by the number of animals used for each stage followed by normalization to the abundance of adult brain. The experiments were performed in three biological replicates, and the mean and standard deviation were reported. One way ANOVA was used to determine the difference among four developmental stages. For concentration comparisons, the quantitative values obtained above were further divided by the sizes of brain of each stage.</p>

PubMed Author Manuscript

Effects of the presence and absence of amino acids on translation,\nsignaling and long-term depression in hippocampal slices from\nFmr1 knockout mice

Fragile X syndrome (FXS) is caused by silencing of the FMR1 gene and consequent absence of its protein product, fragile X mental retardation protein (FMRP). FMRP is an RNA-binding protein that can suppress translation. The absence of FMRP leads to symptoms of FXS including intellectual disability and has been proposed to lead to abnormalities in synaptic plasticity. Synaptic plasticity, protein synthesis and cellular growth pathways have been studied extensively in hippocampal slices from a mouse model of FXS (Fmr1 KO). Enhanced metabotropic glutamate receptor 5 (mGluR5)-dependent long-term depression (LTD), increased rates of protein synthesis, and effects on signaling molecules have been reported. These phenotypes were found under amino acid starvation, a condition that has widespread, powerful effects on activation and translation of proteins involved in regulating protein synthesis. We asked if this nonphysiological condition could have effects on Fmr1 KO phenotypes reported in hippocampal slices. We performed hippocampal slice experiments in the presence and absence of amino acids. We measured rates of incorporation of a radiolabeled amino acid into protein to determine protein synthesis rates. By means of Western blots, we assessed relative levels of total and phosphorylated forms of proteins involved in signaling pathways regulating translation. We measured evoked field potentials in area CA1 to assess the strength of the long-term depression response to mGluR activation. In the absence of amino acids, we replicate many of the reported findings in Fmr1 KO hippocampal slices, but in the more physiological condition of inclusion of amino acids in the medium, we did not find evidence of enhanced mGluR5-dependent LTD. Activation of mGluR5 increased protein synthesis in both wild type and Fmr1 KO. Moreover, mGluR5-activation increased eIF2\xce\xb1 phosphorylation and decreased phosphorylation of p70S6k in slices from Fmr1 KO. We propose that the eIF2\xce\xb1 response is a cellular attempt to compensate for the lack of regulation of translation by FMRP. Our findings call for a re-examination of the mGluR theory of FXS.

effects_of_the_presence_and_absence_of_amino_acids_on_translation,\nsignaling_and_long-term_depressi

Introduction<!>Chemical, reagents and drugs<!>Animals<!>Hippocampal Slice Preparation<!>Measurement of Rates of Protein Synthesis<!>Western Blot<!>Electrophysiology Recordings<!>Statistical Analysis<!>Protein synthesis response of hippocampal slices to group 1 mGluR-activation\nby DHPG depends on amino acids in the incubation medium<!>Effects of group 1 mGluR-activation with DHPG on the mTOR pathway in\nhippocampal slices from WT and Fmr1 KO mice depends on the presence of amino\nacids in the medium<!>Effects of group 1 mGluR-activation with DHPG on phosphorylation of\neIF2\xce\xb1 in hippocampal slices from control and Fmr1 KO mice depends on the\npresence of amino acids in the medium<!>Effects of group 1 mGluR-activation with DHPG on LTD in hippocampal slices\nfrom control and Fmr1 KO mice depends on the presence of amino acids in the\nmedium<!>Discussion

<p>In fragile X syndrome (FXS), FMR1, the gene coding for fragile X mental retardation protein (FMRP), is silenced due to an expanded CGG repeat sequence in the 5'untranslated region of the gene. The absence of FMRP results in a constellation of symptoms including intellectual disability, sensory hypersensitivity, hyperactivity, autistic-like behavior and susceptibility to seizures. FMRP contains RNA-binding motifs including two K homology (KH) domains and an arginine/glycine-rich RNA-binding region (RGG box) and associates with actively translating ribosomes (Ashley et al. 1993; Feng et al. 1997; Siomi et al. 1993). In vitro evidence supports a role for FMRP as a suppressor of translation (Laggerbauer et al. 2001; Li et al. 2001).</p><p>Our in vivo studies of regional rates of cerebral protein synthesis (rCPS) in adult Fmr1 knockout (KO) and wild type (WT) mice indicated elevated hippocampal rCPS in Fmr1 KO mice (Qin et al. 2005). In accord with these findings, experiments in hippocampal slices from Fmr1 KO mice showed increased translation rates over WT (Dölen et al. 2007; Osterweil et al. 2010). Experiments in hippocampal slices designed to test the efficacy of a plasticity response demonstrated enhanced group 1 metabotropic glutamate receptor (mGluR)-dependent long-term depression (LTD) in slices from Fmr1 KO mice (Huber et al. 2002). These findings led to the advancement of the mGluR theory of FXS (Bear et al. 2004) which proposed that many of the symptoms of FXS might be explained by excessive protein synthesis downstream of group 1 mGluR activation. Further exploration of the mGluR theory in hippocampal slice experiments found that activation of group 1 mGluRs with dihydroxyphenylglycine (DHPG) resulted in increased rates of translation in WT but not Fmr1 KO slices and increased phosphorylation of extracellular signal-regulated kinase (ERK1/2) in both WT and Fmr1 KO slices (Osterweil et al. 2010). A thorough examination of effects of mGluR activation on components of the phosphoinositide 3-kinase (PI3K)-protein kinase B (Akt)-mammalian target of rapamycin (mTORC1) signaling pathways showed no differences in slices from WT and Fmr1 KO mice (Osterweil et al. 2010). Ultimately, it was proposed that treatment of FXS with an mGluR negative allosteric modulator would rescue the excessive protein synthesis and ameliorate symptoms in this disease.</p><p>Accordingly, we have attempted to establish an in vitro assay for rates of translation in hippocampal slices to facilitate screening of compounds as potential therapeutics in this disease. In the development of our assay we made specific modifications to published methods with the intention of creating a more physiologically relevant system. Previous experiments in the hippocampal slice preparation were performed in the absence of amino acids in the medium; we denote this condition as amino acid starvation (AAS). In our preparation, we adjusted the incubation medium to include unlabeled essential amino acids at physiological concentrations; we use the term amino acid replete (AAR) to describe this condition. In addition, we used radiolabeled leucine (L-[4,5-3H]leucine) as the tracer amino acid to enable the calculation of a rate of incorporation of leucine into protein. Our results do not agree with published reports, and we present herein an explanation for the differences between our results and those published.</p><p>Our experiments indicate that the presence or absence of essential amino acids in the medium have a profound influence on translation and the differential effects of activation of group 1 mGluRs on hippocampal slices from WT and Fmr1 KO mice. In the presence of amino acids, we found no difference in baseline rates of translation in hippocampal slices from WT and Fmr1 KO mice. Activation of group 1 mGluRs resulted in increases in translation rates in both genotypes. Effects on signaling molecules involved in the regulation of translation indicate that, following treatment with DHPG, phosphorylation of the alpha subunit of eukaryotic initiation factor 2 (eIF2α) and p70S6k were increased and decreased respectively in slices from Fmr1 KO but not WT mice. These results indicate that pathways regulating translation in response to group 1 mGluR activation are differentially affected in the absence of FMRP. We further tested for enhanced group 1 mGluR-activated LTD in Fmr1 KO mice as reported previously (Huber et al. 2002), and found no difference in LTD between slices from WT and Fmr1 KO mice in AAR conditions.</p><!><p>All reagents used in the current experiment were purchased from Sigma-Aldrich (St. Louis, MI, USA) unless noted otherwise. Reagents used in the preparation of artificial cerebrospinal fluid (aCSF) included; sucrose (Cat. No. 84097), sodium chloride (Cat. No. 57653), potassium chloride (Cat. No. P5405), NaH2PO4 (Cat. No. 7892), NaHCO3 (Cat. No. S6014), D-(+)-glucose (Cat. No. G7021), HEPES (Cat. No. H4034), (+)-sodium L-ascorbate (Cat. No. A4034), thiourea (Cat. No. T7875), sodium pyruvate (Cat. No. P2256), N-acetyl-L-cysteine (Cat. No. A7250), MEM Amino Acids Solution (50X) (Cat. No. 11130), NaOH (Cat. No. S8045). Drug treatments used in our experiments were (RS)-3,5-DHPG; (Tocris; Ellisville, MO, US) (Cat. No. 0342/1) and 2-p-methoxyphenylmethyl-3-acetoxy-4-hydroxypyrrolidine (anisomycin; Tocris) (Cat. No. 1290/10). Other reagents used include ethylenedinitrilotetraacetic acid (EDTA; 0.5 M, pH 8.0; Quality Biological, Gaithersburg, MD, USA) (Cat. No. 351–027), ethylene glycol-bis(2-aminoethylether)-N,N,N′,N′-tetraacetic acid (EGTA) (Cat. No. E3889), Triton™ X-100 Cat. No. (X100), Halt™ Protease Inhibitor Cocktail (100X; Thermo Fisher Scientific, Waltham, MA) (Cat. No. 78425), trichloroacetic acid (TCA) (Cat. No. T6399), Halt Protease and Phosphatase Inhibitor Cocktail, 100X(Thermo Fisher Scientific, Waltham, MA, USA) (Cat.No. 78441), T-PER™ Tissue Protein Extraction Reagent (Thermo Fisher Scientific, Waltham, MA, USA) (Cat. No. 78510), and 2-mercaptoethanol (Cat. No. M6250).</p><!><p>Male Fmr1 KO (Jackson Labs: B6.129P2-Fmr1tm1Cgr/J Stock No. 003025) (n = 116) and WT (n = 121) mice on a C57BL/6J background were generated from heterozygous Fmr1 KO female and WT male breeder pairs in-house. Animals were weaned at 21 days of age. Genotyping was performed on tail biopsies as previously described (Qin et al. 2005). The following primers were used: 5-ATCTAGTCATGCTATGGATATCAGC-3 and 5-GTGGGCTCTATGGCTTCTGAGG-3primers were used to screen for the presence or absence of the mutant allele. Animals were group housed (2–5/cage) in a climate-controlled facility with access to food and water ad libitum. All experiments were performed in accordance with the National Institutes of Health Guidelines on the Care and Use of Animals and approved by the National Institute of Mental Health Animal Care and Use Committee (LCM-07). This study was not pre-registered. Animal assignment was not randomized as it was not relevant for the experimental goals. Additionally, animals were used as soon as available. Numbers were assigned to animals by another researcher so that those conducting the experiment were blinded to animal genotype during analysis of rates of protein synthesis, Western Blots, and LTD recordings. All efforts were made to minimize animal suffering.</p><!><p>The compositions of media in which hippocampal slices were prepared, incubated, and dissected are given in Table 1. For experiments under AAR conditions all media included a full complement of essential amino acids. Similarly, for experiments under amino acid starvation (AAS) conditions, all media lacked amino acids. Unanesthetized animals (postnatal day (P)30-P51) were euthanized by decapitation between 9 and 11:30 am, and brains were rapidly removed and placed in ice cold Sucrose aCSF for 45 s (Figure 1). The frontal cortex was removed, the remaining tissue was glued to a puck, and slices (350 μm in thickness) were prepared in ice cold Sucrose aCSF, bubbled with 95% O2 / 5% CO2, by means of a Leica VT1000 S vibratome (Leica, Deerfield, IL, USA). Hippocampal slices were used in three independent sets of experiments as described below. Each set of experiments had its own endpoint. The first set of experiments was designed to measure the rate of incorporation of leucine into protein in slices from both genotypes and under different conditions. The second set of experiments was designed to assess levels of phosphorylated and unphosphorylated forms of signaling molecules involved in the regulation of translation. Relative levels of these proteins were assessed in both genotypes and under different conditions. The third set of experiments was designed to assess the strength of the LTD response following mGluR5 activation.</p><!><p>Five slices, containing ventral and dorsal hippocampus were transferred by means of a fine paintbrush to 100 μm Falcon™ cell strainers (Corning Inc., Corning, NY, USA) in room temperature Sucrose aCSF, bubbled with 95% O2 / 5% CO2, for 10 min. Slices were then transferred to Standard aCSF, bubbled with 95% O2 / 5% CO2, for 4 h of recovery at 37°C. Strainers containing brain slices were then transferred to new baths of 125 mL Standard aCSF for drug/condition treatment. Slices treated with DHPG were incubated in Standard aCSF with 100 μM DHPG for 5 min before transfer to Standard aCSF containing L-[4,5-3H]leucine (SA, 60–120 Ci/mmol, MORAVEK, Inc., Brea, CA) (Cat. No. MT 672), at the concentration of 4 μCi/mL, for 30 min. Following incubation in [3H]leucine, slices were transferred to ice cold Dissection aCSF, and whole hippocampi were dissected and transferred to Precellys™ soft tissue homogenizing CK14 tubes (Bertin Corporation, Rockville, MD).</p><p>Dissected hippocampi were homogenized in 10 mM HEPES, 2 mM EDTA, 2 mM EGTA, 17 mM Triton X-100, 0.01% by volume Halt™ Protease Inhibitor Cocktail (100X), by means of a Precellys Evolution Homogeniser (Bertin Corporation) before addition of TCA to a final concentration of 10%. Precipitated protein was centrifuged and washed 5 times with 10% TCA to remove contaminating free [3H]leucine. Protein pellets were rinsed 3 times with acetone ( −40°C) and air dried for 1 hr. Pellets were resuspended in 0.1 M NaOH and shaken for 45 min at 37°C. Concentrations of [3H] and protein in each sample were determined by liquid scintillation counting and Pierce™ BCA protein assay (Thermo-Scientific, Waltham, MA, USA), respectively. The specific activity of [3H]leucine in the incubation medium was determined in each experiment. In AAS experiments, we used the specific activity of the [3H]leucine from the supplier as the specific activity of leucine in the medium ( 1.55–2.27 x 105 dpm/pmol ). The rate of PS in each sample was calculated as follows: PS=(DPMmg protein)solubilizedpellet(DPMpmolleucine)aCSFx30min in which PS is the rate of incorporation of leucine into protein in pmol/mg protein /min and DPM is disintegrations per minute of 3H.</p><!><p>Samples for Western blot analysis were sliced and recovered as described above. Strainers containing brain slices were then transferred to new baths of 125 mL Standard aCSF for drug/condition treatment. Slices treated with DHPG were incubated in Standard aCSF with 100 μM DHPG for 5 min. Following incubation, slices were transferred to ice cold Dissection aCSF, and whole hippocampi were dissected and immediately frozen on dry ice in pre-weighed Precellys tubes and stored in −80°C until processing. On the day of processing, hippocampal slices were thawed and homogenized two times for 30 s at 5,000 rpm in 10% (weight/volume) solution of T-Per protein extraction reagent with 1% EDTA and 1% Halt Protease and Phosphatase Inhibitor Cocktail by means of Precellys soft tissue homogenizing tubes with ceramic beads (Bertin Corporation). Homogenates were centrifuged at 15,000 x g for 15 min at 4°C, supernatant fractions collected, and protein concentrations measured by means of a BCA protein assay.</p><p>Protein in the supernatant (10 μg) was denatured at 95°C for 5 min with an equal volume of 2x Laemmli buffer (5% 2-mercaptoethanol) and electrophoresed on a 4–15% Mini-PROTEAN TGX Stain-Free gel (Bio-Rad Laboratories, Hercules, CA, USA) and transferred to a nitrocellulose membrane (Bio-Rad Laboratories). The membrane was incubated overnight at 4°C in the primary antibody solution followed by 1 h at room temperature in secondary antibody (goat anti-rabbit horseradish peroxidase-linked 1:10,000 (Bio-Rad Laboratories)). Antibody staining was visualized after incubation in Clarity substrate (Bio-Rad Laboratories) on a ChemiDoc MP Imager (Bio-Rad Laboratories) (RRID: SCR 014210). For normalization of Western blots, we employed the Stain-Free technology (Bio-Rad Laboratories). Primary antibodies (Cell Signaling, Danvers, MA) were: p-ERK1/2 (4370) (RRID:AB_2315112), ERK (4695) (RRID:AB_390779), p-eIF2α (3398) (RRID:AB_2096481), eIF2α (5324) (RRID:AB_10692650), p-mTOR (5536) (RRID: AB_10691552), mTOR (2983) (RRID: AB_2105622), p-p70 S6K (9234) (RRID: AB_2269803), p70 S6K (2708) (RRID: AB_390722), p-S6 235/236 (2211) (RRID: AB_331679), p-S6 240/244 (2215) (RRID: AB_331682), S6 (2217) (RRID: AB_331355), p-Akt (4060) (RRID: AB_2315049), Akt (9272) (RRID:AB_329827), p-GCN2 (OABF01173; Aviva Systems Biology, San Diego, CA, US) (RRID:AB_2801396) and GCN2 (ab137543; Abcam, Cambridge, UK) (RRID:AB_2801397). Antibodies were diluted 1:1000.</p><p>At the outset nine animals per group were examined, however some results were excluded because of artifacts on the blot or because values were greater than two standard deviations from the mean and considered outliers. Details of each removed value are included in the figure legends.</p><!><p>Brain slices (400 μm in thickness) prepared in a similar manner as described above were recovered in Standard aCSF at 32°C for 30 min before being moved to room temperature Standard aCSF for 1.5 h. Following recovery slices were transferred to an interface recording chamber where they were perfused with bubbled (95% O2 / 5% CO2) Standard aCSF at 32°C at a rate of 2 mL/min. Field-recordings were performed by placing a stimulating electrode in the CA3 region of the hippocampus to stimulate the Schaffer collateral pathway and a recording electrode in the CA1 region. Recording electrodes (1 – 2 MΩ) generated the day of recordings by a micropipette puller (Sutter Instrument Co, USA) were filled with Standard aCSF. Field excitatory post-synaptic potentials (fEPSP) were elicited every 15 s through a Master-8 stimulator (A.M.P.I, Israel), recorded with a Multiclamp 700B amplifier, and digitized using Digidata 1440A (Molecular Devices, San Jose, CA, USA). After baseline recording, 100 uM DHPG was perfused on slices for 15 min, then slices were returned to Standard aCSF, and recording continued for 90 min. fEPSP slopes were analyzed using the Clampfit 9 software (Molecular Devices), and every four fEPSPs were averaged to generate an average slope per minute. A grand average of 30 min of baseline fEPSPs was generated and every fEPSP recorded after DHPG treatment was normalized to this mean. Only slices with fEPSP covariances less than 10% during baseline recording were used for final analysis. The number of successful recordings that met baseline criteria are as follows; 17 LTD recordings generated from 12 WT mice under AAS, 16 LTD recordings generated from 14 Fmr1 KO mice under AAS, 11 LTD recordings generated from 11 WT mice under AAR, 10 LTD recordings generated from 5 Fmr1 KO mice under AAR.</p><!><p>Data are expressed as means ± SEMs. Our sample sizes were based on numbers used in published studies. We did not test for the normality of data. We did test for outliers; if points were greater than two standard deviations from the mean of all samples they were excluded. Specific explanations of exclusions are presented in the figure legends. For the statistical analysis of PS results, we analyzed the data from AAR and AAS conditions separately because the values differed by two orders of magnitude. Results of PS studies under AAR and AAS conditions were analyzed by means of 2-way ANOVA with genotype and treatment as between subjects factors. Western blot results were analyzed by means of 3-way ANOVA with amino acid condition, genotype and treatment as between subject factors. For the statistical analysis of electrophysiology results, responses were averaged over 10 min epochs for 90 mins following DHPG treatment. These results were analyzed by means of a 3-way repeated measures ANOVA with epoch as a within subjects variable and genotype and amino acid condition as between subjects variables. When appropriate, we further probed for differences by means of Bonferroni t-tests. These analyses were made by means of SPSS version 21 (IBM, Armonk, NY, USA).</p><!><p>We measured PS rates in slices from both genotypes incubated in AAR or in AAS conditions (see Table 1). The lack of amino acids in the incubation medium drastically reduced the rate of PS by 99% regardless of genotype (Fig. 2A). Under AAR conditions, neither the genotype x condition interaction nor the main effect of genotype was statistically significant (Table 2), but the main effect of DHPG treatment was (p = 0.003). Following DHPG treatment, rates of PS increased 22 and 13% in WT and Fmr1 KO slices, respectively. Similarly, under AAS conditions, neither the genotype x condition interaction nor the main effect of genotype was statistically significant (Table 2) but the main effect of DHPG treatment was (p < 0.001). Treatment with 100 μM DHPG resulted in decreased (46 and 57%) PS rates in slices from both WT and Fmr1 KO, respectively (Fig. 2B). In separate experiments, we confirmed that 95–97% of the incorporation of radiolabeled leucine into the acid precipitable fraction was blocked by treatment with anisomycin in both genotypes under both AAR and AAS conditions (Table S1) indicating that our assay procedure measured de novo PS.</p><p>Taken together our results indicate that in the absence of amino acids in the medium, rates of PS are reduced to 1% of rates in the presence of amino acids. In neither AAR nor AAS conditions did we detect a genotype difference in basal rates, but we did find effects of DHPG treatment. The effects of DHPG were present in both genotypes but differed in AAR and AAS conditions.</p><!><p>Signaling through the PI3K-Akt-mTOR pathway is thought to be required for mGluR-dependent LTD in the hippocampal slice (Hou & Klann 2004). Moreover, mTOR activity is reported to be elevated in lysates of hippocampus from young Fmr1 KO mice (Sharma et al. 2010). We examined the abundance of mTOR and the phosphorylation of mTOR at Ser-2448, an indicator of mTOR activity (Hay & Sonenberg 2004) in slices from both genotypes and incubated in either AAR or AAS conditions. For total mTOR and phosphorylation of mTOR at Ser-2448 only the main effects of amino acid condition were statistically significant (Table S1 and Fig. S1). Levels of mTOR (Fig. S1A) and p-mTOR (Fig. S1B) were lower in AAS versus AAR conditions (p < 0.001, p = 0.007, respectively). These results indicate that in the hippocampal slice preparation 3–4 h of AAS conditions lowers mTOR capacity but does so regardless of genotype or treatment. We did not detect genotype differences or effects of DHPG stimulation under either amino acid condition. There were no statistically significant interactions or main effects for the ratio of p-mTOR to total (Fig. S1C).</p><p>As another marker of mTOR activity, we examined the kinase p70S6k downstream of mTOR for effects of DHPG stimulation. The kinase, p70S6k, is involved in the regulation of ribosomal maturation through its role in catalyzing the phosphorylation of ribosomal protein S6. Phosphorylation of p70S6k at Thr-389 is the site of mTOR-dependent regulation (Burnett et al. 1998). We found no statistically significant effects on abundance of p70S6k (Fig. 3A), but for both p-p70S6k (Thr-389; Fig. 3B) and the ratio of the phosphorylated form to total (Fig. 3C) the amino acid condition x genotype x treatment interactions were statistically significant (Table 2). These results indicate that a genotype-specific response to DHPG is affected by amino acid conditions. In the case of p-p70S6k, treatment with DHPG had no effect on slices from WT mice in either amino acid condition, but in slices from Fmr1 KO mice, DHPG treatment decreased p-p70S6k in AAR conditions (25% of WT, p = 0.042) and increased p-p70S6k in AAS conditions (29% of WT, p = 0.030). A decrease in the ratio of p-p70S6k to total in slices from Fmr1 KO mice in response to DHPG-treatment was also seen (40% of WT, p = 0.013).</p><p>The role of Akt in regulating mTOR and cell growth is complex. Akt is phosphorylated at Thr-308 by 3-phosphoinositide-dependent protein kinase-1 (PDK1) downstream of PI3K activation (Alessi et al. 1997) and at Ser-473 by mTORC2 (Sarbassov et al. 2005). Activated Akt stimulates mTORC1 by its inhibition of tuberous sclerosis complex 2 (TSC2) which relieves TSC suppression of mTORC1 activity. In our hippocampal slice preparation, we found a statistically significant main effect of genotype (p < 0.001) for total Akt (Fig. S2A), indicating that, regardless of amino acid condition or treatment, slices from Fmr1 KO mice had higher levels of Akt than WT. For p-Akt (Thr-473), levels were higher in AAS conditions regardless of genotype or treatment (Fig. S2B; main effect of amino acid condition, p < 0.001), and levels were higher in slices from Fmr1 KO mice regardless of amino acid condition or treatment (main effect of genotype, p = 0.018). For the ratio of p-Akt to total (Fig. S2C), main effects of both amino acid condition (p < 0.001) and treatment were statistically significant (p = 0.043) indicating higher Akt activation in AAS conditions compared to AAR and decreased activity with DHPG treatment. These results suggest that Akt is activated in AAS conditions likely through mTORC2 signaling and that activation is suppressed with activation of group 1 mGluR receptors likely through feedback by mTORC1.</p><p>Another pathway involved in the regulation of cell growth and protein synthesis in response to extracellular cues is the Ras-ERK pathway. There is cross-talk and compensation between the Ras-ERK and PI3K-mTOR pathways (Mendoza et al. 2011), so it is of interest to consider responses together. Moreover, the ERK pathway has been reported to be implicated in FXS (Osterweil et al. 2010; Osterweil et al. 2013; Sawicka et al. 2016). In our experiment, we found a statistically significant main effect of amino acid condition such that total ERK was lower in AAS compared with AAR regardless of genotype or treatment (Table S1, Fig. S3A). With p-ERK the amino acid condition x treatment interaction was statistically significant indicating that regardless of genotype, p-ERK increased significantly with DHPG treatment in AAS conditions (p < 0.001, Table S1, Fig. S3B); effects in AAR conditions were not statistically significant. Regardless of genotype p-ERK was statistically significantly increased in AAS compared with AAR conditions with (p < 0.001) and without (p < 0.001) DHPG treatment. The treatment x amino acid condition interaction was statistically significant for the ratio of p-ERK to total (p = 0.001) (Fig. S3C). In this case, increases with DHPG treatment regardless of genotype were statistically significant under both AAS (p < 0.001) and AAR (p = 0.040) conditions, but effects were greater in the AAS condition. Ratios of p-ERK to total were higher in AAS conditions compared to AAR regardless of genotype with (p < 0.001) and without (p = 0.001) DHPG. Differences were greater with DHPG treatment.</p><!><p>General protein synthesis is suppressed in response to diverse forms of stress through the phosphorylation of eIF2α (Pain 1996). Phosphorylation of eIF2α by one of several kinases generates the integrated stress response which in addition to decreasing general protein synthesis also increases translation of a subset of transcripts containing upstream open reading frames (uORF). One of the stressors known to induce this response is a limitation in availability of amino acids through general control nonderepressible 2 (GCN2) which senses uncharged tRNAs. Other kinases acting on eIF2α are double-stranded RNA-activated protein kinase (PKR), hemin-regulated inhibitor kinase (HRI), and misfolded proteins in the endoplasmic reticulum (PERK). Work from the Sonenberg lab reports that phosphorylation of eIF2α is involved in the long-term phase of LTP (Costa-Mattioli et al. 2009). The hippocampal response to group 1 mGluR activation and the late phase of LTD is also thought to involve phosphorylation of eIF2α (Di Prisco et al. 2014). We measured eIF2α activity in lysates of hippocampal slices from WT and Fmr1 KO mice under the AAR and AAS conditions with and without treatment with DHPG. For total eIF2α (Fig. 4A) and p-eIF2α (Fig.4B), only the main effects of amino acid condition (p < 0.001) and DHPG treatment (p = 0.014), respectively, were statistically significant. For the p-eIF2α/total response, the amino acid condition x genotype x treatment interaction was statistically significant, so we probed for individual differences (Table 2, Fig. 4C). With AAR, the ratio of p-eIF2α to total under basal conditions was lower in slices from Fmr1 KO mice compared to WT (30% of WT, p = 0.051). Moreover, in AAR conditions the response to DHPG differed between WT and Fmr1 KO mice. Whereas slices from WT mice were unaffected by DHPG, p-eIF2α/total increased (by 50% of basal, p = 0.008) with DHPG treatment in slices from Fmr1 KO mice. In AAS, we found no effects of DHPG on p-eIF2α/total in either genotype.</p><p>Considering the effect on p-eIF2α we examined one of the kinases known to phosphorylate eIF2α, GCN2 (Table S1, Fig S4). The only statistically significant effect on GCN2 was a main effect of amino acid condition on the ratio of p-GCN2 to total indicating that in AAS conditions the ratio decreases regardless of genotype or treatment (Fig. S2C). This result is in accord with the lack of activation of eIF2α following 4–5 h of AAS conditions (Fig. 4C).</p><!><p>Previous research has shown that DHPG-induced, group 1 mGluR-dependent LTD is enhanced in Fmr1 KO hippocampal slices (Huber et al. 2002). mGluR-dependent LTD is a complex process that is dependent on PS (Huber et al. 2000). Therefore, it is likely that absence of amino acids in the medium may influence this type of LTD. We examined this question in hippocampal slice preparations from WT and Fmr1 KO mice that were maintained in AAS or AAR condition. In both WT and Fmr1 KO slices, fEPSPs decreased after DHPG treatment regardless of AAR or AAS, indicating that LTD is induced. The genotype x amino acid condition x epoch interaction was not statistically significant (F(4.143,207.129) = 1.441, p = 0.22), but the genotype x amino acid condition interaction was (F(1,50) = 5.523, p = 0.023; Fig. 5). Further testing indicates that, regardless of epoch, there was no difference in DHPG-induced LTD in WT slices between AAR and AAS (mean fEPSPs between 50–60 min post DHPG treatment: WT-AAR = 71% and WT-AAS = 72% of baseline; p = 0.950; Fig. 5). In contrast, in Fmr1 KO slices, fEPSPs after DHPG treatment were statistically significantly lower in AAS than in AAR (mean fEPSPs between 50–60 min post DHPG treatment: Fmr1 KO-AAR = 72% and Fmr1 KO-AAS = 59% of baseline; p = 0.002; Fig. 5), indicating that LTD is enhanced in AAS condition. In AAS, the fEPSP decrease after DHPG treatment was greater in Fmr1 KO slices than in WT slices (mean fEPSPs between 50–60 min post DHPG treatment: WT-AAS = 72 % and Fmr1 KO-AAS = 59% of baseline; p = 0.004; Fig. 5), while in AAR, the fEPSP decrease was comparable in WT and Fmr1 KO slices (mean fEPSPs between 50–60 min post DHPG treatment: WT-AAR = 71% and Fmr1 KO-AAR = 71% of baseline; p = 0.568; Fig. 5). These results confirm the previous finding that in AAS, LTD is enhanced in slices from Fmr1 KO compared to WT slices (Huber et al. 2002). By contrast, under more physiological conditions (AAR), there is no difference between Fmr1 KO and WT in DHPG-induced LTD (Fig. 5).</p><!><p>To our knowledge, our study is the first to evaluate hippocampal slices from WT and Fmr1 KO mice under AAR conditions. The presence of amino acids in aCSF more closely mimics physiological conditions and is the relevant condition for understanding in vivo states. Addition of amino acids to the incubation media is especially important in studies of PS-dependent processes in hippocampal slices. Research in yeast has characterized the stress response that results from AAS. This stress response leads to alterations in cellular growth pathways and an inhibition in cap-dependent translation which is also involved in synaptic processes such as mGluR-dependent LTD (Yang et al. 2016; Di Prisco et al. 2014). In the current experiment, exploration of how the presence of amino acids affects PS-dependent processes in Fmr1 KO mice revealed distinct effects on general rates of PS, activation of cellular growth proteins, and mGluR-dependent LTD.</p><p>The presence or absence of amino acids in aCSF has a profound effect on general rates of PS. A lack of amino acids in aCSF reduced overall rates in WT and Fmr1 KO hippocampal slices to about 1% of rates in AAR aCSF. The effect suggests potential alterations in any process associated with or dependent on PS in AAS. Indeed, DHPG activation of group 1 mGluRs had amino acid-dependent effects on PS. In the presence of amino acids, DHPG increased general rates of PS, whereas in the absence of amino acids DHPG decreased general rates of PS. These results show that the integrated stress response to the lack of amino acids likely influences the downstream synthesis of proteins activated by group 1 mGluR activation. The amino acid condition also affected LTD and did so in a differential manner between the two genotypes. Elevated mGluR-dependent LTD in Fmr1 KO hippocampal slices in AAS is a well-established phenotype of Fmr1 KO mice (Huber et al. 2002). In slices from WT animals, mGluR-dependent LTD is dependent on de novo protein synthesis (Huber et al. 2000), whereas in slices from Fmr1 KO mice it is not (Nosyreva & Huber 2006). This result was interpreted to indicate that "LTD proteins" were already in excess at the synapse due to dysregulated PS allowing LTD to persist in the absence of newly synthesized proteins. In the light of our present study, we think that all these results were influenced by the integrated stress response provoked by the AAS conditions of the experiments.</p><p>We considered that the lack of enhanced mGluR-dependent LTD in Fmr1 KO slices in the presence of amino acids was due to differential activity of signaling pathways regulating PS. In support of differential effects on signaling pathways is our observation that PS increased with DHPG treatment in AAR and decreased with DHPG treatment in AAS. One of the most likely candidates is through regulation of the phosphorylation of eIF2α at serine-51. eIF2α is part of the regulatory subcomplex of eIF2B, a guanine nucleotide exchange factor, that regulates the formation of the ternary complex (eIF2 • GTP • Met-tRNA) and therefore initiation of cap-dependent translation. Phosphorylation of eIF2α is a critical step in the integrated stress response. In the hippocampal slice in AAR and under basal conditions, phosphorylation of eIF2α is lower in slices from Fmr1 KOs compared to WT. With DHPG treatment, phosphorylation of eIF2α is elevated over basal in Fmr1 KOs but not in WT. This result implies that regulation of translation initiation by eIF2 is altered in Fmr1 KOs. It is conceivable that tight local control of translation by FMRP is especially critical with activation of mGluRs. In the Fmr1 KO, this system of local control is lost and may be supplanted by an integrated stress response. Phosphorylation of eIF2α in slices from Fmr1 KO mice following DHPG treatment is in accord with this scenario. Moreover, the decrease in p-p70S6k activity suggests that other systems are also working to quell an unbridled PS response.</p><p>Our study provides an overview of the effects of amino acid conditions on phenotypes observed in hippocampal slices from Fmr1 KO mice. We have not, in this initial report, done a comprehensive study of all possible parameters and mechanisms underlying these phenotypes. Our electrophysiological studies did not address possible genotype and condition differences in baseline responses and in paired-pulse facilitation responses. These properties have been investigated previously (Parradee et al., 1999; Huber et al., 2002) and found not to be appreciably affected by a lack of FMRP in the mouse. Whereas these properties have not been investigated in slices incubated in medium containing amino acids, it seems unlikely that the presence or absence of amino acids would affect the protein synthesis-independent properties. The effects of AAR conditions on these responses will be the subject of future investigations. Nevertheless, our results demonstrate that the presence or absence of amino acids in the incubation medium will affect the measured LTD response in Fmr1 KO slices.</p><p>Our report brings to light several key methodological differences that have large effects on results of studies of PS and PS-dependent processes. First and foremost is the presence of amino acids in the incubation media. Another difference is the use of an inhibitor of transcription. We did not include actinomycin-D in our incubation medium, because we did not wish to eliminate control through the synthesis of transcription factors. Actinomycin-D was included in the media of experiments in which an increase in PS in slices from Fmr1 KO mice was reported in AAS (Dölen et al. 2007; Osterweil et al. 2010). Phosphorylation of eIF2α results in inhibition of general PS at the same time instituting translation of a set of mRNAs with ORFs. These include transcription factors including ATF4 (or CREB2), ATF6, CHOP, XBP-1 and others known to induce transcription of genes coding for endoplasmic reticulum chaperones and enzymes to cope with an excess of unfolded proteins or a lack of amino acids (Wek et al. 2006). In previous studies (Osterweil et al. 2010), treatment with DHPG occurred in the presence of amino acid tracer, capturing the early PS response to DHPG. Under these conditions DHPG was seen to increase PS in WT slices, but it did not affect PS in Fmr1 KO slices. The early response to DHPG treatment may have a disproportionate effect on the measured PS response and may account for our measurement of a decreased rate of PS in AAS with DHPG treatment.</p><p>A puzzling result is our observation that neither p-GCN2 nor p-eIF2α increased in AAS conditions compared with AAR. As GCN2 is known to sense AAS and eIF2α is the direct target of GCN2 activity these results are the opposite of the reported roles these proteins play in the integrated stress response to amino acid deficit. It is possible that the lack of activation of GCN2 and eIF2α is due to the amount of time hippocampal slices were maintained in AAS. In studies of mouse embryonic stem cells, it was shown that the phosphorylation of eIF2α lessened over time during leucine deprivation (Zhang et al. 2002). It is possible that a similar mechanism is occurring here. There may have been an immediate increase in the activity of GCN2 and eIF2α in response to the absence of amino acids that dissipated over the 4–5 h incubation period. This could explain why it was observed that protein synthesis did not reach stable levels until 4 hours after incubation in Standard aCSF lacking amino acids (Osterweil et al. 2010). Future experiments will need to examine the time course of activity of GCN2 and eIF2α in hippocampal slices during the recovery period following slice preparation in AAS and AAR conditions.</p><p>Another important concern is that the current experiment did not replicate previously seen increases in global rates of PS in the hippocampus of Fmr1 KO mice. Increased rates of global PS in the hippocampus of Fmr1 KO mice have been shown both in vivo (Qin et al. 2005) and in vitro (Osterweil et al. 2010) in hippocampal slices, whereas in the present study we saw no genotype differences in PS under either amino acid condition. One consideration is the use of anesthesia. In the present study, mice were euthanized by rapid decapitation; no anesthesia was used. Previous in vivo PS measurements were made in mice after surgical implantation of vascular catheters; surgery was carried out under isoflurane anesthesia 24 h prior to measurements (Qin et al. 2005). In previous in vitro studies, hippocampi were dissected after mice were given an overdose of Nembutal (Osterweil et al. 2010). Our study of the effects of propofol anesthesia on PS rates measured in vivo indicated that Fmr1 KO, but not WT, mice respond to propofol anesthesia with altered rates of global PS (Qin et al. 2013). Moreover, work of others has shown lasting effects of quickly cleared anesthetics such as isoflurane and halothane in WT mice on the activation of proteins such as ERK, mTOR, eIF2α, and p70S6K (Qin et al. 2013; Antila et al. 2017; Kang et al. 2017; Palmer et al. 2005). These proteins are involved in the regulation of PS and are thought to be involved in the group 1 mGluR response. The use of anesthetics may confound results of experiments designed to measure effects of lack of FMRP on PS and PS-dependent processes.</p><p>Our studies call for a re-examination of the mGluR theory of FXS. Whereas our present results do not show enhanced DHPG-induced LTD in hippocampal slices from Fmr1 KO mice under normal physiological conditions (AAR), they do show that the response to DHPG stimulation is abnormal in other respects. In Fmr1 KO slices, the response to DHPG includes phosphorylation of eIF2α indicating that an integrated stress response has been mounted. We propose that this integrated stress response may be a cellular attempt to compensate for the lack of regulation of translation by FMRP.</p>

PubMed Author Manuscript

Curvature Constrained Splines for DFTB Repulsive Potential Parametrization

The Curvature Constrained Splines (CCS) methodology has been used for fitting repulsive potentials to be used in SCC-DFTB calculations. The benefit of using CCS is that the actual fitting of the repulsive potential is performed through quadratic programming on a convex objective function. This guarantees a unique (for strictly convex) and optimum two-body repulsive potential in a single shot, thereby making the parametrization process robust, and with minimal human effort. Furthermore, the constraints in CCS give the user control to tune the shape of the repulsive potential based on prior knowledge about the system in question. Herein, we developed the method further with new constraints and the capability to handle sparse data. We used the method to generate accurate repulsive potentials for bulk Si polymorphs and demonstrate that for a given Slater-Koster table, which reproduces the experimental band structure for bulk Si in its ground state, we are unable to find one single two-body repulsive potential that can accurately describe the various bulk polymorphs of silicon in our training set. We further demonstrate that to increase transferability, the repulsive potential needs to be adjusted to account for changes in the chemical environment, here expressed in the form of a coordination number. By training a near-sighted Atomistic Neural Network potential, which includes many-body effects but still essentially within the first-neighbor shell, we can obtain full transferability for SCC-DFTB in terms of describing the energetics of different Si polymorphs.

curvature_constrained_splines_for_dftb_repulsive_potential_parametrization

Introduction<!>DFTB<!>Vrep Using Curvature Constrained Splines<!>Constraints<!>Repulsive and Monotonous Constraints<!><!>Switch Constraint<!>Sparsity Constraint<!>Computational Details<!>Results and Discussion<!>Structures and Training Set<!><!>Transferability in Electronic Parameters<!><!>Two-Body Hyperparameter Rcut<!><!>Two-Body Hyperparameter Rcut<!>Repulsive Potential Fitting Using CCS<!><!>Repulsive Potential Fitting Using CCS<!>Transferability of the Repulsive Potential within a Polymorph<!><!>Exploring Limits of Transferability Using Two-Body Repulsive Potentials<!><!>Exploring Limits of Transferability Using Two-Body Repulsive Potentials<!>Beyond Two-Body Repulsive Potentials: Atomistic Neural Networks<!><!>Conclusion<!>

<p>Within the field of material science, Density Functional Theory (DFT)1,2 has become one of the main working horses owing to its wide range of applicability and its favorable scaling behavior with system size. Despite its success, DFT is not computationally efficient for systems containing a large number of atoms, sampling of complex energy landscapes, and for high-throughput screening purposes.3</p><p>In regard to challenges with computational efficiency, a force-field (FF) based approach would be ideal. However, the lack of an electronic description makes these approaches unable to estimate electronic properties (e.g., band structure, band gap, etc.). In general, the gap between DFT and FF-methods is filled by semiempirical methods, which strikes the right balance between accuracy and computational cost.</p><p>Self-consistent charge density functional tight binding (SCC-DFTB)4 is a semiempirical method, an approximate and parametrized DFT method, about 2 orders of magnitude faster than DFT when using local or semilocal density functionals. Compared to hybrid density functionals, the gain is even larger. The method is applicable to a wide range of problems within chemistry and physics, including redox chemistry of oxides,6 van der Waals interactions,7 electron transport,8 etc. The method can also be systematically extended to overcome limitations in DFT, such as the problem with underestimated band gaps.9,10</p><p>While SCC-DFTB is a very capable tool that can be used to calculate both geometries and electronic properties at a relatively low computational cost, there is a rather substantial effort required in terms of parametrization when applied to new systems. There are primarily two types of parameters in the SCC-DFTB method: (i) those related to the electronic structure and electronic energy of the valence electrons, the so-called compression radii, and (ii) those related to the repulsion between the ionic cores (the core electrons and the nuclei). Generally, the electronic parameters are optimized first to obtain an accurate electronic structure description. Thereafter, empirical short ranged two-body potentials are used to correct for the remaining missing contributions. We will refer to these potentials as repulsive potentials. Ideally, these potentials should be monotonic, smooth, and not vary too rapidly.11 Moreover, transferability is limited, due to the pairwise additive nature of the repulsive potentials, which means that when we extend beyond the range for which a particular parameter set was developed, we must be prepared to reparametrize the potential. Hence, the parametrization of the repulsive potentials is often a laborious and tedious task, calling for the development of novel methods utilizing robust, efficient, and fast mathematical routines to make the generation of reliable parameter sets more efficient.</p><p>In recent years, a number of initiatives to develop automated fitting procedures and protocols have been developed (see, e.g., refs (9 and 11−26)). Some of these focused exclusively on obtaining the parameters for the repulsive potential.12−16,18,21,23,25,26 A key feature among these developments is the use of splines to represent the repulsive potential. Here, the initial steps toward a semiautomated repulsive potential fitting scheme was done by Knaup et al.12 by their use of cubic splines combined with an evolutionary algorithm to find the best repulsive potential. A more rigorous least-squares optimization using fourth-order splines was done by Gaus et al.14 A drawback with this procedure was the difficulty in determining the knot position of the splines. Later, Chou et al.9 presented an automated Multi-Objective Particle Swarm Optimization (MOPSO) approach for optimizing both the knot positions of the splines for the repulsive potential as well as electronic parameters in the form of l-dependent compression radii. However, even though splines satisfy the criteria of smoothness, their flexibility can in some cases lead to oscillations in the repulsive potential. This prompted Bodrog et al.13 to instead define the repulsive potential as a linear combination of higher-order polynomials (excluding zeroth and first-degree terms) or of exponential basis functions . Indeed, the exponential basis functions (monotonous and smooth) are an ideal choice to approximate the repulsive potential, but in reality, the repulsive potential need not always be strictly repulsive. Therefore, the rigid exponential form for the repulsive potential becomes a drawback. Furthermore, the user has to identify the choice of basis functions required for a specific system, which could be problematic in some cases.</p><p>On a similar note, Hellström et al.16 found that the problem of incorrect energetics for various polymorphs of ZnO using the znorg-0.1 parameter set by Moreira et al.27 could be alleviated by a reparametrization of the repulsive potential using a training set of structures with different coordination numbers. For the new repulsive potential, a four-range Buckingham potential was chosen, an analytical function which the authors found to constitute a fair balance between flexibility and smoothness. The advantage of such an approach is that it avoid problems with oscillations that can occur for splines or high order polynomials. Clearly, the rigid parametric functional form of the four-range Buckingham is a drawback. More recently, Panosetti et al.26 introduced a Gaussian Process Regression (GPR) machine learning approach for fitting the two-body repulsive potential. Even though GPR is a nonparametric approach, the optimization of hyperparameters can be a challenge. Another problem is that GPR models require a large number of data points to train the potential against and are known to have poor extrapolation capabilities.</p><p>In this work, we demonstrate how the curvature constrained cubic splines (CCS) methodology developed in ref (28) can be used to build repulsive potentials without the need for adjustable parameters. The aforementioned problems with splines (oscillatory behavior and nonmonotonocity) are alleviated by defining a set of intuitive constraints on the curvature of the repulsive potential. An additional benefit of using the CCS methodology is that the optimization problem, i.e., the actual fitting of the repulsive potential, is a convex problem which can be solved using quadratic programming, which guarantees a unique (if strictly convex) and optimum two-body repulsive potential in a single shot.</p><p>The methodology is tested for various polymorphs of Si. We examine the flexibility of the CCS method in terms of adopting different shapes to best reproduce the energetics for a number of silicon polymorphs (3D and 2D). Since the CCS method is virtually free from meta-parameters, apart from the cutoff radius in the two-body interaction, it reduces the number of parameters in a SCC-DFTB parametrization to merely the electronic ones. As such, the CCS method is ideal to use in conjunction with global optimization techniques like the MOPSO method presented in ref (9).</p><p>The outline of the paper is as follows: In Section 2, we first briefly introduce the SCC-DFTB formalism and the CCS methodology and show in detail how it can be used for fitting the repulsive potential. Then, in Section 3, we describe the computational details concerning our DFT and SCC-DFTB calculations. In Section 4, we discuss the results obtained, and Section 5 concludes the paper.</p><!><p>The SCC-DFTB method is based on a Taylor expansion of the Kohn–Sham energy functional about a reference density, ρ0, taken to be a superposition of pseudoatomic densities.4 The total energy expression in SCC-DFTB is normally truncated at the second-order and thus becomes1E1 is obtained from the occupied eigenstates. More precisely, it is computed by2By the use of a linear combination of atomic orbitals (LCAO) ansatz and a two-center approximation, all required entries for solving eq 2 can be conveniently precalculated and stored in so-called Slater–Koster tables. The second-order term, E2, describes the energy originating from density fluctuations about the reference ρ0. The term E0 primarily describes the ionic core–ionic core repulsion. However, the term essentially includes all remaining energy contributions not captured by the other two terms. The total repulsive energy of a system in SCC-DFTB is a sum of contributions of repulsive potentials Vrep(r) from each atom pair3where i and j run over the atom indices in the system, and rij is the distance between pair of atoms. The Vrep is usually short-ranged and smoothly decaying to zero at a certain cutoff distance (rcut). For a comprehensive description of the SCC-DFTB method and its capabilities, we refer to refs (11) and29.</p><!><p>The repulsive potential in the actual SCC-DFTB is often constructed using cubic splines. Cubic splines are flexible and easy to use, and their coefficients can be optimized by performing least-squares fitting to reference values. Here, we use the Curvature Constrained cubic Splines (CCS) methodology to provide the best possible repulsive potential. The reason is that traditional cubic spline methods can lead to repulsive potentials with spurious oscillations due to overfitting unless a small number of knots (number of spline intervals) are used, but a small number of knots lead to a poor linear approximation of the repulsive potential's Hessian. To improve the description of the Hessian, Gaus et al.14 suggested the fourth-order spline fitting (a quadratic approximation for the Hessian) as an alternative. However, the former approach, though an improvement, still needs manual intervention to decide the number of knots below which overfitting can be avoided. In the CCS methodology, constraints are imposed in such a way that overfitting is prevented irrespective of the number of knots used.28 Hence, the Hessian can be approximated with arbitrary accuracy by increasing the number of knots. Moreover, constraints can be applied to each spline coefficient to have various shape-preserving properties. Consider a pair of atoms, the repulsive potential (Vrep) between these atoms is defined from an interval of interatomic distances r ranging from (0,rcut). We subdivide the interval into N subintervals In = [xn–1,xn], for n = 1, ..., N. On each subinterval, we define a cubic function4To determine the so-called spline coefficients, we impose interpolation conditions for the second derivative of the spline and continuity conditions for the spline function itself as well as its first derivative. We remark that this treatment is different from the standard. A typical approach is to impose interpolation conditions on the spline function itself and the continuity conditions on its first and second derivatives. The reason for this treatment is that here we are interested in stipulating the curvature or the second derivative of the spline function at each subinterval's end points. That is, we impose the 2N conditions5Later, we will use the curvatures as the unknowns in an optimization problem to determine the best potential. (We remark that the above conditions ensure that fn″(xn) = fn+1″(xn) for n = 1, ..., N – 1.) Moreover, we impose the following continuity conditions6at the interior interval end points. This gives us 2N – 2 additional conditions for the 4N spline coefficients. Finally, to close the system, we also require the spline to have zero value and gradient at the xN = rcut, that is,7By using the above relations (6 and 7), we can show that the coefficients a = [a1,a2,...,aN]T, b = [b1,b2,...,bN]T, and d = [d1,d2,...,dN]T are linearly dependent on the imposed curvatures c = [c0,c1,...,cN]T.</p><p>We want to express the pair potential function Vrep(r) as a spline function. So, in eq 3, the pair potential function Vrep(r) is substituted with eq 4 and can be written as follows:8A detailed derivation of vector v can be found in ref (28). In this work, we add an additional one-body term to the repulsive potential, to get the correct energies at the dissociation limits. This is shown below9where ϵ and w are the vectors containing one-body energy terms and number of atoms, respectively. The usage of one-body terms has earlier been found to improve geometries and reaction energies.13 In this work, the one-body terms were used to aid the fitting process and to investigate the lack of transferability of the two-body potential (see Section 4.5). The repulsive potential energy (Erep) has a linear dependence on the unknown coefficients c (see eq 8). As is detailed below, this implies that the coefficient vector c can be solved via the least-squares regression method. To get an accurate repulsive potential, the difference between the reference energies and corresponding DFTB electronic energies are minimized over a set of diverse chemical configurations ranging from k ∈{1,...,K}, where K denotes the number of configurations in the training set. The objective function (J) can be written down as follows10whereThe W in eq 10 is a matrix for heteroatomic systems and a vector for homoatomic systems. On a similar note, ϵ is a vector for heteroatomic systems and a scalar for homoatomic systems. The objective function in eq 10 can be written as11where12The determination of the best vector x can be written as the standard Quadratic Programming (QP) problem13where P = MTM and q = −MTe. The details for constraint matrices G and h are discussed in the next section. By construction, the matrix P in eq 13 is at least positive semidefinite, and hence the QP problem is convex. A convex optimization problem has the advantage that all local minima are global minima. If P is positive definite, then the problem is strictly convex, and thus only has one minimum.</p><!><p>The highlight of the CCS method is that the shape of the optimized potential can be tuned via constraints on the curvature at the knot intervals. Additionally, the user is free to constrain the potential based on prior information about the system. We have developed new constraints exclusively for SCC-DFTB repulsive potential fitting (see Sections 2.3.1 and 2.3.3). For simplicity of notation, we omit constraints on ϵ, so for the discussion below x = c.</p><!><p>The repulsive constraint ensures that the spline approximated repulsive potential has a strictly positive curvature. However, such repulsive potentials can still have oscillations in the second derivative which may lead to poor forces and frequencies. In such cases, it would be ideal to have a tighter set of constraints with monotonically decreasing curvature values (see Figure 1). The corresponding constraint matrices G and h (shown in eq 13) are given by14in which 0(2N+1)×1 is the zero matrix of dimension (2N + 1) × 1 and15where IN+1 is the identity matrix of dimension (N + 1) × (N + 1), and Grepulsive has dimension N × (N + 1).</p><!><p>A schematic illustration of all the constraints used in CCS. The x-axis is divided into intervals, and on each interval we define a cubic spline uniquely determined by the c coefficients. The top panels (a) and (b) show the repulsive and monotonous constraints on c coefficients. Panel (c) depicts the switch constraint with red and black dots, respectively, indicating negative and positive values for c coefficients. The switching point Nswitch is at the kth knot. The bottom panel (d) represents the sparse constraint. The knot point with an open circle indicates a bin with no data points. The c coefficient of the spline here is undetermined and can without loss of generality be set to an average value of neighboring c coefficients. This is equivalent to merging of the bins or intervals.</p><!><p>Due to the approximations in SCC-DFTB, the repulsive potential at times can have some attractive regions. This cannot be captured by the repulsive constraints discussed above. By instead adding a switch constraint, we allow the curvature values to change sign once (see Figure 1c) at a certain knot position called Nswitch. This allows the repulsive potential to have at most one minimum (for more details see ref (28)). The corresponding constraint matrices are given by16where N1 = Nswitch, N2 = N + 1 – Nswitch, and 0N1×N2 is the zero matrix of dimension N1 × N2.</p><!><p>The success of the CCS method lies in its flexibility, which can adopt the shape of the repulsive potential to arbitrary precision under the given constraints by gradually increasing the number of knots in the spline table. However, such a procedure can lead to a situation in which our problem becomes underdetermined, a problem of sparsity that needs to be handled. Thus, if we choose a fine mesh of knots, we can resolve the curvature in each part of the interval very well. However, in some regions where data is sparse, for example, at distances in between the first and second coordination shells, we have many more knots than there is information available. In other words, the curvature at certain knots cannot be uniquely determined. This problem can be handled by removing redundant knots by subinterval merging. The procedure is illustrated in Figure 1d and ensures that the curvature changes linearly over coherent undetermined subintervals. With this technique, we avoid any ambiguity in the optimization, and in the limit of an infinitely fine mesh, the method would correspond to one having freely adjustable knot positions.</p><!><p>Our primary reference method in the validation and testing of our new spline method is density functional theory in the implementation with plane waves and pseudopotentials. More specifically, the electronic wave functions were expanded in a plane-wave basis set with a kinetic energy cutoff of 600 eV. The core–valence interactions were modeled with pseudopotentials generated within the Projector Augmented Wave (PAW) scheme proposed by Blöchl.30 In the calculations, we explicitly treated four electrons for each Si atom. Furthermore, we used the PBE functional31 as a reference to generate a training set for repulsive potential fitting (i.e., energies). However, as semilocal DFT functionals generally give poor band gap estimates, we used a modified HSE06 functional32,33 (denoted as HSE06') for this purpose. Instead of the normal 25% nonlocal Fock exchange, we used 10%, which previously has been shown to yield electronic band gaps in better accord to experiments.34 All DFT calculations were performed with the Vienna Ab-initio Simulation Package (VASP).35−38 All SCC-DFTB calculations were done using the DFTB+ software.29,39 The repulsive potential fitting was performed using a modified version of the CCS package.28</p><!><p>In this section, we will demonstrate some key features of CCS when used in conjunction with SCC-DFTB. We start by introducing our training set, which consists of various Si polymorphs. Before fitting the repulsive potentials, we start discussing transferability concerning electronic properties. Next, we demonstrate the flexibility of CCS in terms of adapting to different shapes when fitted to data for Si in different chemical environments, here expressed in terms of varying coordination numbers. We further utilize key features of the CCS method which allow us to address the question of the apparent lack of transferability of SCC-DFTB and propose possible solutions to this issue.</p><!><p>As the training set, we have chosen crystalline phases of Si that lack internal parameters to ensure that we probe effects due to an isotropic chemical environment. There is, however, no technical hindrance to also add structures with internal parameters. The following polymorphs of Si were considered (with corresponding coordination numbers): graphene (3 coordinated), diamond (4 coordinated), simple cubic (6 coordinated), and body-centered cubic (8 coordinated). A schematic illustration of the polymorphs used is shown in Figure 2. The training set comprises Energy-Volume (E-V) scans for all the polymorphs. The volumes in the training set correspond to nearest-neighbor distances between Si from 2.1 to 3.3 Å in steps of 0.1 Å. The nearest-neighbor distance distributions for the different polymorphs are shown in Figure 4.</p><!><p>Diversity in the local chemical environment for Si polymorphs expressed in terms of a coordination number: (a) graphene, (b) diamond, (c) simple cubic, and (d) body-centered cubic.</p><!><p>Before fitting repulsive potentials, the quality of the electronic parameters of existing Slater-Koster tables available in the SCC-DFTB community is validated by comparing computed electronic properties toward hybrid-DFT data. In the literature, the following Slater-Koster tables for Si are available: pbc-0.3,40 matsci,41 and siband.10,42 The pbc-0.3 and matsci sets are known to give a poor description of the band structure for Si polymorphs.9 In contrast, the electronic parameters of the siband set were optimized by Markov et al.10 toward experimental Si and SiO2 band structure data.</p><p>The electronic structures in terms of bandwidths and band gaps for the different Si polymorphs in our training set calculated using hybrid DFT and SCC-DFTB with pbc-0.3 and siband are shown in Figure 3. The energies in the plots are aligned through the lowest lying occupied Si state, i.e., the bottom of the valence band. We note that the bandwidths of the valence bands and conduction bands are in good agreement between our modified hybrid DFT calculations and the SCC-DFTB using the siband set for all polymorphs. From these data, it is further clear that the siband set is superior to the pbc-0.3 set and that the transferability when it comes to SCC-DFTB electronic properties, here in terms of bandwidth and band gap, is rather good.</p><!><p>A schematic description of the electronic structure for Si polymorphs including graphene, diamond, simple cubic, and body-centered cubic. The valence band (VB) and the conduction band (CB) are colored gray and red, respectively. The gap between VB and CB for the nonmetallic polymorphs (diamond and graphene) indicates the band gap. All energies are in eV.</p><!><p>Before generating the repulsive potentials using the CCS scheme, we need to determine the two-body hyperparameter Rcut and how the changes in this parameter affect the accuracy of the resulting potential. The repulsive potential in SCC-DFTB is usually short-ranged, and in general, we use a small cutoff value for Rcut. Typically, the range of the first nearest-neighbor distances in the training set (refer to Figure 4 top panel) is used. However, using CCS the ideal cutoff could be determined from a simple grid search. For this purpose, we made the following training sets: i) a set containing E-V scans for all the polymorphs and ii) sets containing E-V scans for individual polymorphs. The Rcut values were varied from 2.38 to 6.42 Å, and optimization was performed using the switch constraint (see Section 2.3.2).</p><!><p>Top panel shows the range of first and second nearest-neighbor distances for graphene (orange), diamond (blue), SC (red), and BCC (green) in the training set. The middle panel shows the variation of RMSE as a function of Rcut for both diamond (blue) and all polymorphs (black). The bottom panel shows the variation of RMSE as a function of Rcut for individual polymorphs. The dashed vertical line at 3.3 Å indicates the largest nearest-neighbor distance in the training set.</p><!><p>In Figure 4 (middle and bottom panels), we show the variation of the training set error as a function of the Rcut values. We infer that the training set error converges immediately after the first nearest-neighbor distances, except for the SC polymorph. The convergence occurs at 4.3 Å for the SC polymorph. Overall, this suggests that the assumption of a short-ranged behavior for the repulsive potential seems valid in this case. The root-mean-squared-error (RMSE) of individual polymorphs is less than 10–2 eV/atom at Rcut greater than 3.4 Å. Hence, we have chosen a Rcut value of 3.4 Å for the Si–Si repulsive potential. The training set error for individual polymorphs converges toward zero, whereas a nonzero convergence is seen for the training set including all polymorphs. A nonzero value for the training set error convergence indicates the limit of accuracy for the two-body approximation. We remark that the choice of electronic parameters might influence this value. In principle, for a given training set, one could search for a set of electronic parameters that minimize the converged error in the two-body approximation. This could be done by combining CCS with a global search algorithm, e.g., MOPSO.</p><!><p>Having established a scheme for obtaining the optimal Rcut value, we move on to discuss the generation of the repulsive potentials for the silicon polymorphs in our training set (see Section 4.1). The electronic energies from SCC-DFTB are obtained using the siband Slater-Koster tables of Markov et al.10 The CCS method was used to optimize the repulsive parameters. Two different types of fitting procedures (optimizations) were done, one using the original constraints on the curvature (strictly positive) and one in which the curvature is allowed to change sign once (switch constraint, see Section 2.3.2). The resulting repulsive potentials are shown in Figure 5.</p><!><p>Left panel in a) compares DFT energies (black dotted lines) with SCC-DFTB for various polymorphs of Si, with a repulsive constraint. The corresponding repulsive potentials are shown to the right. Panels in b) show a corresponding comparison for the switch constraint. Panels c) and d) show the best approximate potential for all Si polymorphs with a repulsive and switch constraint, respectively.</p><!><p>Our first observation is that there is no good repulsive potential that can simultaneously reproduce the energetics for all the polymorphs in the training set with an acceptable accuracy. This is not completely unexpected; in fact, similar transferability issues have been reported previously in the literature. For example, see the incorrect 2D-3D transition in boron clusters for the borg-0-1 set43 and coordination dependence of repulsive potential for different polymorphs of ZnO.16 A similar trend in repulsive potentials for silicon polymorphs was also observed by Chou et al.9 They showed that the accuracy can be improved by increasing the cutoff of the repulsive potential from 3.5 to 6.3 Å (up to the fourth nearest neighbor for diamond). However, within the limits of the constraints used here, we see no significant improvement in extending the cutoff radii beyond 3.4 Å. It should be pointed out that Chou et al.9 also optimized the electronic parameters along with the repulsive potential.</p><p>We further note that there exists a repulsive potential for each individual polymorph that leads to an almost perfect agreement between the DFT and DFTB energies, see Figure 5. The variations in the shape of the repulsive potentials, across different polymorphs and with different constraints, can be appreciated by looking at Figure 5. Clearly, the repulsive curvature constraint leads to smoother repulsive potentials, but this comes at the expense of a slightly worse fit to the target energies in the training set. This is illustrated by the incorrect shape of the E-V curve for high-coordinated phases like SC and BCC. The aforementioned problem can be rectified by the use of a softer single minimum constraint (at most one minimum). The occurrence of attractive regions in the repulsive potential for condensed Si phases was also reported by Chou et al.9</p><p>The silicon example demonstrated here clearly indicates the problems in transferability of SCC-DFTB parametrizations. This issue will be discussed in more detail in the following.</p><!><p>The results from the above section show that it is possible to get a smooth repulsive potential that can describe the isotropic E-V curves for individual polymorphs. Here, we look at the transferability of the obtained repulsive potential for nonisotropic deformations. For this purpose, we consider E-V curves for the diamond (4C) structure along one axis keeping the other two constant. The results obtained are shown in Figure 6. The results indicate that the repulsive potential fitted on isotropic deformations is transferable for nonisotropic deformations.</p><!><p>A comparison of energy-volume curve DFT (black) and SCC-DFTB (blue) for nonisotropic deformation of diamond.</p><!><p>The results from Section 4.4 suggest that there does not exist a single repulsive energy expression (One-fits-all), consisting of one-body and two-body energy terms under the given constraints, that leads to a satisfactory fit across the various polymorphs of Si with different coordination numbers. At the same time, for each polymorph, we can readily obtain a repulsive potential which fits the data with minimal errors. Here, we will analyze the one-body and two-body contributions in detail, before discussing possible extensions of the method that could allow for a single energy expression to fit the whole data set.</p><p>As an example, let us consider the diamond (4C) and simple cubic (6C) structures of Si. We may express the two fitting approaches adopted so far in the following way. First, for the "One-fits-all" procedure, we may write17where the subscripts indicate that there is a single one-body (ϵ) and single two-body (V) for both the 4C and 6C structures. Second, for the individual fits, we may write18where all energy terms are strictly zero when the coordination number does not match that of the subscript. From the previous sections, we know that the quality of the latter is superior to the former—but what about the other combinations of these ϵ's and V's? Using the same notation as before these would correspond to19a19b</p><p>Next, we consider a training set solely comprising of diamond (4C) and simple cubic (6C) E-V scans. We again use the siband Slater-Koster tables of Markov et al.10 for the electronic SCC-DFTB energies. Figure 7 shows a boxplot comparison using all four expressions above for repulsive fitting, with their combinations of ϵ's and V's for the 4C and 6C Si polymorphs. Additionally, we performed a similar analysis with other combinations of polymorphs corresponding to graphene + diamond and simple cubic + body centered cubic (see Figure 7).</p><!><p>Absolute error per atom (y axis) for structures in the training set for different repulsive models presented in Section 4.5. The black dot indicates the mean absolute error. The upper and lower whiskers indicate the maximum and minimum errors for the training set.</p><!><p>Although the magnitude of the one-body term is large, there is little improvement in the fits when multiple ϵ's are used compared to the corresponding fits with a single ϵ. Instead, the results indicate that we need to go beyond the simple two-body repulsive potential to reach transferability in SCC-DFTB. One such solution will be presented in the following section.</p><!><p>It is clear that to describe a transferable repulsive potential with a single energy expression, we need to go beyond the one-body and two-body contributions. Ideally, we need a model that captures the local chemical environment with reasonable accuracy. Recently, machine learning models like Atomistic Neural Networks (ANN) have gained popularity for accurately describing the short-ranged interactions. However, a pure ANN approach fails to account for long-range interactions, even with large cutoff radii. In the case of SCC-DFTB, we have a good description of the long-range interactions but clearly lack transferability in the short-range description. The combination of the two is therefore appealing, and indeed such combined approaches have been presented in the literature using a Deep Tensor Neural Network together with SCC-DFTB.23</p><p>The Behler–Parinello Neural Network (BPNN)44 is a popular ANN architecture that has been proven to work well for molecular and solid systems. Here, we used a BPNN potential to approximate the repulsive potential. The general idea is to represent the local chemical environment of an atom using a set of radial and angular symmetry functions. Our network architecture comprises of four radial and four angular symmetry functions, two hidden layers with two nodes per layer, with a cutoff radius of merely 4 Å (typical cutoff radii are 6–10 Å45), and a hyperbolic tangent activation function. This is a much smaller and more nearsighted neural network representation as compared to pure ANN approaches. Figure 8 shows a schematic comparing a pure ANN approach and our DFTB+ANN approach. For the generation of the ANN potentials, we used the PROPhet package,46 which is an open source implementation of the BPNN method. The network was trained on the same data set (E-V curves of polymorphs) as used in section 4.4 and was optimized using a resilient backpropagation algorithm. The E-V scans for DFT and the corresponding SCC-DFTB+ANN methods are shown in Figure 9. Indeed, a nearsighted ANN repulsive potential can be used to get a "One-fits-all" repulsive potential. However, one should bear in mind that the increased transferability comes with a cost. The extrapolating power of using "chemically intuitive" functions is lost when using the ANN repulsive potential, which implies that the transferability of the repulsive potential to systems beyond the training set must be carefully investigated.</p><!><p>Schematic illustration of a conventional ANN (gray) and SCC-DFTB+ANN (blue) approach. The value of r indicates the radius of the cutoff sphere, and N is the expected number of neighboring atoms within the cutoff. The cutoff value can be kept small for the SCC-DFTB+ANN approach because of the inbuilt long-range interactions of the SCC-DFTB method.</p><p>Comparison between DFT (black dotted lines) and SCC-DFTB+ANN energies for Si polymorphs using a neural network as the repulsive potential.</p><!><p>The goal in this work was to develop a fast and robust machinery to obtain SCC-DFTB repulsive potentials without having to resort to nonlinear fitting procedures. For this purpose, we used a scheme called CCS to generate the repulsive potentials. The CCS scheme was augmented with new constraints (repulsive and monotonous) and was successfully applied to create accurate repulsive potentials for Si polymorphs. The key features of the augmented CCS method include the following: i) Its ability to adopt various shapes without producing spurious oscillations, ii) The method is compatible with sparse data sets, and iii) The lack of hyperparameters reduces the global search/optimization space allowing us to fully concentrate on the electronic parameters. For individual polymorphs of Si, accurate parametrization of the repulsive potential was possible using CCS. However, due to the approximations in SCC-DFTB, global transferability is limited.</p><p>Regarding transferability, we show that a description beyond a two-body additive repulsive potential is required. In this respect, we further demonstrated that a nearsighted nonlinear ANN model can be a viable solution. The generalized repulsive potential approach by Kranz et al.,21 the multicenter tight binding approach of Goldman et al.,25 and the Deep Tensor Neural Network (DTNN) based many body repulsive potential of Stöhr et al.23 are steps along this path.</p><!><p>The authors declare no competing financial interest.</p>

PubMed Open Access

Visible light switching of metallosupramolecular assemblies

A photoswitchable ligand and palladium(II) ions form a dynamic mixture of self-assembled metallosupramolecular structures. The photoswitching ligand is an ortho-fluoroazobenzene with appended pyridyl groups. The E-isomer is combined with palladium(II) salts affords a double-walled triangle with composition [Pd3L6] 6+ and a distorted tetrahedron [Pd4L8] 8+ (1:2 ratio at 298 K). Irradiation with 410 nm light generates a photostationary state with ~80% of the E-isomer of the ligand which results in the selective disassembly of the tetrahedron, the more thermodynamically stable structure, and the formation of the triangle, the kinetic product. The triangle is then slowly transformed back into the tetrahedron over 2 days at 333 K. The Z-isomer of the ligand does not form any well-defined structures and has a thermal half-life of 25 days at 298K. This approach shows how a thermodynamically preferred self-assembled structure can be reversibly pumped to a kinetic trap by small perturbations of the isomer distribution using non-destructive visible light.

visible_light_switching_of_metallosupramolecular_assemblies

INTRODUCTION<!>RESULTS AND DISCUSSION<!>CONCLUSION

<p>The structure and function of self-assembled species, such as molecular cages, can be controlled using stimuliresponsive components. Different stimuli have been used to perturb metal-template supramolecular assemblies 1 including light, 2 guest molecules, 3 pH changes, 4 competing ligands 5 and changes to solvent. 6 Light, especially the visible spectrum, 7 is appealing due to its easy use, potential for highly specific targeting, and the high resolution of spatial and temporal application. 8 Molecular photoswitches, 9 can be isomerized reversibly by light, with each isomer having different geometries and electronic properties. These differences in properties have been used to control the properties of gels, 10 polymers assemblies, 11 or liquid crystals, 12 and to perform functions including acting as light-activated receptors 13 or pharmaceticals, 14 or pumping systems away from thermodynamic equilibrium. 15 The most studied photoswitches are those based on azobenzene, 16 which can be isomerized between a stable E-isomer and a metastable Z-isomer. However, unsubstituted azobenzene requires potentially destructive UV light to form the meta-stable Z-isomer that has a thermal half-life at room temperature of only 2 days. Significant advances have been made in developing azobenzene-type molecules that operate effectively with visible light, 9c,17 with one of the most successful modifications being the introduction of ortho-fluoro substituents (Figure 1a). 18 These ortho-fluoroazobenzenes allow bidirectional visible-light switching with thermal half-lives that can exceed 2 years and have been incorporated into MOFs 18g and discrete self-assembled structures, 18h and have been used to control molecular folding 18f or the function of enzymes. 18j Conceptually there are two approaches for combining photoswitches with discrete self-assembled structures: encapsulation or direct incorporation as part of the struc-ture. The first strategy involves binding the photoactive unit inside a cavity, such as encapsulating azobenzene type derivatives. 19 Encapsulation a photoswitch can also restrict switching or perturb the balance of isomers. 19b,20 Using photoswitches as structural components of selfassembled structures has proven more difficult. Although there are many examples of large photoswitchable assemblies, 21 such as micelles, vesicles or liquid crystals 12 formed with polymers, 11 there are relatively few examples of photoswitches being self-assembled into well-defined, discrete structures. In a key example, pyridine-based ligands and palladium(II) were self-assembled into a [Pd12L24] 24+ molecular sphere with endohedral azobenzene groups 2a which could be switched with UV to increase the hydrophilicity of the sphere's cavity. Some other examples of pyridyl-functionalized switches include [M2L4] 4+ cages formed with stiff-stilbenes and palladium(II), 22 chiral [M6L3] 6+ metallocycles formed from dithienylethene (DTEs) 23 and platinum(II), 24 and related ligands reacted with iron(II/III) to form [Fe2L3] n+ helicates. 25 The first example of a molecular cage with functioning azobenzene-type photoswitches as linkers used cyclotriguaiacylene units with three appended pyridyl-azo-phenyl photoswitches and iridium(III) complexes to form [Ir3L2] 3+ cages. 26 The flexible linkers allowed photoswitching to occur without disrupting the cage topology. The most wellstudied photoswitchable cages are based on pyridylfunctionalized DTE photoswitches assembled with palladium(II) ions reported by the Clever group. 27 The difference in geometries has been exploited for selective guest uptake, 27a,27d,27e and control over macromolecular properties when incorporated into gels. 28 Photoswitching units can also modulate the topology of metallosupramolecular structures; however, this usually leads to the assembly of new non-discrete structures. 30 There are few reported examples of modulation between discrete metallosupramolecular structures, with some key examples represented in Figure 1b,c. 27b, 29 One example used azobenzene or stilbene based ligands to form [M2M'2L4] 8+ (M = Pd, M' = Pd or Re) macrocycles where UV irradiation isomerize the azo unit to contract the macrocycle to the smaller [M2L2] 4+ species. 29,31 Other examples use DTE-based ligands. 27b,27e For one system, the open and closed isomers give rise to a double-walled triangle (as the major component) and a cuboctahedral sphere, respectively. 27b These species can be interconverted using UV and green light, giving reversible control over the structure by external stimuli although the conversion was relatively slow, with a full cycle taking over 3 days. A more recent example was able to eject one ligand from a Pd2L4 cage upon irradiation. 27e Despite these examples, there are no reports of metallosupramolecular structures which can be reversibly rearranged using visible light only.</p><p>Herein we report a system of two discrete metallosupramolecular assemblies, formed from an orthofluoroazobenzene ligand (Figure 1d). The system can be driven out-of-equilibrium with visible light due to the different kinetic labilities of the structures. To the best of our knowledge this is one of the only examples of light-induced topology changes and the first example of all-visible light switching between discrete structures.</p><!><p>We synthesized substituted ortho-fluoroazobenzenes in moderate yield over three steps from commercially available 4-bromo-2,6-difluoroaniline (see SI-1,2 for details). 32 Compound 1 was obtained in 65% yield using a methodology previously used to generate unsymmetrical azobenzene derivatives, 18a,18b,33a,33b Boronic ester substituted ortho-fluoroazobenzene 2 has been previously reported, 18c but use of microwave heating allowed us to reduce the reaction time to 15 minutes with a trivial work-up that excluded chromatography. Suzuki coupling gave the photoswitchable ligand 3 (53% yield) and the control compound, phenyl derivative 4 (20% yield). The second coupling reaction did not always reach completion despite the arylhalide being in excess, with the mono-substituted product being identified and characterized (see SI-2.4, SI-16 for details). This suggests the second coupling reaction is considerably more difficult than the first. The compounds were isolated as mixtures of the thermodynamically favored E-isomer and the metastable Z-isomer. Heating a solution of 3 in DMSO-d6 generated the pure E-3 isomer as observed by 1 H and 19 F NMR spectra (Figure 2b). 34 The UV-vis absorption of photoswitchable ligand E-3 (Figure 2c) extends into the visible, with a visible absorption maximum at 466 nm assigned as the n-π* band and a band at 356 nm assigned to the π-π* transition (in DMSO at 298 K). Both transitions are red-shifted relative to the parent ortho-fluoroazobenzene, which has an n-π* transition at 460 nm and a π-π* transition at 314 nm (in DMSO at 298 K). 18a The larger red-shift of the n-π* band compared to the π-π* was also reported for 2,2',6,6'tetrafluoro-4,4'-diacetamidoazobenzene, 18a suggesting this effect is due to substitution with electron donating groups. Photoswitchable ligand 3 undergoes reversible photoswitching with visible light (Figure 2c). Irradiation of 3 with an LED centered at 530 nm generated a photostationary state comprising 80% Z-3 (calculated from 19 F NMR signal integrations, Figure 2b and SI-4.1). Subsequent irradiation at 410 nm generated a new photostationary state comprising 85% E-3. The calculated absorption spectrum 35 of Z-3 shows an n-π* transition with an absorption maximum at 432 nm, slightly red-shifted compared to unsubstituted or ester substituted orthofluoroazobenzenes (λmax = 417-421 nm). 18a The separation between the n-π* bands for the two isomers of 3 (Δλn-π* = 33 nm) is similar to that found for other orthofluoroazobenzenes with electron-donating groups in the para position, 18a but less than that for the parent orthofluoroazobenzenes or examples with electron withdrawing groups (Δλn-π* = 30 to 50 nm). 18b Nonetheless, selective photoswitching is still achieved between the isomers. Photoswitchable ligand Z-3 has a thermal half-life of ≈ 25 days at 298 K (thermal barrier of 110 kJmol -1 , measured at 333 K in DMSO, see SI-4.3). Photoswitch 4 has similar properties to photoswitchable ligand 3.</p><p>For example, photoswitch 4 has an n-π* absorption band at 462 nm and a π-π* band at 360 nm, and Z-4 has a thermal half-life of ≈37 days at 298 K (see SI-5 for details). Photoswitches 3 and 4 both have shorter half-lives compared to the parent ortho-fluoroazobenzene which has a half-life of 700 days (thermal barrier of 117 kJmol -1 , measured at 333-373 K in DMSO) 18b Having characterized photoswitchable ligand 3, we next investigated its self-assembly with palladium(II) ions. When [Pd(CH3CN)4](BF4)2 was added to E-3 in DMSO-d6 a new species was immediately formed as observed by 1 H and 19 F NMR spectroscopy (Figure 3b, SI-6). Equilibration in the dark at room temperature over 10 days led to the formation of a new, lower symmetry, assembly comprising 69% of the mixture (Figure 3c). Using 1 H NMR diffusion (Figure 4a) and ROESY NMR (Figure 4b) data we identified two separate species, with the higher symmetry species having a smaller hydrodynamic diameters (27 vs 31 Å). Similar self-assembly using a more soluble palladium salt 36 gave the same two structures albeit with a slightly different relative abundance, see SI-7. Characteristic downfield shifts of the 1 H NMR signals for pyridyl protons indicate coordination to the metal ion (see SI-8.1 for full details). 37 In the initially formed species the ligand retains its original symmetry and the 1 H NMR signal for H a (see Fig. 3 for atom labels) shifts upfield by ≈ 0.2 ppm, consistent with shielding effects commonly seen for related structures. 38,37b The lower symmetry species has a doubling of all ligand signals (Figure 3c), with a significant upfield shift (≈ 0.4 ppm) of the H e ' proton compared to the symmetric species. The 19 F NMR spectrum confirms the reduced symmetry with two peaks observed for the lower symmetry species. The significant peak shifts observed in the NMR spectra did not allow unambiguous assignment of the E/Z-isomerization state. The UV-visible absorption spectrum of the mixture was also unhelpful for assigning the E or Z isomer composition (SI-15.1). Therefore, a degradation experiment was performed. 4-Dimethylaminopyridine (DMAP) was added to the equilibrated mixture in the dark which rapidly disassembled the structures to form exclusively E-3 and [Pd(DMAP)4](BF4)2 as seen by 1 H NMR spectroscopy (See SI-12). Due to the long thermal half-life of Z-3, this degradation experiment indicates that the observed 1 H NMR peak shifts and changes in the UV-visible absorption spectra are due to the constrained local environment or distortions of the E-3 ligand imposed by the structure, rather than isomerization of the ligand.</p><p>High resolution electrospray ionization mass spectrometry (ESI-MS) identified two major species, a [Pd3(3)6] 6+ and a [Pd4(3)8] 8+ assembly (Figure 5, SI-9) with a range of charge states corresponding to sequential loss of BF4anions from these structures. The combination of NMR and MS data, together with preliminary molecular modelling, was used to propose the topologies of the self-assembled structures for [Pd3(3)6] 6+ and a [Pd4(3)8] 8+ (Figure 6b and 6c). For the [Pd3(3)6] 6+ species, the NMR spectra indicates a highly symmetrical structure, which is assigned as a double-walled triangle. 39 For the species with composition [Pd4(3)8] 8+ , several possibilities can be considered (Figure 6a): a doublewalled square, 39b,39e an interpenetrated double cage 40 or a distorted tetrahedron. 39a,39h,41 The double-walled square would nominally have D4h symmetry with all pyridyl rings being equivalent. This is inconsistent with the observed number of signals in the NMR spectra. The interpenetrated double cage structure would show a doubling of the 1 H and 19 F NMR signals as observed. However, in previous reports of such topologies the transient formation of a [Pd2L4] 4+ cage was observed in the 1 H NMR spectrum and by ESI-MS. 40b Such species were not observed for the current system and molecular modelling also suggests significant strain would be required in the [Pd2(3)4] 4+ subunit. The structure is therefore proposed as a distorted tetrahedron with C2v symmetry. Ligands with 3-pyridyl groups bridged by phenyl 39a,41b or BINOL linkers 41a have been previously assembled into analogous distorted tetrahedra with palladium(II), but the topology remains rare. 39a,41a,41b,41d,39h For [Pd4(3)8] 8+ the groups of signals from the non-equivalent ligands were assigned using 2D NMR techniques and by comparing to previously reported examples. 39a The local environment for the double-bridged ligands is similar to that observed for the double-walled triangle [Pd3(3)6] 6+ species. The single-bridged ligands are more similar to free E-3, especially the phenyl proton (H a' ) which is distal to coordinating pyridines. The molecular model suggests a longest axis (28 Å) in agreement with the calculated hydrodynamic diameter (31 Å) from the diffusion NMR data.</p><p>Variable temperature 1 H NMR spectra (SI-8.3) confirmed the two species were in equilibrium. Increasing the temperature to 333 K gave a mixture containing 63% of the smaller [Pd3(3)3] 6+ species. This is ascribed to entropic considerations, as proposed in other systems. 42 The system initially remained out of equilibrium upon cooling to 298 K, reaching the original distribution after 18 hours in the dark. This indicates that the double-walled triangle acts as a kinetic trap for the system, consistent with the initial observations upon combination of E-3 and [Pd(CH3CN)4](BF4)2.</p><p>Having investigated the self-assembly properties of E-3, we next investigated the behavior of the Z-3 isomer. A sample of 3 was enriched to 80% Z-3 by irradiation with 530 nm light, then combined with [Pd(CH3CN)4](BF4)2 in DMSO-d6. The resulting poorly resolved 1 H NMR spectrum suggests the formation of non-distinct or polymeric products, which do not significantly resolve over time (see SI-13). To understand the self-assembly behavior, we next considered the binding affinity of the ligand for palladium(II) centers. To the best of our knowledge, and despite their widespread use in supramolecular self-assembly, quantitative binding constants for simple pyridine derivatives to palladium(II) ions do not appear to be reported. To study a single 1:1 binding event, we used a palladium(II) complex with a tridentate terpyridine ligand (ttpy = 4'-(para-tolyl)-2,2':6',2''-terpyridine), [Pd(ttpy)(DMSO)](BF4)2, which has a weakly bound solvent molecule that can be readily exchanged for the other ligands. We used 3-methylpyridine as a simple monodentate ligand (SI-3 for synthetic details). Isothermal titration calorimetry (ITC) was used to measure the 1:1 binding constant (see SI-14.1). The relative binding constant is 1.73 ×10 4 mol -1 in DMSO, equivalent to a binding energy of just 24 kJ•mol -1 at 298 K. Similar ITC measurements with 3 and [Pd(ttpy)(DMSO)](BF4)2 indicated only weaker binding (Ka <1000), although solubility difficulties prevented quantitative measurements. Competitive binding experiments monitored by 1 H NMR spectroscopy confirmed that 3 is nearly completely displaced from [Pd(ttpy)(DMSO)](BF4)2 when one equivalent of 3methylpyridine is added (see SI-14.2), consistent with 3 being a surprisingly poor ligand for palladium(II). We also investigated the influence of palladium(II) ions on the photoswitching behavior of ligand 3. When 100 equivalents of [Pd(ttpy)(DMSO)](BF4)2 was added to ligand 3 and the sample was irradiated with 530 nm light, the same thermal Z→E half-life in the dark was measured by UV-vis absorption, (see SI-4.3, SI-4.4). As ligand 3 has only weak affinity for palladium(II), its ability to assemble into discrete structures suggests that cooperativity is responsible for stabilizing the resulting self-assembled structures.</p><p>The distribution between [Pd3L6] 6+ and [Pd4L8] 8+ can be pumped away from equilibrium using light, even though Z-3 did not self-assemble into well-defined structures with palladium(II) ions. After irradiating a mixture of [Pd3(3)6] 6+ and [Pd4(3)8] 8+ in DMSO-d6 with 410 nm light for 10 minutes, 1 H and 19 F NMR spectroscopy reveals a significant increase in the population of [Pd3(3)6] 6+ , while also showing the concomitant decrease of [Pd4(3)8] 8+ (Figure 7b, ii). No new signals were observed, suggesting no new well-defined self-assembled species were formed. This observation was reaffirmed by high-resolution ESI-MS, the relative population of [Pd3(3)6] 6+ increased after irradiation with 410 nm light (see SI-15). Irradiating the sample with 530 nm light for 10 minutes resulted in the deformation of the two species as seen in the poorly defined 1 H and 19 F NMR spectra, suggesting the formation of polymeric species, or other low symmetry species (Figure 7b, iii). The large population of [Pd3(3)6] 6+ could be recovered by irradiating the system again with 410 nm light for 10 minutes (Figure 7b, iv), demonstrating selective and reversible assembly and disassembly of the triangle species. After heating the sample at 60 °C for 2 days the original distribution was largely recovered ([Pd3(3)6] 6+ : [Pd4(3)8] 8+ = 3:4, Figure 7b, v)), although some chemical shift changes and peak broadening had occurred. The broad peaks observed in the 1 H and 19 F NMR spectra after irradiation are consistent with the involvement of Z-3 within the self-assembled species (Figure 7b, iii), either as structural components or as guest molecules. This effect is far more pronounced within [Pd4(3)8] 8+ , supporting the notion that [Pd4(3)8] 8+ is more flexible and able to accommodate the mismatched ligand whereas [Pd3(3)6] 6+ is more rigid and well-defined. 3)6] 6+ ); ii) the same sample after irradiation with 410 nm light for 10 minutes (82% [Pd3(3)6] 6+ ); iii) the same sample after irradiation with 530 nm light for 10 minutes; iv) the same sample after irradiation with 410 nm light for 10 minutes again; and v) the same sample after 2 days of being heated at 60 °C followed by 6 h of equilibrating at room temperature.</p><p>The selective disassembly of [Pd4(3)8] 8+ can be rationalized by considering the composition of ligands, the rate of ligand exchange for each species, and the constraints imposed on the photoswitching of ligand 3 while assembled. Variable temperature NMR experiments confirm the struc-tures are dynamic with exchange of ligands and solvent molecules, as is common for palladium(II)-pyridyl assemblies (see SI-8.3). 43 If photoisomerization is suppressed within the self-assembled structures, as observed for a DTE-based cage, 27b ligand 3 can only isomerize after dissociating from palladium. For the tetrahedron [Pd4(3)8] 8+ , a E-3 ligand can dissociate and photoisomerize, but the newly generated Z-3 ligand cannot reassemble into the same original structure. We propose that a metastable [Pd4(3)7] 8+ structure is formed and the ligands rapidly rearrange to form the double-walled triangle, [Pd3(3)6] 6+ . As [Pd3(3)6] 6+ is more inert, any free E-3 in solution will be kinetically trapped as [Pd3(3)6] 6+ . As such, irradiation with 410 nm light continuously pumps the system out-ofequilibrium to favour the formation of the less thermodynamically preferred [Pd3(3)6] 6+ . The PSS generated when irradiated with 530 nm light comprises only 20% E-3, which appears too low to form a significant amount of [Pd3(3)6] 6+ . This finding is consistent with our experiments using a sample of enriched Z-3 and palladium(II) which also resulted in the same ill-defined mixtures.</p><p>The observed behavior is surprising as it results from a relatively small change (~20%) in the isomer distribution caused by irradiating with 410 nm light. Typically, stimuli responsive architectures are designed to maximize the proportion of components that are switched. This work offers a different approach, where small changes in isomer distribution can be amplified to significant changes within the system, similar to the sergeants-and-soldiers concept 44 in self-sorting. To the best of our knowledge, this is the first example of a self-assembled system where the configuration can be controlled using only visible light and the resultant distribution contains the same sub-components as the equilibrium distribution.</p><!><p>We have shown that building visible-light switchable ofluoroazobenzenes into palladium(II)-pyridyl selfassemblies leads to visible-light responsive systems. Irradiating with visible light reversibly redistributes the subcomponents, driving the system out-of-equilibrium to form the higher energy, but less labile, structure. Unlike previous examples, the distinct assemblies contain the same photoisomer of the ligand. This approach of pumping systems to metastable states exploits kinetic effects to amplify small changes in photoisomer distributions to generate large changes in structural distributions.</p>

ChemRxiv

Structural characterization and bioactivity analysis of two-component lantibiotic flv system from a ruminant bacterium

Summary The discovery of new ribosomally synthesized and post-translationally modified peptide natural products (RiPPs) has greatly benefitted from the influx of genomic information. The lanthipeptides are a subset of this class of compounds. Adopting the genome mining approach revealed a novel lanthipeptide gene cluster encoded in the genome of Ruminococcus flavefaciens FD-1, an anaerobic bacterium that is an important member of the rumen microbiota of livestock. The post-translationally modified peptides were produced via heterologous expression in Escherichia coli. Subsequent structural characterization and assessment of their bioactivity revealed features reminiscent of and distinct from previously reported lanthipeptides. The lanthipeptides of R. flavefaciens FD-1 represent a unique example within two-component lanthipeptides, consisting of a highly conserved \xce\xb1-peptide and a diverse set of eight \xce\xb2-peptides.

structural_characterization_and_bioactivity_analysis_of_two-component_lantibiotic_flv_system_from_a_

Introduction<!>Bioinformatic analysis of the FlvA substrates<!>Attempted detection of modified FlvA production in R. flavefaciens FD-1<!>Heterologous production of modified FlvA peptides<!>FlvM1 is selective for FlvA1 and FlvM2 is selective for the FlvA2 peptides<!>Partial FlvA2.x leader peptide removal and structural analysis<!>Proteolytic removal of the remaining portion of the FlvA2.x leader peptides<!>Antimicrobial activity of the modified FlvA core peptides<!>Discussion<!>Significance<!>Experimental Procedures

<p>The human microbiome has received much attention in recent years for its connection to human health (Garrett, 2015; Wlodarska et al., 2015; Yurkovetskiy et al., 2015). Increasingly, the mutualistic relationship between the microbiome and host has been demonstrated (Hooper et al., 2012; Maynard et al., 2012; Nicholson et al., 2005; Tremaroli and Backhed, 2012). One of the beneficial roles of the gut microbiota of a healthy individual is believed to be the resistance that is provided against colonization by pathogens (Vogt et al., 2015). Similarly, the microbiota of livestock is critical for animal health, and a better understanding of the mechanisms that confer pathogen resistance in this setting is desired. Ruminant animals such as cattle and sheep have a symbiotic relationship with ruminal microorganisms that can degrade cellulose and/or hemicellulose (Hungate, 1966; White et al., 2014). Among the bacteria with cellulose activity in the rumen are the anaerobic ruminococci (Weimer, 2015). Some Ruminococcus strains have been reported to produce bacteriocins (Chen et al., 2004; Dabard et al., 2001; Gomez et al., 2002; Marcille et al., 2002; Pujol et al., 2011), a class of ribosomally produced antimicrobial compounds that may be important for maintaining a niche in the competitive microbial environment of the rumen as well as in the human gastrointestinal tract (Russell and Mantovani, 2002). Ruminococcus flavefaciens FD-1 has a particularly high cellulolytic activity (Shi and Weimer, 1996). It has not been reported to produce any antimicrobial peptides, but its genome was recently sequenced (Berg Miller et al., 2009), providing a view of the genetic capability to produce such compounds. Here we show that its genome encodes an unusual group of lanthipeptides that are composed of four highly conserved copies of a peptide that likely binds lipid II and a diverse set of eight additional peptides, some of which act synergistically with the lipid II-binding peptide. The possible functional implications of the system in the context of the rumen environment are discussed.</p><p>Lanthipeptides are a class of ribosomally synthesized and post-translationally modified peptides (RiPPs) that are characterized by the presence of thioether linkages. These crosslinks are formed by the Michael-type addition of cysteine thiols onto dehydroamino acids, and the resulting linkages are critical for their bioactivity. Lanthipeptides that display antibacterial activity are called lantibiotics, and interest in these compounds has stemmed from their potent antibacterial activity and low propensity for the development of resistance (Lubelski et al., 2008). Although the prospects of RiPP production in anaerobic organisms has been highlighted (Letzel et al., 2014), relatively few such compounds produced by the genus Ruminococcus have been reported thus far (Crost et al., 2011; Dabard et al., 2001; Kalmokoff and Teather, 1997). We designate the lanthipeptide biosynthetic gene cluster in R. flavefaciens FD-1 flv (Fig. 1A).</p><p>The LanA substrate peptides that are converted into lanthipeptides are divided into an N-terminal leader and C-terminal core peptide (Knerr and van der Donk, 2012). The leader peptide does not undergo post-translational modification during lanthipeptide maturation, whereas select Ser, Thr, and Cys residues within the core peptide are enzymatically cyclized to generate thioether linkages. Crosslink formation involves the dehydration of Ser and Thr residues, resulting in dehydroalanine (Dha) and dehydrobutyrine (Dhb), respectively. Subsequent attack of Cys thiols onto a Dha results in a lanthionine (Lan), whereas cyclization with a Dhb residue yields a methyllanthionine (MeLan). Both dehydration and cyclization reactions are catalyzed by lanthipeptide synthetases whose functional domains form the basis for a classification scheme for lanthipeptide systems. In the case of class II lanthipeptides, a single bifunctional enzyme termed LanM carries out the dehydration and cyclization reactions within the core peptide of the LanA substrate (Siezen et al., 1996; Xie et al., 2004). Compared to known lanthipeptide biosynthetic gene clusters, the flv cluster possesses unusual characteristics that drew our attention. The cluster encodes twelve putative LanA genes and two putative LanM genes (Fig. 1A). The large number of substrates is reminiscent of the prochlorosin-like systems (Li et al., 2010; Zhang et al., 2014b) and the presence of two LanM synthetases resembles two-component lantibiotic systems such as lacticin 3147 and haloduracin (McAuliffe et al., 2000; McClerren et al., 2006). Most of the two-component lanthipeptides investigated to date consist of a lipid II-binding α-peptide and a β-peptide that acts synergistically to affect pore formation via a poorly defined mechanism (Begley et al., 2009; McAuliffe et al., 1998; Oman et al., 2011b; Oman and van der Donk, 2009; Wiedemann et al., 2006). The gene clusters for two-component lantibiotics usually encode a single copy of an α and a β-peptide, and hence the twelve FlvA peptides encoded within the flv gene cluster are unprecedented. In light of these unusual characteristics, we set out to obtain the modified FlvA peptides. Systematic production of these peptides facilitated their structural characterization and bioactivity analysis.</p><!><p>Inspection of the FlvA amino acid sequences revealed that four out of the twelve FlvA peptides possessed nearly identical amino acid sequences, with a single and two amino acid differences in the leader and core peptide, respectively (Fig. 1B). Querying the non-redundant protein database from the National Center for Biotechnology Information using the Basic Local Alignment Search Tool (Altschul et al., 1990) with the core peptides from this nearly identical set of FlvA peptides returned the LicA1 peptide sequence (Fig. 1C) (Begley et al., 2009; Dischinger et al., 2009; Shenkarev et al., 2010). The modified LicA1 core peptide forms the α-peptide of the two-component lantibiotic lichenicidin. An alignment of the lichenicidin peptides and all of the FlvA peptides using the alignment program Clustal Omega (Sievers et al., 2011) indicated that the latter could be roughly divided into a set of putative α and β precursor peptides (Fig. 1C,D). The four nearly identical FlvA peptides (Fig. 1B) make up the putative α-peptides and were designated FlvA1.x (where x = a–d), whereas the remaining eight FlvA peptides were hypothesized to be the corresponding β-peptides and were designated FlvA2.x (where x = a–h). Whereas the FlvA1.x core peptide amino acid sequences resemble that of lichenicidin α, the core peptides of FlvA2.x exhibited more limited sequence similarity with lichenicidin β, although some of the Ser, Thr, and Cys residues in the C-termini were similarly positioned (Fig. 1D). In general, the core peptides of the FlvA peptides are longer than those of the previously characterized α and β-peptides.</p><p>Further comparison of the FlvA peptides with other two-component lantibiotics revealed additional shared and unusual features. The FlvA1.a peptide has Ser20 and Cys30 positioned to form a putative lipid II binding motif (Fig. 1C, underlined) found in the lantibiotics mersacidin, lacticin 481, and nukacin ISK-1 (Dufour et al., 2007; Hsu et al., 2003; Islam et al., 2012; Szekat et al., 2003) and in the α-peptide of two-component lantibiotics (Begley et al., 2009; Dischinger et al., 2009; Martin et al., 2004; Oman and van der Donk, 2009). However, compared to the other α-peptides, the FlvA1.x peptides contain more Cys residues in the core peptide (Fig. 1C). The involvement of these Cys residues in thioether rings or disulfide bridges could endow the FlvA1.a peptide with increased stability, which may be important in the ruminal environment. Similar to the FlvA1.x peptides, many of the FlvA2.x core peptides (c-e, g, and h) also have additional Ser and Thr residues compared to other two-component β-peptides (Fig. 1D). Notably, some of these residues form a putative Dhx-Dhx-Xxx-Xxx-Cys motif (where Dhx = Dha or Dhb and Xxx = any amino acid), which has been shown in some lanthipeptides to result in MeLan rings with LL stereochemistry (Lohans et al., 2014; Tang et al., 2015; Tang and van der Donk, 2013). Corollary to the structural diversity of the FlvA2.x peptides would be a high substrate tolerance of the FlvM enzyme(s) responsible for installing the (Me)Lan rings.</p><!><p>Liquid cultures of anaerobically grown R. flavefaciens FD-1 were desalted and analyzed for the presence of FlvA peptides. MALDI-TOF-MS analysis did not provide clear indication of masses corresponding to putative peptide products of the flv cluster. In alternative attempts to elicit peptide production, R. flavefaciens FD-1 and Ruminococcus albus 7 were co-cultured as well as grown in proximity to one another on solid media, an approach that has been successful for other genera (Traxler et al., 2013; Yang et al., 2011). However, no new ions with m/z > 2000 could be observed by MALDI-TOF MS. These observations are representative of a commonly encountered hurdle in natural product research: the environmental triggers for expression of biosynthetic genes are generally poorly understood and often cannot be replicated in the laboratory (Bode and Müller, 2005; Winter et al., 2011). The inability to detect potential products of the flv cluster prompted consideration of alternate strategies to produce and assess the products generated from the flv genes.</p><!><p>To access the products of the Flv system, a validated heterologous production strategy using Escherichia coli was employed. This approach has been successfully applied to the production of several lantibiotics and involves the co-expression of a LanA peptide and cognate LanM enzyme in E. coli, which typically results in full posttranslational modification of the peptide. The modified full length LanA peptide is then purified from E. coli and the leader peptide is removed in vitro using a variety of proteolytic strategies (Basi-Chipalu et al., 2015; Caetano et al., 2011; Garg et al., 2012; Lin et al., 2011; Nagao et al., 2005; Okesli et al., 2011; Shi et al., 2011; Tang and van der Donk, 2013; Wang et al., 2014b). The sequence similarity of FlvM1 with characterized two-component LanM1 enzymes is greater than the similarity of the latter enzymes with FlvM2. Thus, we hypothesized that FlvM1 was responsible for the modification of the putative FlvA1.x peptides and FlvM2 for the FlvA2.x peptides. In order to probe this hypothesis, pilot co-expression studies of an arbitrarily chosen peptide FlvA2.g with either FlvM1 or FlvM2 were undertaken. Co-expression of FlvA2.g with FlvM2 resulted in dehydrated peptides whereas co-expression of FlvA2.g with FlvM1 yielded no dehydration (Fig. S1A). Further experiments with all of the FlvA2.x peptides showed they were also dehydrated upon co-expression with FlvM2 in E. coli. Similarly, FlvA1.a was dehydrated upon co-expression with FlvM1 (Fig. S1B). Identical results were obtained in vitro with enzymes and substrates that were expressed as His6-tagged fusion proteins and purified by immobilized affinity chromatography (e.g. Fig. 2A and S2). In all cases the number of dehydrations was observed to be less than the total number of Ser and Thr residues within the core peptide (Table 1, Fig. 2B, Fig. S3), but Ser and Thr residues that escape dehydration are not unusual in lanthipeptides (Rink et al., 2005) including the lichenicidins (Caetano et al., 2011; Shenkarev et al., 2010).</p><!><p>In order to assess the substrate tolerance of FlvM1, FlvA2.a and FlvA2.g were also incubated in vitro with FlvM1 under identical conditions. Analysis of the assay revealed little to no modification of either peptide by FlvM1 (Fig. S2). Similarly, FlvM2 was unable to modify FlvA1.a to any appreciable extent when exposed to the peptide in vitro under the standard reaction conditions (Fig. S2). The inability of FlvM1 to modify FlvA2.a and 2.g and FlvM2 to modify FlvA1.a suggests a high degree of substrate selectivity that is probably governed by recognition of the different leader peptides (Fig. 1B) (Thibodeaux et al., 2015; Yang and van der Donk, 2013).</p><!><p>The alignment of the amino acid sequences of the FlvA peptides shows a conserved GA/GG motif located in the middle of the peptides (Fig. 1B). Based on previous studies of other class II lanthipeptides, this motif likely marks the boundary between the FlvA leader and core peptides (Uguen et al., 2005). Usually, leader peptides ending in the GA/GG motif are removed by the Cys protease domain of LanT transporters (Håvarstein et al., 1995), and indeed such a protein is present in the gene cluster (FlvT, Fig. 1A). Evidence that the GA/GG sequence is indeed the cleavage motif for the FlvA peptides was provided by co-expression of FlvT with FlvM2 and FlvA2.g in E. coli. The supernatant of the co-expression was analyzed by MALDI-TOF MAS, which resulted in detection of masses corresponding to cleavage of modified FlvA2.g after the GA sequence (Fig. S6H). These masses were not observed in the culture co-expressing only FlvM2 and FlvA2.g. This observation is consistent with the previously reported removal of the leader peptide of modified LicA2 upon expression of the lic gene cluster, which contains licT, in E. coli (Caetano et al., 2011). Although the use of FlvT confirmed the predicted leader peptide cleavage site, this strategy unfortunately did not result in producing sufficient quantities of the Flvα and β-peptides for structure determination, in part because of the difficulty associated with purifying the desired, modified peptide from partially dehydrated peptides that were also secreted.</p><p>We therefore turned towards in vitro removal of the putative leader peptides from the His6-tagged FlvA2.x peptides after co-expression with FlvM2 and purification by immobilized nickel affinity chromatography. The purified peptides were first subjected to proteolysis by endoproteinase Glu-C. Although FlvA2.f-2.h have Glu residues within their core peptides, appreciable proteolysis at these sites was not observed, probably because they are protected by the post-translational modifications. Instead, treatment with Glu-C resulted primarily in fragments arising from proteolysis within the leader peptide sequence, specifically at Glu−5 for FlvA2.a, Glu−6 for FlvA2.b through FlvA2.e and FlvA2.g, Glu−7 for FlvA2.h, and Glu−10 for FlvA2.f (for sequences, see Fig. 1B). The Glu-C-generated C-terminal fragments of the modified FlvA2.x peptides were purified by reversed-phase high performance liquid chromatography (RP-HPLC) and analyzed by electrospray ionization tandem mass spectrometry (ESI-MS/MS) (e.g. Fig. 3A).</p><p>A lack of fragmentation was observed in the C-termini of all of the modified FlvA2.x peptides analyzed, suggesting the presence of thioether rings (Fig. 3A and S4). ESI-MS/MS analysis of modified FlvA1.a also indicated the absence of fragmentation in the C-terminus of the peptide, consistent with cyclization in this region of the peptide (Fig. S4). To confirm the presence of (Me)Lan residues within the FlvM-modified FlvA1.a and FlvA2.x peptides, the peptides were hydrolyzed and the resulting residues were derivatized to the corresponding pentafluoropropionamide methyl esters. These volatile derivatives were then separated by chiral gas chromatography monitored by mass spectrometry (GC/MS), which confirmed that the modified FlvA1.a and FlvA2.x peptides indeed contained (Me)Lan residues (Fig. 3B and Fig. S5). The stereochemistry of the observed (Me)Lan residues was determined by using synthetic standards with known stereochemistries (Ross et al., 2010), and is listed in Table 1. As anticipated based on previous predictions (Tang and van der Donk, 2013), peptides containing a Dhx-Dhx-Xxx-Xxx-Cys motif resulted in LL-Lan whereas all other Lan and MeLan residues had the DL stereochemistry.</p><p>The possibility of incomplete cyclization was assessed by iodoacetamide (IAA) assays; iodoacetamide selectively alkylates free Cys residues and not thioethers. For all FlvM-modified peptides except FlvA1.a, treatment with IAA did not result in appreciable formation of adducts, which is indicative of complete cyclization; di-alkylation was observed with FlvM1-modified FlvA1.a (Fig. S6C). Based on the fragmentation pattern (Fig. S6D) and the alignment with other α-peptides (Fig. 1C), the di-alkylation product most likely involves alkylation at Cys8 and Cys29. Two non-cyclized Cys are also present in haloduracin α, but the Cys residues involved do not align with those of Flvα.a. The two free Cys residues in modified FlvA1.a presented the possibility of a disulfide bond in the structure of mature FlvAα.a. The antimicrobial activity of some lantibiotics has been reported to depend on the presence of such disulfide bonds (Kabuki et al., 2009; Lin et al., 2011; Wang et al., 2014b; Zhang et al., 2014a). Modified FlvA1.a was therefore incubated with oxidized and reduced glutathione in an attempt to oxidatively fold the peptide (Oman et al., 2011a). However, the peptide remained reduced under these conditions, as observed by ESI-MS, suggesting that unlike lantibiotics that contain disulfide bonds, the two Cys in modified FlvA1.a are not in a conformation in which they readily form a disulfide bond (Basi-Chipalu et al., 2015; Lin et al., 2011).</p><!><p>Complete removal of the leader peptide was challenging as introduction of artificial protease cleavage sites resulted in either incomplete modification by the FlvM enzymes, incomplete proteolysis because of the post-translational modifications near the cleavage site, or cleavage in the core peptides, which are problems that have been previously observed for other lanthipeptides (Garg et al., 2012; Plat et al., 2011; Tang and van der Donk, 2012). We therefore resorted to treating the Glu-C-generated FlvA2.x peptide fragments with aminopeptidase (Bindman and van der Donk, 2013; Majchrzykiewicz et al., 2010; Shi et al., 2012). It was anticipated that the proteolytic activity of aminopeptidase would terminate upon encountering the dehydrated residues featured at or near the predicted N-termini of most of the FlvA2.x peptides (Fig. 5). MALDI-TOF-MS analysis of aminopeptidase reactions indicated that for FlvA2.b, the predicted native product was formed. For FlvA2.c, FlvA2.e, and FlvA2.g, aminopeptidase removed all remaining residues of the leader peptides as well as the predicted first residue of the core peptides, resulting in N-terminal lanthionines. We term these products Δ1-Flvβ.c, Δ1-Flvβ.e, and Δ1-Flvβ.g. For FlvA2.d and FlvA2.h, the aminopeptidase treatment left two amino acids of the leader peptide on the final products, likely because of the high Gly content at the junction between the core and leader peptides, making these peptides less efficient substrates for aminopeptidase (Velásquez et al., 2011). We term these products +2Flvβ.d and +2Flvβ.h. For FlvA2.a, three residues remained after aminopeptidase treatment, resulting in +3Flvβ.a. Finally, in the case of FlvA2.f, seven residues from the leader peptide remained after aminopeptidase treatment (Fig. S6G; +7Flvβ.f).</p><p>This sequential digestion approach works reasonably well when the N-terminus of the lanthipeptide is blocked by post-translational modifications, but if the N-terminus is a linear sequence as predicted for the product of FlvA1.a, the method does not work. Instead, removal of the FlvA1.a leader peptide was achieved via a single step proteolysis using chymotrypsin (Fig. S6B). Assuming that the GA/GG sequence marks the end of the leader peptide, chymotrypsin cleaves after Trp2 in the core peptide of FlvM1-modified FlvA1.a, and therefore the resulting peptide was designated Δ2-Flvα.a. In an effort to avoid removal of the two N-terminal amino acid residues of the FlvA1.a core peptide, a mutant FlvA1.a peptide with Glu inserted before the first residue of the core peptide was generated (termed FlvA1.a(A−1insE)), and co-expressed with FlvM1. Doing so resulted in six-fold dehydrated FlvA1.a(A−1insE), and treatment of this mutant peptide with Glu-C yielded the desired Flvα.a (Fig. S6E).</p><!><p>The peptides produced by in vitro leader peptide removal were purified by RP-HPLC, and the modified FlvA peptides were first assayed for antimicrobial activity against the lantibiotic-sensitive, aerobic microorganism Micrococcus luteus DSM 1790. In an agar diffusion test, Δ1-Flvβ.c and Δ1-Flvβ.e exhibited antimicrobial activity that was not synergistically enhanced by Δ2−Flvα.a (Fig. 4). In contrast, combinations of Flvβ.b and Δ1-Flvβ.g with Δ2−Flvα.a displayed the synergistic antimicrobial activity that is characteristic of two-component lantibiotics. A comparable level of synergistic, antimicrobial activity was also observed for a combination of Flvα.a and Flvβ.b (Fig. S6F), suggesting that the removal of the additional two amino acids at the N-terminus of Δ2−Flvα.a is not detrimental for bioactivity. Since we could obtain considerably larger quantities of Δ2−Flvα.a, this peptide was used for all subsequent bioassays. Δ1-Flvβ.g also displayed activity by itself that was lower than that observed when spotted with Δ2−Flvα.a. The remainder of the Flvβ-peptides (+3Flvβ.a, +2Flvβ.d, +7Flvβ.f, and +2Flvβ.h) did not display antimicrobial activity, whether tested separately or in combination with Δ2−Flvα.a (Fig. 4). Enhanced activity beyond what was observed in pairwise combination was also not detected by combining all peptides. In light of their antimicrobial activity, Flvα.a, β.b, β.c, β.e, and β.g were designated flavecins.</p><p>M. luteus is often used as an indicator strain to test for antimicrobial activity of lipid II-targeting peptides such as two-component lantibiotics, but it is not a rumen bacterium. To investigate the activity of the peptides against bacteria that would be more relevant in the ruminal context, the flavecins were assayed against Ruminococcus albus 7 and R. flavefaciens C94 under anaerobic conditions. Interestingly, very similar patterns of activity were observed with Δ2−Flvα.a, Flvβ.b, and Δ1-Flvβ.e (Fig. S6A). As with M. luteus, the activity was weak, suggesting that these organisms are not the intended target of these compounds. Alternatively, the weak antimicrobial activity exhibited by the flavecins may suggest that these peptides mediate more subtle microbial interactions (D'Onofrio et al., 2010).</p><!><p>The sequence similarity of LanM enzymes was used to survey genomes for unusual class II lanthipeptide biosynthetic systems, which resulted in the identification of the novel flv gene cluster in R. flavefaciens FD-1. Recently, Singh and Sareen also bioinformatically identified the flv cluster via the sequence similarity of FlvT to HalT (Singh and Sareen, 2014), but this previous study did not characterize the enzymes or the products of the cluster nor discussed its unusual characteristics. Some features of the flv cluster are reminiscent of two-component lantibiotics but the high number of substrate genes with diverse sequences is not consistent with a typical two-component system. Previous reports on antimicrobial peptides produced by species from the Ruminococcus genus (Chen et al., 2004; Odenyo et al., 1994; Russell and Mantovani, 2002) suggested that the mature FlvA peptides could represent antimicrobial defenses of R. flavefaciens FD-1. In addition, previous studies of the lantibiotics ruminococcin A and butyrivibriocin OR79A (Dabard et al., 2001; Kalmokoff et al., 1999; Marcille et al., 2002) suggested that anaerobic bacteria associated with the digestive system are a source of bioactive lanthipeptides. Thus, characterization of the flv gene cluster and the resulting modified FlvA peptides was undertaken.</p><p>The structure of Flvα.a suggested by sequence homology and tandem MS data is consistent with those of other α-peptides of two-component systems. Although a general lack of fragmentation from the middle of the peptide to the C-terminus makes definitive assignment of the ring structure in this region difficult (Fig. S4I), the ring pattern is very likely the same as those in Halα, Ltnα, and Licα (Fig. 1C,D) (Cooper et al., 2008; Martin et al., 2004; Shenkarev et al., 2010). The most N-terminal Cys residue does not appear to be cyclized since fragmentation is observed on either side of the residue, and IAA assays showed the presence of two free Cys (Fig. S4I and S6C). The possibility that these two Cys might be engaged in a disulfide bond was not supported by attempts to oxidatively cyclize them.</p><p>Based on the results obtained from tandem MS and GC/MS, the modified FlvA2 peptides can be divided into two groups (Fig. 5). The first group features an N-terminal ring, and the second category does not have the residues required to form this N-terminal ring. The assignment of Flvβ-peptides into the former group was based on the lack of fragmentation at the N-terminal extremity of these peptides as well as the LL stereochemistry of the Lan rings observed by GC/MS. Such a ring is also present in the β-peptides of haloduracin and lichenicidin (Cooper et al., 2008; Martin et al., 2004; Shenkarev et al., 2010). The central portion of all Flvβ-peptides contains a ring consisting of seven residues. The assignment of this ring is based on the tandem MS data that show a clear lack of fragmentation for all Flvβ-peptides in this region (Fig. S4). This ring also nearly aligns to the B ring of other two-component peptides, but the latter feature a smaller five-amino-acid ring in this region (Fig. 1D). The size of this central ring is not entirely clear for Flvβ.b, β.c, β.e, and β.g since a six-amino-acid ring is also possible. However, formation of a seven-amino-acid ring is more likely based on Flvβ.a, β.d, β.f and β.h, which unambiguously contain such a ring. Also, comparison of the fragmentation patterns of Flvβ.b, β.c, β.e and β.g imply that one Ser/Thr within this region consistently escapes modification (Fig. 5), suggesting the Ser/Thr that could result in a six-amino-acid ring is actually not dehydrated.</p><p>Two combinations of peptides resulted in synergistic bioactivity. Adding Δ2-Flvα.a to either Flvβ.b or Δ1-Flvβ.g resulted in clearly enhanced zones of growth inhibition. A comparison of the structures determined for Flvβ.b and Δ1-Flvβ.g indicates that the similarities are all in the C-terminal region, suggesting that this segment of the peptide is responsible for the interactions of the Flvβ peptides with the Flvα peptide. This observation is in agreement with the extensive mutagenesis data on the β-peptide of lacticin 3147 (Cotter et al., 2006). Two possibilities were considered for the relatively low observed antibacterial activity. Perhaps the small deviations between the predicted leader peptide removal sites and the products obtained after in vitro leader peptide removal resulted in imperfect interactions of the peptides. We did not find this explanation very likely since the N-termini of the eight Flvβ-peptides (and other two-component lanthipeptides) are already highly heterogeneous by sequence (Figure 1D), and more importantly, the antimicrobial activity observed for the predicted Flvα.a and β.b core peptides was comparable to the combination of Δ2−Flvα.a and β.b (Fig. S6F). Alternatively, perhaps the flv cluster is a pseudo cluster. We find this possibility also highly unlikely. It seems no coincidence that the four copies of the α-peptide are highly conserved, consistent with the general model of two-component lantibiotics in which this peptide recognizes a structurally conserved target (lipid II). The observation that the cluster encodes four such copies and eight copies of the β-peptides in light of a 1:2:2 stoichiometry observed for lipid II:Halα:Halβ (Oman et al, 2011b) also seems non-coincidental, but attempts to obtain increased activity by various combinations of peptides have thus far not been successful.</p><p>The weak observed activity may instead reflect our inability to identify the physiological target organism(s). Ruminal two-component lantibiotics may be much more fine-tuned for specific target organisms than observed in other environments. It has been reported that bacterial diversity in vertebrate-associated communities is quite different from that in soil or aquatic environments, with 16S rRNA gene-based trees indicating extensive diversity at the species level rather than at higher levels as observed in the latter environments (Backhed et al., 2005; Dethlefsen et al., 2007; Kim et al., 2011; Ley et al., 2008). Perhaps such an environment has promoted the thus-far unique example of a multi-component lantibiotic system with eight diverse β-peptides. The corollary of this hypothesis is that it may be very difficult to identify the specific physiological target organism of these peptides, especially if the flavecins exhibit the narrow target specificity that is not unusual for RiPPs (Scholz et al., 2011; Wang et al., 2014a). In turn, this may explain the rather weak activity observed against the organisms tested thus far. At present, still very few lanthipeptide gene clusters are known from ruminant environments (Azevedo et al, 2015) and we have not observed any other two-component systems in the available genomes. As discussed below, this may be the result of the still limited available genome sequence information, and as more genomes are sequenced, additional such examples may be uncovered.</p><p>The current work on the genetic capacity of R. flavefaciens FD-1 to produce a diverse set of eight different β-peptides has similarities but also important differences compared to previous examples of combinatorial biosynthesis of lanthipeptides. In 2010, the first example was reported of a single enzyme making 30 different lanthipeptides in the cyanobacterium Prochloroccocus MIT 9313 (Li et al., 2010). At the time this was a curious finding isolated to one genome, but in the intervening time as more genomes have been sequenced, this combinatorial biosynthesis has been found in many different cyanobacteria genera (Zhang et al., 2014b). To date, none of the products has shown any antimicrobial activities and the products have all had very different structures. In contrast, the example described here is clearly different. Firstly, the four copies of the α-peptide are essentially identical and all contain the highly conserved lipid II binding motif (Fig. 1C), strongly suggesting that the products are antimicrobial peptides. Secondly, the system contains two enzymes that have clearly defined roles, unlike the one enzyme in the cyanobacterial systems that has evolved for promiscuity. Thirdly, unlike the one cyanobacterial enzyme, which in order to be able to accommodate a highly diverse set of substrates is a very slow enzyme (Thibodeaux et al., 2014), the FlvM2 enzyme is very efficient in processing the eight FlvA2.x substrates that are diverse but that have maintained a certain core set of rings at the C-terminus. All of these observations point at a two-component lantibiotic system where the diversification of the β-peptides has been beneficial in the specific growth environment of the producer organism.</p><p>In conclusion, we present the characterization of the ring topology and stereochemistry of nine different lanthipeptides encoded in the genome of R. flavefaciens FD-1. A subset of the modified Flvβ core peptides (Flvβ.b and Δ1-Flvβ.g) demonstrated enhanced antibacterial activity when used in combination with Flvα.a, demonstrating that at least a subset of these peptides constitute two-component lantibiotic systems. Although the FlvA2 peptides display appreciable variation in their amino acid sequence, their post-translational modification is orchestrated by only one FlvM2 enzyme that is responsible for the modification of eight FlvA2 peptides with quite diverse sequences and final ring topologies. Elucidation of the features of the Flv system sets the stage for more in depth investigation of the biological roles of this curious set of lanthipeptides.</p><!><p>A heterologous expression system was used to access post-translationally modified lanthipeptides from the anaerobic Ruminococcus flavefaciens FD-1, thus enabling characterization of their structure and activity. FlvM2 is a substrate-tolerant enzyme that dehydrates and cyclizes eight different FlvA2 peptides. A subset of the FlvA2.x modified peptides displays synergistic antimicrobial activity in the presence of the FlvM1-modified FlvA1.a peptide. Thus, the flv system is the first example of a two-component lantibiotic gene cluster with several unique β-peptides.</p><!><p>For cloning and biochemical detailed procedures, see the Supplemental Experimental Procedures. Expression and purification of the FlvA peptides was carried out as previously described with modifications to the protocol (Li et al., 2009). The LanM in vitro, IAA assay, and oxidative folding reaction conditions were adapted from previous reports. MALDI-TOF-MS analysis of desalted samples was completed using a Bruker UltrafleXtreme MALDI-TOFTOF-MS maintained in the UIUC School of Chemical Sciences Mass Spectrometry Laboratory.</p><p>Proteolysis reactions ranged in volume from 20 μL to 20 mL and contained final concentrations of 250 mM Tris-HCl, pH 7.5, 2 mg/mL of peptide, and 0.2–1 μg/mL of protease. After approximately 1 h incubation at room temperature, the reactions were desalted by ZipTip, as per the manufacturer's instruction (EMD Millipore). The content of the ZipTip was directly eluted onto a Bruker MTP 384 polished steel target plate with 1 μL of sinapic acid matrix. The sample was analyzed by MALDI-TOF-MS for the complete consumption of full length substrate peptide. The remainder of the peptide in the proteolysis reaction was desalted by SPE.</p><p>The pellet was removed from cultures obtained after the co-expression of FlvM2, FlvA2.g and FlvT by centrifugation at 11,900×g. The supernatant was then decanted and subjected to ZipTip and MALDI-TOF-MS analysis as stated above.</p><p>Analytical scale RP-HPLC was completed using an Agilient 1260 Infinity equipped with a Phenomenex Luna column (10 μm, C18(2), 100 Å, 250×4.6 mm) and managed using Agilent Instrument 1 (Online) software. Semi-preparative scale RP-HPLC was performed using a Shimadzu Prominence equipped with a Phenomenex Luna column (10 μm, C18(2), 100 Å, 250×10.0 mm), managed using the Shimadzu program LC Real Time Analysis. Peptide hydrolysis, Lan derivatization, and GC/MS was adapted from a previous report (Ross et al., 2010). A Waters Synapt ESI-QTOF coupled to an Acquity ultra high performance liquid chromatography (UPLC) system was used for tandem mass spectrometry analysis.</p><p>The liquid media used to culture R. flavefaciens FD-1 consisted of 4.0 g of cellobiose, 2.0 g of Bacto-tryptone, 50 mL mineral solution 1, 50 mL mineral solution 2, and 10 mL volatile fatty acid solution in 820 mL of distilled water (see Supplemental Experimental Procedures for solution compositions). After autoclaving but prior to inoculation, 50 mL of an 8% sodium carbonate solution and 20 mL of a 1.25% cysteine-sulfide solution was added.</p><p>Culturing procedures were performed using anaerobic technique and as previously described (Berg Miller et al., 2009). M. luteus was maintained as 40% glycerol stocks at −80 °C. The glycerol stocks were revived by first streaking onto solid media and incubating at 37 °C for 16–20 h. A single colony of the organism was transferred into 5 mL of liquid media and incubated at 37 °C, 200–220 rpm for 12–16 h. Agar diffusion assay plates were prepared by seeding 20 mL of molten solid media (~42 °C) with 50 μL of overnight liquid culture. Following thorough mixing, the seeded solution was poured into a petri dish and allowed to solidify at room temperature. Samples to be assayed were dissolved in DMSO and 1 μL aliquots were spotted directly onto the solidified media. When two peptide solutions were combined, only 0.5 μL aliquots of each solution were spotted in a single area. The Δ2−Flvα.a peptide was assayed at a concentration of 1 mM while the FlvA2.x peptides were assayed at a concentration of 0.5 mM. Nisin was assayed at a concentration of 100 μM and 0.2 μL of the solution was spotted on the plate. Once all spots were dry, the petri dish was incubated at 37 °C for 16–20 h prior to visualization using a Bio-Rad Molecular Imager XR+ Gel Doc Imaging System and the accompanying Quantity One 4.6.9 program.</p>

PubMed Author Manuscript

“Broken-hearted” carbon bowl <i>via</i> electron shuttle reaction: energetics and electron coupling

Unprecedented one-step C]C bond cleavage leading to opening of the buckybowl (p-bowl), that could provide access to carbon-rich structures with previously inaccessible topologies, is reported; highlighting the possibility to implement drastically different synthetic routes to p-bowls in contrast to conventional ones applied for polycyclic aromatic hydrocarbons. Through theoretical modeling, we evaluated the mechanistic pathways feasible for p-bowl planarization and factors that could affect such a transformation including strain and released energies. Through employment of Marcus theory, optical spectroscopy, and crystallographic analysis, we estimated the possibility of charge transfer and electron coupling between "open" corannulene and a strong electron acceptor such as 7,7,8,8tetracyanoquinodimethane. Alternative to a one-pot solid-state corannulene "unzipping" route, we reported a nine-step solution-based approach for preparation of novel planar "open" corannulene-based derivatives in which electronic structures and photophysical profiles were estimated through the energies and isosurfaces of the frontier natural transition orbitals.

“broken-hearted”_carbon_bowl_<i>via</i>_electron_shuttle_reaction:_energetics_and_electron_coupling

Introduction<!>Results and discussion<!>Conclusions

<p>Unzipping nanotubes, 1-5 nanosheets, 6,7 buckyballs, [8][9][10][11] or annulenes [12][13][14][15][16] is driven by the renewed interest in fundamental understanding and practical access to novel structural transformations, 17,18 leading to materials with unique optical and electronic proles. For instance, cutting and unravelling of nanotubes resulted in nanoribbons having electronic properties that can be varied as a function of their width, and therefore, applied in a variety of electronic devices including eld-effect transistors, light-emitting diodes, and transparent conductive electrodes. 19,20 In addition, hydrogenation of graphene nanoribbons led to enhanced photoluminescent properties that could pave the way for the development of optically active graphene nanoribbon-based devices. 21 However, promotion of selective C]C bond cleavage in graphitic materials is challenging, [22][23][24] and although there have been examples of structural changes due to periphery modications of buckybowls (p-bowls), [25][26][27][28] ring expansion, 22 or opening of the strained pbowl, 29,30 these accounts are overall very limited. Pursuing the route of C-C bond activation in curved p-bowl-containing systems is advantageous as this could lead to addressing challenges such as selective sphere opening for preparation of endohedral fullerenes, shortening carbon nanotubes (CNTs), guest integration within the CNT body, as well as access to a class of materials that has not been prepared through "wetchemistry, conventional" routes. One strategy to facilitate C-C bond activation is to employ, for instance, strain energy as a variable, release of which could energetically promote such transformations. 31 Indeed, as presented in this report, release of strain energy can be the driving force for planarization of the naturally curved buckybowl surface (e.g., C 20 H 10 , corannulene), since there is no direct route to cleave a C]C bond, except through uncontrollable ash vacuum pyrolysis 22,27,29 or addition of a directing group (and a catalyst). 32 Although there are numerous reports of catalytic hydrocracking of planar polycyclic aromatic hydrocarbons (PAHs) i.e., increasing the ratio of hydrogen-to-carbon, [33][34][35][36] there are very few accounts on C-C bond cleavage following the hydrogenation step. 37 The literature precedent for C-C bond scission primarily relies on the assistance of transition metal catalysts, high hydrogen pressure, elevated temperatures, or a combination of all three parameters. [38][39][40][41][42][43][44] Therefore, unexpected C-C bond cleavage (discovered from photophysical studies of alignment of electron donor (corannulene) and acceptor (7,7,8,8-tetracyanoquinodimethane, TCNQ) in the solid state) reported herein led us to probe mechanistic pathways to determine the feasibility for p-bowl planarization and factors that could affect such a transformation including strain energy (E s ) and released energy (E 0 , Scheme 1, see more details in the ESI †). The electron coupling and charge transfer (CT) rates between "open" corannulene (or parent corannulene) and TCNQ were evaluated by applying Marcus theory. In addition to the solid-state reaction, we also offer more "conventional" solution-based nine-step synthetic routes for the preparation of novel "open" corannulene analogs. In the reported ndings, we also discuss the electronic structure and photophysical proles of the synthesized "open" analogs through estimation of their energies and isosurfaces of the frontier highest occupied and lowest unoccupied natural transition orbitals (HONTO and LUNTO).</p><!><p>The reductive C]C bond cleavage and consecutive corannulene planarization to form 5,6-dimethyl-benzo[ghi]uoranthene (planar corannulene analog (P-C 20 H 14 ), Scheme 1) was achieved through a one-pot solid-state reaction, in which corannulene (15 mg, 0.060 mmol), TCNQ (an electron shuttle; 14 mg, 0.068 mmol), and zinc powder (a reducing agent; 50 mg, 0.76 mmol) were ground together (further experimental details can be found in the ESI †). Aer that, the reaction mixture was placed in a glass tube, a drop of hydrochloric acid (proton source) was added, and the glass tube was ame-sealed under dynamic vacuum (4 Â 10 À5 mbar). Heating the reaction mixture at 200 C for six days resulted in the formation of dark brown needle-shaped crystals suitable for single-crystal X-ray diffraction analysis (Scheme 1). As shown in Scheme 1, such treatment resulted in planarization of the corannulene bowl through partial hydrogenation and formation of P-C 20 H 14 . X-ray crystallographic studies of (C 20 H 10 )$(P-C 20 H 14 )$(TCNQ) (1) cocrystals revealed that the packing consists of alternating columns of TCNQ and P-C 20 H 14 along the c-axis direction (Fig. S1 and S2 †). Furthermore, neither mass spectrometry nor spectroscopic studies identied the presence of any other partially hydrogenated products (Fig. S1-S3 †).</p><p>To gain insight into a plausible mechanism of such p-bowl opening during the one-pot solid-state synthesis (Scheme 1), we initially tested the hypothesis of whether all components of the reaction mixture were essential to perform the solid-state C]C bond cleavage. Our results illustrated that the absence of one of the components of the reaction mixture resulted in either no transformation or formation of (corannulene) 2 $(TCNQ) cocrystals, previously reported in the literature (CCDC 1037414) 45 and also detected in our studies (Fig. S4 †). Utilization of a different redox mediator rather than TCNQ (e.g., methyl viologen) did not lead to corannulene opening despite previous reports in which TCNQ and methyl viologen have both been used as electron shuttles in various biological applications. [46][47][48][49] Variation of synthetic conditions, for instance, replacement of the zinc powder with sodium dithionite 50 as a reducing agent did not lead to hydrogenated products (see ESI † for more details).</p><p>Utilization of more conventional solution-based routes through heating the same reagents (C 20 H 10 /TCNQ/Zn/HCl) in a series of organic solvents was also attempted. We varied the reaction media starting with the solvents possessing low boiling points (e.g., dichloromethane or methanol), transitioning to dichloromethane/water or methanol/water mixtures, and nally attempting heating in the higher boiling glycerol (b.p. ¼ 290 C) or ethylene glycol (b.p. ¼ 197 C) to more closely match the reaction temperature (200 C) of the solid-state synthesis. In all reactions, no evidence of P-C 20 H 14 was detected according to the 1 H nuclear magnetic resonance (NMR) spectroscopic or mass spectrometry analysis. Notably, the reported hydrogenation reactions of corannulene typically occurred under relatively harsh conditions (e.g., electron bombardment, alkyllithium reagents, or alkali metals), and even despite them, reactions typically led to hydrogenation of one or two rim C]C bonds without carbon-carbon bond cleavage. [51][52][53][54][55][56][57] Since the developed conditions (C 20 H 10 /TCNQ/Zn/HCl) required the presence of zinc, we also probed the Clemmensen reduction that uses zinc amalgam and concentrated hydrochloric acid. 58 Mass spectrometry and 1 H NMR spectroscopy studies of reaction products detected the presence of only pristine corannulene and did not detect any traces of corannulene hydrogenation. As a logical progression, we surveyed an electrochemical method suitable for arene reductive transformations, 59 but proved unsuccessful. Finally, attempts to electrochemically cleave the C]C bond by bulk electrolysis were performed in anhydrous N,N-dimethylformamide or acetonitrile for up to two days, but were also not successful. In line with these studies, we probed the reaction conditions previously utilized for the ring-opening of other nonplanar structures such as o-carborane. [60][61][62] For that, we used a triosmium carbonyl complex, Os 3 (CO) 10 (NCMe) 2 ; however, no ring-opening of corannulene was observed, while successful ocarborane opening occurred at 150 C in a nonane reux. 62 Further experimental investigations were pursued to rule out aromaticity stabilization as a main factor by performing reactions with signicantly less strained PAHs including pyrene (0.0 kJ mol À1 ) 31 or phenanthrene (0.0 kJ mol À1 ) 63 under experimental conditions similar to those used for the reaction with highly-strained corannulene (101 kJ mol À1 ; 31 see ESI † for more details). As a result, no bond cleavage was detected in any of these systems, and formation of only PAH$TCNQ complexes was observed (e.g., phenanthrene$TCNQ complex (similar to the structure reported in the literature 64 ) or pyrene$TCNQ complex, Fig. S17; see more details in the ESI †). Thus, this experimental evidence suggests that one of the driving forces for the observed solid-state reaction could potentially be a release of energy through buckybowl planarization (Fig. 2a). To prove this hypothesis, we estimated released energy, E 0 , for PAHs and carbon p-bowls as shown in Scheme 1. For instance, E 0 for corannulene was calculated to be 202.0 kJ mol À1 (Fig. S9, B3LYP/6-31+G*, see the ESI for more details †). In contrast, E 0 calculated for the PAHs in Scheme 1 was found to be less than 135.7 kJ mol À1 (Table S3 †). Therefore, corannulene opening is much more energetically favorable in comparison with the PAHs shown in Scheme 1. A similar statement is also valid for a family of extended p-bowls for which the estimated E 0 was even higher than that of corannulene (Scheme S3 †).</p><p>The estimated enthalpy of the reaction (eqn (1)),</p><p>was found to be À179.5 kJ mol À1 (À239 and 59.5 kJ mol À1 for only the electronic and the ZPE-corrected electronic energies, respectively, using density functional theory (DFT, Table S2, see the ESI for more details †)). Thus, a combination of two parameters, strain energy (E s ) and released energy (E 0 ), highlights the unique nature of buckybowls in comparison with the considered PAHs (Scheme 1). As a next step, we took a closer look at a possible mechanism for p-bowl hydrogenation and C-C bond cleavage. On the basis of our theoretical calculations, experimental results, and literature reports, we hypothesized that the transformation of C 20 H 10 to P-C 20 H 14 occurs in a series of reactions that is rst initiated by a sequence of electron and proton transfers in which hydrochloric acid acts as the proton source (Fig. 1). Moreover, probing the strength of the C-C bond revealed a signicantly weaker bond (115 kJ mol À1 for C 20 H 12 and 9 kJ mol À1 for C 20 H 12 À c, Fig. 1) than a typical C-C bond in RCH 2 -CH 2 R systems, allowing for bond cleavage to occur. 65,66 For instance, if R is a substituent on a pyrene or coronene core then the electronic energy of the C-C bond would be approximately 350 kJ mol À1 (Fig. S14 †) and 302 kJ mol À1 , respectively (Fig. S15 †).</p><p>Comprehensive analysis of photophysical data for the obtained crystals of 1 revealed properties that are uncharacteristic of the individual components i.e., corannulene and TCNQ themselves. Based on photoluminescence and epiuorescence microscopy studies, the obtained crystals of 1 exhibited red emission (l max ¼ 705 nm, l ex ¼ 365 nm) in contrast to their constituents (l max (TCNQ crystal) ¼ undetectable emission and l max (corannulene crystal) ¼ 490 nm, l ex ¼ 365 nm, Fig. S5 †). Furthermore, in contrast to diffuse reectance (DR) proles of pristine corannulene and TCNQ (Fig. S6 †), the appearance of a new red-shied band (550 nm) in the DR prole of 1 was detected. Based on our theoretical calculations using timedependent density functional theory (TDDFT), the new band (550 nm) is characteristic of CT complex formation (see the ESI for more details †) that is in line with a previous report on PAHs and TCNQ co-crystals. 24 In particular, according to our studies using the B3LYP-D3/6-311+G** level of theory (Fig. 2b), both a bathochromic shi and new band appearance could be attributed to CT 66,67 between the HOMOÀ2 and LUMO of a TCNQ/P-C 20 H 14 "stack" (Fig. 2c). To further shed light on the experimental changes of the emission prole, we examined optical excitations of isolated corannulene, P-C 20 H 14 , TCNQ, and the relevant dimers through the TDDFT calculations based on the B3LYP-D3/6-311+G** method. The considered TCNQ/P-C 20 H 14 "stack" is the only species with excitation energies of appreciable strength around 690 nm (1.8 eV; Fig. S7 †), which is in agreement with the experimentally observed red emission at l max ¼ 705 nm (Fig. S5 †). The lowest excitations for TCNQ, pbowl, and P-C 20 H 14 are 413, 288, and 344 nm (3.0, 4.3, and 3.6 eV), respectively (Fig. S7 †).</p><p>To further probe the idea that CT is more effective in an exclusively planar TCNQ/P-C 20 H 14 "stack" rather than in a TCNQ/C 20 H 10 "stack", that encounters steric hindrance from the curved surface of the p-bowl, 45 we employed Marcus theory 68 to compare the electron coupling (that is proportional to CT rate) between TCNQ/C 20 H 10 and TCNQ/P-C 20 H 14 using eqn (2):</p><p>where k ¼ charge transfer rate, V c ¼ electron coupling, l ¼ reorganization energy of the system, and DG 0 ¼ energy difference between the initial and nal states (see the ESI for more details †). According to the Marcus theory model, TCNQ/P-C 20 H 14 could result in ca. 128-fold increase in electron coupling compared to TCNQ/C 20 H 10 (Fig. S8 and Table S5, see the ESI for more details †). Since electron coupling is related to the electron transfer rate, we can surmise that there is likely an increased electron transfer rate as well. 66 The charge on the TCNQ molecule was evaluated by applying the Kistenmacher relationship (i.e., correlation between TCNQ intramolecular bond distances and charge on TCNQ) 69 using the crystallographic data of 1 and (corannulene) 2 $TCNQ co-crystals. 45 In the case of 1, the charge on TCNQ was estimated to be À0.84 and for (corannulene) 2 -$TCNQ co-crystals was found to be À0.20, suggesting more effective CT can occur in 1 between "open" corannulene (P-C 20 H 14 ) and TCNQ. We calculated electronic transitions corresponding to the ground state, rst and second excited singlet states of P-C 20 H 14 in THF (Fig. 3 and S29 †). Delocalization of excited energy levels in P-C 20 H 14 was slightly enhanced, leading to optical transitions of the rst and second excited states with values of 399 and 409 nm, respectively (Fig. S29 †).</p><p>As an alternative scalable approach to access a family of "open" corannulene-containing derivatives, we report a solution-phase route (see ESI †). 70 Despite a number of required steps (Schemes S1 and S2 †), in comparison with the one-step solid-state synthesis, the "solution" approach has some advantages since it does not rely on selection of the specic substrate/electron shuttle/reducing agent system and also provides a scalable route for the synthesis of a library of new planar corannulene-type analogs. The synthetic details for preparation of 5-methylbenzo[ghi]uoranthene (C 19 H 12 , X, Fig. 4a) and 5-ethyl-6-methylbenzo[ghi]uoranthene (C 21 H 16 , X 0 , Fig. 4b), using this approach, are provided in ESI. † Sublimation of X (Scheme S1 †) allowed for the formation of single crystals of X suitable for X-ray diffraction (Fig. S24 †). The structure of X 0 was conrmed using 1 H and 13 C NMR spectroscopy and mass spectrometry (Fig. S21 †). As in the case of solid-state "open" P-C 20 H 14 , both X and X 0 structures possess a planar geometry (Fig. S24 and S25 †). The emission studies of the prepared X and X 0 compounds revealed a red-shied emission (l max ¼ 548 nm and 573 nm, l ex ¼ 365 nm, respectively, Fig. S26 †) in contrast to pristine corannulene (l max ¼ 490 nm, l ex ¼ 365 nm, Fig. S5 †). The emission maxima of X and X 0 in THF was found to be 479 nm and 502 nm, respectively, (l ex ¼ 365 nm) and is hypsochromically shied compared to the solid-state 548 and 573 nm-centered emission, respectively (l ex ¼ 365 nm, Fig. 4a, b, S27, and S28 †). In a similar vein to TDDFT calculations of P-C 20 H 14 , we determined the optical transitions corresponding to the ground state, rst and second excited singlet states of X and X 0 (Fig. S27 and S28 †). While the electronic transition of the rst excited singlet state for both X and X 0 did not differ from the optical transitions corresponding to the ground state (351 nm for X and 361 nm for X 0 ), optical transitions for the second excited singlet states were determined to be 395 nm and 412 nm for X and X 0 , respectively. Electronic transitions corresponding to the second excited singlet state can be associated with the emission proles that are similar to the experimental data (see the ESI for more details).</p><!><p>To summarize, we report the rst example of a unique one-step C]C bond cleavage in the traditionally very robust p-bowl occurring via an electron shuttle reaction. Such ring opening is unprecedented in the literature and has not been observed for pristine p-bowls (e.g., corannulene) to date (with the exception of uncontrollable brute force vacuum pyrolysis 22 ). PAH hydrogenation has been previously observed under harsh experimental conditions (e.g., high hydrogen pressure or extreme temperatures above 1000 C), [38][39][40][41][42][43][44] therefore the formation of P-C 20 H 14 in a one-pot synthesis under relatively mild conditions is an unexpected and remarkable result. Through employment of Marcus theory, optical spectroscopy, and crystallographic analysis, we estimated the electron coupling between "open" corannulene and a strong electron acceptor, TCNQ. A solutionphase route was employed for preparation of two novel "open" corannulene-based derivatives with the corresponding spectroscopic analysis of their properties experimentally and theoretically. Furthermore, through a combination of theoretical modeling with experimental results, mechanistic studies were undertaken to shed light on possible factors (such as strain energy) that could act as a driving force for the observed p-bowl opening. Our studies highlight the possibility to implement novel synthetic routes for p-bowl transformations, that are drastically different from the conventional approaches toward derivatization of traditional PAHs. Thus, the presented solidstate, solution, and theoretical methodology are the rst steps Fig. 4 (a) (Top left) Single-crystal X-ray structure of X. (Bottom left) Optical transition strengths computed at the ground state optimal geometry for X in THF (blue) and at the second singlet excited state optimal geometry for X in THF (red). (Right) Energies and isosurfaces of the HONTO and LUNTO of X in the ground and the second singlet excited states. S 0 and S 2 are the ground and excited states for X of the ground state. S 0 and S 2 0 are the ground and excited state intermediates for the minimum energy geometry of the second excited singlet state. The black solid and wavy arrows indicate absorption (S 0 / S 2 ) or emission (S 2 0 / S 0 0 ) and vibrational relaxation (S 2 / S 2 0 and S 2 0 / S 2 ), respectively. The theory level is TDDFT/RPA based on the B3LYP-D3/ 6-31+G* method. (b) (Top left) Geometrically optimized structure of X 0 based on B3LYP-D3/6-31+G* level of theory. (Bottom left) Optical transition strengths computed at the ground state optimal geometry for X 0 in THF (blue) and at the second singlet excited state optimal geometry for X 0 in THF (red). (Right) Energies and isosurfaces of HONTO and LUNTO of X 0 in the ground and the second singlet excited states. S 0 and S 2 are the ground and excited states for X 0 of the ground state. S 0 ' and S 2 0 are the ground and excited state intermediates for the minimum energy geometry of the second excited singlet state. The black solid and wavy arrows indicate absorption (S 0 / S 2 ) or emission (S 2 0 / S 0 0 ) and vibrational relaxation (S 2 / S 2 0 and S 2 0 / S 2 ), respectively. The theory level is TDDFT/RPA based on the B3LYP-D3/ 6-31+G* method.</p><p>toward understanding possible avenues to prepare barely accessible structures by "unlocking" the corannulene core and application of the latter for molecular electronic development.</p>

Royal Society of Chemistry (RSC)

Biosynthesis of (-)-(1S,2R)-allocoronamic acyl thioester by an FeII-dependent halogenase and a cyclopropane-forming flavoprotein

The biosynthetic gene cluster for the kutzneride family of hexapeptidolactones includes the four-gene cassette ktzABCD postulated to generate a nonproteinogenic amino acid. Encoded by this cassette are the nonheme FeII-dependent halogenase KtzD and the acyl-CoA dehydrogenase-like flavoprotein KtzA, proposed to work in conjunction with adenylating protein KtzB and carrier protein KtzC. In the present work, we report the in vitro reconstitution of this four-protein system and identify the final product as (1S, 2R)-allocoronamic acid bound in thioester linkage to KtzC. Further analysis of KtzD and KtzA support a biosynthetic pathway that involves KtzD-mediated generation of a \xce\xb3-chloroisoleucyl intermediate which is cyclized to the final product by KtzA without redox participation of the bound flavin cofactor. This work introduces a new monomer for potential incorporation into nonribosomal peptides and validates the unique strategy for its biosynthesis.

biosynthesis_of_(-)-(1s,2r)-allocoronamic_acyl_thioester_by_an_feii-dependent_halogenase_and_a_cyclo

<p>Nonribosomal peptides constitute a diverse class of secondary metabolites with important biological activities.1 In addition to the twenty proteinogenic amino acids, nonribosomal peptide synthetases make use of a large number of modified amino acid and hydroxy acid building blocks.2 Understanding the chemical pathways and mechanisms used in the biosyntheses of these monomers is a first step towards expanding the pool of new molecules available through metabolic engineering.</p><p>Towards this end, we recently used halogenase-directed probes to clone the gene cluster3 responsible for the biosynthesis of a series of hexadepsipeptides, kutznerides, isolated from the soil actinomycete kutzneria4,5 (Figure S1). These molecules contain a number of unusual monomers whose biosyntheses have not been described. Bioinformatic analysis of the gene cluster identified three potential halogenases. KtzQ and KtzR are shown in separate work to be flavin-dependent tryptophan halogenases which can generate the 6,7-dichlorotryptophan residue postulated in kutzneride biosynthesis.6 KtzD was predicted to be an FeII-dependent halogenase capable of chlorinating an unactivated carbon center.</p><p>The four-protein cassette KtzABCD could potentially function to generate a novel amino acid via a cryptic chlorination pathway. KtzC and KtzB were assigned roles as carrier protein and adenylation protein, respectively. KtzA is a flavoprotein homologous to members of the acyl-CoA dehydrogenase (ACAD) family of enzymes. We initially hypothesized that these four proteins produce the nonproteinogenic amino acid 2-(1-methylcyclopropyl)glycine (MecPG),3,7 a component of all kutzernide molecules reported to date. Herein, we report the biochemical characterization of KtzABCD and demonstrate that they instead function to produce an alternative cyclopropyl-containing amino acid, (1S, 2R)-allocoronamic acid (alloCMA), tethered to KtzC. To our knowledge, this is the first identification of alloCMA in a natural source.</p><p>To study the role of the KtzABCD cassette, the individual genes were cloned into E. coli expression vectors as His-tag fusions. Overexpression and Ni-NTA purification provided soluble protein for each construct. The carrier protein KtzC was isolated in apo form and phosphopantetheinylated via incubation with coenzyme A and Sfp. KtzA was isolated with FAD bound at 50% occupancy. After initial purification of the halogenase KtzD, the His tag was proteolytically cleaved and the enzyme purified by gel filtration and reconstituted anaerobically with FeII and α-ketoglutarate to obtain the active form of the protein.</p><p>In accordance with its previously described adenylating activity,3 KtzB efficiently activates and loads both l-Ile and l-allo-Ile to holo-KtzC as the phosphopantetheinyl thioesters. The protein-tethered amino acids can be identified by enzymatic hydrolysis with the Type II thioesterase TycF8 and conversion to readily detectable isoindole derivatives using established procedures.9 Incubation of l-Ile-S-KtzC with KtzD for 1 hour results in complete consumption of starting material and production of a new aminoacyl thioester product (Figure 1A). MS analysis on the released isoindole derivative shows this product results from a single chlorination event. Treatment of l-allo-Ile-S-KtzC with KtzD does not lead to the formation of a new product (Figure S3), illustrating the substrate specificity of the halogenase.</p><p>Exposure of the chlorinated intermediate to KtzA again leads to complete conversion to a new product. MS analysis shows the released amino acid product to have a m/z ratio consistent with dehydrochlorination of the intermediate and net two-electron oxidation of the starting isoleucine. Supplementing the reaction with excess FAD had no effect on product formation (data not shown). To confirm the structure of the final product, the isoindole derivative was purified by HPLC and analyzed by 1H- and COSY-NMR. Additionally, 13C6-Ile was used as the starting material, and the isotope-enriched product analyzed by HMQC and 1H-NMR spectroscopy. The spectral data support the atom connectivity of alloCMA (Figures S7/S8, Table S1).</p><p>The stereochemistry of the product was deduced from two observations. First, the observed product fails to co-elute with an authentic sample of coronamic acid under the typical HPLC conditions (Figure S9). Second, the γ-CH2 protons show 1H-NMR resonances at 1.47 and -0.14 ppm. The unusual upfield resonance is best explained by strong shielding of the proton as a result of its location directly above the aromatic ring of the isoindole group. For this to occur on the rigid cyclopropyl skeleton, the ethyl side chain and isoindole group must be in a cis arrangement. Based on the observed alloCMA structure, we propose a biosynthetic scheme which involves γ-CH3 chlorination of l-Ile-S-KtzC followed by KtzA-mediated α,γ cyclization to alloCMA-S-KtzC via an intermediate Cα-carbanion (Figure 1B).</p><p>To confirm the regiochemistry of γ-chlorination by KtzD, we synthesized substrates deuterated at the γ-CH3 position (Ile-d3) or at the β-CH,γ-CH2, and δ-CH3 positions (Ile-d6). Treatment of Ile-d6-S-KtzC with KtzD produces a chlorinated product that retains all deuterium labels. Identical treatment of Ile-d3-S-KtzC produces a chlorinated product which retains only two deuterium labels (Figure 2). Additionally, analysis of the reaction at partial conversion shows that the rate of chlorination of Ile-d3 is retarded relative to Ile and Ile-d6 (Figure S4). This kinetic isotope effect for H/D abstraction is consistent with the isotope effect observed in stopped flow studies of the related halogenase CytC3.10</p><p>Cyclization of 2 to alloCMA-S-KtzC is mediated by the ACAD-like flavoprotein KtzA. ACADs catalyze dehydrogenation of acyl/aminoacyl CoA thioesters by abstraction of the acidic α- proton and transfer of a β-hydride to the FAD cofactor to yield the α,β-unsaturated thioester and FADH2.11 In contrast, KtzA need only abstract the alpha proton of 2 to generate a stabilized thioester enolate which can collapse by intramolecular attack at Cγ to form the cyclopropane. Loss of Cl- constitutes the two-electron oxidation of substrate and avoids formation of the FADH2 oxidation state.12 To verify that KtzA can catalyze reversible α-proton abstraction as the initial event in cyclization, unlabeled Ile-S-KtzC was incubated with KtzA in the presence of D2O. Recovery of the amino acid and analysis by HPLC and MS showed incorporation of a single deuterium atom (Figures S5/S6). The yield is similar to that for untreated Ile-S-KtzC indicating that there is no significant loss of amino acid through hydrolysis of a dehydrogenated intermediate. Attempts to generate apo-KtzA by treatment of Ni-bound protein with 2M KBr and 2M urea gave negligible yields of protein; thus it is possible the flavin cofactor is needed to maintain the tertiary structure of KtzA.</p><p>The rerouting of a carbanion intermediate in a flavoprotein active site to intramolecular elimination rather than two-electron transfer to the FAD coenzyme has previously been observed for the flavoprotein D-alanine oxidase when presented with a non-natural substrate.13 In that case, β-chloroalanine can be partitioned between αβ-dehydrochlorination to dehydroalanine and Cα oxidation to β-chloro-iminopyruvate. In KtzA catalysis, the net removal of Hα and Clγ appears to be the physiological reaction. We have previously reported a related four protein pathway in coronamic acid biosynthesis where l-allo-Ile is comparably converted to (1S,2S)-coronamic acid via the same cryptic γ-chlorination strategy, but in that case the cyclization catalyst is the Zn-dependent enzyme CmaC.9,14 KtzA and CmaC thus represent two distinct enzymatic strategies for generation of the thioester enolate needed to initiate cyclopropane formation.</p><p>In conclusion, we have characterized the activities of the KtzABCD proteins and identified the novel metabolite alloCMA as the protein-tethered product of these enzymes. Deuterium-labeling studies confirm the regioselective chlorination of l-Ile-S-KtzC at the γ-CH3 position by KtzD and support the unusual role of the flavoprotein KtzA as the catalytic base which induces cyclization without redox participation of bound FAD. This cassette provides a biosynthetic route to a novel amino acid monomer that is a stereochemical complement to the previously described coronamic acid. At the same time, the results suggest future directions for research on kutzneria. Identifying additional nonribosomal peptides which may contain alloCMA and locating the gene cassette which provides for MecPG biosynthesis are now key avenues for metabolomic and genomic approaches.</p>

PubMed Author Manuscript

Synthesis and SAR study of modulators inhibiting tRXR\xce\xb1-dependent AKT activation

RXR\xce\xb1 represents an intriguing and unique target for pharmacologic interventions. We recently showed that Sulindac and a designed analog could bind to RXR\xce\xb1 and modulate its biological activity, including inhibition of the interaction of an N-terminally truncated RXR\xce\xb1 (tRXR\xce\xb1) with the p85\xce\xb1 regulatory subunit of phosphatidylinositol-3-OH kinase (PI3K). Here we report the synthesis, testing and SAR of a series of novel analogs of Sulindac as potential modulators for inhibiting tRXR\xce\xb1-dependent AKT activation. A new compound 30 was identified to have improved biological activity.

synthesis_and_sar_study_of_modulators_inhibiting_trxr\xce\xb1-dependent_akt_activation

1. Introduction<!>2. Results and discussion<!>3. Chemistry<!>4. Conclusion<!>5.1.1. General methods<!>5.1.2. General Procedure A: the synthesis of acrylic acid from aromatic aldehyde<!>5.1.3. 3-(4-Fluorophenyl)-2-methylacrylic acid (32)<!>5.1.4. General Procedure B: the synthesis of propanoic acid derivative from acrylic acid by Pd/C-catalyzed reduction<!>5.1.5. 3-(4-Fluorophenyl)-2-methylpropanoic acid (33)<!>5.1.6. General Procedure C: the synthesis of indenone from propanoic acid derivative by F-C acylation<!>5.1.7. 6-Fluoro-2-methyl-2,3-dihydroinden-1-one (34)<!>5.1.8. General Procedure D: the synthesis of inden-3-yl acetate from indenone<!>5.1.9. Ethyl 2-(5-fluoro-2-methyl-1H-inden-3-yl)acetate (35)<!>5.1.10. General Procedure E: the synthesis of indene derivative from appropriate inden-3-yl acetate<!>5.1.11. (Z)-2-(1-Benzylidene-5-fluoro-2-methyl-1H-inden-3-yl) acetic acid (3)<!>5.1.12. (Z)-2-(5-Fluoro-2-methyl-1-(4-trifluoromethylbenzylidene)-1H-inden-3-yl)acetic acid (4)<!>5.1.13. (Z)-2-(1-(4-tert-Butylbenzylidene)-5-fluoro-2-methyl-1H-inden-3-yl)acetic acid (5)<!>5.1.14. (Z)-2-(5-Fluoro-2-methyl-1-(4-(pyridin-2-yl)benzylidene)-1H-inden-3-yl)acetic acid (6)<!>5.1.15. (Z)-2-(5-Fluoro-1-(4-methoxybenzylidene)- 2-methyl-1H-inden-3-yl)acetic acid (7)<!>5.1.16. (Z)-2-(1-(4-Ethoxybenzylidene)-5-fluoro-2-methyl-1H-inden-3-yl)acetic acid (8)<!>5.1.17. (Z)-2-(1-(4-Cyanobenzylidene)-5-fluoro-2-methyl-1H-inden-3-yl)acetic acid (9)<!>5.1.18. (Z)-2-(1-(4-(Dimethylamino)benzylidene)-5-fluoro-2-methyl-1H-inden-3-yl)acetic acid (10)<!>5.1.19. (Z)-2-(5-Fluoro-2-methyl-1-(3-trifluoromethylbenzylidene)-1H-inden-3-yl)acetic acid (11)<!>5.1.20. (Z)-2-(5-Fluoro-1-(3-methoxybenzylidene)- 2-methyl-1H-inden-3-yl)acetic acid (12)<!>5.1.21. (Z)-2-(1-(3-Cyanobenzylidene)-5-fluoro-2-methyl-1H-inden-3-yl)acetic acid (13)<!>5.1.22. (Z)-2-(1-(4-Acetamidobenzylidene)-5-fluoro-2-methyl-1H-inden-3-yl)acetic acid (14)<!>5.1.23. Methyl 3-(5-fluoro-2-methyl-1H-inden-3-yl)propanoate (38)<!>5.1.24. (Z)-3-(5-Fluoro-1-(4-isopropylbenzylidene)-2-methyl-1H-inden-3-yl)propanoic acid (15)<!>5.1.25. 3-(5-Fluoro-2-methyl-1H-inden-3-yl)propanenitrile (39)<!>5.1.26. (Z)-3-(5-Fluoro-1-(4-isopropylbenzylidene)-2-methyl-1H-inden-3-yl)propanenitrile (16)<!>5.1.27. 2-(5-Fluoro-1-(4-isopropylbenzylidene)-2-methyl-1H-inden-3-yl)acetamide (17 and 18)<!>5.1.28. 2-(5-Fluoro-1-(4-isopropylbenzylidene)-2-methyl-1H-inden-3-yl)-N-methylacetamide (19 and 20)<!>5.1.29. 2-(4-Fluorlbenzylidene)butanoic acid (40)<!>5.1.30. 2-(4-Fluorlbenzyl)butanoic acid (41)<!>5.1.31. 2-Ethyl-6-fluoro-2,3-dihydro-1H-inden-1-one (42)<!>5.1.32. Ethyl-2-(2-ethyl-5-fluoro-1H-inden-3-yl)acetate (43)<!>5.1.33. (Z)-2-(2-Ethyl-5-fluoro-1-(4-isopropylbenzylidene)-1H-inden-3-yl)acetic acid (21)<!>5.1.34. 6-Fluoro-2-isobutyl-2,3-dihydro-1H-inden-1-one (44)<!>5.1.35. Ethyl-2-(5-fluoro-2-isobutyl-1H-inden-3-yl) acetate (45)<!>5.1.36. (Z)-2-(5-Fluoro-2-isobutyl-1-(4-isopropylbenzylidene)-1H-inden-3-yl) acetic acid (22)<!>5.1.37. 12-Methyl-3-phenylacrylic acid (47a)<!>5.1.38. 2-Benzylpropanoic acid (48a)<!>6.1.39. 2,3-Dihydro-2-methylinden-1-one (49a)<!>5.1.40. Ethyl 2-(2-methyl-1H-inden-3-yl) acetate (50a)<!>5.1.41. (Z)-2-(1-(4-Isopropylbenzylidene)-2-methyl-1H-inden-3-yl) acetic acid (23)<!>5.1.42. 2-Methyl-3-p-tolylacrylic acid (47b)<!>5.1.43. 2-Methyl-3-p-tolylpropanoic acid (48b)<!>5.1.44. 2,6-Dimethyl-2,3-dihydro-1H-inden-1-one (49b)<!>5.1.45. Ethyl 2-(2,5-dimethyl-1H-inden-3-yl) acetate (50b)<!>5.1.46. (Z)-2-(1-(4-iso-Propylbenzylidene)-2,5-dimethyl-1H-inden-3-yl) acetic acid (24)<!>5.1.47. 3-(4-Methoxyphenyl)-2-methylacrylic acid (47c)<!>5.1.48. 3-(4-Methoxyphenyl)-2-methylpropanoic acid (48c)<!>5.1.49. 6-Methoxy-2-methyl-2,3-dihydro-1H-inden-1-one (49c)<!>5.1.50. Ethyl 2-(5-methoxy-2-methyl-1H-inden-3-yl) acetate (50c)<!>5.1.51. (Z)-2-(1-(4-Isopropylbenzylidene)-5-methyl-2-methyl-1H-inden-3-yl) acetic acid (25)<!>5.1.52. 3-(4-Ethylphenyl)-2-methylacrylic acid (47d)<!>5.1.53. 3-(4-Ethylphenyl)-2-methylpropanoic acid (48d)<!>5.1.54. 6-Ethyl-2-methyl-2,3-dihydro-1H-inden-1-one (49d)<!>5.1.55. Ethyl 2-(5-ethyl-2-methyl-1H-inden-3-yl) acetate (50d)<!>5.1.56. (Z)-2-(5-Ethyl-1-(4-isopropylbenzylidene)-2-methyl-1H-inden-3-yl) acetic acid (26)<!>5.1.57. 3-(4-Ethoxyphenyl)-2-methylacrylic acid (47e)<!>5.1.58. 3-(4-Ethoxyphenyl)-2-methylpropanoic acid (48e)<!>5.1.59. 6-Ethoxy-2-methyl-2,3-dihydro-1H-inden-1-one (49e)<!>5.1.60. Ethyl 2-(5-ethoxy-2-methyl-1H-inden-3-yl) acetate (50e)<!>5.1.61. (Z)-2-(5-Ethoxy-(4-isopropylbenzylidene)-2-methyl-1H-inden-3-yl) acetic acid (27)<!>5.1.62. 3-(4-iso-Propylphenyl)-2-methylacrylic acid (47f)<!>5.1.63. 3-(4-iso-Propylphenyl)-2-methylpanoic acid (48f)<!>5.1.64. 6-iso-Propyl-2-methyl-2,3-dihydro-1H-inden-1-one (49f)<!>5.1.65. Ethyl 2-(5-isopropyl-2-methyl-1H-inden-3-yl) acetate (50f)<!>5.1.66. (Z)-2-(5-iso-Propyl-1-(4-isopropylbenzylidene)-2-methyl-1H-inden-3-yl) acetic acid (28)<!>5.1.67. 3-(4-Chlorophenyl)-2-methylacrylic acid (47g)<!>5.1.68. 3-(4-Chlorophenyl)-2-methylpropanoic acid (48g)<!>5.1.69. 6-Chloro-2-methyl-2,3-dihydro-1H-inden-1-one (49g)<!>5.1.70. Ethyl 2-(5-chloro-2-methyl-1H-inden-3-yl) acetate (50g)<!>5.1.71. (Z)-2-(5-Chloro-1-(4-isopropylbenzylidene)-2-methyl-1H-inden-3-yl) acetic acid (29)<!>5.1.72. Ethyl 2-(5-Fluoro-1H-inden-3-yl)acetate (52)<!>5.1.73. (E)-2-(5-Fluoro-1-(4-isopropylbenzylidene)-1H-inden-3-yl)acetic acid (30)<!>5.2.1. Ligand-binding Competition Assay<!>5.2.2. Cell Culture and Transfection<!>5.2.3. MTT assay<!>5.2.4 Western Blotting<!>5.2.5. Transient transfection and reporter assay<!>5.2.6. RXR\xce\xb1 siRNA and transfection<!>5.3. Ligand Docking

<p>Retinoid X receptor-α (RXRα) is a unique member of the nuclear receptor (NR) superfamily, playing an important role in many biological processes ranging from apoptosis, cell differentiation and growth to lipid metabolism [1–3]. RXRα acts primarily as a ligand-dependent transcription factor through forming homodimer with itself or heterodimer with other members of the NR family. Structurally RXRα shares a modular organization with other nuclear receptors, consisting of three main functional domains: an N-terminal region where the ligand-independent transcriptional activation function (AF-1) is located, a DNA-binding domain and a ligand-binding domain (LBD) [2]. The transcriptional activity is directly mediated by the LBD and thus the LBD has been the most studied domain. The LBD possesses a ligand-binding pocket (LBP) for the binding of small molecule ligands, a transactivation function domain termed AF-2 composed of Helix 12 (H12) of the LBD, a coregulator binding surface, and a dimerization surface. Numerous ligands targeting the LBP have been designed and reported [4, 5]. Natural RXRα ligand 9-cis-Retinoic Acid (9-cis-RA) and synthetic RXR ligands (rexinoids) have been effective in preventing tumorigenesis in animals [6] and RXRα has been a drug target for therapeutic applications, especially in the treatment of cancer [7]. Targretin, a synthetic RXR-selective retinoid, was approved for treating cutaneous T-cell lymphoma [8, 9], and it has also been explored for the treatment of other form of cancer such as lung cancer, breast cancer, and prostate cancer [10–12].</p><p>Sulindac, a nonsteroidal antiinflammatory drug (NSAID) drug, has been investigated as a cancer chemopreventive agent, because of its potent induction of apoptosis and inhibition of cancer cell growth [13–16]. It has been documented that the anti-cancer effect of Sulindac can be mediated through COX-2-independent mechanisms [14, 15, 17]. We recently reported that Sulindac induces apoptosis in several cancer cell lines and primary tumors by binding to an N-terminally-truncated RXRα (tRXRα) [18]. Tumor necrosis factor-α (TNFα) promoted tRXRα interaction with the p85α subunit of phosphatidylinositol-3-OH kinase (PI3K), activating PI3K/AKT signaling. When combined with TNFα, Sulindac inhibited TNFα-induced tRXRα/p85α interaction, leading to activation of the death receptor-mediated apoptotic pathway [18]. Furthermore, we showed, a designed Sulindac analog K-80003 (2) (Fig. 1) exhibits increased affinity to RXRα without COX inhibitory activity, and displays enhanced efficacy in inhibiting tRXRα-dependent AKT activation and tRXRα tumor growth in animals, demonstrating the feasibility of developing a new generation of RXRα-specific molecules for therapeutic application or mechanistic studies of RXRα. Here we present the synthesis, SAR studies and biological evaluation of a series of K-80003 derivatives and the discovery of a new scaffold 30.</p><!><p>Compared to Sulindac (1, Fig. 1), compound 2 displays an increased binding to RXRα and potency in inhibiting tRXRα-dependent AKT activation [18]. 1 and 2 differ in the replacement of the sulfide group in 1 by an isopropyl group in 2 at R1 (Fig. 2). Thus, for the first round of SAR study we investigated the effects of various substituents of R1 (Fig. 2) on the binding affinity to the RXRα LBD. Scheme 1 outlines the synthetic chemistry used for the preparation of this group of compounds (3–14) and the testing results are listed in Table 1. The designed compounds 3–14 provide an opportunity to study the effect of the size of the group and the influence of electron-deficient and electron-rich groups. It seems that the binding capability is sensitive to the size of R1 group at position 4. Compound 3 with R1 = H displayed weaker binding whereas compound 4 with R1= CF3 exhibited slightly weaker binding and when R1= C(CH3)3 the compound 5 showed no binding. Replacing R1 of –CH(CH3)2 in 2 with either electron-donating (compounds 7 and 8, 10, 14) diminished the compounds' binding. Electron-withdrawing group of -CN (compounds 9) also abolished the binding. Similar trend was observed when the substitution was moved from position 4 to 3 (11–13, table 1).</p><p>Our previous molecular docking study showed that the carboxylate group of 2 formed charge-charge interaction with Arg316 in the ligand-binding pocket of RXRα in a similar fashion to the carboxylate group found in other RXRα ligands [18]. This binding model is consistent with the SAR study of R2 group as shown in Table 2. Although extending the carboxylate group by one carbon (15) weakens the binding, replacing the carboxylate group with non-charged groups (16, 17 and 19) resulted in the loss of the binding activity. Compounds 15–17 and 19 were synthesized according to Scheme 2.</p><p>A few substituents at R3 and R4 were also examined and the binding results are outlined in Table 3. It shows that R3 can tolerate bulkier groups. For example, replacing methyl at R3 with ethyl (21) or isobutyl group (22) did not affect the binding dramatically. However, R4 is sensitive to different substituents. Except for the ethoxyl group, replacing floride by other groups including hydrogen, chloride, methyl (23–26, 28, 29) caused steep drop in binding. Compounds 21–29 were prepared according to the procedure outlined in Scheme 3.</p><p>Overall, the analogs synthesized for this SAR study didn't improve the binding activity compared to the original lead 2. Therefore we decided to explore the E-isomer of 2. Compound 30 (Figure 3) was prepared according to Scheme 4 and was found to exhibit slightly tighter binding with an IC50 of 1.6 μM. Compound 30 was then evaluated for the effect on RXRα transactivation activity by employing the Gal4 reporter assay. The LBD of RXRα was cloned as a Gal4 fusion and the resulting Gal4-RXRα/LBD chimera and Gal4 reporter system were used to evaluate the effect of compound 30. Gal4-RXRα/LBD strongly activated the Gal4 reporter in the presence of 9-cis-RA, which was inhibited by BI-1003, a RXRα antagonist [19]. As shown in Fig. 4, treatment of cells with 2 or 30 resulted in inhibition of 9-cis-RA-induced reporter activity in a dose dependent manner. So, Like 2, 30 acts as a RXRα antagonist, however 30 showed stronger antagonism activity.</p><p>30, being an E-isomer of 2, displays a different shape from 2 due to the different orientation of the isopropyl benzene motif. With such a difference in shape, it would be expected intuitively that 30 would not be tolerated in the same pocket where 2 binds. Thus we were intrigued to understand how 2 and 30 bind to the same LBP. Docking study was performed to explore the potential binding modes of 2 and 30. RXRα/antagonist complex structure 3A9E [20] from Protein Data Bank (PDB) was used and Glide docking program [21] from Schrodinger was applied. The docked binding mode of 2 suggested that 2 bound to the LBP of RXRα in a similar mode as previously proposed for Sulindac [18], in which the carboxylate interacts with Arg316 of RXRα and the isopropyl benzene portion of the compound interacts with hydrophobic side chains residing in H3, H5 and H7 (Fig. 5A). For 30, it was found that 30 could be tolerated and docked into the same pocket (Fig. 5B). This is most likely explained by the large size and the hydrophobic nature of the LBP. However, the docked 30 adopts a different orientation from 2 and forms different interactions with the protein (Fig. 5C). In the docked mode, the carboxylate group of 30 is not close to Arg316 to make the same interaction as seen in 2. Instead, 30 makes more extensive hydrophobic interactions with hydrophobic side chains in H3, H5, H7 and H11 (Fig. 5B). Recent crystal structures of RXRα in complex with antagonists have demonstrated the significance of the hydrophobic interactions that were found dominant in the ligand binding [22, 23]. Therefore in the case of 30, it is conceivable that even though the acid group of 30 may not contribute as much to the binding as that of 2, the hydrophobic interactions play a key role in the binding.</p><p>We further tested 30 in other biological assays for its effect on the apoptosis of cancer cells and its ability to inhibit the PI3K/AKT activation. In the MTT assay, 30 could dose-dependently induce growth inhibition in some cancer cell lines such as PC3 prostate cancer cells and ZR75-1 breast cancer cells (Figure 6A). In the induction of PARP cleavage, 30 was more effective than 1 or 2 (Figure 6B).</p><p>We previously demonstrated that inhibition of AKT activation by the Sulindac/TNFα combination was closely associated with its apoptotic effect [18]. We then investigated whether compound 30 could inhibit TNFα-induced AKT activation. In agreement with previous studies, treatment of A549 lung cancer cells with TNFα led to a strong AKT activation [18], which was inhibited by 1, 2 or 30 (Figure 7A). Such effects were also observed in HCT-116 colon cancer cells and HepG2 liver cancer cells (data not shown). Consistently, compound 30 showed a better effect on the inhibition of TNFα-induced AKT activation. Knocking down tRXRα by siRNA significantly impaired the inhibitory effect of 30 on AKT activation (Figure 7B). These results indicated that inactivation of AKT by 30 was tRXRα-dependent.</p><p>We also examined whether 30 could enhance the TNFα-induced apoptosis. Figure 8 showed that 30 could significantly enhance the PARP cleavage in combination with TNFα, suggesting that 30 could activate TNFα-dependent apoptotic pathway. The observed synergistic effect of 30/TNFα on the TNFα-induced apoptosis was stronger than that of 2/TNFα or 1/TNFα.</p><!><p>The synthesis started from the Perkin reaction of 4-fluorobenzaldehyde 31 with propionate anhydride [24], in which K2CO3 was used as a base to substitute hygroscopic sodium propionate, providing the desired product 32 in 83% yield (Scheme 1). Catalytic hydrogenation in the presence of Pd/C and under 10 atm of hydrogen gave carboxylic acid 33 in 90% yield. Polyphosphoric acid (PPA)-promoted intermolecular Friedel-Crafts acylation reaction produced indenone 34 in 74% yield. Treatment of indenone 34 with the enolate generated from ethyl acetate and LDA gave the corresponding β-hydroxy ester, which was treated with a mixture of HOAc and concentrated H2SO4 (v/v 10:1) to yield the indene 35 in 80% yield. Finally, two methods were used for the Claisen-Schmidt condensation reactions of compound 35 with differently substituted benzaldehydes to give compounds 3~14, respectively. The electronic properties of the substituents on the aromatic aldehydes were found to have an impact on the reaction, and slightly different conditions should be used for the synthesis of a specific compound. The results of the reactions are summarized in Table 4.</p><!><p>In conclusion, we have described the synthesis and SAR studies on a series of novel analogs of Sulindac as potential modulators for inhibiting tRXRα-dependent AKT activation. Compound 30, a geometric isomer of the original lead 2 and with better binding activity and improved biological effects, could bind to the LBP of RXRα in a different mode from 2, which offers a new design strategy. 30 is a promising lead for further optimization studies and may find application as a small molecule probe in studying the mechanism of the tRXRα-dependent AKT signaling.</p><!><p>Melting points (M.p.) were determined on a Yanaco MP-500 micro melting point apparatus and were uncorrected. Infrared spectra were measured with a Nicolet Avatar 360 FT-IR spectrometer using film KBr pellet techniques. 1H and 13C NMR spectra were recorded in CDCl3 or CD3OD on a Bruker 400 spectrometer with tetramethylsilane as an internal standard. Chemical shifts are expressed in δ (ppm) units downfield from TMS. Mass spectra were recorded by a Bruker Dalton ESquire 3000 plus liquid chromatography-mass spectrum (direct injection). Optical rotations were measured with a Perkin-Elmer 341 automatic polarimeter. Diastereoselectivities and enantioselectivities were determined by chiral HPLC analysis using a Shimadzu LC-10AT VP series and a Shimadzu SPD-M10Avp photo diode array detector (190–370 nm) with a Chiralcel OJ-H column using n-hexane/i-PrOH (98:2, v/v) as a mobile phase. Flash column chromatography was carried out with silica gel (300–400 mesh). THF was distilled over sodium benzophenone ketyl under N2.</p><!><p>Appropriate anhydride (300 mmol, 1.6 equiv.) was added to potassium carbonate (224 mmol, 1.2 equiv.) at 0 °C. After stirring for 5 min to mix up, appropriate aromatic aldehyde (186 mmol, 1.0 equiv.) was added. The mixture was heated to reflux for 12 h. After cooling with an ice bath, to the reaction mixture was added water and solid Na2CO3 (30 g). After the resultant yellow precipitate was filtered, the reaction mixture was acidified to pH 6.0 using concentrated HCl to afford acrylic acid as a solid.</p><!><p>Compound 32 [24] was synthesized according to the general procedure A. Pale yellow crystals, yield: 83%. M.p. 155–158 °C (MeOH); IR (film): νmax 3429, 3076, 2972, 1665, 1596, 1508, 1425, 1313, 1298, 1224 cm−1; 1H NMR (400 MHz, DMSO-d6) δ 2.00 (d, J = 1.2H Hz, 3H, C 3), 7.19–7.25 (m, 2H, Ar-H), 7.46–7.52 (m, 2H, 2H, Ar-H), 7.58 (s, 1H, vinyl-H), 12.50 (br s, 1H, COOH) ppm; 13C NMR (100 MHz, DMSO-d6) δ 14.2, 115.8 (d, JC-F = 21.0 Hz), 129.0, 132.2 (d, JC-F = 9.0 Hz), 132.46 (d, JC-F = 3.0 Hz), 136.9, 162.3 (d, JC-F = 245.0 Hz), 169.7 ppm; MS (ESI) m/z 179 (M+H+).</p><!><p>A mixture of acrylic acid (55 mmol, 1.0 equiv.) and Pd/C (10%) in methanol (70 mL) was hydrogenated under 10 atm of hydrogen for 24 h. The catalyst was filtered off and the filtrate concentrated to afford propanoic acid, which was used in the next step as it was. An analytical sample of compound was obtained by flash column chromato- graphy on silica gel.</p><!><p>Compound 33 [24] was synthesized according to the general procedure B. Colorless oil, yield: 90%. IR (film): νmax 3406, 2972, 2933, 1701, 1560, 1509, 1460, 1406, 1223 cm−1; 1H NMR (400 MHz, CDCl3) δ 1.12 (d, J = 6.7 Hz, 3H, CH3), 2.60 (dd, J = 13.0, 7.9 Hz, 1H, CH2CH), 2.66 (ddq, J = 7.9, 6.0, 6.7 Hz, CHCH3), 2.99 (dd, J = 13.0, 6.0 Hz, 1H, CH2CH), 6.90–7.00 (m, 2H, Ar-H), 7.06–7.14 (m, 2H, Ar-H), 9.80 (br s, 1H, COOH) ppm; 13C NMR (100 MHz, CDCl3) δ 16.5, 38.6, 41.8, 115.1 (d, JC-F = 21.0 Hz), 130.35 (d, JC-F = 8.0 Hz), 134.87 (d, JC-F = 3.0 Hz), 161.6 (d, JC-F = 243.0 Hz), 182.3 ppm; MS (ESI) m/z 181 (M+H+).</p><!><p>A mixture of the crude propanoic acid derivative (42.0 mmol, 1.0 equiv.) and polyphosphoric acid (400 mmol, 9.5 equiv.) was stirred at 80 °C for 12 hours. The resulting mixture was poured into ice water and extracted with EtOAc (30 mL × 3). The combined extracts were washed with a saturated aqueous NaHCO3 (10 mL × 3) to remove the starting acids, and then washed with brine, dried over anhydrous Na2SO4, filtered, and concentrated under reduced pressure. The residue was purified by flash column chromatography to afford indenone.</p><!><p>Compound 34 [24] was synthesized according to the general procedure C, and purified by flash column chromatography on silica gel (eluent: ethyl acetate: petroleum ether = 1:40). Pale yellow oil, yield: 74%. IR (film) νmax 3064, 2968, 2932, 2873, 1716, 1611, 1509, 1486, 1444, 1264, 1158 cm−1; 1H NMR (400 MHz, CDCl3) δ 1.32 (d, J = 7.4 Hz, 3H, CH3), 2.70 (dd, J = 16.7, 3.9 Hz, 1H, CH2CH), 2.74–2.82 (m, 1H, CHCH3), 3.37 (dd, J = 16.7, 7.6 Hz, 1H, CH2CH), 7.26–7.33 (m, 1H, Ar-H), 7.36–7.44 (m, 2H, Ar-H) ppm; 13C NMR (100 MHz, CDCl3) δ 16.2, 34.4, 42.9, 109.7 (d, JC-F = 22.0 Hz), 122.3 (d, JC-F = 24.0 Hz), 127.85 (d, JC-F = 8.0 Hz), 138.1, 148.8, 162.3 (d, JC-F = 247.0 Hz), 208.4 ppm; MS (ESI) m/z 187 (M+Na+).</p><!><p>To a solution of LDA or LHMDS (48.0 mmol, 2.0 equiv.) in anhydrous THF (100 mL) was added EtOAc (61.0 mmol, 2.5 equiv.) at −78 °C. The mixture was stirred at −78 °C for 30 min. To the resulting mixture was added dropwise a solution of indenone (24.0 mmol, 1.0 equiv.) in anhydrous THF (20 mL). The mixture was stirred at −78 °C for another 4 hr and then quenched with a saturated aqueous NH4Cl. The mixture was extracted with EtOAc (20 mL × 3). The combined organic layers were dried over anhydrous Na2SO4, filtered, and concentrated under reduced pressure. To the residue was added HOAc/H2SO4 (10/1, 40 mL). After stirring for 4 hr at room temperature, the mixture was extracted with EtOAc (15 mL × 3). The combined extracts were washed successively with water, saturated NaHCO3, and brine, dried over Na2SO4, filtered, and concentrated under reduced pressure. The residue was purified by flash column chromatography on silica gel to afford inden-3-yl acetate.</p><!><p>Compound 35 [24] was synthesized according to the general procedure D, and purified by flash column chromatography on silica gel (eluent: ethyl acetate: petroleum ether = 1:50). Colorless oil, yield: 80%. IR (film) νmax 2981, 2911, 1736, 1614, 1590, 1473, 1368, 1329, 1308, 1256, 1154, 1034 cm−1; 1H NMR (400 MHz, CDCl3) δ 1.25 (t, J = 7.1 Hz, 3H, OCH2CH3), 2.12 (s, 3H, C=CCH3), 3.29 (s, 2H, ArCH2C=C), 3.48 (s, 2H, CH2COOEt), 4.14 (q, J = 7.1 Hz, 2H, OCH2CH3), 6.77–6.83 (m, 1H, Ar-H), 6.94–6.99 (m, 1H, Ar-H), 7.23–7.27 (m, 1H, Ar-H) ppm; 13C NMR (100 MHz, CDCl3) δ 14.15, 14.26, 31.5, 42.1, 60.9, 105.8 (d, JC-F = 23.0 Hz), 110.3 (d, JC-F = 23.0 Hz), 123.7 (d, JC-F = 9.0 Hz), 129.6, 137.19 (d, JC-F = 2.0 Hz) 144.5, 147.87 (d, JC-F = 9.0 Hz), 162.4 (d, JC-F = 239.0 Hz), 170.7 ppm; MS (ESI) m/z 257 (M+Na+).</p><!><p>To a solution of indene-3-yl acetate 35 (1.3 mmol, 1.0 equiv.) in MeOH (4.0 mL) was added 2.5 N NaOMe (4.0 mmol, 3.0 equiv.) at room temperature to get an orange mixture. After stirring for 30 min, to the mixture was added appropriate aromatic aldehyde (1.3~2.0 mmol, 1.0~1.5 equiv.). The resulting mixture was refluxed at 80 °C for 4 h. After concentrated under reduced pressure, the residue was acidified with a 1N HCl solution to pH 4.0~6.0. After stirring for another 0.5 hr at room temperature, the mixture was extracted with EtOAc (15 mL × 3). The combined organic layers were dried over anhydrous Na2SO4, filtered, and concentrated under reduced pressure. The residue was purified by flash chromatography to afford indene derivative. An analytical sample of compound was obtained by recrystallization.</p><!><p>Compound 3 was synthesized according to the general procedure E, and purified by flash column chromatography on silica gel (eluent: ethyl acetate: petroleum ether = 1:3). Yellow solid, yield: 77%. M.p. 175–176 °C (hexane/EtOAc); IR (film): νmax 3430, 3021, 2918, 1705, 1604, 1467, 1415, 1302, 1168 cm−1; 1H NMR (400 MHz, CDCl3) δ 2.10 (s, 3H, C=CCH3), 3.50 (s, 2H, CH2COOH), 6.43–6.50 (m, 1H, Ar-H), 6.75–6.84 (m, 1H, Ar-H), 7.13 (s, 1H, vinyl-H), 7.25–7.43 (m, 6H, Ar-H) ppm; 13C NMR (100 MHz, CDCl3) δ 10.5, 31.4, 105.7 (d, JC-F = 24.0 Hz), 110.6 (d, JC-F = 22.0 Hz), 123.8 (d, JC-F = 9.0 Hz), 128.0, 128.2, 128.5 (2C), 129.2 (2C), 129.4, 129.8, 130.1, 130.7, 136.5, 138.8, 140.2, 146.2 (d, JC-F = 8.0 Hz), 163.1 (d, JC-F = 245.0 Hz), 176.6 ppm; MS (ESI) m/z 317.1 (M+Na+, 100%); HRMS (ESI) calcd for C19H15FNaO2+ [M+Na+]: 317.0948; found: 317.0951.</p><!><p>Compound 4 was synthesized according to the general procedure E, and purified by flash column chromatography on silica gel (eluent: ethyl acetate: petroleum ether = 1:3). Yellow solid, yield: 62%. M.p. 188–189 °C (hexane/EtOAc); IR (film): νmax 3435, 2918, 1708, 1604, 1467, 1321, 1165, 1122, 1065, 1016 cm−1; 1H NMR (400 MHz, CDCl3) δ 2.21 (s, 3H, C=CCH3), 3.60 (s, 2H, CH2COOH), 6.54–6.61 (m, 1H, Ar-H), 6.86–6.91 (m, 1H, Ar-H), 7.08–7.14 (m, 1H, Ar-H), 7.18 (s, 1H, vinyl-H), 7.57–7.63 (m, 2H, Ar-H), 7.67–7.74 (m, 2H, Ar-H) ppm; 13C NMR (100 MHz, CDCl3) δ 10.5, 31.3, 106.1 (d, JC-F = 23.0 Hz), 110.9 (d, JC-F = 22.0 Hz), 122.7, 123.7 (d, JC-F = 9.0 Hz), 125.46, 125.49, 128.3, 129.4, 129.5, 131.1, 138.6, 140.3, 141.6, 146.4 (d, JC-F = 8.0 Hz), 163.4 (d, JC-F = 245.0 Hz), 176.0 ppm; MS (ESI) m/z 385.1 (M+Na+, 100%); HRMS (ESI) calcd for C20H14F4NaO2+ [M+Na+]: 385.0822; found: 385.0819.</p><!><p>Compound 5 was synthesized according to the general procedure E, and purified by flash column chromatography on silica gel (eluent: ethyl acetate: petroleum ether = 1:3). Yellow solid, yield: 72%. M.p. 187–188 °C (hexane/EtOAc); IR (film): νmax 3420, 2964, 1708, 1604, 1503, 1464, 1412, 1363, 1266, 1168 cm−1; 1H NMR (400 MHz, CDCl3) δ 1.39 (s, 9H, C(CH3)3), 2.20 (s, 3H, C=CCH3), 3.60 (s, 2H, CH2COOH), 6.56–6.64 (m, 1H, Ar-H), 6.86–6.93 (m, 1H, Ar-H), 7.20 (s, 1H, vinyl-H), 7.38–7.52 (m, 5H, Ar-H) ppm; 13C NMR (100 MHz, CDCl3) δ 10.5, 31.3, 31.4, 34.8, 105.6 (d, JC-F = 24.0 Hz), 110.6 (d, JC-F = 22.0 Hz), 123.8 (d, JC-F = 9.0 Hz), 125.4 (2C), 129.2 (2C), 129.7, 129.8, 130.9, 133.4, 139.0, 139.7, 146.2 (d, JC-F = 9.0 Hz), 151.6, 163.1 (d, JC-F = 244.0 Hz), 176.3 ppm; MS (ESI) m/z 373.2 (M+Na+, 100%); HRMS (ESI) calcd for C23H23FNaO2+ [M+Na+]: 373.1574; found: 373.1571.</p><!><p>To a solution of indene 35 (150 mg, 0.64 mmol) in toluene (4.0 mL) was added DBU (0.9 mL, 6.4 mmol) at room temperature. After stirring for 30 min at 80 °C, to the mixture was added the solution of aromatic aldehyde (168 mg, 0.96 mmol) in toluene (2.0 mL). The resulting mixture was heated at 80 °C for 36 h, then quenched with a saturated aqueous NH4Cl. The mixture was extracted with EtOAc (20 mL × 3). The combined organic layers were dried over anhydrous Na2SO4, filtered, and concentrated under reduced pressure. Then to the residue in MeOH (3 mL) was added 2N NaOH (2 mL). The mixture was stirring for 2 hr at 65 °C. After concentrated under reduced pressure, the residue was acidified with a 1N HCl solution to pH 6.0~7.0. The residue was purified by flash chromatography on silica gel (eluent: 1H NMR (400 MHz, DMSO-d6) δ 2.18 (s, 3H, C=CCH3), 3.59 (s, 2H, CH2COOH), 6.70–6.77 (m, 1H, Ar-H), 7.00–7.05 (m, 1H, Ar-H), 7.31–7.36 (m, 1H, Ar-H), 7.36–7.40 (m, 1H, Ar-H), 7.40 (s, 1H, vinyl-H), 7.64–7.70 (m, 2H, Ar-H), 7.89–7.95 (m, 1H, Ar-H), 8.04–8.08 (m, 1H, Ar-H), 8.20–8.25 (m, 2H, Ar-H), 8.68–8.72 (m, 1H, Ar-H), 12.45 (br s, 1H, COOH) ppm; 13C NMR (100 MHz, DMSO-d6) δ 10.3, 31.1, 105.9 (d, JC-F = 24.0 Hz), 110.3 (d, JC-F = 23.0 Hz), 120.3, 122.9, 123.2 (d, JC-F = 9.0 Hz), 126.6 (2C), 129.6, 129.7, 130.4 (2C), 132.1, 136.8, 137.3, 138.0, 138.4, 139.8, 146.96 (d, JC-F = 8.0 Hz), 149.6, 155.3, 162.42 (d, JC-F = 242.0 Hz), 171.6 ppm; MS (ESI) m/z 372.1 (M+H+, 100%); HRMS (ESI) calcd for C24H19FNO2+ [M+H+]: 372.1394; found: 372.1395.</p><!><p>Compound 7 was synthesized according to the general procedure E, and purified by flash column chromatography on silica gel (eluent: ethyl acetate: petroleum ether = 1:4). Yellow solid, yield: 78%. M.p. 182–183 °C (hexane/EtOAc); IR (film): νmax 3418, 2927, 2833, 1708, 1601, 1507, 1464, 1296, 1250, 1171, 1028 cm−1; 1H NMR (400 MHz, CDCl3) δ 2.20 (s, 3H, C=CCH3), 3.60 (s, 2H, CH2COOH), 3.89 (s, 3H, OCH3), 6.55–6.63 (m, 1H, Ar-H), 6.85–6.91 (m, 1H, Ar-H), 6.92–6.99 (m, 2H, Ar-H), 7.18 (s, 1H, vinyl-H), 7.36–7.43 (m, 1H, Ar-H), 7.44–7.50 (m, 2H, Ar-H) ppm; 13C NMR (100 MHz, CDCl3) δ 10.6, 31.3, 55.3, 105.6 (d, JC-F = 24.0 Hz), 110.5 (d, JC-F = 22.0 Hz), 113.9 (2C), 123.6 (d, JC-F = 9.0 Hz), 128.7, 129.5, 129.8, 130.8, 130.9 (2C), 139.0, 139.2, 146.1 (d, JC-F = 9.0 Hz), 159.7, 163.0 (d, JC-F = 245.0 Hz), 175.6 ppm; MS (ESI) m/z 347.1 (M+Na+, 100%); HRMS (ESI) calcd for C20H17FNaO3+ [M+Na+]: 347.1054; found: 347.1060.</p><!><p>Compound 8 was synthesized according to the general procedure E, and purified by flash column chromatography on silica gel (eluent: ethyl acetate: petroleum ether = 1:4). Yellow solid, yield: 66%. M.p. 186–187 °C (hexane/EtOAc); IR (film): νmax 3410, 2976, 2921, 1705, 1601, 1507, 1464, 1296, 1247, 1168, 1043 cm−1; 1H NMR (400 MHz, CDCl3) δ 1.47 (t, J = 7.0 Hz, 3H, OCH 2CH3), 2.21 (s, 3H, C=CCH3), 3.59 (s, 2H, CH2COOH), 4.11 (q, J = 7.0 Hz, 2H, OCH2CH3), 6.56–6.63 (m, 1H, Ar-H), 6.86–6.99 (m, 3H, Ar-H), 7.18 (s, 1H, vinyl-H), 7.40–7.49 (m, 3H, Ar-H) ppm; 13C NMR (100 MHz, CDCl3) δ 10.6, 14.8, 31.4, 63.5, 105.5 (d, JC-F = 23.0 Hz), 110.5 (d, JC-F = 22.0 Hz), 114.4 (2C), 123.5 (d, JC-F = 9.0 Hz), 128.5, 129.4, 129.8, 130.88, 130.94 (2C), 139.0, 139.1, 146.1 (d, JC-F = 9.0 Hz), 159.1, 163.0 (d, JC-F = 244.0 Hz), 176.9 ppm; MS (ESI) m/z 361.1 (M+Na+, 100%); HRMS (ESI) calcd for C21H19FNaO3+ [M+Na+]: 361.1210; found: 361.1212.</p><!><p>Compound 9 was synthesized according to the general procedure E, and purified by flash column chromatography on silica gel (eluent: ethyl acetate: petroleum ether = 1:4). Yellow solid, yield: 73%. M.p. 199–201 °C (hexane/EtOAc); IR (film): νmax 3434, 2915, 2223, 1705, 1601, 1467, 1496, 1314, 1266, 1226, 1165, 1131, 1113, 1016 cm−1; 1H NMR (400 MHz, CDCl3) δ 2.20 (s, 3H, C=CCH3), 3.60 (s, 2H, CH2COOH), 6.53–6.60 (m, 1H, Ar-H), 6.85–6.90 (m, 1H, Ar-H), 7.04–7.10 (m, 1H, Ar-H), 7.12 (s, 1H, vinyl-H), 7.59–7.62 (m, 2H, Ar-H), 7.70–7.75 (m, 2H, Ar-H) ppm; 13C NMR (100 MHz, CDCl3) δ 10.5, 31.3, 106.3 (d, JC-F = 24.0 Hz), 111.0 (d, JC-F = 22.0 Hz), 111.7, 118.6, 123.7 (d, JC-F = 9.0 Hz), 127.6, 129.2, 129.9 (2C), 131.5, 132.3 (2C), 138.5, 141.4, 142.1, 146.5 (d, JC-F = 8.0 Hz), 163.4 (d, JC-F = 246.0 Hz), 176.0 ppm; MS (ESI) m/z 342.1 (M+Na+, 100%); HRMS (ESI) calcd for C20H14FNNaO2+ [M+Na+]: 342.0901; found: 342.0902.</p><!><p>Compound 10 was synthesized according to the general procedure E, and purified by flash column chromatography on silica gel (eluent: ethyl acetate: petroleum ether = 1:4). Yellow solid, yield: 20%. M.p. 174–175 °C (hexane/EtOAc); IR (film): νmax 3415, 2911, 1708, 1594, 1522, 1464, 1363, 1189, 1162, 1135, 1061 cm−1; 1H NMR (400 MHz, CDCl3) δ 2.21 (s, 3H, C=CCH3), 3.04 (s, 6H, N(CH3)2), 3.60 (s, 2H, CH2COOH), 6.58–6.65 (m, 1H, Ar-H), 6.72–6.77 (m, 2H, Ar-H), 6.87–6.92 (m, 1H, Ar-H), 7.17 (s, 1H, vinyl-H), 7.45–7.51 (d, 2H, Ar-H), 7.64–7.70 (m, 1H, Ar-H) ppm; 13C NMR (100 MHz, CDCl3) δ 10.7, 31.3, 40.3, 105.3 (d, JC-F = 23.0 Hz), 110.3 (d, JC-F = 23.0 Hz), 111.7 (2C), 123.4 (d, JC-F = 9.0 Hz), 123.8, 128.3, 130.0, 131.3 (2C), 132.1, 137.2, 139.2, 145.8 (d, JC-F = 10.0 Hz), 150.4, 162.8 (d, JC-F = 247 Hz), 175.8 ppm; MS (ESI) m/z 338.2 (M+H+, 100%); HRMS (ESI) calcd for C21H21FNO2+ [M+H+]: 338.1551; found: 338.1549.</p><!><p>Compound 11 was synthesized according to the general procedure E, and purified by flash column chromatography on silica gel (eluent: ethyl acetate: petroleum ether = 1:4). Yellow solid, yield: 71%. M.p. 188–190 °C (hexane/EtOAc); IR (film): νmax 3433, 2921, 1708, 1604, 1464, 1409, 1330, 1165, 1263, 1122, 1208, 1068 cm−1; 1H NMR (400 MHz, CDCl3) δ 2.21 (s, 3H, C=CCH3), 3.60 (s, 2H, CH2COOH), 6.53–6.61 (m, 1H, Ar-H), 6.86–6.92 (m, 1H, Ar-H), 7.07–7.13 (m, 1H, Ar-H), 7.18 (s, 1H, vinyl-H), 7.53–7.59 (m, 1H, Ar-H), 7.62–7.72 (m, 2H, Ar-H), 7.74–7.79 (m, 1H, Ar-H) ppm; 13C NMR (100 MHz, CDCl3) δ 10.5, 31.4, 106.1 (d, JC-F = 24.0 Hz), 110.9 (d, JC-F = 23.0 Hz), 122.6, 123.6 (d, JC-F = 9.0 Hz), 124.8, 125.3, 126.12, 128.3, 129.0, 129.4, 131.0, 132.5, 137.3, 138.6, 141.5, 146.4 (d, JC-F = 9.0 Hz), 163.3 (d, JC-F = 245.0 Hz), 176.5 ppm; MS (ESI) m/z 385.1 (M+Na+, 100%); HRMS (ESI) calcd for C20H14F4NaO2+ [M+Na+]: 385.0822; found: 385.0825.</p><!><p>Compound 12 was synthesized according to the general procedure E, and purified by flash column chromatography on silica gel (eluent: ethyl acetate: petroleum ether = 1:4). Yellow solid, yield: 70%. M.p. 132–133 °C (hexane/EtOAc); IR (film): νmax 3418, 2936, 2833, 1708, 1598, 1464, 1424, 1275, 1159, 1049 cm−1; 1H NMR (400 MHz, CDCl3) δ 2.21 (s, 3H, C=CCH3), 3.59 (s, 2H, CH2COOH), 3.82 (s, 3H, OCH3), 6.53–6.61 (m, 1H, Ar-H), 6.86–6.91 (m, 1H, Ar-H), 6.91–6.96 (m, 1H, Ar-H), 7.01–7.10 (m, 2H, Ar-H), 7.20 (s, 1H, vinyl-H), 7.27–7.37 (m, 2H, Ar-H) ppm; 13C NMR (100 MHz, CDCl3) δ 10.5, 31.3, 55.3, 105.7 (d, JC-F = 23.0 Hz), 110.7 (d, JC-F = 23.0 Hz), 114.27, 114.30, 121.6, 124.0 (d, JC-F = 9.0 Hz), 129.6, 129.8, 130.2, 130.4, 137.9, 138.8, 140.4, 146.3 (d, JC-F = 9.0 Hz), 159.7, 163.2 (d, JC-F = 245.0 Hz), 175.9 ppm; MS (ESI) m/z 347.1 (M+Na+, 100%); HRMS (ESI) calcd for C20H17FNaO3+ [M+Na+]: 347.1054; found: 347.1054.</p><!><p>Compound 13 was synthesized according to the general procedure E, and purified by flash column chromatography on silica gel (eluent: ethyl acetate: petroleum ether = 1:4). Yellow solid, yield: 60%. M.p. 187–189 °C (hexane/EtOAc); IR (film): νmax 3433, 2917, 2226, 1711, 1601, 1464, 1409, 1311, 1271, 1168, 1131, 1037 cm−1; 1H NMR (400 MHz, CDCl3) δ 2.20 (s, 3H, C=CCH3), 3.59 (s, 2H, CH2COOH), 6.54–6.61 (m, 1H, Ar-H), 6.85–6.91 (m, 1H, Ar-H), 6.97–7.02 (m,1H, Ar-H), 7.09 (s, 1H, vinyl-H), 7.52–7.58 (m, 1H, Ar-H), 7.65–7.70 (m, 1H, Ar-H), 7.71–7.78 (m, 2H, Ar-H) ppm; 13C NMR (100 MHz, CDCl3) δ 10.5, 31.3, 106.3 (d, JC-F = 23.0 Hz), 111.0 (d, JC-F = 23.0 Hz), 112.9, 118.3, 123.5 (d, JC-F = 9.0 Hz), 127.0, 129.2, 129.4, 131.4, 131.6, 132.6, 133.5, 138.0, 138.4, 142.0, 146.5 (d, JC-F = 8.0 Hz), 163.4 (d, JC-F = 246.0 Hz), 175.6 ppm; MS (ESI) m/z 342.1 (M+Na+, 100%); HRMS (ESI) calcd for C20H14FNNaO2+ [M+Na+]: 342.0901; found: 342.0899.</p><!><p>Compound 14 was synthesized according to the general procedure E, and purified by flash column chromatography on silica gel (eluent: ethyl acetate: petroleum ether = 1:4). Yellow solid, yield: 22%. M.p. 240–242 °C (hexane/EtOAc); IR (film): νmax 3411, 3296, 2921, 1662, 1595, 1476, 1406, 1381, 1318, 1220, 1171, 1122, 1037 cm−1; 1H NMR (400 MHz, DMSO-d6) δ 2.08 (s, 3H, COCH3), 2.14 (s, 3H, C=CCH3), 3.57 (s, 2H, CH2COOH), 6.70–6.79 (m, 1H, Ar-H), 6.97–7.04 (m, 1H, Ar-H), 7.28 (s, 1H, vinyl-H), 7.32–7.37 (m, 1H, Ar-H), 7.45–7.53 (m, 2H, Ar-H), 7.65–7.73 (m, 2H, Ar-H), 10.13 (s, 1H, NHCOCH3), 12.40 (s, 1H, COOH) ppm; 13C NMR (100 MHz, DMSO-d6) δ 10.2, 24.0, 31.1, 105.7 (d, JC-F = 23.0 Hz), 110.1 (d, JC-F = 22.0 Hz), 118.6 (2C), 122.9 (d, JC-F = 9.0 Hz), 129.6, 130.0 (2C), 130.4, 130.9, 131.5, 137.9, 138.7, 139.5, 146.7 (d, JC-F = 9.0 Hz), 162.2 (d, JC-F = 241.0 Hz), 168.5, 171.6 ppm; MS (ESI) m/z 374.1 (M+Na+, 100%); HRMS (ESI) calcd for C21H18FNNaO3+ [M+Na+]: 374.1163; found: 374.1162.</p><!><p>A solution of compound 34 (164.0 mg, 1.0 mmol), iso-propanol (0.38 mL, 5.0 mmol), and methyl acrylate (0.9 mL, 10 mmol) in THF (4 mL) was purged with argon for 20 min and cooled to 0 °C. A SmI2 (3.0 mmol) solution in THF (30 mL) was added through transfer needle. After 5 min, the reaction was quenched with saturated Na2CO3 (3 mL). The resulting mixture was extracted with Et2O (5 mL × 3). The combined organic phases were washed with brine, dried over anhydrous Na2SO4, filtered and concentrated under reduced pressure. To a solution of the residue in CH3OH (4.0 mL) was added p-TsOH (cat.), then the mixture was refluxed for 3 h. The reaction was quenched with a saturated aqueous NaHCO3 (2.0 mL). The resulting mixture was extracted with EtOAc (5 mL × 3). The combined organic phases were washed with brine, dried over anhydrous Na2SO4, filtered and concentrated under reduced pressure. The residue was purified by flash column chromatography on silica gel (eluent: ethyl acetate: petroleum ether = 1:10) to afford compound 38 (171 mg, 73%) as a colorless oil. IR (film): νmax 2948, 1735, 1610, 1589, 1473, 1430, 1281, 1171, 1046 cm−1; 1H NMR (400 MHz, CDCl3) δ 2.09 (s, 3H, C=CCH3), 2.53 (t, J = 7.8 Hz, 2H, CH2CH2COOMe), 2.82 (t, J = 7.8 Hz, 2H, CH2CH2COOMe), 3.23 (s, 2H, ArCH2C=C), 3.68 (s, 3H, COOCH3), 6.76–6.82 (m, 1H, Ar-H), 6.89–6.93 (m, 1H, Ar-H), 7.23–7.28 (m, 1H, Ar-H) ppm; 13C NMR (100 MHz, CDCl3) δ 14.0, 20.6, 32.8, 42.0, 51.6, 105.2 (d, JC-F = 23.0 Hz), 110.0 (d, JC-F = 23.0 Hz), 123.75 (d, JC-F = 9.0 Hz), 134.85 (d, JC-F = 3.0 Hz), 137.56 (d, JC-F = 2.0 Hz), 142.3, 147.9 (d, JC-F = 9.0 Hz), 162.4 (d, JC-F = 240.0 Hz), 173.4 ppm; MS (ESI) m/z 257.1 (M+Na+, 100%); HRMS (ESI) calcd for C14H15FNaO2+ [M+Na+]: 257.0948; found: 257.0946.</p><!><p>Compound 15 was synthesized according to the general procedure E, and purified by flash column chromatography on silica gel (eluent: ethyl acetate: petroleum ether = 1:4). Yellow solid, yield: 86%. M.p. 130–131 °C (hexane/EtOAc); IR (film): νmax 3426, 2961, 1711, 1601, 1464, 1412, 1290, 1193, 1138 cm−1; 1H NMR (400 MHz, CDCl3) δ 1.31 (d, J = 6.9 Hz, 6H, CH(CH3)2), 2.18 (s, 3H, C=CCH3), 2.62 (t, J = 7.8 Hz, 2H, CH2CH2COOH), 2.90 (t, J = 7.8 Hz, 2H, CH2CH2COOH), 2.97 (sept, J = 6.9 Hz, 1H, CH(CH3)2), 6.55–6.62 (m, 1H, Ar-H), 6.82–6.87 (m, 1H, Ar-H), 7.14 (s, 1H, vinyl-H), 7.27–7.31 (m, 2H, Ar-H), 7.35–7.40 (m, 1H, Ar-H), 7.42–7.46 (m, 2H, Ar-H) ppm; 13C NMR (100 MHz, CDCl3) δ 10.3, 20.7, 23.9, 32.7, 34.0, 105.3 (d, JC-F = 23.0 Hz), 110.3 (d, JC-F = 23.0 Hz), 123.8 (d, JC-F = 9.0 Hz), 126.5, 129.4, 129.9, 130.24 (d, JC-F = 3.0 Hz), 134.0, 136.06 (d, JC-F = 2.0 Hz), 136.7, 140.0, 146.36 (d, JC-F = 8.0 Hz), 149.1, 163.1 (d, JC-F = 244.0 Hz), 178.3 ppm; MS (ESI) m/z 373.2 (M+Na+, 100%); HRMS (ESI) calcd for C23H23FNaO2+ [M+Na+]: 373.1574; found: 373.1572.</p><!><p>A solution of compound 34 (300.0 mg, 1.8 mmol), and iso-propanol (0.7 mL, 9.0 mmol), and acrylonitrile (1.2 mL, 18.0 mmol) in THF (4 mL) was purged with argon for 20 min and cooled to 0 °C. A SmI2 (5.4 mmol) solution in THF (54 mL) was added through transfer needle. After 5 min, the reaction was quenched with saturated aqueous Na2CO3 (10 mL). The resulting mixture was extracted with Et2O (15 mL × 3). The combined organic phases were washed with brine, dried over anhydrous Na2SO4, filtered and concentrated under reduced pressure. To the residue was added HOAc/H2SO4 (10/1, 3.0 mL). After stirring for 4 hr at room temperature, the mixture was extracted with EtOAc (15 mL × 3). The combined extracts were washed successively with saturated NaHCO3 and brine, dried over Na2SO4, filtered, and concentrated under reduced pressure. The residue was purified by flash column chromatography on silica gel (eluent: ethyl acetate: petroleum ether = 1:4) to afford compound 39 as a white solid (108 mg, 30%). M.p. 91–92 °C (hexane/EtOAc); IR (film): νmax 2915, 2247, 1610, 1592, 1476, 1275, 1190, 1165 cm−1; 1H NMR (400 MHz, CDCl3) δ 2.16 (s, 3H, C=CCH3), 2.57 (t, J = 7.3 Hz, 2H, CH2CH2CN), 2.86 (t, J = 7.3 Hz, 2H, CH2CH2CN), 3.31 (s, 2H, CH2C=C), 6.79–6.88 (m, 2H, Ar-H), 7.27–7.32 (m, 1H, Ar-H) ppm; 13C NMR (100 MHz, CDCl3) δ 14.3, 16.6, 21.3, 41.2, 104.8 (d, JC-F = 24.0 Hz), 110.5 (d, JC-F = 23.0 Hz), 119.2, 124.17 (d, JC-F = 9.0 Hz), 132.8, 137.5, 144.6, 146.9 (d, JC-F = 9.0 Hz), 162.4 (d, JC-F = 241.0 Hz) ppm; MS (ESI) m/z 224.1 (M+Na+, 100%); HRMS (ESI) calcd for C13H12FNNa+ [M+Na+]: 224.0846; found: 224.0848.</p><!><p>Compound 16 was synthesized according to the general procedure E, and purified by flash column chromatography on silica gel (eluent: ethyl acetate: petroleum ether = 1:20). Yellow solid, yield: 53%. M.p. 108–109 °C (hexane/EtOAc); IR (film): νmax 2957, 2927, 2866, 2247, 1598, 1464, 1199, 1162, 1138, 1055, 1016 cm−1; 1H NMR (400 MHz, CDCl3) δ 1.32 (d, J = 6.9 Hz, 6H, CH(CH3)2), 2.24 (s, 3H, C=CCH3), 2.60 (t, J = 7.4 Hz, 2H, CH2CH2CN), 2.93 (t, J = 7.4 Hz, 2H, CH2CH2CN), 2.98 (sept, J = 6.9 Hz, 1H, CH(CH3)2), 6.58–6.65 (m, 1H, Ar-H), 6.75–6.80 (m, 1H, Ar-H), 7.21 (s, 1H, vinyl-H), 7.28–7.33 (m, 2H, Ar-H), 7.39–7.48 (m, 3H, Ar-H) ppm; 13C NMR (100 MHz, CDCl3) δ 10.5, 16.6, 21.6, 23.9, 34.0, 104.7 (d, JC-F = 23.0 Hz), 110.6 (d, JC-F = 22.0 Hz), 119.1, 124.0 (d, JC-F = 8.0 Hz), 126.5 (2C), 129.4 (2C), 130.2, 131.1, 133.68, 133.78, 138.2, 139.6, 145.47 (d, JC-F = 8.0 Hz), 149.3, 163.0 (d, JC-F = 244.0 Hz) ppm; MS (ESI) m/z 354.2 (M+Na+, 100%); HRMS (ESI) calcd for C23H22FNNa+ [M+Na+]: 354.1628; found: 354.1625.</p><!><p>A solution of compound 3 (Z : E = 2.5 : 1) (140.0 mg, 0.42 mmol), HOBt (72 mL, 0.53 mmol), and EDCI (101 mg, 0.53 mmol) in CH2Cl2 (4 mL) was stirred at room temperature under argon for 1 hr and cooled to 0 °C. Then to the solution was added NH3·H2O (0.1 mL), and stirred for 12 hr at room temperature. The reaction was quenched with 1.0 N citric acid, extracted with CH2Cl2 (5 mL × 3). The combined organic layers were washed with brine, saturated NHCO3, dried over anhydrous Na2SO4, filtered, and concentrated under reduced pressure. The residue was purified by flash chromatography on silica gel (eluent: ethyl acetate: petroleum ether = 1:1) to give compound (Z)-17 as a yellow solid (82 mg, 58%), and compound (E)-18 as a yellow solid (26 mg, 19%). The data for (Z)-17: M.p. 138–139 °C (hexane/EtOAc); IR (film) νmax 3393, 3195, 2960, 2860, 1659, 1601, 1464, 1409, 1272, 1165, 1131, 1052, 1016 cm−1; 1H NMR (400 MHz, CDCl3) δ 1.32 (d, J = 6.9 Hz, 6H, CH(CH3)2), 2.21 (s, 3H, C=CCH3), 2.98 (sept, J = 6.9 Hz, 1H, CH(CH3)2), 3.51 (s, 2H, CH2CO), 5.75 (s, 1H, CONH2), 6.28 (s, 1H, CONH2), 6.57–6.64 (m, 1H, Ar-H), 6.85–6.90 (m, 1H, Ar-H), 7.22 (s, 1H, vinyl-H), 7.28–7.32 (m, 2H, Ar-H), 7.40–7.48 (m, 3H, Ar-H) ppm; 13C NMR (100 MHz, CDCl3) δ 10.5, 23.8, 33.2, 33.9, 105.5 (d, JC-F = 24.0 Hz), 110.8 (d, JC-F = 23.0 Hz), 123.79 (d, JC-F = 9.0 Hz), 126.5 (2C), 129.3 (2C), 129.85, 129.87, 131.2, 133.5, 138.8, 139.5, 145.9 (d, JC-F = 8.0 Hz), 149.3, 163.0 (d, JC-F = 244.0 Hz), 172.4 ppm; MS (ESI) m/z 358.2 (M+Na+, 100%); HRMS (ESI) calcd for C22H22FNNaO+ [M+Na+]: 358.1578; found: 358.1572. The data for (E)-18: M.p. 169–170 °C (hexane/EtOAc); IR (film): νmax 3393, 3192, 2961, 2918, 2872, 1656, 1607, 1461, 1397, 1257, 1147, 1113 cm−1; 1H NMR (400 MHz, CDCl3) δ: 1.29 (d, J = 6.9 Hz, 6H, CH(CH3)2), 1.90 (s, 3H, C=CCH3), 2.96 (sept, J = 6.9 Hz, 1H, CH(CH3)2), 3.50 (s, 2H, CH2CO), 5.60 (s, 1H, CONH2), 6.87 (s, 1H, CONH2), 6.83–6.90 (m, 1H, Ar-H), 6.90–6.95 (m, 1H, Ar-H), 7.23–7.33 (m, 4H, Ar-H), 7.49–7.54 (m, 1H, Ar-H), 7.63 (s, 1H, vinyl-H) ppm; 13C NMR (100 MHz, CDCl3) δ 14.2, 23.9, 33.2, 33.9, 105.4 (d, JC-F = 24.0 Hz), 111.3 (d, JC-F = 23.0 Hz), 119.5 (d, JC-F = 9.0 Hz), 126.2 (2C), 129.6 (2C), 130.3, 133.3, 133.5, 135.3, 136.2, 138.8, 143.07 (d, JC-F = 9.0 Hz), 149.2, 163.16 (d, JC-F = 243.0 Hz), 171.9 ppm; MS (ESI) m/z 358.2 (M+Na+, 100%); HRMS (ESI) calcd for C22H22FNNaO+ [M+Na+]: 358.1578; found: 358.1581.</p><!><p>Following the procedure described for 17 and 18, compounds 19 and 20 were synthesized respectively, and purified by flash column chromatography on silica gel (eluent: ethyl acetate: petroleum ether = 1:2). (Z)-19 is a yellow solid (63%). The data for (Z)-19: M.p. 199–200 °C (hexane/EtOAc); IR (film): νmax 3289, 2957, 2869, 1647, 1601, 1467, 1409, 1262, 1162, 1055 cm−1; 1H NMR (400 MHz, CDCl3) δ 1.31 (d, J = 6.9 Hz, 6H, CH(CH3)2), 2.20 (s, 3H, C=CCH3), 2.76 (d, J = 4.8 Hz, 3H, NCH3), 2.97 (sept, J = 6.9 Hz, 1H, CH(CH3)2), 3.51 (s, 2H, CH2CO), 5.85 (q, J = 4.8 Hz, 1H, CONH), 6.57–6.64 (m, 1H, Ar-H), 6.83–6.88 (m, 1H, Ar-H), 7.22 (s, 1H, vinyl-H), 7.27–7.32 (m, 2H, Ar-H), 7.40–7.48 (m, 3H, Ar-H) ppm; 13C NMR (100 MHz, CDCl3) δ: 10.5, 23.8, 26.4, 33.4, 33.9, 105.5 (d, JC-F = 23.0 Hz), 110.7 (d, JC-F = 22.0 Hz), 123.76 (d, JC-F = 9.0 Hz), 126.5 (2C), 129.3 (2C), 129.85, 129.87, 131.1 (2C), 133.5, 139.0, 139.5, 146.0 (d, JC-F = 9.0 Hz), 149.4, 163.0 (d, JC-F = 244.0 Hz), 169.9 ppm; MS (ESI) m/z 372.2 (M+Na+, 100%); HRMS (ESI) calcd for C23H24FNNaO+ [M+Na+]: 372.1734; found: 372.1733. (E)-20 is a yellow solid (23%). The data for (E)-20: M.p. 135–136 °C (hexane/EtOAc); IR (film): νmax 3296, 2957, 1644, 1598, 1470, 1409, 1202, 1150, 1049, 1013 cm−1; 1H NMR (400 MHz, CDCl3) δ 1.29 (d, J = 6.9 Hz, 6H, CH(CH3)2), 1.89 (s, 3H, C=CCH3), 2.75 (d, J = 4.8 Hz, 3H, NCH3), 2.95 (sept, J = 6.9 Hz, 1H, CH(CH3)2), 3.51 (s, 2H, CH2CO), 5.60 (q, J = 4.8 Hz, 1H, CONH), 6.84–6.92 (m, 2H, Ar-H), 7.23–7.33 (m, 4H, Ar-H), 7.50–7.56 (m, 1H, Ar-H), 7.63 (s, 1H, vinyl-H) ppm; 13C NMR (100 MHz, CDCl3) δ 14.2, 23.9, 26.5, 33.5, 33.9, 105.5 (d, JC-F = 23.0 Hz), 111.35 (d, JC-F = 23.0 Hz), 119.5 (d, JC-F = 9.0 Hz), 126.2 (2C), 129.6 (2C), 130.2, 133.3, 133.5, 135.3, 136.4, 138.8, 143.18 (d, JC-F = 9.0 Hz), 149.2, 163.2 (d, JC-F = 244.0 Hz), 169.7 ppm; MS (ESI) m/z 372.2 (M+Na+, 100%). HRMS (ESI) calcd for C23H24FNNaO+ [M+Na+]: 372.1734; found: 372.1732.</p><!><p>Compound 40 was synthesized according to the general procedure A. White solid, yield: 50%. M.p. 113–114 °C (hexane/EtOAc); IR (film): νmax 3425, 2957, 1668, 1595, 1506, 1424, 1308, 1257, 1223, 1162, 1141 cm−1; 1H NMR (400 MHz, CDCl3) δ 1.22 (t, J = 7.4 Hz, 3H, CH2CH3), 2.56 (q, J = 7.4 Hz, 2H, CH2CH3), 7.06–7.15 (m, 2H, Ar-H), 7.35–7.45 (m, 2H, Ar-H), 7.76 (s, 1H, vinyl-H) ppm; 13C NMR (100 MHz, CDCl3) δ 13.6, 20.5, 115.66 (d, JC-F = 22.0 Hz), 131.3 (d, JC-F = 8.0 Hz), 131.5 (d, JC-F = 4.0 Hz), 133.7, 139.6, 162.8 (d, JC-F = 249.0 Hz), 174.0 ppm; MS (ESI) m/z 193.1 (M−H+); HRMS (ESI) calcd for C11H10FO2− [M−H+]: 193.0670; found: 193.0667.</p><!><p>Compound 41 was synthesized according to the general procedure B. Colorless oil, yield: 90%. IR (film): νmax 3415, 2969, 1705, 1598, 1509, 1458, 1415, 1229, 1149 cm−1; 1H NMR (400 MHz, CDCl3) δ 0.96 (t, J = 7.4 Hz, 3H, CH2CH3), 1.52–1.72 (m, 2H, CH2CH3), 2.52–2.63 (m, 1H, CH2CH), 2.73 (dd, J = 13.8, 6.6 Hz, 1H, CH2CH), 2.93 (dd, J = 13.8, 8.2 Hz, 1H, CH2CH), 6.92–7.00 (m, 2H, Ar-H), 6.09–7.17 (m, 2H, Ar-H), 10.40 (br s, 1H, COOH) ppm; 13C NMR (100 MHz, CDCl3) δ 11.6, 24.8, 36.9, 49.1, 115.2 (d, JC-F = 21.0 Hz) (2C), 130.26 (d, JC-F = 8.0 Hz) (2C), 134.78 (d, JC-F = 3.0 Hz), 162.6 (d, JC-F = 243.0 Hz), 181.7 ppm; MS (ESI) m/z 195.1 (M−H+); HRMS (ESI) calcd for C11H12FO2− [M−H+]: 195.0827; found: 195.0824.</p><!><p>Compound 42 was synthesized according to the general procedure C, and purified by flash column chromatography on silica gel (eluent: ethyl acetate: petroleum ether = 1:40). Colorless oil, yield: 85%. IR (film): νmax 2967, 2930, 2872, 1717, 1613, 1482, 1439, 1287, 1263, 1229, 1162 cm−1; 1H NMR (400 MHz, CDCl3) δ 1.00 (t, J = 7.4 Hz, 3H, CH2CH3), 1.48–1.61 (m, 1H, CH2CH3), 1.90–2.01 (m, 1H, CH2CH3), 2.62–2.70 (m, 1H, CH2CH), 2.78 (dd, J = 17.0, 3.7 Hz, 1H, CH2CH), 3.28 (dd, J = 17.0, 7.8 Hz, 1H, CH2CH), 7.25–7.32 (m, 1H, Ar-H), 7.34–7.38 (m, 1H, Ar-H), 7.39–7.44 (m, 1H, Ar-H) ppm; 13C NMR (100 MHz, CDCl3) δ 11.5, 24.4, 31.7, 49.6, 109.5 (d, JC-F = 22.0 Hz), 122.2 (d, JC-F = 23.0 Hz), 127.85 (d, JC-F = 8.0 Hz), 138.6 (d, JC-F = 7.0 Hz), 149.1, 162.25 (d, JC-F = 246.0 Hz), 207.9 ppm; MS (ESI) m/z 201.1 (M+Na+, 100%); HRMS (ESI) calcd for C11H11FNaO+ [M+Na+]: 201.0686; found: 201.0688.</p><!><p>Compound 43 was synthesized according to the general procedure D, and purified by flash column chromatography on silica gel (eluent: ethyl acetate: petroleum ether = 1:50). Colorless oil, yield: 67%. IR (film): νmax 2970, 2930, 1735, 1613, 1592, 1473, 1372, 1321, 1260, 1150, 1089, 1043 cm−1; 1H NMR (400 MHz, CDCl3) δ 1.18 (t, J = 7.6 Hz, 3H, CH2CH3), 1.25 (t, J = 7.1 Hz, 3H, OCH2CH3), 2.53 (q, J = 7.6 Hz, 2H, CH2CH3), 3.32 (s, 2H, ArCH2C=C), 3.50 (s, 2H, CH2CO), 4.14 (q, J = 7.1 Hz, 2H, OCH2CH3), 6.78–6.85 (m, 1H, Ar-H), 6.97–7.04 (m, 1H, Ar-H), 7.26–7.31 (m, 1H, Ar-H) ppm; 13C NMR (100 MHz, CDCl3) δ 14.0, 14.1, 21.9, 31.5, 39.3, 60.9, 105.9 (d, JC-F = 24.0 Hz), 110.4 (d, JC-F = 23.0 Hz), 123.84 (d, JC-F = 9.0 Hz), 128.8, 137.24 (d, JC-F = 2.0 Hz), 147.86 (d, JC-F = 9.0 Hz), 150.5, 162.4 (d, JC-F = 240.0 Hz), 170.7 ppm; MS (ESI) m/z 271.1 (M+Na+, 100%); HRMS (ESI) calcd for C15H17FNaO2+ [M+Na+]: 271.1105; found: 271.1104.</p><!><p>Compound 21 was synthesized according to the general procedure E, and purified by flash column chromatography on silica gel (eluent: ethyl acetate: petroleum ether = 1:4). Yellow solid, yield: 45%. M.p. 160–161 °C (hexane/EtOAc); IR (film): νmax 3426, 2967, 2939, 2866, 1707, 1598, 1464, 1412, 1299, 1174, 1055 cm−1; 1H NMR (400 MHz, CDCl3) δ 1.22 (t, J = 7.6 Hz, 3H, CH2CH3), 1.32 (d, J = 6.9 Hz, 6H, CH(CH3)2), 2.65 (q, J = 7.6 Hz, 2H, CH2CH3), 2.98 (sept, J = 6.9 Hz, 1H, CH(CH3)2), 3.60 (s, 2H, CH2CO), 6.56–6.63 (m, 1H, Ar-H), 6.89–6.94 (m, 1H, Ar-H), 7.22 (s, 1H, vinyl-H), 7.27–7.32 (m, 2H, Ar-H), 7.38–7.43 (m, 1H, Ar-H), 7.44–7.49 (m, 2H, Ar-H) ppm; 13C NMR (100 MHz, CDCl3) δ 15.6, 18.3, 23.9, 31.3, 34.0, 105.8 (d, JC-F = 23.0 Hz), 110.7 (d, JC-F = 22.0 Hz), 123.96 (d, JC-F = 9.0 Hz), 126.5 (2C), 129.0, 129.5 (2C), 129.9, 130.9, 133.8, 138.1, 145.0, 146.1 (d, JC-F = 9.0 Hz), 149.3, 163.1 (d, JC-F = 244.0 Hz), 176.4 ppm; MS (ESI) m/z 373.2 (M+Na+, 100%); HRMS (ESI) calcd for C23H23FNaO2+ [M+Na+]: 373.1574; found: 373.1574.</p><!><p>To the solution of 6-fluoro-1-indanone 34 (600 mg, 4.0 mmol), KOH (336 mg, 6.0 mmol), and Pd/C (60 mg, 10 %) in EtOH (40 mL), was added i-PrCHO (0.55 mL, 6.0 mmol) at 0 °C. After stirring for 1 hr at room temperature, the catalyst was filtered off and the mixture was acidified with a 6N HCl solution to pH 7.0, concentrated, extracted with EtOAc (20 mL × 3). The combined organic layers were dried over anhydrous Na2SO4, filtered, and concentration under reduced pressure. The residue was purified by flash chromatography on silica gel (eluent: petroleum ether) to afford compound 44 as white solid (511 mg, 62%). M.p. 42–43 °C (hexane/EtOAc); IR (film): νmax 2961, 2930, 2872, 1717, 1607, 1488, 1436, 1281, 1260, 1162, 1034 cm−1; 1H NMR (400 MHz, CDCl3) δ 0.97 (dd, J = 6.2, 1.3 Hz, 6H, CH(CH3)2), 1.24–1.37 (m, 1H, CH2CH(CH3)2), 1.70–1.87 (m, 2H, CH(CH3)2 and CH2CH(CH3)2), 2.71–2.80 (m, 2H, ArCH2CH and ArCH2CH), 3.30 (dd, J = 17.6, 8.6 Hz, 1H, ArCH2CH), 7.26–7.32 (m, 1H, Ar-H), 7.36–7.38 (m, 2H, Ar-H) ppm; 13C NMR (100 MHz, CDCl3) δ 21.7, 23.4, 26.5, 32.8, 40.6, 46.9, 109.6 (d, JC-F = 22.0 Hz), 122.2 (d, JC-F = 24.0 Hz), 127.8 (d, JC-F = 7.0 Hz), 138.4 (d, JC-F = 7.0 Hz), 148.93 (d, JC-F = 2.0 Hz), 162.29 (d, JC-F = 247.0 Hz), 208.26 (d, JC-F = 3.0 Hz) ppm; MS (ESI) m/z 229.1 (M+Na+, 100%); HRMS (ESI) calcd for C13H15FNaO+ [M+Na+]: 229.0999; found: 229.0996.</p><!><p>Compound 45 was synthesized according to the general procedure D, and purified by flash column chromatography on silica gel (eluent: ethyl acetate: petroleum ether = 1:50). Yellow oil, yield: 57%. IR (film): νmax 2957, 1738, 1610, 1586, 1476, 1366, 1263, 1153, 1031 cm−1; 1H NMR (400 MHz, CDCl3) δ 0.94 (d, J = 6.6 Hz, 6H, CH(CH3)2), 1.24 (t, J = 7.1 Hz, 3H, OCH2CH3), 1.85–2.00 (m, 1H, CH(CH3)2), 2.38 (d, J = 7.4 Hz, 2H, CH2CH), 3.31 (s, 2H, ArCH2C=C), 3.51 (s, 2H, CH2COOEt), 4.14 (q, J = 7.1 Hz, 2H, OCH2CH3), 6.78–6.85 (m, 1H, Ar-H), 6.98–7.04 (m, 1H, Ar-H), 7.25–7.30 (m, 1H, Ar-H) ppm; 13C NMR (100 MHz, CDCl3) δ 14.1, 22.7, 28.9, 31.6, 38.1, 40.2, 60.8, 106.0 (d, JC-F = 23.0 Hz), 110.4 (d, JC-F = 22.0 Hz), 123.7 (d, JC-F = 9.0 Hz), 130.32, 130.35, 137.3, 147.79 (d, JC-F = 8.0 Hz), 148.3, 162.4 (d, JC-F = 240.0 Hz), 170.7 ppm; MS (ESI) m/z 299.1 (M+Na+, 100%); HRMS (ESI) calcd for C17H21FNaO2+ [M+Na+]: 299.1418; found: 299.1419.</p><!><p>Compound 22 was synthesized according to the general procedure E, and purified by flash column chromatography on silica gel (eluent: ethyl acetate: petroleum ether = 1:10). Yellow solid, yield: 75%. M.p. 135–136 °C (hexane/EtOAc); IR (film): νmax 3425, 2957, 2927, 2866, 1714, 1604, 1467, 1412, 1278, 1165, 1049 cm−1; 1H NMR (400 MHz, CDCl3) δ 0.80 (d, J = 6.6 Hz, 6H, CH(CH3)2), 1.32 (d, J = 6.9 Hz, 6H, PhCH(CH3)2), 1.88–2.01 (m, 1H, CH(CH3)2), 2.51 (d, J = 7.3 Hz, 2H, CH2CH), 2.98 (sept, J = 6.9 Hz, 1H, PhCH(CH3)2), 3.63 (s, 2H, CH2CO), 6.57–6.64 (m, 1H, Ar-H), 6.88–6.94 (m, 1H, Ar-H), 7.20 (s, 1H, vinyl-H), 7.27–7.33 (m, 2H, Ar-H), 7.36–7.42 (m, 1H, Ar-H), 7.43–7.48 (m, 2H, Ar-H) ppm; 13C NMR (100 MHz, CDCl3) δ 22.8 (2C), 23.9 (2C), 30.4, 31.6, 34.0, 34.2, 105.85 (d, JC-F = 23.0 Hz), 110.66 (d, JC-F = 23.0 Hz), 123.85 (d, JC-F = 9.0 Hz), 126.5 (2C), 129.4 (2C), 129.9, 130.4, 131.4, 133.8, 139.0, 142.4, 145.97 (d, JC-F = 9.0 Hz), 149.3, 163.0 (d, JC-F = 244.0 Hz), 176.5 ppm; MS (ESI) m/z 401.2 (M+Na+, 100%); HRMS (ESI) calcd for C25H27FNaO2+ [M+Na+]: 401.1887; found: 401.1887.</p><!><p>Compound 47a [25] was synthesized according to the general procedure A. Yellow solid, yield: 40%. M.p. 80–82 °C (hexane/EtOAc); IR (film): νmax 3415, 3425, 3040, 1662 (CO), 1609, 1443, 1407,1260 cm−1; 1H-NMR (400 MHz, CDCl3) δ 2.15 (s, 3H, CH=CC H3), 7.30–7.45 (m, 5H, Ar-H), 7.85 (s, 1H, CH=CCH3) ppm; 13C-NMR (100 MHz, CDCl3) δ 1 3.7, 127.6, 128.4, 128.7, 129.8, 135.6, 141.1, 174.3 ppm; MS (ESI) m/z 163 (M+H+).</p><!><p>Compound 48a [25] was synthesized according to the general procedure B. Colorless oil, yield: 100%. IR (film): νmax 3401, 2977, 2664, 1705, 1455, 1292, 1240 cm−1; 1H NMR (400 MHz, CDCl3) δ 1.17 (d, J = 6.9 Hz, 3H, CHCH3), 2.66 (dd, J = 13.3, 8.0 Hz, 1H, CH2CH), 2.71–2.81 (ddq, J = 8.0, 6.3, 6.9 Hz, 1H, CHCH3), 3.07 (dd, J = 13.3, 6.3 Hz, 1H, CH2CH), 7.16–7.31 (m, 5H, Ar-H), 11.03 (br s, 1H, COOH) ppm; 13C NMR (100 MHz, CDCl3) δ 16.4, 39.3, 41.3, 126.4, 128.4, 129.0, 139.0, 182.6 ppm; MS (ESI) m/z 165 (M+H+).</p><!><p>Compound 49a [26] was synthesized according to the general procedure C, and purified by flash column chromatography on silica gel (eluent: ethyl acetate: petroleum ether = 1:40). Yellow oil, yield: 83%. IR (film): νmax 3072, 2929, 1709, 1605, 1462, 1288, 1201 cm−1; 1H NMR (400 MHz, CDCl3) δ 1.31 (d, J = 7.3 Hz, 3H, CHCH3), 2.66–2.76 (m, 2H, CH2CH+CH2CH), 3.40 (dd, J = 17.9, 8.8 Hz, 1H, CH2CH), 7.33–7.78 (m, 4H, Ar-H) ppm; 13C NMR (100 MHz, CDCl3) δ 16.2, 34.9, 41.9, 123.9, 126.5, 127.3, 134.6, 136.3, 153.4, 209.4 ppm; MS (ESI) m/z 169 (M+Na+).</p><!><p>Compound 50a [26] was synthesized according to the general procedure D, and purified by flash column chromatography on silica gel (eluent: ethyl acetate: petroleum ether = 1:50). Yellow oil, yield: 72%. IR (film): νmax 2976, 2903, 1732, 1610, 1470, 1394, 1366, 1308, 1257, 1156, 1037 cm−1; 1H NMR (400 MHz, CDCl3) δ 1.25 (t, J = 7.1 Hz, 3H, OCH2CH3), 2.14 (s, 3H, C=CCH3), 3.35 (s, 2H, ArCH2C=C), 3.54 (s, 2H, CH2COOEt), 4.15 (q, J = 7.1 Hz, 2H, OCH2CH3), 7.11–7.16 (m, 1H, Ar-H), 7.24–7.32 (m, 2H, Ar-H), 7.36–7.40 (m, 1H, Ar-H) ppm; 13C NMR (100 MHz, CDCl3) δ 14.13, 14.16, 31.6, 42.7, 60.7, 118.4, 123.1, 123.9, 126.1, 129.8, 141.95, 142.0, 145.9, 171.0 ppm; MS (ESI) m/z 239.1 (M+Na+, 100%).</p><!><p>Compound 23 was synthesized according to the general procedure E, and purified by flash column chromatography on silica gel (eluent: ethyl acetate: petroleum ether = 1:10). Yellow solid, yield: 78%. M.p. 133–134 °C (hexane/EtOAc); IR (film): νmax 3420, 2964, 1705, 1601, 1452, 1409, 1299, 1217, 1162, 1052, 1016 cm−1; 1H NMR (400 MHz, CDCl3) δ 1.31 (d, J = 6.9 Hz, 6H, CH(CH3)2), 2.20 (s, 3H, C=CCH3), 2.97 (sept, J = 6.9 Hz, 1H, CH(CH3)2), 3.63 (s, 2H, CH2COOH), 6.89–6.95 (m, 1H, Ar-H), 7.14–7.19 (m, 2H, Ar-H), 7.21 (s, 1H, vinyl-H), 7.27–7.31 (m, 2H, Ar-H), 7.45–7.51 (m, 2H, Ar-H) ppm; 13C NMR (100 MHz, CDCl3) δ 10.4, 23.9, 31.3, 34.0, 117.9, 122.8, 124.5, 126.4, 127.7, 129.4, 130.6, 131.0, 134.1, 136.9, 140.7, 143.8, 149.1, 176.1 ppm; MS (ESI) m/z 341.1 (M+Na+, 100%); HRMS (ESI) calcd for C22H22NaO2+ [M+Na+]: 341.1512; found: 341.1512.</p><!><p>Compound 47b was synthesized according to the general procedure A. White solid, yield: 30%. M.p. 163–164 °C (hexane/EtOAc); IR (film): νmax 3410, 2924, 1662, 1601, 1412, 1318, 1263, 1208, 1122 cm−1; 1H NMR (400 MHz, CDCl3) δ 2.16 (s, 3H, C=CCH3), 2.39 (s, 3H, Ar-CH3), 7.19–7.40 (m, 4H, Ar-H), 7.82 (s, 1H, vinyl-H) ppm; 13C NMR (100 MHz, CDCl3) δ 13.8, 21.4, 126.6, 129.2, 130.0, 132.8, 139.0, 141.2, 174.4 ppm; MS (ESI) m/z 175.1 (M−H+); HRMS (ESI) calcd for C11H11O2− [M−H+]: 175.0765; found: 175.0762.</p><!><p>Compound 48b was synthesized according to the general procedure B. Colorless oil yield: 98%. IR (film): νmax 3400, 2979, 1702, 1516, 1461, 1412, 1290, 1241, 1198, 1119, 1037 cm−1; 1H NMR (400 MHz, CDCl3) δ 1.18 (d, J = 6.9 Hz, 3H, CHCH3), 2.33 (s, 3H, Ar-CH3), 2.65 (dd, J = 13.4, 8.0 Hz, 1H, CHCH2), 2.74 (ddq, J = 8.0, 6.4, 6.9 Hz, 1H, CHCH3), 3.04 (dd, J = 13.4, 6.4 1H, CHCH2), 7.06–7.13 (m, 4H, Ar-H) ppm; 13C NMR (100 MHz, CDCl3) δ 16.4, 21.0, 38.9, 41.3, 128.9, 129.1, 135.87, 135.92, 182.7 ppm; MS (ESI) m/z 177.1 (M−H+); HRMS (ESI) calcd for C11H13O2− [M−H+]: 177.0921; found: 177.0919.</p><!><p>Compound 49b was synthesized according to the general procedure C, and purified by flash column chromatography on silica gel (eluent: ethyl acetate: petroleum ether = 1:40). Yellow oil yield: 87%. IR (film): νmax 2960, 2924, 2869, 1710, 1610, 1494, 1281, 1150, 1116, 1034 cm−1; 1H NMR (400 MHz, CDCl3) δ 1.30 (d, J = 7.3 Hz, 3H, CHCH3), 2.40 (s, 3H, Ar-CH3), 2.67 (dd, J = 17.0, 3.9 Hz, 1H, CHCH2), 2.67–2.75 (m, 1H, CHCH3), 3.34 (dd, J = 17.0, 7.9 Hz, 1H, CHCH2), 7.30–7.35 (m, 1H, Ar-H), 7.38–7.42 (m, 1H, Ar-H), 7.54–7.57 (m, 1H, Ar-H) ppm; 13C NMR (100 MHz, CDCl3) δ 16.3, 21.1, 34.6, 42.3, 123.9, 126.2, 135.9, 136.5, 137.2, 150.8, 209.6 ppm; MS (ESI) m/z 183.1 (M+Na+, 100%); HRMS (ESI) calcd for C11H12NaO+ [M+Na+]: 183.0780; found: 183.0772.</p><!><p>Compound 50b was synthesized according to the general procedure D, and purified by flash column chromatography on silica gel (eluent: ethyl acetate: petroleum ether = 1:50). Colorless oil yield: 79%. IR (film) νmax 2982, 2915, 1735, 1616, 1479, 1366, 1253, 1153, 1037 cm−1; 1H NMR (400 MHz, CDCl3) δ 1.26 (t, J = 7.1 Hz, 3H, OCH2CH3), 2.12 (s, 3H, C=CCH3), 2.39 (s, 3H, Ar-CH3), 3.30 (s, 2H, CH2C=C), 3.52 (s, 2H, CH2COOEt), 4.15 (q, J = 7.1 Hz, 2H, OCH2CH3), 6.92–6.97 (m, 1H, Ar-H), 7.09–7.13 (m, 1H, Ar-H), 7.23–7.27 (m, 1H, Ar-H) ppm; 13C NMR (100 MHz, CDCl3) δ 14.2 (2C), 21.5, 31.6, 42.3, 60.7, 119.2, 122.8, 124.6, 129.7, 135.7, 139.1, 142.2, 146.1, 171.1 ppm; MS (ESI) m/z 253.1 (M+Na+, 100%); HRMS (ESI) calcd for C15H18NaO2+ [M+Na+]: 253.1199; found: 253.1199.</p><!><p>Compound 24 was synthesized according to the general procedure E, and purified by flash column chromatography on silica gel (eluent: ethyl acetate: petroleum ether = 1:10). Yellow solid, yield: 70%. M.p. 143–144 °C (hexane/EtOAc); IR (film): νmax 3416, 2961, 2924, 1705, 1607, 1467, 1415, 1311, 1159, 1046 cm−1; 1H NMR (400 MHz, CDCl3) δ 1.31 (d, J = 6.9 Hz, 6H, CH(CH3)2), 2.19 (s, 3H, C=CCH3), 2.33 (s, 3H, Ar-CH3), 2.97 (sept, J = 6.9 Hz, 1H, CH(CH3)2), 3.61 (s, 2H, CH2COOH), 6.71–6.76 (m, 1H, Ar-H), 6.97–7.02 (m, 1H, Ar-H), 7.14 (s, 1H, vinyl-H), 7.27 (m, 2H, Ar-H), 7.34–7.39 (m, 1H, Ar-H), 7.44–7.50 (m, 2H, Ar-H) ppm; 13C NMR (100 MHz, CDCl3) δ 10.5, 21.7, 23.9, 31.3, 34.0, 118.8, 122.6, 125.1, 126.4, 129.5, 130.0, 130.4, 131.5, 134.3, 137.2, 137.7, 140.6, 144.1, 148.9, 176.3 ppm; MS (ESI) m/z 355.2 (M+Na+, 100%); HRMS (ESI) calcd for C23H24NaO2+ [M+Na+]: 355.1699; found: 355.1664.</p><!><p>Compound 47c [25] was synthesized according to the general procedure A. White solid, yield: 32%. M.p. 154–155 °C (hexane/EtOAc); IR (film): νmax 3390, 2945, 2836. 1662, 1598, 1513, 1424, 1281, 1257, 1177, 1037 cm−1; 1H NMR (400 MHz, CDCl3) δ 2.16 (s, 3H, C=CCH3), 3.85 (s, 3H, OCH3), 6.90–7.00 (m, 2H, Ar-H), 7.40–7.48 (m, 2H, Ar-H), 7.79 (s, 1H, vinyl-H) ppm; 13C NMR (100 MHz, CDCl3) δ 13.7, 55.3, 113.9, 125.1, 128.2, 131.7, 140.8, 160.0, 174.5 ppm; MS (ESI) m/z 215.1 (M+Na+, 100%).</p><!><p>Compound 48c [25] was synthesized according to the general procedure B. Colorless oil, yield: 99%. IR (film): νmax 3380, 2933, 1711, 1613, 1513, 1461, 1247, 1299, 1183, 1116, 1037 cm−1; 1H NMR (400 MHz, CDCl3) δ 1.17 (d, J = 6.9 Hz, 3H, CHCH3), 2.63 (dd, J = 13.4, 7.9 Hz, 1H, CHCH2), 2.72 (ddq, J = 7.9, 6.3, 6.9 Hz, 1H, CHCH3), 3.02 (dd, J = 13.4, 6.3 1H, CHCH2), 3.79 (s, 3H, OCH3), 6.81–6.87 (m, 2H, Ar-H), 7.08–7.14 (m, 2H, Ar-H) ppm; 13C NMR (100 MHz, CDCl3) δ 16.4, 38.4, 41.5, 55.2, 113.8, 129.9, 131.1, 158.1, 182.6 ppm; MS (ESI) m/z 217.1 (M+Na+, 100%).</p><!><p>Compound 49c [27] was synthesized according to the general procedure C, and purified by flash column chromatography on silica gel (eluent: ethyl acetate: petroleum ether = 1:40). Colorless oil, yield: 50%. IR (film): νmax 2961, 2927, 1708, 1619, 1488, 1436, 1278, 1244, 1171, 1028 cm−1; 1H NMR (400 MHz, CDCl3) δ 1.30 (d, J = 7.4 Hz, 3H, CHCH3), 2.64 (dd, J = 16.6, 3.6 Hz, 1H, CHCH2), 2.68–2.78 (m, 1H, CHCH3), 3.32 (dd, J = 16.6, 7.6 Hz, 1H, CHCH2), 3.83 (s, 3H, OCH3), 7.15–7.20 (m, 2H, Ar-H), 7.31–7.35 (m, 1H, Ar-H) ppm; 13C NMR (100 MHz, CDCl3) δ 16.3, 34.3, 42.8, 55.6, 105.1, 124.1, 127.2, 137.4, 146.2, 159.4, 209.5 ppm; MS (ESI) m/z 199.1 (M+Na+, 100%).</p><!><p>Compound 50c [27] was synthesized according to the general procedure D, and purified by flash column chromatography on silica gel (eluent: ethyl acetate: petroleum ether = 1:50). White solid, yield: 75%. M.p. 36–37 °C (hexane/EtOAc); IR (film): νmax 2985, 2933, 2908, 1729, 1619, 1583, 1479, 1284, 1244, 1202, 1153, 1092, 1040 cm−1; 1H NMR (400 MHz, CDCl3) δ 1.25 (t, J = 7.1 Hz, 3H, OCH2CH3), 2.12 (s, 3H, C=CCH3), 2.29 (s, 2H, ArCH2C=C), 3.50 (s, 2H, CH2COOEt), 3.83 (s, 3H, OCH3), 4.14 (q, J = 7.1 Hz, 2H, OCH2CH3), 6.66–6.71 (m, 1H, Ar-H), 6.85–6.88 (m, 1H, Ar-H), 7.22–7.26 (m, 1H, Ar-H) ppm; 13C NMR (100 MHz, CDCl3) δ 14.19, 14.27, 31.6, 42.0, 55.5, 60.8, 104.5, 109.4, 123.5, 129.7, 134.1, 143.6, 147.4, 158.9, 171.0 ppm; MS (ESI) m/z 269.1 (M+Na+, 100%).</p><!><p>Compound 25 was synthesized according to the general procedure E, and purified by flash column chromatography on silica gel (eluent: ethyl acetate: petroleum ether = 1:10). Yellow solid, yield: 76%. M.p. 117–119 °C (hexane/EtOAc); IR (film) νmax 3411, 2957, 1705, 1598, 1473, 1214, 1162, 1037 cm−1; 1H NMR (400 MHz, CDCl3) δ 1.31 (d, J = 6.9 Hz, 6H, CH(CH3)2), 2.19 (s, 3H, C=CCH3), 2.97 (sept, J = 6.9 Hz, 1H, CH(CH3)2), 3.60 (s, 2H, CH2COOH), 3.79 (s, 3H, OCH3), 6.41–6.47 (m, 1H, Ar-H), 6.74–6.77 (m, 1H, Ar-H), 7.10 (s, 1H, vinyl-H), 7.26–7.29 (m, 2H, Ar-H), 7.37–7.41 (m, 1H, Ar-H), 7.44–7.49 (m, 2H, Ar-H) ppm; 13C NMR (100 MHz, CDCl3) δ 10.6, 23.9, 31.3, 34.0, 55.4, 104.7, 108.9, 123.6, 126.4, 126.9, 129.1, 129.5, 130.1, 134.3, 138.4, 140.2, 145.8, 148.9, 160.0, 175.7 ppm; MS (ESI) m/z 371.2 (M+Na+, 100%); HRMS (ESI) calcd for C23H24NaO3+ [M+Na+]: 371.1618; found: 371.1620.</p><!><p>Compound 47d was synthesized according to the general procedure A. White solid, yield: 41%. M.p. 126–127 °C (hexane/EtOAc); IR (film): νmax 3411, 2957, 1671, 1607, 1424, 1314, 1269, 1180, 1132 cm−1; 1H NMR (400 MHz, CDCl3) δ 1.28 (t, J = 7.6 Hz, 3H, CH2CH3), 2.18 (s, 3H, C=CCH3), 2.70 (q, J = 7.6 Hz, 2H, CH2CH3), 7.23–7.30 (m, 2H, Ar-H), 7.37–7.44 (m, 2H, Ar-H), 7.85 (s, 1H, vinyl-H) ppm; 13C NMR (100 MHz, CDCl3) δ 13.7, 15.3, 28.7, 126.6, 128.0, 130.1, 133.0, 141.2, 145.2, 174.6 ppm; MS (ESI) m/z 189.1 (M−H+); HRMS (ESI) calcd for C12H13O2− [M−H+]: 189.0921; found: 189.0919.</p><!><p>Compound 48d was synthesized according to the general procedure B. Colorless oil, yield: 100%. IR (film): νmax 3401, 2967, 2933, 1708, 1595, 1516, 1464, 1418, 1293, 1238, 1196, 1122 cm−1; 1H NMR (400 MHz, CDCl3) δ 1.18 (d, J = 6.9 Hz, 3H, CHCH3), 1.23 (t, J = 7.6 Hz, 3H, CH2CH3), 2.63 (q, J = 7.6 Hz, 2H, CH2CH3), 2.65 (dd, J = 13.4, 8.2 Hz, 1H, CHCH2), 2.74 (ddq, J = 8.2, 6.3, 6.9 Hz, 1H, CHCH3), 3.06 (dd, J = 13.4, 6.3 1H, CHCH2), 7.08–7.16 (m, 4H, Ar-H) ppm; 13C NMR (100 MHz, CDCl3) δ 15.5, 16.5, 28.4, 38.9, 41.3, 127.9, 128.9, 136.2, 142.3, 182.4 ppm; MS (ESI) m/z 191.1 (M−H+); HRMS (ESI) calcd for C12H15O2− [M−H+]: 191.1078; found: 191.1074.</p><!><p>Compound 49d [28] was synthesized according to the general procedure C, and purified by flash column chromatography on silica gel (eluent: ethyl acetate: petroleum ether = 1:40). Yellow oil, yield: 91%. IR (film): νmax 2967, 2930, 2872, 1711, 1613, 1494, 1452, 1275, 1153, 1122 cm−1; 1H NMR (400 MHz, CDCl3) δ 1.25 (t, J = 7.6 Hz, 3H, CH2CH3), 1.30 (d, J = 7.3 Hz, 3H, CHCH3), 2.64–2.76 (m, 4H, CHCH3, CHCH2, CH2CH3), 3.35 (dd, J = 17.2, 8.0 Hz, 1H, CHCH2), 7.33–7.39 (m, 1H, Ar-H), 7.40–7.46 (m, 1H, Ar-H), 7.56–7.61 (s, 1H, Ar-H) ppm; 13C NMR (100 MHz, CDCl3) δ: 15.5, 16.3, 28.5, 34.6, 42.3, 122.6, 126.3, 135.0, 136.5, 143.7, 151.1, 209.6 ppm; MS (ESI) m/z 197.1 (M+Na+, 100%).</p><!><p>Compound 50d was synthesized according to the general procedure D, and purified by flash column chromatography on silica gel (eluent: ethyl acetate: petroleum ether = 1:50). Colorless oil, yield: 84%. IR (film): νmax 2957, 2924, 1732, 1613, 1479, 1366, 1305, 1263, 1144, 1034 cm−1; 1H NMR (400 MHz, CDCl3) δ 1.25 (t, J = 7.1 Hz, 3H, OCH2CH3), 1.26 (t, J = 7.6 Hz, 3H, CH2CH3), 2.12 (s, 3H, C=CCH3), 2.69 (q, J = 7.6 Hz, 2H, CH2CH3), 3.30 (s, 2H, ArCH2C=C), 3.52 (s, 2H, CH2COOEt), 4.15 (q, J = 7.1 Hz, 2H, OCH2CH3), 6.94–7.00 (m, 1H, Ar-H), 7.11–7.16 (m, 1H, Ar-H), 7.26–7.30 (m, 1H, Ar-H) ppm; 13C NMR (100 MHz, CDCl3) δ 14.2 (2C), 16.2, 29.1, 31.7, 42.4, 60.7, 118.0, 123.0, 123.6, 129.8, 139.4, 142.2, 142.4, 146.2, 171.1 ppm; MS (ESI) m/z 267.1 (M+Na+, 100%); HRMS (ESI) calcd for C16H20NaO2+ [M+Na+]: 267.1356; found: 267.1352.</p><!><p>Compound 26 was synthesized according to the general procedure E, and purified by flash column chromatography on silica gel (eluent: ethyl acetate: petroleum ether = 1:10). Yellow solid, yield: 76%. M.p. 111–112 °C (hexane/EtOAc); IR (film): νmax 3413, 2957, 2933, 2872, 1705, 1604, 1506, 1467, 1409, 1293, 1055 cm−1; 1H NMR (400 MHz, CDCl3) δ 1.23 (t, J = 7.6 Hz, 3H, CH2CH3), 1.32 (d, J = 6.9 Hz, 6H, CH(CH3)2), 2.20 (s, 3H, C=CCH3), 2.64 (q, J = 7.6 Hz, 2H, CH2CH3), 2.98 (sept, J = 6.9 Hz, 1H, CH(CH3)2), 3.63 (s, 2H, CH2COOH), 6.75–6.80 (m, 1H, Ar-H), 7.01–7.06 (m, 1H, Ar-H), 7.16 (s, 1H, vinyl-H), 7.27–7.31 (m, 2H, Ar-H), 7.38–7.43 (m, 1H, Ar-H), 7.46–7.51 (m, 2H, Ar-H) ppm; 13C NMR (100 MHz, CDCl3) δ: 10.5, 15.7, 23.9, 29.1, 31.4, 34.0, 117.6, 122.7, 124.0, 126.4, 129.5, 130.0, 130.5, 131.7, 134.3, 137.2, 140.7, 144.1, 144.2, 148.9, 176.9 ppm; MS (ESI) m/z 369.2 (M+Na+, 100%); HRMS (ESI) calcd for C24H26NaO2+ [M+Na+]: 369.1825; found: 369.1818.</p><!><p>Compound 47e was synthesized according to the general procedure A. White solid, yield: 30%. M.p. 166–167 °C (hexane/EtOAc); IR (film): νmax 3385, 2930, 1671, 1601, 1509, 1424, 1256, 1180, 1122, 1043 cm−1; 1H NMR (400 MHz, CDCl3) δ 1.44 (t, J = 7.0 Hz, 3H, OCH2CH3), 2.16 (d, J = 1.3 Hz, 3H, C=CCH3), 4.07 (q, J = 7.0 Hz, 2H, OCH2CH3), 6.91–6.95 (m, 2H, Ar-H), 7.39–7.45 (m, 2H, Ar-H), 7.78 (q, J = 1.3 Hz, 1H, vinyl-H) ppm; 13C NMR (100 MHz, CDCl3) δ 13.8, 14.8, 63.5, 114.4, 124.9, 128.1, 131.8, 140.8, 159.4, 174.0 ppm; MS (ESI) m/z 205.1 ([M−H+); HRMS (ESI) calcd for C12H13O3−[M−H+]: 205.0870; found: 205.0866.</p><!><p>Compound 48e was synthesized according to the general procedure B. White solid, yield: 98%. M.p. 56–57 °C (hexane/EtOAc); IR (film) νmax 3378, 2973, 2933, 1705, 1610, 1509, 1378, 1244, 1177, 1119, 1046 cm−1; 1H NMR (400 MHz, CDCl3) δ 1.17 (d, J = 6.9 Hz, 3H, CHCH3), 1.41 (t, J = 7.0 Hz, 3H, OCH2CH3), 2.62 (dd, J = 13.4, 7.9 Hz, 1H, CHCH2), 2.72 (ddq, J = 7.9, 6.4, 6.9 Hz, 1H, CHCH3), 3.02 (dd, J = 13.4, 6.4, 1H, CHCH2), 4.02 (q, J = 7.0 Hz, 2H, OCH2CH3), 6.80–6.87 (m, 2H, Ar-H), 7.06–7.13 (m, 2H, Ar-H), 10.33 (br s, 1H, COOH) ppm; 13C NMR (100 MHz, CDCl3) δ 14.8, 16.4, 38.5, 41.5, 63.3, 114.4, 129.9, 130.9, 157.5, 182.6 ppm; MS (ESI) m/z 207.1 (M−H+); HRMS (ESI) calcd for C12H15O3− [M−H+]: 207.1027; found: 207.1023.</p><!><p>Compound 49e was synthesized according to the general procedure C, and purified by flash column chromatography on silica gel (eluent: ethyl acetate: petroleum ether = 1:40). White solid, yield: 50%. M.p. 49–50 °C (hexane/EtOAc); IR (film): νmax 2969, 2927, 1705, 1616, 1491, 1446, 1278, 1241, 1174, 1116, 1040 cm−1; 1H NMR (400 MHz, CDCl3) δ 1.30 (d, J = 7.4 Hz, 3H, CHCH3), 1.41 (t, J = 7.0 Hz, 3H, OCH2CH3), 2.64 (dd, J = 16.6, 3.6 Hz, 1H, CHCH2), 2.67–2.78 (m, 1H, CHCH3), 3.31 (dd, J = 16.6, 7.6 Hz, 1H, CHCH2), 4.05 (q, J = 7.0 Hz, 2H, OCH2CH3), 7.14–7.19 (m, 2H, Ar-H), 7.29–7.34 (m, 1H, Ar-H) ppm; 13C NMR (100 MHz, CDCl3) δ 14.7, 16.3, 34.2, 42.7, 63.8, 105.7, 124.5, 127.2, 137.4, 146.1, 158.7, 209.5 ppm; MS (ESI) m/z 213.1 (M+Na+, 100%); HRMS (ESI) calcd for C12H14NaO2+[M+Na+]: 213.0886; found: 213.0886.</p><!><p>Compound 50e was synthesized according to the general procedure D, and purified by flash column chromatography on silica gel (eluent: ethyl acetate: petroleum ether = 1:50). White solid, yield: 82%. M.p. 37–38 °C (hexane/EtOAc); IR (film): νmax 2979, 2909, 1732, 1607, 1467, 1394, 1259, 1208, 1153, 1086, 1034 cm−1; 1H NMR (400 MHz, CDCl3) δ 1.24 (t, J = 7.1 Hz, 3H, COOCH2CH3), 1.42 (t, J = 7.0 Hz, 3H, OCH2CH3), 2.12 (s, 3H, C=CCH3), 3.27 (s, 2H, ArCH2C=C), 3.49 (s, 2H, CH2COOEt), 4.05 (q, J = 7.0 Hz, 2H, OCH2CH3), 4.14 (q, J = 7.1 Hz, 2H, COOCH2CH3), 6.65–6.69 (m, 1H, Ar-H), 6.84–6.87 (m, 1H, Ar-H), 7.21–7.24 (m, 1H, Ar-H) ppm; 13C NMR (100 MHz, CDCl3) δ 14.19, 14.27, 14.9, 31.6, 42.0, 60.7, 63.6, 105.3, 110.1, 123.5, 129.8, 134.1, 143.5, 147.3, 158.2, 171.0 ppm; MS (ESI) m/z 283.1 (M+Na+, 100%); HRMS (ESI) calcd for C16H20NaO3+ [M+Na+]: 283.1305; found: 283.1303.</p><!><p>Compound 27 was synthesized according to the general procedure E, and purified by flash column chromatography on silica gel (eluent: ethyl acetate: petroleum ether = 1:10). Yellow solid, yield: 67%. M.p. 146–147 °C (hexane/EtOAc); IR (film) νmax 3410, 2957, 2930, 1705, 1610, 1464, 1385, 1211, 1156, 1119, 1040 cm−1; 1H NMR (400 MHz, CDCl3) δ 1.31 (d, J = 6.9 Hz, 6H, CH(CH3)2), 1.39 (t, J = 7.0 Hz, 3H, OCH2CH3), 2.19 (s, 3H, C=CCH3), 2.97 (sept, J = 6.9 Hz, 1H, CH(CH3)2), 3.59 (s, 2H, CH2COOH), 4.02 (q, J = 7.0 Hz, 2H, OCH2CH3), 6.40–6.46 (m, 1H, Ar-H), 6.72–6.78 (m, 1H, Ar-H), 7.09 (s, 1H, vinyl-H), 7.24–7.30 (m, 2H, Ar-H), 7.36–7.40 (m, 1H, Ar-H), 7.43–7.49 (m, 2H, Ar-H) ppm; 13C NMR (100 MHz, CDCl3) δ 10.5, 14.8, 23.9, 31.3, 34.0, 63.5, 105.3, 109.5, 123.6, 126.4, 126.7, 128.9, 129.5, 130.1, 134.3, 138.3, 140.2, 145.7, 148.8, 159.3, 176.1 ppm; MS (ESI) m/z 385.2 (M+Na+, 100%); HRMS (ESI) calcd for C24H26NaO3+ [M+Na+]: 385.1774; found: 385.1777.</p><!><p>Compound 47f was synthesized according to the general procedure A. White solid, yield: 30%. M.p. 86–87 °C (hexane/EtOAc); IR (film) νmax 3380, 2961, 1677, 1619, 1421, 1360, 1272, 1217, 1122, 1046 cm−1; 1H NMR (400 MHz, CDCl3) δ 1.28 (d, J = 6.9 Hz, 6H, CH(CH3)2), 2.17 (d, J = 1.3 Hz, 3H, C=CCH3), 2.95 (sept, J = 6.9 Hz, 1H, CH(CH3)2), 7.26–7.31 (m, 2H, Ar-H), 7.38–7.43 (m, 2H, Ar-H), 7.83 (q, J = 1.3 Hz, 1H, vinyl-H) ppm; 13C NMR (100 MHz, CDCl3) δ: 13.8, 23.8, 34.0, 126.6, 130.1, 130.4, 133.1, 141.1, 149.8, 174.4 ppm; MS (ESI) m/z 203.1 (M−H+); HRMS (ESI) calcd for C13H15O2−[M−H+]: 203.1078; found: 203.1075.</p><!><p>Compound 48f was synthesized according to the general procedure B. Colorless oil, yield: 97%. IR (film): νmax 3365, 2957, 1705, 1513, 1461, 1418, 1284, 1238, 1193, 1116, 1052 cm−1; 1H NMR (400 MHz, CDCl3) δ 1.19 (d, J = 6.9 Hz, 3H, CHCH3), 1.25 (d, J = 6.9 Hz, 6H, CH(CH3)2), 2.64 (dd, J = 13.4, 7.1 Hz, 1H, CHCH2), 2.76 (ddq, J = 7.1, 6.2, 6.9 Hz, 1H, CHCH3), 2.89 (sept, J = 6.9 Hz, 1H, CH(CH3)2), 3.08 (dd, J = 13.4, 6.2, 1H, CHCH2), 7.10–7.19 (m, 4H, Ar-H), 9.70 (br s, 1H, COOH) ppm; 13C NMR (100 MHz, CDCl3) δ 16.5, 24.0, 33.7, 38.9, 41.2, 126.4, 128.9, 136.3, 146.9, 182.6 ppm; MS (ESI) m/z 205.1 (M−H+); HRMS (ESI) calcd for C13H17O2− [M−H+]: 205.1234; found: 205.1230.</p><!><p>Compound 49f was synthesized according to the general procedure C, and purified by flash column chromatography on silica gel (eluent: ethyl acetate: petroleum ether = 1:40). Yellow oil, yield: 87%. IR (film): νmax 2960, 2927, 2866, 1708, 1622, 1494, 1436, 1259, 1174, 1119 cm−1; 1H NMR (400 MHz, CDCl3) δ 1.26 (d, J = 6.9 Hz, 6H, CH(CH3)2), 1.30 (d, J = 7.3 Hz, 3H, CHCH3), 2.64–2.76 (m, 2H, CHCH2, CHCH3), 2.97 (sept, J = 6.9 Hz, 1H, CH(CH3)2), 3.35 (dd, J = 17.0, 7.9 Hz, 1H, CHCH2), 7.34–7.39 (m, 1H, Ar-H), 7.45–7.49 (m, 1H, Ar-H), 7.61–7.64 (m, 1H, Ar-H) ppm; 13C NMR (100 MHz, CDCl3) δ 16.3, 23.9, 33.9, 34.6, 42.3, 121.1, 126.3, 133.8, 136.5, 148.5, 151.3, 209.7 ppm; MS (ESI) m/z 211.1 (M+Na+, 100%). HRMS (ESI) calcd for C13H16NaO+ [M+Na+]: 211.1093; found: 211.1095.</p><!><p>Compound 50f was synthesized according to the general procedure D, and purified by flash column chromatography on silica gel (eluent: ethyl acetate: petroleum ether = 1:50). Colorless oil, yield: 82%. IR (film): νmax 2954, 1741, 1610, 1482, 1366, 1302, 1253, 1156, 1031 cm−1; 1H NMR (400 MHz, CDCl3) δ 1.28 (t, J = 7.2 Hz, 3H, OCH2CH3), 1.32 (d, J = 6.8 Hz, 6H, CH(CH3)2), 2.16 (s, 3H, C=CCH3), 2.99 (sept, J = 6.8 Hz, 1H, CH(CH3)2), 3.33 (s, 2H, ArCH2C=C), 3.57 (s, 2H, CH2COOEt), 4.18 (q, J = 7.2 Hz, 2H, OCH2CH3), 7.02–7.07 (m, 1H, Ar-H), 7.18–7.24 (m, 1H, Ar-H), 7.30–7.36 (m, 1H, Ar-H) ppm; 13C NMR (100 MHz, CDCl3) δ 14.2, 24.3 (2C), 31.7, 34.3, 42.3, 60.7, 116.5, 122.1, 122.9, 129.8, 139.6, 142.2, 146.1, 147.0, 171.1 ppm; MS (ESI) m/z 281.2 (M+Na+, 100%); HRMS (ESI) calcd for C17H22NaO2+ [M+Na+]: 281.1512; found: 281.1517.</p><!><p>Compound 28 was synthesized according to the general procedure E, and purified by flash column chromatography on silica gel (eluent: ethyl acetate: petroleum ether = 1:10). Yellow solid, yield: 78%. M.p. 125–126 °C (hexane/EtOAc); IR (film) νmax 3405, 2954, 1714, 1607, 1507, 1467, 1406, 1299, 1213, 1052, 1013 cm−1; 1H NMR (400 MHz, CDCl3) δ 1.24 (d, J = 6.9 Hz, 6H, CH(CH3)2), 1.32 (d, J = 6.9 Hz, 6H, CH(CH3)2), 2.19 (s, 3H, C=CCH3), 2.89 (sept, J = 6.9 Hz, 1H, CH(CH3)2), 2.97 (sept, J = 6.9 Hz, 1H, CH(CH3)2), 3.63 (s, 2H, CH2COOH), 6.78–6.83 (m, 1H, Ar-H), 7.03–7.08 (m, 1H, Ar-H), 7.15 (s, 1H, vinyl-H), 7.26–7.31 (m, 2H, Ar-H), 7.38–7.42 (m, 1H, Ar-H), 7.45–7.50 (m, 2H, Ar-H) ppm; 13C NMR (100 MHz, CDCl3) δ 10.5, 23.9, 24.0, 31.4, 34.0, 34.3, 116.3, 122.4, 122.7, 126.4, 129.4, 130.0, 130.6, 131.9, 134.3, 137.1, 140.7, 144.1, 148.88, 148.92, 176.6 ppm; MS (ESI) m/z 383.2 (M+Na+, 100%). HRMS (ESI) calcd for C25H28NaO2+ [M+Na+]: 383.1982; found: 383.1982.</p><!><p>Compound 47g [25] was synthesized according to the general procedure A. White solid, yield: 55%. M.p. 164–165 °C (hexane/EtOAc); IR (film): νmax 3360, 2957, 2826, 1671, 1491, 1446, 1424, 1308, 1287, 1263, 1214, 1132, 1092, 1012 cm−1; 1H NMR (400 MHz, CDCl3) δ 2.13 (d, J = 1.4 Hz, C=CCH3), 7.35–7.41 (m, 4H, Ar-H), 7.79 (q, J = 1.4 Hz, 1H, vinyl-H) ppm; 13C NMR (100 MHz, CDCl3) δ 13.7, 128.1, 128.7, 131.1, 134.0, 134.7, 139.7, 173.7 ppm; MS (ESI) m/z 197.0 (M+Na+, 100%).</p><!><p>A mixture of the acrylic acid 47g (1.5 g, 7.63 mmol) and 10 % Pd/C (150 mg) in EtOAc (30 mL) was hydrogenated under 3 atm of hydrogen for 12 h. The catalyst was filtered off and the filtrate concentrated to afford compound 48g as a white solid (1.46 g, 96%). M.p. 52–53 °C (hexane/EtOAc); IR (film): νmax 3351, 2976, 2933, 1698, 1494, 1461, 1406, 1296, 1232, 1196, 1089, 1019 cm−1; 1H NMR (400 MHz, CDCl3) δ 1.19 (d, J = 6.8 Hz, 3H, CHCH3), 2.66 (dd, J = 13.3, 7.6 Hz, 1H, CHCH2), 2.74 (ddq, J = 7.6, 6.5, 6.8 Hz, 1H, CHCH3), 3.02 (dd, J = 13.3, 6.5, 1H, CHCH2), 7.09–7.14 (m, 2H, Ar-H), 7.24–7.28 (m, 2H, Ar-H) ppm; 13C NMR (100 MHz, CDCl3) δ: 16.5, 38.6, 41.1, 128.5, 130.3, 132.3, 137.4, 182.2 ppm; MS (ESI) m/z 221.0 (M+Na+, 100%).</p><!><p>Compound 49g was synthesized according to the general procedure C, and purified by flash column chromatography on silica gel (eluent: ethyl acetate: petroleum ether = 1:40). White solid, yield: 78%. M.p. 62–64 °C (hexane/EtOAc); IR (film): νmax 3079, 2967, 1930, 1774, 1704, 1604, 1455, 1433, 1256, 1235, 1193, 1104, 1064 cm−1; 1H NMR (400 MHz, CDCl3) δ 1.30 (d, J = 7.3 Hz, 3H, CHCH3), 2.68 (dd, J = 16.8, 3.9 Hz, 1H, CHCH2), 2.71–2.79 (m, 1H, CHCH3), 3.36 (dd, J = 16.8, 7.6, 1H, CHCH2), 7.35–7.40 (m, 1H, Ar-H), 7.50–7.55 (m, 1H, Ar-H), 7.67–7.71 (m, 1H, Ar-H) ppm; 13C NMR (100 MHz, CDCl3) δ 16.2, 34.5, 42.6, 123.8, 127.8, 133.7, 134.7, 137.9, 151.5, 208.0 ppm; MS (ESI) m/z 203.0 (M+Na+, 100%).</p><!><p>Compound 50g was synthesized according to the general procedure D, and purified by flash column chromatography on silica gel (eluent: ethyl acetate: petroleum ether = 1:50). Colorless oil, yield: 67%. IR (film): νmax 2976, 2906, 1732, 1601, 1464, 1363, 1305, 1256, 1156, 1098, 1074, 1034 cm−1; 1H NMR (400 MHz, CDCl3) δ 1.27 (t, J = 7.1 Hz, 3H, OCH2CH3), 2.14 (s, 3H, C=CCH3), 3.32 (s, 2H, ArCH2C=C), 3.50 (s, 2H, CH2COOEt), 4.17 (q, J = 7.1 Hz, 2H, OCH2CH3), 7.08–7.13 (m, 1H, Ar-H), 7.24–7.29 (m, 2H, Ar-H) ppm; 13C NMR (100 MHz, CDCl3) δ 14.15, 14.2, 31.4, 42.3, 60.9, 118.8, 123.7, 124.0, 129.4, 132.2, 140.2, 144.1, 147.7, 170.7 ppm; MS (ESI) m/z 273.1 (M+Na+, 100%); HRMS (ESI) calcd for C14H15ClNaO2+ [M+Na+]: 273.0653; found: 273.0659.</p><!><p>Compound 29 was synthesized according to the general procedure E, and purified by flash column chromatography on silica gel (eluent: ethyl acetate: petroleum ether = 1:10). Yellow solid, yield: 77%. M.p. 182–183 °C (hexane/EtOAc); IR (film): νmax 3392, 2961, 2930, 1705, 1595, 1452, 1412, 1308, 1217, 1086, 1022 cm−1; 1H NMR (400 MHz, CDCl3) δ 1.31 (d, J = 6.9 Hz, 6H, CH(CH3)2), 2.21 (s, 3H, C=CCH3), 2.97 (sept, J = 6.9 Hz, 1H, CH(CH3)2), 3.60 (s, 2H, CH2COOH), 6.86–6.90 (m, 1H, Ar-H), 7.14–7.16 (m, 1H, Ar-H), 7.23 (s, 1H, vinyl-H), 7.27–7.31 (m, 2H, Ar-H), 7.35–7.39 (m, 1H, Ar-H), 7.42–7.47 (m, 2H, Ar-H) ppm; 13C NMR (100 MHz, CDCl3) δ 10.5, 23.9, 31.2, 34.0, 118.3, 123.6, 124.1, 126.6, 129.4, 129.7, 131.9, 132.3, 133.5, 133.7, 138.6, 139.7, 145.6, 149.4, 176.0 ppm; MS (ESI) m/z 375.1 (M+Na+, 100%); HRMS (ESI) calcd for C22H21ClNaO2+ [M+Na+]: 375.1122; found: 375.1125.</p><!><p>Compound 52 was synthesized according to the general procedure D, and purified by flash column chromatography on silica gel (eluent: ethyl acetate: petroleum ether = 1:50). Colorless oil, yield: 55%. IR (film): νmax 3054, 2982, 2931, 1704, 1636, 1486, 1446, 1369, 1345, 1288, 1276 cm−1; 1H NMR (400 MHz, CDCl3) δ 1.28 (t, J = 7.2 Hz, 3H, CH2CH3), 3.35 (s, 2H, CH2CO), 3.56 (m, 2H, CH2CH), 4.19 (q, J = 7.2 Hz, 2H, CH2CH3), 6.52 (s, 1H, C=CH), 6.87–6.94 (m, 1H, Ar-H), 7.03–7.09 (m, 1H, Ar-H), 7.33–7.39 (m, 1H, Ar-H) ppm; 13C NMR (100 MHz, CDCl3) δ 14.2, 34.1, 37.4, 61.0, 106.5 (d, JC-F = 24.0 Hz), 111.5 (d, JC-F = 23.0 Hz), 124.4 (d, JC-F = 9.0 Hz), 134.1, 136.36 (d, JC-F = 3.0 Hz), 139.24 (d, JC-F = 2.0 Hz), 146.34 (d, JC-F = 8.0 Hz), 162.3 (d, JC-F = 241.0 Hz), 170.7 ppm; MS (ESI) m/z 243.1 (M+Na+); Anal. Calcd for C13H13FO2: C, 70.90; H, 5.95. Found: C, 71.30; H, 6.23.</p><!><p>Compound 30 was synthesized according to the general procedure E, and purified by flash column chromatography on silica gel (eluent: ethyl acetate: petroleum ether = 1:10). Yellow solid, yield: 78%. M.p. 128–129 °C (hexane/EtOAc); IR (film): νmax 3426, 2960, 1708, 1604, 1513, 1461, 1418, 1247, 1159, 1049 cm−1; 1H NMR (400 MHz, CDCl3) δ 1.29 (d, J = 6.9 Hz, 6H, CH(CH3)2), 2.96 (sept, J = 6.9 Hz, 1H, CH(CH3)2), 3.69 (s, 2H, CH2COO), 6.90–6.98 (m, 1H, Ar-H), 7.01–7.06 (m, 1H, Ar-H), 7.09 (s, 1H, CH2C=CH), 7.27–7.32 (m, 2H, Ar-H), 7.40 (s, 1H, vinyl-H), 7.50–7.55 (m, 2H, Ar-H), 7.59–7.64 (m, 1H, Ar-H) ppm; 13C NMR (100 MHz, CDCl3) δ 23.8, 33.8, 34.0, 106.5 (d, JC-F = 23.0 Hz), 111.96 (d, JC-F = 23.0 Hz), 119.97 (d, JC-F = 9.0 Hz), 126.9 (2C), 127.0, 129.2, 130.2 (2C), 133.6, 134.2, 137.0, 137.6, 142.96 (d, JC-F = 9.0 Hz), 149.7, 162.9 (d, JC-F = 243.0 Hz), 176.1 ppm; MS (ESI) m/z 345.1 (M+Na+, 100%); HRMS (ESI) calcd for C21H19FNaO2+ [M+Na+]: 345.1261; found: 345.1262.</p><!><p>The GST-tagged human RXRα-LBD (223–462) was incubated with unlabeled 9-cis-RA or different concentrations of compounds in 200 μL binding buffer [0.15 M KCl, 10 mMTris·HCl (pH7.4), 8% glycerol, and 0.5% CHAPS detergent] at 4°C for 1 h. [3H]-9-cis-RA was added to the final concentration of 7.5 nM and final volume of 300 μL and incubated overnight at 4°C. The RXRα-LBD was captured by Glutathione sepharose beads. Bound [3H]-9-cis-RA was quantitated by liquid scintillation counting [29, 30].</p><!><p>HeLa cervix cancer, HCT-116 colon cancer and A549 lung cancer cells were cultured in Dulbecco modified Eagle's medium supplemented with 10% fetal bovine serum (FBS). PC3 prostate cancer and ZR-75-1 breast cancer cells were grown in RPMI1640 medium containing 10% FBS. The cells were maintained at 5% CO2 at 37°C.</p><!><p>Cells cultured in 96-well dishes were treated with various concentrations of compound 30 for 24 hr. The cells were then incubated with 2 mg/mL MTT for 4 hr at 37°C and dissolved by addition of 150 μL DMSO each well. Absorbance was measured at 570 nm [30].</p><!><p>Cells were lysed and equal proteins were electrophoresed on 10% SDS-PAGE gels and transferred onto PVDF membranes (Millipore). The membranes were blocked in 5% skimmed milk in TBST [50 mmol/L Tris-HCl (pH7.4),150 mmol/L NaCl and 0.1% Tween20] for 1 hr, then incubated with primary antibodies and secondary antibodies and detected using ECL system(Thermo). The dilutions of the primary antibodies were anti-RXRα(N197, Santa Cruz) in 1:1000, anti-PARP(H-250, Santa Cruz) in 1:3000, anti-p-AKT (D9E, Cell Signaling Technology) in 1:1000, anti-AKT1/2/3 (H-136, Santa Cruz) in 1:1000, anti-β-actin (Sigma) in 1:5000.</p><!><p>HCT-116 colon cancer cells were transfected with pG5 luciferase reporter vector (50 ng/well) and pGAL-4-RXRα-LBD expression vector (50ng/well) for 24hr. Cells were incubated with varied concentrations of compounds for another 12 hr. Luciferase activities were measured using the Dual-Luciferase Assay System Kit (Promega).</p><!><p>RXRα siRNA used in the experiments were obtained from Dharmacon Research, Inc. A 2.5-μl aliquot of 20 μmol of siRNA/well was transfected into cells grown in 12-well plates by using oligofectamine reagent (Invitrogen) according to the manufacturer's recommendations. Two days after transfection, the cells were harvested for Western blotting [18].</p><!><p>Schrodinger's (Portland, OR) (www.schrodinger.com) GLIDE[21], a grid-based docking program, was used for docking studies of the small molecule ligands to the protein. The crystal structure of RXRα LBD in complex with antagonist LG100754 (Protein Data Bank code 3A9E) was used. The GLIDE GScore was used as docking score to rank the docking results. Visual inspection was done to pick the docked pose from the ranked results. Schrödinger's Maestro 6.5 was used to prepare Figure 5.</p>

PubMed Author Manuscript

Open Characterization of Vaping Liquids in Canada: Chemical Profiles and Trends

Currently, there is a lack of comprehensive data on the diversity of chemicals present in vaping liquids. To address this gap, a non-targeted analysis of 825 vaping liquids collected between 2017 and 2019 from Canadian retailers was conducted. Prior to mass spectrometry analysis, samples were diluted 1:500 v/v with methanol or acetonitrile. Chemical compound separation and analysis was carried out using gas chromatography and triple quadrupole mass spectrometry (GC-MS/MS) systems operated in the full scan mode and mass range of 35–450 m/z. Mass spectrum for each sample was obtained in electron ionization at 70 eV and processed. Non-targeted identification workflow included use of automated mass spectral deconvolution and identification system (AMDIS), where required, as well as a number of commercially available spectral libraries. In order to validate identities, an in-house database of expected compounds previously detected in vaping liquids was used along with genuine analytical standards for compounds of interest. This resulted in a dataset of over 1,500 unique detected chemicals. Approximately half of these chemical compounds were detected only once in a single product and not in multiple products analyzed. For any sample analyzed, on average, 40% of the chemical constituents appeared to have flavouring properties. The remainder were nicotine and related alkaloids, processing, degradation or indirect additives, natural extractives and compounds with unknown roles. Data published here from the project on the Open Characterization of vaping liquids is unique as it offers a detailed understanding of products’ flavour chemical profiles, the presence and frequency of chemicals of potential health concern, as well as trends and changes in products’ chemical complexity over a three-year period. Non-targeted chemical surveillance such as this present valuable tools to public health officials and researchers in responding to emergent issues such as vaping associated lung injury or informing chemical based strategies which may be aimed at addressing product safety or appeal.

open_characterization_of_vaping_liquids_in_canada:_chemical_profiles_and_trends

1 Introduction<!>2.1 Chemicals and Reagents<!>2.2 Samples<!>2.2.1 Vaping Liquid Flavour Classification<!>2.3 Sample Preparation<!>2.4 GC MS/MS Analysis<!>2.5 Non-Targeted Workflow<!><!>2.5 Non-Targeted Workflow<!>2.6 Data Processing and Chemical Roles<!>3.1 Workflow and Method Challenges<!>3.2 Chemical Space<!><!>3.3 Chemical Roles<!><!>3.3 Chemical Roles<!><!>3.3 Chemical Roles<!>3.4 Flavour Chemicals of Concern<!><!>3.4 Flavour Chemicals of Concern<!>3.5 Chemicals of Health Concern<!><!>3.5 Chemicals of Health Concern

<p>Nicotine containing vaping products are a less harmful source of nicotine for people who smoke and are unable to cease the use of traditional tobacco products such as combustible cigarettes (Government of Canada, 2020a). Vaping products are not free from harm, in fact, for people who do not smoke, inhalation of vaping aerosol represents an unnecessary source of exposure to chemicals of potential health concern. Vaping products are a highly varied (Office of the Surgeon General, 2016) class of consumer products that continue to rapidly evolve and exhibit dynamic changes in product design and performance. This lack of product homogeneity as well as high variability in product use behaviors are thought to be one of the main reasons for not more fully understanding the harms and benefits of vaping products. The chemical exposure profile depends on vaping device parameters and design, user behavior and vaping liquid chemical composition. Elucidating the chemical composition of vaping liquids informs not only on the product's safety and health risks relative to smoking, it can also provide information on aspects of product appeal and addiction liability among the products studied.</p><p>Nearly all vaping products intended for use with nicotine contain a liquid made up of approximately 90% carrier solvents (humectants-propylene glycol and glycerol), 0–6% nicotine with the remainder comprised of flavouring agents, processing aids, contaminants and water. The chemical heterogeneity of the vaping products originates from the variability among flavouring and processing agents used and presence of contaminants and post-formulation chemical transformations due to product storage and ageing. The traditional approach to analyzing chemicals in products is through targeted chemical analysis, wherein known chemicals are examined using optimized laboratory methods. Data generated using these methods offer an important support for decisions and actions but are limited to the known chemical space for which reference standards exist. In comparison to traditional chemical analytical methods, non-targeted analysis (NTA) methods aim to discover and prioritize total chemical exposures from as many as possible sources of chemicals present in the products. These methods use advanced analytical equipment, chemical libraries, and software based workflows to handle large datasets and detect as many chemicals as possible, including those previously unknown or understudied. The main aim of our study is to create a foundational library of chemicals present in Canadian vaping products using data collected from an analysis of 825 vaping liquids. This work can be used to better understand health risks, appeal and addiction associated with vaping products. In the current report we outline the study design, details of the non-targeted approach applied, large dataset organization and preliminary data analysis.</p><!><p>99.7% pure propylene glycol and 99.2% pure glycerol were purchased from Sigma-Aldrich (Oakville, ON, Canada). HPLC grade methanol and acetonitrile were purchased from Fisher Scientific (Ottawa, ON, Canada). For a full list of individual compounds used to detect select chemicals refer to Supplementary Table S1.</p><!><p>A diverse sample of 825 vaping liquids were collected from vaping stores and physical retailers in seven cities across Canada and from online Canadian retailers, between 2017 and 2019. The samples included liquids of various nicotine concentrations (0–59 mg/ml) as well as varying proportions of propylene glycol (PG) and vegetable glycerine (VG) (0/100% to 100/0% PG/VG). Overall, samples represented 182 different brands. While 8% of samples had no declared product origin, a majority of products were formulated in Canada (82.5%), followed by United States (7.3%), and elsewhere (2.2%). Ninety-seven percent of products collected were packaged in refillable bottle format (30 or 60 ml, glass or plastic), while the rest were in plastic pod based format.</p><!><p>Flavour–related information from product packaging and from product descriptions on manufacturer websites were used to inform the primary, intended flavour of the vaping liquid and systematically classify each sample into one of 18 flavour categories in a modified vaping liquid flavour wheel (Krüsemann et al., 2019), adapted for vaping liquid flavours available in the Canadian market. The following 18 flavour categories were used for product classification: Fruit (N = 108), Desserts (N = 76), Tobacco (N = 134), Mint/menthol (N = 97), Coffee (N = 33), Tea (N = 35), Energy Drinks (N = 19), Confectionary (N = 49), Savoury (N = 24), Spices (N = 19), Herbal/floral (N = 7), Nuts (N = 21), Alcohol (N = 34), Breakfast cereals (N = 33), Soft drinks (N = 29), Milk/cream/yogurt (N = 26), Unflavoured (N = 26), and Other (N = 55).</p><!><p>Following thorough sample mixing, 40 µl of each vaping liquid was diluted to 20 ml with methanol (Quantum TSQ GC MS/MS methodology) or acetonitrile (7000C GC MS/MS methodology). Diluted samples were vortex mixed and 1 µl was injected and analyzed using gas chromatography mass spectrometry. Solvent blank (methanol or acetonitrile) was injected after each sample to ensure no carryover between samples. Matrix blank consisting of propylene glycol and glycerol was used during the method development process to assess possibility of PG/VG thermal degradation during GC analysis.</p><!><p>Two instruments (Quantum TSQ and 7000C GC MS/MS) were used to acquire data, as such, two different methods were optimized. The acquisition mode for the both instruments was full-scan acquisition mode. The Quantum TSQ MS/MS instrument was coupled to a Trace GC Ultra gas chromatograph (Thermo Electron Corp.). The oven ramp for this instrument was set as followed: 65°C hold for 1 min, followed by an increase of 5°C/min to 280°C and held for 3 min thereafter. The source temperature and interface were held at 200°C and 250°C, respectively. The MS was operated in Electron Ionization, full-scan mode with scan range 35–450 m/z and emission current set at 100 µA. Source temperature was set to 200°C, while GC interface temperature was 250°C. The second instrument was a 6890N gas chromatograph coupled to a 7000C MSMS detector (Agilent Technologies Inc.). The GC oven programming was started at 50°C and held for 2 min, followed by a ramp at 5°C/min to 240°C where it was held for 3 min. Both source and the interface temperature were held at 280°C. The MS was operated in a full-scan acquisition mode and scan range 30–450 m/z. GC analyte separation was performed using the Zebron ZB-5HT GC capillary column (30 m × 0.25 mm × 0.25 µm) from Phenomenex (CA, United States) on both instruments. The injector temperature was set at 280°C for both GCs with splitless injection mode for GC Ultra and pulsed splitless mode for 6890N GC. In both cases GC carrier gas was helium operated in constant flow mode at 1 ml/min rate.</p><!><p>Immediately following the sample analysis the chromatograms were processed as described in Figure 1. In some instances, where peak separation was poor, automated mass spectral deconvolution and identification system (AMDIS) (NIST, National Institute of Standards and Technology, 2019) was used for peak deconvolution. In general, the spectrum of individual compounds was matched against spectra from the National Institute of Standards and Technology (NIST 17) library reference peaks. In addition, the Agilent GC MS/MS instrument was also equipped with Wiley's library of Mass Spectra of Flavors and Fragrances of Natural and Synthetic Compounds (FFNSC), 3rd Edition, while the Quantum GC also, used the Wiley Registry of Mass Spectral Data, 11th Edition for improved detection and confirmation. The peaks at signal intensity higher than signal to noise 3:1 are at first tentatively identified. In general, the compounds which score higher when matched against spectral libraries (>70 Agilent, >700 Quantum) and have an appropriate Retention Index, where available, are considered to be a good fit. In order to improve the analyte identification "starting confidence" or "prior probability" was utilized as previously described (Stein 2012). A database of previously detected and reported chemical compounds in vaping liquids from other published sources (N = 151, Supplementary Table S2 was used to develop categories of expected chemical compounds in vaping liquids (Table 1). Moreover, the same expected chemical compounds list served as the basis to set up an internal mass spectral database using genuine analytical standards of individual chemical compounds.</p><!><p>Non-targeted workflow.</p><p>Identification of detected chemicals.</p><!><p>The chemical compounds with poor matching were compiled and a follow up analysis (e.g., accurate mass determination) will be performed in the future, if required.</p><!><p>Each identified chemical was assigned one or more roles in order to have a better understanding of the function they may have within a vaping liquid formulation. A literature synthesis was conducted which involved drawing from a variety of sources including published literature, open source websites and databases (e.g. PubChem (NIH, National Institutes of Health, 2021a), Chem Spider (Royal Society of Chemistry, 2021), The Human Metabolome Database (HMDB) (Wishart et al., 2018), Flavor DB (Garg et al., 2018), FooDB (Harrington et al., 2019)), manufacturer specifications, patents, Safety Data Sheets (SDS) and others, in order to aid in data processing and assignment of roles. Each chemical was classified into at least one of the six (6) roles: nicotine and related alkaloids, processing chemicals, natural extracts, flavours or fragrances, indirect additives and chemicals with unknown role. Supplementary information provides more information on specific functional role categories.</p><!><p>A number of challenges, which were successfully resolved, were encountered during this project. During the method development stages significant amount of time was invested in optimizing methodology as to minimize any compounds that may form during chemical analysis and degradation of product carrier solvents. More details and discussion are provided on method validation in Supplementary Section S2. Simple matrix blanks of PG and VG were put through dilution and analysis and no detected chemical compounds were formed during the analysis run time. Of note is that there was no carryover between samples analyzed as observed through testing of analytical blank samples between each injected sample. Simple dilution prior to mass spectrometry analysis did not result in any background contamination either. The 500 times solvent dilution often resulted in a broad glycerol peak and challenging chromatographic separation that, at times, would overlap with a signal for another chemical compound. In those instances, AMDIS was applied successfully, Supplementary Section S3. Processing of the resulting chromatograms was time consuming task, but was simplified using genuine analytical standards and established retention times for the group of chemical compounds previously reported to be present in vaping products (Table 1; Supplementary Table S1). This project was a significant undertaking (development of NTA methodologies and processing of large dataset with over 14,000 chemical compounds identified), it required diverse skillsets and frequent literature reviews to better elucidate chemical information such as functional groups and possible functional roles. While some parts of this process were automated, many steps still required manual quality control and review of results to ensure accuracy and completeness. Searching for individual chemical characteristics was done using Chemical Abstracts Services (CAS number) as provided in the mass spectral libraries. Significant data clean-up was performed in order to remove duplicate CAS numbers as some compounds may have multiple CAS numbers (e.g. menthol) and different mass spectral libraries may have preferences for CAS number provided as primary one.</p><!><p>The actual chemical space of all products tested was 1,507 unique chemical compounds. Since some chemical compounds were detected in more than one product, total number of chemicals detected in 825 samples was over 14,000. Close to 50% (734/1,507) of all chemicals were detected in just one vaping liquid, illustrating the heterogeneity of this class of consumer products and infrequency of occurrence among chemical compounds used. Only four chemical compounds were detected in over 50% of all products studied. These include nicotine, the carrier solvents propylene glycol and glycerol, as well as β-Nicotyrine, a nicotine oxidation by-product that may form during storage (Wada et al., 1959). Seven hundred and thirty-eight products were labelled as nicotine-containing, however, among these products 14 were found not to contain any detectable nicotine. The lack of detection of nicotine in these samples was not due to the sensitivity of analytical method as this scan method is able to detect nicotine down to 0.03 mg/ml. The majority of samples with this discrepancy were, in fact, labelled to contain nicotine at over 9 mg/ml. Out of 87 products labelled as nicotine free, one product was detected to contain nicotine. These discrepancies on nicotine presence are likely due to poor manufacturing practices or lack of nicotine stability, as noted elsewhere (Goniewicz et al., 2015; Kavvalakis et al., 2015). Of note is that all samples in question were collected prior to September 2018 and, when labelled, were marked as manufactured prior to this date. These products likely precede the Government of Canada's Tobacco and Vaping Products Act (Government of Canada, 2018b) which includes limits on nicotine concentrations and brings forward compliance and enforcement of the same.</p><p>All chemical compounds detected in the course of the study can be classified into one of 170 chemical classes. The most frequently detected chemical classes are alcohol, organooxygen, carboxylic acid and derivatives, and esters, Figure 2.</p><!><p>Detected chemical classes.</p><!><p>There were 87 (0.6%) chemical compounds for which it was not possible to assign or determine their identity using the mass spectral libraries available. In the future, samples with these compounds may be analyzed using different analytical approaches to identify them. Each chemical compound with a known identity was assigned at least one of the six functional roles using the various sources of peer-reviewed literature and supporting materials. Although identity was determined for the vast majority of detected chemicals, a functional role was not assigned to 8% of the chemicals detected as no supporting materials were found. Of note, a larger number of the chemicals with unknown roles have been previously detected in yeast (University of Washington, 2018). At this time, it is not known what the exact role or origin of yeast related chemicals in vaping liquids is. Autolyzed yeast extract is used as a flavour enhancer in foods and beverages (U.S. Food and Drug Administration, 2010; U.S. Food and Drug Administration, 2010), while microbial contamination of vaping products has been reported previously (Lee et al., 2019). Six percent of all chemicals were assigned indirect additive roles with supporting materials (Food and Drug Administration, 2017) often found among records on indirect additives on foods or food contact materials. It is likely these are found in products as a result of leaching into the vaping liquid during processing or packaging. Alkaloid roles were assigned to 10% of chemicals, which in the majority of cases included nicotine and related minor alkaloids. Thirteen percent of chemicals were found to have the natural extract role while 27% of chemicals were likely used as processing chemicals in the formulation. Examples of processing roles include emulsifiers, humectants, diluents and others. Forty-three percent of all chemicals detected were assigned a flavour or fragrance role. The number of individual chemicals per vaping liquid sample ranged between 4 and 66 compounds with a mean of 18 chemical compounds detected per product. Although a lower number of nicotine-salt based products were analyzed (N = 116) when compared to free-base nicotine products (N = 623), nicotine-salt products were found to contain a lower number of chemicals, with a mean of 16 chemicals detected per product. The number of chemical substances present in vaping liquids (e-liquids) can be used as one of the indicators of potential toxicity of the product, as reported previously by the group of researchers from North Carolina (Sassano et al., 2018) who concluded that increasing chemical numbers were associated with increasing toxicity when compared to solvent (PG/VG) vehicle in high-throughput in-vitro toxicity testing. In addition to nicotine type used in the product, the number of chemicals detected varied with the liquid' flavour categories, Figure 3.</p><!><p>Number of detected chemicals per flavour category.</p><!><p>As expected, the unflavoured products appeared to have the least complex chemical profiles (mean number of nine chemicals), followed by the tobacco flavour category (mean number of 14 chemicals). The most complex chemical profiles were found in the categories of milk/cream (e.g., Yogurt) and spices (e.g. cinnamon), each with a mean of 22 detected chemicals. On average per product flavour category, the unflavoured category had the lowest proportion of flavour chemicals (15% of total chemicals), and energy drinks had the highest proportion of flavour chemicals (58% of total chemicals). Flavour categories such as fruit, confectionary and dessert, which may have a higher preference among youth, had higher proportions of flavour chemicals on average (48, 54 and 55% of all chemicals, respectively). This proportion of flavour compounds is somewhat lower compared to proportions (63%) reported by the Dutch study from European vaping products (Krüsemann et al., 2021). The differences could be due to the origins of the chemical datasets, as Dutch data is based on a reporting system where manufacturers provide information on ingredients added, while the non-targeted analysis based dataset results from chemical analysis which may detect impurities, indirect additives, as well as compounds that result from chemical reactions post-product formulation and product ageing (degradation, leaching and transformations). These additional compounds would increase the total number of compounds known to be present in the product, thereby decreasing the percentage of flavouring compound in the final composition.</p><p>Of note is that the mean number of chemicals detected per product has in fact changed over the years; products collected in 2017 and 2018 appear to have a significantly higher number of chemical compounds when compared to those collected in 2019. This trend is observed regardless of flavour category analyzed, Figure 4A. When the trend is examined for the number of flavour compounds over this time period and in the same products a similar trend emerges, suggesting a decrease in the chemical flavour complexities among this group of products, Figure 4B.</p><!><p>Overall chemicals (A) and Flavour chemicals (B) per popular flavour category, 2017–2019.</p><!><p>This trend could be in part explained by the higher frequency of nicotine-salt based products post 2018 which on average appear to contain a lower number of chemicals. Nicotine-salts are perceived to provide a less harsh and smoother sensory experience for the product users (Leventhal et al., 2021), thus it is likely they require less flavouring agents to mask the sensory experience normally associated with free-base nicotine products.</p><!><p>Vaping products on the Canadian market come in a variety of flavour categories. In the past few years, youth vaping prevalence has increased in Canada (Government of Canada, 2020a) and flavours play an important role in attracting youth to vaping products. Recent evidence suggests that youth prefer flavour categories such as fruit, confectionary and dessert (Government of Canada, 2018a; O'Connor et al., 2019). The chemicals detected in products are used to better understand flavour chemicals and their role in imparting intended or declared product flavours. Vaping product formulations are the manufacturer's interpretation of the intended or declared flavour. Our data indicates that the chemical space of each flavour category is diverse and there is a high degree of chemical overlap between flavour categories. Similar to previously published studies (Tierney et al., 2016; Omaiye et al., 2019), our data shows that vaping liquids contain some of the same flavour chemicals despite their flavour category. Except for mint/menthol, herbal/floral and unflavoured category, across all other products, the top five most frequently detected flavour chemicals (Table 2) were vanillin, ethyl maltol, ethyl vanillin, vanillin propylene glycol acetal and cyclotene. Vanillin, ethyl maltol and ethyl vanillin were in the top five flavouring chemicals for more than half of the flavour categories studied. Collectively, the top five chemicals have flavour descriptors such as "sweet," "creamy" and "vanilla" (Good Scents Company, 2021). Vanillin and ethyl maltol, but not ethyl vanillin, were the most frequently detected flavour chemical in the three categories likely to be more appealing to youth. Ethyl maltol is a sweetener, with a sweet, caramellic, jammy, strawberry-like odor description and sweet, burnt cotton candy, caramel-like taste. Perception of sweet flavour in vaping products has been shown to produce greater appeal and perceived sweetness ratings among young vapers (Goldenson et al., 2016). Moreover, sweet perception and appealing flavours can enhance nicotine reward reinforcing effects in vaping and other tobacco products (Kroemer et al., 2018; Patten and De Biasi 2020).</p><!><p>The top five most frequently identified chemicals in all flavour categories and the flavour/odour description from the Good Scents Company website.</p><p>The Good Scents Company Information System (Good Scents Company, 2021).</p><!><p>In published studies, concentrations of ethyl maltol in vaping liquids range between undetectable to 4,200 μg/ml (Aszyk et al., 2017; Behar et al., 2018), compared to average maximum concentration ranges of 12.4–152 μg/ml in non-alcoholic beverages and baked goods, respectively, on which Flavor Extract Manufacturers Association (FEMA, The Flavor and Extract Manufacturers Association of the United States, 2021) Expert Panel based its' judgments that this substance is safe for ingestion (Oser and Ford 1977). Although generally recognized as safe for ingestion, the health effects of ethyl maltol, and more broadly the majority of flavour compounds, have not been assessed for the inhalation route (Flavor and Extract Manufacturers Association, 2021). Currently, published studies on vaping flavours focus on cytotoxic and mutagenic effects in cell models (Behar et al., 2018; Muthumalage et al., 2018); translating these study findings into a real-life setting is challenging. While inhalation toxicity data is scarce for some compounds, certain vaping flavour compounds are recognized as those of concern for human health. For example, diacetyl and 2,3 pentanedione are two buttery flavours, shown to cause lung and respiratory airways damage in animal models and are associated with respiratory disease and decreased lung function in occupationally exposed employees of food flavouring and food manufacturing facilities (NIOSH, The National Institute for Occupational Safety and Health, 2016). While diacetyl was detected in two vaping liquids acquired prior to 2018, 2,3 pentanedione was not detected in any vaping liquids analyzed in the Open Characterization dataset. Another flavour, the monoterpene pulegone typically found in extracts of mint oil, has been previously detected in vaping products (Hutzler et al., 2014; Geiss et al., 2015). This chemical has been shown to induce some carcinogenic effects in mice and rats (National Toxicology Program, 2011). In the Open Characterization analysis, 11 out of 825 (1.3%) products were found to contain pulegone at unknown concentration levels, mainly mint/menthol flavoured products (9/11 products). Currently, no evidence is available that pulegone has any vaping-related health effects in humans.</p><!><p>Within this dataset, the quantification of all chemicals identified is untenable given the targeted study method developments may take years to complete. Chemical prioritization or screening based on known hazards was used to develop a list of chemicals for quantification. Providing exposure estimates through targeted analytical studies focused on these prioritized chemicals will provide sufficient information to better elucidate the risk. The majority of studies provide results on relative risk and comparison to tobacco cigarettes. Vaping products in fact infrequently contain tobacco specific toxicants and even in cases when they do, these are often present at much lower concentrations as observed in the exposure studies on product users (Goniewicz et al., 2018; Engineering, and Medicine National Academies of Sciences, 2018). For example, in our study there was only one product that was found to contain N-Nitrosodimethylamine (NDMA); no other nitrosamines were detected. In addition to NDMA, 9 out of 93 US FDA's Harmful and Potentially Harmful Constituents (HPHC) (FDA, US Food and Drug Administration, 2012) were detected in Open Characterization samples (Table 3).</p><!><p>Established list of constituents identified by US FDA as harmful and potentially harmful constituents and their detection frequency in vaping liquids.</p><!><p>The reasons behind the higher frequency of detection of naphthalene compared to other HPHC chemicals are unclear at this time; this Polycyclic Aromatic Hydrocarbon is normally present in tobacco smoke, but also in the extracts of various fruits and other plants (Gómez et al., 1993; Paris et al., 2018), so it is possible that naphthalene originates from the natural extracts used to flavour the products. Of note is that other methylated and naphthalene-related structural analogs, not on the HPHC list, were also detected in vaping products studied. For example, 1-methyl naphthalene, a flavour and fragrance agent normally found in fruits (Good Scents Company, 2021), is also detected in 12% of products analyzed. Exposure of laboratory animals to 1- and 2-methylnaphtalene resulted in spleen and organ damage while mice exposed dermally for 30 weeks developed pulmonary alveolar proteinosis. Humans exposed to this compound developed skin irritation and skin photosensitization (NIH, National Institutes of Health, 2021b). In 2019, USFDA proposed the addition of 19 chemical compounds to an existing HPHC list of 93 (Food and Drug Administration, 2017), mainly to reflect potentially harmful chemicals present in vaping products. The first proposed chemical is glycidol, a probable human carcinogen (International Agency for Research on Cancer, 2000) thought to result from thermal degradation of glycerol. Glycidol has been previously detected in vaping product emissions (Sleiman et al., 2016) and was found in 3% of the liquids tested. Non-targeted studies such as this provide datasets that can inform future steps and ultimately characterize product-use specific harms. The prioritization can consider chemicals with already established health effects of concern, detection frequency or chemical presence in products with high market share. In our dataset, most chemicals of concern were not detected in the majority (>50%) of products studied, indicating that the chemicals of concern can be used to identify products for which the ingredients used may be a cause for concern. The goal is to provide information that would lead to products which minimize the risk of vaping products for consumers looking to completely switch from combustible tobacco products.</p><p>In comparison to traditional chemical analytical methods, non-targeted analysis (NTA) methods aim to discover as many chemicals as possible in products, including those previously unknown or with limited data. To date, there has been only one published study using non-targeted screening of Canadian vaping liquids (Czoli et al., 2019). One hundred and sixty-six vaping liquids collected in 2015 were analyzed using a gas chromatography mass spectrometry instrument with limited sensitivity and resolution. Similarly, a U.S. dataset generated by the Centre for Tobacco Regulatory Science and Lung Health (Center for Tobacco Regulatory Science and Lung Health, 2021), chemically characterized approximately 300 vaping product samples; significantly fewer than the Canadian Open Characterization dataset (N = 825). Closed pod-system brands that make up a majority of the vaping market in Canada were not included in the U.S. dataset. In addition, limited information is available on the products tested in the U.S. including classification by flavour categories, as their product names are not self-explanatory (e.g. Carnage, Magic Dragon, etc.). Finally, it is unknown how many of these U.S. products are available for sale in Canada. These factors present challenges in comparing the two datasets. Overall, valuable information can be determined by evaluating different market datasets, however direct comparisons are challenging given the heterogeneity of vaping products within and between different regions.</p>

PubMed Open Access

O-(Triazolyl)methyl carbamates as a novel and potent class of FAAH inhibitors

Inhibition of fatty acid amide hydrolase (FAAH) activity is under investigation as a valuable strategy for the treatment of several disorders, including pain and drug addiction. A number of potent FAAH inhibitors belonging to different chemical classes have been disclosed. O-aryl carbamates are one of the most representative families. In the search for novel FAAH inhibitors, we synthesized a series of O-(1,2,3-triazol-4-yl)methyl carbamate derivatives exploiting the copper-catalyzed [3 + 2] cycloaddition reaction between azides and alkynes (click chemistry). We explored structure-activity relationships within this new class of compounds and identified potent inhibitors of both rat and human FAAH with IC50 values in the single-digit nanomolar range.

o-(triazolyl)methyl_carbamates_as_a_novel_and_potent_class_of_faah_inhibitors

Introduction<!>Chemistry<!>Structure\xe2\x80\x93Activity Relationship (SAR) and Stability Studies<!>Conclusion<!>Chemistry<!>General procedure (1) for the synthesis of triazoles (17-30, 37-40)<!>General procedure (2) for the synthesis of double triazoles carbamates (34a-b)<!>(3-phenylphenyl)methyl N-cyclohexylcarbamate (4)<!>(1-phenyltriazol-4-yl)methyl N-cyclohexylcarbamate (17)<!>(1-benzyltriazol-4-yl)methyl N-cyclohexylcarbamate (18)<!>(1-phenethyltriazol-4-yl)methyl N-cyclohexylcarbamate (19)<!>(1-benzhydryltriazol-4-yl)methyl N-cyclohexylcarbamate (20)<!>[1-(2-naphthylmethyl)triazol-4-yl]methyl N-cyclohexylcarbamate (21)<!>[1-[(2-cyanophenyl)methyl]triazol-4-yl]methyl N-cyclohexylcarbamate (22a)<!>[1-[(3-cyanophenyl)methyl]triazol-4-yl]methyl N-cyclohexylcarbamate (22b)<!>[1-[(4-cyanophenyl)methyl]triazol-4-yl]methyl N-cyclohexylcarbamate (22c)<!>[1-[(2-fluorophenyl)methyl]triazol-4-yl]methyl N-cyclohexylcarbamate (23a)<!>[1-[(3-fluorophenyl)methyl]triazol-4-yl]methyl N-cyclohexylcarbamate (23b)<!>[1-[(4-fluorophenyl)methyl]triazol-4-yl]methyl N-cyclohexylcarbamate (23c)<!>[1-[(2-chlorophenyl)methyl]triazol-4-yl]methyl N-cyclohexylcarbamate (24a)<!>[1-[(3-chlorophenyl)methyl]triazol-4-yl]methyl N-cyclohexylcarbamate (24b)<!>[1-[(4-chlorophenyl)methyl]triazol-4-yl]methyl N-cyclohexylcarbamate (24c)<!>[1-(o-tolylmethyl)triazol-4-yl]methyl N-cyclohexylcarbamate (25a)<!>[1-(m-tolylmethyl)triazol-4-yl]methyl N-cyclohexylcarbamate (25b)<!>[1-(p-tolylmethyl)triazol-4-yl]methyl N-cyclohexylcarbamate (25c)<!>[1-[(2-methoxyphenyl)methyl]triazol-4-yl]methyl N-cyclohexylcarbamate (26a)<!>[1-[(3-methoxyphenyl)methyl]triazol-4-yl]methyl N-cyclohexylcarbamate (26b)<!>[1-[(4-methoxyphenyl)methyl]triazol-4-yl]methyl N-cyclohexylcarbamate (26c)<!>[1-[(3,5-dimethoxyphenyl)methyl]triazol-4-yl]methyl N-cyclohexylcarbamate (27)<!>[1-[(2,6-difluorophenyl)methyl]triazol-4-yl]methyl N-cyclohexylcarbamate (28)<!>[1-[(3,5-difluorophenyl)methyl]triazol-4-yl]methyl N-cyclohexylcarbamate (29)<!>[1-[(2-fluoro-3-methoxy-phenyl)methyl]triazol-4-yl]methyl N-cyclohexylcarbamate (30)<!>2-(1-phenyltriazol-4-yl)ethyl N-cyclohexylcarbamate (32)<!>2-(1-benzyltriazol-4-yl)ethyl N-cyclohexylcarbamate (33)<!>[1-[(2-methoxyphenyl)methyl]triazol-4-yl]methyl N-[[1-[(2-methoxyphenyl)methyl]triazol-4-yl]methyl]carbamate (34a)<!>[1-[(3-methoxyphenyl)methyl]triazol-4-yl]methyl N-[[1-[(3-methoxyphenyl)methyl]triazol-4-yl]methyl]carbamate (34b)<!>[1-[(2-methoxyphenyl)methyl]triazol-4-yl]methyl N-[[1-[(3-methoxyphenyl)methyl]triazol-4-yl]methyl]carbamate (37)<!>[1-[(2-methoxyphenyl)methyl]triazol-4-yl]methyl N-[[1-[(4-methoxyphenyl)methyl]triazol-4-yl]methyl]carbamate (38)<!>[1-[(3-methoxyphenyl)methyl]triazol-4-yl]methyl N-[[1-[(2-methoxyphenyl)methyl]triazol-4-yl]methyl]carbamate (39)<!>[1-[(3-methoxyphenyl)methyl]triazol-4-yl]methyl N-[[1-[(4-methoxyphenyl)methyl]triazol-4-yl]methyl]carbamate (40)<!>(1-(1-phenyl-1,2,4-triazol-3-yl)methyl N-cyclohexylcarbamate (43)<!>Benzyl-1,2,4-triazol-3-yl)methyl N-cyclohexylcarbamate (46)<!>In vitro rat FAAH radiometric assay<!>In vitro human FAAH fluorescent assay<!>Ex vivo FAAH inhibition assay<!>MGL activity assay<!>In vitro Rat Plasma stability assay<!>In vitro Mouse Liver Microsomes (MLM) stability assay<!>Aqueous Kinetic Solubility assay

<p>Fatty acid amide hydrolase (FAAH)[1] is a membrane-bound serine hydrolase that catalyses the hydrolytic cleavage of endogenous biologically active fatty acid ethanolamides (FAEs), such as anandamide (AEA), an agonist of cannabinoid receptors,[2] and palmitoylethanolamide (PEA)[3] and oleoylethanolamide (OEA),[4] which are agonists of type-α peroxisome proliferator-activated receptors (PPAR-α).[5] These natural FAAH substrates may play important roles in the central nervous system (CNS) and in peripheral tissues, where they are involved in a variety of physiological processes.[6]</p><p>Substantial efforts have been dedicated to the discovery of potent and selective FAAH inhibitors, with the objective of developing therapeutic approaches for pathologic conditions such as pain, drug addiction, anxiety, and depression.[6-7] Different classes of molecules are known to increase intracellular FAE levels through FAAH inhibition, including carbamates[8] and piperidine/piperazine ureas[9] that covalently bind to FAAH,[10] and α-keto heterocycles-based inhibitors,[11] which inhibit FAAH by reversible hemiketal formation with the catalytic serine of the enzyme.[12] Among them, the O-arylcarbamate series, exemplified by URB524[8a, 8b, 13] (1a, Figure 1), URB597[8a, 8b, 13] (1b, Figure 1) and URB694[8c, 14] (1c, Figure 1), has been extensively investigated.[15] In particular, it was shown that compound 1b exerts a combination of anxiolytic-like, antidepressant-like, and analgesic effects, because of its ability to inhibit FAAH activity in the CNS and peripheral tissues.[16]</p><p>The active site of FAAH is characterized by an atypical catalytic triad, consisting of Ser241-Ser217-Lys142, which is capable of hydrolyzing amide and ester bonds at similar rates.[17] Several studies, including computational modeling,[18] supported by the resolution of the crystal structure of humanized rat FAAH in complex with 1b,[19] indicate that O-arylcarbamates bind covalently to FAAH and cause its irreversible inhibition. In particular, it has been proposed that this class of molecules is attacked at the carbonyl group by Ser 241, leading to the formation of carbamoylated, catalytically inactive FAAH and releasing the O-biphenyl moiety as the leaving group.</p><p>O-arylcarbamates such as 1b are selective for FAAH, but can also interact with select liver carboxylesterases, at least at high concentrations, and have limited plasma stability.[9a] Recently, however, highly potent O-arylcarbamates with markedly improved selectivity for FAAH were identified.[8c] The insertion of an electron-donating substituent, such as a hydroxy or amino group, in the para position of the proximal phenyl ring of 1a did not significantly affect inhibitory potency in vitro, but caused a marked increase in the stability of the compounds in plasma, in comparison to other molecules in the series. The compound URB694 (1c) was identified as a potent FAAH inhibitor with improved plasma stability, prolonged half-life in vivo, and decreased activity towards liver carboxylesterases in comparison to 1b.</p><p>In the search for better FAAH inhibitors with improved stability, we designed a novel class of carbamates where the O-aryl moiety is replaced by an O-(triazol-4-yl)alkyl group (Figure 2). We expected those compounds to be more stable than their O-aryl analogues, since the aliphatic alcohol resulting from the nucleophilic attack on the carbamate is a worse leaving group than a phenol. Although carbamate-based FAAH inhibitors containing an O-(heteroaryl)alkyl moiety have been reported in the patent literature,[12] no O-(triazol-4-yl)alkyl carbamate derivatives have been described to date.</p><p>We report here the synthesis and characterization of a series of O-(1,2,3-triazol-4-yl)alkyl carbamates, prepared by copper-catalyzed [3 + 2] cycloaddition reaction between azides and alkynes.[20] The fast and versatile synthesis via click chemistry allowed us to prepare a number of analogues in a quick and reliable manner, and rapidly explore the SAR within this new class of FAAH inhibitors.</p><!><p>The (3-phenylphenyl)methyl N-cyclohexylcarbamate (4) was prepared from commercially available 3-phenylbenzoic acid (2) by lithium aluminum hydride reduction to 3-phenylbenzyl alcohol (3) followed by reaction with commercial cyclohexyl isocyanate (Scheme 1).</p><p>The preparation of 1,4-disubstituted-1,2,3-triazoles was accomplished by [3 + 2] cycloaddition reaction between azides and alkynes, in the presence of copper (I) salts (click chemistry).[20]</p><p>We designed a versatile synthetic strategy that allowed us to generate a first set of molecules, bearing an O-(1,2,3-triazol-4-yl)methyl moiety. The desired compounds (17-30, 32, 33) were synthesized as shown in Scheme 2.</p><p>The aromatic azide 5a was prepared from aniline by a diazotation-azidation protocol,[21] while 5b, 5c, and 6-16 were obtained in good to excellent yields by reacting the corresponding halides with sodium azide.[22] Final compounds 17, 18, and 19 were prepared via click chemistry, starting from prop-2-yn-1-yl N-cyclohexylcarbamate,[23] prepared by reaction of cyclohexylamine with the commercially available prop-2-ynyl chloroformate, and the azides 5a, 5b, and 5c, respectively (Scheme 2). Then, copper catalyzed [3 + 2] cycloaddition reaction between azides 5a and 5b with the commercially available but-3-yn-1-ol, allowed us to obtain compounds 31a and 31b in acceptable yields.[24] Finally, compounds 32 and 33 were prepared by coupling alcohols 31a and 31b, respectively, with commercial cyclohexyl isocyanate (Scheme 2).</p><p>The O-(1,2,3-triazol-4-yl)methyl carbamate derivatives 20-30 (Scheme 2) were prepared by reaction of the azides 6-16 with prop-2-yn-1-yl N-cyclohexylcarbamate, under click chemistry conditions.</p><p>A second set of analogues was synthesized, as reported in Scheme 3, in order to explore region A (Figure 2). Prop-2-ynyl-N-prop-2-ynyl carbamate,[25] obtained by reaction of the commercially available propargyl amine with prop-2-ynyl chloroformate, was reacted with aromatic azides 12a and 12b thus affording compounds 34a and 34b, which bear the same substituent on both aromatic rings (Scheme 3).</p><p>Bis-(triazol-4-yl)methyl carbamates with different substitution pattern on the aromatic rings were synthesized according to Scheme 3. Aromatic azides 12a-c were reacted with propargyl amine, under click chemistry conditions, to give the corresponding aminomethyl triazoles 35a-c, which were subsequently coupled with prop-2-ynyl chloroformate to afford N-(triazol-4-yl)methyl-O-propargyl carbamates 36a-c. The latter compounds were reacted with aromatic azides 12a-b, under click chemistry conditions, to afford N-(triazol-4-yl)methyl-O-(triazol-4-yl)methyl carbamates 37-40.</p><p>The compounds bearing a (1,2,4-triazol-3-yl-)methyl moiety, 43 and 46, were synthesized as reported in Scheme 4. Compound 43 was prepared by lithium aluminum hydride reduction of the commercially available 1-phenyl-1,2,4-triazole-3-carboxylic acid 41 followed by reaction with cyclohexyl isocyanate. Then, the commercially available methyl 1H-1,2,4-triazole-3-carboxylate 44 was reacted with benzyl bromide in presence of potassium carbonate to obtain compound 45. Reduction by lithium aluminum hydride followed by reaction with cyclohexyl isocyanate afforded compound 46.</p><!><p>The compounds were tested for their ability to inhibit the hydrolysis of [3H]anandamide by FAAH prepared from rat brain homogenates. Median inhibitory concentration (IC50) values are reported in Tables 1-4.</p><p>In a first attempt to improve the stability of O-biphenyl carbamate FAAH inhibitors, we replaced the O-(3-phenylphenyl) residue of 1a with an O-(3-phenylphenyl)methyl group, as in compound 4. This change caused an almost complete loss of activity, as 4 showed only 65% inhibition of FAAH activity at 100 µM (Table 1). Interestingly, the substitution of the O-(3-phenylphenyl)methyl group with an O-(1-phenyl-1,2,3-triazol-4-yl)methyl residue, as in compound 17 (IC50 = 381 nM), allowed to recover the FAAH inhibitory activity. Encouraged by this result, we synthesized a set of close analogues of 17, compounds 18-19 and 32-33, to identify the best substituents at position 1 and 4 of the triazole ring for FAAH inhibition. The results are reported in Table 1.</p><p>The replacement of the phenyl group at position 1 of the triazole with a benzyl residue, compound 18 (IC50 = 26 nM), led to a ca. 15-fold increase in potency. It is interesting to note that this compound was only 4.6-fold less potent than 1a (URB524). When the methylene linker at position 1 or 4 of the triazole ring was substituted by an ethylene moiety, as in compounds 19 and 33, respectively, a drop in potency with respect to 18 was observed. The replacement of the 4-methylene residue in compound 17 with a 4-ethylene one, leading to compound 32 (IC50 = 1278 nM), resulted in a ca. 3-fold decrease in potency.</p><p>We then asked whether 1-phenyl- or 1-benzyl-(1,2,4-triazol-3yl)methyl moieties worked as a replacement of the O-(3-phenylphenyl) residue of 1a and whether they were interchangeable with the isomeric 1-phenyl- and 1-benzyl-(1,2,3-triazol-4-yl)methyl residues. Compounds 43 and 46 showed a dramatic decrease in potency with respect to 1a and, surprisingly, were much less potent than 17 and 18. We speculate that the loss in potency might be the ascribed to the different electronic properties of the 1,2,4-triazole ring with respect to the isomeric 1,2,3-triazole counterpart, leading to an unfavourable interaction with the active site of the enzyme.</p><p>To test our hypothesis that O-(triazol-4-yl)alkyl carbamate derivatives are more stable than O-aryl carbamates, we compared the rat plasma stability of compounds 1a, 17 and 18. The results are reported in Table 1. Consistent with our expectation, the O-(1,2,3-triazol-4-yl)methyl carbamates 17 and 18 showed a significantly higher plasma stability than did 1a. The latter compound displayed half-life of 62 min[26], but was no longer detectable after 7 hours incubation with rat plasma. In contrast, ca. 90% of the initial amount of compound 17 and 18 was still detectable after 7 hours.</p><p>The limited decrease in potency of compound 18 vs. 1a, coupled with its significantly higher plasma stability, prompted us to explore further this chemical class. We first investigated region C (Figure 2) by preparing a series of compounds bearing variously substituted benzyl residues at position 1 of the triazole ring. The results are summarized in Table 2.</p><p>Replacement of the benzyl residue in 18 with a benzhydryl moiety, compound 20 (73% inhibition at 100 µM), or a 2-naphthylmethyl group, compound 21 (IC50 = 2.0 µM), led to a significant decrease in FAAH inhibitory potency, indicating that bulky arylmethyl groups linked to the triazole were not tolerated, most likely because of steric clash at the active site of the enzyme.</p><p>Interestingly, the nature of the substituent on the phenyl ring appeared to have a limited effect on the potency of the compounds as FAAH inhibitors. In fact, benzyl residues bearing both electron-withdrawing (CN, F, Cl) and electron-donating (Me, OMe) substituents at ortho or meta position all led to low-nanomolar inhibitors. Among them, the ortho-methoxybenzyl derivative 26a showed the highest potency (IC50 = 1.4 nM). All the para-substituted derivatives were less potent than the corresponding ortho- or meta-substituted analogues, irrespective of the electronic properties of the susbtituent. In particular, compound 22c (IC50 = 2282 nM), bearing a para-cyano group, showed the highest loss in potency within this sub-set of analogues. As for the naphthylmethyl compound 21, we interpret this finding as the result of an unfavorable steric interaction between the para-substituted phenyl ring and the active site of the enzyme. Together, from this small series of derivatives the rank order of potency ortho- > meta- > para-substituted compounds clearly emerged.</p><p>The excellent potency of benzyl derivatives bearing a fluorine or a methoxyl group at position ortho and meta led us to synthesize the di-substituted compounds 27-30 to verify whether any additive effect on potency was observed. With the exception of the 2-fluoro-3-methoxy-derivative 30 (IC50 = 44.6 nM), all the compounds retained an excellent potency, with IC50 in the range 10.4 - 11.9 nM, but none of them improved significantly over the corresponding mono-substituted analogue.</p><p>The most potent compound, 26a, was effective at inhibiting FAAH activity ex vivo. One hour after systemic administration of 26a (3 mg/kg, intraperitoneally) to CD1 mice, FAAH activity measured ex vivo in brain tissue was reduced by 78% (n=3) with respect to control.</p><p>As the next step in the investigation of the SAR of this new class of FAAH inhibitors, we conducted a preliminary exploration of region A (Figure 2). Previous studies on O-arylcarbamates showed that replacement of the cyclohexyl group of 1a with an arylalkyl moiety led to inhibitors of greater potency.[15a] Exploiting the click chemistry approach, we replaced the cyclohexyl group with a [1-(methoxybenzyl)triazol-4-yl]methyl residue, as in compounds 34a and 34b, and 37-40. The results are reported in Table 3.</p><p>Compound 34a, bearing an ortho-methoxybenzyl residue on both triazolyl rings turned out to be the least potent derivative (IC50 = 154 nM). Moving the methoxyl group to the meta position of the N-(1-benzyltriazol-4-yl)methyl moiety led to derivative 37 (IC50 = 3.2 nM), which showed a 48-fold increase in potency vs. 34a and confirmed as the most potent compound of this small series. Consistent with the previous finding, replacement of the ortho-methoxybenzyl residue at position 1 of the triazole ring in region C with a meta-methoxybenzyl group, as in compound 39 (IC50 = 9.8 nM), was accompanied by a ca. 15-fold increase in potency with respect to 34a. Introduction of a meta-methoxybenzyl group at position 1 on both triazolyl rings led to the potent inhibitor 34b (IC50 = 3.9 nM). Interestingly, a para-methoxybenzyl group at position 1 of the (triazol-4-yl)methyl moiety in region A was not detrimental for potency, as compounds 38 and 40 inhibited FAAH activity with IC50 of 7.6 and 5.8 nM, respectively.</p><p>The most interesting compounds identified from the SAR exploration were tested for their inhibitory activity against human FAAH-1 (h-FAAH-1). A comparison between the inhibitory potency of selected compounds on r-FAAH vs. h-FAAH-1 is reported in Table 4.</p><p>The series of substituted O-(1-benzyltriazol-4-yl)methyl N-cyclohexylcarbamate derivatives resulted to be generally less active at inhibiting h-FAAH-1 than r-FAAH, displaying a 14- to 240-fold drop in potency. The only exception was represented by 27 (IC50 = 3.6 nM), which showed 3-fold higher potency on h-FAAH-1 with respect to r-FAAH. Moreover, the ortho-substituted analogues 23a, 24a, and 26a, which displayed single-digit nanomolar r-FAAH inhibition, and the 2,6-dibustituted derivative 28, suffered the most marked loss of potency: 62, 172, 240, and 73-fold respectively. The meta-substituted analogues 23b, 24b, and 26b, and the 3,5-disubstituted derivative 29 were more potent inhibitors than their ortho-substituted analogues, thus reversing the preference for ortho-substituted benzyl residues observed with r-FAAH.</p><p>By contrast, compounds bearing substituted (1-benzyltriazol-4-yl)methyl residues at both region A and C (Figure 2) displayed a minor decrease in activity on h-FAAH-1 compared to the N-cyclohexylcarbamate derivatives. Indeed, compounds 37, 38, and 39 showed 4 to 10-fold lower potency, but retained double-digit nanomolar potency on h-FAAH-1. The most promising molecules in terms of potency were 34b and 40 that possessed IC50 for h-FAAH-1 inhibition of 4.2 nM and 9.4 nM respectively. Interestingly, both of them are characterized by a meta-methoxybenzyl group on region C (Figure 2), suggesting that the binding site on h-FAAH-1 prefers meta-substituents on that benzyl group.</p><p>The most potent compounds against both rat and human FAAH, i.e. 27, 34b, and 40, were tested for their selectivity versus monoacylglycerol lipase (MGL), a serine hydrolase that inactivates the endocannabinoid 2-arachidonoylglycerol (2-AG).[27] None of the compounds inhibited MGL activity when tested at concentrations up to 100 μM (Figure S1). The selective inhibition of FAAH activity by compounds 27, 34b, and 40 is in agreement with previous observations with the O-arylcarbamates 1a and 1b.[28]</p><p>Finally, compounds 27, 34b, and 40 were further characterized by determining their rat plasma and mouse liver microsomal (MLM) stability, and their kinetic solubility in buffer. A comparison of the overall profile of O-arylcarbamate 1a with compounds 27, 34b, and 40 is reported in Table 5.</p><p>The selected O-[(1-benzyltriazol-4-yl)methyl]carbamate derivatives 27, 34b, and 40 showed FAAH inhibitory activity and MLM stability comparable to those of the O-arylcarbamate 1a. However, they displayed much higher rat plasma stability than 1a, as ca. 90% of the initial amount of the compounds was still present after 7 hours incubation with rat plasma. Moreover, with the exception of compound 27, the kinetic solubility in buffer also improved significantly.</p><!><p>In the present study, we report the synthesis and characterization of O-(1,2,3-triazol-4-yl)alkyl carbamates as a novel class of FAAH inhibitors. In these compounds, an O-(triazol-4-yl)methyl group replaces the O-aryl moiety of known and potent FAAH inhibitors such as compound 1a (URB524) and 1b (URB597). A number of compounds were prepared by copper-catalyzed [3 + 2] cycloaddition reactions between azides and alkynes (click chemistry). Exploiting the same chemistry, we also synthesized carbamates bearing a substituted (1-benzyltriazol-4-yl)methyl moiety at both the O and N end. The click chemistry approach allowed us rapidly to explore the structure-activity relationships within the class. Several single-digit nanomolar inhibitors of rat FAAH were obtained, including the potent derivative 26a, which showed an IC50 value of 1.4 nM and inhibited brain FAAH activity in vivo. Some of these compounds potently inhibited human FAAH-1. In particular, compounds 34b and 40, bearing a [1-[(methoxyphenyl)methyl]triazol-4-yl]methyl group at both the O and N end of the carbamate function, displayed single-digit nanomolar IC50 values for both rat and human FAAH. In addition, they showed a remarkable improvement in rat plasma stability and kinetic solubility in buffer with respect to URB524.</p><p>The dramatic decrease in FAAH inhibitory activity of 1-phenyl- or 1-benzyl-substituted O-(1,2,4-triazol-3-yl)mehtyl carbamates 43 and 46 demonstrated that the 1-substituted-(1,2,3-triazol-4-yl)methyl core structure, easily accessible by click chemistry, was essential for obtaining potent inhibition of FAAH activity.</p><p>In conclusion, exploiting a click chemistry approach we prepared a novel series of potent and drug-like FAAH inhibitors containing an O-(1,2,3-triazol-4-yl)alkyl carbamate moiety. The compounds described in the present study represent a promising starting point for the development of new FAAH inhibitors with improved drug-like properties.</p><!><p>Chemicals, Materials and Methods. Solvents and reagents were obtained from commercial suppliers and were used without further purification. For simplicity, solvents and reagents were indicated as follows: acetonitrile (CH3CN), benzyl bromide (BnBr), cyclohexane (Cy), dichloromethane (DCM), diethyl ether (Et2O), 4-(dimethylamino)-pyridine (DMAP), ethanol (EtOH), ethyl acetate (EtOAc), hydrochloric acid (HCl), methanol (MeOH), N,N'-dimethylformamide (DMF), room temperature (rt), sodium sulfate (Na2SO4), sodium bicarbonate (NaHCO3), sulfuric acid (H2SO4), tert-butanol (t-BuOH), tetrahydrofuran (THF), triethylamine (Et3N).</p><p>Automated column chromatography purifications were performed by using a Teledyne ISCO apparatus (CombiFlash® Rf) with pre-packed silica gel columns of different sizes (from 4 g to 120 g). Mixtures of increasing polarity of Cy and EtOAc or DCM and MeOH were used as eluents. NMR experiments were run on a Bruker Avance III 400 system (400.13 MHz for 1H, and 100.62 MHz for 13C), equipped with a BBI probe and Z-gradients. Spectra were acquired at 300 K, using deuterated dimethylsulfoxide ([D6]DMSO) or deuterated chloroform (CDCl3) as solvents. Chemical shifts for 1H and 13C spectra were recorded in parts per million using the residual non-deuterated solvent as the internal standard (for CDCl3: 7.26 ppm, 1H and 77.16 ppm, 13C; for [D6]DMSO: 2.50 ppm, 1H; 39.52 ppm, 13C). UPLC/MS analyses were run on a Waters ACQUITY UPLC/MS system consisting of a Single Quadropole Detector (SQD) Mass Spectrometer (MS) equipped with an Electrospray Ionization (ESI) interface and a Photodiode Array (PDA) Detector. PDA range was 210-400 nm. ESI in positive and negative mode was applied. Mobile phases: (A) 10mM NH4OAc in H2O, pH 5; (B) 10mM NH4OAc in MeCN/H2O (95:5) pH 5. Analyses were performed either with method A or B. Method A: gradient 5 to 95% B over 3 min; flow rate 0.5 mL/min; temperature 40 °C. Pre column: Vanguard BEH C18 (1.7µm 2.1x5mm). Column: BEH C18 (1.7µm 2.1x50mm). Method B: gradient: 50 to 100% B over 3 min, flow rate 0.5 mL/min; temperature 40 °C. Pre column: Vanguard BEH C18 (1.7µm 2.1x5mm). Column: BEH C18 (1.7µm 2.1x50mm). Accurate mass measurement (HMRS) was performed on a Synapt G2 Quadrupole-Tof Instrument (Waters, USA), equipped with an ESI ion source.</p><p>All final compounds (4, 17-30, 32, 33, 34a-b, 37-40, 43 and 46) showed ≥ 95% purity by NMR and UPLC/MS analysis. The syntheses of reaction intermediates 3, 5a-c, 6-16, 31a-b, 35a-c, 36a-c, 41, 42a-b, and 45 are described in the Supporting Information.</p><!><p>1 equiv. of the ethynyl derivatives and 1 equiv. of the azido compounds were suspended in a solution of water / t-BuOH 1:1. Sodium ascorbate (0.1 eq) of a freshly prepared 1 M solution in water was added, followed by the addition of copper (II) sulfate pentahydrate (0.01 eq). The resulting reaction vigorously stirred for 3 h at rt. The reaction mixture was then diluted with water, cooled on ice, and the precipitate was collected by filtration. When addition of water failed to precipitate the desired triazoles, evaporation of the solvent allowed the recovery of the crude products. Purifications were performed by column chromatography.</p><!><p>0.5 equiv. of the ethynyl derivatives and 1 equiv. of the azido compounds were suspended in a solution of water / t-BuOH 1:1. Sodium ascorbate (0.1 eq) of a freshly prepared 1 M solution in water was added, followed by the addition of copper (II) sulfate pentahydrate (0.01 eq). The resulting reaction vigorously stirred for 8 h rt. Afterwards, evaporation of the solvent allowed the recovery of the crude products. Purification were performed by column chromatography.</p><!><p>(3-phenylphenyl)methanol (3, 0.125 g, 0.68 mmol) were dissolved in dry CH3CN (5 mL) while stirring at rt. Then, DMAP (0.08 g, 0.68 mmol) and cyclohexyl isocyanate (0.09 g, 0.75 mmol) were added and the reaction mixture was stirred at 80 °C for 3 h. Afterwards, the reaction mixture was diluted with EtOAc and washed once with 2N HCl, and once with brine. The organic layer was dried over Na2SO4 and concentrated in vacuo. The crude product was purified by chromatography employing a gradient of MeOH in DCM from 0% to 2%, to afford compound 4 as a white powder (0.13 g; 61%): 1H NMR (400 MHz, [D6]DMSO): δ=7.69 – 7.61 (m, 3H), 7.60 (dt, J=7.9, 1.4 Hz, 1H), 7.47 (q, J=7.3 Hz, 3H), 7.42 – 7.31 (m, 2H), 7.19 (d, J=7.9 Hz, 1H), 5.07 (s, 2H), 3.31 – 3.22 (m, 1H), 1.76 (dd, J=12.6, 3.6 Hz, 2H), 1.67 (dq, J=12.5, 3.7 Hz, 2H), 1.53 (dq, J=11.5, 3.7 Hz, 1H), 1.31 – 1.00 ppm (m, 5H); 13C NMR (100 MHz, [D6]DMSO): δ=155.2, 140.2, 139.9, 138.0, 128.9, 128.9, 127.5, 126.7, 126.6, 126.0 (2C), 65.0, 49.5, 32.6 (2C), 25.1, 24.6 ppm (2C); UPLC-MS: Method B, Rt 1.97, ionization: m/z 310 [M+H]+; HRMS–ESI: m/z [M+Na]+ calcd for C20H23NO2Na: 332.1626, found: 332.1622.</p><!><p>The reaction was carried out following general procedure (1), using prop-2-ynyl N-cyclohexylcarbamate (0.23 g, 1.26 mmol) and azidobenzene (0.15 g, 1.26 mmol), sodium ascorbate (0.02, 0.12 mmol), and copper (II) sulfate pentahydrate (0.003 g, 0.01 mmol) in water / t-BuOH 1:1 (3 mL). Purification was performed by flash chromatography (SiO2) eluting with a gradient from 0 to 2% MeOH in DCM, to afford compound 17 as a white powder (0.22 g; 59%): 1H NMR (400 MHz, [D6]DMSO): δ=8.82 (s, 1H), 7.91 (m, 2H), 7.61 (m, 2H), 7.52 (m, 1H), 7.20 (d, J=8.0 Hz, 1H), 5.13 (s, 2H), 3.28 (m, 1H), 1.76 (m, 2H), 1.67 (dt, J=12.2, 3.8 Hz, 2H), 1.54 (m, 1H), 1.18 ppm (m, 5H); 13C NMR (100 MHz, [D6]DMSO): δ=155.4, 144.3, 137.0, 130.3 (2-C), 129.2, 123.2, 120.6 (2-C), 57.0, 50.0, 33.1 (2-C), 25.6, 25.0 ppm (2-C); UPLC-MS: Method A, Rt 2.42, ionization: m/z 301 [M+H]+; HRMS–ESI: m/z [M+H]+ calcd for C16H20N4O2: 301.1665, found: 301.1666.</p><!><p>The reaction was carried out following general procedure (1), using prop-2-ynyl N-cyclohexylcarbamate (0.15 g, 0.82 mmol) and azidomethylbenzene (0.11 g, 0.82 mmol), sodium ascorbate (0.016, 0.08 mmol), and copper (II) sulfate pentahydrate (0.002 g, 0.008 mmol) in water / t-BuOH 1:1 (3 mL). Purification was performed by flash chromatography (SiO2) eluting with a gradient from 0 to 50% EtOAc in Cy, to afford compound 18 as a white powder (0.18 g; 71%): 1H NMR (400 MHz, [D6]DMSO): δ=8.15 (s, 1H), 7.35 (m, 5H), 7.13 (d, J=7.9 Hz, 1H), 5.60 (s, 2H), 5.02 (s, 2H), 3.25 (m, 1H), 1.69 (m, 4H), 1.53 (dt, J=12.7, 3.8 Hz, 1H), 1.15 ppm (m, 5H); 13C NMR (100 MHz, [D6]DMSO): δ=155.4, 143.5, 136.5, 129.2 (2-C), 128.6, 128.4 (2-C), 125.0, 57.2, 53.2, 49.9, 33.0 (2-C), 25.6, 25.0 ppm (2-C); UPLC-MS: Method A, Rt 2.37, ionization: m/z 315 [M+H]+; HRMS–ESI: m/z [M+H]+ calcd for C17H22N4O2: 315.1821, found: 315.1826.</p><!><p>The reaction was carried out following general procedure (1), using prop-2-ynyl N-cyclohexylcarbamate (0.15 g, 0.82 mmol) and 2-azidoethylbenzene (0.12 g, 0.82 mmol), sodium ascorbate (0.016, 0.08 mmol), and copper (II) sulfate pentahydrate (0.002 g, 0.008 mmol) in water / t-BuOH 1:1 (3 mL). Purification was performed by flash chromatography (SiO2) eluting with a gradient from 0 to 2% MeOH in DCM, to afford compound 19 as a white powder (0.135 g; 50%): 1H NMR (400 MHz, [D6]DMSO): δ=8.03 (s, 1H), 7.29 (m, 2H), 7.21 (m, 3H), 7.13 (d, J=7.9 Hz, 1H), 4.99 (s, 2H), 4.61 (dd, J=7.9, 6.8 Hz, 2H), 3.24 (m, 1H), 3.16 (t, J=7.4 Hz, 2H), 1.69 (m, 4H), 1.54 (d, J=13.3 Hz, 1H), 1.18 ppm (m, 5H); 13C NMR (100 MHz, [D6]DMSO): δ=154.6, 142.5, 137.5, 128.6 (2-C), 128.3 (2-C), 126.5, 124.4, 56.6, 50.3, 49.4, 35.6, 32.6 (2-C), 25.1, 24.5 ppm (2-C); UPLC-MS: Method A, Rt 2.44, ionization: m/z 329 [M+H]+; HRMS–ESI: m/z [M+H]+ calcd for C18H24N4O2: 329.1978, found: 329.1982.</p><!><p>The reaction was carried out following general procedure (1), using prop-2-ynyl N-cyclohexylcarbamate (0.15 g, 0.82 mmol) and [azido(phenyl)methyl]benzene (0.17 g, 0.82 mmol), sodium ascorbate (0.016, 0.08 mmol), and copper (II) sulfate pentahydrate (0.002 g, 0.008 mmol) in water / t-BuOH 1:1 (3 mL). Purification was performed by flash chromatography (SiO2) eluting with a gradient from 0 to 40% EtOAc in Cy, to afford compound 20 as a white powder (0.23 g; 71%): 1H NMR (400 MHz, [D6]DMSO): δ=8.08 (s, 1H), 7.39 (m, 6H), 7.31 (s, 1H), 7.21 (m, 4H), 7.14 (d, J=7.9 Hz, 1H), 5.02 (s, 2H), 3.22 (m, 1H), 1.67 (m, 4H), 1.52 (d, J=12.5 Hz, 1H), 1.12 ppm (m, 5H); 13C NMR (100 MHz, [D6]DMSO): δ=155.4, 144.3, 138.6 (2-C), 128.7 (4-C), 128.2 (2-C), 127.9 (4-C), 66.5, 56.7, 49.4, 32.5 (2-C), 25.1, 24.5 ppm (2-C); UPLC-MS: Method A, Rt 2.81, ionization: m/z 391 [M+H]+; HRMS–ESI: m/z [M+H]+ calcd for C23H26N4O2: 391.2134, found: 391.2132.</p><!><p>The reaction was carried out following general procedure (1), using prop-2-ynyl N-cyclohexylcarbamate (0.15 g, 0.82 mmol) and 2-(azidomethyl)naphthalene (0.15 g, 0.82 mmol), sodium ascorbate (0.016, 0.08 mmol), and copper (II) sulfate pentahydrate (0.002 g, 0.008 mmol) in water / t-BuOH 1:1 (3 mL). Purification was performed by flash chromatography (SiO2) eluting with a gradient from 0 to 2% MeOH in DCM, to afford compound 21 as a white powder (0.16 g; 54%): 1H NMR (400 MHz, [D6]DMSO): δ=8.19 (s, 1H), 7.92 (m, 3H), 7.86 (s, 1H), 7.54 (m, 2H), 7.44 (dd, J=8.5, 1.6 Hz, 1H), 7.10 (d, J=7.8 Hz, 1H), 5.76 (s, 2H), 5.01 (s, 2H), 3.22 (m, 1H), 1.66 (dd, J=26.8, 12.6 Hz, 4H), 1.51 (d, J=12.5 Hz, 1H), 1.11 ppm (m, 5H); 13C NMR (100 MHz, [D6]DMSO): δ=154.9, 143.0, 133.4, 132.7, 132.4, 128.4, 127.7, 127.5, 126.9, 126.5, 126.4, 125.7, 124.6, 56.7, 52.9, 49.4, 32.5 (2-C), 25.1, 24.5 ppm (2-C); UPLC-MS: Method A, Rt 2.65, ionization: m/z 365 [M+H]+; HRMS–ESI: m/z [M+H]+ calcd for C21H24N4O2: 365.1978, found: 365.1975.</p><!><p>The reaction was carried out following general procedure (1), using prop-2-ynyl N-cyclohexylcarbamate (0.15 g, 0.82 mmol) and 2-(azidomethyl)benzonitrile (0.13 g, 0.82 mmol), sodium ascorbate (0.016, 0.08 mmol), and copper (II) sulfate pentahydrate (0.002 g, 0.008 mmol) in water / t-BuOH 1:1 (3 mL). Purification was performed by flash chromatography (SiO2) eluting with a gradient from 0 to 2% MeOH in DCM, to afford compound 22a as a white powder (0.15 g; 53%): 1H NMR (400 MHz, [D6]DMSO): δ=8.19 (s, 1H), 7.92 (dd, J=7.7, 1.0 Hz, 1H), 7.72 (td, J=7.7, 1.2 Hz, 1H), 7.57 (td, J=7.7, 0.9 Hz, 1H), 7.36 (d, J=7.8 Hz, 1H), 7.14 (d, J=7.8 Hz, 1H), 5.81 (s, 2H), 5.03 (s, 2H), 3.25 (m, 1H), 1.68 (dd, J=27.0, 12.5 Hz, 4H), 1.53 (d, J=12.7 Hz, 1H), 1.13 ppm (dtd, J=31.1, 24.1, 12.1 Hz, 5H); 13C NMR (100 MHz, [D6]DMSO): δ=154.9, 143.1, 138.7, 133.8, 133.3, 129.4, 129.2, 125.0, 116.9, 111.2, 56.6, 50.9, 49.4, 32.6 (2-C), 25.1, 24.5 ppm (2-C); UPLC-MS: Method A, Rt 2.25, ionization: m/z 340 [M+H]+; HRMS–ESI: m/z [M+H]+ calcd for C18H21N5O2: 340.1773, found: 340.1779.</p><!><p>The reaction was carried out following general procedure (1), using prop-2-ynyl N-cyclohexylcarbamate (0.15 g, 0.82 mmol) and 3-(azidomethyl)benzonitrile (0.13 g, 0.82 mmol), sodium ascorbate (0.016, 0.08 mmol), and copper (II) sulfate pentahydrate (0.002 g, 0.008 mmol) in water / t-BuOH 1:1 (3 mL). Purification was performed by flash chromatography (SiO2) eluting with a gradient from 0 to 2% MeOH in DCM, to afford compound 22b as a white powder (0.17 g; 62%): 1H NMR (400 MHz, [D6]DMSO): δ=8.21 (s, 1H), 7.82 (m, 2H), 7.62 (m, 2H), 7.12 (d, J=7.8 Hz, 1H), 5.67 (s, 2H), 5.01 (s, 2H), 3.24 (m, 1H), 1.68 (dd, J=25.4, 12.5 Hz, 4H), 1.52 (d, J=12.6 Hz, 1H), 1.13 ppm (m, 5H); 13C NMR (100 MHz, [D6]DMSO): δ=154.9, 143.2, 137.5, 132.9, 131.9, 131.6, 130.0, 124.8, 118.3, 111.6, 56.6, 51.7, 49.4, 32.6 (2-C), 25.1, 24.5 ppm (2-C); UPLC-MS: Method A, Rt 2.25, ionization: m/z 340 [M+H]+; HRMS–ESI: m/z [M+H]+ calcd for C18H21N5O2: 340.1773, found: 340.1781.</p><!><p>The reaction was carried out following general procedure (1), using prop-2-ynyl N-cyclohexylcarbamate (0.15 g, 0.82 mmol) and 4-(azidomethyl)benzonitrile (0.13 g, 0.82 mmol), sodium ascorbate (0.016, 0.08 mmol), and copper (II) sulfate pentahydrate (0.002 g, 0.008 mmol) in water / t-BuOH 1:1 (3 mL). Purification was performed by flash chromatography (SiO2) eluting with a gradient from 0 to 2% MeOH in DCM, to afford compound 22c as a white powder (0.21 g; 77%): 1H NMR (400 MHz, [D6]DMSO): δ=8.21 (s, 1H), 7.86 (d, J=8.2 Hz, 2H), 7.45 (d, J=8.2 Hz, 2H), 7.14 (d, J=7.8 Hz, 1H), 5.72 (s, 2H), 5.02 (s, 2H), 3.24 (m, 1H), 1.69 (dd, J=25.1, 12.6 Hz, 4H), 1.53 (d, J=12.4 Hz, 1H), 1.14 ppm (m, 5H); 13C NMR (100 MHz, [D6]DMSO): δ=154.9, 143.2, 141.4, 132.7 (2-C), 128.6 (2-C), 124.9, 118.4, 110.9, 56.6, 52.1, 49.4, 32.6 (2-C), 25.1, 24.5 ppm (2-C); UPLC-MS: Method A, Rt 2.22, ionization: m/z 340 [M+H]+; HRMS–ESI: m/z [M+Na]+ calcd for C18H21N5O2Na: 362.1593, found: 362.1594.</p><!><p>The reaction was carried out following general procedure (1), using prop-2-ynyl N-cyclohexylcarbamate (0.15 g, 0.82 mmol) and 1-(azidomethyl)-2-fluoro-benzene (0.12 g, 0.82 mmol), sodium ascorbate (0.016, 0.08 mmol), and copper (II) sulfate pentahydrate (0.002 g, 0.008 mmol) in water / t-BuOH 1:1 (3 mL). Purification was performed by flash chromatography (SiO2) eluting with a gradient from 0 to 2% MeOH in DCM, to afford compound 23a as a white powder (0.17 g; 61%): 1H NMR (400 MHz, [D6]DMSO): δ=8.11 (s, 1H), 7.42 (m, 1H), 7.34 (td, J=7.6, 1.4 Hz, 1H), 7.23 (m, 2H), 7.13 (d, J=7.8 Hz, 1H), 5.66 (s, 2H), 5.00 (s, 2H), 3.24 (m, 1H), 1.67 (m, 4H), 1.52 (d, J=12.5 Hz, 1H), 1.12 ppm (m, 5H); 13C NMR (100 MHz, [D6]DMSO): δ=160.0 (d, J=246.7 Hz), 155.0, 143.0, 130.7 (d, J=4.6 Hz), 130.7, 124.8 (d, J=3.4 Hz), 124.7, 122.8 (d, J=14.7 Hz), 115.6 (d, J=20.8 Hz), 56.7, 49.4, 46.8 (d, J=3.7 Hz), 32.6 (2-C), 25.1, 24.5 ppm (2-C); UPLC-MS: Method A, Rt 2.37, ionization: m/z 333 [M+H]+; HRMS–ESI: m/z [M+H]+ calcd for C17H21FN4O2: 333.1727, found: 333.1732.</p><!><p>The reaction was carried out following general procedure (1), using prop-2-ynyl N-cyclohexylcarbamate (0.15 g, 0.82 mmol) and 1-(azidomethyl)-3-fluoro-benzene (0.12 g, 0.82 mmol), sodium ascorbate (0.016, 0.08 mmol), and copper (II) sulfate pentahydrate (0.002 g, 0.008 mmol) in water / t-BuOH 1:1 (3 mL). Purification was performed by flash chromatography (SiO2) eluting with a gradient from 0 to 50% EtOAc in Cy, to afford compound 23b as a white powder (0.17 g; 63%): White powder; yield 63%; 1H NMR (400 MHz, [D6]DMSO): δ=8.18 (s, 1H), 7.42 (m, 1H), 7.16 (m, 4H), 5.62 (s, 2H), 5.01 (s, 2H), 3.24 (m, 1H), 1.68 (dd, J=26.2, 12.5 Hz, 4H), 1.52 (d, J=12.5 Hz, 1H), 1.12 ppm (m, 5H); 13C NMR (100 MHz, [D6]DMSO): δ=162.5 (d, J=244.3 Hz), 155.4, 143.6, 139.2, 131.3 (d, J=8.3 Hz), 125.2, 124.4 (d, J=2.7 Hz), 115.3 (m, 2-C), 57.1, 52.5, 49.9, 33.0 (2-C), 25.6, 25.0 ppm (2-C); UPLC-MS: Method A, Rt 2.40, ionization: m/z 333 [M+H]+; HRMS–ESI: m/z [M+H]+ calcd for C17H21FN4O2: 333.1727, found: 333.1731.</p><!><p>The reaction was carried out following general procedure (1), using prop-2-ynyl N-cyclohexylcarbamate (0.15 g, 0.82 mmol) and 1-(azidomethyl)-4-fluoro-benzene (0.12 g, 0.82 mmol), sodium ascorbate (0.016, 0.08 mmol), and copper (II) sulfate pentahydrate (0.002 g, 0.008 mmol) in water / t-BuOH 1:1 (3 mL). Purification was performed by flash chromatography (SiO2) eluting with a gradient from 0 to 2% MeOH in DCM, to afford compound 23c as a white powder (0.13 g; 49%): 1H NMR (400 MHz, [D6]DMSO): δ=8.14 (s, 1H), 7.38 (ddd, J=8.4, 5.3, 2.5 Hz, 2H), 7.20 (m, 2H), 7.12 (d, J=7.5 Hz, 1H), 5.58 (s, 2H), 5.00 (s, 2H), 3.23 (m, 1H), 1.67 (m, 4H), 1.52 (d, J=12.5 Hz, 1H), 1.11 ppm (m, 5H); 13C NMR (100 MHz, [D6]DMSO): δ=161.8 (d, J=244.5 Hz), 154.9, 143.0, 132.2, 130.2 (d, J=8.4 Hz), 124.5, 115.5 (d, J=21.6 Hz), 56.7, 51.9, 49.4, 32.6 (2-C), 25.1, 24.5 ppm (2-C); UPLC-MS: Method A, Rt 2.39, ionization: m/z 333 [M+H]+; HRMS–ESI: m/z [M+H]+ calcd for C17H21FN4O2: 333.1727, found: 333.1731.</p><!><p>The reaction was carried out following general procedure (1), using prop-2-ynyl N-cyclohexylcarbamate (0.15 g, 0.82 mmol) and 1-(azidomethyl)-2-chloro-benzene (0.14 g, 0.82 mmol), sodium ascorbate (0.016, 0.08 mmol), and copper (II) sulfate pentahydrate (0.002 g, 0.008 mmol) in water / t-BuOH 1:1 (3 mL). Purification was performed by flash chromatography (SiO2) eluting with a gradient from 0 to 2% MeOH in DCM, to afford compound 24a as a white powder (0.19 g; 68%): 1H NMR (400 MHz, [D6]DMSO): δ=8.11 (s, 1H), 7.52 (dd, J=7.7, 1.5 Hz, 1H), 7.38 (m, 2H), 7.22 (dd, J=7.4, 1.7 Hz, 1H), 7.14 (d, J=7.7 Hz, 1H), 5.70 (s, 2H), 5.02 (s, 2H), 3.24 (m, 1H), 1.68 (dd, J=26.2, 12.5 Hz, 4H), 1.53 (m, 1H), 1.11 ppm (m, 5H); 13C NMR (100 MHz, [D6]DMSO): δ=154.9, 142.9, 133.2, 132.6, 130.4, 130.2, 129.6, 127.6, 124.9, 56.7, 50.5, 49.4, 32.6 (2-C), 25.1, 24.5 ppm (2-C); UPLC-MS: Method A, Rt 2.49, ionization: m/z 349 [M+H]+; HRMS–ESI: m/z [M+H]+ calcd for C17H21ClN4O2: 349.1431, found: 349.1435.</p><!><p>The reaction was carried out following general procedure (1), using prop-2-ynyl N-cyclohexylcarbamate (0.15 g, 0.82 mmol) and 1-(azidomethyl)-3-chloro-benzene (0.14 g, 0.82 mmol), sodium ascorbate (0.016, 0.08 mmol), and copper (II) sulfate pentahydrate (0.002 g, 0.008 mmol) in water / t-BuOH 1:1 (3 mL). Purification was performed by flash chromatography (SiO2) eluting with a gradient from 0 to 50% EtOAc in Cy, to afford compound 24b as a white powder (0.24 g; 83%): 1H NMR (400 MHz, [D6]DMSO): δ=8.19 (s, 1H), 7.40 (dd, J=6.1, 2.3 Hz, 3H), 7.27 (dq, J=5.9, 2.8 Hz, 1H), 7.13 (d, J=7.9 Hz, 1H), 5.61 (s, 2H), 5.01 (s, 2H), 3.25 (m, 1H), 1.68 (dd, J=26.4, 12.4 Hz, 4H), 1.52 (d, J=12.5 Hz, 1H), 1.13 ppm (m, 5H); 13C NMR (100 MHz, [D6]DMSO): δ=154.9, 143.1, 138.4, 133.2, 130.6, 128.1, 127.8, 126.6, 124.7, 56.6, 51.9, 49.4, 32.6 (2-C), 25.1, 24.5 ppm (2-C); UPLC-MS: Method A, Rt 2.55, ionization: m/z 349 [M+H]+; HRMS–ESI: m/z [M+H]+ calcd for C17H21ClN4O2: 349.1431, found: 349.1436.</p><!><p>The reaction was carried out following general procedure (1), using prop-2-ynyl N-cyclohexylcarbamate (0.15 g, 0.82 mmol) and 1-(azidomethyl)-4-chloro-benzene (0.14 g, 0.82 mmol), sodium ascorbate (0.016, 0.08 mmol), and copper (II) sulfate pentahydrate (0.002 g, 0.008 mmol) in water / t-BuOH 1:1 (3 mL). Purification was performed by flash chromatography (SiO2) eluting with a gradient from 0 to 50% EtOAc in Cy, to afford compound 24c as a white powder (0.13 g; 47%): 1H NMR (400 MHz, [D6]DMSO): δ=8.15 (s, 1H), 7.44 (m, 2H), 7.33 (d, J=8.5 Hz, 2H), 7.12 (d, J=7.8 Hz, 1H), 5.60 (s, 2H), 5.00 (s, 2H), 3.23 (m, 1H), 1.67 (m, 4H), 1.52 (m, 1H), 1.11 ppm (m, 5H); 13C NMR (100 MHz, [D6]DMSO): δ=154.9, 143.1, 135.0, 132.8, 129.8 (2-C), 128.7 (2-C), 124.6, 56.6, 51.9, 49.4, 32.6 (2-C), 25.1, 24.5 ppm (2-C); UPLC-MS: Method A, Rt 2.55, ionization: m/z 349 [M+H]+; HRMS–ESI: m/z [M+H]+ calcd for C17H21ClN4O2: 349.1431, found: 349.1427.</p><!><p>The reaction was carried out following general procedure (1), using prop-2-ynyl N-cyclohexylcarbamate (0.37 g, 2.04 mmol) and 1-(azidomethyl)-2-methyl-benzene (0.3 g, 2.04 mmol), sodium ascorbate (0.040, 0.2 mmol), and copper (II) sulfate pentahydrate (0.005 g, 0.02 mmol) in water / t-BuOH 1:1 (3 mL). Purification was performed by flash chromatography (SiO2) eluting with a gradient from 0 to 2% MeOH in DCM, to afford compound 25a as a white powder (0.13 g; 47%): 1H NMR (400 MHz, [D6]DMSO): δ=8.03 (s, 1H), 7.29 – 7.12 (m, 4H), 7.07 (d, J=7.5 Hz, 1H), 5.60 (s, 2H), 5.00 (s, 2H), 3.22 (dt, J=10.7, 5.8 Hz, 1H), 2.30 (s, 3H), 1.82 – 1.58 (m, 4H), 1.52 (d, J=12.5 Hz, 1H), 1.31 – 0.95 ppm (m, 5H); 13C NMR (100 MHz, [D6]DMSO): δ=155.4, 143.3, 136.7, 134.5, 130.8, 129.1, 128.8, 126.7, 125.1, 57.2, 51.3, 49.9, 33.0 (2-C), 25.5, 25.0 (2-C), 19.0 ppm; UPLC-MS: Method A, Rt 2.5, ionization: m/z 329 [M+H]+; HRMS–ESI: m/z [M+H]+ calcd for C18H24N4O2: 329.1978, found: 329.1977.</p><!><p>The reaction was carried out following general procedure (1), using prop-2-ynyl N-cyclohexylcarbamate (0.37 g, 2.04 mmol) and 1-(azidomethyl)-3-methyl-benzene (0.3 g, 2.04 mmol), sodium ascorbate (0.040, 0.2 mmol), and copper (II) sulfate pentahydrate (0.005 g, 0.02 mmol) in water / t-BuOH 1:1 (3 mL). Purification was performed by flash chromatography (SiO2) eluting with a gradient from 0 to 2% MeOH in DCM, to afford compound 25b as a white powder (0.17 g; 65%): 1H NMR (400 MHz, [D6]DMSO): δ=8.12 (s, 1H), 7.25 (t, J=7.8 Hz, 1H), 7.13 (m, 4H), 5.54 (s, 2H), 5.00 (s, 2H), 3.24 (m, 1H), 2.28 (s, 3H), 1.68 (dd, J=25.9, 12.5 Hz, 4H), 1.52 (d, J=12.5 Hz, 1H), 1.12 ppm (m, 5H); 13C NMR (100 MHz, [D6]DMSO): δ=154.9, 142.9, 137.9, 135.8, 128.7, 128.6, 128.5, 125.0, 124.5, 56.7, 52.7, 49.4, 32.5 (2-C), 25.1, 24.5 (2-C), 20.8 ppm; UPLC-MS: Method A, Rt 2.52, ionization: m/z 329 [M+H]+; HRMS–ESI: m/z [M+H]+ calcd for C18H24N4O2: 329.1978, found: 329.1981.</p><!><p>The reaction was carried out following general procedure (1), using prop-2-ynyl N-cyclohexylcarbamate (0.37 g, 2.04 mmol) and 1-(azidomethyl)-4-methyl-benzene (0.3 g, 2.04 mmol), sodium ascorbate (0.040, 0.2 mmol), and copper (II) sulfate pentahydrate (0.005 g, 0.02 mmol) in water / t-BuOH 1:1 (3 mL). Purification was performed by flash chromatography (SiO2) eluting with a gradient from 0 to 2% MeOH in DCM, to afford compound 25c as a white powder (0.18 g; 67%): 1H NMR (400 MHz, [D6]DMSO): δ=8.10 (s, 1H), 7.28 – 7.08 (m, 5H), 5.53 (s, 2H), 4.99 (s, 2H), 3.30 – 3.15 (m, 1H), 2.27 (s, 3H), 1.67 (dd, J=24.7, 12.5 Hz, 4H), 1.52 (d, J=12.2 Hz, 1H), 1.15 ppm (dt, J=37.6, 12.2 Hz, 5H); 13C NMR (100 MHz, [D6]DMSO): δ=155.4, 143.4, 137.9, 133.4, 129.7 (2-C), 128.4 (2-C), 124.9, 57.2, 53.0, 49.9, 33.0 (2-C), 25.5, 25.0 (2-C), 21.1 ppm; UPLC-MS: Method A, Rt 2.52, ionization: m/z 329 [M+H]+; HRMS–ESI: m/z [M+H]+ calcd for C18H24N4O2: 329.1978, found: 329.1978.</p><!><p>The reaction was carried out following general procedure (1), using prop-2-ynyl N-cyclohexylcarbamate (0.15 g, 0.82 mmol) and 1-(azidomethyl)-2-methoxy-benzene (0.13 g, 0.82 mmol), sodium ascorbate (0.016, 0.08 mmol), and copper (II) sulfate pentahydrate (0.002 g, 0.008 mmol) in water / t-BuOH 1:1 (3 mL). Purification was performed by flash chromatography (SiO2) eluting with a gradient from 0 to 2% MeOH in DCM, to afford compound 26a as a white powder (0.18 g; 64%): 1H NMR (400 MHz, [D6]DMSO): δ=7.98 (s, 1H), 7.35 (m, 1H), 7.09 (m, 3H), 6.93 (m, 1H), 5.52 (s, 2H), 4.99 (s, 2H), 3.82 (s, 3H), 3.23 (m, 1H), 1.67 (m, 4H), 1.52 (d, J=12.6 Hz, 1H), 1.12 ppm (m, 5H); 13C NMR (100 MHz, [D6]DMSO): δ=156.8, 155.0, 142.7, 130.0, 129.6, 124.6, 123.5, 120.5, 111.2, 56.7, 55.5, 49.4, 48.2, 32.6 (2-C), 25.1, 24.5 ppm (2-C); UPLC-MS: Method A, Rt 2.43, ionization: m/z 345 [M+H]+; HRMS–ESI: m/z [M+H]+ calcd for C18H24N4O3: 345.1927, found: 345.1930.</p><!><p>The reaction was carried out following general procedure (1), using prop-2-ynyl N-cyclohexylcarbamate (0.15 g, 0.82 mmol) and 1-(azidomethyl)-3-methoxy-benzene (0.13 g, 0.82 mmol), sodium ascorbate (0.016, 0.08 mmol), and copper (II) sulfate pentahydrate (0.002 g, 0.008 mmol) in water / t-BuOH 1:1 (3 mL). Purification was performed by flash chromatography (SiO2) eluting with a gradient from 0 to 70% EtOAc in Cy, to afford compound 26b as a white powder (0.14 g; 50%): 1H NMR (400 MHz, [D6]DMSO): δ=8.14 (s, 1H), 7.28 (td, J=7.5, 1.8 Hz, 1H), 7.12 (d, J=7.8 Hz, 1H), 6.90 (d, J=6.2 Hz, 2H), 6.85 (d, J=7.8 Hz, 1H), 5.55 (s, 2H), 5.00 (s, 2H), 3.73 (s, 3H), 3.24 (d, J=7.6 Hz, 1H), 1.68 (dd, J=25.7, 12.5 Hz, 4H), 1.52 (d, J=12.4 Hz, 1H), 1.13 ppm (m, 5H); 13C NMR (100 MHz, [D6]DMSO): δ=159.4, 154.9, 143.0, 137.4, 129.8, 124.6, 119.9, 113.7, 113.4, 56.7, 55.0, 52.6, 49.4, 32.6 (2-C), 25.1, 24.5 ppm (2-C); UPLC-MS: Method A, Rt 2.38, ionization: m/z 345 [M+H]+; HRMS–ESI: m/z [M+H]+ calcd for C18H24N4O3: 345.1927, found: 345.1929.</p><!><p>The reaction was carried out following general procedure (1), using prop-2-ynyl N-cyclohexylcarbamate (0.15 g, 0.82 mmol) and 1-(azidomethyl)-4-methoxy-benzene (0.13 g, 0.82 mmol), sodium ascorbate (0.016, 0.08 mmol), and copper (II) sulfate pentahydrate (0.002 g, 0.008 mmol) in water / t-BuOH 1:1 (3 mL). Purification was performed by flash chromatography (SiO2) eluting with a gradient from 0 to 60% EtOAc in Cy, to afford compound 26c as a white powder (0.15 g; 54%): 1H NMR (400 MHz, [D6]DMSO): δ=8.08 (s, 1H), 7.29 (d, J=8.6 Hz, 2H), 7.11 (d, J=7.8 Hz, 1H), 6.92 (d, J=8.6 Hz, 2H), 5.50 (s, 2H), 4.99 (s, 2H), 3.73 (s, 3H), 3.22 (m, 1H), 1.68 (dd, J=25.3, 12.6 Hz, 4H), 1.52 (d, J=12.6 Hz, 1H), 1.13 ppm (m, 5H); 13C NMR (100 MHz, [D6]DMSO): δ=159.1, 154.9, 142.9, 129.5 (2-C), 127.9, 124.2, 114.0 (2-C), 56.7, 55.1, 52.2, 49.4, 32.6 (2-C), 25.1, 24.5 ppm (2-C); UPLC-MS: Method A, Rt 2.35, ionization: m/z 345 [M+H]+; HRMS–ESI: m/z [M+H]+ calcd for C18H24N4O3: 345.1927, found: 345.1924.</p><!><p>The reaction was carried out following general procedure (1), using prop-2-ynyl N-cyclohexylcarbamate (0.15 g, 0.82 mmol) and 1-(azidomethyl)-3,5-dimethoxy-benzene (0.16 g, 0.82 mmol), sodium ascorbate (0.016, 0.08 mmol), and copper (II) sulfate pentahydrate (0.002 g, 0.008 mmol) in water / t-BuOH 1:1 (3 mL). Purification was performed by flash chromatography (SiO2) eluting with a gradient from 0 to 40% EtOAc in Cy, to afford compound 27 as a white powder (0.17 g; 55%): 1H NMR (400 MHz, [D6]DMSO): δ=8.14 (s, 1H), 7.12 (d, J=7.8 Hz, 1H), 6.46 (s, 3H), 5.50 (s, 2H), 5.01 (s, 2H), 3.72 (s, 6H), 3.24 (m, 1H), 1.67 (m, 4H), 1.52 (d, J=12.4 Hz, 1H), 1.12 ppm (m, 5H); 13C NMR (100 MHz, [D6]DMSO): δ=160.6, 154.9, 143.0, 138.0, 124.6, 106.0, 99.5, 56.7, 55.2, 52.7, 49.4, 32.6 (2-C), 25.1, 24.5 ppm (2-C); UPLC-MS: Method A, Rt 2.42, ionization: m/z 375 [M+H]+; HRMS–ESI: m/z [M+H]+ calcd for C19H26N4O4: 375.2032, found: 375.2047.</p><!><p>The reaction was carried out following general procedure (1), using prop-2-ynyl N-cyclohexylcarbamate (0.15 g, 0.82 mmol) and 2-(azidomethyl)-1,3-difluoro-benzene (0.14 g, 0.82 mmol), sodium ascorbate (0.016, 0.08 mmol), and copper (II) sulfate pentahydrate (0.002 g, 0.008 mmol) in water / t-BuOH 1:1 (3 mL). Purification was performed by flash chromatography (SiO2) eluting with a gradient from 0 to 40% EtOAc in Cy, to afford compound 28 as a white powder (0.16 g; 56%): 1H NMR (400 MHz, [D6]DMSO): δ=8.10 (s, 1H), 7.51 (m, 1H), 7.18 (t, J=8.1 Hz, 2H), 7.13 (d, J=7.8 Hz, 1H), 5.66 (s, 2H), 4.99 (s, 2H), 3.23 (m, 1H), 1.68 (dd, J=26.0, 12.6 Hz, 4H), 1.52 (d, J=12.6 Hz, 1H), 1.12 ppm (m, 5H); 13C NMR (100 MHz, [D6]DMSO): δ=161.2 (d, J=249.2 Hz), 155.4, 143.3, 132.1 (t, J=10.4 Hz), 125.1, 112.3 (d, J=24.4 Hz), 111.7 (t, J=19.3 Hz), 57.1, 49.9, 41.2 (t, J=3.7 Hz), 33.0 (2-C), 25.6, 25.0 ppm (2-C); UPLC-MS: Method A, Rt 2.38, ionization: m/z 351 [M+H]+; HRMS–ESI: m/z [M+H]+ calcd for C17H20F2N4O2: 351.1633, found: 351.1631.</p><!><p>The reaction was carried out following general procedure (1), using prop-2-ynyl N-cyclohexylcarbamate (0.15 g, 0.82 mmol) and 1-(azidomethyl)-3,5-difluoro-benzene (0.14 g, 0.82 mmol), sodium ascorbate (0.016, 0.08 mmol), and copper (II) sulfate pentahydrate (0.002 g, 0.008 mmol) in water / t-BuOH 1:1 (3 mL). Purification was performed by flash chromatography (SiO2) eluting with a gradient from 0 to 2% MeOH in DCM, to afford compound 29 as a white powder (0.22 g; 77%): 1H NMR (400 MHz, [D6]DMSO): δ=8.21 (s, 1H), 7.23 (tt, J=9.4, 2.3 Hz, 1H), 7.13 (d, J=7.8 Hz, 1H), 7.04 (t, J= 6.4 Hz, 2H), 5.64 (s, 2H), 5.02 (s, 2H), 3.23 (m, 1H), 1.68 (dd, J=26.4, 12.4 Hz, 4H), 1.52 (d, J=12.6 Hz, 1H), 1.13 ppm (m, 5H); 13C NMR (100 MHz, [D6]DMSO): δ=162.8 (dd, J=247.1, 13.2 Hz), 155.4, 143.7, 140.6 (t, J=9.4 Hz), 125.4, 111.7 (m), 104.1 (t, J=25.7 Hz), 57.1, 52.1, 49.9, 33.0 (2-C), 25.5, 25.0 ppm (2-C); UPLC-MS: Method A, Rt 2.46, ionization: m/z 351 [M+H]+; HRMS–ESI: m/z [M+H]+ calcd for C17H20F2N4O2: 351.1633, found: 351.1634.</p><!><p>The reaction was carried out following general procedure (1), using prop-2-ynyl N-cyclohexylcarbamate (0.15 g, 0.82 mmol) and 1-(azidomethyl)-2-fluoro-3-methoxy-benzene (0.15 g, 0.82 mmol), sodium ascorbate (0.016, 0.08 mmol), and copper (II) sulfate pentahydrate (0.002 g, 0.008 mmol) in water / t-BuOH 1:1 (3 mL). Purification was performed by flash chromatography (SiO2) eluting with a gradient from 0 to 2% MeOH in DCM, to afford compound 30 as a white powder (0.23 g; 77%): 1H NMR (400 MHz, [D6]DMSO): δ=8.09 (s, 1H), 7.14 (m, 3H), 6.84 (m, 1H), 5.64 (s, 2H), 5.00 (s, 2H), 3.83 (s, 3H), 3.24 (m, 1H), 1.68 (dd, J=25.9, 12.5 Hz, 4H), 1.52 (d, J=12.5 Hz, 1H), 1.12 ppm (m, 5H); 13C NMR (100 MHz, [D6]DMSO): δ=154.9, 149.5 (d, J=246.8 Hz), 147.3 (d, J=10.0 Hz), 142.9, 124.7, 124.6 (d, J=4.6 Hz), 123.5 (d, J=11.9 Hz), 121.1 (d, J=2.0 Hz), 114.0 (d, J=1.3 Hz), 56.7, 56.1, 49.4, 46.7 (d, J=4.6 Hz), 32.6 (2-C), 25.1, 24.5 ppm (2-C); UPLC-MS: Method A, Rt 2.37, ionization: m/z 363 [M+H]+; HRMS–ESI: m/z [M+H]+ calcd for C18H23FN4O3: 363.1832, found: 363.1834.</p><!><p>2-(1-phenyltriazol-4-yl)ethanol (31a, 0.24 g, 1.26 mmol) was dissolved in dry CH3CN (5 mL) under stirring. Then, DMAP (0.15 g, 1.26 mmol) and cyclohexyl isocyanate (0.17 g, 1.38 mmol) were added and the reaction mixture was stirred overnight at 80 °C. The mixture was then diluted with EtOAc and washed once with 2N HCl, and once with brine. The organic layer was dried over sodium sulfate and concentrated in vacuo. The crude product was purified by by flash chromatography (SiO2) eluting with a gradient from 0 to 2% MeOH in DCM. Purification was performed by flash chromatography (SiO2) eluting with a gradient from 0 to 2% MeOH in DCM, to afford compound 31a as a white powder (0.25 g; 63%): 1H NMR (400 MHz, [D6]DMSO): δ=8.62 (s, 1H), 7.87 (m, 2H), 7.60 (m, 2H), 7.49 (m, 1H), 7.06 (d, J=8.0 Hz, 1H), 4.25 (t, J=6.7 Hz, 2H), 3.24 (m, 1H), 3.02 (t, J=6.7 Hz, 2H), 1.69 (m, 4H), 1.53 (d, J=13.3 Hz, 1H), 1.15 ppm (m, 5H); 13C NMR (100 MHz, [D6]DMSO): δ=155.7, 145.1, 137.2, 130.3 (2-C), 128.9, 121.4, 120.3 (2-C), 62.7, 49.8, 33.1 (2-C), 26.0, 25.6, 25.0 ppm (2-C); UPLC-MS: Method A, Rt 2.39, ionization: m/z 315 [M+H]+; HRMS–ESI: m/z [M+H]+ calcd for C17H22N4O2: 315.1821, found: 315.1829.</p><!><p>It was synthesized according to the procedure employed for 32, starting from 2-(1-benzyltriazol-4-yl)ethanol (31b, 0.23 g, 1.12 mmol), cyclohexyl isocyanate (0.15 g, 1.23 mmol), and DMAP (0.14 g, 1.12 mmol) in dry CH3CN (5 mL). Purification was performed by flash chromatography (SiO2) eluting with a gradient from 0 to 2% MeOH in DCM, to afford compound 33 as a white powder (0.2 g; 54%): 1H NMR (400 MHz, [D6]DMSO): δ=7.95 (s, 1H), 7.33 (m, 5H), 7.01 (d, J=7.96 Hz, 1H), 5.55 (s, 2H), 4.16 (t, J=6.83 Hz, 2H), 3.21 (m, 1H), 2.90 (t, J=6.82 Hz, 2H), 1.67 (m, 4H), 1.52 (m, 1H), 1.12 ppm (m, 5H); 13C NMR (100 MHz, [D6]DMSO): δ=155.0, 144.1, 136.6, 129.1 (2-C), 128.5, 128.3 (2-C), 123.1, 99.9, 62.8, 53.1, 49.8, 33.1 (2-C), 26.0, 25.6, 25.0 ppm (2-C); UPLC-MS: Method A, Rt 2.33, ionization: m/z 329 [M+H]+; HRMS–ESI: m/z [M+H]+ calcd for C18H24N4O2: 329.1978, found: 329.1983.</p><!><p>The reaction was carried out following general procedure (2), using prop-2-ynyl N-prop-2-ynylcarbamate (0.11 g, 0.767 mmol) and 1-(azidomethyl)-2-methoxy-benzene (0.25 g, 1.53 mmol), sodium ascorbate (0.030, 0.15 mmol), and copper (II) sulfate pentahydrate (0.004 g, 0.002 mmol) in water / t-BuOH 1:1 (3 mL). Purification was performed by flash chromatography (SiO2) eluting with a gradient from 0 to 4% MeOH in DCM, to afford compound 34a as a white amorphous solid (0.2 g; 58%): 1H NMR (400 MHz, [D6]DMSO): δ=8.00 (s, 1H), 7.80 (s, 1H), 7.68 (t, J=5.7 Hz, 1H), 7.34 (m, 2H), 7.08 (m, 4H), 6.93 (m, 2H), 5.51 (s, 2H), 5.48 (s, 2H), 5.03 (s, 2H), 4.21 (d, J=5.8 Hz, 2H), 3.81 (s, 3H), 3.80 ppm (s, 3H); 13C NMR (100 MHz, [D6]DMSO): δ=156.83, 156.8, 155.9, 145.0, 142.5, 130.0, 129.9, 129.6, 129.5, 124.7, 123.6, 123.4, 122.8, 120.5 (2C), 111.2, 111.1, 57.1, 55.5 (2C), 48.2, 48.1, 35.9 ppm; UPLC-MS: Method A, Rt 2.18, ionization: m/z 464 [M+H]+; HRMS–ESI: m/z [M+H]+ calcd for C23H25N7O4: 464.2046, found: 464.2056.</p><!><p>The reaction was carried out following general procedure (2), using prop-2-ynyl N-prop-2-ynylcarbamate (0.11 g, 0.767 mmol) and 1-(azidomethyl)-3-methoxy-benzene (0.25 g, 1.53 mmol), sodium ascorbate (0.030, 0.15 mmol), and copper (II) sulfate pentahydrate (0.004 g, 0.002 mmol) in water / t-BuOH 1:1 (3 mL). Purification was performed by flash chromatography (SiO2) eluting with a gradient from 0 to 5% MeOH in DCM, to afford compound 34b as a white amorphous solid (0.29 g; 81%): 1H NMR (400 MHz, [D6]DMSO): δ=8.15 (s, 1H), 7.95 (s, 1H), 7.71 (t, J=5.7 Hz, 1H), 7.28 (td, J=8.2, 7.5, 2.1 Hz, 2H), 6.89 (dd, J=5.8, 2.9 Hz, 4H), 6.85 (d, J=7.3 Hz, 2H), 5.55 (s, 2H), 5.51 (s, 2H), 5.04 (s, 2H), 4.22 (d, J=5.8 Hz, 2H), 3.73 ppm (s, 6H); 13C NMR (100 MHz, [D6]DMSO): δ=159.4 (2C), 155.9, 145.3, 142.8, 137.5, 137.4, 129.9, 129.8, 124.6, 122.8, 120.0 (2C), 113.7, 113.7, 113.4, 113.4, 57.1, 55.0 (2C), 52.7, 52.6, 36.0 ppm; UPLC-MS: Method A, Rt 2.12, ionization: m/z 464 [M+H]+; HRMS–ESI: m/z [M+H]+ calcd for C23H25N7O4: 464.2046, found: 464.2052.</p><!><p>The reaction was carried out following general procedure (1), using prop-2-ynyl N-[[1-[(3-methoxyphenyl)methyl]triazol-4-yl]methyl]carbamate (36b, 0.15 g, 0.5 mmol) and 1-(azidomethyl)-2-methoxy-benzene (0.08 g, 0.5 mmol), sodium ascorbate (0.010, 0.05 mmol), and copper (II) sulfate pentahydrate (0.001 g, 0.005 mmol) in water / t-BuOH 1:1 (3 mL). Purification was performed by flash chromatography (SiO2) eluting with a gradient from 0 to 2% MeOH in DCM, to afford compound 37 as a white amorphous solid (0.13 g; 58%): 1H NMR (400 MHz, [D6]DMSO): δ=8.00 (s, 1H), 7.95 (s, 1H), 7.70 (t, J=5.7 Hz, 1H), 7.34 (m, 1H), 7.27 (m, 1H), 7.11 (dd, J=7.5, 1.4 Hz, 1H), 7.05 (d, J=8.2 Hz, 1H), 6.92 (m, 3H), 6.85 (d, J=7.6 Hz, 1H), 5.51 (s, 4H), 5.03 (s, 2H), 4.21 (d, J=5.8 Hz, 2H), 3.80 (s, 3H), 3.73 ppm (s, 3H); 13C NMR (100 MHz, [D6]DMSO): δ=159.8, 157.3, 156.4, 145.8, 142.9, 137.9, 130.4, 130.3, 130.1, 125.2, 123.9, 123.3, 120.9, 120.4, 114.2, 113.8, 111.6, 57.6, 56.0, 55.5, 53.1, 48.7, 36.4 ppm; UPLC-MS: Method A, Rt 2.15, ionization: m/z 464 [M+H]+; HRMS–ESI: m/z [M+H]+ calcd for C23H25N7O4: 464.2046, found: 464.2062.</p><!><p>The reaction was carried out following general procedure (1), using prop-2-ynyl N-[[1-[(4-methoxyphenyl)methyl]triazol-4-yl]methyl]carbamate (36c, 0.15 g, 0.5 mmol) with 1-(azidomethyl)-2-methoxy-benzene (0.08 g, 0.5 mmol), sodium ascorbate (0.010, 0.05 mmol), and copper (II) sulfate pentahydrate (0.001 g, 0.005 mmol) in water / t-BuOH 1:1 (3 mL). Purification was performed by flash chromatography (SiO2) eluting with a gradient from 0 to 2% MeOH in DCM, to afford compound 38 as a colourless amorphous solid (0.15 g; 63%): 1H NMR (400 MHz, [D6]DMSO): δ=8.00 (s, 1H), 7.89 (s, 1H), 7.68 (t, J=5.7 Hz, 1H), 7.34 (m, 1H), 7.28 (d, J=8.5 Hz, 2H), 7.11 (dd, J=7.4, 1.3 Hz, 1H), 7.05 (d, J=8.2 Hz, 1H), 6.92 (m, 3H), 5.51 (s, 2H), 5.46 (s, 2H), 5.03 (s, 2H), 4.20 (d, J=5.8 Hz, 2H), 3.80 (s, 3H), 3.73 ppm (s, 3H); 13C NMR (100 MHz, [D6]DMSO): δ=159.5, 157.3, 156.3, 145.8, 142.9, 130.4, 130.1, 130.0 (2C), 128.4, 125.2, 123.9, 122.9, 120.9, 114.5 (2C), 111.7, 57.6, 56.0, 55.5, 52.7, 48.7, 36.4 ppm; UPLC-MS: Method A, Rt 2.13, ionization: m/z 464 [M+H]+; HRMS–ESI: m/z [M+H]+ calcd for C23H25N7O4: 464.2046, found: 464.2052.</p><!><p>The reaction was carried out following general procedure (1), using prop-2-ynyl N-[[1-[(2-methoxyphenyl)methyl]triazol-4-yl]methyl]carbamate (36a, 0.15 g, 0.5 mmol) and 1-(azidomethyl)-3-methoxy-benzene (0.08 g, 0.5 mmol), sodium ascorbate (0.010, 0.05 mmol), and copper (II) sulfate pentahydrate (0.001 g, 0.005 mmol) in water / t-BuOH 1:1 (3 mL). Purification was performed by flash chromatography (SiO2) eluting with a gradient from 0 to 2% MeOH in DCM, to afford compound 39 as a colourless amorphous solid (0.15 g; 66%): 1H NMR (400 MHz, [D6]DMSO): δ=8.15 (s, 1H), 7.80 (s, 1H), 7.69 (t, J=5.7 Hz, 1H), 7.34 (t, J=7.8 Hz, 1H), 7.28 (m, 1H), 7.06 (dd, J=14.4, 7.8 Hz, 2H), 6.89 (m, 4H), 5.54 (s, 2H), 5.48 (s, 2H), 5.04 (s, 2H), 4.21 (d, J=5.8 Hz, 2H), 3.80 (s, 3H), 3.73 ppm (s, 3H); 13C NMR (100 MHz, [D6]DMSO): δ=159.4, 156.8, 155.9, 145.0, 142.8, 137.4, 129.9 (2-C), 129.5, 124.6, 123.6, 122.8, 120.5, 120.0, 113.7, 113.4, 111.1, 57.1, 55.5, 55.0, 52.7, 48.1, 35.9 ppm; UPLC-MS: Method A, Rt 2.15, ionization: m/z 464 [M+H]+; HRMS–ESI: m/z [M+H]+ calcd for C23H25N7O4: 464.2046, found: 464.2044.</p><!><p>The reaction was carried out following general procedure (1), using prop-2-ynyl N-[[1-[(4-methoxyphenyl)methyl]triazol-4-yl]methyl]carbamate (36c, 0.15 g, 0.5 mmol) with 1-(azidomethyl)-3-methoxy-benzene (0.08 g, 0.5 mmol), sodium ascorbate (0.010, 0.05 mmol), and copper (II) sulfate pentahydrate (0.001 g, 0.005 mmol) in water / t-BuOH 1:1 (3 mL). Purification was performed by flash chromatography (SiO2) eluting with a gradient from 0 to 2% MeOH in DCM, to afford compound 40 as a colourless amorphous solid (0.16 g; 68%): 1H NMR (400 MHz, [D6]DMSO): δ=8.15 (s, 1H), 7.89 (s, 1H), 7.69 (t, J=5.7 Hz, 1H), 7.28 (m, 3H), 6.90 (m, 4H), 6.85 (d, J=7.7 Hz, 1H), 5.55 (s, 2H), 5.46 (s, 2H), 5.04 (s, 2H), 4.20 (d, J=5.8 Hz, 2H), 3.73 ppm (s, 6H); 13C NMR (100 MHz, [D6]DMSO): δ=159.4, 159.1, 155.9, 145.3, 142.8, 137.4, 129.9, 129.6 (2C), 127.9, 124.6, 122.4, 120.0, 114.0 (2C), 113.7, 113.4, 57.1, 55.1, 55.0, 52.6, 52.2, 35.9 ppm; UPLC-MS: Method A, Rt 2.09, ionization: m/z 464 [M+H]+; HRMS–ESI: m/z [M+H]+ calcd for C23H25N7O4: 464.2046, found: 464.2052.</p><!><p>It was synthesized according to the procedure employed for 32, starting from (1-phenyl-1,2,4-triazol-3-yl)methanol (42a, 0.05 g, 0.28 mmol), cyclohexyl isocyanate (0.04 g, 0.30 mmol), and DMAP (0.03 g, 0.28 mmol) in dry CH3CN (5 mL). Purification was performed by flash chromatography (SiO2) eluting with a gradient from 0 to 100% EtOAc in Cy, to afford compound 43 as a white powder (0.05 g; 58%): 1H NMR (400 MHz, [D6]DMSO): δ=9.26 (s, 1H), 7.89 – 7.77 (m, 2H), 7.62 – 7.49 (m, 2H), 7.48 – 7.37 (m, 1H), 7.26 (d, J=7.8 Hz, 1H), 5.08 (s, 2H), 3.26 (ddd, J=10.5, 7.2, 3.2 Hz, 1H), 1.86 – 1.60 (m, 4H), 1.53 (d, J=12.8 Hz, 1H), 1.30 – 1.01 ppm (m, 5H); 13C NMR (100 MHz, [D6]DMSO): δ=161.0, 155.2, 143.5, 137.0, 130.2, 128.3, 119.7, 58.8, 50.0, 33.1 (2C), 25.6, 25.0 ppm (2C); UPLC-MS: Method A, Rt 2.34, ionization: m/z 301 [M+H]+; HRMS–ESI: m/z [M+H]+ calcd for C16H20N4O2: 301.1665, found: 301.1674.</p><!><p>It was synthesized according to the procedure employed for 32, starting from (1-benzyl-1,2,4-triazol-3-yl)methanol (42b, 0.06 g, 0.32 mmol), cyclohexyl isocyanate (0.044 g, 0.35 mmol), and DMAP (0.04 g, 0.32 mmol) in dry CH3CN (5 mL). Purification was performed by flash chromatography (SiO2) eluting with a gradient from 0 to 100% EtOAc in Cy, to afford compound 46 as a white powder (0.07 g; 68%): 1H NMR (400 MHz, [D6]DMSO): δ=8.62 (s, 1H), 7.42 – 7.25 (m, 5H), 7.17 (d, J=7.7 Hz, 1H), 5.37 (s, 2H), 4.93 (s, 2H), 3.27 – 3.15 (m, 1H), 1.68 (dd, J=28.7, 12.5 Hz, 4H), 1.52 (d, J=12.6 Hz, 1H), 1.13 ppm (tdd, J=32.8, 24.1, 12.1 Hz, 5H); 13C NMR (100 MHz, [D6]DMSO): δ=159.7, 154.7, 145.0, 136.1, 128.6, 127.9, 127.9, 58.4, 52.1, 49.4, 32.5 (2C), 25.1, 24.5 ppm (2C); UPLC-MS: Method A, Rt 2.27, ionization: m/z 315 [M+H]+; HRMS–ESI: m/z [M+H]+ calcd for C17H22N4O2: 315.1821, found: 315.1823.</p><!><p>Rat FAAH was prepared from male Sprague Dawley rat brains, homogenized in a potter in 20 mM of Tris HCl pH 7.4, 0.32 M sucrose. The radiometric assay used to measure FAAH activity was performed in Eppendorf tubes: 50 µg of total rat brain homogenate were pre-incubated in 445.5 µL of assay buffer (50 mM Tris-HCl pH 7.4, 0.05% Fatty acid-free -bovine serum albumin (BSA)) with 4.5 µL of inhibitor (at appropriate concentration in DMSO) or DMSO alone (to measure FAAH total activity) for 10 min at 37 °C. The blank (no activity control) was prepared using 445.5 µL of assay buffer and 4.5 µL of DMSO without the 50 µg of total rat brain homogenate. After 10 min of pre-incubation with test compounds, the reaction was started by adding 50 µL of substrate and incubating for 30 min at 37 °C. The substrate was prepared in assay buffer in order to achieve the final concentration of 1 µM arachidonoyl ethanolamide (Cayman Chemical N. 90050) and 0.6nM anandamide [ethanolamine-1-3H] (American Radiolabeled Chemicals Inc., ART. 0626, conc. 1 mCi/mL, S.A. 60 Ci/mmol). The reaction was stopped by adding cold 1:1 CHCl3/methanol. After 10 min of centrifugation (845xg at 4 °C) 600 µL of aqueous phase were transferred into scintillation vials previously filled with 3mL of scintillation fluid (Ultima GoldTM, Perkin Elmer Inc., Cat. 6013329). Radioactivity was measured by liquid scintillation counting (MicroBeta2 LumiJET Perkin Elmer Inc.).</p><!><p>Human recombinant FAAH was obtained from a HEK-293 cell line stably overexpressing human FAAH-1 enzyme. Cells were grown in Dulbecco's Modified Eagle Medium (DMEM) containing 10% FBS, 1% pen/strep, 1% glutamine and 500 µg /mL G418. To obtain membrane preparation cells were scraped off with cold PBS and collected by centrifugation (500xg, 10 minutes, 4°C); the cell pellet was re-suspended in 20mM Tris-HCl pH 7.4, 0.32M sucrose, disrupted by sonication (10 pulses, 5 times) and centrifuged (800xg, 15 minutes, 4°C); the collected supernatant was centrifuged at 105,000xg for 1h at 4°C and the pellet was re-suspended in PBS. The fluorescent assay to measure FAAH activity was performed in 96 wells black plates: 2.5 µg of human FAAH-1 membrane preparation were pre-incubated for 50 min at 37 °C, in 180 µL of assay buffer (50mM Tris-HCl pH 7.4, 0.05% Fatty acid-free BSA) with 10 µL of inhibitor or 10 µL DMSO to measure FAAH total activity. The background (no activity) samples were prepared using 180 µL of assay buffer without human FAAH-1 and 10 µL of DMSO. The reaction was then started by addition of 10 µL of the substrate (7-amino-4-methyl coumarin-arachidonamideN. 10005098, Cayman Chemical) dissolved in ethanol and used at a final concentration of 2 µM. The reaction was carried out for 30 min at 37 °C and fluorescence was measured with a Tecan Infinite M200 nanoquant plate reader (excitation wavelength 350 nm / emission wavelength 460nm). IC50 values (concentrations causing half-maximal inhibition) were determined by non-linear regression analysis of the Log [concentration]/response curves generated with mean replicate values using a four parameter Hill equation curve fitting with GraphPad Prism 5 (GraphPad Software Inc., CA – USA).</p><!><p>CD1 male mice from Charles River Italia were treated intraperitoneally (i.p.) with the test compound (3 mg/kg) or vehicle (1:1:8, PEG400, Tween® 80 and Saline 0.9%). One hour after treatment, the animals were killed by decapitation and the brain and liver were collected. Samples were homogenized in 1.5 mL of 20 mM Tris-HCl buffer pH 7.4, containing 0.32 M sucrose and the homogenates were centrifuged at 1000xg for 10 min (4°C). The supernatants were collected and the protein concentration was measured by Bradford method (Bio Rad Protein Assay Kit). FAAH activity was measured using 50 µg of total brain or liver homogenate in 450 µL of assay buffer (50 mM Tris-HCl pH 7.4, 0.05% Fatty acid-free BSA); the blank (no activity sample) was prepared with 450 µL of assay buffer. The reaction was started by adding 50 µL of substrate for 30 min at 37 °C. The substrate was prepared in assay buffer in order to obtain a final concentration of 1 µM arachidonoyl-ethanolamide (N.90050, Cayman Chemical) and 0.6 nM anandamide [ethanolamine-1-3H] (American Radiolabeled Chemicals Inc., ART.0626, 1 mCi/mL, specific activity 60 Ci/mmoL). The reaction was stopped by adding cold 1:1 chloroform/methanol. After 10 min centrifugation (845xg at 4°C) 600 µL of the aqueous phase was transferred into scintillation vials previously filled with 3 mL of scintillation fluid (ULTIMA GOLD, Cat.6013329, Perkin Elmer). Radioactivity was measured by liquid scintillation counting (Microbeta2 Lumijet, Perkin Elmer Inc.).</p><!><p>10 ng of the purified rat MGL was pre-incubated with appropriate drug for 10 min at 37 °C in 50 mM Tris-HCl, pH 8.0 containing 0.5 mg/mL fatty acid-free bovine serum albumin (BSA, Sigma-Aldrich). The final concentration of vehicle (1% DMSO) had no effect on MGL activity. Then, 2-oleoylglycerol (2-OG, 10 μM final) was added to the mixture and incubated for additional 30 min at 37 °C. Reactions were stopped by adding chloroform:methanol (2:1, vol:vol), containing heptadecanoic acid (5 nmol/sample) as an internal standard. After centrifugation at 2,000 x g at 4 °C for 10 min, the organic layers were collected and dried under a stream of N2. The lipid extracts were then suspended in chloroform:methanol (1:3, vol:vol) and analyzed by a liquid chromatography/mass spectrometry (LC/MS) method.</p><!><p>Compounds were diluted in rat plasma added with 10 % DMSO to help solubilization. Plasma was already pre-heated at 37° C (30 min). The final compound concentration was 1.0 μM. At time points (immediately after dilution, 30, 60, 120, 240, 360 and 420 min) a 40 µL aliquot of the incubation solution was diluted in 150 µL of cold CH3CN spiked with 200 nM warfarin as internal standard. After vortexing for 30 s, the solution was centrifuged at 3500 g for 15 min at 4 °C and the supernatant transferred for LC-MS analysis on a Waters ACQUITY UPLC/MS TQD system consisting of a TQD (Triple Quadrupole Detector) Mass Spectrometer equipped with an Electrospray Ionization interface. Briefly, 3.0 µL of the supernatant were injected on a reversed phase column (BEH C18 1.7 µm 2.1X50 mm) and separated with a linear acetonitrile gradient. Compounds were quantified on the basis of their MRM (Multiple Reaction Monitoring) peak areas. The response factors, calculated on the basis of the internal standard peak area, were then plotted over time. For each compound, analyses were conducted in triplicate: compound remaining (%) with corresponding standard deviation at 420 minutes is reported.</p><!><p>Compounds were preincubated with microsomes in 100mM TRIS-HCl pH 7.4 for 15 minutes. At time zero, cofactors were added. The final incubation conditions for each sample were: 1.25mg/mL liver microsomes, 5μM compound (final DMSO 0.1%), NADP 1mM, G6P 20mM, MgCl2 2mM, G6P dehydrogenase 2 Units. The mixture was kept at 37°C under shaking. Aliquots (30μL) were taken at various time points (typically 0, 5, 15, 30, and 60 minutes) and crashed with 200μL of acetonitrile spiked with 200nM warfarin (internal standard). A reference incubation, with microsomes but without cofactors, was kept at 37°C and sampled at the end of the time course. After vortexing and centrifugation, 3μL of superrnatant were analyzed by LC-MS/MS by multiple reaction monitoring (MRM).</p><!><p>The kinetic solubility in Phosphate Buffered Saline (PBS) at pH 7.4 was determined starting from a 10 mM DMSO solution of the test compounds. The study was performed by incubation of an aliquot of 10 mM DMSO solution in PBS (pH 7.4) at 25°C for 24h, under shaking, followed by centrifugation and quantification of dissolved compound in the supernatant by UPLC/MS. The compound target concentration in the solutionwas 250 µM, resulting in a final DMSO concentration of 2.5%. The supernatant was analyzed by UPLC/MS and the quantification of the dissolved compound was determined by monitoring the UV trace at 215nm. The kinetic solubility (µM) was calculated by dividing the peak area of the test compound in the supernatant by the peak area of a reference solution (250 μM) of the test compound in 1:1 CH3CN-H2O, and further multiplied by the concentration of the test compound reference and the dilution factor. The UPLC/MS analyses were performed on a Waters ACQUITY UPLC/MS system consisting of a SQD (Single Quadrupole Detector) Mass Spectrometer equipped with an Electrospray Ionization interface and a Photodiode Array Detector. The PDA range was 210-400 nm. The analyses were run on an ACQUITY UPLC BEH C18 column (50x2.1 mmID, particle size 1.7 µm) with a VanGuard BEH C18 pre-column (5x2.1 mmID, particle size 1.7 µm). The mobile phase was 10 mM NH4OAc in H2O at pH 5 adjusted with AcOH (A) and 10 mM NH4OAc in CH3CN-H2O (95:5) at pH 5 (B). Electrospray ionization in positive and negative mode was applied in the mass scan range 100-500Da.</p>

PubMed Author Manuscript

Inhibition of African swine fever virus protease by myricetin and myricitrin

AbstractAfrican swine fever (ASF) caused by the ASF virus (ASFV) is the most hazardous swine disease. Since a huge number of pigs have been slaughtered to avoid a pandemic spread, intense studies on the disease should be followed quickly. Recent studies reported that flavonoids have various antiviral activity including ASFV. In this report, ASFV protease was selected as an antiviral target protein to cope with ASF. With a FRET (Fluorescence resonance energy transfer) method, ASFV protease was assayed with a flavonoid library which was composed of sixty-five derivatives classified based on ten different scaffolds. Of these, the flavonols scaffold contains a potential anti-ASFV protease activity. The most prominent flavonol was myricetin with IC50 of 8.4 μM. Its derivative, myricitrin, with the rhamnoside moiety was also showed the profound inhibitory effect on ASFV protease. These two flavonols apparently provide a way to develop anti-ASFV agents based on their scaffold.

inhibition_of_african_swine_fever_virus_protease_by_myricetin_and_myricitrin

Introduction<!>Protein expression and purification<!>FRET protease assays with ASFV protease<!>FRET protease assays with ASFV protease in the presence of triton X-100<!>Absorption spectroscopic studies based on tryptophans of ASFV protease<!>Results<!><!>Results<!><!>Discussion<!><!>Discussion<!>Conclusion<!><!>Disclosure statement

<p>African swine fever (ASF) is the most dangerous swine disease because of its large sanitary and socioeconomic impact. ASF generally has symptoms of high fever, inactivity, bleeding of the skin and internal organs and can result in death within 2–10 days, with mortality rates rising to 100%1. The ASF began in 1921, when it was first recorded in the Montgomery region of Kenya, Africa. Then, ASF has spread across the continent to not only Europe but also to Russia and more recently to Asia. Nowadays, ASF is pandemic in China and Hong-Kong2–5. It is threating to South Korea, too.</p><p>ASF is caused by the ASF virus (ASFV), a large, enveloped virus containing a 170–193 kbp double-stranded DNA encoding more than 150 genes. It is the only member of Asfarviridae family6. ASFV employs a polyprotein processing at Gly-Gly-Xaa sites to produce several core components of viral particles. The virus gene S273R encodes 31 kDa protein containing a "core areas" with the characteristics of SUMO-1 specific protease and adenovirus protease. The S273R product (ASFV protease) is known as the protease involved in the processing of the ASFV polyproteins pp220 and pp62. Therefore, ASFV protease is a good drug target for anti-ASFV infection7.</p><p>Recently, P1192R of ASFV has been proven to code for a functional type II DNA topoisomerase (ASFV-Topo II)8. The study of ASFV-Topo II with enrofloxacin suggested its key role both at intermediate and late stages of viral infection9. Since ASFV-Topo II plays an integral role in virus genome cloning and in the transcription process, this enzyme has been a target for the virus control. Topoisomerase poisons and inhibitors such as coumermycin A1, doxorubicin and amsacrine displayed a higher sensitivity against ASFV-Topo II10. A potent anti-ASFV effect of six fluoroquinolones also has been reported11.</p><p>In this study, we employed a proteolytic method to probe inhibitory compounds against ASFV protease. A synthetic peptide labelled with an EDANS-DABCYL FRET (Fluorescence resonance energy transfer) pair was used to search ASFV protease inhibitory compounds against a flavonoid library. Since the FRET pair was connected by a peptide including the ASFV protease recognition site, an increased intensity of fluorescence could be a sign to judge the presence of the catalytic activity of ASFV protease in this design.</p><p>With the FRET pair, a flavonoid library was screened to search ASFV protease inhibitory compounds. Recent studies showed that flavonoids have antiviral activity in some viruses including ASFV12–14. However, none of the antiviral studies with specific target protease at the molecular level has been reported. In this study, ASFV protease was selected as an antiviral target protein and assayed with a flavonoid library to find a potent inhibitory compound. The effects of flavonoids according to their scaffolds were analysed and the most promising flavonoid was suggested to be developed as a potent antiviral agent.</p><!><p>The coding sequence of pS273R, african swine fever virus (NCBI Ref. seq. NP_042804.1) was synthesised chemically by Bioneer (Daejeon, Korea) and cloned into a bacteriophage T7-based expression vector. The plasmid DNA was transformed into E. coli BL21 (DE3) for protein expression. E. coli BL21 (DE3) cells were grown on Luria–Bertani (LB) agar plates containing 150 μg mL−1 ampicillin. Several colonies were picked and grown in capped test tubes with 10 mL LB broth containing 150 μg mL−1 ampicillin. A cell stock composed of 0.85 mL culture and 0.15 mL glycerol was prepared and frozen at 193 K for use in a large culture. The frozen cell stock was grown in 5 mL LB medium and diluted into 1000 mL fresh LB medium. The culture was incubated at 310 K with shaking until an OD600 of 0.6–0.8 was reached. At this point, the expression of ASFV protease was induced using isopropyl-β-d-1-thiogalactopyranoside (IPTG) at a final concentration of 1 mM. The culture was further grown at 310 K for 3 h in a shaking incubator. Cells were harvested by centrifugation at 7650 g (6500 rev min−1) for 10 min in a high-speed refrigerated centrifuge at 277 K. The cultured cell paste was resuspended in 25 mL of a buffer consisting of 50 mM Tris–HCl pH 8.0, 100 mM NaCl, 10 mM imidazole, 1 mM phenylmethylsulfonyl fluoride (PMSF), 10 μg mL−1 DNase I. The cell suspension was disrupted using an ultrasonic cell disruptor (Digital Sonifier 450, Branson, USA). Cell debris was pelleted by centrifugation at 24 900 g (15 000 rev min-1) for 30 min in a high-speed refrigerated ultra-centrifuge at 277 K.</p><p>The protein was purified by affinity chromatography using a 5 mL Hi-Trap His column (GE Healthcare, Piscataway, New Jersey, USA). The column was equilibrated with a buffer consisting of 50 mM Tris–HCl pH 8.0, 300 mM NaCl and 10 mM imidazole. The target protein was eluted with a buffer consisting of 50 mM Tris–HCl pH 8.0 and 100 mM NaCl with a gradient from 10 to 500 mM imidazole. The purified protein was buffer exchanged into 20 mM Tris pH 7.5 using Vivaspin 20 MWCO 10 kDa (GE Healthcare), a centrifugal device. SDS–PAGE showed one band around 32 kDa (31550.22 Da), corresponding to the molecular weight of ASFV protease. The protein was concentrated to 1.38 mg mL−1 for the protease assay in a buffer consisting of 20 mM Tris pH 7.5.</p><!><p>The custom-synthesised fluorogenic substrate, DABCYL-KVGGNETQ-EDANS (ANYGEN, Gwangju, Korea), was used as a substrate for the proteolytic assay using ASFV protease7. This substrate contains Gly-Gly-Xaa site and the proteolytic cleavages occur after the second Gly of the consensus sequence Gly-Gly-Xaa which is recognised as a cleavage site in the maturation of the protein. The peptide was dissolved in distilled water and incubated with each protease. A SpectraMax i3x Multi-mode microplate reader (Molecular Devices) was used to measure spectral-based fluorescence. The proteolytic activity was determined at 310 K by following the increase in fluorescence (λexcitation = 340 nm, λemission = 490 nm, bandwidths = 9, 15 nm, respectively) of EDANS upon peptide hydrolysis as a function of time. Assays were conducted in black, 96-well plates (Nunc) in 350 μL assay buffers containing protease and substrate as follow; For the ASFV protease assay, 8 μL of 0.044 mM protease containing 20 mM Tris pH 7.5 was incubated with 3.5 μL of 0.1 mM substrate at 310 K for 2 h before measuring Relative Fluorescence Units (RFU). Before the assay, the emission spectra of 65 flavonoids were surveyed after illuminating at 340 nm to avoid the overlapping with the emission spectrum of EDANS. Every compound was suitable to be tested. The final concentration of the protease, peptide and chemical used at the assay was 1 μM, 1 μM and 20 μM each. At first, ASFV protease and chemical were mixed and preincubated at room temperature for 1 h. The reaction was initiated by the addition of the substrate and each well was incubated at 310 K for 2 h. After that, we measured the fluorescence of the mixture on the black 96-well plate using the endpoint mode of SpectraMax i3x where the excitation wavelength was fixed to 340 nm and the emission wavelength was set to 490 nm using 9, 15 nm bandwidth, respectively. All reactions were carried out in triplicate. Among the first sixty-five flavonoids (Supplementary Table 1), one of them was picked up to further assay at a concentration range of 1 μM ∼ 40 μM. The IC50 value which is the value causing 50% inhibition of ASFV protease was calculated by nonlinear regression analysis using GraphPad Prism 7.03 (GraphPad Software, San Diego, CA, USA).</p><!><p>The proteolytic assay using ASFV protease in the presence of Triton X-100 has been performed to differentiate the artificial inhibitory activity of chemicals through non-specific binding with proteases by forming aggregate or complexation. The concentration used in this study was 0.01%.</p><!><p>To confirm the feasibility of the assay method independently, the fluorescence spectra from the tryptophan of ASFV protease with candidate inhibitors were investigated15. The fluorescence measurements were recorded with a SpectraMax i3x Multi-mode microplate reader (Molecular Devices) at excitation and emission wavelengths of 290 nm and 300–500 nm, respectively. The optimal excitation and emission wavelengths were determined by SoftMax Pro. Four tryptophan of ASFV protease showed a fluorescence emission with a peak at 350 nm after the excitation at the wavelength of 290 nm. In contrast, the flavonoids were almost non-fluorescent under the same experiment condition. Each 20 μM and 40 μM chemical was incubated with 1 μM ASFV protease for 1 h and the fluorescence intensity of the mixture was measured.</p><!><p>The cell yield harvested for purification of ASFV protease was 3.1 g per 2000 mL of E. coli culture. The amount of purified protein synthesised with His6-tag was 2.76 mg. For storage and assay, the protein solution was concentrated to 1.38 mg mL−1. The concentrate was diluted to 1 μM when the inhibitory assay was carried out.</p><p>A flavonoid library consisting of ten different scaffolds was also built (Figure 1). It contains five isoflavones, one isoflavane, seventeen flavones, twelve flavonols, seven flavanols, seven flavanones, four flavanonol, one prenylflavonoid, nine chalcones and two unclassified flavonoids (Supplementary Table 1). We applied the library to assay ASFV protease. Using sixty-five flavonoids, an inhibitory effect of each compound at 20 μM was tested. Among them, myricetin (3,3′,4′,5,5′,7-Hexahydroxyflavone) were found to have prominent inhibitory activity. The binding affinity data were plotted as log inhibitor concentration versus percent fluorescence inhibition (Figure 2). The compound showed the severely reduced fluorescent intensity and thus represented their ASFV protease inhibitory activity. The IC50 value was calculated from the dose-dependent inhibitory curve of myricetin. The measured values were 8.4 μM. Since flavonoids are known to be aggregated through complexity and thus non-specifically inhibit various proteases, the assay in the presence of Triton X-100 was also performed16. Before the examination, we tested the effects of Triton X-100 on ASFV protease. As shown in Supplementary Figure 1, only a slight increase in catalyst activity was observed up to 0.1% Triton X-100. Therefore, the assay was performed at a concentration of 0.01% Triton X-100 with no significant interference detected.</p><!><p>The example skeleton structures of flavonoids. Ten different scaffolds classified in this study were displayed with example flavonoid derivatives. Basic skeletons and their carbon atoms numbering pattern were drawn.</p><p>Results from the FRET method. Each data point represents the effect of myricetin against ASFV protease compared to the control. The RFU are plotted against the log-concentration of inhibitory compounds. Each dot is expressed as the mean ± standard error of the mean (n = 3). RFU: Relative Fluorescence Units.</p><!><p>To independently confirm the inhibitory activity of flavonoids, a general tryptophan based assay method was employed. Tryptophan was well known to emit its fluorescence. Therefore, if tryptophan is positioned adequately in proteins, the change of fluorescence intensity can reflect the binding state of chemicals and be used to judge the interaction between proteins and chemicals. The ASFV protease contains four tryptophan residues. Therefore, its fluorescence change can reflect the environmental variation of the protein. The ASFV protease used in this study displays a fluorescence peak at 340 nm after the tryptophan excitation wavelength of 295 nm. We monitored the change of the fluorescence intensities depending on the presence or absence of all flavonoids. Since each compound in the flavonoid library was almost non-fluorescent under the experiment condition, a change of fluorescence intensity reflects interactions between the protein and chemical. The decreased emission intensity confirmed the complex formation between the ASFV protease and the inhibitory compound (Figure 3).</p><!><p>Fluorescence quenching spectra of ASFV protease. A solution containing 1 μM ASFV protease showed a strong fluorescence emission (the solid line) with a peak at 340 nm at the excitation wavelength of 295 nm. After adding 20 μM (the dashed line) and 40 (one dotted line) μM inhibitory compound, fluorescence quenching spectra were obtained.</p><!><p>ASFV has caused devastating problems from European continent in 2007 to Asian continent now2–5. Pig farmers desperately tried to prevent the spread of ASFV and the economy of affected countries was negatively influenced. Considering the absence of vaccine, the development of therapeutic agents is crucial to prevent pandemic of ASFV. Flavonoids belong to a class of plant secondary metabolites with a polyphenolic structure widely found in fruits and vegetables. They have a wide range of biochemical and pharmacological effects including antioxidants, anti-inflammatory, anti-mutagens and anti-cancer-causing properties combined with the ability to control major cellular enzyme functions. Intriguingly, some flavonoids also have antiviral activity17. Specifically, ASFV was reported to be inhibited by apigenin12 and genkwanin14 and thus its infection in Vero cells was severely reduced. The former affected on its protein synthesis and viral factory formation and the latter for its entry and egress stages. In other cases such as HCV NS3 serine protease and dengue‐2 virus NS3 protease, main proteases responsible for viral processes were inhibited directly by some flavonoids18,19. Therefore, ASFV protease can be a promising target to suppress the pathogenicity of ASFV.</p><p>In this study, the systematic analysis using various scaffolds of flavonoids targeting ASFV protease was performed. In order to find the best scaffold to inhibit the function of ASFV protease, an assay with various flavonoid derivatives classified in ten scaffolds was built and performed. Among them, the flavone, chalcone and flavonol scaffolds displayed potential inhibitory effects in order. Although the backbone skeleton of flavonol is different from that of chalcone, it is quite similar to that of flavone except the presence of the additional 3-hydroxyl group in its chromen-4-one ring (Figure 1). Since the inhibitory activity is clearly better with flavonol than flavone derivatives, the 3-hydroxyl group of flavonol seems to contribute to interact with ASFV protease (Figure 4). The best inhibitory compound was myricetin with the flavonol scaffold (Figure 2). The comparison with other homologues showed that the three hydroxyl groups of the 3,4,5-trihydroxylphenyl ring of myricetin is essential for its inhibitory activity. For example, quercetin with a 3,4,-dihydroxylphenyl ring clearly showed diminished activity. Its activity, however, is similar to those of herbacetin, kaempferol and morin (Figure 5). It implies that the 3,4,5-trihydroxyphenyl group of myricetin is important for its inhibitory function.</p><!><p>Schematic representation of flavonols. Structures of inhibitory compounds discussed in this study were listed and the positions of substituted hydroxyl groups were demonstrated. Since they belong to the flavonol family, the hydroxyl group at the carbon atom position 3 is conserved. In myricitrin and quercitrin, the L-rhamnoside moiety was placed instead of the hydroxyl group.</p><p>The effects of flavonoids on the ASFV protease activity. Each bar represents the inhibitory activity of compounds with 0.01% Triton X-100. The first black bar represents the control. Inhibitory compounds were used at 40 μM concentration. Each bar is expressed as the mean ± standard error of the mean (n = 3). RFU: Relative Fluorescence Units.</p><!><p>In order to find out better inhibitory derivatives of myricetin, a commercially available compound, myricitrin, was additionally investigated. IC50s of myricetin and myricitrin were 8.4 μM and 2.7 μM, respectively. The extra rhamnoside of myricitrin might perform a positive contribution for interaction to ASFV protease. The closest myricitrin homologue, quercitirn, which also contains the extra rhamnoside was reviewed. Interestingly, quercitirn did not showed prominent inhibitory activity. The comparison again confirmed the importance of 3,4,5-trihydroxyphenyl group of myricetin by representing its better affinity (Figure 5).</p><p>The natural polyphenol found with hydroxyl groups at 3, 5, 7, 3′, 4′ and 5′ positions of flavonol, myricetin, is very common in berries, vegetables, and in teas and wines produced from various plants. It may possess anticancer activity against hepatic, skin, pancreatic and colon cancer cells. It also has beneficial biological functions such as anti-inflammatory, anti-hypertensive properties20, and anti-HIV activities21. We first discovered and reported its potential anti-ASFV activity. Myricetin and its derivative, myricitrin, clearly suggested that their basic skeleton can be used as a reference scaffold to develop better chemicals as anti-ASFV agents. Unfortunately, a crystal structure of ASFV protease is not yet been known. Therefore, a further study is going on to determine the X-ray crystal structure of ASFV protease together with its complex structures with myricetin derivatives.</p><!><p>We formed a flavonoid library to systematically investigate ASFV protease inhibitory compounds by FRET method. Among them, the flavone, chalcone and flavonol scaffolds produced potential inhibitory effects in sequence. Myricetin and its derivative, myricitrin, with the flavonol scaffold were the best inhibitory compounds against ASFV protease in the flavonoid library. The binding of the flavonoids was independently demonstrated by a tryptophan-based fluorescence method. The three hydroxyl groups of the 3,4,5-trihydroxylphenyl ring of myricetin are essential for its inhibitory activitie and the additional rhamnoside of myricitrin can positively contribute to the interaction with ASFV protease. Since the crystal structure of ASFV protease is not yet known, further research will be conducted to determine the X-ray crystal structure of ASFV protease with myricetin derivatives.</p><!><p>Click here for additional data file.</p><!><p>No potential conflict of interest was reported by the author(s).</p>

PubMed Open Access

Gas-phase Fragmentation of Deprotonated p-Hydroxyphenacyl Derivatives

Electrospray ionization of methanolic solutions of p-hydroxyphenacyl derivatives HO-C6H4-C(O)-CH2-X (X = leaving group) provides abundant signals for the deprotonated species which are assigned to the corresponding phenolate anions \xe2\x88\x92O-C6H4-C(O)-CH2-X. Upon collisional activation in the gas phase, these anions inter alia undergo loss of a neutral \xe2\x80\x9cC8H6O2\xe2\x80\x9d species concomitant with formation of the corresponding anions X\xe2\x88\x92. The energies required for the loss of neutral roughly correlate with the gas phase acidities of the conjugate acids (HX). Extensive theoretical studies performed for X = CF3COO in order to reveal the energetically most favorable pathway for the formation of neutral \xe2\x80\x9cC8H6O2\xe2\x80\x9d suggest three different routes of similar energy demands, involving a spirocyclopropanone, epoxide formation, and a diradical, respectively.

gas-phase_fragmentation_of_deprotonated_p-hydroxyphenacyl_derivatives

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

<p>p-Hydroxyphenacyl derivatives HO-C6H4-C(O)-CH2-X (1) bearing a suitable leaving group X undergo an efficient photoreaction in water to yield the phenylacetic acid (2) via initial formation of a biradical (3), followed by ring closure to a cyclopropanone (4), and subsequent hydrolysis to 2. Overall, the sequence 1 → 2 corresponds to a photo-Favorskii rearrangement. A competing reaction, the decarbonylation to the quinomethane 5, also takes place, followed by hydration to yield the corresponding p-hydroxybenzyl alcohol 6 (Scheme 1).1</p><p>Owing to its high photoefficiency and compatibility with aqueous solvents, the reaction is of potential interest for the photorelease of compounds for medical or biological purposes2 and has therefore been the subject of intense research in the last years.3,4 While transient absorption spectroscopy has provided deep insight into the photochemical reaction,1–4 the precise mechanism of the complex rearrangement is still a matter of debate.5</p><p>Herein, we report the results of some mass spectrometric experiments which deal with free −O-C6H4-C(O)-CH2-X ions in the gas phase.6,7 Consideration of the unsolvated ion represents, of course, a major change when comparison is made with condensed-phase chemistry in many respects. Moreover, in the studies described below we use collisional heating, rather than the photoexcitation applied in Scheme 1. Most important in the specific context is that water has been shown to play a critical role in the photocleavage of the title compounds in solution, whereas the gaseous ions investigated below are unsolvated. Notwithstanding these differences, the gas-phase experiments can allow for the determination of the intrinsic reaction mechanism as well as their energetics and to directly compare these with theoretical results.8 Our original intention was actually to explore the gas-phase chemistry of the ionized compounds (either cations or anions) in order to find a suitable system for probing the Favorskii-type photochemistry that occurs in solution in a "gas-phase photochemical experiment".9 It was found, however, that simple deprotonation to the corresponding phenolate ion and their subsequent collision-induced fragmentation reactions bear an analogy with the formal Favorskii-type process observed in solution-phase photolysis. Furthermore, regardless of their possible relevance to condensed-phase processes, the fragmentations of the phenolate anions are of interest on their own.</p><!><p>Upon collision-induced dissociation (CID) of the mass-selected ions in an ion-trap mass spectrometer, each of the p-hydroxyphenacyl derivatives studied that bears a reasonably good anionic leaving group showed dissociation according to reaction (1).</p><p>The mass spectrometric experiments reported here do not reveal any information about the nature of the neutral species "C8H6O2", which does not even need to be an intact entity. A more definitive assignment by mass spectrometric means would require more sophisticated methods,10,11 which cannot be applied in this particular case due to instrumental limitations. By reference to the already existing mechanistic knowledge summarized in Scheme 1, the "C8H6O2" species could thus correspond to either the biradical 3, the cyclopropanone 4, the quinomethane 5 as the product of decarbonylation, or as other alternatives, such as the oxirane 7 or the ketene 8, which can be formed via hydrogen migration following a Favorskii-type rearrangement. These structural alternatives are shown in Scheme 2; note that the location of the radical centers in 3 is only formal and the spin is likely to be located within the aromatic ring system.12</p><p>When comparisons are made with condensed-phase experiments, it is important to note that water is not involved in these gas-phase studies. Though traces of water are present as a background in the ion-trap mass spectrometer and react with the primary product ions at extended timescales, the energizing collisions in the CID experiments occur with the helium bath gas. Therefore, the subsequent hydrolysis reactions of neutral "C8H6O2" to phenylacetic acid 2 or of the quinomethane 5 to 6 do not play a role for the gas-phase fragmentations.</p><p>As an example, the ion abundances in the energy-resolved CID spectra of mass selected −O-C6H4-C(O)-CH2-OP(O)(OEt)2 are summarized in Figure 1. At low collision energies, no fragmentation takes place. Starting from an appearance energy of about AE = 2.4 eV, the formation of −OP(O)(OEt)2 concomitant with loss of neutral C8H6O2 is observed and then rapidly gains in intensity. Slightly above 3 eV collision energy, the parent ion nearly disappears. With a branching ratio (BR) of a 2 % for X = OP(O)(OEt)2, the loss of the corresponding acid (i.e. HOP(O)(OEt)2) with concomitant formation of an ion with m/z 133 (i.e. deprotonated C8H6O2) is observed. Within the experimental error the fragmentation thresholds of these two channels are identical, which might hint at the proposed mechanism (see below). Of particular note is the relatively steep rise of the fragmentation yield with increasing collision energy, indicating that the fragmentation reaction involved in reaction (1) is not subject to any pronounced entropic or kinetic restrictions that are likely to arise from atom- or group transfer reactions, more complex structural rearrangements, or changes in spin state.</p><p>As detailed in the introduction, the title compounds undergo a very effective photolytic cleavage in the condensed phase. Given the fact that the energies of about 2 eV required for fragmentation of the isolated ions in the gas phase are much lower than the energy of the mid-UV photons (about 4 eV) typically used in photodeprotection, two obvious questions arise.</p><!><p>Could the reaction be initiated by a photo-induced deprotonation of the phenol, followed by a nucleophilic attack starting from the phenoxide ion to displace the leaving group and form 4 or 7?</p><p>Can intersystem crossing of thermally excited phenoxide ions serve as a plausible explanation for the facile fragmentation of gaseous −O-C6H4-C(O)-CH2-X ions?</p><!><p>Similar to the example shown in Figure 1, we investigated several other −O-C6H4-C(O)-CH2-X ions made by ESI of the corresponding p-hydroxyphenacyl derivatives in the negative ion mode. For all ions investigated, the CID breakdown behavior can be reproduced reasonably well with simple sigmoid fits and the relevant branching ratios and appearance energies are summarized in Table 1.</p><p>Interestingly, all fragmentation channels for an ion with a given X start at the same appearance energies (within the experimental error), suggesting that the threshold values are subject to a common, kinetically-controlled step (e.g. the barrier of a Favorskii-type process). In this context, we note that the appearance energies were derived via linear extrapolations of sigmoid fits of the fragment ion abundances to the energy axis. While this is not the most elegant approach and, furthermore, we are assuming energy-independent branching ratios where multiple fragmentations are observed, the data do provide a semi-quantitative comparison of the effect of the leaving group on the fragmentation process. This is illustrated by the reasonable correlation between the appearance energies for the formation of X− and the gas-phase acidities of the conjugate acids (HX) in Figure 2. It should be noted, however, that some of the literature ΔHacid(HX) values are associated with considerable error bars (e.g. ± 28 kJ mol−1 for MsOH) and that several others had to be estimated.</p><p>From this correlation we can roughly estimate that the occurrence of reaction (1) for the simple carboxylates (i.e. X = HCOO, CH3COO, and H2N(CH2)3COO) should have onsets at about 3 eV. Obviously, the lower thresholds of their competitive and specific fragmentations (i.e. loss of CO for formate, loss of ketene for acetate, and loss of C4H7NO for γ-aminobutyrate) suppress the formation of X− via reaction (1).13</p><p>To gain more insight into the reaction mechanism and the possible neutral products (Scheme 3), we have investigated reaction (1) for X = CF3COO by means of density functional theory, where all energies are given relative to the deprotonated starting compound [1-H]− with X = CF3COO. For an initial pre-optimization, B3LYP calculations with the small 6–31+G* were performed, while the discussion refers to the results obtained with cc-pVTZ basis sets. The choice of X = CF3COO for the computational study was due to the good quality of the experimental data and negligible side reactions in conjunction with the moderate size and the absence heavier elements like phosphorus or sulfur.</p><p>We have first explored possible fragmentations of [1-H]− in the singlet ground state. Two possible scenarios are proposed. The elimination of the CF3COO− can be assisted by the closing of either an oxirane ring or a cyclopropanone ring. The first reaction pathway is associated with an energy barrier of 0.99 eV (TS 11/19) and leads to a van der Waals complex 19. The break-up of the complex 19 yields the oxirane product 17 concomitant with free CF3COO−. The concurrent fragmentation of [11-H]− leads via a larger energy barrier of 1.39 eV (TS 11/110) to a van der Waals complex 110, which after elimination of CF3COO−, yields the spirodione 14. The spirodione can further undergo a facile decarbonylation via the transition state TS 14/15 which is only 0.37 eV higher in energy and provide the enone product 15. Interestingly, both primary "C6H8O2" products 14 and 17 are almost isoenergetic and the barriers en route to these neutral fragments are lower than the exit channels themselves, i.e. there are no barriers in excess of the overall reaction endothermicities. Accordingly, the fragmentation is likely to be under thermochemical control and hence 14 and 17 both should be formed during the dissociation of [1-H]−.</p><p>We explored the possible involvement of the biradical fragment 33. The formation of the biradical (1[1-H]− → 33 + X−)14 requires only slightly more energy (2.10 eV, Table 2) than those of the singlet products 14 and 17 (Figure 3). Given that the calculated energy demands for the formations of 33, 14, as well as 17 are all within the range of the experimental estimate of AE = (2.2 ± 0.2) eV for X = CF3COO (Table 1), the apparent CID threshold from experiment cannot be used as a criterion to distinguish these channels. A reasonable agreement between the computed and experimentally found endothermicities associated with the formation of CF3COO− also lends further support the calibration of the energy scale used in the CID experiments.15</p><p>The major difference among the possible products concerns their spin multiplicities. Given that the reactant ion 1[1-H]− is a singlet, formation of 33 would require a spin flip along the reaction coordinate ("two-state reactivity"),16 whereas formation of 14 and 17 can proceed on a single spin surface as depicted in Figure 4. Though it is often implied that effective spin changes require the presence of heavy elements and/or transition metals, some rather simple CHO compounds have been shown to obey two-state scenarios in their dissociation (e.g. CH3O+, C2O2, and C2O42+).17,18 Hence, the spin argument cannot be used to a priori exclude the high-spin product 33.19 If we consider the singlet and triplet potential energy curves calculated along the elongated C–C bond of the spirodione and the shortened the C–C bond of the biradical (Figure 4), we find that the crossing of both curves is located about 0.5 eV above the respective minima. Another aspect, perhaps unusual for synthetic chemists, is that structure 3 does exist only as a triplet and structure 4 only as a singlet, in that neither 13 nor 34 are minima on the respective potential energy surface. Intersystem crossing of 33 to the singlet surface is thus associated with a ring closure to afford 14 and the reverse process corresponds to ring opening (Scheme 3).</p><p>The high strain energy of the cyclopropanone 14 prompted us to probe its rearrangement via hydrogen migration. Though the resulting ketene 18 is significantly more stable than 33 as well as 14 (Table 2), the energy barrier resulting from the associated transition structure TS 14/18 (Erel = 3.48 eV) excludes this as a relevant intermediate in the experiment. Likewise, a tautomer of 18, (4-hydroxyphenyl) ketene 18′, is rather stable, but not accessible in a direct rearrangement from either 33 or 14. Furthermore, decarbonylation of the diradical 3 and the spiro compound 14 that affords quinomethane 5 in either spin state is quite favorable if the singlet products are formed, but the barriers involved (TS 33/35, Erel = 2.80 eV and TS 14/15, Erel = 2.36 eV) appear too large to account for the experimental findings.</p><p>Correlating the experimental data (apparent thresholds and energy behavior) as well as the theoretical findings (minima, barriers, and spin states), we conclude that close to the threshold reaction (1) for X = CF3COO can lead to the cyclopropanone 14 or the oxirane 17 with nearly equal probability. At elevated energies the triplet biradical 33 may also contribute. We note in passing that B3LYP/cc-pVTZ predicts the gas-phase acidity of CF3COOH in the correct range (1371 kJ mol−1 compared to the experimental value of 1355 ± 12 kJ mol−1).</p><p>In addition, some phenacyl compounds with differently substituted aromatic moieties were probed in order to elucidate possible electronic effects. The influence of methoxy substitution was examined for X = (EtO)2(O)PO, where AE(X−) = 2.43 eV was obtained for the unsubstituted parent p-hydroxyphenacyl compound (Table 1), whereas the 2-methoxy- and for 3,5-dimethoxy phenacyl ions gave AEs of 2.42 eV and 2.34 eV, respectively. These differences are, however, within the experimental error margins of about ± 0.05 eV in the relative and ± 0.2 eV in the absolute CID thresholds. For cyano-substituted phenacyl derivatives with X = CH3COO as potential leaving groups, again, no formation of X− was observed and loss of ketene prevailed instead with thresholds of 2.90 eV for the 2-cyano and 2.96 eV for the 3-cyano compared to only 2.52 eV for the all-H compound. The emerging trends for substitutions of the phenacyl core can by and large be attributed to mere thermodynamic effects operating on the respective reactant anions [1-H]−. Thus, donors such as methoxy substituents, are expected to decrease the electron affinity of the neutral counterpart [1-H]• and thus also decrease the stability of the [1-H]− anion. However, in the products resulting from dissociation of [1-H]− according to reaction (1), the substituents in the phenacyl core end up in the neutral product and mere electron-withdrawing or -donating effects thus become much less important. Hence, for a given X, the dissociation threshold is likely to be determined more by the thermochemical properties of the precursor ion [1-H]−. Consistent with this line of reasoning are the significantly elevated AEs of the cyano-substituted derivatives, where the electron withdrawing substituents primarily stabilize the precursor [1-H]−, whereas the corresponding effect on the neutral product of reaction (1) is much smaller. In fact, these findings parallel the photochemistry in the condensed phase. In a comparative investigation of a broad range of substituent effects that included F, MeO, CN, CO2R, CONH2, and CH3 on the photochemical rearrangements of p-hydroxyphenacyl esters, there was little effect on either the rate or the quantum efficiency for the photo-Favorskii rearrangement and the release of the acid leaving group or on the lifetimes of the reactive triplet state.20</p><!><p>Collision-induced dissociation of mass-selected p-hydroxyphenacyl derivatives as their conjugate bases, i.e. −O-C6H4-C(O)-CH2X, with various leaving groups X were probed by means of ion-trap mass spectrometry. The losses of X− concomitant with formal generation of "C8H6O2" have surprisingly small thresholds and occur with high efficiencies once the reaction endothermicity is provided in the collision experiments for most leaving groups X. Simple carboxylates such as X = HCOO or CH3COO, however, follow different fragmentation pathways. Detailed computational studies using density functional theory for X = CF3COO reveal a competition of two pathways and two spin surfaces for the loss of trifluoroacetate from the conjugate base of p-hydroxyphenacyl. In fact, we face a situation in which C–X bond heterolysis leads to cyclopropanone 14 and oxirane 17 on the singlet surface and to the biradical 33 for triplet spin. The energy demands of about 2 eV for all three pathways are too close to each other to allow an assignment of the preferred route with regard to the errors in both theory and experiment. Nevertheless, crossing between 33 and 14 is expected to be facile at ambient temperatures, and structure 4 may be regarded as a direct precursor for the formation of p-hydroxyphenylacetic acid and quinomethane 5 that hydrolyzes to 6.</p><p>Last, but not least, let us phenomenologically try to extrapolate the gas-phase data to the condensed phase. At first, it is obvious that the phenoxide group will experience a significant stabilization upon solvation, such that the activation energy associated with C X bond heterolysis is expected to be larger in solution. Next, several rearrangements which are relatively difficult in the gas phase, in particular the tautomerizations 4 → 8 and 8 → 8′, will proceed much more efficiently in protic solvents. In fact, an intimate interaction of water molecules with intermediates such as 4 may very much facilitate the interconversion to derivatives of phenylacetic acid. In future studies, we will try to address the latter aspect by investigation of the gas-phase chemistry of partially solvated ions derived from p-hydroxyphenacyl precursors using not only CID, but also experiments at elongated reaction times in the presence of water and possibly also spectroscopic methods.9</p><!><p>The measurements were performed with an ion-trap mass spectrometer (IT-MS) which has been described elsewhere21 by electrospray ionization of dilute methanolic solutions of the p-hydroxyphenacyl derivatives HO-C6H4-C(O)-CH2-X. In brief, our IT-MS bears a conventional ESI source consisting of the spray unit (typical flow rate 5 μl/min., typical spray voltage 5 kV) with nitrogen as a sheath gas, followed by a heated transfer capillary (kept at 200 °C), a first set of lenses which determines the soft- or hardness of ionization by variation of the degree of collisional activation in the medium-pressure regime,22–24 two transfer octopoles, and a Paul ion-trap with ca. 10−3 mbar helium for ion storage, manipulation, and a variety of MSn experiments.21,25 In the ESI source, the sample solution is evaporated at room temperature in the presence of a nitrogen stream and assistance by a kV voltage. The spray then passes through a high-temperature zone (typically operated at 200 °C) that facilitates droplet shrinkage and propels the spray into the differential pumping system. The nascent ions are then transferred to the mass spectrometer. For detection, the ions are ejected from the trap to an electron multiplier. Note that in the standard mode of operation, the IT-MS used has a low-mass cut-off at m/z 50.13 In the present experiments, the instrument was operated in negative ion mode, the corresponding deprotonated ions −O-C6H4-C(O)-CH2-X (X = leaving group) were mass-selected, and then subjected to collision-induced dissociation (CID) by application of an excitation AC voltage to the end caps of the trap to induce collisions of the isolated ions with the helium buffer gas. In a CID experiment, the ion of interest is gradually excited kinetically and allowed to collide with helium as a non-reactive collision partner which causes ion fragmentations. Note that rearrangements themselves are "invisible" in standard MS because these are not associated with mass changes. When the CID experiments are conducted at variable collision energies, quantitative reaction thresholds can be derived.26 In the present IT-MS experiments, we used an ion-excitation period of 20 ms, for which we have recently introduced a phenomenological linear conversion factor for the extraction of approximate appearance energies (AEs) based on comparisons with reference molecules of known bond strengths,15 i.e. Ag(CH3OH)+, Ag(CH3OH)2+, and Fe(C5H5)2+ as suggested by O'Hair and coworkers27 as well as Cu(C5H5N)2+.28 In brief, the AEs are derived from a sigmoid fit of the fragment ion abundances as a function of the collision energy.29 We note that the agreement between the AE derived experimentally and the theoretical prediction for X = CF3COO (see below) can be regarded as an additional test for the conversion scheme from relative to absolute collision energies in IT-MS and its applicability to gaseous anions.15 Furthermore, it is important to recognize that the threshold energies are determined by the highest point in energy the system has to pass prior to fragmentation and AEs can thus reflect either thermodynamic or kinetic control. Due to the presence of some residual water in the ion trap, some very reactive ionic species tend to rapidly associate with this background water during the time of trapping.30 These adducts are summed accordingly into the primary fragments.</p><p>Ab initio calculations were performed using the Gaussian03 package31 at the B3LYP level of theory with the cc-pVTZ basis set.32 The minimum geometries were obtained from ab initio optimization and further checked by frequency analysis with the same method and basis set. At this level of theory, all stationary points discussed had only positive frequencies and thus correspond to genuine minima. Harmonic frequency analysis was also used to obtain thermochemical data. All calculations refer to the gaseous state in that additional solvation, aggregation etc. is deliberately not included in order to match the present experimental conditions.</p><!><p>Breakdown graph for the CID of the mass-selected phenolate ion −O-C6H4-C(O)-CH2-OP(O)(OEt)2 () as a function of the collision energy (in eV) to yield the fragments −OP(O)(OEt)2 (, m/z 153) and C8H5O2− (, m/z 133). Both fragmentation channels have the same appearance energy of AE = (2.4 ± 0.2) eV. In addition, the graph shows the characteristic energy E1/2 at which 50 % of the parent ion fragments and the appearance energy AE derived from linear extrapolation of the slope at E1/2 to the baseline.</p><p>Correlation of the experimentally determined appearance energies (in eV) for the formation of X− upon CID of mass-selected−O-C6H4-C(O)-CH2-X ions and the gas-phase acidities of the conjugate acids (HX).</p><p>B3LYP/cc-pVTZ singlet potential energy surface for the fragmentation and rearrangements of [1-H]−. Energies are given in eV and refer to 0 K (the energies in brackets are obtained at B3LYP/6–31+G* level of theory). Only the most relevant channels are shown; for others see Table 2.</p><p>Estimated surface crossing (energies in eV) between the opening spirodione (blue, singlet) and the closing diradical (red, triplet) derived from B3LYP/6–31+G* geometry optimizations with fixed length of the (opening/closing) C–C bond.</p><p>Photo-induced Favorskii rearrangement of p-hydroxyphenacyl derivatives.</p><p>Relevant isomeric structures of the intermediate "C8H6O2" species (structure 5 has lost CO).</p><p>Possible neutral "C8H6O2" products for loss of X− upon CID of −O-C6H4-C(O)-CH2-X and their notation in the theoretical studies including the spin states (X = CF3COO); ISC stands for intersystem crossing between the spin states.</p><p>Summarized appearance energies (AE)a and branching ratios (BR) in the dissociation of mass-selected −O-C6H4-C(O)-CH2-X ions together with the gas-phase acidities (ΔHacid in kJ mol−1) of the conjugate acids of the leaving groups.b The last column gives the major "other" fragments (see supplementary material).</p><p>The relative error of the AEs with regard to the quality of the fits is below ± 0.05 eV in all cases, whereas the absolute error is estimated as ± 0.2 eV.</p><p>Values taken from the NIST database in kJ mol−1. Values in italics are estimated from analogous compounds which were in the database.</p><p>The references refer to the preparation of the various precursors; the synthesis of new derivatives is described in the Supplementary Information.</p><p>According to the mass differences observed, the majority of these side reactions are due to losses of HF (Δm = −20), phenol (Δm = −94), and HX (Δm = −162).</p><p>According to the mass differences observed, these major fragmentation reactions are due to losses of CH3• (Δm = −15), CH3• + phenol (Δm = −109), CH3OC6H4O• (Δm = −123), and "C9H8O3" (Δm = −164).</p><p>Calculated energetics (relative energies at 0 K in eV) of −O-C6H4-C(O)-CH2-OOCCF3 and its possible fragments at the B3LYP/cc-pVTZ level of theory.</p><p>Total energy: −985.384563 Hartree.</p>

PubMed Author Manuscript

Vapor- and Liquid-Phase Adsorption of Alcohol and Water in Silicalite-1 Synthesized in Fluoride Media

In this work, batch-adsorption experiments and molecular simulations are employed to probe the adsorption of binary mixtures containing ethanol or a linear alkane-1,n-diol solvated in water or ethanol onto silicate-1. Since the batch-adsorption experiments require an additional relationship to determine the amount of solute (and solvent adsorbed, as only the bulk liquid reservoir can be probed directly, molecular simulations are used to provide a relationship between solute and solvent adsorption for input to the experimental bulk measurements. The combination of bulk experimental measurements and simulated solute-solvent relationship yields solvent and solute loadings that are self-consistent with simulation alone, and allow for an assessment of the various assumptions made in literature. At low solution concentrations, the solute loading calculated is independent of the assumption made. At high concentrations, a negligent choice of assumption can lead to systematic overestimation or underestimation of calculated solute loading.

vapor-_and_liquid-phase_adsorption_of_alcohol_and_water_in_silicalite-1_synthesized_in_fluoride_medi

Introduction<!>Methodology<!>Molecular Simulations<!>Adsorbent Characterization<!>Unary Adsorption Experiments<!>Results and Discussion<!>Conclusions

<p>Adsorption from liquid mixtures onto solids is exploited in a wide range of chemical processes, ranging from fixed-bed adsorption and membrane separations to heterogeneous catalysis and their hybrids. A primary characteristic of these processes is their equilibria. However, multicomponent adsorption equilibria are often not available because of inherent challenges in their measurements [1]. As a specific example, the equilibria of diols in zeolites are important for heterogeneous catalysis [2–4] and separation [5–9] applications. Campaigns to produce high-value chemicals by renewable routes [10, 11] are opening up others.</p><p>The physical adsorption of both solute and solvent (i.e., the total uptake) from liquid solutions onto solids is difficult to determine in widely-adopted static (i.e., batch) experiments, because only the bulk liquid reservoir can be probed directly. The consequences of this limitation are best explained by the mass balances. The total mass balance for uptake from a binary mixture can be expressed as (1)Vinρin=Veqρeq+mz(QA+QS) where Vin and Veq are the initial and equilibrium solution volumes, respectively, ρin and ρeq are the respective values for the solution (volumetric mass) density, mz is the mass of the adsorbent, and QA and QS are the loadings (in mass units) of solute A and solvent S, respectively. The mass balance on the solute is (2)VinCA,in=VeqCA,eq+mzQA where CA,in and CA,eq are the solute (volumetric mass) concentrations in the initial and equilibrated solutions, respectively. In a typical experiment, Vin, CA,in, CA,eq, and mz are measured. Together, Eqs. 1 and 2 have three unknowns (Veq, QA, QS), giving them no unique solution. As a result, an additional approximation must be made in order to estimate the solute and solvent loadings. Different options typically employed in the literature are listed in Table 1.</p><p>Here, Vp, ρA, and ρS are the micropore volume of the adsorbent, the density of the liquid-phase of the solute, and the density of the solvent, respectively. The two simpler approaches are termed excess (XS) adsorption and " no-solvent" (NS) adsorption. The XS adsorption is obtained by assuming that the volume of solution does not change upon adsorption. To address the inherent volume change (VC) of solution upon adsorption, the VC assuming ideal solution can be estimated from the amount of adsorbed solute and its liquid density. A fourth method, the pore filling (PF) model, assumes that the solution adsorbed is ideal with a volume equivalent to Vp. For nanoporous materials that allow for highly selective adsorption, however, the occupiable pore volume depends on the guest molecule used in its determination [39]. The reasons for this dependence are that a certain fraction of smaller pores may only be accessible to smaller guest molecules and that, for nonspherical and/or hydrogen-bonding guest molecules, their packing is influenced by the accessible orientations in elongated channels. For example, the pore volume of silicalite-1 obtained from the saturation loadings of water or nitrogen [40] ranges from 0.125 to 0.186 mL/g (a difference of a factor of 1.5, and alcohol adsorption yields intermediate Vp values, see Table S2 in the Supporting Information, SI). Therefore, it is necessary to assess the significance of the choice of pore volume on the uptake predicted.</p><p>More complicated methods of obtaining total uptake exist, but are scarcely used. The addition of another component to the solution phase which does not adsorb in the micropores will allow for closure of mass balances. This is commonly referred to as the nonadsorbing solvent method [42–43]. However, this approach which is commonly used for liquid mixtures of hydrocarbons [44,45] breaks down for nonideal mixtures, since sorbate fugacities can be quite different in the presence of a nonadsorbing solvent component [46]. In addition, it is sometimes difficult to find an inert compound that does not adsorb after long times [35]. Alternatively, Farhapour and Bono [47] designed a special pycnometer to measure the total uptake of ethanol/water mixtures onto silicalite-1 and demonstrated that the PF model is inadequate because it does not account for VCs of mixing (i.e., nonideal solutions) in the micropores. Yu et al. [46] presented the density bottle method, which measures adsorption by the apparent density change of zeolite crystals, and showed that this method exhibits qualitative agreement with ideal adsorbed solution theory [48] (IAST) for acetone/methanol mixtures while the nonadsorbing solvent method does not. However, as the authors state in the paper [46], the density bottle method is highly dependent on the solution density, which could introduce large uncertainties in the calculation of total uptake. Moreover, the method requires a large amount of material, which could represent an issue for assessing the performance of new advantageous materials that cannot yet be synthesized in large scales. In subsequent work, the density bottle method was used to measure adsorption of liquid benzene/hydrocarbon mixtures onto silicalite-1 and NaX zeolites [49]. Bowen and Vane [50] used the density bottle method to determine the total uptake from ethanol/water solutions onto silicalite-1 and ZSM-5. However, they were unable to rationalize the finding that a zeolite with acidic aluminum sites was more selective for ethanol adsorption over water than an all-silica zeolite of the same framework type. Another rarely used approach is to determine the relative or absolute solute and solvent loadings in the porous material by spectroscopic means [51]. While there is no consensus on how to measure total uptake, predicting both solute and solvent adsorption using unary adsorption data and IAST is fraud with error for strongly associating mixtures [52–54]. Molecular simulations are useful, if not necessary, in this regard.</p><p>Recently, we reported adsorption equilibria of aqueous solutions of linear alkane-α, ω-diols (diols) with three to six carbons onto silicalite-1, and observed great agreement for solute loading between independently-conducted simulations and experiments [8]. Similar agreement between the two approaches was obtained from single-component gas-phase adsorption of methanol, ethanol, and water in silicalite-1 [53]. We realized that the simulations allow direct observation of the adsorbed phase, and that this information can be supplied to the experiments. For hydrophobic all-silica zeolites, simulations show that water adsorption primarily results from coadsorption with adsorbed alcohols; it reaches a maximum at intermediate alcohol loadings, and then decreases as the diol loading approaches saturation. In that case, an attractive option is to assume that the coadsorption in the experimental system matches that predicted by simulation. Mathematically, this can be expressed for the aqueous mixtures as (7)QS=∑k=02akQAk where {ak} (with 1 ≤ k ≤ 2) are coefficients determined from simulation, and a0 is the loading of neat solvent taken either from experiment or simulation. To this point, solutions of ethanol (E), butane-1,4-diol (B), or pentane-1,5-diol (P) in water (W) were considered. To investigate the transferability of the results to solution phases with diverse chemical characters, a solution of P in a strongly adsorbing solvent, E, was also included. The coadsorption method was then used to compare solute and solvent uptakes calculated from various approaches reported in the literature. The results demonstrate that systematic underestimation or overestimation of solute/solvent loadings can be made without careful consideration of the adsorbent and mathematical approach used. This work also opens new opportunities for the simulation of other material frameworks of varied properties, which could aid the integration between computational and experimental research in the field.</p><!><p>Certain commercially available items may be identified in this paper. This identification does not imply recommendation by NIST, nor does it imply that it is the best available for the purposes described.</p><!><p>Configurational-bias Monte Carlo simulations [55] in the isobaric-isothermal version of the Gibbs ensemble [56–58] were employed to obtain adsorption equilibria for the P/E mixture from solution at T=323 K and p=1.0 bar with a total of N=100 or 500 molecules (where the larger number was needed at lower solution concentrations to ensure sufficiently large equilibrium box sizes for the solution phase). Sampling of phase transfers was enhanced through identity switch moves [59] between P and E molecules. The interaction of the solute and solvent molecules and the zeolite were described by the TraPPE-UA [60,61] and TraPPE-Zeo [62] models, respectively. The rigid all-silica zeolite framework (silicalite-1) used was based on the structure with orthorhombic symmetry and P nma space group resolved by van Koningsveld et al [63]. Its unit cell was replicated two, two, and three times in a, b, and c, respectively, to obtain the entire simulation box representing the zeolite phase.</p><p>Eight independent simulations were carried out at each state point, and the statistical uncertainties are reported as the 95% confidence intervals estimated by multiplying the standard error of the mean by a factor of 2.4. The number of Monte Carlo Cycles (MCCs), each consisting of Ntot randomly selected trial moves, for equilibration ranged from 100,000 to 850,000 MCCs, with longer periods being required for those at low concentrations and/or high loadings. All production periods consisted of 100,000–450,000 MCCs.</p><p>The simulation data for the adsorption of the E/W, B/W, and P/W mixtures in silicalite-1 at T = 323 K and p = 1.0 bar with a total of Ntot = 1,100 molecules were taken from prior work [8,54].</p><!><p>All-silica MFI hydrophobic zeolite was synthesized by the fluoride method reported elsewhere, referred herein as MFI-F. Characterization of the material properties can be found in the same reference [32]. All-silica MFI zeolite was synthesized on a large scale using a conventional approach in alkaline medium, referred as MFI-OH. The material had its structural properties characterized. Powder X-ray diffraction (PXRD) patterns were collected using an X'Pert X-ray powder diffractometer with an X'celerator detector. Samples were scanned at 45 kV and 40 mA using Co Kα radiation (λ = 1.789 Å) and a step size of 2θ = 0.02° (50.0 s/step) over a 2θ range of 3 to 50°. Ar adsorption data were collected at T = 87 K using an Autosorb 2 from Quantachrome. Samples were outgassed at T = 573 K overnight before the measurements. Scanning electron microscopy (SEM) images were obtained on a JEOL 6500 instrument at an accelerating voltage of 5 kV. 29Si and 1H MAS NMR were performed using a Bruker DSX-500 and a Bruker MAS probe. The powder was packed into a 4 mm rotor and was spun at 14 kHz for 1H and 8 kHz for 29Si MAS NMR at room temperature. The spectral frequencies were 500.2 and 99.5 MHz, for 1H and 29Si, respectively. NMR signals were collected after 4 microsecond 90 degree pulse for both nuclei. NMR spectra were referenced to TMS for both nuclei. Characterization data can be found in the SI.</p><!><p>Water and ethanol unary sorption isotherms were conducted at 298 K. The experiments were conducted on a TA instruments VTI-SA +vapor sorption analyzer located at the Facility for Adsorbent Characterization and Testing at the National Institute of Standards and Technology. The instrument is a dynamic vapor sorption system that obtains the desired relative humidity or partial pressure value by continuously mixing a dry nitrogen flow with a humid nitrogen flow. The sample (≈25 mg) was activated in-situ at 413 K for up to 8 hr under a constant flow of pure nitrogen before starting each experiment.</p><p>Batch-adsorption experiments were conducted at T = 323 ± 0.5 K. Solution concentrations were analyzed with an Ailgent 7890B gas chromatograph equipped with a fused silica column (Rtx-VMS, Restek) and a flame ionization detector. The relative signal intensities of the adsorbate and a 1-butanol (99.5%, Aldrich) internal standard were used to determine the concentrations. The reported uncertainties in concentrations represent one standard deviation from multiple GC injections.</p><p>The initial diol solution to adsorbent mass was 4 mL/g. Approximately 100 mg of zeolite with appropriate amount of diol solution was added to glass vials (C4011–1, crimp seal, Thermo Scientific™) and then the vials were rotated at 20 r pm in a ProBlot12 hybridization oven until equilibrium was reached. The supernatant solutions were filtered using a Monoject syringe fitted with a 0.2μm hydrophilic polypropylene (GHP) syringe filter to remove the zeolite particles.</p><p>To investigate the influence of the choice of pore volume on the calculated uptake, values of pore volumes (Vp) were estimated from nitrogen adsorption [40], the solute saturation loading and liquid density, and the solvent saturation loading and liquid density. Section S1 of the SI provides additional details on the determination of Vp and the equations used to calculate the solute and solvent loadings from Eqs. 1 and 2 and one of Eqs. 3 to 6. The analytical forms of the equations used to calculate the solute loadings and solvent loadings are shown in Section S1.3 of the SI.</p><p>The densities of pure components and of aqueous solutions were obtained by extrapolation of the relationships reported in the open literature [64–66] to 323 K (see Figure S2 in the SI), allowing solution nonidealities to be captured. In the absence of reported values, the density of the P/E solutions was calculated assuming ideal solution.</p><!><p>The MFI-F and MFI-OH materials in this work are assigned monoclinic symmetry (see Section S2.1 in the SI). The gas-phase unary water and ethanol adsorption isotherms on these materials at T =298 K are shown in Figure 1 for the MFI-F and MFI-OH materials investigated in this work. The results are compared to the defect-free silicalite-1 simulations reported previously [53]. The very low water adsorption onto defect-free silicalite-1 synthesized via the fluoride route confirms its hydrophobicity. However, unary water adsorption onto MFI-OH shows high adsorption reaching ≈ 32 molecules per unit cell, indicating a very hydrophilic material. Indeed, 29Si Solid-state NMR experiments confirmed the presence of ≈ 8 % Q3 sites in MFI-OH (see Section S2.1 in the SI). As expected, the use of ethanol as adsorbate leads to stronger adsorption due to the hydrophobic aliphatic chain in the solute. There is great agreement between simulated and experimental isotherms for MFI-F, which gives us confidence that our simulation methodology is accurate.</p><p>Next, we focus on the competitive adsorption in binary solutions. As discussed previously, solution-phase batch-adsorption experiments require one assumption to obtain a unique solution to the mass balances. We propose to use the predicted coadsorption by simulation coupled with the bulk solute concentration measurements to determine the solvent loadings in the liquid phase. For most liquid-phase binary systems, this would be challenging to validate. However, the use of ethanol/water solutions is convenient because their isotherms can be obtained independently from vapor-phase experiments.</p><p>In Figure 2, we compare the liquid-phase ethanol/water adsorption isotherm onto MFI-F with simulated data. Both solvent and solute adsorption are self-consistent between simulations and our combined approach (referred to as the coadsorption method from now on), which includes experimental bulk measurements and simulated solvent loadings, especially at low concentration. As the ethanol concentration increases, the uncertainty in the measurement of solute concentration by GC increases, leading to a larger relative fluctuation in determined water loading than ethanol loading. As the saturation loading is reached, the ethanol loading calculated by the combined approach becomes ≈ 12 molecules per unit cell experimentally. This saturation loading is in line with the vapor-phase ethanol saturation of ≈ 14 molec/uc.</p><p>Next, we implement this approach to solutions of high boiling point solutes, specifically linear alkane-α, ω-diols, whose adsorption cannot be easily measured by vapor-phase unary experiments. Liquid-phase adsorption is much different in the presence of a coadsorbing solvent and a strongly adsorbing solvent. The difference lies in the way the pores are filled as a function of increasing solution concentration, as compared between P/W in Figure 3a–c and P/E in Figure 3d–f. At low concentrations, water is scarcely observed in the pores (Figure 3a), while the channels are filled by ethanol (Figure 3d). As the diol concentrations increase, the pores in equilibrium with the P/W mixture become more occupied as the loadings of both P and W increase (Figure 3b), while the occupation of the pores in equilibrium with the P/E mixture does not change significantly (Figure 3e). Instead, the adsorption of diol from the ethanolic mixture is associated with a decrease in solvent adsorption. In this sense, the diol displaces the solvent as the concentration is increased. At high concentrations, where the diol loadings approach saturation, the occupation of the pore volumes is similar, with the smaller differences being a result of different hydrogen-bonding networks and packing efficiencies from the diols with different chain lengths. The contrasting solvent effects on the total uptake of these two mixtures makes them prime candidates for assessment of the adsorption uptakes of solutions with diverse chemical character.</p><p>We then applied the coadsorption method to P/W solutions, i.e., we used simulations to obtain a relationship between adsorbed P and adsorbed W, which we used along with the measured P concentrations to solve the mass balance equations. In Figure 4a, the diol loadings obtained by simulation and the coadsorption method are presented as a function of equilibrium solution concentration for P/W mixtures. The two experimental materials were synthesized differently and possess different crystal symmetries. The simulations agree very well with the results measured experimentally for MFI-F. However, the adsorption of P obtained from MFI-OH, which was synthesized in an alkaline medium on a large scale, is different than the obtained for MFI-F. First, the adsorption step is delayed to higher P concentrations. P saturation is somewhat similar to the ones obtained for MFI-F and predicted by simulations. As will be shown in the following sections, the solute loadings calculated at low solution concentrations are independent of the calculation approach. These observations lead to the conclusion that the difference in uptakes at low/intermediate concentrations must be due to the different synthesis protocol, that is, defects introduced by the non-fluoride synthesis route.</p><p>The water loadings predicted by the coadsorption method and simulation are presented in Figure 4b. The water loadings obtained by simulations exhibit their characteristic peak, [8] which results from exclusion of water at high diol loadings where diol packing is more efficient and water coadsorption requires a larger entropic penalty. Calculation of the water adsorption by the coadsorption method reveals a similar peak for MFI-F, which again represents consistency for the hydrophobic material used in both experiments and simulations. The water loading for MFI-OH shows a broader peak, which could be due to defects in the framework. However, at low diol concentrations, it predicts that very low water amount is adsorbed on MFI-OH. This is not consistent with the unary data presented in Figure 1. The coadsorption coefficients obtained from simulations for a defect-free material can only be used for a self-consistent calculation of adsorbed water for a defect-free MFI-F, and not for MFI-OH with high silanol density. Similar observations can be made for B/W solutions, as shown in Figure S7.</p><p>The coadsorption model for predicting the adsorption of P and W onto MFI-OH must be different. Since the water adsorption is relatively large near the saturation pressure of water (see Figure 1a), the coadsorption between P and W is expected to be similar to pore-filling. Therefore, we adopt a coadsorption model with a similar mathematical form to the PF model, with a linear relationship between QP and QW. We choose two different CA models to investigate the range of loadings thatmay be calculated. The first one (Model H) assumes a high neat solvent loading of 32 molec/uc, while the second (Model L) assumes a low neat solvent loading of 20 molec/uc. For both models, the rate of solvent loading change with increasing solute loading is chosen to be-2.4 molec/uc so that the W loading at P saturation (8 molec/uc) is 0.8 molec/uc for Model L, which is consistent with simulation of the defect-free structure. The resultant isotherms for the two models are shown in Figure S9. Models L and H do not show a considerable difference for the diol isotherm except at high concentration, where a difference of ≈2 molec P/uc can be observed. The associated water loading of Model H is shifted upward from that of Model L by a constant value of ≈ 12 molec/uc at concentrations below ≈ 20 g/L, and by a slightly smaller amount at higher concentrations as the associated diol loadings from the two models begin to deviate from one another. The relatively large deviation in water loading between the two models associated with no deviation in diol loading at concentrations below 400 g/L suggests that a consistent diol loading can be calculated below high concentrations, while the associated water loading remains elusive. Simulations with frameworks containing defects could assist in validating the calculation of water loading, which represent a challenge for future studies.</p><p>We then attempted to use the density-bottle method to calculate the total adsorption uptake for P/W onto MFI-OH. The results can be found in Table S3 in the SI. Although the diol loadings agree with the ones obtained by the coadsorption method, the solvent loadings show high uncertainty, probably because they are strongly dependent on the solution density used [46], being inconclusive in our hands.</p><p>In Figure 5, the coadsorption method is used as a means to assess the performance of conventional methods used in the literature for calculating uptakes. The differences in diol and water loadings between conventional methods and those predicted by the coadsorption method are presented for five characteristic concentrations (increasing from left-to-right) in Figure 5a, b, respectively. The associated equilibrium values for solution concentration and loadings are depicted in Table 3. At low concentrations, the difference in diol loading calculated is immaterial to the calculation approach applied, and all approaches are in agreement with the coadsorption method. However, the corresponding water loadings differ by almost two orders-of-magnitude. The PF models overestimate the water loading by ≈ 40 to ≈ 60 molec/uc. The VC method underpredicts the water adsorption, while the NS and XS adsorption predict unphysical values, as their water adsorption is zero and negative, respectively.</p><p>As the solution concentration (or loading) increases (moving to the right in Figure 5), the deviation in P loading increases in magnitude while the deviation in water loading decreases in magnitude (an exception is the deviation in XS water loading, which reaches a minimum and then continues to be a large, negative value). The sign in the deviation of all terms does not change, except for the water loadings predicted by the PF-W method. This results from the volume which P occupies at saturation in the zeolite being larger than the volume which water can occupy, and also corresponds to negative water loadings (see Table S21 in the SI).</p><p>At high concentration, the choice of uptake calculation method can either overestimate or underestimate those obtained from the coadsorption method. Specifically, the NS, PF-P, PF-W, and VC methods perform well, while the PF-N and XS methods perform poorly. The PF-Napproach predicts water loadings and diol loadings which are approximately 10 and 1 molec/uc higher than that yielded by the coadsorption approach, resulting in an underestimation of the ratio water/diol adsorbed. The XS approach, which unfortunately is the most commonly employed method, performs the worst, underestimating the diol loading by almost three molec/uc and the water loading by almost 30 molec/uc.</p><p>The coadsorption method was also applied for P/E mixtures. The isotherm is presented in Figure S8. Good agreement is obtained between simulation and experimental data for MFI-F. The diol isotherm obtained from MFI-OH is somewhat similar to MFI-F. However, it underestimates the amount of ethanol adsorbed at low concentrations, indicating the need for simulations using a defects-containing framework. We then compared the coadsorption method for P/E over MFI-F to the other conventional methods in Figure 6. The associated equilibrium values for solution concentration and loadings are depicted in Table 4. Unlike water, ethanol adsorbs strongly at low solute concentrations. As the solute concentration increases, the solute (P) loading increases while the solvent (E) loading decreases (see Figure S8 in the SI). In fact, the number of ethanol molecules per unit cell decreases linearly with increasing number of diol molecules per unit cell (see Figure S1 in the SI). This suggests that the factors contributing to adsorption in ethanolic solutions are much different than those in the aqueous solutions. P molecules cannot adsorb at very low concentrations; instead, they adsorb at higher concentrations where P molecules must displace adsorbed ethanol as the concentration is increased.</p><p>The trends for deviation in diol loading for the P/E mixture are shown in Figure 6a. At low concentrations, the choice of calculation method is immaterial to the equilibria predicted. However, the deviations increase in magnitude as the concentrations are increased. The VC and NS methods are equivalent when the total mixture density is assumed to be an ideal solution (see Section S1.5 in the SI); they both slightly underestimate the P loading at intermediate concentrations and above. The most significant deviations are for PF-N (overestimation by ≈ 0.9 molec/uc) and XS (underestimation by ≈ 3.8 molec/uc). At high solution concentrations, the PF-P and PF-E diol loadings are essentially the same as the coadsorption method, with no statistically significant deviations. Less approaches are in agreement with the coadsorption method for the P/E mixture, suggesting that it is more difficult to accurately calculate the diol adsorption from ethanol solution.</p><p>The solvent deviation for the ethanolic solution is much different than for the aqueous solution. This is because ethanol saturates the zeolite at low diol solution concentrations. As a result, the pore-filling methods are much more realistic (although PF-N still overestimates by ≈ 5 molec/uc).</p><p>Since the VC and NS methods both yield zero solvent loading, and the solvent adsorption predicted by the XS method is always negative, only the PF models predict finite selectivities. The PF-N method slightly underestimates the selectivity at low concentrations, and the extent of underestimation increases as the concentrations increase. At all concentrations observed, the selectivities predicted by PF-P and PF-E are statistically equivalent to those predicted by the coadsorption method. This is expected based off of the `displacement adsorption' mechanism described earlier, and serves as additional validation of the coadsorption approach.</p><p>The results for the four different options for calculating the total uptake from liquid solution are summarized in Table 5 for the two different solutions and concentration regimes. At low concentrations the useful result observed is that the solute loading calculated is inconsequential to the method chosen. The solvent loadings can be underestimated or overestimated depending on the approach used. Specifically, the pore-filling model can introduce great uncertainties for aqueous solutions where the co-adsorption mechanism occurs. It may be applicable, however, to solutions where the displacement adsorption mechanism is more favorable, as the case of ethanolic solutions. It is also worthy mentioning that the choice of pore-volume is critical for this type of approach and should be used carefully. At high diol concentrations, different approaches can either underestimate or overestimate the loadings for both solute and solvent.</p><!><p>In order to predict the total uptake in a solution-phase batch-adsorption experiment, an assumption must be made to close the mass-balance equations. In this work, a new option for closure of the mass balances, referred to as the coadsorption method, is presented. This approach matches the coadsorption of experiment to that yielded by molecular simulation, which can directly relate the uptake of both solute and solvent in the adsorbed phase. The coadsorption method was found to be self-consistent with simulation of defect-free silicalite-1 crystals. In addition, it was validated by comparing the calculated adsorption of ethanol/water mixtures from liquid-phase to that measured independently in single-component adsorption from vapor-phase. The coadsorption method was then used to determine the adsorption of different binary diol solutions and as a basis to assess the accuracy of conventional approaches for the calculation of uptakes and selectivities. This work is, to our knowledge, the first systematic investigation of the effect of different assumptions on the resultant equilibria.</p><p>The results demonstrate that the excess adsorption model, while mathematically simple and extremely popular in the literature, is a poor choice for calculating adsorption from solution-phase in batch experiments above low concentrations. It leads to a systematic underestimation of solute adsorption above low concentrations. In addition, using the pore-filling model with a pore volume estimated from the adsorption of a small molecule (e.g., N2), which is perhaps the second-most-popular approach, leads to a systematic overestimation of solute loading above low concentrations. Using the pore-filling model with a pore volume estimated from the saturated loading and liquid-phase density of the pure solute or solvent leads to more accurate solute loadings. For pentane-1,5-diol/ethanol solutions, this approach also led to accurate solvent loadings. For the adsorption of diols from aqueous solutions onto a hydrophobic, all-silica MFI-type zeolite, batch-adsorption measurements with conventional uptake calculation approaches could not accurately predict the selectivity without input from molecular simulation. Although the coadsorption method with simulation of defect-free MFI did not accurately represent the solvent adsorption onto hydrophilic MFI, the results present new opportunities for the simulation of imperfect crystals as a way to investigate adsorption phenomena from liquid-solutions coupled with experimental data.</p>

PubMed Author Manuscript

A femtosecond transient absorption spectroscopic study on a carbonyl-containing carotenoid analogue, 2-(all-trans-retinylidene)-indan-1,3-dione

The photophysical properties of a carbonyl-containing carotenoid analogue in an s-cis configuration, relative to the conjugated \xcf\x80 system, 2-(all-trans-retinylidene)-indan-1,3-dione (C20Ind), were investigated by femtosecond time-resolved spectroscopy in various solvents. The lifetime of the optically forbidden S1 state of C20Ind becomes long as solvent polarity increases. This trend is completely opposite to the situation of S1-ICT dynamics of carbonyl-containing carotenoids, such as peridinin and fucoxanthin. Excitation energy dependence of the transient absorption measurements shows that the transient absorption spectra in non-polar solvents were originated from two distinct transient species, while those in polar and protic solvents are due to a single transient species. By referring to the results of MNDO-PSDCI (modified neglect of differential overlap with partial single- and double-configuration interaction) calculations, we conclude: (1) In polar and protic solvents, the S1 state is generated following excitation up to the S2 state; (2) In non-polar solvents, however, both the S1 and 1n\xcf\x80* states are generated; and (3) C20Ind does not generate the S1-ICT state, despite the fact that it has two conjugated carbonyl groups.

a_femtosecond_transient_absorption_spectroscopic_study_on_a_carbonyl-containing_carotenoid_analogue,

Introduction<!>Sample Preparation and Solvents Used in This Study<!>Steady-state Absorption and Resonance Raman Spectroscopic Measurements<!>Femtosecond Transient Absorption Spectroscopic Measurements<!>Quantum Chemical Calculation<!>Steady-state Absorption and Resonance Raman Spectra<!>Solvent Dependence of Femtosecond Transient Absorption Spectra and Kinetics of C20Ind<!>The Excitation Energy Dependence of The Femtosecond Transient Absorption Spectra and Their Kinetics for C20Ind in n-hexane and Acetone<!>MNDO-PSDCI Calculations of C20Ind<!>Conclusions<!>

<p>Carotenoids are naturally occurring chromophores containing a long-chain polyene backbone and play an important role in photosynthesis acting as accessory light harvesting molecules.1,2 The singlet excited states of carotenoids without carbonyls are designated by referring to an idealized C2h point symmetry group.2,3 The lowest optically allowed singlet excited-state is the 11Bu+ state (S2), and an optically forbidden 21Ag− state (S1) exists below S2. The presence of an intramolecular charge transfer (SICT) state coupled to the S1 state (S1-ICT) have been reported for carbonyl-containing carotenoids, such as peridinin, in solution and in pigment-protein complexes.4–9 The S1-ICT state plays a key role in the photophysical properties of carbonyl-containing carotenoids as was shown by transient absorption spectroscopic measurements.10–13 The lifetime of the S1 was found to become shorter as the length of the π-conjugation increases10–13 similar to the behavior of carotenoids without carbonyls.14,15 In contrast, the S1-ICT state lifetimes are invariant with respect to the conjugated chain-length.10–13 These experimental findings are supported by quantum-chemical calculations.16–18 Models of the S1-ICT state have been proposed based on evolution of the S1 state.4–13,16–18 Recently it was reported that the S1 and SICT states of fucoxanthin can be generated distinctly by selecting two-photon excitation.19 The situation of peridinin, however, is more complicated and a distinction of S1 and S1-ICT is not straightforward.9 In the case of 3′-hydroxyechinenone (heCN), its S1-ICT state was not found to be generated in solution, but it is formed in the orange carotenoid protein (OCP) from the cyanobacterium Arthrospira maxima.20 The carbonyl group of heCN in OCP is locked in an s-trans configuration to its polyene backbone.20 Furthermore, canthaxanthin and rhodoxanthin with terminal ring structures containing carbonyl groups do not produce an S1-ICT state in solution because of the twisted geometry of the terminal rings with respect to the conjugated polyene backbone.21 These phenomena suggest that the conformation of a carbonyl group to the π-conjugated system in polyene backbone plays an important role in the generation of the S1-ICT state.</p><p>Moreover, there is another excited singlet state that complicates the photophysics of these polyenes. The presence of a low-lying 1nπ* state is a general feature of polyenes with a carbonyl group, such as benzophenone, trans-β-hydrindanone and linear polyenals.22–24 This state plays important role in intersystem crossing from the singlet to triplet state as well as the trans to cis isomerization.22,25 Presence of 1nπ* state between the S1 and ground state (S0) was reported in retinal having 5 conjugated double bonds and an aldehyde-type carbonyl group.26–29 The lifetime of 1nπ* state of retinal in protic solvents is shorter than that in non-polar solvents.29 However, as yet there is no clear evidence for the presence of the 1nπ* state in carotenoids with carbonyl groups except for the case of siphonaxanthin.30</p><p>In a previous study, we synthesized a carotenoid analogue with 2 carbonyl groups, 2-(all-trans-retinylidene)-indan-1,3-dione (abbreviated as C20Ind, see Figure 1 for chemical structure).31 The conjugation of C20Ind extended onto both the carbonyls but not onto the indan ring. It has six carbon-carbon conjugated double bonds attached onto a branched double bond.31 This conjugation pattern is very similar to that of peridinin. The structure of C20Ind has been determined by X-ray crystallography.31 The carbonyl groups of C20Ind were locked in the s-cis conformation in relation to the conjugated polyene chain. This suggests that C20Ind represents a simplified model to help examine the role of the carbonyl group in the generation of the S1-ICT and/or 1nπ* states.</p><p>The aim of this present study is to investigate the formation of the S1-ICT and/or 1nπ* states in a short carotenoid with two carbonyl groups in an s-cis conformation. We present here a combined experimental and theoretical study of C20Ind. Our experimental methods include electronic and Raman spectroscopy as well as femtosecond time-resolved absorption spectroscopy in a variety of solvents. Our theoretical method is MNDO-PSDCI (modified neglect of differential overlap with partial single- and double-configuration interaction) theory including full single and double configuration interaction (CI) within the π system.</p><!><p>C20Ind was synthesized and purified as previously reported.31 Retinyl acetate was purchased from BASF (Switzerland) and purified by silica gel column chromatography and subsequent recrystallization from n-hexane. Indan-1,3-dione was purchased from Tokyo Kasei (Japan) and used as received. All-trans retinal was prepared by the hydrolysis of retinyl acetate followed by oxidation using MnO2 (Aldrich Chemicals, U.S.A.). C20Ind was prepared by means of a modified Knoevenagel condensation of retinal and indan-1,3-dione in methanol.</p><p>C20Ind was dissolved in non-polar, n-hexane (Kishida Chemical, Japan) and toluene (WAKO Chemical, Japan), in polar, THF (WAKO Chemical, Japan), acetone (Kishida Chemical, Japan) and acetonitrile (Kishida Chemical, Japan), and in protic, methanol (Kishida Chemical, Japan), solvents for the following studies.</p><!><p>The optical density of C20Ind was adjusted to 1 at the maximum of the steady-state absorption spectrum. The steady-state absorption spectra were measured using a conventional spectrophotometer (V-670, Jasco, Japan). A diode-pumped solid state CW-laser with output at 532nm (SDL-532-SLM-030T, Shanghai Dream Lasers Technology, China) was used as the excitation light-source to generate resonance Raman spectra of C20Ind. The Raman scattering was detected by using a liquid nitrogen cooled CCD camera (LN/CCD-576-E/1, Roper Scientific, Japan) attached to a spectrometer (U1000, Horiba-Jovin-Ybon, Japan). The back-scattering optical geometry was used and all measurements were performed at room temperature.</p><!><p>The absorbance of C20Ind was adjusted to 0.5 at the maximum of steady-state absorption spectra in 1 mm optical path-length quartz cell. The experimental setup of the femtosecond transient absorption measurements was as described in a previous paper.32 A mode-locked Ti:Sapphire oscillator and a 1 kHz regenerative amplifier (Hurricane-X, Spectra Physics, U.S.A.) provided the excitation and probe pulses. Excitation pulses were obtained by sum-frequency mixing the output of an optical parametric amplifier (OPA-800CF, Spectra Physics, U.S.A.) with a residual fundamental pulse in a 1.0 mm BBO crystal. The excitation energy was adjusted to 20 nJ/pulse. A white light continuum probe pulse, generated using a 5.0 mm sapphire plate, was detected by a photo-diode array (1024 pixels NMOS linear image sensor S3903-1024G, Hamamatsu, Japan) through a spectrometer (Acton SP275i, Princeton Instruments, U.S.A.). The excitation pulses were modulated at 500 Hz by an optical chopper (C-995 Optical Chopper, Terahertz Technologies Inc., U.S.A.) and the data output was synchronized with the laser repetition of 1 kHz using home-built electronic circuitry.32 The instrumental response function of the system determined by cross-correlation between the excitation and probe pulses was better than 100 fs. The cross-correlation function was used to determine the zero time delay at each probe energy. After chirp compensation, the uncertainty in the zero time delay was less than 20 fs.</p><!><p>The ground-state equilibrium geometry and ground state properties of C20Ind were calculated by using Gaussian 03, the B3LYP density functional, and a 6-31G(d) basis set.33 The excited-state electronic properties were calculated by using MNDO-PSDCI molecular orbital theory and an AM1 Hamiltonian. Quantum chemical calculations using MNDO-PSDCI, SAC-CI, and TDDFT methods have been applied to a carbonyl containing carotenoid fucoxanthin.18 None of each calculation can fully explain the properties of the SICT state of fucoxanthin, but some parts of the experimental results were satisfactorily interpreted using the results of these calculations. Among these three calculation methods SAC-CI is the highest end, but it needs sufficiently high computer resource and huge amount of calculation time for the practical usage. Nevertheless, calculations using this method cannot correctly predict the values of parameter |Δμ| (change of static dipole moment of molecules upon photoexcitation) that is important to discuss the property of SICT. TDDFT method is useful to theoretically predict |Δμ| values, but it does not correctly predict the ordering of the singlet excited-states of fucoxanthin. MNDO-PSDCI method is handy to use and can correctly reproduce the ordering of the S1 (21Ag−), S2 (11Bu−), and S3 (11Bu+) states as can be done using SAC-CI method. Therefore, we have adopted the MNDO-PSDCI method in this present study. MNDO-PSDCI calculations using an AM1 Hamiltonian were applied to C20Ind whose structure was optimized in the S0 ground state using DFT method. These methods have been applied successfully to the study of several comparable systems including carbonyl-containing carotenoids.17,18,34–43 We carried out full single and double CI calculations within the eight highest energy filled π orbitals, the eight lowest energy unfilled π orbitals and the two highest energy filled n orbitals. The MNDO-PSDCI programs are available by contacting R. R. Birge (rbirge@uconn.edu).</p><!><p>Figure 2 shows steady-state absorption spectra of C20Ind in several solvents. In all cases, the vibronic development of the S0 → S2 transition cannot be resolved due to inhomogeneous broadening. However, it is clearly seen that the energy of the maximum absorbance is red-shifted when increasing solvent polarity. This red shift is caused by an electrostatic solvent effect as has already been observed in other carbonyl-containing carotenoids.4,5,10–13 Figure 3 shows resonance Raman spectra of C20Ind in the same set of solvents. Based on the assignment of the Raman bands in the previous papers, the Raman lines observed for C20Ind can be empirically assigned as below.44–49 The Raman lines in the 959 – 967 spectral region are assigned to the C-C-H out-of-plane bending vibration. Those around 1007 cm−1 are assigned to methyl in-plane rocking vibrations. Those around 1270 cm−1 are assigned to the coupling mode between C-C-H bendings and the carbon-carbon single and double bond stretching vibrations. Those in the 1200 - 1150 and 1550 - 1530 cm−1 spectral regions are assigned to the carbon-carbon single and double bond stretching vibrations, respectively. Those in the 1375 - 1328 cm−1 region and around 1445 cm−1 are assigned to the symmetric and asymmetric methyl deformation, respectively. Those in the 1730 - 1715 cm−1 region are assigned to the carbonyl C=O stretchings. In all solutions, the Raman lines of the carbon-carbon double bond stretching vibrations (1550 - 1530 cm−1 spectral region) are the strongest compared to the other Raman lines. Based on the results of resonance Raman spectra, the system origin and the full width at half maximum (FWHM) of the steady-state absorption spectra were determined by a Franck-Condon analysis as described below.50</p><p>For transitions from the ground state, |S0>, to the lowest optically allowed excited state, |S2>, with energies {E2} the linear absorption coefficient, α(ω), defined as the fraction of energy absorbed in passing through an isotropic material, is</p><p>Flm=|<l|m>|2 is a Franck-Condon factor, the square magnitude of the instantaneous overlap of the mth nuclear wavefunction of the ground state, |m>, and the lth nuclear wavefunction of the excited state, |l>. μ̂ is the transition dipole operator for the transition from |S0> to |S2>. Ω02 = (E2 − E0/Ħ) is the angular transition frequency of the state |S2>. ω is the angular frequency of incident light. ων is the angular molecular vibrational frequency. Assuming that the steady-state absorption spectra of C20Ind in solution are convoluted with Gaussian sub-bands, we have</p><p> α(ω)∝∑l∣〈S0∣μ^∣S2〉∣2·Flm·exp(−(ω−Ω02−l·ωv)22·Γ2),Γ=σ22ln2 where σ is a full width at half maximum (FWHM) of a Gaussian function. At the temperature of the system T = 0, the lowest vibrational state (m = 0) of the ground state is occupied. We define the Franck Condon factor as</p><p> Fl0=∣〈l∣0〉∣2=Sl·exp(−S)l!, where S is a Huang-Rhys factor. Then, we obtain the fitting model function of the steady-state absorption spectra, A(ω), as</p><p>Based on the observed resonance Raman spectra, the frequency of carbon-carbon double bond stretching mode is dominant and hence used as ων. S, Ω02 and Γ are determined by spectral fitting.</p><p>The calculations (solid-lines in Figure 2) agree well with the experimental results. Table 1 summarizes the fitting parameters determined by the Franck-Condon analysis. The large Huang-Rhys factor is due to the large displacement of equilibrium nuclear positions between ground and excited states caused by strong electron-phonon coupling. Figure 4 shows the plots of the 0-0 transition energy and FWHM as a solvent polarity factor, Pf = (ε−1)/(ε+2)−(n2−1)/(n2+2). This parameter can be derived to combine the Lippert and Mataga equation51 and the effect of elliptical cavity formed by the surrounding solvent molecules.52 It is appropriately applicable for the case of carotenoids to explain the solvent dependent energy-shift of optical absorption.10,52 It is suggested that the system origins of the steady-state absorption spectra are blue-shifted and that the spectral line widths are broadened as solvent polarity increases. This trend is similar to that observed with other carbonyl-containing carotenoids.4,5,10–13</p><!><p>Figure 5 shows femtosecond transient absorption spectra of C20Ind in several solvents. In all cases, blue-shifting and narrowing of the transient absorption spectra are observed at early delay times after excitation (0.1, 0.5, and 2 ps). A typical relaxation pathway of carotenoids in non-polar solvent has been proposed as S2 → hot S1 → S1 → S0.53–55 Based on this idea, the spectral change observed here is assignable to the vibrational relaxation in the S1 state. In the transient absorption spectra in polar solvents (THF, acetone and acetonitrile) at 5 ps after excitation (dashed double-dotted lines in Figure 5), the peak energy of transient absorption is blue-shifted with an increase of solvent polarity. The spectral band shape does not show significant difference at delay times longer than 2 ps after excitation (Figure 5(c), 5(d) and 5(e)). A similar trend can be seen in a protic solvent (Figure 5(f)). In carbonyl-containing carotenoids such as peridinin and fucoxanthin, a transient absorption of the S1-ICT can be seen at the lower energy side of the S1 → Sn absorption transition band in polar and protic solvents.4,5,10–13 In Figure 5, however, a new transient absorption band of C20Ind in polar and protic solvents is not detected compared to that in non-polar solvents. This phenomenon suggests that the transient absorption of the S1-ICT state of C20Ind has not been detected. The spectral behavior in non-polar solvents, however, is quite different from that in polar and protic solvents. The spectral bandwidths of the transient absorption of C20Ind in non-polar solvents are apparently broader than those in polar solvents. In order to gain insight into the origin of the spectral broadening, the excitation energy dependence of the transient absorption spectra and their kinetics was investigated, for C20Ind in n-hexane and acetone (see next section).</p><p>Figure 6 shows the solvent dependence of the transient absorption kinetics of C20Ind. In all cases, the kinetics are well fitted by multi-exponential functions. Table 2 summarizes the results of the fitting. Based on the relaxation pathway of typical carotenoids, such as β-carotene, lycopene, zeaxanthin and spheroidene,53–55 τ1, τ2 and τ3 correspond to the S2, hot S1 and S1 lifetimes, respectively. 10-ns is regarded as infinity compared to the time scale in which we are observing the kinetics using femtosecond time-resolved absorption spectroscopy. This value is conventionally adopted and is fixed for the fitting of the kinetics traces. Also the 10-ns decay component shows good agreement with the bleaching signal of S0 → S2 absorption in the spectral domain. The degradation of C20Ind by direct photoexcitation has already been reported by Fujii et al.56 Therefore we assigned the origin of this 10-ns decay component to the bleaching of the S0 → S2 absorption caused by the sample decomposition. The S1 → S0 internal conversion time constant (τ3 in Table 2) becomes long as solvent polarity increases. This trend is completely opposite to the situation of S1-ICT dynamics of carbonyl-containing carotenoids, such as peridinin and fucoxanthin.4,5,10–13 Rather, this trend is reminiscent of the S1 dynamics of a carotenoid with an aldehyde-type carbonyl group.29</p><!><p>Figure 7 shows the excitation energy dependence of the transient absorption spectra of C20Ind in n-hexane and acetone. In all cases, the maxima of transient absorptions show blue shifts between 0.1 and 2 ps. This phenomenon is interpreted as vibronic relaxation in the excited state. 53–55 The transient absorption spectra at longer than 2 ps after excitation in n-hexane depend strongly on excitation energy, whereas those in acetone do not. In order to gain insight into the origin of the spectral difference, spectral decomposition of the transient absorption spectra of C20Ind was carried out.</p><p>Figure 8 shows the excitation energy dependence of the normalized transient absorption spectra and the results of the spectral decomposition at delay times after excitation of 3 ps, 5 ps and 10 ps in both solvents. Table 3 shows the relative intensities, peak energies, and FWHM of the transient absorptions determined by spectral decomposition. In the cases following excitation at 2.25 and 2.58 eV in n-hexane, the transient absorption spectra are well reproduced using one or two Gaussian functions together with a bleaching component corresponding to bleaching of the steady-state absorption spectrum, as shown in Figure 8(a) and 8(b). In the case of both these excitation energies in acetone, three Gaussian functions (the lowest, middle, and the highest energies) and one bleaching component are required to achieve satisfactory fitting of the transient absorption spectra (Figure 8(c) and 8(d)). Transient absorptions following excitation at 2.25 eV in n-hexane (Figure 8(a)) show a similar spectral pattern at each delay time. Whereas, with those following excitation at 2.58 eV in n-hexane the relative intensity of each Gaussian sub-band apparently changes depending on a delay time (see relative intensity of the "lowest energy" and "middle energy" positions following excitation at 2.58 eV in n-hexane in Table 3). This suggests that at least two excited states must be involved in the dynamics of these transient absorption spectra at times longer than 3 ps after excitation at 2.58 eV.</p><p>Transient absorption band shapes in acetone are independent of the delay time and excitation energy. This suggests that the same single excited state was generated following the relaxation from the S2 state after excitation at 2.25 eV and 2.58 eV in acetone. Figure 9 shows the excitation energy dependence of the transient absorption kinetics of C20Ind in n-hexane and acetone. The sharp peak that appeared instantaneously following excitation has been assigned to nonlinear optical effects.57,58 In all cases, the kinetics are well fitted by multi-exponential functions. Table 4 summarizes the results of this fitting. The lower energy (LE) and higher energy (HE) bands generated following excitation at 2.58 eV in n-hexane have different decay time constants (5.4 and 5.0 ps, respectively). Furthermore, the decay time constant of the HE band is close to that of the excited state generated following excitation at 2.25 eV in n-hexane. These results suggest that the HE band is the same excited state that is generated following excitation at 2.25 eV. The LE band is generated in a time regime shorter than the instrument response function (~100 fs). Therefore, it can be tentatively concluded that the LE band was generated either by direct photoexcitation or by relaxation from the higher vibrational excited state of the S2 state. The rise component of the HE band is not synchronous with the decay component of the LE band.</p><!><p>Figure 10 shows the ground state chemical structure of C20Ind determined by density functional B3LYP/6-31G(d) calculation and the solvent dependence of the energies of the S2 (11Bu+), S1 (21Ag−) and 1nπ* states predicted by MNDO-PSDCI calculations. It should be noted that the energies of the S2 and 1nπ* states were determined to be very close (~0.1 eV). Consequently, in the case following excitation at 2.25 eV in n-hexane, it is suggested that only the S2 was generated, whereas in the case following excitation at 2.58 eV in n-hexane, both the S2 and 1nπ* states were generated. The S2 of carotenoids instantaneously relaxes to the S1.53–55 Therefore, it is suggested that the HE band produced following excitation at 2.58 eV and the excited state produced following excitation at 2.25 eV in n-hexane can be unequivocally assigned to the S1 → Sn transition. The LE band produced following excitation at 2.58 eV in n-hexane can then be assigned to the 1nπ* state → Sm (a higher lying excited state) transition.</p><p>Figure 11 shows a schematic description of the relaxation pathways of C20Ind in n-hexane based on the experimental results and the calculations. In the case of excitation at 2.25 eV, the relaxation pathway is proposed as S2 → hot S1 → S1 → S0. In the case of excitation at 2.58 eV, two cases are considered for the relaxation pathways of C20Ind. One is that both hot S2 → S2 → hot S1 → S1 → S0 and 1nπ* → S0 occur simultaneously, the other is that both hot S2 → S2 → hot S1 → S1 → S0 and hot S2 → 1nπ* → S0 occur simultaneously. The latter idea, the branching pathway from the S2 state, has been reported in previous papers.59,60 The lifetimes of each excited state of C20Ind were experimentally determined (Table 4). In the previous reports, 1nπ* state was reported to be generated in the relaxation process of the S1 of retinal having a short polyene chain conjugated with an aldehyde-type carbonyl group.26–29 On the other hand, it is reported that the nearest underlying state to the S2 is the 1nπ* state for siphonaxanthin, a carbonyl-containing carotenoid with longer conjugated chain length than retinal.30 The relaxation pathway of siphonaxanthin following excitation to the S2 is proposed as S2 → 1nπ* state → S1 → S0.30 The generation and relaxation process of the 1nπ* state of C20Ind is clearly different from those of retinal and siphonaxanthin.</p><p>Based on the calculations, it is expected that both the S2 and the nπ* states are generated by the excitation at 2.58 eV in acetone. However, the transient absorption spectra and kinetics are independent of excitation energy, suggesting the production of a single excited-state. The lifetime of the 1nπ* state of retinal in protic solvents is shorter than that in non-polar solvents.29 By inference, based on this observation, it can be suggested that the 1nπ* state of C20Ind in acetone is too short-lived to be detected in the present study.</p><!><p>The steady-state and femtosecond transient absorption spectra of C20Ind in various solvents were investigated. The peak shift and the broadening of the steady-state absorption spectra with increasing solvent polarity is similar to those observed in other carbonyl-containing carotenoids such as peridinin and fucoxanthin.4,5,10–13 In contrast, the trend observed for the solvent dependence of the transient absorptions was unusual. The observed solvent dependence was similar to that observed for the S1 state of retinal.29 The femtosecond transient absorption spectra of C20Ind following excitation at 2.58 eV in non-polar n-hexane were composed of the two transient absorption bands (the HE and LE bands), whereas in the case of excitation at 2.25 eV only the HE band was observed. The HE and LE bands have been assigned to S1 → Sn, and 1nπ* states → Sm (higher lying excited states), respectively, based on the MNDO-PSDCI calculations. The kinetics of the LE band indicate that the 1nπ* state of C20Ind was generated by either direct photoexcitation or relaxation from the higher vibrational excited states of the S2. The present results suggest that an S1-ICT does not form in a carotenoid with two carbonyl groups in an s-cis conformation relative to the polyene backbone.</p><!><p>Chemical structure of 2-(all-trans-retinylidene)-indan-1,3-dione (C20Ind)</p><p>Steady-state absorption spectra of C20Ind in (a) n-hexane, (b) toluene, (c) THF, (d) acetone, (e) acetonitrile, and (f) methanol. Pf is the solvent polarity factor, which was determined from the dielectric constant ε and the refractive index n of the solvents using the expression, Pf = (ε−1)/(ε+2)−(n2−1)/(n2+2). Circles indicate the experimental data. Solid-lines indicate the results of spectral fitting based on a Franck-Condon analysis. All the spectra are normalized at the absorption maximum.</p><p>Spontaneous resonance Raman spectra of C20Ind in (a) n-hexane, (b) toluene, (c) THF, (d) acetone, (e) acetonitrile, and (f) methanol observed at 532nm excitation. Raman lines due to solvents are subtracted. All the spectra are normalized at the highest peak signals.</p><p>Solvent polarity dependences of the 0-0 transition energy (top) and a full width at half maximum (FWHM) (bottom) of the absorption bands of C20Ind. Open circles show the experimental results. Solid-lines show the results of the least square fittings except for the data points in methanol ((f) in this figure).</p><p>Femtosecond transient absorption spectra of C20Ind following excitation at 2.43eV in (a) n-hexane, (b) toluene, (c) THF, (d) acetone, (e) acetonitrile, and (f) methanol. Dotted lines show the steady-state absorption spectra. Dashed, solid-, dashed-dotted, dashed double-dotted and long-dashed lines show the femtosecond transient absorption spectra recorded at 0.1 ps, 0.5 ps, 2.0 ps, 5.0 ps, and 10 ps after excitation, respectively.</p><p>The normalized transient absorption kinetics of C20Ind in (a) n-hexane, (b) acetone, (c) THF, (d) acetone, (e) acetonitrile, and (e) methanol. All kinetic traces were normalized using the amplitude at 3 ps after excitation. Open circles, open tetragons, and open triangles show the experimental data. Solid-lines show the results of fitting using multi-exponential functions. The insets show the same plots of a short time regime after excitation.</p><p>Excitation energy dependence of the transient absorption spectra of C20Ind in n-hexane (left-hand side) and acetone (right-hand side). (a) and (b) show the transient absorption spectra in n-hexane for 2.25eV and 2.58eV excitaion, respectively. (c), and (d) show those in acetone for 2.25eV, 2.58eV excitaion, respectively. Dotted lines show the steady-state absorption spectra. Dashed, solid-, dashed-dotted, dashed double-dotted and long-dashed lines show the femtosecond transient absorption spectra recorded at 0.1 ps, 0.5 ps, 2.0 ps, 5.0 ps, and 10 ps after excitation, respectively.</p><p>Delay time dependence and spectral decomposition of the normalized transient absorption spectra of C20Ind in n-hexane (left-hand side) and acetone (right-hand side). (a) and (b) show the results of spectral decomposition in n-hexane following excitation at 2.25eV and 2.58eV, respectively. (c) and (d) show those in acetone following excitation at 2.25eV and 2.58eV, respectively. Dashed, dashed dotted and dashed double-dotted lines show the femtosecond transient absorption spectra recorded at 3 ps, 5 ps, and 10 ps after excitation, respectively. Solid-lines show the results of spectral decomposition determined by two or three Gaussian functions together with the bleaching of the steady-state absorption spectra. Dotted lines show the Gaussian sub-bands determined by the spectral decomposition. Dashed and solid arrows in (b) indicate the lower energy (LE) and the higher energy (HE) band, respectively.</p><p>Excitation energy dependence of the kinetics of transient absorbance change of C20Ind (a) in n-hexane following excitation at 2.25eV, (b) in n-hexane following excitation at 2.58eV, (c) in acetone following excitation at 2.25eV, and (d) in acetone following excitation at 2.58eV. Open circles, open tetragons, and open triangles show the experimental data. Solid-lines show the results of fitting using multi-exponential functions. The insets show the same plots of a short time regime after excitation. "HE band" and "LE band" in (b) correspond to the higher and lower energy sub-bands, respectively, in Figure 8(b).</p><p>(Top) A ground state chemical structure of C20Ind determined by density functional B3LYP/6-31G(d) calculation. (Bottom) Energy diagram of three lowest singlet excited states of C20Ind in (a) vacuum, (b) isopentane, (c) cyclohexane, (d) diethyl ether, (e) acetonitrile, (f) benzene, (g) decaline, (h) 3-methylpentane, and (i) ethanol predicted by MNDO-PSDCI calculations. Dotted, dashed and solid-lines show the transition energy of the 21Ag−, the 11Bu+, and the 1nπ* states of C20Ind, respectively.</p><p>Schematic drawings of the relaxation pathways of C20Ind in n-hexane. Dashed arrows show the excitation by incident light. Thin solid arrows indicate the vibrational relaxation in the S2 or internal conversion from the vibrational excited state of S2 to the 1nπ* state. Thick solid arrows show the relaxation pathways with experimentally determined lifetimes.</p><p>A Huang-Rhys factor, the 0-0 transition energy, the vibrational energy of the C=C stretching mode and a full width at half maximum (FWHM) determined by Franck-Condon analysis of the steady-state absorption spectra of C20Ind in various solvents</p><p>Solvent dependence of decay time constants determined by the fitting of transient absorption kinetics following excitation at 2.43 eV using multi-exponential functions of C20Ind. Plus and minus signs on the right-hand side of decay time constants indicate rise and decay phases, respectively.</p><p>The relative intensities, peak energy and full width at half maximum (FWHM) of Gaussian profiles of the excited-state absorption components and relative bleaching intensity of the steady-state absorption component determined by the convolution of the transient absorption spectra of C20Ind.</p><p>Excitation energy dependence of decay time constants determined by fitting of transient absorption kinetics of C20Ind using multi-exponential functions. Plus and minus signs on the right-hand side of decay time constants indicate rise and decay phases, respectively.</p>

PubMed Author Manuscript

Virtual Screening and Free Energy Estimation for Identifying Mycobacterium Tuberculosis Flavoenzyme DprE1 Inhibitors

In Mycobacterium tuberculosis (MTB), the cell wall synthesis flavoenzyme decaprenylphosphoryl-β-dribose 2'-epimerase (DprE1) plays a crucial role in host pathogenesis, virulence, lethality and survival under stress. The emergence of different variants of drug resistant MTB are one of the major threats worldwide which essentially requires more effective new drug molecules with no major side effects.Here, we used structure based virtual screening of bioactive molecules from ChEMBL database targeting DprE1, having bioactive 78,713 molecules known for anti-tuberculosis activity. An extensive molecular docking, binding affinity and pharmacokinetics profile filtering results in the selection four compounds, C5 (ChEMBL2441313), C6 (ChEMBL2338605), C8 (ChEMBL441373) and C10 (ChEMBL1607606) which may explore as potential drug candidates. The obtained results were validated with thirteen known DprE1 inhibitors. We further estimated the free-binding energy, solvation and entropy terms underlying the binding properties of DprE1-ligand interactions with the implication of MD simulation, MM-GBSA, MM-PBSA and MM-3D-RISM. Interestingly, we find that C6 shows highest ΔG values (-41.28±3.51, -22.36±3.17, -10.33±5.70 kcal mol -1 ) in MM-GBSA, MM-PBSA and MM-3D-RISM assay, respectively. Whereas, the minimum ΔG scores (-35.31±3.44, -13.67±2.65, -3.40±4.06 kcal mol -1 ) observed for CT319, the inhibitor co-crystallized with DprE1. Collectively, the results demonstrated that hit-molecules C5, C6, C8 and C10 having better free binding energy and molecular affinity as compared to CT319. Thus, we proposed that selected compounds may be explored as lead molecules in MTB therapy.

virtual_screening_and_free_energy_estimation_for_identifying_mycobacterium_tuberculosis_flavoenzyme_

Introduction<!>Protein preparation<!>Ligand preparation<!>Virtual Screening and Scoring<!>ADME studies<!>In-silico drug-likeness and toxicity prediction<!>Molecular dynamics (MD) simulation<!>Binding free energy calculation<!>Generalized Born Surface Area (GBSA)<!>Poisson Boltzmann Surface Area (PBSA)<!>Three-dimensional reference interaction site model with KH closure (3D-RISM-KH)<!>Results and Discussion<!>Virtual screening and docking analysis against DprE1<!>ADME property analysis against DprE1 receptor<!>In silico drug-likeness and toxicity predictions<!>Molecular Interactions<!>The spatial orientation of C10 (3-(3-hydroxypropyl)-7-(2-thiophen-3-ylethynyl)isochromen-1-one)<!>Conformational dynamics and stability of protein-ligand complex<!>Hydrogen bond analysis<!>Essential Dynamics<!>Binding free energy analysis<!>Conclusion

<p>Mycobacterium tuberculosis (MTB) is a slow growing and widely spread pathogen, survive in both, intra-cellular and extracellular systems of patients, and infection may result in chronic and complex disease state. During the treatment, it can go to latency which revert to exponential growth on the immune defiance conditions of hosts [1,2]. In recent years, WHO reports suggested that around 10.0 million (range, 9.0-11.1 million) individuals infected and 1.3 million (range, 1.2-1.4 million) people died from tuberculosis (TB) [1]. Moreover, the infection of MTB is one of the major causes of death worldwide, possessing the global health crisis, especially for the immunocompromised and HIV patients [3].</p><p>Although, the specific treatment may cure MTB, however, it requires multiple drug therapy for a longer period [1,3]. Furthermore, the development of multi-and extensively-drug-resistant (MDR-TB and XDR-TB) MTB strains are the big challenges to control TB infections [4,5]. In several conditions, it may turn into totally drug-resistant (TDR) tuberculosis which may worsen the condition of patients and therapy [2,6]. Thus, the potential drug candidates, having minimal or no side effects are highly sought in MTB therapy [1,2].</p><p>In recent years, several proteins involved in MTB survival and metabolism have been explored as potential drug targets and are progress in the drug development. During the evolution, mycobacteria have developed well-orchestrated and complex biosynthetic pathways to sustain a unique and thick cell wall which helps in maintaining the cellular integrity, survival under stress and dormancy, and eluding the host's immune systems. In MTB, the cell wall consists of the polymers of mycolyl-arabinogalactanpeptidoglycan, covalently connected with peptidoglycan and trehalose dimycolate that protects from stress, antibiotics and the hots immune systems [7]. The flavoenzyme decaprenylphosphoryl-β-d-ribose 2'-epimerase (DprE1) involve in the biosynthesis of cell wall, plays critical role in formation of peptidoglycan-arbinogalactan-mycolic acid complex (PAM) and arabinogalactan and lipoarabinomannan (LAM) which are the essential building blocks and play crucial role in survival and host pathogenesis, virulence, and lethality. DprE1 catalyses the first stage of epimerization reaction especially in the presence of FAD, it oxidizes C2' hydroxyl site of DPR to produce the keto intermediary decaprenyl-2'-keto-D-arabinose(DPX) and then DPA is formed by using decaprenyl-phosphoryl-D-2keto-erythro-pentose reductase (DprE2) and reduced form of nicotinamide adenine dinucleotide (NADH) as a cofactor [8][9][10]. Thus, the catalytic activity of DprE1 is one of the potential drug targets in the development of tuberculosis therapy [2,4,7]. Recently, the benzothiazinones (BTZs) derivatives have shown higher potency for inhibition of DprE1, and efficacy against XDR and MDR mycobacterium clinical isolates.</p><p>To improve the pharmacological properties of the compounds, chemical scaffold piperazine was added to BTZ. Further, the lead optimization of PBTZ derivatives results in the discovery of more potent compounds which are currently in clinical trials [5,8,11]. In this view, several structurally distinct chemical scaffolds are in drug screening as DprE1 inhibitors. Broadly, these inhibitors can be categorized as covalent or noncovalent, distinctly involved in interaction at the catalytic domain of DprE1 [8,11].</p><p>To elucidate the action and interaction of BTZs compounds, Batt el al., solved the X-ray crystal structure of DPrE1 in both, ligand free and bound form. He found that the structure of DprE1 consists of two functional domains, FAD binding domain and substrate binding domain. The co-factor was buried deeply in highly conserved FAD domain. The substrate binding extended for FAD, decorated largely with antiparallel β-strands (β10-16) and included disordered loops at surface which govern the wide and open active site. The nitroaromatic inhibitors (e.g., BTZ, VI-9376, nitroimidazole 377790) possesses nitro moiety which involved in covalent interaction at C387, whereas, the noncovalent inhibitors (e.g., TCA1, 1,4-azaindoles, pyrazolopyridones, 4-aminoquinolone piperidine amides, Ty38c) potentially inhibit the enzymatic function of DprE1 showed that hydrophobic, electrostatic, and van der Waals interactions are critical for the spatial stability of inhibitors at the active site of DprE1 [5,11]. Thus, the exploration of crystal structure of DprE1 has been largely facilitated the drug discovery efforts to tend the molecules effective against MDR and XDR strains [2,5,8,11].</p><p>Recent studies on the development of DprE1 inhibitors suggested a major contribution of molecular modelling, high throughput screening, docking, functional genomics and proteomics in paradigm of identifying novel chemical scaffolds as potential molecules for TB chemotherapy [5,8,11]. Although, molecular docking programs provide the description of protein-ligand interactions. However, a better understanding of protein-ligands interactions requires an accurate description of the spatial orientation of ligands at the active site of protein, conformational dynamics of protein and active sites residues, interaction energy and molecular stability [12][13][14]. In this view, MD simulation is an efficient and wellestablished computational method which mimics the flexible nature of bio-molecules, protein conformational changes, protein-ligand interactions, structural perturbation and provide more realistic picture with atomic details in reference to time [4,15,16]. Moreover, the free binding energy estimation, effect of solvation and thermodynamic integration is the central focus to understand the molecular interactions which can be well achieved by the implication of MM-GBSA, MM-PBSA and MM-3D-RISM using the trajectories obtained from MD simulation [12,[17][18][19].</p><p>In this context, we employed the structure based virtual screening for identification of promising chemical entities as DprE1 inhibitors from the ChEMBL database. We find that 78,713 small molecules at ChEMBL database suggested for the anti-mycobacterial activity. The three steps molecular docking and binding affinity estimation process lead to the selection of 10 hit-molecules. Similar procedures were applied on the selected 13 DprE1 inhibitors for the comparison of results with hit-molecules. Multiple MD simulations were performed on the DprE1 complex with hit-molecules and inhibitor (CT319) and the spatial stability of ligand molecules at active site of protein was estimated in terms of binding free energy using MM/PBSA/GBSA, and MM-3D-RISM [15,17,18]. The extensive evaluation of pharmacokinetic profile and drug-likeness properties analyses suggested that four chemical entities, compounds C5 (ChEMBL2441313), C6 (ChEMBL2338605), C8 (ChEMBL441373) and C10 (ChEMBL1607606) may be explored as potential lead molecules for the development of promising DprE1 inhibitors in MTB therapy.</p><!><p>The X-ray structure of DprE1 with inhibitor CT319 and cofactor FAD (PDB ID: 4FDO) was taken from the protein data bank (www.rcsb.org) [20]. The structure of DprE1 consist of two domains, the FAD binding domain comprised with α/β folds (residues 7-196, 413-461) and another domain, substrate binding includes extended conformation and antiparallel β-sheets (residues 197-412). In the crystal structure, the spatial orientation of FAD-binding domain and residues involved in interactions were highly conserved and critical, the cofactor is deeply buried, with the isoalloxazine at the interface to the substrate-binding domain [11]. And, the substrate-binding domain orientated towards the interface of flavin binding at centre. Thus, to prepare the protein files for molecular docking studies, the bound complex of protein with cofactor FAD was used. The other heteroatoms, co-crystallised inhibitor (CT319) and water molecules were removed. The structural regions lacking for low electron density were prepared, using Chimera tools [21]. Protein preparation wizard of Glide was used to assign hydrogen atoms and examine the structural correctness [22]. Finally, the optimized coordinates of DprE1</p><p>with FAD was used to carry out molecular docking and virtual screening against the selected antimycobacterial compounds from ChEMBL database.</p><!><p>Here, we used the anti-tuberculosis compounds taken from the publicly available chemical compounds database, ChEMBL. It consists of 2,101,843 compounds, out of these 78,713 compounds were bioactive molecules, having anti-tuberculosis activity downloaded from the ChEMBL database [23]. After sorting of compounds based on the repetitive entries, 30,789 ChEMBL compounds were found unique which were used for the ligand preparation. The SMILES (simplified molecular-input line-entry system) strings formats of compounds were converted to 3D SDF format, missing hydrogen atoms were added, and the structures were optimized using CORINA v2.64 software package [24]. The module, Ligprep of Schrodinger suite 2017-3 used to generate compounds with low energy 3D structures [25]. The ionization/ tautomeric states of the selected compounds were taken care of by Epik parameters. The compounds chirality was taken from the original state. All the conformations were minimized and produced at a maximum of 32 conformations per ligand using the OPLS-2005 force field at a pH 7+ 2 [26][27][28].</p><!><p>The structure based virtual screening of compounds against DprE1 was performed using Glide, Schrodinger, LLC [29,30]. The grid box define over the active site of DprE1, having outer box size X= 30, Y= 30, Z= 30 with grid center, X=40.1971, Y= 16.829, and Z= 9.172 [11]. The high throughput virtual screening involved three step filtering processes (i) selection of top ten percent ligand molecules using standard precision, (ii) then, docking of molecules by the extra precision mode of Glide (Glide XP) which allow the flexibility of ligands and (iii) the best-docked compounds were chosen using a Glide Emodel energy, Glide energy and Glide score function. The Glide Emodel includes a combination of Coulombic and van der Waals interaction energy, Glide score and strain energy of ligands which were used to selected lowest energy docked complexes on which the post-docking analyses were performed.</p><p>Further analysis involved the re-scoring of selected docked complex using X-score v1.2.1 [31] and the protein-ligand molecular interactions were examined using Ligplot [32] and Discovery Studio Visualizer (Accelrys, San Diego, CA, USA).</p><!><p>The bioavailability of selected lead molecules, ADME (adsorption, distribution, metabolism and excretion) properties were calculated using module Qikprop v5.7 of Schrodinger 2018-3. ADME descriptors includes, central nervous system (CNS), molecular weight (MW), prediction of octanol/water partition coefficient (QPlogPo/w), aqueous solubility (QPlogS), IC50 value for blockage of HERG K+ channels (QPlogHERG), gut blood barrier (QPPCaco), brain/blood partition coefficient (QPlogBB), binding to human serum albumin (QPlogKhsa), Lipinski's rule of five (RO5) and percentage of human oral absorption (% of Human oral absorption).</p><!><p>The molecular properties predictor tool, OSIRIS was used for the prediction of side effect risks of the hit-molecules, such as mutagenicity, tumorigenicity, irritant and reproductive effects. It also calculates the drug-relevant properties: cLogP, solubility (LogS), molecular weight (MW) and based on overall drug-score suggested the drug-likeness properties of molecules [33].</p><!><p>MD simulation was performed on the coordinates of DprE1 and DprE1-ligand complexes, using Amber16 biosimulation package. The force field ff14SB with TIP3P water model was used for the solvation of prepared systems. The charges, parameters and force field for cofactor (FAD) and ligands were defined by AM1-BCC charges and force field GAFF, using Antechamber tool. Here, six independent MD runs were carried out for the prepared systems, DprE1-FAD, DprE1-FAD-CT319, DprE1-FAD-ChEMBL1607606, DprE1-FAD-ChEMBL2338605, DprE1-FAD-ChEMBL2441313 and DprE1-FAD-ChEMBL441373 complexes. And, the systems were prepared using tleap tool of Amber16 with buffer distance (12 Å) in the octahedral simulation box. To neutralize the system 0.15 M counter ions (Na + and Cl -) were added [34]. Bonds involving hydrogens were treated with SHAKE algorithm and the long-range electrostatic forces were handled using Particle mesh Ewald summation. During the simulation, Berendson's barostat and Langevin thermostat were used to maintain the Pressure and temperature, respectively. The energy minimization processes involved two phases. First phase included 3000 minimization steps, which involved 2500 steps of steepest gradient and remaining 500 steps of conjugate gradient algorithm. The solute atoms were restrained (100 kcal mol -1 Å -2 ) and the only movements of water molecules and counter ions were allowed. The second phase of minimization included 5000 minimization steps (steepest gradient: 4500 steps and conjugate gradient: 500 used) without restraints on any atom. Minimization step followed by heating equilibration of system from 0 to 298 K with a time step of 1fs for 30 ps and consecutive equilibration run of 100 ps using time step of 2 fs with NPT ensemble. Using pmemd.cuda, the production run was performed on NPT ensemble for period of 100 ns and the time step was set to 2 fs. All files, trajectories, velocity and energy were saved at a gap of every 10 ps. The simulation trajectories were analysed using cpptraj tool available in Amber16.</p><!><p>Free energy change of a protein-ligand binding can be represented as follows:</p><p>ΔG=⟨GRL⟩-⟨GR⟩-⟨GL⟩ eq. ( 2)</p><p>where, ΔG, GRL, GR and GL represent the binding free energy of protein-ligand system, free energy of protein complexed with ligand, free energy of protein and free energy of ligand, respectively. The angular brackets represent ensemble average. Neglecting the entropy change of protein and ligand as a result of binding, equation (eq.) 2 can approximately be written as:</p><p>where, ΔE is the interaction energy change (gas-phase) upon ligand binding. ΔGSOLV is the solvation free energy change on ligand binding. Here, ΔE can be computed using molecular-mechanics force field and the second term, ΔGSOL can be estimated with the help of a proper solvation model.</p><p>Solvation models can be categorized in one of the two classes, implicit solvation model and explicit solvation model. Implicit solvation models consider the solvent molecules and dissolved salt ions as a mean field dielectric continuum. In contrary to implicit solvation models, explicit solvation models define solvent species at atomistic detail. Generalized Born Surface Area (GBSA), Poisson-Boltzmann Surface area (PBSA) are the two most common implicit solvation models used for solvation free energy calculation. Whereas, using the first principles 3D-RISM-KH (three-dimensional reference interaction site model with Kovalenko-Hirata) provides a 3D maps of solvation structure, thermodynamics and, more accurately predicting the parameters accounts for the ligands binding interactions and affinities.</p><!><p>In GBSA approach, the solvation free energy of a solute molecule, G SOL is calculated in two parts: polar or electrostatics (G SOL−GB ) and nonpolar or non-electrostatics</p><p>G SOL−GB is estimated using the following generalized Born expression:</p><p>where qi, qj represent charges on solute atoms i, j respectively and r ij is the distance between them, Ri represents effective Born radii (estimated using van der Waals radius and burial of atom),</p><p>and the summation runs over all pairs of atoms in the solute molecule. The screening effect produced by the monovalent salt ions is incorporated in eq. 4 through the Debye-Huckel screening parameter κ. The nonpolar or non-electrostatics contribution to solvation free energy (also known as cavitation term),G SURF is determined on the basis of solvent accessible surface area (SASA) of molecules.</p><!><p>In PBSA model, the solute is represented in atomic detail with molecular mechanics force field and the solvent molecules along with dissolved electrolytes is represented as a dielectric continuum. This approach considers the solute molecule as a dielectric object whose shape is determined by the atomic coordinates and their radii. Electric charges present on atoms of solute molecule produce electric field and in response the solvent also produce a reaction field. The electrostatic potentialϕ(r)at a point satisfies the Poisson-Boltzmann (PB) equation and can be computed by solving it:</p><p>Where, ε(r) is the dielectric constant, ϕ(r) is the electrostatic potential, ρ(r) is the solute charge density at position r, zi is the charge on ion i, ci is the bulk number density of ion i, k is the Boltzmann constant, and T is the absolute temperature; the summation is over all different ion types. when the solute does not carry a high charge, the second term in eq. 5 can be linearized and it results into linearized PB equation.</p><p>In Amber PB equation is solved numerically for solutes of arbitrary shape which gives electrostatic potentialϕ(r)at each point of the system. Once we know the electrostatic potential ϕ(r) at each point, we can calculate the polar solvation free energy (G SOL−PB ) of the solute by multiplying each solute charge q i by electrostatic potential ϕ(r i ) at that point. Here, the nonpolar part of solvation free energy G NPOL is also calculated through SASA of solute, as described in GBSA analyses.</p><!><p>3DRISM is a semi analytical theory based on statistical mechanical Ornstein-Zernike (OZ) equation which is a contrary to MM/GBSA and MM/PBSA, considers molecular structure of solvent and salt ions.</p><p>OZ equation splits the total correlation between two molecules into direct correlation between them and indirect correlation, which comes from other particles present in the system. In this theory, the molecular interactions are converted to sum of site-site interactions where atomic centres are taken as sites. The three-dimensional solvent distribution functions are obtained from the solution of following 3DRISM integral equation:</p><p>Where, h γ (𝐫) = g γ (r − 1) and c γ (𝐫) are the total and direct correlations of solvent in 3D and summation is taken over all interaction sites of all solvent species. The susceptibility function χ αγ (r′)for solvent was calculated using dielectrically consistent reference interaction site model (DRISM) theory and used as input to 3DRISM calculation. As eq. 6 involves two variables; hand c, therefore, we need another closure relation to solve it. Here, we have used Kovalenko-Hirata (KH) closure which is as following:</p><p>and g(𝐫) = 1 − βu γ (r) + h γ (r) − c γ (r) for g γ (r) > 1 (eq. 7)</p><p>The one-dimensional site-site solvent susceptibility of solvent is defined in two parts as follows:</p><p>Where, the intramolecular correlation function ω αγ (r) incorporates the molecular geometry of solvent and h αγ (r) is the total correlation between solvent sitesαand γ.</p><p>In Amber, eqs. ( 6) and ( 7) are solved numerically to obtain three-dimensional solvent distribution functions around a fixed solute geometry and solvation free energy of the solute is calculated using the following equation which is an extension of Singer-Chandler formula:</p><p>Where, Θ is the heaviside function and the summation runs over all the solvent sites i. The nonpolar part of GSOL-NPOL is calculated by assigning all solute charges zero. Further, the solvation free energy GSOL can also be decomposed into energetic and entropic components (∆GSOL-E and -∆GSOL-TS) using temperature derivatives.</p><!><p>The drug development process involves several expansive steps and complex strategies. Recent advancement in the computational modelling techniques, molecular docking, high-throughput virtual screening, pharmacokinetic profile (ADME), toxicity and bioavailability analyses of the molecules have been perceived as well-established techniques to accelerate the drug development processes [13,31,[35][36][37]. Further, the integration of MD simulation and estimation of free-binding energy provide an accuracy on the spatial fitting, interaction stability and binding affinity of ligands at the active site of protein [15,[17][18][19]38]. Herein, we systematically utilized the structure based virtual screening of compounds from ChEMBL database, having bioactive 78,713 molecules known for anti-tuberculosis activity (Figure 1).</p><p>The initial sorting of molecules leads to selection of 30,789 molecules which were subjected for molecular docking against the protein DprE1, the oxidoreductase enzyme involved in cell wall synthesis of MTB. The extensive ADME, toxicity and pharmacokinetic profile analyses were performed on hitmolecules which results in the selection of four ChEMBL compounds as potential lead molecules for DprE1 inhibitors. To improvise the molecular docking results, multiple MD simulations carried out on DprE1, DprE1-CT319 and DprE1--hit molecules (C5, C6, C8, C10) complexes. Typically, MD simulation deciphers the structural stability of protein-ligand interaction, conformational orientation, stability and molecular interactions of ligands at active site [11,22,39]. Moreover, the obtained MD trajectories utilized to calculate MM/PBSA, MM/GBSA and 3D-RISM-KH, which provide a robust estimation of free-binding energy, contacts and effect of solvent underlying the binding affinity of ligand molecules [12,17,18].</p><!><p>The structure based virtual screening was performed on 30,789 small molecules taken from ChEMBL database, having anti-tuberculosis biological activity. Glide based molecular docking involves various filtering steps for the high throughput virtual screening (HTVS) [29]. During the initial step, docking leads to the selection of 3,078 compounds, top scored 10 % compounds. These top scored 10 % compounds are subjected for standard precision (SP) docking which results in the selection of 307 compounds. The extra precision (XP) filtering is applied on another top scored 10 % (307 molecules) hit-molecules. Finally, the best scored top 10 hit-molecules are selected for the comparative studies with known DprE1 inhibitors and drug molecules taken from the recent literatures [5,8,11]. The 2D interactions of top 10 hit-molecules and 13 inhibitors are shown in Supplementary information S1A and S1B. The same procedures, SP followed by XP were applied for molecular docking of DprE1 inhibitors and X-score re-scoring method was applied to measure the binding affinity of molecules with DprE1 [31]. Glide docking scores and binding affinity (X-score) of top 10 hit-molecules and DprE1 inhibitors are summarized in Table 1 and 2, respectively. The DprE1 inhibitor, CT319 shows highest docking score -5.48 kcal mol -1 , whereas, BTZ-N3 shows the lowest docking score -1.82 kcal mol -1 . However, the X-score results show highest binding affinity of BTz043 (-9.62 kcal mol -1 ) with DprE1, whereas, the lowest binding affinity found with TBA-7371 (-7.87 kcal mol -1 ). Among the compounds taken from ChEMBL database, compound ChEMBL2323138 (C1) shows highest docking score of (-10.198 kcal mol -1 ) with DprE1and the minimum docking score -8.795 kcal mol -1 is obtained for ChEMBL1607606 (C10). Whereas, the X-score results show highest binding affinity -10.74 kcal mol -1 with ChEMBL2338605 (C6) and lowest affinity -8.60 kcal mol -1 for ChEMBL1607606 (C10).</p><!><p>Another filtering method involves the pharmacokinetic properties (ADME) analysis of hit-molecules [35,40,41]. Predicting the bioavailability, toxicity and safety of compounds is an important and integral component of drug designing process [13,35]. We employed QikProp v5.7 available with Schrodinger 2018-3 to analyse the ADME properties of compounds and compared with DprE1 inhibitors. Results</p><p>show that compounds, C5, C6, C8 and C10 having the CNS activity with the potential range of drug molecules -1 to 0 (Table 3). All hit-molecules having molecular weight <500. The optimal range value recommended for the lipophilicity (QPlogPo/w) of compound is between -2.0 -6.5. Result shows that all 10 molecules having QPlogPo/w <6.5, however, the lowest value (0.199) is observed for C4. Whereas, the higher QPlogPo/w value 5.108 is obtained for C5. The QPlogS (potential range -6.5 -0.5) defines the aqueous solubility of compounds which are observed within the favourable range for all 10 compounds. The recommended range for predicting IC50 value for blockage of hERG K + channel is QPlogHERG <-5 which is well satisfied by all compounds. The compounds having the predicted apparent Caco-2 cell permeability test (QPlogCaco) > 500 is recommended. Out of 10 molecules, only four compounds, C5, C6, C8 and C10 having a value range > 500. The recommended range for QPlogBB is -3.0 -1.2 which is observed favourable for all molecules. Similarly, it is observed that all hit-molecules obeyed the drug likeness properties RO5 (Lipinski's rule of five) and found within the recommended range for human oral absorption (PHOA). The ADME analysis of DprE1 inhibitors shows that all 13 compounds are lying within the recommended ranges for predicted ADME descriptors (Table 4).</p><!><p>The physiochemical properties, toxicity, tumorigenicity and mutagenesis risk of the compounds are investigated by OSIRIS Property Explorer [33] and compared with the DprE1 inhibitors (Table 5 and 6). Results show that out of 10 hit-molecules, 9 compounds are estimated as no risk for mutagenicity (MUT and tumorigenicity (TUMO), whereas, C2 shows higher risk for both MUT and TUMO test (Table 5). All 10 compounds show no risk for irritation (IR), however, reproductive development (REP) toxicity result shows high risk for C6, whereas, C9 shows medium risk for REP. The drug score (DS) is representing the combined score value of compounds solubility, polar surface area, toxicity, druglikeness and CLogP which define the overall sensitivity of drug molecules. Result shows highest DS value 0.79 for the hit-molecule, C4 and the least observed DS score (0.14) for C2. The four compounds, C5, C6, C8 and C10 which successfully cross the ADME test, show moderate DS score ranges 0.26 -0.42. The compound C8 shows higher DS score 0.42, C5, C8 and C10 predicted as no risk for toxicity parameters. Whereas, C6 predicted as high toxic risk for reproductive effect (REP), as the chemical scaffold contains ketone moiety. Table 6 shows the toxicity and drug-likeness parameters index of DprE1 inhibitors. The inhibitors, Ty38c and 4AQs predicted as higher risk for mutagenesis, whereas, VI-9376 shows medium risk and other inhibitors observed as no risk. All selected inhibitors having no risk for TUMO. Only, two inhibitors show medium risk for IR and all are predicted as no risk for REP. The predicted DS score of inhibitors ranges 0.22 -0.88. The inhibitor CT319, co-crystallized with DprE1 X-ray structure shows moderate range of DS score 0.37, however, no risk is observed for toxicity parameters.</p><!><p>The crystal structure of DprE1 consists of well separated, conversed FAD-domain and the substrate binding domain. The deeply buried FAD-domain is composed of an α/β fold, formed by the residues belonging to N-terminus (residues 7-196) and C-terminus (413-461). The substrate binding domain is extended from flavin at centre to surface, orchestrated by anti-parallel β-sheets (β10-16) and helices (α5, 9 and 10). The wide-open active site of DprE1 is governed by two loops at surface which facilitate the accessibility and flexibility of ligand binding. We find that the top 10 hit-molecules (C1-C10) taken from ChEMBL shared a common interaction with active site residues which is summarized in Table 1.</p><p>The 2D molecular interaction of hit-molecules C1-C10 and inhibitors at the active site of DprE1 are shown in Supplementary Figure S2 and S3. The active site amino acid residues involved in interactions with DprE1 inhibitors are tabulated in Table 2.</p><p>The co-crystallized structure of DprE1-CT319 shows that the inhibitor at the active site is largely stabilized by the hydrophobic interactions. Trifluoromethyl moiety of CT319 involved in hydrophobic interaction with Lys134 and Tyr314 and the nitro-benzene is stabilized with Lys317 and Val365. The nitro (NO2)-group shows H-bond with His312 and Lys418, and the phenylethyl moiety forms hydrophobic interaction with Tyr60, Trp230, Phe320, Leu363, Val365 and FAD present in the vicinity of catalytic domain. Furthermore, the structural studies of BTZs derivative inhibitors demonstrated that the hydrophobic amino acid residues at the active site, Trp60, Gly117, His132, Gly133, Lys134, Ser228, Phe231, Tyr314, Leu317, Phe320, Gln321, Trp323, Asn324, Gln334, Q336, Leu363, Val365, Phe366, Lys367, Phe369, Asn385, Ile386, Cys387, Asp389 and Lys418 are critical for the ligand recognition.</p><p>And, some inhibitors are covalently linked with C387. Molecular docking result shows that at the active site of DprE1, CT319 forms H-bind with Gln336, Asn385, Lys418, alkyl interactions with Tyr314, 𝜋alkyl interaction with Leu317, Leu363, Cys387 and the 𝜋-Sigma bond with Val365 (Figure 2). The trifluoromethyl moiety of CT319 interacting with Pro116, Gly133, Lys134 and Tyr314, whereas, the both benzene rings are stabilized with hydrophobic and van der Waals interaction interacted with residues: Tyr60, Gly117, His132, Lys134, Ser228, Phe320, Gly321, Lys367, Asp389 which is observed consistent with the co-crystal structure [8,11]. show 𝜋-alkyl interaction, Cys387 engaged in 𝜋-sulfur interaction and Lys418 and Val365 are involved in 𝜋-cation and 𝜋-Sigma interactions, respectively. And, the hydrophobic and van der Waals interactions of Gly117, His132, Gly133, Lys134, Ser228, Phe320, Gly321, Gln334, Gln336, Phe369, Asn385, Asp389 provided the additional stability to C5 at the active site of DprE1 (Supplementary Figure 4).</p><p>The molecular binding of C8 (1-cyclohexyl-5-oxo-N-(3-phenylphenyl)pyrrolidine-3-carboxamide) at the active site of shows H-bond interaction with Lys418, 𝜋-donor H-bonding with Tyr60, the alkyl and 𝜋-alkyl interactions with Lys134, Lys367 and Leu317, Leu363, Val365, respectively. And, the hydrophobic and van der Waals interactions with Pro116, Gly117, His132, Gly133, Ser228, Tyr314, Phe231, Phe320, Gly321, Trp323, Asn324, Gln334, Gln336, Phe366, Phe369, Asn385, Ile386, Asp389 at the active of DprE1 (Supplementary Figure 5).</p><!><p>shows H-bond interaction with Asn324, Cys387, the carbon hydrogen bond with Phe230 and 𝜋-alkyl interaction with Leu317, Leu363, Lys367. The thiophen moiety of C10 involve in 𝜋-sulfur interactions with His132, Phe369, whereas, Val365 and Lys418 show 𝜋-sigma interaction and 𝜋-cation interaction.</p><p>And, the ligand is stabilized by the hydrophobic and van der Waals interactions of residues: Tyr60, Trp66, Gly117, Gly133, Lys134, Gly321, Glu322, Trp323, Arg325, Gln336, Asn385, Val388, Asn389</p><p>which is shown in Supplementary Figure 6. These molecular docking results suggested that apart from the conventional H-bonding at the active site of DprE1, several other interactions, 𝜋-alkyl (Leu317, Lys367), 𝜋-sulfur (Cys387), 𝜋-Sigma and the van der Waals interactions of amino residues His132, Gly133, Lys134 and Asn385 are critically involved in interactions with lead-molecules which observed consist with the co-crystalized structure of DprE1-CT319. Thus, the molecular docking results, binding affinity scores and pharmacokinetic analysis of hit-molecules suggested that compounds, C5 (docking score -9.248 kcal/mol, X-score -9.66 kcal/mol), C6 (-9.211 kcal/mol, X-score -10.74 kcal/mol), C8 (-9.106 kcal/mol, X-score -10.64 kcal/mol) and C10 (docking score -8.795 kcal/mol, X-score -8.60 kcal/mol) may be explored as promising candidates for further lead optimization as DprE1 inhibitors.</p><!><p>The solvent environment around the protein influences the molecular interaction. Thus, the various interactions observed during the molecular docking may or may not exist during the simulation [15,42].</p><p>To examine the conformational stability, dynamics and structural integrity of DprE1 complex with novel hit-molecules, multiple MD simulations were performed in aqueous environment for the period of 100 ns, at 300 K. The conformational dynamics of DprE1 and DprE1-CT319 during the MD simulation used as a control to elucidate the structural stability of DprE1 complexed with ChEMBL compounds.</p><p>Trajectories obtained from the simulation were further used for the binding free energy estimation of molecules, using MM/GBSA, MM/PBSA and MM/3DRISM [15,17,19,43].</p><p>To determine the conformational stability of DprE1 with hit-molecules, we measured all atom Cα-RMSD of protein-ligand complexes and compared the results in reference of DprE1-CT319 complex (Figure 4). Results show that the structure of DprE1 remains stable with an average change in RMSD value 3.57±0.49 Å. The trajectory of DprE1 archives equilibrium at ~25 ns and a continuous stable equilibrium can be seen up to 100 ns of simulation. RMSD plot of DprE1 complex with inhibitor CT319 shows that trajectory achieves equilibrium at ~15 ns and the complex structure remains stable for the remaining period of simulation with change in RMSD 2.80±0.26 Å. The RMSD plot of DprE1-C5 shows a continuous increase in trajectory during initial 0-35 ns. The complex structure remains stable for the period of 35-55 ns and a small drift of 0.5 Å is observed at 60 ns. The RMSD trajectory during 60-100 ns suggested a stabilized structure of DprE1-C5 complex for the last 40 ns of simulation. The trajectory of DprE1-C6 achieves equilibrium in 0-15 ns and remains stable till the simulation finished at 100 ns.</p><p>We observed a consistent and overlapped RMSD plot of DprE1-C6 with DprE1-CT319 complex. The RMSD plot of DprE1-C8 shows initial perturbation in structure during 0-25 ns, however, reaches to equilibrium at ~30 ns, after that the conformational dynamics remains stable around RMSD 3.66±0.37</p><p>Å. Although the structure of DprE1-C10 achieves equilibrium earlier at ~10 ns and remains stable up to 70 ns, however, we observed structural adjustment with the drift of ~1 Å at 75 ns and the structural dynamics remains stable with an average change of RMSD value 3.18±0.47 Å.</p><p>To understand the spatial stability of ligand molecules at active site of DprE1, we also calculated the time evolution plot of distance of hit-molecules and inhibitor CT319 from the centre of binding pocket, as shown in Figure 5. We observed that the average distance of CT319, C6 and C8 remains quite stable suggesting that ligand is spatially well occupied at active site and stabilized with molecular interaction, during the simulation. The compound C5 shows continuous drop down in distance during 0-40 ns and it stabilized at distance ~4 Å which is seen up to 100 ns. The distance plot of C10 shows fluctuating behaviour which suggested the spatial adjustment at the binding pocket, thus, we observed a small drift of 1 Å in RMSD plot of DprE1-C10 complex.</p><p>The conformational order parameter, radius of gyration (Rg) represents structural compactness and integrity of a protein structure [44]. The Rg plot shows that all five complexes are stabilized around average Rg value ~21-23 Å, suggesting that all the ligand molecules were well occupied at the binding pocket of DprE1 during the simulation (Figure 6). Similar to RMSD results, we observed slightly higher Rg value 22.02±0.13 Å for the ligand unbound structure of DprE1. The complex of DprE1-C6 shows lowest Rg value 21.75±0.06 Å and highest 22.28±0.17 Å for DprE1-C5, whereas, the structure of DprE1-CT319 is stabilized around Rg score 22.10±0.07 Å. The marginal differences in Rg value of DprE, ligand bound, and unbound structures suggested the stable interaction of novel hits molecules at the active site of protein. These results provide a clear evidence that the selected ligands are well accommodated in binding pocket, having consistent interactions with active site residues as observed during the molecular docking.</p><p>We further investigated the conformational fluctuations and local dynamics through the calculation of average fluctuation of each amino acid residue of DprE1. The RMSF plot of all C α -atoms of DprE1 and docked complexes with ligands are shown in Figure 7. In this Figure 7, we can see the comparative results of DprE1, and each ligand bind complex with DprE1. Results show that the average fluctuation of residues are reduced on binding of ligands at the binding pocket of DprE1 which is suggesting a favourable molecular interaction. We observed that average fluctuation of C-terminal residues is increased on binding of CT319 and C5. The RMSF plots show higher mobility for residues 150-200, belonging to β8-10, which can be seen in CT319, C5 and C6. Whereas, DprE1-C8 and DprE1-C10 show lower average fluctuation in compared to FAD bound DprE1. However, the secondary structure analysis plots using DSSP suggested that no significant conformational changes are observed in the secondary structure, upon the binding of ligands during simulation which provide an elegance evidence of stable molecular interactions of ligands with DprE1 (Supplementary Figure 7).</p><!><p>The structure of protein is largely stabilized with the network of H-bond which plays critical role in the conformational adaptability, mobility and interaction with biomolecules. Apart from the structural stability, H-bond interactions are crucial in molecular recognition and protein-ligand interactions. Thus, we analysed H-bond interactions between protein and ligands, using cpptraj module of Amber with distance cut off 3.5 Å and angle cut off 135°. Results show that average two to four H-bond interactions are involved in DprE1 interactions with ligands (Figure 8). (Figure 8C). In C5, the H-bond formed between N atom and Asn324 is broken and new H-bonds are formed between this N-atom and H atoms at zeta position of Lys418 during the simulation (Figure 8B).</p><p>The O atom present in C5 was also found to form H-bonds with H atoms at epsilon position of His132,</p><!><p>We further performed essential dynamics (ED) analysis to understand the dynamics of protein-ligand complexes. ED analysis involves representation of collective motion of the most variable region of protein in terms of two principal components PC1 and PC2. The projection of each protein-ligand complex trajectory along with native protein onto two principal components PC1 and PC2 is shown in These observations thus support the idea of decrease in flexibility in the presence of DprE1 inhibitor, CT319 and compounds, C6 and C10. In the presence of CT319, DprE1 is restricted to small excursions slightly away from its initial conformation. The finding of a strong restriction in the size of the explored conformational space with only a minor reduction in RMSF which can be seen in Figure 7, indicates that local fluctuations take place but that collective motions have been compromised or more likely slowed down in the presence of CT319, C6 and C10 (Figure 9A, 9C and 9E) as compared to DprE1 complex with C5 and C8 (Figure 9B and 9D). Thus, the ED results along with H-bond interactions suggested that protein-ligands interactions remain consistent during the simulation, however, the most stable conformational dynamics is observed for CT319, C6 and C10.</p><!><p>The quantitative assessment of molecular binding interaction of DprE1 inhibitor CT319 and lead molecules C5, C6, C8 and C10 are estimated using three different methods for the molecular theory of solvation, MM/GBSA, MM/PBSA and MM/3DRISM-KH. MM/GBSA and MM/PBSA calculations are performed on the 5000 frames taken from the last 50 ns of the MD simulation [12,18,43]. Considering the large computational cost, 100 equally spaced frames taken from the last 50 ns of simulation are used for MM/3DRISM-KH analysis. The result of these three calculations are given in Furthermore, MM/3DRISM method also gives the energetic and entropic component of the solvation free energy (Table 9) which shows that it is the entropic part of solvation free energy not energetic, which favours the protein-ligand binding for all the ligands. Thus, the all three methods do not give the same order of binding free energy for the protein-ligands, but the comparison of predicted binding free energies provide an important clue to evaluate the relative stabilities and flexibilities of compounds at the active site of DprE1. In the table 7, results of MM-GBSA analyses show the higher value of combined ΔG (-41.28±3.51 kcal mol -1 ) for compound C6, whereas, lowest estimated ΔG for CT319 (-35.31±3.44 kcal mol -1 ). The compound C5 and C10 shared the almost similar binding free energies (ΔG ~36 kcal mol -1 ) and the slightly better binding energy predicted for C8 (-40.75±3.86 kcal mol -1 ).</p><p>Moreover, the binding energies estimated by MM-PBSA (Table 8) also suggested the major contribution of non-polar solvation energies in the molecular interaction of compounds at the active site of DprE1.</p><p>We find that the ΔG values for lead compounds ranges -16.08 --22.36 kcal mol -1 , whereas, the estimated ΔG -13.67±2.65 kcal mol -1 for CT319. In another analysis, the molecular theory solvation, MM-RISM-KH which yields the broader picture of molecular interactions on solvation structure and implication of thermodynamics from the first principles, accounts for solvent and biomolecules to describe the relative binding affinities [17]. Table 9 shows that binding interaction of compounds stayed stable as perceived from the molecular docking, however, results again reveal the highest ΔG value -10.33±5.70 kcal mol -1 for C6 and lowest for CT319 (ΔG value -3.40±4.06 kcal mol -1 ). Furthermore, the binding free energy approximation by all three methods suggested the larger contribution of van der Waals energies for ligands interactions and stability at the active site of DprE1. Collectively, the results demonstrated that all four hit-molecules C5, C6, C8 and C10 have better binding affinity with DprE1 as compared to inhibitor CT319 (Figure 10). Thus, the lead optimization of selected four compounds from ChEMBL chemical database may provide a new endeavour for the development of DprE1 inhibitors in MTB therapy.</p><!><p>In conclusion, we have explored structure based virtual screening for the identification of promising chemical entities as DprE1 inhibitors from ChEMBL database. Initial sorting of compounds results in the selection of 30,789 small molecules which are suggested for the anti-mycobacterial activity. The three steps molecular docking and binding affinity estimation processes lead to the selection of bioactive 10 hit-molecules. Similar procedures were applied on the selected 13 DprE1-inhibitors which were used to compare the results with hit-molecules. The extensive evaluation of pharmacokinetic profile and druglikeness properties analyses using ADME, toxicity and OSIRIS properties explorer suggested that four chemical entities, C5 (ChEMBL2441313), C6 (ChEMBL2338605), C8 (ChEMBL441373) and C10 (ChEMBL1607606) may be explored as potential candidates for the lead optimization as DprE1 inhibitors. To determine the conformational stability of hit-molecules at the active site of DprE1 in aqueous environment, multiple MD simulation were performed on the complex of DprE1 with lead molecules and inhibitor CT319. The binding free energy estimation using MM/PBSA, MM/GBSA and 3D-RISM-KH revealed that compounds C5, C6, C8 and C10 show better binding affinity as compared to DprE1 inhibitors. Thus, our comparative studies suggested that the selected compounds (C5, C6, C8</p><p>and C10) could be further investigated as novel lead molecules for the rational drug designing of DprE1inhibitors in MTB therapy.</p>

ChemRxiv

A Fast Approximation for Adaptive Wavelength Selection for Infrared Chemical Sensors †

Active mid-infrared spectroscopy with tunable lasers is a leading technology for standoff detection and identification of trace chemicals. Information-theoretic optimal selection of the laser wavelength offers the promise of increased detection confidence at lower abundances and with fewer wavelengths. Reducing the number of wavelengths required enables faster detections and lowers sensor power consumption while keeping the optical power under eye safety limits. This paper presents an approximation to the mutual information which operates ∼40 000× faster than traditional techniques, thereby making near-optimal real-time sensor control computationally feasible. Application of this technique to synthetic data suggests it can reduce the number of wavelengths needed by a factor of two relative to an evenly-spaced grid, with even higher gains for chemicals with weak signatures.

a_fast_approximation_for_adaptive_wavelength_selection_for_infrared_chemical_sensors_†

<!>Description of the Algorithm<!>Parameter Tuning

<p>Active mid-infrared (MIR) spectroscopy 1 is a popular technique for detection and identification of trace chemicals (both surface residues and vapors/aerosols) at distances of up to tens of meters. [2][3][4][5][6][7][8][9][10][11][12][13] The MIR reflectance spectrum can be obtained either by illuminating the target with a broadband source and dispersing the received light at the detector, or by scanning the wavelength 14 of a tunable, narrowband source such as a quantum cascade laser. 10,11 Use of a tunable source offers many advantages, including higher optical throughput (for a given level of illumination, such as dictated by eye safety), simpler detectors (no need for a Fourier-transform infrared or other complex spectrometer), and the ability to select only wavelengths which help discriminate between target chemicals (i.e., no photons need be emitted at uninformative wavelengths).</p><p>A block diagram of a laser-based chemical sensor is shown in Figure 1. The active IR Figure 1: Block diagram of an active MIR chemical sensor. An active IR sensor interrogates the target with a tunable laser and reports the reflectance spectrum to the detection and control algorithm. The detection algorithm compares the measured spectrum to a library of spectral signatures, and the control algorithm selects additional wavelengths to measure until the desired level of confidence is reached.</p><p>sensor comprises a tunable laser (which can be raster-scanned over the target to build up an image) and a broadband imager. The active sensor reports the reflectance spectrum, in the form of a hyperspectral image (HSI), to the detection and control algorithm. The detection algorithm [15][16][17][18] compares the measured spectrum to a library of spectral signatures.</p><p>If the results of the detection algorithm are inconclusive, the control algorithm selects further wavelengths to measure. This process repeats until the desired detection confidence is obtained.</p><p>Information theory is often used for feature selection in the machine learning community, 19 and has been applied to the optimization of various chemical sensors, [20][21][22] the characterization of spectral variability in hyperspectral images, 23 and the characterization of limits of detection in chemical sensors. 24 In particular, previous work has presented techniques for information-theoretic optimal wavelength selection (OWLS). 25 Specifically, given candidate wavelengths Ω, OWLS seeks</p><p>where R(Λ) is the reflectance at the wavelengths Λ, Y is the identity of the chemical, and</p><p>is the mutual information between R and Y . 26 The wavelengths selected with this scheme are optimal in the sense that maximizing I(R; Y ) minimizes the mis-classification rate. 27 The simplest way to use OWLS is a priori: the sequence of wavelengths which is expected to deliver the best accuracy is selected offline, before any data have been measured. It is desirable, however, to select the next wavelength(s) to measure adaptively based on the data which have been measured so far, as described in Algorithm 1.</p><p>For offline use, I(R; Y ) is estimated at each step (line 6) using the Kozachenko-Leonenko (KL) estimator. 28 Computing the first 10 wavelengths (from 500 candidates) to optimally discriminate between 67 chemicals takes six minutes on a typical laptop, and is therefore infeasible for real-time, adaptive wavelength selection. Instead, this paper presents a geometric approximation which exploits the structure of the signature model to deliver comparable results in under 10 ms: a 40 000× speedup.</p><!><p>In order to enable real-time adaptive OWLS, we need an algorithm which has the following properties:</p><p>Algorithm 1 Adaptive OWLS (Algorithm 2 of Ref. 25)</p><p>1: function AdaptiveOWLSDetection(Λ, Ω, α, q) Λ is the initial set of wavelengths Ω is the set of candidate wavelengths (Λ ∩ Ω = ∅) α is the required level of detection confidence q is the quantile of chemicals to retain at each step Ŷ ← {y | s y ≥ q(s)} Ŷ are the most likely chemicals 6:</p><p>return arg max s Return chemical with highest detection score</p><p>• Selects wavelengths which contain comparable mutual information to those selected by more rigorous means (such as the KL estimator).</p><p>• Can incrementally add new wavelengths to the existing set Λ.</p><p>• Can select wavelengths in less time than it would take to simply measure all of the candidate wavelengths. (Current systems can measure as fast as 1 ms per wavelength. 13 )</p><p>The geometric approximation is inspired by the structure of the signature model illustrated in Figure 2: the cluster corresponding to each chemical forms a line emanating from a single point (the reflectance of the bare substrate). For double pass absorption (such as from a thin film on a metallic surface), the linear relationship is exact in the absorbance domain (i.e., the logarithm of reflectance). For more general cases, this linear behavior is a reasonable approximation for low abundances (where the signal-to-noise ratio (SNR) is low and the most benefit can be gleaned from OWLS). Therefore, each chemical can be thought of as corresponding to a vector which begins at the point corresponding to bare substrate and ends at the point corresponding to a typical abundance. A possible way to pick informative wavelengths is then to find the set Λ which maximizes the squared Euclidean distance between the endpoints of all pairs of chemicals a and b:</p><p>where R a (λ) is the reflectance at wavelength λ for a typical concentration of chemical a. Simply picking wavelengths which maximize the sum of this quantity over all pairs of chemicals will not provide a good approximation to the wavelengths selected using an actual calculation of mutual information, however: given an initial set of wavelengths, the next wavelength should be the one which contributes the most to separating the chemical pairs which are not already well-separated. Therefore, when assessing the potential gain from adding a wavelength, the objective function should include some form of discounting to reduce the weight of gains which come from chemical pairs which are already well-separated. This suggests a function of the form</p><p>which has two parameters:</p><p>• δ ≥ 0 sets how strongly gains for chemical pairs which already have large separation are discounted. Setting δ = 0 ignores how well the current set Λ already separates any given pair of chemicals.</p><p>• p selects how the discounted gains for each chemical pair are aggregated. The extremes are p = 1 (sum of discounted gains) and p = ∞ (maximum of discounted gains), with intermediate values interpolating between the two behaviors.</p><p>The full procedure for geometric real-time optimal wavelength selection (GROWLS) is shown in Algorithm 2. In practice, this function replaces the computation of I(R; Y ) on line 6 of Algorithm 1.</p><!><p>In order for GROWLS to be effective, the parameters δ and p must be tuned so that it best approximates the behavior of the KL estimator. One way of selecting the parameters a priori is to minimize the sum of the difference in mutual information captured by the KL estimator and GROWLS:</p><p>where k = |Λ| is the number of wavelengths selected. Figure 3(a) shows ∆ as a function of δ and p for k = 50. For very low δ, too many wavelengths which only help easily-separable Algorithm 2 Geometric Real-Time Optimal Wavelength Selection (GROWLS)</p><p>1: function GROWLS(Ω, δ, p, k) Ω is the set of candidate wavelengths δ is the discount exponent p is the norm order k ≤ |Ω| is the number of wavelengths to select 2:</p><p>Λ ← {λ new } 4:</p><p>while |Λ| < k do 6:</p><p>return Λ chemical pairs are selected, which results in performance which is worse than using evenlyspaced wavenumbers. For all three values of p there is a broad minimum near δ = 1.6. There is only a weak dependence on p, but the minimum for p = 1 is slightly lower than for p = 2 and p = ∞. Therefore, the parameters p = 1, δ = 1.6 were used for the remainder of this work.</p><p>Figure 3(b) shows the trajectories of mutual information accumulation for evenly-spaced wavenumbers, the optimal wavelengths computed using the KL estimator, and the approximate wavelengths computed using GROWLS. The geometric approximation nearly matches the performance of the KL estimator, indicating that this scheme should deliver comparable detection performance to the full mutual information calculation. Figure 3(c) shows the spectral library of 67 chemicals together with the wavelengths selected by the three techniques.</p><p>Both the KL estimator and GROWLS pick wavelengths which are focused on informative parts of the spectrum, ignoring the uninformative regions that the evenly-spaced scheme samples. In order to illustrate the utility of adaptive OWLS using GROWLS ("adaptive GROWLS"), we generated noisy synthetic data corresponding to liquid films on metallic substrates and assessed the chemical identification accuracy. As a simplification to the industry-standard adaptive cosine estimator (ACE), 15 we used cosine similarity to match the simulated absorbance data a to the library signatures a y :</p><p>Figure 4(a) shows the contours corresponding to 90% chemical identification accuracy for evenly-spaced wavenumbers, a priori OWLS (i.e., wavelengths selected a priori using the KL estimator), and adaptive GROWLS. For this example, six wavelengths were added at each step and GROWLS used q = 0.8 (i.e., the top 20% of chemicals were retained at each iteration of Algorithm 1), p = 1, δ = 1.6. The two OWLS schemes deliver comparable performance to each other, and require roughly half the number of wavelengths as the evenly-spaced scheme for moderate abundances.</p><p>The time to reach a detection of a given confidence is given by</p><p>where t d is the time-to-detection, N b is the number of batches of N λ wavelengths which are measured, t m is the time to measure a single wavelength (∼1 ms), t c is the time to run the detection/identification algorithm to determine the detection confidence (and, for adaptive GROWLS, to rank the chemicals) after each batch (∼10 µs for cosine similarity with 67 candidates), and t s is the time to select the next N λ wavelengths (∼2 ms for GROWLS with Despite the additional computational overhead, adaptive GROWLS delivers comparable per-Figure 4: (a) Contours corresponding to 90% accuracy, averaged over all 67 chemicals in the library. Adaptive GROWLS delivers comparable average performance to a priori OWLS, and both OWLS schemes out-perform the evenly-spaced wavenumbers, thereby permitting successful detection at lower concentrations while using fewer wavelengths. (b) Time-todetection when 10 µg/cm 2 of the target is present, averaged over all 67 chemicals. Despite the additional computational overhead, adaptive GROWLS delivers accuracy above 90% within about 20 ms, comparable to the performance of a priori OWLS.</p><p>formance to a priori OWLS: both OWLS schemes reach 90% accuracy in less than half the time it takes evenly-spaced wavenumbers to reach this point.</p><p>The comparable performance of a priori OWLS and adaptive GROWLS is expected at moderate concentrations: we are adding six wavelengths at each step, but often a few dozen wavelengths are sufficient at moderate concentrations, so adaptive GROWLS does not have much of a chance to make a difference. As noted previously, however, adaptive GROWLS makes a substantial difference for chemicals with low absorbance. 25 This is illustrated by Figure 5, which shows the detection performance when the target is 10 µg/cm 2 of benzene.</p><p>Adaptive GROWLS obtains an accuracy of 90% after just 15 ms, compared to 30 ms for a priori OWLS and 85 ms for evenly-spaced wavenumbers: adaptive GROWLS is able to substantially reduce the time-to-detection in this case.</p><p>This paper has presented a geometric approximation to mutual information which runs approximately 40 000× faster than the standard KL estimator, thereby enabling real-time adaptive selection of wavelengths to optimize detection accuracy. It was shown that, for moderate abundances, this approach halves the number wavelengths needed to reach an identification accuracy of 90%. For weak absorbers, the gains are even greater: benzene can be identified 6× faster, even after accounting for the overhead to compute the next batch of wavelengths. These simulated results indicate that adaptive GROWLS can enable sensors with tunable sources to obtain more rapid detections. Furthermore, the approximation presented here is applicable to any situation where the classes have the general structure illustrated in Figure 2, enabling efficient feature selection in a wide variety of machine learning contexts.</p>

ChemRxiv

Aryl Phosphonate-Directed Ortho C\xe2\x80\x93H Borylation: Rapid Entry into Highly-Substituted Phosphoarenes

Phosphonate-directed ortho C\xe2\x80\x93H borylation of aromatic phosphonates is reported. Using simple starting materials and commercially accessible catalysts, this method provides steady access to ortho-phosphonate aryl boronic esters bearing pendant functionality and flexible substitution patterns. These products serve as flexible precursors for a variety of highly-substituted phosphoarenes, and in situ downstream functionalization of the products is described.

aryl_phosphonate-directed_ortho_c\xe2\x80\x93h_borylation:_rapid_entry_into_highly-substituted_phosp

<p>Aryl phosphonates are useful motifs in organic synthesis as their versatility allows for transformation into a number of functional groups, including phosphines, phosphoramidates, and phosphorus oxides.1–2 Additionally, some aryl phosphonates and related compounds have also shown interesting bioactivities,3–5 including in the clinical treatment of various cancers.6</p><p>Due to their value, several strategies have been reported to prepare aryl phosphonates. However, current methods to generate functionalized aryl phosphonates typically involve the construction of the C–P bond via pre-functionalized starting materials such as organolithium or Grignard reagents,7 (pseudo)halides,8–12 boronic acids,13–15 carboxylic acids,16 or hypervalent iodine compounds.17–18 These methods are often limited by the availability of the starting materials, and thus only aryl phosphonates with simple structures can be synthesized readily. More complex aryl phosphonates, such as the ones bearing ortho-substitutions, are often very challenging to access for this reason. One attractive route to the synthesis of highly-substituted phosphonates is via ortho-boryl aryl phosphonates, as these products are readily diversified and are able to be converted into a range of phosphoarenes. However, like the other methods mentioned above, current routes to these intermediates require multiple steps to access (Figure 1).19</p><p>Over the last couple decades, directed C–H borylation has emerged as a powerful tool to prepare ortho-substituted aryl boronic esters from simple starting materials.20–22 A variety of directing groups have been demonstrated in this field, including a number of carbonyls,23–27 Lewis basic nitrogens,23, 26–32 phenols,33 silanes,34–35 thianes,36–39 ethers,23, 26, 40 and halides.23 We envisioned phosphonate-directed ortho-borylation as an attractive platform for the preparation of a variety of ortho-functionalized aryl phosphonates (Figure 1). A general method to convert widely accessible simple aryl phosphonates into ortho-boryl phosphonates, when combined with the vast chemistry of aryl boronates, would allow for the facile preparation of a wide variety of ortho-functionalized aryl phosphonates and their downstream derivatives.</p><p>Previously, we have shown that benzyl phosphines can be used as directing groups for C–H borylation, which provides access to ortho-borylated benzylic phosphines.41–42 Later, Shi and Takaya independently reported C–H borylation conditions to prepare ortho-borylated phosphines via the use of an aryl phosphine as a directing group.43–44 However, these latter methods are limited to phosphines bearing a single type of aromatic group. Moreover, no reactivity was observed with substrates bearing pre-existing ortho-substitutions. Thus, there is a clear need to have an alternative way to prepare these ortho-borylated phosphoryl aromatic compounds that can overcome these limitations. In this paper, we describe the successful realization of that goal, and report conditions for the preparation of ortho-phosphonate aryl boronic esters via a C–H borylation strategy using phosphonate as a directing group. Using easily accessed starting materials and commercially available catalytic components, we show that the reaction enjoys broad functional group tolerance and is highly flexible in regard to substitution patterns.</p><p>Our studies began with the borylation of commercially available diethyl phenylphosphonate, and the preliminary screen of iridium catalysts revealed (COD)Ir(acac) as the optimal iridium source (see Supporting Information). The use of [(COD)2Ir]BF4 provided comparable results, but the selection of (COD)Ir(acac) is advantageous because it is bench stable and less expensive. We then examined the effects of ligand with the model substrate 1, an aryl phosphonate substituted at the ortho position (Table 1). The ortho-borylation of 1 would produce the highly-substituted phosphonate 2, whose substitution pattern would be difficult to access using other methods.</p><p>Various ligands were examined in conjunction with (COD)Ir(acac) as catalyst. We began by investigating conditions without added external ligand, such as those that had shown success in our previous work on benzyl phosphine-directed C–H borylation.41 However, with this catalyst system we did not observe any ortho-borylation; instead, the borylation took place at the meta position to give 3 (Table 1, entry 1).</p><p>Moving forward, very little reactivity was observed with 2-picolylamine as ligand, the ligand used in our benzyl amine-directed C–H borylation (entry 2).30 Although we found that the borylation was more efficient with bipyridine ligands, such as dtbpy and tmphen that have often been used in aryl borylation, there was still no formation of the desired ortho-borylated product (entry 3–4). The selective formation of the meta-borylation isomer is in itself notable. This result suggests that the primary selectivity is steric in nature, but the fact that borylation para to the phosphonate is not observed also suggests a strong electronic differentiation between the methyl and phosphonate groups.</p><p>Excitingly, however, when the phosphine/silane-based bidentate ligand previously reported by Smith, Maleczka, and Krska was used, we observed 31% yield of 2 (entry 5).26 Despite the encouraging result using this ligand, we continued to explore other options as we recognized that use of a commercially available ligand would make the method more broadly useful. Therefore, we explored simple phosphine ligands as an alternative. Although PPh3 did not provide any of the ortho-borylation product (entry 6), with more electron-deficient triaryl phosphines we started to observe the desired ortho selectivity (entry 7–8). Following this trend, we finally identified the use of P[3,5-(CF3)2-C6H3]3 in this reaction, which has the optimal balance of electronic and steric properties, affording 2 in 84% yield (entry 9).24,45</p><p>With optimized reaction conditions developed, the scope of this reaction was then investigated (Scheme 1). The borylated phosphonate 2 was isolated in 78% yield on 1 mmol scale. During purification, we observed some loss of product due to the unstable nature of boronic esters, but such an effect can be minimized by running the reaction at larger scales. For example, scaling the same reaction up to gram scale provided 87% isolated yield of 2.</p><p>Substrates without ortho-substitution next to the phosphonate require somewhat more stringent conditions, including use of a higher catalyst loading, use of iridium pre-catalyst with a less coordinating counter-ion [(COD)2Ir]BF4, and elevated reaction temperature (4). Under these conditions, however, we observed a significant amount of bis-borylation at both ortho positions.46 To avoid the bis-borylation, we hypothesized that the use of larger directing group could restrict the conformational rotation of the phosphonate group in the mono-borylated product and prevent the second C–H activation at the other ortho position. This idea is supported by the use of diisopropyl phosphonate, only mono-borylated product is produced (5).</p><p>In contrast, substrates bearing varied substitutions at the pre-existing ortho position participate in this reaction under standard conditions and result in borylated products efficiently (6-13), including ethers (6, 9), trifluoromethyl (7), aniline (8), and halogens (10-13). With a bulky substitution at the ortho position, such as bromine, we found that the reaction requires higher catalyst loading and temperature to proceed (11), presumably due to the limitation of conformational rotation of phosphonate by steric congestion. We found, however, that simply switching to the smaller dimethyl phosphonate group overcomes this limitation and higher yield is observed under milder conditions (12). Phosphonates with aromatic substitutions were also easily borylated under the standard conditions (13); however, product instability precluded isolation.</p><p>With meta-substituted aryl phosphonates, the borylation only occurs at the less sterically hindered site (14). These reactions are most efficient with weakly donating or electron-withdrawing groups at the meta position (15). With a strong electron-donating group at the meta position, yield of the borylated product is lower (16).47 However, these electronic effects appear to be very sensitive. Simple inclusion of a weak electron-withdrawing group in the substrate mitigates the effects of the meta donor group (17-18).</p><p>Aryl phosphonates bearing di-substitutions at other positions can also participate in this reaction (19-20), and even highly-substituted substrates could be converted to the borylated products with good efficiencies (21-22). As with other directed C–H borylation reactions, this reaction is sensitive to steric effects. We did not observe reactivity with alkyl substitution next to the reactive C–H bond (23). However, with a smaller substitution, such as fluorine, high yields of the desired product are observed (24). A naphthyl substrate also works well under these reaction conditions and can be borylated in 90% isolated yield (25). One limitation that we have found with this reaction is that Lewis basic heterocycles, such as pyridine, were not tolerated and resulted in only minimal conversion (26). However, less-basic heterocycles are tolerated (27-28).</p><p>As one of the most versatile functional groups in organic synthesis, boronic esters can be readily converted into a range of other structures. For example, we were rapidly able to prepare aryl phosphonates containing ortho phenols (29),48 chlorides (30),49 bromides (31),49 and nitriles (32)50 by applying known methods from the literature (Scheme 2). It is noteworthy that the purification of the boronic esters is not required prior to many downstream functionalizations. This is particularly important because the purification of some borylated products can cause significant material loss due to their instabilities.</p><p>In addition, aryl boronic esters are reliable precursors for substituted biaryl compounds. Previously, Westcott and co-workers reported conditions for Suzuki coupling to transform ortho-boryl aryl phosphonates into biarylphosphonate substrates, such as 33.19 The further functionalization of 33 led to the synthesis of a dialkyl biarylphosphine ligand 34, showing the utility of ortho-borylated aryl phosphonate products in ligand synthesis (Scheme 3).51</p><p>Finally, we wished to compare the relative directing group ability of phosphonates to esters (Scheme 4).24 Under the standard and related conditions (see Supporting Information), borylation ortho to the ester (36) is preferred and mono-borylation ortho to the phosphonate was not observed. Interestingly, however, subsequent borylation at that site appears to be competitive (37).</p><p>In conclusion, we developed conditions for C–H borylation using phosphonate as a directing group, which allows the preparation of ortho-phosphonate aryl boronic esters from simple starting materials. This reaction uses commercially available catalytic components and has a broad functional group tolerance. Using this method, highly-substituted aryl phosphonates can be prepared, which is a known synthetic challenge in this field.</p>

PubMed Author Manuscript

Trace determination of lenalidomide in plasma by non-extractive HPLC procedures with fluorescence detection after pre-column derivatization with fluorescamine

BackgroundLenalidomide (LND) is a new potent drug used for treatment of multiple myeloma. For its pharmacokinetic studies and therapeutic monitoring, a proper analytical method was required.ResultsIn this study, a non extractive and simple pre-column derivatization procedures have been proposed, for the for trace determination of lenalidomide (LND) in human plasma by HPLC with fluorescence detection. Plasma samples were treated with acetonitrile for protein precipitation then treated with copper acetate to form stable complexes with the biogenic amines and mask their interference with the derivatization reaction of LND. Treated plasma samples containing LND was derivatized with fluorescamine (FLC) in aqueous media at ambient temperature. Separation of the derivatized LND was performed on Hypersil BDS C18 column (250 × 4.6 mm, 5 μm particle size) using a mobile phase consisting of phosphate buffer (pH 4):methanol: tetrahydrofuran (70:10:20, v/v) at a flow rate of 1.0 mL/min. The derivatized samples were monitored at an emission wavelength of 495 nm after excitation at a wavelength of 382 nm. Under the optimum chromatographic conditions, a linear relationship with good correlation coefficient (r = 0.9997, n = 9) was found between the peak area and LND concentrations in the range of 2–100 ng/mL. The limits of detection and quantitation were 0.8 and 2.30 ng/mL, respectively. The intra- and inter-assay precisions were satisfactory and the accuracy of the method was proved. The recovery of LND from the spiked human plasma was 99.30 ± 2.88.ConclusionsThe proposed method had high throughput as the analysis involved simple sample pre-treatment procedure and a relatively short run-time (< 15 min). The results demonstrated that the method would have a great value when it is applied in the therapeutic monitoring and pharmacokinetic studies for LND.

trace_determination_of_lenalidomide_in_plasma_by_non-extractive_hplc_procedures_with_fluorescence_de

Background<!>Experimental<!>Chemicals and materials<!>Lenalidomide (LND) standard solution<!>Fluorescamine solution<!>Phosphate buffer solution<!>Chromatographic system<!>General procedure and construction of the calibration curve<!>Design and strategy for assay development<!>Optimization of reaction variables<!><!>Optimization of reaction variables<!>Method development<!>Selectivity, linearity, limit of detection and limit of quantitation<!><!>Selectivity, linearity, limit of detection and limit of quantitation<!><!>Accuracy and precision<!><!>Robustness and ruggedness<!>Conclusions<!>Abbreviations<!>Competing interests<!>Authors’ contributions

<p>Multiple myeloma (MM) is a B-cell malignancy of the plasma cell and represents the second most common haematological malignancy (about 10%), with non-Hodgkin's lymphoma being the most common. It is estimated that approximately 21,500 new cases of multiple myeloma are diagnosed per annum with approximately 16,000 deaths from the disease annually within the European Union [1]. Multiple myeloma is characterized by an asymptomatic or subclinical phase before diagnosis (possibly for several years), a chronic phase lasting several years and an aggressive terminal phase. Multiple myeloma is primarily a disease of the elderly, with a median age at diagnosis of 68 years [2]. The disease leads to progressive morbidity and eventual mortality by lowering resistance to infection and causing significant skeletal destruction (with bone pain, pathological fractures, and hypercalcaemia, anaemia, renal failure [3], and, less commonly, neurological complications and hyperviscosity. From the time of diagnosis, the survival without treatment is between 6 to 12 months and extends to 3 years with chemotherapy. Approximately 25% of patients survive 5 years or longer, with fewer than 5% surviving longer than 10 years. MM is characterized with the production of a homogeneous immunoglobulin fraction, called myeloma protein, by the malignant plasma cells [4]. The classical triad of symptoms is plasmacytosis (> 30% of plasma cells in the bone marrow), myeloma protein either in the urine or blood, and lytic bone lesions [4,5].</p><p>In the 1990s, thalidomide (Thalomid®, Celgene Corporation) was used empirically in treatment of MM based on its antiangiogenic activity and clinical activity in refractory or relapsed myeloma [6]. However, thalidomide has significant and dose-limiting side effects such as sleepiness, constipation, neuropathy and teratogenicity [7]. These toxic effects promoted the search for more potent but less toxic thalidomide derivatives [8]. Lenalidomide (LND) is a potent novel thalidomide analog which demonstrated remarkable clinical activity against myeloma cells [8-13], via a multiple-pathways mechanism [7,8,14-19].</p><p>LND has a more improved side effects profile than its parent compound thalidomide, nevertheless, it causes some dose-dependent side effects such as thrombocytopenia, venous thromboembolism, and myelosuppression [20,21]. These side effects can be managed by combination therapy and/or careful dose adjustment [22,23]. Because LND is primarily excreted via kidneys, patients with renal insufficiency or failure must be dose adjusted to prevent the exacerbation of its myelosuppressive effects [10,18]. Because of the structural relation of LND to thalidomide, a teratogenic effect can not ruled out, thus effective contraception must be used by female patients [24,25]. Furthermore, studies showed large inter-individual pharmacokinetic variability with concentration–toxicity relationship [26]. For these reasons, a risk management, monitoring blood counts, and therapeutic drug monitoring are required to achieve the highest therapeutic benefits of LND and prevent its fatal complications [9,27,28]. Nevertheless, the therapeutic profile of LND is anticipated to encourage the development of new pharmaceutical preparations for LND. As a consequence, there is an increasing demand for proper analytical technologies for determination of pharmacokinetic parameters in bioequivalence studies for LND, as well as in its therapeutic monitoring.</p><p>Extensive literature survey showed that there were only two methods for the determination of LND in plasma [29,30]. These two methods involved liquid chromatography-coupled with mass spectrometric detectors (LC-MS). These two methods offered adequate sensitivities; however they employed the expensive mass detectors that are not available in most laboratories, and involved laborious liquid-liquid sample extraction procedures that negatively affected the accuracy of the results, and limit the throughput of the procedures in screening of large number of specimens. Accordingly, the development of a new alternative analytical method for the determination of LND in plasma with adequate sensitivity, improved simplicity, lower cost, and higher throughput is urgently needed.</p><p>Fluorescence-based HPLC has been used as a sensitive and less costly alternative approach to LC-MS. For these reasons, the present research proposal was directed towards the development and validation of a new simple and sensitive HPLC method with fluorescence detection for the determination of LND in plasma samples. The method involved a very simple non-extractive isolation of LND from plasma samples using protein precipitation with acetonitrile followed by masking biogenic amines by treatment with copper acetate, and derivatization with fluorescamine (FLC). The method was successfully applied to determination of LND in spiked human plasma samples.</p><!><p>The experimental research that is reported in this manuscript did not get any approval of ethics committee as the research has not been carried out on humans or animals.</p><!><p>Lenalidomide (LND), Free Base (3-(4' aminoisoindoline-1'-one)-1-piperidine-2, 6-dione) was purchased from LC Laboratories (Woburn, MA, USA), 165 New Boston Street Woburn, MA 01801, USA) and used as received. Fluorescamine was purchased from Sigma Chemical Co. (St. Louis, MO, USA). Copper acetate AR was obtained from BDH, Poole, UK. Human plasma samples were collected from normal healthy volunteers at King Khaled University Hospital (Riyadh, Saudi Arabia), and they were kept frozen at −20°C until the time of analysis. Acetonitrile, tetrahydrofuran and all the other solvents were of HPLC grade (Merck, Darmstadt, Germany). Water was Millipore filtered. All other materials were of analytical grade.</p><!><p>An accurately weighed amount of LND was quantitatively transferred into a calibrated volumetric flask, dissolved in methanol and completed to volume with the same solvent to produce a stock solution of 1 mg/mL and kept in the refrigerator. On the day of analysis, the stock solution was further diluted stepwise with water to obtain a working standard solution containing 1.0 μg/mL.</p><!><p>An accurately weighed amount (5 mg) of Fluorescamine was transferred into a 10-mL volumetric flask, dissolved in acetonitrile and completed to volume with the same solvent to produce a stock solution of 0.05% (w/v). The solution was freshly prepared and kept at -20°C pro-tected from light to be used within seven days.</p><!><p>Weighed amounts of di-sodium hydrogen orthophosphate (5.04 g) and potassium di-hydrogen orthophosphate (3.01 g) were dissolved in about 700 mL distilled water. The pH of the solution was adjusted to 4.0±0.1 with glacial acetic acid using a calibrated pH-meter (Microprocessor pH meter BT-500, Boeco, Germany).</p><!><p>HPLC apparatus consisted of a Shimadzu system (Shimadzu Corporation, Kyoto, Japan) equipped with two solvent delivery systems (LC-20 AD VP) with FCV-12AH high pressure flow channel changeover valve, SIL-20A auto-sampler, CTO-10A column oven, SPD-10A UV-visible detector,, RF-10A XL fluorescence detector, and SCL-10A vp system controller. The chromatographic separations were performed on an analytical column Hypersil BDS C18 (250 mm length × 4.6 mm i.d., 5 μm particle diameter) manufactured by Hypersil, ThermoQuest Corporation (England). The column temperature was kept constant at 25 ± 2°C. Separations were performed in isocratic mode. The mobile phase used for separation consisted of phosphate buffer (pH 4): methanol: tetrahydrofuran (70: 10: 20, v/v) pumped at flow rate of 1.0 mL/min. The mobile phase was filtered by a Millipore vacuum filtration system equipped with a 0.45 μm pore size filter, degassed by ultrasonication, and further by bubbling with helium gas. The samples (50 μL each) were injected by the aid of the auto-sampler. The fluorescence detector was set at 382 nm as an excitation wavelength and 495 nm as an emission wavelength. The system control and data acquisition were performed by Shimadzu CLASS-VP software, version 5.032 (Shimadzu Corporation, Kyoto, Japan). The relation between the area of LND peak and its concentration was used as the basis for the quantification. Alternatively, the regression equation was derived.</p><!><p>Accurately measured aliquots of LND working stock solution (1.0 μg/mL) ranging from 20–400 μL were transferred into eight separate reaction vials each containing 500 μl plasma and 100 μl of a 0.2% solution of copper acetate. The volumes in all vials were adjusted, as necessary, to 1.0 mL with water. The vials were heated in a boiling water-bath for 15 min. and cooled. 1.0 mL of acetonitrile was added to each vial and centrifuged for 10 min at 10000 rpm. Portions of 500 μL of the supernatant solutions were transferred into each of a set of eight vials followed by 300 μL of water. 200 μL of FLC solution (0.05% w/v) was added to all vials. This resulted in a series of LND standard solutions covering the working range of 5–100 ng/mL in the final reaction mixture. The reaction in each vial was allowed to proceed at room temperature for 5 min. before injecting 50 μL into the HPLC system. Peak area values of the reaction product between LND and FLC obtained at retention time of around 11.6 min. were plotted versus the LND concentration. A reagent blank treated in the same manner but using water instead of the LND working stock solution was prepared. Also another blank was prepared omitting the addition of copper acetate.</p><!><p>LND contains a weakly absorbing chromphore in its molecule. This weak absorptivity does not facilitate the adequate sensitivity for determination of LND levels in plasma without pre-concentration of the samples, particularly at the initial and elimination pharmacokinetic phases. On the other hand, LND has no native fluorescence and therefore a pre-colmn derivatization procedure was necessary. Fluorescamine (4′-phenylspiro[2-benzofuran-3,2′-furan]-1,3′-dione) is a fluorogenic reagent which reacts almost instantaneously with a wide variety of nucleophiles including primary amines, even at very low concentrations, forming fluorescent pyrrolinone moieties. The reaction is almost instantaneous at room temperature in aqueous media. The products are highly fluorescent, whereas the reagent and its degradation products are nonfluorescent [31]. Fluorescamine does not react with secondary amines and forms a non-fluorescent adduct upon binding to free NH3. LND contains a primary aromatic amino group which makes it a good candidate readily reacting with fluorescamine giving a highly fluorescent adduct.</p><!><p>From our previous study of the reaction between LND and FLC [32], we found that the reaction was dependent on FLC concentration, pH of the reaction medium, the nature of the diluting solvent and the duration of the reaction time. The investigations revealed that the best conditions for the reaction were found by maintaining a final concentration of 0.0025% w/v of FLC, a neutral pH using simple water as diluting solvent and leaving the reaction to proceed for five minutes at room temperature. Regarding the stoichiometry of the reaction and by analogy to our study mentioned above, the reaction is proposed to proceed as shown in Figure 1.</p><!><p>Scheme for the reaction pathway between LND and FLC.</p><!><p>The excitation and emission spectra of the LND-FLC derivative formed under the optimized conditions were investigated using RF-5301 PC spectrofluorimeter (Shimadzu, Kyoto, Japan). The reaction product was found to be fluorescent showing the highest fluorescence intensity at λex of 382 nm and λem of 495 nm. The present study was devoted to adopt the above reaction in the developing a sensitive HPLC method with fluorescence detection for determination of LND in plasma.</p><!><p>Plasma is highly rich in nitrogen containing compounds such as amino acids. Consequently, interference from such endogenous amino acids is obviously expected. In our present study, this interference was eliminated through a selective complex formation between those α-amino acids and copper ions furnished by addition of copper acetate [33], a modification from the method which describes the use copper hydroxide as the source of copper ions [34]. Copper acetate did not interfere with the determination of LND based on the following facts: (1) copper forms a stable complex with α-amino acids via binding of two amino acids by an amino nityrogen and a carboxylic oxygen (i.e. NNOO coordination bonding); (2) The presence of carboxylic group in α position to the primary amino group is essential for the formation of the complex; (3) LND contains a primary amino group however it does contain the carboxylic oxygen to form the copper complex. For these reasons, copper acetate did not interfere with the determination of LND in the method described herein. Acetonitrile is added to precipitate the large molecule plasma proteins which are then separated by centrifugation. The centrifugation also helps separation of the amino acid-copper complexes. The supernatant layers which contain the LND were transferred into the set of vials ready for the derivatization procedure as explained above.</p><!><p>The selectivity of the method was evaluated by carrying out blank experiments in the mobile phase to identify the reagent peaks and the peaks due to the derivatized plasma components. Typical chromatograms obtained from FLC-derivatized blank plasma before and after treatment with copper acetate are given in Figure 2A and Figure 2B, respectively. The chromatogram of the derivatization product between LND-spiked plasma (50 ng/mL) after treatment with copper acetate and FLC reagent is given in Figure 2C. The effect of treatment of the plasma with copper acetate before derivatization with FLC is clearly observed in from the chromatograms. The chromatogram given in Figure 2A shows how great is the interference from the endogenous amines if the plasma is not treated with copper acetate before derivatization procedure. Very high intensity peaks cover almost the whole range including the area where the derivatized LND peak appears (~ 11.6 min). On the other hand, the chromatogram of the plank plasma derivatized with FLC after pretreatment with copper acetate (Figure 2B) is absolutely free from any interference around the retention time of LND. These results demonstrated the great importance of treating the plasma samples with copper acetate before derivatization with FLC.</p><!><p>Chromatogram obtained from FLC-derivatized blank plasma before treatment with copper acetate (A), after treatment with copper acetate (B), derivatization product of LND-spiked plasma (50 ng mL−1) pretreated with copper acetate (C). The peak of retention time 11.6 min corresponds to LND-FLC fluorescent product.</p><!><p>The Chromatographic parameters and performance data of the proposed HPLC method for determination of LND in plasma are presented in Table 1. Under the optimized conditions, a linear relationship with good correlation coefficient (r = 0.9993 ± 0.00027, n = 6) was found between the peak area of LND-FLC complex (Y) versus LND concentration (X) in the range of 5–100 ng/mL. The experiments were performed using eight-point standard series. The mean regression equation of the calibration curve obtained was Y = 1.27 × 105 + 3.11 × 105 X. The % RSD value for the slopes of the calibration curves was 1.89% (n = 6). The limit of detection (LOD) and limit of quantitation (LOQ) were calculated according to the ICH guidelines for validation of analytical procedure based on the standard deviation of the response and the slope of the calibration curve [35] using the formula: LOD or LOQ = κσ/S, where κ = 3.3 for LOD and 10 for LOQ, σ is the standard deviation of the response, and S is the slope of the calibration curve. Calculations on 6 replicate experimental injections, the LOD and LOQ were 0.8 and 2.3 ng/mL respectively, and the relative standard deviations (RSD) did not exceed 2%. The LOD achieved in this method was lower than that achieved in our previous method involving the same derivatization procedure [32]. This is attributed due to the difference in the detection systems between the conventional fluorimeter (in the previous study) and the fluorescence HPLC detector in the method described herein.</p><!><p>Chromatographic parameters and performance data of the proposed HPLC method for determination of LND in plasma</p><!><p>The accuracy and precision of the proposed method was determined by intra-day and inter-day replicate analysis of plasma spiked with different concentrations of LND covering the working linear range. The inter-day assays were carried out on five different days at the same concentration levels for spiked plasma samples. The recovery values from the intra-day analysis were 97.93 -103.65% with a mean value of 100.85 ± 2.07 whereas the values for the inter-day analysis were 97.85 – 104.60 with a mean value of 100.91 ± 1.87 (Table 2), indicating the accuracy and precision of the proposed method.</p><!><p>Intra-assay and inter-assay precision and accuracy for determination of LND in spiked human plasma</p><p>a : Average of three replicates.</p><p>b : Average of four replicates.</p><p>* : %RSD not exceeding 2%.</p><!><p>The robustness of the method was evaluated by making minor changes on the parameters affecting the reaction between the analyte and the reagent in addition to the chromatographic parameters. In order to measure the extent, the most critical parameters were interchanged while keeping the other parameters unchanged. The chromatographic parameters were interchanged within the range of 1-10% of the optimum recommended conditions. The parameters involved were: the pH of the phosphate buffer used in the mobile phase, the ratio between the components of the mobile phase, pH of the derivatization reaction mixture, and column temperature. The chromatographic profile including: capacity factor (k'), retention time (Rt), peak asymmetry, resolution and column efficiency were calculated and compared with those of the system suitability (Table 1). The results revealed that the method was robust for these small changes in the parameters. However, increasing the pH value above 7.3 resulted in marked decrease in the detector signal. The increase in the column temperature generally decreases the k' values, and the column temperature has to be maintained at 25±2°C.</p><p>The ruggedness of the method was evaluated by applying the recommended analytical procedures on the same HPLC system on different days on the analysis of series of LND-spiked plasma samples. The values of the capacity factor (k'), retention time (Rt) and peak areas obtained each time were not significant.</p><!><p>A simple, accurate and precise HPLC method with fluorescence detector for trace determination of LND after its pre-column derivatization with FLC has been successfully developed and validated. The sample preparation procedure was very simple and robust as it did not involve any liquid-liquid extraction of the sample. It was based on selective complexation of the endogenous α- amino acids with copper ions and precipitation of the proteins with acetonitrile before the derivatization reaction. The derivatized sample was directly injected into the HPLC system. The chromatographic separation was based on a reversed phase mechanism carried out under isocratic elution mode for only less than 15-min total run time. The analytical results demonstrated that the proposed method is suitable for the accurate quantification of LND in human plasma at concentrations as low as 2.3 ng/mL, with a wide linear range. The simple procedure involved in the sample preparation and the short run-time added the property of a higher throughput to the method. It is valuable for the combined pharmacokinetic and bioavailability studies of LND in human subjects.</p><!><p>MM: Multiple myeloma; LND: Lenalidomide; FLC: Fluorescamine; HPLC: High-performance liquid chromatography; LOD: Limit of detection; LOQ: Limit of quantification; SD: Standard deviation; RSD: Relative standard deviation</p><!><p>The authors declare that they have no competing interests.</p><!><p>NYK conducted the method development and validation. IAD proposed the subject, designed the study, participated in the results discussion and revised the manuscript. TAW participated in method development and validation. AAA participated in results discussion and writing the manuscript. All authors read and approved the final manuscript.</p>

PubMed Open Access

Interactions of Alamethicin with Model Cell Membranes Investigated Using Sum Frequency Generation Vibrational Spectroscopy in Real Time in Situ

Structures of membrane associated peptides and molecular interactions between peptides and cell membrane bilayers govern biological functions of these peptides. Sum frequency generation (SFG) vibrational spectroscopy has been demonstrated to be a powerful technique to study such structures and interactions at the molecular level. In this research, SFG has been applied, supplemented by attenuated total reflectance Fourier transform infrared spectroscopy (ATR-FTIR), to characterize interactions between alamethicin (a model for larger channel proteins) and different lipid bilayers in the absence of membrane potential. The orientation of alamethicin in lipid bilayers has been determined using SFG amide I spectra detected with different polarization combinations. It was found that alamethicin adopts a mixed \xce\xb1-helical and 310- helical structure in fluid-phase lipid bilayers. The helix (mainly \xce\xb1-helix) at the N-terminus tilts at about 63\xc2\xb0 versus the surface normal in a fluid-phase 1,2-Dimyristoyl-D54-sn-Glycero-3-Phosphocholine-1,1,2,2-D4-N,N,N-trimethyl-D9 (d-DMPC)/1,2-Dimyristoyl-sn-Glycero-3-Phosphocholine (DMPC) bilayer. The 310 helix at the C-terminus (beyond the Pro14 residue) tilts at about 43\xc2\xb0 versus the surface normal. This is the first time to apply SFG to study a 310 helix experimentally. When interacting with a gel-phase lipid bilayer, alamethicin lies down on the gel-phase bilayer surface and/or aggregates, which does not have significant insertion into the lipid bilayer.

interactions_of_alamethicin_with_model_cell_membranes_investigated_using_sum_frequency_generation_vi

1. Introduction<!>2.1. Materials<!>2.1.1 Preparation of lipid bilayers<!>2.2 Polarized ATR-FTIR Experiments<!>2.3 SFG<!>2.4.1 Orientation angle of peptides deduced from ATR-FTIR<!>2.4.2 Orientation angle of peptides deduced from SFG<!>3.1.1 SFG results<!>A) Orientation of 310 helix containing residues 14\xe2\x80\x9320 (\xce\xb82)<!>B) Orientation of the helix containing residues 1 to 13 (\xce\xb81)<!>3.1.2 ATR-FTIR results<!>3.2 Interaction between Alamethicin and Different Lipid Bilayers<!>4. Conclusion<!>

<p>Ion channels represent an important class of transmembrane proteins that regulate the ionic permeability in cell membranes. They are key elements in signaling and sensing pathways, as well as connecting the inside of the cell to its outside in a selective fashion.1 They play crucial roles in normal and pathophysiological functions of cells. Defective ion channels can lead to many diseases such as cystic fibrosis, cardiac arrhythmias, and Parkinson's disease.2–9 Investigations on structures of these membrane proteins will aid in the understanding of disease mechanisms and provide important clues to cure such diseases.</p><p>Alamethicin is a 20-residue hydrophobic antibiotic peptide extracted from the fungus Trichoderma viride that can form voltage-gated ion channels in membranes.10–23 It has been used frequently as a model for larger channel proteins.10–18 In addition to the regular amino acids, the peptide contains eight aminoisobutyric acid units. The molecular structure and conformational features of alamethicin have been studied extensively.10–27 The crystal structure of alamethicin crystallized from methanol determined by X-ray diffraction is predominantly helical, with an N-terminal α-helix and a C-terminal domain beyond Pro14 residue that contains 310-helical element.24 Pro14 residue acts as a bend in the helix and the bend angle between the two helical axes is about 20° to 35°.24</p><p>Extensive research has been performed to examine the mechanisms of alamethicin's action on cell membranes.10–23 It is currently believed that alamethicin interacts with cell membranes through the barrel-stave mode with the resulting conducting pores in the membrane formed by parallel bundles of three to twelve helical alamethicin monomers surrounding a central, water-filled pore.12,13,22–24,28–30 However, further details on the mechanism of alamethicin channel formation at the molecular level are still far from completion.31,32 In addition, contradictory orientations of alamethicin in the cell membranes in the absence of membrane potential have been reported. Alamethicin has been suggested to adopt a transmembrane orientation,25–27, 33–37 lie down on the membrane surface,38–40 or both (depending on the experimental conditions).41,42 A continuous distribution of orientations has also been proposed.43</p><p>A detailed characterization of interactions between alamethicin and model membranes without a transmembrane potential is a fundamental step in the understanding of the operational mechanisms of ion channels. However, as mentioned above, different results have been reported in the literature. In this research, we applied sum frequency generation (SFG) vibrational spectroscopy, supplemented by attenuated total reflectance Fourier transform infrared spectroscopy (ATR-FTIR), to investigate the interactions between alamethicin and different lipid bilayers. SFG is a nonlinear optical laser technique which provides vibrational spectra of surfaces and interfaces.44–66 It has several advantages over other analytical techniques: it is intrinsically surface sensitive, requires small amount of samples, and can probe surfaces and interfaces in situ in real-time. As a polarized vibrational spectroscopy, SFG permits the identification of interfacial molecular species (or chemical groups), and also provides information about the interfacial structure, such as the orientation and the orientation distribution of functional groups on a surface or at an interface.44–55 SFG has been applied to study the structure and orientation of various biomolecules (including peptides and proteins) in interfacial environments.56–64 Here we deduced the orientation of alamethicin by analyzing the polarized SFG amide I signals without the presence of membrane potential, which will serve as a basis for future SFG studies on alamethicin under membrane potential. In addition, we also investigated the lipid chain length effect on the interactions of alamethicin with model membranes using SFG.</p><!><p>Alamethicin from Trichoderma viride was purchased from Sigma-Aldrich (St. Louis, MO), with a minimum purity of 90%. Different lipids (listed in Table 1) were purchased from Avanti Polar Lipids (Alabaster, AL). Deuterated water (D2O) was ordered from Aldrich (Milwaukee, WI). Right-angle CaF2 prisms were purchased from Altos (Trabuco Canyon, CA).</p><p>All of the chemicals were used as received. CaF2 prisms were thoroughly cleaned using a procedure with several steps: They were first soaked in toluene for 24 h and then sonicated in Contrex AP solution from Decon Labs (King of Prussia, PA) for 1 h. After that, they were rinsed with deionized (DI) water before soaking in methanol for 10 minutes. All of the prisms were then rinsed thoroughly with an ample amount of DI water and cleaned inside a glow discharge plasma chamber for 4 min immediately before depositing lipid molecules on them. Substrates were tested using SFG, and no signal from contamination was detected.</p><!><p>Single lipid bilayers which can have two different leaflets were prepared on CaF2 substrates. Langmuir-Blodgett and Langmuir-Schaefer (LB/LS) methods were used to deposit the proximal and then the distal leaflets, respectively. A KSV2000 LB system and ultrapure water from a Millipore system (Millipore, Bedford, MA) were used throughout the experiments for bilayer preparation. The detailed procedure was reported previously65,66 and will not be repeated here.</p><p>The bilayer was immersed in water inside a 1.6-mL reservoir throughout the entire experiment and a small amount of water could be added to the reservoir to compensate for evaporation when needed for long timescale experiments. For alamethicin-bilayer interaction experiments, ~15 μL alamethicin solution (in methanol with a concentration of 2.5 mg/mL) was injected into the reservoir. A magnetic micro stirrer was used to ensure a homogeneous concentration distribution of peptide molecules in the subphase below the bilayer.</p><!><p>A Nicolet Magna-IR 550 spectrometer was used to collect attenuated total reflectance-Fourier transform infrared (ATR-FTIR) spectra with a standard 45° ZnSe ATR cell and a ZnSe grating polarizer (from Optometrics LLC). The ZnSe crystal (Specac Ltd. Woodstock, GA) was cleaned using the same procedures as the CaF2 prisms. The lipid bilayers were prepared onto the ZnSe crystal surface using the LB/LS method mentioned above. After the lipid bilayer was deposited onto the crystal, the water that kept the bilayer hydrated was flushed multiple times with D2O to avoid signal confusion between the O-H bending mode and the peptide amide I mode and to ensure a better S/N ratio in the peptide amide I band frequency region. After at least 2 h to allow equilibration, a background polarized spectrum of the lipid bilayer/D2O interface was recorded. Then about 15 μL alamethicin solution (in methanol with a concentration of 2.5 mg/mL) was injected into the above small reservoir of 1.6 mL. After at least 1 h to allow the alamethicin adsorption to reach equilibrium, a polarized spectrum was collected. Finally, the amide I signal of alamethicin on bilayer in D2O was obtained by subtracting the background spectrum of the bilayer/D2O interface from the later collected spectrum after alamethicin was adsorbed and equilibrated. All the spectra collected here were averages of 256 scans with a 2 cm−1 resolution.</p><!><p>SFG is a second-order nonlinear optical spectroscopic technique that has submonolayer surface sensitivity.44–66 Details regarding SFG theories and measurements have been extensively published44–66 and will not be repeated here. The SFG experimental setup was similar to that described in our earlier publications and will not be presented.65,66 In this research, all of the SFG experiments were carried out at the room temperature (25 °C). SFG spectra from interfacial alamethicin with different polarization combinations including ssp (s-polarized SF output, s-polarized visible input, and p-polarized infrared input) and ppp were collected using the near total internal reflection geometry.</p><!><p>ATR-FTIR spectroscopy has been widely used to analyze peptide/protein secondary structures on surfaces or at interfaces and determine the orientation of such secondary structures.67 In ATR-FTIR studies, the tilt angle (θ) of the helices can be determined from the measured infrared linear dichroic ratio (R) in ATR-FTIR using p- and s-polarized IR beams:67,68</p><p> (1)R=A//A⊥=Ez2kz+Ex2kxEy2ky. where Ei (i=x,y,z) is the electric field amplitude of the evanescent wave at the surface of the internal reflection element, and ki (i=x,y,z) is a component of the integrated absorption coefficient in the lab fixed coordinate system. Ei (i=x,y,z) depends on the incidence angle of the IR beam at the solid-liquid interface, and the refractive indices of the internal reflection element (ATR crystal), the thin film (bilayer), and the bulk contacting medium (D2O). We calculated the value of Ei (i=x,y,z) using the formula published in the literature.67,68 If we model the orientation distribution of a helix in the lab-fixed coordinate system as a Gaussian distribution ( f=12πσe−(x−θ)22σ2), ki (i=x,y,z) is given as follows:68</p><p> (2)〈kx〉=〈ky〉=cos(α)2(12−cos(2θ)2e2σ2)2+sin(α)24+sin(α)2(12+cos(2θ)2e2σ2)2, (3)〈kz〉=cos(α)2(12+cos(2θ)2e2σ2)+sin(α)2(12−cos(2θ)2e2σ2)2, where θ and σ are the tilt angle between the helix's principal axis and the surface normal and the orientation distribution width respectively; α is the angle between the transition dipole moment and the principal axis of the helix, which equals to 38° for α-helix and 45° for 310 helix.69,70 The bracket denotes the time and ensemble average. When σ = 0, the orientation distribution is a δ-distribution. Since ATR-FTIR only provides one experimentally measured parameter (R), based on the eqns (1) to (3), the tilt angle θ can be determined by knowing the value of Ei (i=x,y,z), and α, and assuming a certain value of σ. Here we used a delta distribution.</p><!><p>The molecular orientation information can be obtained by relating SFG susceptibility tensor elements χijk(i, j,k = x, y, z) to the SFG molecular hyperpolarizability tensor elements βlmn (l,m,n = a,b,c).44–66 Our lab has developed a methodology to determine the orientation of α-helical structure using SFG amide I spectra collected with different polarization combinations. This method has been introduced in our previous papers 66, 71–75 and will not be repeated here.</p><p>Similar to the method for an α-helix, we developed the orientation analysis method for a 310-helix in membrane.72 We deduced the relation between the χppp/χssp value and the 310-helix orientation with a δ- or Gaussian distribution using different hyperpolarizability tensor elements with the adoption of the bond additivity model. Thus the orientation angle (θ) of 310 helix can be deduced by measuring the ppp and ssp spectral intensity ratio of the peptide amide I signals.</p><!><p>SFG ssp and ppp spectra of alamethicin in a d-DMPC/DMPC bilayer are shown in Fig. 1. The spectra were collected after 15 μL alamethicin/methanol solution was injected into the subphase (~1.6 mL) of a d-DMPC/DMPC bilayer at pH 6.7. The SFG spectra are dominated by two peaks at 1635 cm−1 and 1670 cm−1, which is consistent with the results of previous FTIR and Raman studies in which the amide I peaks centered at 1639 and 1662 cm−1 were observed in membrane-incorporated alamethicin.76–78 The frequency of the 1662 cm−1 peak in IR spectra is higher than those normally found for soluble or membrane-inserted α-helices (usually at about 1650 cm−1). Chapman et al. conclude that this higher frequency is an indication for a 310-helical structure connected to the α-helix in alamethicin in lipid bilayers.76 Therefore we believe that the 1670 cm−1 peak observed in the SFG spectra is contributed by a helical structure dominated by α-helix but with a 310-helix part. Peak assignments in the literature indicate that the 1635 cm−1 peak is due to the 310-helix.76–79</p><p>Alamethicin is consisted of two helical segments because of the presence of the helix-breaking Pro14 residue.24 According to the above discussion, the 310-helix formed by residues 14 to 20 contributes to the signal at 1635 cm−1; the α-helical/310-helical structure (mainly α-helical component) which contains residues 1 to 13 contributes to the signal at 1670 cm−1. Here, we define the tilt angle between the principal axis of the helix with the residues 1 to 13 and the d-DMPC/DMPC bilayer surface normal to be θ1, while the tilt angle between the principal axis of the helix composed of residues 14 to 20 and the d-DMPC/DMPC bilayer surface normal to be θ2.</p><!><p>Using the relation between the measured ppp and ssp spectral intensity ratio of the peak at 1635 cm−1, we should be able to determine the orientation angle (θ2). We deduced the hyperpolarizability tensor elements for a 310 helix that consists of 7 amino acid residues and obtained βaca =0.54βccc and βaac= 1.1βccc. Details regarding the deduction of specific tensor elements of 310 helix were reported previously72 and will not be repeated here. As a result of the deduction, a relation between the measured χppp/χssp ratio and the 310 helix orientation for a 310 helix with 7 amino acids is presented in Fig. 2.72 Here the experimental measured χppp/χssp ratio of the peak at 1635 cm−1 in the d-DMPC/DMPC bilayer is 1.80±0.15, therefore the orientation angle θ2 is about 43° (between 39° and 47°). The number of amino acid residues in an ideal 310 helix should be multiples of 3. There are some concerns that the 310-helical symmetry may be slightly broken when the number of the amino acids in a 310 helix deviates from multiples of 3. In that case, the relation between the χppp/χssp value and 310 helix orientation may vary when the number of the amino acids in a 310 helix changes. To address the above concern, we calculated βaac/βccc, βaca/βccc for 310 helices having 3 to 7 amino acids.72 Using the calculated parameters, the ratio of χppp/χssp as a function of 310-helix orientation angle (θ2) can be deduced. The results indicated that the orientation determination is not noticeably affected by the numbers of amino acids in a 310-helix when the tilt angle is smaller than 50°.</p><!><p>The orientation angle (θ1) can be deduced by using the relation between the measured ppp and ssp spectral intensity ratio of the peak at 1670 cm−1. We deduced the hyperpolarizability tensor elements for the helix that consists of residues 1 to 13, which contains both α- and 310- helical structures (mainly α-helical elements). Recently, Salnikov et al. studied the structure and alignment of uniformly 15N-labeled alamethicin in POPC and DMPC membranes using oriented 15N and 31P solid state NMR spectroscopy.27 A model structure with α-helix formed by the first ten residues and 310-helix formed by the next ten residues was found to be able to predict the features in the observed NMR spectrum reasonably well.27 Here we propose that the helix formed by residues 1 to 13 has two portions, an α-helix formed by the first ten residues and a 310-helix formed by the residues 11 to 13. The relation between the χppp/χssp ratio of this structure and its membrane orientation is presented in Fig. 3. The experimentally measured χppp/χssp ratio for the peak at 1670 cm−1 in the d-DMPC/DMPC bilayer is about 2.45±0.15, yielding an orientation angle θ1 of about 63° (between 57° and 68°). Although this value of θ1 is deduced from the above proposed structure with α-helix formed by the first ten residues and 310 helix formed by residues 11 to 13, θ1 does not substantially change when different structures were used in orientation determination, as shown in Table 2.</p><p>We note that the bend angle between the two helical components in alamethicin (θ1–θ2) is about 20° in this study, assuming the plane containing both helical components are perpendicular to the membrane surface. This is in excellent agreement with previous results: 17° was reported in DMPC-bilayer associated alamethicin,26 and 20–35° in crystallized alamethicin.24</p><!><p>ATR-FTIR was used as a supplemental technique to substantiate SFG results. ATR-FTIR polarized spectra of alamethicin in a d-DMPC/DMPC bilayer are displayed in Fig. 4. According to the previous results in the literature, we fit these spectra using three peaks centered at 1623, 1635 and 1660 cm−1. The intensity ratio (R) of the signal measured using the p- versus s- polarized beam is 1.7 for the 1635 cm−1 peak and 1.6 for the 1660 cm−1 peak. From this R value, the orientation angle can be calculated to be 55° for θ1 (between 52° and 58°) and 49° for θ2 (between 45° and 53°), assuming a δ-orientation distribution; they are not very different from the SFG results of θ1 =63° and θ2 = 43° presented above. The difference between the SFG and ATR-FTIR results might be due to the fact that the orientation distribution is not a delta-distribution.</p><p>Our SFG and ATR-FTIR studies both indicated that alamethicin molecules exhibit a large tilt angle versus the surface normal in a d-DMPC/DMPC bilayer. These results agree with those from recent ATR-FTIR research on fluid-phase lipid membrane associated alamethicin.32,36 Marsh reported that the tilt angle of alamethicin is 67° in 1,2-didecanoyl-sn-glycero-3-phosphocholine membranes, and 51° in 1,2-diundecanoyl-sn-glycero-3-phosphocholine membranes at 36 °C.36 In addition, Stella et al. observed that the tilt angle of alamethicin is about 60° in a POPC bilayer membrane.32</p><p>In contrast, our results are quite different from those obtained from labeled alamethicin in NMR or electron paramagnetic resonance (EPR) studies. Site-specific 15N-labeled alamethicin was found to be more or less parallel to the DMPC bilayer normal,25 or slightly tilted (10° to 20°) determined by solid-state 15N NMR.26 In addition, it was shown by solid-state 15N and 31P NMR spectroscopy that uniformly 15N-labeled alamethicin orients parallel to the POPC and DMPC membrane surface normal.27 EPR studies indicated that TOAC-substituted alamethicin orients with a tilt angle varying from 23° to 13° in fluid-phase diacyl phosphatidylcholine bilayer at 75 °C.35 It has been suggested that the measured orientation of alamethicin on membranes depends on the experimental conditions.41,42 We also agree that the different methods may lead to varying results. NMR studies require complicated sample preparation procedures as well as exogenous labels. The effective order parameters measured by EPR with TOAC-substituted alamethicin are relative to the local membrane director.35 This local tilt of transmembrane polypeptides has been augmented by thermally excited elastic bending fluctuations of the entire membrane.80,81 These fluctuations can change the elastic modulus for membrane area expansion,82 facilitating the insertion of transmembrane proteins/peptides, and also give rise to a net inclination of the local membrane normal (or director), to which the molecular tilt of the peptide is referred.82</p><!><p>It has been shown that the membrane lipid chain length affects the interactions between alamethicin and cell membranes.21,34–36,83,84 Recently, Marsh et al. investigated the dependence of the incorporation of spin-labeled alamethicin into fluid phosphatidylcholine membrane bilayers on lipid chain length using EPR. Their results suggested the orientation and aggregation of alamethicin are related to the chain length of the lipid. Lipid chain length can modulate the activity of transmembrane proteins/peptides by the mismatch between the hydrophobic span of the protein/peptide and the lipid membrane.34–36 In this study, we observed markedly different SFG signal intensities from alamethicin in bilayers with lipids of different chain lengths. The different lipids examined and varied SFG results are listed in Table 3. It is well known that the lipid chain length is one of the factors that determine the phase of the lipid bilayer at the room temperature: Similar lipids with longer chains tend to exist in the gel phase, whereas with shorter chains are likely in the fluid phase.</p><p>Fig. 5 shows the ppp SFG spectra collected after 15 μL alamethicin/methanol solution was injected into the subphase (~ 1.6 mL) of various bilayers. In the fluid-phase lipid bilayers (Fig. 5a–d), strong SFG amide I signals of alamethicin were observed, dominated by two peaks at 1635 and 1670 cm−1. These two peaks are contributed by the 310-helical structure and the α-helical structure of alamethicin. Using the orientation analysis method discussed above, the orientations of alamethicin in different lipids were investigated. The deduced results indicated that the orientations of alamethicin in different fluid-phase lipid bilayers (POPC/POPC, POPC/POPG, DMPC/DMPC bilayers) are similar to each other, and thus also similar to the one presented above regarding the d-DMPC/DMPC bilayer. Therefore it is concluded that alamethicin has a highly ordered orientation in fluid-phase lipid bilayers with a large tilt angle.</p><p>When alamethicin was interacting with gel-phase lipid bilayers (Fig. 5e–g), only two weak SFG peaks at 1685 and 1720 cm−1 were observed. We believe that the peak at 1685 cm−1 is contributed by the antiparallel β-sheet or aggregated strand of peptides.61,67 The 1720 cm−1 signal is not the amide I signal from alamethicin; instead it originates from the carbonyl groups of the lipid bilayer. The absence of the strong alamethicin helical signal on gel-phase lipid bilayer indicates that both the helical structures lie down on the lipid bilayer surface and/or may be changed into other secondary structures, e.g., antiparallel β-sheet or aggregated strand. This is consistent with the results obtained using other analytical tools in the literature.38–40 Therefore, it is evident that alamethicin interacts with gel-phase and fluid-phase lipid bilayers differently.</p><!><p>We applied SFG to investigate the molecular interactions between alamethicin, an important model for larger channel proteins, and different lipid bilayers in situ and in real time without exogenous labeling. It was found that alamethicin interacts differently with gel-phase versus fluid-phase lipid bilayers. When alamethicin molecules interact with gel-phase lipid bilayers, they lie down and/or aggregate on the gel-phase bilayer surface. We believe that they do not have significant insertion into the lipid bilayers. Differently, alamethicin molecules can insert into fluid-phase lipid bilayers with large tilt angels in the absence of membrane potential. The orientation of alamethicin in fluid-phase lipid bilayers was determined by deducing the orientations of the α- and 310- helical structural segments using polarized SFG amide I spectra, and substantiated by the polarized ATR-FTIR studies. According to our knowledge, this is the first time to successfully apply SFG to determine orientation of a 310- helix experimentally. The formation of channels by alamethicin under membrane potential will be investigated in the future to elucidate molecular mechanisms of these channels.</p><!><p>SFG spectra of alamethicin in a d-DMPC/DMPC bilayer at pH=6.7. Top: ssp spectrum; bottom: ppp spectrum.</p><p>The relation between SFG susceptibility tensor component ratio and the 310-helix orientation angle. The measured orientation angle is about 43° (between 39° and 47°) assuming a narrow angle distribution.</p><p>The relation between SFG susceptibility tensor component ratio for the helix containing 1–13 residues and the helix orientation angle. The measured orientation angle is about 63° (between 57° and 68°) assuming a narrow angle distribution.</p><p>Polarized ATR-FTIR spectra of alamethicin in a d-DMPC/DMPC bilayer at pH=6.7. Top: p-polarized spectrum; bottom: s-polarized spectrum.</p><p>SFG ppp spectra of alamethicin in different bilayers at pH=6.7. a) POPC/POPC; b) POPC/POPG; c) d-DMPC/DMPC; d) d-DMPC/d-DMPC; e) d-DPPC/DPPC; f) d-DPPG/DPPG; g) d-DSPC/DSPC.</p><p>Lipids studied in this paper</p><p>The orientation angle θ1 of different proposed structures of the helix containing residues 1 to 13 in alamethicin</p><p>Interactions between alamethicin and different lipid bilayers.</p><p> http://www.avantilipids.com/PhaseTransitionTemperaturesForGlycerophospholipids.html </p>

PubMed Author Manuscript

Understanding how Lewis acids dope organic semiconductors: a “complex” story

We report on computational studies of the potential of three borane Lewis acids (LAs) (B(C 6 F 5 ) 3 (BCF), BF 3 , and BBr 3 ) to form stable adducts and/or to generate positive polarons with three different semiconducting p-conjugated polymers (PFPT, PCPDTPT and PCPDTBT). Density functional theory (DFT) and timedependent DFT (TD-DFT) calculations based on range-separated hybrid (RSH) functionals provide insight into changes in the electronic structure and optical properties upon adduct formation between LAs and the two polymers containing pyridine moieties, PFPT and PCPDTPT, unravelling the complex interplay between partial hybridization, charge transfer and changes in the polymer backbone conformation. We then assess the potential of BCF to induce p-doping in PCPDTBT, which does not contain pyridine groups, by computing the energetics of various reaction mechanisms proposed in the literature. We find that reaction of BCF(OH 2 ) to form protonated PCPDTBT and [BCF(OH)] À , followed by electron transfer from a pristine to a protonated PCPDTBT chain is highly endergonic, and thus unlikely at low doping concentration. The theoretical and experimental data can, however, be reconciled if one considers the formation of [BCF(OH)BCF] À or [BCF(OH)(OH 2 )BCF] À counterions rather than [BCF(OH)] À and invokes subsequent reactions resulting in the elimination of H 2 .

understanding_how_lewis_acids_dope_organic_semiconductors:_a_“complex”_story

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

<p>Molecular doping [1][2][3][4][5][6] is a paramount topic in the organic semiconductor community, where it can enhance charge-carrier density and therefore electrical conductivity, improve charge injection and lower contact resistance, or increase charge mobility thanks by lling traps. The most straightforward approach to p-or n-doping is to use simple one-electron oxidants or reductants that react with the semiconductor to generate radical cations or anions (positive or negative polarons). A less intuitive approach to doping involves Lewis acids (LAs), notably tris(pentauorophenyl)borane (BCF). Depending on the nature of the semiconducting polymers, LAs either effectively act as p-dopants or form Lewis Acid-Base (LAB) adducts. 7 The aim of this computational study is to give insight into these two types of reactivity.</p><p>A decade ago, it was demonstrated that LAs can form physical complexes with semiconducting p-conjugated polymers, 8 a process driven by the interaction between the empty p-orbitals of the centrally electrophilic boron atom in the LA and the electron lone pair of a Lewis base (LB) site on the polymer, such as a pyridyl nitrogen. The formation of a new stable covalent bond yields a LAB adduct with a specic ngerprint in optical absorption 9 and increased charge carrier density with respect to the unbound polymer, [10][11][12] representing a means of postsynthetic engineering. 13 More specically, alternating donoracceptor conjugated copolymers, where the acceptor moiety is pyridylthiadiazole (PT), are able to strongly coordinate LAs, such as BCF, likely resulting in partial ground-state charge transfer (CT). The interaction with BCF has been shown to translate into a red-shied onset in optical absorption of the organic semiconductor by $0.3 eV, a shi primarily due to the effect of the electron-withdrawing LA moiety on the electron affinity in presence of the LA itself. 13 Rather unexpectedly, BCF can also act as an apparent oxidant. Indeed, in the late 1990s Doerrer and Green 14 demonstrated that BCFeither when used intentionally as its 1 : 1 water complex BCF(OH 2 ), which is a strong Brønsted acid, or in the presence of adventitious watercan behave as a strong oxidant, converting metallocenes (MCp 2 , M ¼ Fe, Cr, Co) to the corresponding MCp 2 + . They considered that oxidation likely proceeded by protonation of MCp 2 by BCF(OH 2 ), followed by elimination of H 2 from two MCp 2 H + ions. Interestingly, the products they obtained did not contain the simple [BCF(OH)] À anion (which is known and crystallographically characterized in other contexts 15 ), but rather either [BCF(OH)BCF] À or [BCF(OH)(OH 2 )BCF] À anions. More recently, the oxidizing characteristics of BCF have been rediscovered in the context of the p-doping of organic semiconductors. 16 BCF behaves as a strong oxidant, consistent with the ndings of Doerrer and Green, but inconsistent with a simple one-electron transfer from polymer to BCF. It has been observed that BCF is reduced to the unstable radical anion at ca. À1.7 to À1.8 V versus ferrocene, 17 whereas polymers that have been doped by BCF are oxidized at potentials comparable to, or more positive than, ferrocene, indicating that such an electron transfer would be highly endergonic. Thus, BCF(OH 2 ), or other BCF(OH 2 ) n adducts, which are strong Brønsted acids and are formed by the hygroscopic BCF (unless water is scrupulously excluded), are thought to be the likely oxidant, if not by a direct one-electron transfer manner. In some cases, the use of BCF may be desirable relative to the very widely used 2,3,5,6-tetrauoro-7,7,8,8tetracyanoquinodimethane (F 4 TCNQ), due to its solubility in organic solvents, its lower volatility, and its ability to dope molecular materials with a relative high ionization potential ($5.8 eV). 11,12,18 On the other hand, other p-dopants that act as clean one-electron-oxidants may be more predictable in their behaviour as a consequence of their more straightforward chemistry. 19,20 In any case, Yan et al. have successfully used BCF as molecular dopant in a donor:acceptor planar heterojunction device structure and found that LA doping plays a synergistic role in changing the opto-electronic properties and nanomorphology of the blends leading to improved device performances, even at low doping concentration. [21][22][23] Consistent with the work of Doerrer and Green, 14 it has been suggested that some particular polymers like poly-cyclopentadithiophene-benzothiadiazole (PCPDTBT) can be also oxidized by BCF(OH 2 ) via an initial protonation step of the cyclopentadithiophene (CPDT) unit in the polymer backbone. In ref. 16 it was proposed that the resulting protonated, positively charged, polymer chain would undergo an increase in electron affinity (compared to the pristine polymer) large enough to prompt an electron transfer from another, pristine, polymer chain (or chain section), resulting in the presence of two radical species, i.e., a neutral "protonated radical" and a radical cation (positive polaron). Continuous-wave electronnuclear double resonance (ENDOR) spectroscopy affords a spectrum that is consistent with the presence of both radicals; specically, a structureless spectrum is observed similar to what is expected for the "protonated radical", while the polaron is expected to contribute a much less intense structured pattern. However, in a later work on p-doping of poly(3-hexylthiophene) (P3HT), Arvind et al. could observe only the radical cation using high-resolution electron paramagnetic resonance (EPR) spectroscopy, suggesting either the "protonated radical" does not form or that it is unstable against further chemical reactions. 24 In particular, H 2 elimination, as previously invoked in the contexts of both metallocene oxidation by BCF(OH 2 ) and spiro-OMeTAD p-doping by HN(SO 2 CF 3 ) 2 (another strong Brønsted acid), 25 has been suggested to play a paramount role, but to our knowledge formation of H2 has yet to be observed directly.</p><p>A comprehensive description of how LAs interact with semiconducting p-conjugated polymers is currently lacking. Here, we report on state-of-the-art calculations investigating the potential of three boron-based LAs to either form physical complexes or undergo chemical reactions involving oneelectron oxidation of the semiconductor with three different p-conjugated polymers (Fig. 1). Using density functional theory (DFT) and time-dependent DFT (TD-DFT) calculations based on optimally tuned (OT) range-separated hybrid (RSH) functionals, 26,27 we rst analyse the structural, energetics, and optical signature of ground-state complexes formed between three LAs and poly-uorene-pyridylthiadiazole (PFPT) and poly-cyclopentadithiophene-pyridylthiadiazole (PCPDTPT) tetramers, nding good agreement with experiment and highlighting the factors affecting the changes in optical absorption. Though there is clear experimental evidence that LAs are able to dope some polymer semiconductors, the mechanistic aspects of the doping have not been elaborated yet. We thus move on in investigating the doping mechanisms of a PCPDTBT tetramer by BCF(OH 2 ) from rst-principles. This involves identifying the most likely protonation sites and assessing the energetics of previously proposed reactions. Our results show that those are highly endergonic, mostly due to the thermodynamically unfavourable protonation to form [BCF(OH)] À , thus ruling out all proposed mechanistic scenarios proposed in the literature. Capitalizing on the seminal work by Doerrer and Green, we instead consider reactions leading to the formation of larger complex anions, as observed in the context of metallocene oxidation. 14 Remarkably, we then nd that the resulting protonated PCPDTBT chains can undergo moderately endergonic reactions when eliminating H 2 to produce a single spincarrying charged species.</p><!><p>Gas-phase ground-state equilibrium geometries of two representative tetramers, PFPT and PCPDTPT, were obtained by performing DFT optimization at the RSH functional level of theory, using the exchange-correlation uB97X-D functional 28 and the 6-31G(d,p) split-valence Pople's basis set for all the atomic species. The tetramers containing the PT moiety were optimized as an alternating copolymer of formula H-(-A-B-) 4 -H considering the regiochemical alternation between successive PT groups. For the sake of simplicity and to speed up the calculations, the alkyl chains were substituted with methyl groups in all investigated tetramers, a licit procedure as recently shown in the literature. 29 The same level of theory was used for all the structural optimizations in gas-phase when we introduced the three different LAs to form the LAB adducts with the tetramer PFPT and PCPDTPT. We also checked the inuence of the OT range separation parameter u on the resulting optimized structures. 30 Using a RSH functional oen comes along with a non-empirical tuning of u. In fact, for each specic N-electron system, an optimal value of u can be found by enforcing the exchange-correlation functional to obey the DFT version of Koopman's theorem by aligning the negative energy of the HOMO with the gas-phase vertical IP. In practice, one computes the total energy difference between the N-electron and the (N-1)electron system and tries to minimize the overall error by minimizing the following target function:</p><p>In addition, for a better description of the fundamental gap, the gas-phase vertical EA of the N-electron system can be represented by the vertical IP of the (N+1)-electron system, barring relaxation effects. In order to perform a gap tuning procedure, [31][32][33][34][35][36] the modied target function to minimize is the following:</p><p>By doing that, the difference between the HOMO and LUMO energies of the N-electron systems in OT-RSH functionals provides a good approximation to the fundamental gap, that is the difference between IP and EA. In tuning the u value, we resorted to a polarizable continuum model 37 (PCM) using a screening dielectric constant of 3 ¼ 5.0, with the role of solvent polarity being addressed elsewhere. 29 With this caveat, from now on, we will refer to the highest occupied molecular orbital (HOMO) negative energy as the vertical ionization potential (IP) of the molecule and to the lowest unoccupied molecular orbital (LUMO) negative energy as its vertical electron affinity (EA). For the neat PFPT and PCPDTPT tetramer and their relatives LAB adducts, the absorption spectra were computed with full TD-DFT calculations and a ground-state population analysis was performed by means of the Charge Model 5 (CM5), 38 at the OT-RSH + PCM level of theory.</p><p>In order to identify the most likely protonation site by mimicking the protonation mediated by a Brønsted acid of the PCPDTBT tetramer, we modelled in a rst place a CPDT-BT-CPDT unit (see sketch in Table 3). The pristine and protonated model moieties were tightly optimized in gas-phase at the uB97X-D/6-31G(d,p) level of theory. Proton affinity (P(A)) is dened as the negative of the protonation reaction enthalpy at room temperature (T ¼ 298 K):</p><p>where DZPE is the corrected zero-point vibrational energy (ZPE) of the normal modes, DH 0 elec is the variation in the electronic enthalpy going from the pristine to the protonated model moiety and R is the ideal gas constant. Then, in order to evaluate the thermodynamic properties of all the reactions presented below, each molecule was tightly optimized at the uB97X-D/6-31G(d,p) level of theory in conjunction with PCM and 3 ¼ 5.0. The 3N À 6 frequencies of the vibrational normal modes (all checked to be positive) were computed and scaled by 0.949 in order to correct for anharmonicity effects. 39 In a given reaction, the Gibbs free energy difference DG 0 reads:</p><p>where DH 0 is the enthalpy and DS 0 is the entropy, both Tdependent. Moreover, each contribution can be decomposed in an electronic and a vibrational term (neglecting the rotational and translational ones, as they are not expected to contribute signicantly), so that:</p><p>DS 0 (T)</p><p>Within the harmonic approximation, the vibrational enthalpy H 0 vib (T) and the vibrational entropy S 0 vib (T) can be computed as:</p><p>where n i is the frequency of the i-th normal mode, h is the Planck constant, k B is the Boltzmann constant and both the sums run over the 3N À 6 normal modes. The electronic enthalpy H 0 elec is directly computed at the DFT level, while the electronic entropy S 0 elec can be estimated as:</p><p>where S is the spin multiplicity. Here we present reactions at room temperature that involve neutral (S ¼ 0) and radical (S ¼ 1/2) species: thus, only the latter have an electronic entropic contribution. In each investigated reaction, its DG 0 was computed as an energy difference between the products and the reactants, by calculating the enthalpic and entropic contribution of each species separately. DFT and TD-DFT calculations were performed using the GAUSSIAN16 package, 40</p><!><p>The optimized pristine PFPT oligomer shows a rather twisted structure. Due to the steric repulsion experienced by the nearest hydrogen atoms in the uorene group and the -CH side of the PT moiety (see Fig. S1 and Table S1 in ESI ‡), the dihedral angles between these two groups are 39 , while the lower steric bulk on the N-bearing side of the PT results in a smaller PT/uorene dihedral angles of 17-19 . Irrespective of its nature, the addition of one LA borane molecule with the boron atom in front of the pyridyl nitrogen in the PT group increases the dihedral angle up to 49-52 , while the other dihedrals further away from the LA remain unaltered. Gas-phase LAB adduct binding energies were estimated for the three LAs as total energy differences between the adduct coordinated with a LA and the sum of the isolated neat oligomer and LA molecule. The calculated binding energies prove the higher affinity of BBr 3 (À29.5 kcal mol À1 ), followed by BCF and BF 3 (À22.7 kcal mol À1 and À21.3 kcal mol À1 , respectively), in line with previous theoretical and experimental works. 43,44 The vertical IP and EA values of the neat PFPT oligomer and the corresponding adducts are reported in Table 1 (see also Fig. 2a). A clear stabilization of the charge-transport energy levels is observed in presence of LAs, i.e., both the IP and EA of the LAB adducts are increased. These changes are asymmetric, with a larger impact on EA than IP, resulting in a lowering of the transport gap, E gap . In the case of BCF, the IP increases by 0.14 eV and the EA by 0.39 eV, for an overall reduction in E gap of 0.25 eV. The changes in IP and EA are mostly driven by the partial ground-state CT taking place from the PT group to the LA, with changes across the series BBr 3 , BCF and BF 3 also reecting various degrees of hybridization of the unoccupied electronic levels (see ESI and Fig. S2 ‡). The predicted $0.1 eV change in IP upon complexation with BCF agrees with ultraviolet photoemission data. 13 TD-DFT calculations (Fig. 2b, Table 1) indicate the emergence of a new, red-shied, optical absorption band upon complexation. 45 As detailed below, the additional optical feature at wavelengths above 600 nm directly reects the section of the polymer backbone interacting with the LA, with regions spatially away from the contact points contributing to the feature that is seen at $520-550 nm, slightly blue-shied from that of the neat oligomer. We observe the largest red-shi of the lowest electronic excitation for BBr 3 (0.30 eV), followed by BCF (0.23 eV) and BF 3 (0.15 eV). The predicted red-shi (by 0.23 eV) of the lowest electronic transition is in excellent agreement with experimental optical absorption at 1 molar equivalent and above of BCF, showing a $0.3 eV red-shi of the maximum absorption peak in both lm and solution. 13 Natural transition orbitals (NTOs) pertaining to the lowest electronic excitation of the neat oligomer and the adduct with BCF are reported in Fig. 2c. In the neat PFPT oligomer, the hole density is delocalized over the entire molecular backbone, but the electron density has larger weights on the PT electron-accepting units (with dominant contributions on the two inner rings), consistent with the lowest excited state having signicant intramolecular CT character. When BCF binds a pyridyl nitrogen on the PT group, the hole density distribution remains essentially unaltered (despite the slight increase in IP relative to the pristine oligomer), and the electron density is now fully conned to the PT moiety that is in direct interaction with the LA (as this PT unit is now electron poorer and has higher EA). The lowest electronic excitation NTOs of the adduct with BF 3 and BBr 3 are shown in Fig. S3 in ESI. ‡ In order to assess the inuence of polymer chain length and its potential impact on the nature of the optical excitations, 46 we also modelled a neat PFPT octamer and its LAB adduct with BCF (see Table S2 and Fig. S4 in ESI ‡). By doubling the molecular length, we note that E gap is only slightly reduced (by $0.1 eV), mainly due to a destabilization of the IP. Irrespective of the conjugation length, the lowest electronic transition of the LAB adduct is red-shied by 0.20 eV compared to the neat polymer chain.</p><p>We performed the same analysis for another donor-acceptor oligomer, PCPDTPT, differing from PFPT by the nature of the electron-donating units (see Fig. S5 and Table S3 in ESI ‡). In contrast to PFPT, the PCPDTPT oligomer has a perfectly planar backbone with all dihedral angles equal to 0 in the pristine form. However, the addition of a LA molecule dramatically distorts the structure of the oligomer because of steric effects: the bulkier the LA, the higher the degree of distortion. In particular, the dihedral angle between the LA-bound side of the PT and the CPDT moiety reaches 112 (almost orthogonal orientation) in the adduct formed with BCF, 46 with BBr 3 and 39 with BF 3 . We stress that these substantial changes in the conformation of the molecular backbone are expected to strongly perturb the optical properties of the LAB adduct, as a result of the reduced p-conjugation. A similar effect was also observed by Schier et al. 47 for a quarterthiophene (4T) doped by BCF, with the presence of the LA interacting with the oligomer inducing substantial structural distortions. The calculated IP and EA values of the neat PCPDTPT and its respective LAB adducts, reported in Table 2 and Fig. 3a, show that, upon binding, there is an effective decrease in E gap . However, this effect is far less pronounced than for the PFPT oligomer, with the largest lowering of E gap being 0.14 eV in the case of BBr 3 (versus 0.35 eV for PFPT:BBr 3 ). As in the PFPT case, the IP, EA and E gap values are dictated by a partial ground-state CT and orbital hybridization in the LUMO of the adduct (see ESI and Fig. S6 ‡). We attribute the reduced spectral change to a competition to the opposing effects exerted by electronic CT and hybridization (which tend to reduce the gap) and conformational distortions away from planarity (which tend to increase the gap).</p><p>TD-DFT optical absorption spectra in Fig. 3b (see also Table 2) show that the formation of the LAB adduct is accompanied by the appearance of a new, red-shied, optical transition ngerprint, as in the PFPT case. The largest red-shi is predicted for BBr 3 (0.11 eV), followed by BCF (0.07 eV) and BF 3 (0.04 eV), following the trend of the calculated E gap values and similar to what reported above for PFPT. We also note that optical absorption measurements on PCPDTPT:BCF thin lms point to a larger spectral shi (reaching almost 0.4 eV) 16 than predicted, a discrepancy that could arise from conformational restraints in the solid-state (see Fig. In contrast to the previous two tetramers that were investigated, PCPDTBT does not undergo any binding reaction with LAs, 16 as the benzothiadiazole (BT) moiety lacks a pyridyl nitrogen able to share an electron lone pair with the empty boron p-orbital of the LA. Instead, adding BCF to a PCPDTBT based lm leads to an increase in electrical conductivity and to the formation of positive polarons, i.e., molecular pdoping. 16,24,48 As in the mechanism proposed by Doerrer and Green for oxidation of metallocenes, 14 Yurash et al. suggested that the rst step of this p-doping was the protonation by the highly Brønsted acidic complex BCF(OH 2 ) of the CPDT moiety of the polymer backbone. 16 They further proposed that protonation would increase the EA sufficiently that a nearby neutral chain segment would be able to transfer an electron to the (positively charged) protonated segment (with the segments belonging either to the same or different physical polymer chains, if the process is intrachain or interchain, respectively). This mechanism results in the formation of two radical species: a neutral, "protonated radical" and a radical cation, as shown in Scheme 1:</p><p>The optimized PCPDTBT structure in PCM yields a slightly twisted backbone, with all the dihedral angles of about 20 (see Fig. S9 and Table S5 in ESI ‡). In an attempt to identify the most likely protonation site along the polymer backbone, we performed P(A) calculations. The results reported in Table 3 show that (in contrast to ref. 16 in which position 3 was assumed to be protonated) position 1 (an a-carbon atom) in the CPDT moiety is the most favorable site to be protonated, followed by position 3 (a b-carbon atom) and 2. As a result, the DH 0 elec penalty for the protonation step is signicantly overestimated in the modeling work by Yurash et al. compared to the value reported here (+40.4 kcal mol À1 in ref. 16 versus +22.9 kcal mol À1 here). The addition of one proton (or hydrogen atom) to position 1 on the CPDT group dramatically affects the polymer backbone planarity since it breaks the p-conjugation by introducing sp 3 carbon atoms and the oligomer becomes quite twisted. By computing the thermodynamic properties of all the species (i.e., proposed reactants, intermediates and products) involved in the above reactions, our calculations show that both the protonation and the one electron-transfer processes are substantially endergonic, with DG 0 values of +23.0 and +13.1 kcal mol À1 , respectively (see Scheme 1), implying a total DG 0 of +36.1 kcal mol À1 (or +1.57 eV), thus suggesting the overall reaction to be very unlikely. In a recent study by Arvind et al. 24 on P3HT, EPR measurements performed on BCF-doped samples revealed the formation of free radical cations on the polymer backbone, yet showing no indication for the presence of another radical species (i.e., associated with the "protonated radical"). If BCF doping of PCPDTBT proceeds in analogous fashion to that proposed for the BCF-induced doping of P3HT by Arvind et al. the overall reaction would be that shown in Scheme 2:</p><p>The computed DG 0 value for the overall reaction is +31.5 kcal mol À1 , smaller than that for Scheme 1, but still highly endergonic. As shown in Scheme 3, several possible pathways might lead to the same overall reaction as that shown in Scheme 2:</p><p>In scenario A, the protonation step is followed by electron transfer (as in Scheme 1), but here two neutral "protonated radicals" subsequently react to eliminate H 2 to regenerate two neutral closed-shell polymers (shown for one such radical affording half a molecule of H 2 ), contributing a negative (exergonic) DG 0 ¼ À4.6 kcal mol À1 (or À0.20 eV). Scenario B is a variant of scenario A where H 2 is eliminated from two protonated cationic polymers, contributing with a DG 0 ¼ +8.5 (13.1-4.6) kcal mol À1 (or +0.37 eV). Finally, scenario C is a combination of scenarios A and B, leading, as expected, to a twofold increase in the total DG 0 ¼ +63.0 (2 Â 31.5) kcal mol À1 . We note that reactions of the type shown in Schemes 2 and 3 (and the similar overall reactions involving larger counter-ions that are discussed in the following section) are apparently at odds with the CW ENDOR results of ref. 16. However, although the structureless feature is consistent with that expected for the "protonated radical", it could also in principle arise from dynamic effects leading to loss of the structure expected for the polaron signal, or even from other radicals formed through side reactions. We also reckon that, as observed elsewhere in the literature, 24,[49][50][51] the polymer conjugation length plays a paramount role in the context of molecular doping, since different mechanisms might occur depending on the extension of the polymer backbone. To address this point, Table S6 in ESI ‡ reports the computed DH 0 elec values pertaining to Scheme 1 and Scheme 2, using either a PCPDTBT tetramer or an octamer as representative model. The computed Scheme 1 Reaction mechanism similar to that proposed by Yurash et al., involving a protonation followed by an electron-transfer reaction (this mechanism differs from that in ref. 16 in the position of the protonated site, see below). Calculations reported here yield DG 0 ¼ +23.0 kcal mol À1 (or +1.00 eV) for the protonation and DG 0 ¼ +13.1 kcal mol À1 (or +0.57 eV) for the electron transfer. For the sake of simplicity, the distinct structures are shown for single tetramer repeat unit, while we acknowledge that both spin and charge will be delocalized over multiple repeat units to varying extents. DH 0 elec values are found to be comparable, which comforts our choice of tetramer models as providing a good trade-off between accuracy and computational cost.</p><p>Neither the overall reactions of Scheme 1 nor Scheme 2 appear likely to represent the mechanism responsible for the formation of excess charge carriers in PCPDTBT upon LA doping, since the overall reactions are highly endergonic, with a particularly high energy penalty being associated with the protonation of the pristine polymer chains by BCF(OH 2 ) complex with concomitant formation of [BCF(OH)] À . However, in the previous work on metallocene oxidation by BCF(OH 2 ), 14 [BCF(OH)] À was not observed, but rather [BCF(OH)(OH 2 )BCF] À (Scheme 4), in which [BCF(OH)] À is hydrogen bonded to another BCF(OH 2 ) complex, and À (Scheme 5), where [BCF(OH)] À coordinates a BCF molecule. We reconsidered, therefore, Scheme 2 based on that proposed by Arvind et al. for P3HT, but now forming anions containing two BCF units of the two types observed by Doerrer and Green:</p><p>If we consider the protonation reaction as forming the "fourbody" [BCF(OH)(OH 2 )BCF] À anion of Scheme 4, the highly endergonic (DG 0 ¼ +23.0 kcal mol À1 ) protonation reaction found when [BCF(OH)] À is formed now becomes highly exergonic (DG 0 ¼ À22.2 kcal mol À1 ). Consequently, the overall DG 0 for the reaction presented in Scheme 1 and that for a H 2forming reaction in Scheme 2 is now negative: À9.0 kcal mol À1 (or À0.39 eV) for the former and À13.7 kcal mol À1 (or À0.59 eV) for the latter. Moreover, none of the proposed steps aer protonation is prohibitively endergonic, and thus may be kinetically feasible, while irreversible loss of gaseous H 2 can drive the doping reaction to the right. The greater exergonicity</p><!><p>We modelled the interactions between three boron-based LAs and different semiconducting p-conjugated polymers, performing detailed quantum-chemical calculations of the structural, energetics and optical signatures for ground-state LAB adducts between LAs and either PFPT or PCPDTPT. Our calculations demonstrate that the observed red-shied optical absorption in the adducts results from a complex interplay between hybridization, partial CT and changes in the polymer conformation. In assessing the potential of BCF to induce molecular doping in PCPDTBT based on calculated Gibbs free energies of different proposed reactions, we came to the conclusions that both the overall processes proposed by Yurash et al. 16 and by Arvind et al. 24 are highly endergonic, mostly because of the thermodynamically unfavourable protonation by BCF(OH 2 ). Reconciling theory with experiment requires considering complexation of the [BCF(OH)] À with another BCF or BCF(OH 2 ) moiety to form more stable anions of the stoichiometry and structure observed crystallographically by Doerrer and Green; 14 these offer a dramatic reduction in the DG 0 penalty for forming the protonated intermediates. We propose that this is followed by moderately endergonic reactions resulting in the elimination of H 2 (as also suggested for the case of metallocene oxidation), either directly from two protonated cationic segments of polymer chains, from "protonated radicals" formed by electron transfer between neutral and protonated cationic segments, or from a protonated cation and a protonated radical (Scheme 3), hence explaining why a single spin-carrying species is observed in EPR measurements. Overall, our calculations highlight the necessity of H 2 loss for the overall feasibility of the reaction, and most importantly, the key role played by the formation of diboron-containing bridged anions in the doping mechanism. Those bridged anions were known, as was the monomeric [BCF(OH)] À , but the energetic benets of bridged anion formation, and therefore its effect on overall reaction feasibility, had not been recognized and certainly not quanti-ed, neither in ref. 16 nor in other works dealing with the doping of p-conjugated polymers with LAs. This is the likely mechanism prevailing at dopant concentrations large enough that BCF dopants can encounter and form complex anions derived from two BCF moieties. In addition, at low dopant concentration and if the dopant is rigorously waterfree, it is also possible that highly hygroscopic BCF molecules could free hole carriers from trapping sites associated with water and/or water-oxygen complexes, 52,53 rather than create excess charges through a conventional doping mechanism. Additional experimental and theoretical work is needed to conrm or reject this hypothesis, as well as to unravel the exact nature of the BCF(OH 2 ) adducts present in doping solutions and the anions present in doped solids. However, this is likely to be very challenging as, even in solution, 1 H and 19 F NMR spectroscopies are unable to reliably distinguish between BCF(OH 2 ) n complexes with different n, 54 while neither the 11 B nor 19 F NMR spectra of [BCF(OH)BCF] À differ signicantly from that of [BCF(OH)(OH 2 )BCF] À in solution. 22 Finally we note that the non-straightforward doping nature of the BCF-induced doping process potentially complicates predictions regarding its applicability to other semiconductors. Although variations of the thermodynamic feasibility of the proposed overall p-doping reaction (Scheme 2, but with a complex counterion) for different semiconductors will depend only on the IP of the semiconductor, the kinetic feasibility is expected to depend critically on the ability to protonate the semiconductor. Moreover, different mechanisms may be operative for different semiconductors, for example, if they form substantially more stable "protonated radicals" than PCPDTBT. Finally, the use of BCF as a p-dopant relies on adventitious water and to obtain reproducible doping levels it is likely desirable to use a well-dened and intentionally synthesized BCF(OH 2 ) complex. However, in the presence of additional adventitious water the Brønsted acidity (and thus oxidant strength) of BCF(OH 2 ) is likely decreased. In addition, BCF(OH 2 ) decomposes to (C 6 F 5 ) 2 BOH and C 6 F 5 H on heating, 55 potentially leading to an ill-dened mixture of species in doping solutions or doped lms. It will be useful to carry out further work to identify other Brønsted acids that may be used as effective dopants and that avoid some of these drawbacks.</p>

Royal Society of Chemistry (RSC)

The impact of cross-docked poses on performance of machine learning classifier for protein–ligand binding pose prediction

Structure-based drug design depends on the detailed knowledge of the three-dimensional (3D) structures of protein–ligand binding complexes, but accurate prediction of ligand-binding poses is still a major challenge for molecular docking due to deficiency of scoring functions (SFs) and ignorance of protein flexibility upon ligand binding. In this study, based on a cross-docking dataset dedicatedly constructed from the PDBbind database, we developed several XGBoost-trained classifiers to discriminate the near-native binding poses from decoys, and systematically assessed their performance with/without the involvement of the cross-docked poses in the training/test sets. The calculation results illustrate that using Extended Connectivity Interaction Features (ECIF), Vina energy terms and docking pose ranks as the features can achieve the best performance, according to the validation through the random splitting or refined-core splitting and the testing on the re-docked or cross-docked poses. Besides, it is found that, despite the significant decrease of the performance for the threefold clustered cross-validation, the inclusion of the Vina energy terms can effectively ensure the lower limit of the performance of the models and thus improve their generalization capability. Furthermore, our calculation results also highlight the importance of the incorporation of the cross-docked poses into the training of the SFs with wide application domain and high robustness for binding pose prediction. The source code and the newly-developed cross-docking datasets can be freely available at https://github.com/sc8668/ml_pose_prediction and https://zenodo.org/record/5525936, respectively, under an open-source license. We believe that our study may provide valuable guidance for the development and assessment of new machine learning-based SFs (MLSFs) for the predictions of protein–ligand binding poses. Supplementary InformationThe online version contains supplementary material available at 10.1186/s13321-021-00560-w.

the_impact_of_cross-docked_poses_on_performance_of_machine_learning_classifier_for_protein–ligand_bi

Introduction<!><!>PDBbind-ReDocked<!>CASF-Docking<!>PDBbind-CrossDocked-Core<!>PDBbind-CrossDocked-Refined<!>Feature calculation<!>NNscore<!>ECIF and ELEM<!>E3FP<!>Validation methods<!>Random splitting<!>Refined-core splitting<!>threefold clustered-cross validation (CCV)<!>Model construction<!>Baselines<!>Evaluation metrics<!><!>Comparison based on the random splitting of the re-docked poses<!><!>Comparison based on the threefold clustered-cross validation of the re-docked poses<!><!>Comparison based on the refined-core splitting of the re-docked poses<!><!>Training on re-docked poses and testing on cross-docked poses<!><!>Conclusions<!>

<p>Molecular docking is one of the core technologies in structure-based drug design (SBDD), and it has contributed enormously to drug discovery and development in the past decades [1–3]. Typically, molecular docking has two stages: (1) sampling the pose of the ligand in the binding site of a macromolecular target (usually a protein) and (2) scoring the binding strength of the ligand to the target by using a predefined scoring function (SF). Despite impressive success of molecular docking, the deficiency of SFs remains a major obstacle to the reliability of real-world applications of docking [4, 5].</p><p>Depending on different theoretical principles, existing SFs can be typically divided into four main groups: physics-based, empirical, knowledge-based, and newly-emerged machine learning-based SFs (MLSFs) [6]. The former three can be collectively referred to as classical SFs, since all of them follow an additive formulated hypothesis to represent the relationship between the features that characterize protein–ligand interactions and experimental bioactivities. With the aid of state-of-the-art machine learning (ML) algorithms, MLSFs developed by automatically learning the generalized nonlinear functional forms from the training data have gradually emerged as a promising alternative to overcome the disadvantages of classical SFs. During the past few years, extensive efforts have been made towards the development of MLSFs, ranging from traditional ML-based approaches (e.g., RF-Score [7–9], NNScore [10, 11], MIEC-SVM [12], ΔVinaRF20 [13], and AGL-Score [14]) to recently-emerged deep learning (DL)-based methods (e.g., AtomNet [15], DeepVS [16], KDEEP [17], and PotentialNet [18]), and most of these MLSFs demonstrate remarkably superior performance compared with the classical SFs [19–22].</p><p>Typically, three major metrics are used to evaluate the performance of a certain SF, i.e., the ability to produce binding scores that linearly correlate with experimentally determined affinities (scoring power), the ability to discriminate near-native ligand binding pose from decoys (docking power), and the ability to identify active compounds from decoys (screening power). Classical SFs are usually constructed based on the datasets where the crystal structure and the binding affinity for each protein–ligand complex have been experimentally determined (e.g., PDBbind [23]), and then they can be generalized to either binding pose prediction or structure-based virtual screening (SBVS). However, that is not the case for MLSFs. Though most MLSFs built in a similar way exhibited better scoring power than classical SFs, their docking power and screening power are usually unsatisfactory, implying that the generalization capability of MLSFs may be still questionable [24–27]. Thus, building different MLSFs for specific tasks (i.e., binding pose prediction, binding affinity prediction or virtual screening) with the involvement of decoy poses and/or inactive compounds in the training set is a mainstream strategy rather than building a single generalized MLSF. Recently, the scoring power and screening power of a number of MLSFs have been systematically assessed [20–22, 27–35], and in this study we tend to investigate the capability of MLSFs in binding pose prediction.</p><p>Accurate identification of near-native binding poses from decoy poses is a prerequisite for many downstream simulation tasks, such as binding affinity prediction and SBVS. Over the last few years, a number of MLSFs for binding pose prediction have been reported [26, 36–46]. Some of them were trained to explicitly predict the root-mean-square-deviation (RMSD) values of binding poses, while the others were trained to directly distinguish near-native poses from high-RMSD ones. In 2015, Ashtawy and Mahapatra first developed MLSFs for the prediction of RMSD values, and the top RMSD-based SF can yield a success rate of ~ 80%, significantly higher than 70% of the best empirical SF [36]. They also found that the RMSD-based method can provide more than 120% improvement on docking task over the counterparts trained for binding affinity prediction [38]. In 2017, Ragoza et al. implemented their three-dimension (3D) grid-based convolutional neural network (CNN) architecture to build a ML classifier for pose prediction [37]. They found that the method performed consistently well in an inter-target pose prediction test, but it could hardly beat the classical Autodock Vina in an intra-target pose ranking test, which we are more concerned about. In 2020, Morrone et al. proposed a dual-graph neural network model for pose prediction, which was concatenated by a ligand-only sub-graph to store ligand structures and an interaction-based sub-graph to represent protein–ligand interaction information [40]. Similarly, this model outperformed Autodock Vina in terms of the area under the receiver operating characteristic curve (AUROC) but did not show improved performance regarding the top 1 success rate. Then, they incorporated the docking pose ranks into training as an additional feature, and the retrained model showed better performance than Vina. Despite so, a common defect of the above studies is that they can only be compared with themselves or with the classical SFs but can hardly be compared to each other because the building procedures for these models are quite different, such as different dataset partitioning methods and different pose generation (docking) methods. Besides, in most cases only the re-docked poses (re-docking the co-crystalized ligands into the pockets) were used in model training/testing, but actually, re-docking is just an artificial exercise, which completely neglects the induced fit or conformational change of the targets that occur upon ligand binding. This perfect match between protein and ligand can be hardly obtained in a real-world prospective exercise. Very recently, Francoeur et al. reported a standardized dataset named CrossDocked2020 set with 22.5 million poses generated by docking ligands into multiple similar binding pockets to better mimic the real-world scenarios, and they comprehensively estimated the scoring and docking powers of their grid-based CNN models [42]. Based on the dataset and assessment results, they further released the 1.0 version of GNINA, which could be considered as the first publicly available docking software that integrated an ensemble of CNNs as a SF [46]. However, it seems that the dataset may be not so suitable for the large-scale assessment of SFs due to its complexity and randomness. Moreover, many recent publications put more focus on DL-based models, but some traditional ML-based approaches that exhibit comparative or even better performance in many cases may also deserve attention [47–49].</p><p>In this study, two sets of descriptors that had been well validated in binding affinity prediction tasks, including the NNscore features [10, 11] and Extended Connectivity Interaction Features (ECIF) [47], were used to build the MLSFs for binding pose prediction utilizing the extreme gradient boosting (XGBoost) algorithm. In addition, the impacts of the incorporation of classical energy terms and docking pose ranks as the features on the performance of MLSFs were explored. The MLSFs were dedicatedly validated through three validation strategies, including random splitting, refined-core splitting, and threefold clustered-cross-validation. Besides the routine investigation based on the re-docked poses from PDBbind, several PDBbind-based datasets for cross-docking tests (e.g. PDBbind-CrossDocked-Core and PDBbind-CrossDocked-Refined) were constructed to investigate some important aspects of MLSFs, including their sensitivity to crystal structures, their sensitivity to docking programs, and the impacts of re-docked and cross-docked poses on each other.</p><!><p>The information of the datasets utilized in this study</p><p>aThe number in bracket refers to the number after removing the crystal poses</p><p>bThe core set of original PDBbind 2016 has 290 complexes belonging to 58 clusters, while only 285 are remained when constructing the CASF because there is a duplicated cluster</p><p>cThe set eliminates the cross-native poses</p><!><p>Previous studies were accustomed to using AutoDock Vina/Smina [52, 53] to generate docking poses due to its free of charge and acceptable accuracy, but here we used Surflex-Dock [54], one of the best-performing docking programs in our previous assessments [55, 56] to reproduce the native binding pose when the best pose with the lowest RMSD among the top 20 scoring poses was utilized as the final pose. The docking was conducted with the '-pgeom' mode, and up to 20 poses were generated for each ligand. To guarantee that each complex in the training set has at least one low-RMSD pose, the crystal poses were also mixed into the dataset, thus resulting in a total of 4057 complexes and 83,876 poses. The heavy-atom RMSD between each docking pose and the crystal pose was calculated using the obrms utility implemented in OpenBabel [57], and the poses with RMSD less than 2.0 Å were considered as near-native. Finally, the dataset (https://zenodo.org/record/5525936/files/PDBbind-CrossDocked-Core.tar.bz2) contains 39,978 positives and 43,898 negatives.</p><!><p>Comparative Assessment of Scoring Functions (CASF) [58] benchmark (http://www.pdbbind.org.cn/casf.php) based on a subset of PDBbind (core set) can be considered as a golden standard for the assessment of classical SFs, and it contains four subsets to assess the four aspects of a SF. The subset to assess docking power (CASF-Docking) contains 285 protein–ligand complexes, and ∼1000 docking poses was generated for each complex using three popular docking programs (GOLD, Surflex-Dock, and MOE Dock) to achieve the maximal conformational diversity. Finally, up to 100 poses was selected by clustering for each complex, thus generating a total of 22,777 poses (5494 positives and 17,283 negatives). The details of the pose generation process can be found in Ref [58]. This dataset serves as an external test set. It should be noted that a fairly complete coverage of the possible binding poses is provided in this dataset because multiple docking programs were utilized and a further clustering operation was conducted.</p><!><p>The 285 protein–ligand complexes in the core set of PDBbind were clustered into 57 groups by protein sequence similarity with 5 complexes in each cluster. Therefore, the complexes within each cluster were aligned using the structalign utility in Schrödinger [59], and then cross docking was carried out by docking a certain ligand in a crystal structure into the pockets of the other four crystal structures in the same cluster. In order to explore the sensitivity of MLSFs to different docking programs, besides Surflex-Dock, Glide SP [60] and AutoDock Vina were also used to generate the binding poses. For Glide SP, the receptor grids centered on the co-crystallized ligand were defined with the size of the binding box of 10 × 10 × 10 Å. For AutoDock Vina, the size of the search space was set to 30 × 30 × 30 Å, and the maximum energy difference between the best and the worst binding modes was set to 100 kcal/mol. For both programs, up to 20 poses were generated, and the other parameters were set to default. Meanwhile, the docking results were visually inspected to guarantee that the cross docking was just conducted for the complexes with the ligands in the same pockets and without residue mutations in the pockets. Of course, the complexes failing in docking were removed. The three datasets (https://zenodo.org/record/5525936/files/PDBbind-CrossDocked-Core.tar.bz2) generated by Surflex-Dock (PDBbind-CrossDocked-Core-s), Glide SP (PDBbind-CrossDocked-Core-g) and Vina (PDBbind-CrossDocked-Core-v) contain 1343 complexes and 26,410 poses (8437 positives and 17,973 negatives), 1312 complexes and 22,609 poses (5364 positives and 17,245 negatives), and 1343 complexes and 26,838 poses (1041 positives and 25,797 negatives), respectively. It should be noted that PDBbind-CrossDocked-Core-v is an extremely difficult set because most poses are marked as the negatives. This may be mainly caused by the default post-docking operations in Vina, which clusters the resulting poses using a relatively high RMSD cutoff, and therefore only a few near-native poses are finally obtained.</p><!><p>Based on the Uniprot IDs provided in PDBbind, the refined set excluding the complexes in the core set can be divided into 1302 clusters (in which 749 clusters have only one complex). The re-docking operation is the same as that for PDBbind-ReDocked, while for the clusters with more than one complex, cross docking was carried out by Surflex-Dock, thus generating a mixed dataset composed of both the re-docked and cross-docked poses (PDBbind-CrossDocked-Refined, https://zenodo.org/record/5525936/files/PDBbind-CrossDocked-Refined.tar.bz2), which contains 93,769 complexes and 1,964,686 poses.</p><!><p>Two sets of descriptors that had been well validated [10, 11, 30, 47], i.e., the NNscore features and Extended Connectivity Interaction Features (ECIF), were tested in this study, and the simple element atom-type pairwise counts (ELEM) and extended three-dimensional fingerprint (E3FP) [61] were utilized for comparison. Besides, we also tried to incorporate the Vina energy terms and docking pose ranks into the training of MLSFs.</p><!><p>NNscore proposed by Durrant et al. is a pioneer MLSF [10], and the MLSF reported by our study based on the NNscore descriptors show excellent performance to binding affinity prediction [30]. A total of 348 descriptors are used by the second version of NNscore [11], and they can encode the interaction pattern for a protein–ligand complex from multiple aspects. The five energy terms used by NNscore were directly computed by AutoDock Vina, and the other features were calculated by BINANA [62], which provide 12 different binding characteristics ranging from the number of hydrogen bonds to rough metrics of active-site flexibility.</p><!><p>ELEM is a set of simple protein–ligand atom-type pairwise counts, which was first used by RF-Score [7]. Here four types of protein atoms (C, N, O, and S) and nine types of ligand atoms (C, N, O, S, P, F, Cl, Br, and I) within 6.0 Å around the pockets were considered, and then a total of 36 features could be computed. ECIF is also a set of atom-type pairwise counts but takes each atom's connectivity into account. A total of 22 protein atom types and 71 ligand atom types, defined by atomic element, the number of explicit valences, the number of attached heavy atoms, the number of attached hydrogen atoms, whether is aromatic and whether is in a ring, are used to characterize each atom, resulting in a total of 1562 features. These two types of descriptors were calculated by the scripts based on the RDKit toolkit (version 2019.03.1) [63].</p><!><p>The E3FP [61] fingerprints are developed based on the logic of the extended connectivity fingerprints (ECFP) [64], a class of widely-used topological fingerprints based on the Morgan algorithm. Given a specific conformer, E3FP can generate a 3D fingerprint parameterized by a shell radius multiplier r and the maximum number of iterations L. Here, all the fingerprints were generated based on the docking poses with the default settings, and the hashed fingerprints with 1024 bits were ultimately generated. For this set of features, only the 3D conformations of ligands are needed, thus serving as a reference to judge whether our MLSFs could consistently learn the interaction information.</p><!><p>For the re-docked experiments, three validation methods, i.e., random splitting, refined-core splitting, and threefold clustered-cross validation (CCV), were employed, while for the cross-docked experiments, only the refined-core splitting approach was utilized. It should be noted that all the data was partitioned based on the targets rather than the poses.</p><!><p>This validation approach can best mimic the real-world scenarios because we can hardly judge whether the tested sample is novel enough. Here the whole refined set (4057 complexes) was randomly split into the training and test sets with the ratio of 4:1, and the whole operation was repeated by 10 times to yield a more convincing result.</p><!><p>The core set of PDBbind has been widely used for the evaluation of SFs, and the goal of the refined-core splitting is to have a better comparison between our SFs and the methods reported by other studies. Here the core set (290 complexes) was used as the test set and the remaining (3767 complexes) were used as the training set.</p><!><p>The aforementioned two validation methods may yield over-optimistic performance because some complexes between the training and test sets have high protein/ligand structural similarity. Hence CCV was employed to roughly estimate the generalization capability of the constructed models. The whole dataset was equally clustered into three subsets, where the proteins in different sets should have low sequence similarity and at the same time the ligands should have low structural similarity. The ligand similarity was determined by the Tanimoto similarity based on the RDKit topological fingerprints, while the sequence similarity was measured by computing the pairwise distance matrix using the pairwise2.align.globalxx module implemented in biopython [65]. The similarity thresholds for the ligands and proteins were set to 0.9 and 0.5, respectively, while 0.3 for the proteins if the cognate ligands were similar. As shown in Additional file 1: Figure S1, the samples in different subsets indeed satisfy the requirements of low sequence similarity and low ligand structural similarity. Any two sets were used as the training set and the other one as the test set, and the training and testing process was repeated 3 times. The whole operation was carried out using the script modified from clustering.py provided by Francoeur et al. [42].</p><!><p>The features with the variance less than 0.01 were removed, followed by the standardization of the remaining features using the sklearn.preprocessing [66] module. Extreme gradient boosting (XGBoost) [67], a well-validated ML algorithm that has been widely used in the field of computer-aided drug design (CADD) [28, 29, 31], was utilized to construct the classification models. Some major hyper-parameters (Additional file 1: Table S1) were tuned with the hyeropt [68] package and determined by the AUROC statistic based on the fivefold cross-validation. The maximum iteration was set to 60 with Tree Parzen estimator as the optimization algorithm. Along with the best hyper-parameter combination, the model was trained on the training set and then evaluated on the corresponding test set. XGBoost was implemented by the xgboost package [67]. In addition to the classifiers, we also built several regressors to directly predict the RMSD values. However, a simple experiment based on the refined-core splitting of the PDBbind-ReDocked dataset indicates that most regressors perform no better than their corresponding classifiers (Additional file 1: Table S2), and hence the classifiers were utilized for the following exploration. The seeds for xgboost and hyperopt were fixed to 2399 and 123, respectively, for others to reproduce our work.</p><!><p>Besides the ML-based approaches mentioned above, we also utilized several classical methods for comparison, including the docking scores from Surflex-Dock (or AutoDock Vina or Glide SP), the Vina scores extracted from the NNscore features, empirical SF X-Score [69] and more robust Prime-MM/GBSA [53]. For X-Score, the FixPDB and FixMol2 utilities were first utilized to prepare the protein and ligand files, respectively, and then the average score of the three individual SFs available in X-Score was employed for rescoring the binding poses. Prime-MM/GBSA was executed with the prime_mmgbsa utility implemented in Schrödinger. The rescoring was conducted with the variable-dielectric generalized Born (VSGB) solvation model and OPLS2005 force field.</p><!><p>With the predicted probabilities/scores obtained from the ML classifiers/classical SFs, AUROC and Spearman's rank correlation coefficient (Rs) could be calculated to evaluate the ranking capabilities of the MLSFs. The ROC curve that describes the relationship between true positive rate and false positive rate can indicate how well a model is able to distinguish low-RMSD poses from incorrect poses overall, and the corresponding AUC value ranges from 0 for a complete failure to 0.5 for a random prediction to 1 for a perfect classification, while Rs can quantitatively represent the correlation between the pose ranks predicted by each model/SF and their RMSD values. Here we defined two types of metrics, including inter-target metrics (inter-AUROC and inter-Rs) and intra-target metrics (intra-AUROC and intra-Rs). The former is computed directly based on all the tested complexes and poses, and can to some extent reflect the overall ranking capability of the models/SFs for all the groups of protein–ligand binding poses. The latter is calculated just within a specific protein–ligand complex (here at most 20 poses for a certain ligand), and then the average of all complexes is utilized to represent the final results.</p><p>The most important and intuitive metric for docking power should be the success rate (SR). For a certain complex, if one of the RMSD values of the top-ranked poses is below the predefined cutoff (usually 2.0 Å [58]), this complex can be marked as a successful prediction for the given MLSF. The analysis was performed over all the complexes in the test set, and then an overall SR was obtained by calculating the percentage of the successful cases among all the cases. For both the re-docked and cross-docked poses, the poses generated by a single docking campaign were considered as a single set, and the resulting success rate was used as the main metric. Besides, when evaluating the cross-docked poses, we also tried to consider all the docked poses of a ligand across multiple crystal structures as a single set, using the idea of ensemble docking just as Francoeur et al. did in their study [42]. As all the MLSFs developed in this study were designed for rescoring, we specifically focused on the situation when just the top 1 poses were used to calculate the success rate (SR1). Of course, the SR involving the top 3 poses (SR3) was also provided for reference.</p><p>For the random splitting and threefold CCV, the averages and standard deviations were directly obtained for analyses, while for the refined-core splitting, the random sampling of 1000 redundant copies with replacement was conducted for each statistic and the average score was calculated. In addition, we employed the Wilcoxon signed-rank test to judge whether the difference between any two compared methods was statistically significant. The difference with the P-value less than 0.05 was considered to be statistically significant.</p><!><p>The performance in terms of A, D inter-AUROC and intra-AUROC, B, E inter-Rs and intra-Rs, and C, F top1 and top3 success rates based on the random splitting of the PDBbind-ReDocked dataset. For A–C, the crystal poses in the test set is remained while for D–F is removed. The error bars represent the standard deviation of the 10 repetitions, and the dotted lines in C and F indicate the ceiling of the success rate</p><!><p>As shown in Fig. 1D, E, in terms of AUROC and Rs, the ML-based models own absolute superiority over the classical methods. This can be majorly attributed to the training way of those MLSFs, which is executed by maximizing the overall classification capability to distinguish the correct poses from incorrect ones. However, as for the classical methods, the scores of the poses for a certain complex are often quite close while those of the poses for different complexes vary considerably, and therefore in some cases the score of the best pose for a certain complex is even lower than that of the worst pose for another complex. And it can also partially account for why the inter-statistics values for MLSFs are usually higher than the intra-statistics while those for classical SFs are often in the opposite. Among all the MLSFs, E3FP_XGB inevitably performs the worst, with its inter-AUROC (0.662) slightly higher than randomness but its intra-AUROC (0.536) close to randomness, suggesting that the model can consistently learn the differences between the pure ligand binding poses for different complexes but can hardly learn effective information from the intra-difference among a set of poses for a certain complex. According to the AUROC and Rs values, the inclusion of the classical Vina energy terms can surely improve the performance (e.g., the inter-AUROC values for NNscore_XGB and NNscore-Vina_XGB are 0.863 and 0.824, respectively, and those for ECIF_XGB and ECIF + Vina_XGB are 0.851 and 0.878, respectively), but the further incorporation of the docking pose ranks did not improve the predictions anymore (e.g., the inter-AUROC values for NNscore_XGB and NNscore + Rank_XGB are 0.863 and 0.864, respectively).</p><p>As for the success rate, a more important and intuitive statistics, although it shows a substantially similar trend to AUROC and Rs, some differences can still be observed. Among all the classical methods, Prime-MM/GBSA shows the best performance (SR1 = 0.690). Our MLSFs can hardly beat Prime-MM/GBSA unless the Vina energy terms are included, as shown by NNscore_XGB (SR1 = 0.715) and ECIF + Vina_XGB (SR1 = 0.730). Further inclusion of the docking pose ranks can only slightly improve the prediction, and finally ECIF + Vina + Rank_XGB illustrates the best docking power (SR1 = 0.736). Actually, the Vina energy terms or docking pose ranks may be regarded as a correction for the original Vina SF or pose ranking, thus not only ensuring the bottom line of performance but also having the chance to gain improvements through the interaction information learnt from the training data.</p><!><p>The performance in terms of A inter-AUROC and intra-AUROC, B inter-Rs and intra-Rs, and C top1 and top3 success rates based on the threefold clustered cross-validation (CCV) of the PDBbind-ReDocked dataset. A more intuitive comparison of different validation methods can be found in D–F taking NNscore_XGB as an example. The error bars represent the standard deviation of the 3 repetitions (10 repetitions for 4:1 random splitting). The dotted lines in C indicate the ceiling of the success rate, and those in F for 4:1 random splitting, threefold random CV, and threefold CCV are colored in black, purple, and blue, respectively</p><!><p>The performance decrease is especially prominent for the atomic pairwise counts-dominant methods, such as ECIF_XGB (SR1 = 0.686 vs 0.507), ELEM_XGB (SR1 = 0.559 vs 0.473) and NNscore-Vina_XGB (SR1 = 0.662 vs 0.562). ECIF_XGB and NNscore-Vina_XGB outperform Vina for the random splitting, but here their performance is remarkably poor, highlighting their sensitivity to the similar samples in the training set. Consistent with our previous study focusing on binding affinity prediction [30], the involvement of the Vina energy terms can alleviate this sensitivity, thus guaranteeing the bottom line of performance. Hence, it can also be expected to incorporate the energy terms from a more reliable classical SF rather than Vina in order to further improve the docking power and generalization capability of MLSFs.</p><!><p>The performance in terms of A inter-AUROC and intra-AUROC, B inter-Rs and intra-Rs, and C top1 and top3 success rates based on the refined-core splitting of the PDBbind-ReDocked dataset. The error bars represent the standard deviation of the random sampling of 1000 redundant copies with replacements, and the dotted line in C indicates the ceiling of the success rate</p><p>The performance in terms of A top1, top2 and top3 success rates, and B binding funnel analysis when the models are trained on the PDBbind-ReDocked-Refined set and tested on CASF-Docking set. The RMSDs utilized in the original CASF benchmark are employed here. The methods colored in pink and orange are calculated in this study, while the predicted scores of the others are just copied from original paper. The averages of the random sampling of 1000 redundant copies with replacements are shown in the figure. For binding funnel analysis, the x-axis indicates the RMSD range (e.g., [0–2 Å], [0–3 Å], etc.) where the Spearman correlation coefficients between the RMSD values and the predicted scores are calculated</p><!><p>According to the binding funnel analysis shown in Fig. 4B and Additional file 1: Figure S1B, the superiority of the MLSFs seems more obvious. The aim of the binding funnel analysis is to estimate the rank correlation between the RMSD values and the predicted scores, which is similar to the Rs statistics described above. The only difference is that it further divides RMSD values into several windows, such as [0–2 Å], [0–3 Å], etc., to conduct a more comprehensive analysis. Compared with the top-ranked classical SFs in terms of the top1 success rate, both ECIF + Vina_XGB and NNscore_XGB do not show worse predictions and are significantly better if more high-RMSD poses are involved for analysis.</p><p>Another interesting finding is that the atomic pairwise counts-dominant MLSFs (such as ECIF_XGB and NNscore-Vina_XGB) tend to have better performance when more high-RMSD poses are included, suggesting their capability to recognize those extremely incorrect binding poses (e.g., the poses far from the binding pockets), while the energy term-centered methods (such as Vina_XGB and some of the classical SFs) have higher correlation coefficients among those low-RMSD poses, suggesting that they have better capability to rank high-quality binding poses. Moreover, the combination of these two types of features can result in more powerful classifiers, such as ECIF + Vina_XGB and NNscore_XGB.</p><!><p>The top1 and top3 success rates of the models trained on PDBbind-ReDocked-Refined and tested on PDBbind-CrossDocked-Core-s, based on A all poses, B re-docked poses, and C cross-docked poses. The error bars represent the standard deviation of the random sampling of 1000 redundant copies with replacements, and the dotted line indicates the ceiling of the success rate</p><p>The top1 and top3 success rates of the models trained on PDBbind-ReDocked-Refined and tested on the A cross-docked and D re-docked poses in PDBbind-CrossDocked-Core-s, the B cross-docked and E re-docked poses in PDBbind-CrossDocked-Core-g, and the C cross-docked and F re-docked pose in PDBbind-CrossDocked-Core-v. The error bars represent the standard deviation of the random sampling of 1000 redundant copies with replacements, and the dotted line indicates the ceiling of the success rate</p><!><p>Taken together, it seems that those ML-models trained on the re-docked poses can be well generalized to the re-docked or cross-docked poses generated by the same docking program. For the pose space defined by other docking programs, their performance is limited, especially for the predictions of cross-docked poses. Hence, a feasible strategy is to enlarge the training set, either through the augmentation and the diversification of the pose space for a certain complex or through the involvement of more complexes in the training set.</p><!><p>The impacts of the inclusion of the cross-native poses in training set on the top1 and top3 success rates of the models trained on PDBbind-CrossDocked-Refined and tested on CASF-docking. The error bars represent the standard deviation of the random sampling of 1000 redundant copies with replacements</p><p>The impacts of the contents of the training set on the top1 success rates of the models trained on PDBbind-CrossDocked-Refined and tested on the A cross-docked and D re-docked poses in PDBbind-CrossDocked-Core-s, the B cross-docked and E re-docked poses in PDBbind-CrossDocked-Core-g, and the C cross-docked and F re-docked poses in PDBbind-CrossDocked-Core-v. The error bars represent the standard deviation of the random sampling of 1000 redundant copies with replacements</p><p>The top1 success rates of the models trained on PDBbind-CrossDocked-Refined and tested on the cross-docked poses in A in PDBbind-CrossDocked-Core-s, B PDBbind-CrossDocked-Core-g, and C PDBbind-CrossDocked-Core-v using the ensemble strategy. The error bars represent the standard deviation of the random sampling of 1000 redundant copies with replacements</p><!><p>Herein, several XGBoost-based classifiers designed for protein–ligand pose predictions were carefully validated through three rigorous validation methods. When both the training and test sets contain the re-docked poses, our MLSFs can surely exhibit superior performance to the classical methods, whichever based on the random splitting or refined-core splitting, or even tested on the dataset where the poses own a large coverage of RMSD distribution (i.e., CASF-docking). But as a common feature of the ML-based methods, the sequence/structural similarity of the proteins and ligands between the training and test sets consistently exerts a notable influence on the performance of MLSFs, which is reflected by a significant decrease of the performance of those methods when using the threefold CCV. However, although our best MLSF performs no better than Prime-MM/GBSA, it can still beat other commonly-used classical methods. Then, the models are also estimated with the involvement of the cross-docked poses in either the training or the test set. It seems that the ML models trained on the pure re-docked poses can only be well generalized to the re-docked/cross-docked poses produced by the same docking program used for the training set, but they cannot always outperform the classical methods when tested on the cross-docked poses generated by different docking programs. The incorporation of the cross-docked poses into the training set is favorable to enhance the performance on the cross-docked poses, but for the test sets with only the re-docked poses, the expansion of the training set by adding cross-docked poses sometimes can hardly counteract the influence of the pose quality. Besides the impacts of datasets, this study also demonstrates the importance of the inclusion of the classical energy terms or docking pose ranks as the features in binding pose prediction task, which can not only further improve the docking power to some extents but can ensure the generalization capability of the models.</p><p>Introducing ML technologies into SFs has emerged as a promising trend in recent years, but most relevant studies seem to pay more attention to binding affinity prediction or SBVS, rather than binding pose prediction, which has not been well achieved by traditional methods and has long been an important limiting factor for the further performance improvements of the former two tasks. As a supplement to the study conducted by Francoeur et al. [42], our study adopted a different way to handle the docking poses and employs a more direct way to validate the models. In addition, we further developed several pure PDBbind-based datasets, namely PDBbind-ReDocked (https://zenodo.org/record/5525936/files/PDBbind-CrossDocked-Core.tar.bz2), PDBbind-CrossDocked-Core (https://zenodo.org/record/5525936/files/PDBbind-CrossDocked-Core.tar.bz2), and PDBbind-CrossDocked-Refined (https://zenodo.org/record/5525936/files/PDBbind-CrossDocked-Refined.tar.bz2), for cross-docking experiments, which can be easily combined with the widely-used CASF benchmark/PDBbind dataset to conduct a more comprehensive assessment of SFs. Our study may provide sufficiently valuable guidance for the applications of MLSFs in binding pose prediction. Moreover, our datasets may serve as an important benchmark for further development and assessment of the MLSFs for protein–ligand binding pose prediction.</p><!><p>Additional file 1: Additional Tables and Figures.</p><p>Computer-aided drug design</p><p>Structure-based drug design</p><p>Structure-based virtual screening</p><p>Three-dimensional</p><p>Scoring function</p><p>Machine learning</p><p>Deep learning</p><p>Convolutional neural network</p><p>Machine learning-based scoring function</p><p>Extended Connectivity Interaction Features</p><p>Extended three-dimensional fingerprint</p><p>Extended connectivity fingerprint</p><p>Root-mean-square-deviation</p><p>Comparative Assessment of Scoring Functions</p><p>Clustered-cross validation</p><p>Extreme gradient boosting</p><p>Area under the receiver operating characteristic curve</p><p>Spearman's rank correlation coefficient</p><p>Success rate</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

Direct Mass Spectrometry Analysis of Complex Mixtures by Nano-Electrospray with Simultaneous Atmospheric Pressure Chemical Ionization and Electrophoretic Separation Capabilities

Accuracy and rapid analysis of complex microsamples are challenging tasks in translational research. Nano-electrospray ionization (nESI) is the method of choice for analyzing small sample volumes by mass spectrometry (MS) but this technique works well only for polar analytes. Herein we describe a versatile dual non-contact nESI/nAPCI (nano-atmospheric pressure chemical ionization) source that allows simultaneous detection of both polar and nonpolar analytes in microliter quantities of samples under ambient conditions and without pre-treatment. The same device can be activated to enable electrophoretic separation. The non-contact nESI/nAPCI MS platform was applied to analyze different samples, including high sensitive direct analysis of biofluids and the efficient detection of proteins in buffers with high concentration of nonvolatile salts. Excellent linearity, accuracy and limits of detection were achieved for compounds with different chemical properties in different matrices. The high sensitivity, universality, simplicity and ease of operation make this MS technique promising for use in clinical and forensic applications.

direct_mass_spectrometry_analysis_of_complex_mixtures_by_nano-electrospray_with_simultaneous_atmosph

<!>Non-contact nESI/nAPCI Apparatus.<!>Instrumentation.<!>Chemicals, Reagents, and Samples.<!>Stability of Glass Tip at High Voltages.<!>Detection of Polar and Nonpolar Compounds.<!>Direct Biofluid Analysis.<!>Electrophoretic Separation.<!>CONCLUSIONS

<p>One of the challenges of translational research is the availability of reliable quantitative assays that have the required sensitivity and accuracy to evaluate complex biological samples in a high throughput fashion. Mass spectrometry (MS) has shown great promise in this respect, allowing quantification of both biomolecules and small organic compounds.1-3 It is well recognized that sample preparation is the slowest step of the chemical analysis process and so the recent introduction of ambient ionization methods, which integrate analyte extraction and ionization into a single step, represents an important advancement for achieving rapid analysis. Aside from quantitation and high throughput requirements, another important merit in the biomedical field is the analysis of microsamples (<50 μL) with minimal dilution. Nano-electrospray ionization (nESI) MS is one of the most efficient methods for small volume analysis.4,5 With appropriate modifications, it can be used to analyze untreated complex mixtures.6,7</p><p>An unmet need is for the integrated, robust, and versatile nESI system that can quantitatively and rapidly ionize polar and non-polar organic compounds, and large biomolecules in various matrices. Herein, we describe a simple integrated approach that combines non-contact nESI, a novel nAPCI (nano-atmospheric pressure chemical ionization, explained below), and electrophoretic separation in one complimentary analytical tool to make possible a more complete analysis of structurally diverse compounds from small complex mixtures. The results indicate that a non-contact mode nESI allows the application of >4 kV spray voltage without damaging the tip of the glass emitter. In the presence of an auxiliary electrode (e.g., Ag), placed in close proximity to the emitter tip, the applied high voltage induces corona discharge in ambient air, which allows for simultaneous formation of protonated (M+H)+ and molecular ions (M•+) via atmospheric pressure chemical ionization (APCI). We suggest referring to this ionization process as nAPCI because it ionizes gas-phase species delivered at a ~10 nL/mL flowrate from a conventional direct infusion nESI setup. Unlike the traditional APCI method, the low flowrates in the nAPCI process allows effective desolvation and molecular ion generation without the use of nebulizer gas. The elimination of heated nebulizer gas in turn enabled a more simplified and compact experimental setup, with the corona discharge needle (i.e., the auxiliary Ag electrode) coaxial with the direction of the nanodroplets. Corona is not formed at low spray voltages (≤3 kV), hence protonated (M+H)+ or deprotonated (M−H)− ions are predominantly generated enabling the detection of large biomolecules like proteins by the same ion source.</p><p>Many studies have investigated various aspects of the nESI setup, including (i) the mode by which the analyte solution is electrically charged (i.e., contact versus non-contact),8 (ii) the source/nature of the electrical energy (e.g., piezoelectric discharge, triboelectric nano-generator, pulsed DC/AC voltage and square-wave potential),9-12 (iii) flowrate manipulation to control ion suppression and sample consumption (e.g., via the use of smaller tip size13,14 or on-demand pulsed charges11), (iv) reduction of electrical current (via the use of high input ohmic resistance) to avoid destructive corona discharge phenomenon when electrospraying under high voltage conditions,15,16 and (v) the use of other operational tricks like step voltage and polarity reverse applications.17,18 None of these methods are completely adequate, especially for simultaneous generation of different ion types. There appear to be one case where the alternative operation of independent nESI and APCI sources allowed for the generation of ions of opposite polarities for ion/ion reactions. Unlike our proposed approach, that study used the conventional contact-mode nESI and two separate HV power supplies.19 Hybrid ESI and APCI ion sources have been developed20,21 but they are designed for larger volumes of samples, and often combined with liquid chromatography and/or electrophoresis; the corona discharge is not applied to the mixture of gas analytes derived directly from droplets as is done here. In addition to the integration of experimental steps and small sample consumption, the proposed dual non-contact nESI/nAPCI method is capable of in-situ, in-capillary liquid/liquid extraction,6 which enables part-per-trillion (pg/mL) level sensitivity for cocaine and part-per-billion (ng/mL) detection limit for a non-polar β-estradiol analyte, all from untreated whole human blood. The adapted in-capillary liquid/liquid extraction procedure facilitates microsamples (5 μL) processing and analysis by MS, compared with offline extraction procedures that require sample dilutions and additional transfer steps. Analysis of the dilution sample certainly will require a more sensitive instrument, one that might not be readily available. The potential applications of this non-contact nESI/nAPCI MS approach may include omics fields, in various steps of drug development (pre-clinical research) and in clinical applications (e.g., newborn screening, forensic toxicology, therapeutic drug monitoring, etc.), all requiring minimally invasive microsampling.</p><!><p>The dual non-contact nESI/nAPCI experimental setup is as shown in Figure 1a, and is capable of three spray modes: a) Non-contact nESI in which the analyte solution present in a disposable glass capillary (ID 1.2 mm; ≤5 μm pulled tip, Fig. S1) is electrically charged through electrostatic induction.22,23 That is, the Ag electrode on which the DC high voltage (HV) is applied is not in physical contact with the analyte solution. Instead, a ~1 cm air gap is created, and as little as 1 kV applied voltage is able to induce electrostatic charging, which causes the release of charged droplets from the capillary tip that are sampled by the mass spectrometer. b) Non-contact nESI/nAPCI mode, where both charged droplets and plasma are simultaneously produced when potentials above the breakdown voltage (4 kV) of air are applied. Here, the presence of auxiliary Ag electrode placed in a collimating glass capillary (ID 1.2 mm) allows the exposure of the resultant solvated/gas-phase ions to corona discharge. Note: a single HV power supply (available from the MS instrument) is used, plus no further modification of the conventional nESI source is required except for the attachment of the auxiliary Ag electrode, which does not obstruct nESI performance at low spray voltages. c) Electrophoretic separation spray mode in which polarity reversing (from negative to positive voltage) enables detection of highly resolved multiply-charged protein ions under high voltage conditions in the presence of concentrated inorganic salts.</p><!><p>Sample analysis was carried out with a Velos Pro LTQ mass spectrometer using Xcalibur 2.2 SP1 software (Thermo Fisher Scientific, San Jose, CA, USA). Applied MS parameters were as follows: 400 °C capillary temperature; spray voltage (2kV for nESI, 6 kV for non-contact nESI/nAPCI, and alternate from −5 kV to 2 kV for electrophoretic separation); 60% S-lens voltage; 5 mm distance from an ion source to MS inlet; 3 microscans; 100 ms ion injection time; 30 s spectra recording time. Analytes were identified by tandem MS with collision induced dissociation (CID).</p><!><p>Caffeine (99.0%), 3,5-dimethyl-1-hexyn-3-ol (surfynol 61, 98%), docosanoic acid (99%) ergocalciferol (vitamin D2, 98%), β-estradiol (98%), ethyl acetate (99.8%), 5-fluorouracil (99%), lauric acid (99.5%), linoleic acid (99%), methanol (99.9%), oleic acid (99%), stearic acid (95%), and thymol (98.5%) were all purchased from Sigma Aldrich (St. Louis, MO, USA). Phenol (99%), and stereo 2X-4X microscope (S 15019975) was provided by Fisher Scientific (Pittsburgh, PA, USA). 1.0 mg/mL standard solution of cocaine and 100 μg/mL of cocaine-D3 were acquired from Cerilliant (Round Rock, TX, USA). Di-chloromethane (99.9%) was supplied by Acros Organics (Geel, Belgium). Both single donor human blood and single donor human serum were obtained from Innovative Research (Novi, MI, USA). Bovine blood was provided by LAMPIRE Biological Laboratories (Pipersville, PA, USA). 18.2 MΩ water was used for water solutions (Milli-Q water purification system, Millipore, Billerica, MA, USA). Borosilicate capillaries (ID 1.17 mm) were provided by Sutter Industries (Novato, CA, USA).</p><!><p>Figure 1b compares tip stability under different spray conditions. Joule heating24 (among other factors) generated after applying 5-8 kV to an electrode in contact with analyte solution (conventional nESI) is mainly responsible for the tip burning/breakage observed here. For example, the electric field at the emitter apex is a key parameter in controlling electrospray current, which in turn determines the intensity of heat generated. Therefore, heating can be expected to be more effective at the tip of the glass capillary. To investigate the effect of field strength on heat generation in our experiment, we monitored the extent of tip damage/burning as a function of spray distance (1 – 50 mm) from the glass tip to the inlet of the mass spectrometer while keeping the spray voltage constant at 6 kV in the contact mode nESI experiment. As expected, we observed a reduced burning effect as the spray distance was increased (Fig. S2a). For 1 – 25 mm distances, ion signal was stable for ~30 s, after which an obvious boiling of the aqueous analyte solution occurred (indicated by visible bubble formation due to extensive Joule heating; see Video S1). This period was followed immediately by the burning of the glass tip, resulting in concomitant termination of ion signal (Fig. S2b; also see Video S2). Additional close examinations indicated that tiny bubbles begin to form at ~3 s following the application of voltage (Video S1) and bright glow (electrical discharge) formed at the tip follows after 10 s later. Physical inspections of the tip revealed partial cracks at ~14 s of electrospray under the contact mode conditions. Based on these empirical observations we propose the following mechanism: boiling of analyte solution happens first, followed by burning, which leads to partial breakage (cracking) of the glass tip. These events reduce solvent flow at the tip and cause the observed electrical discharge while Joule heating exacerbates. The combined effects of discharge and heating result in visible damage of the glass tip. Glass tips remained stable at 50 mm spray distance but signal intensity decreased by two orders of magnitude due to inefficient ion transfer. Joule heating significantly reduced in the non-contact spray mode due to the presence of the air gap (resistivity of air is >1.3 × Ω at 200 °C), which led to much more stable tips at the same applied voltages. Interestingly, the glass tips became remarkably stable in the presence of the proximal auxiliary Ag electrode (non-contact nESI/nPACI, Fig. 1b). In this case, the well-known cooling effects of corona discharge further reduces Joule heating by inducing rapid movement of air/droplets around the tip area.25-26 As will be shown, the ability of non-contact nESI/nAPCI setup to tolerate high voltages could be a useful feature in MS analysis.</p><!><p>Figure 2a shows positive-ion mass spectrum recorded after the analysis of a methanol solution containing equimolar (200 μM) mixture of 5-fluorouracil (1), caffeine (2), β-estradiol (3), cocaine (4), and vitamin D2 (5) using the conventional contact mode nESI source at an applied voltage of 2 kV. As can be observed, only the polar cocaine analyte with high proton affinity (930 kJ/mol) was detected at m/z 304. Caffeine (MW 194), another polar analyte was significantly suppressed despite having relatively high proton affinity (914 kJ/mol). Not surprisingly, detectable ion signal was not observed for 1, 3 and 5, even from individual solutions (i.e., in the absence of other analytes) at 10 ppm concentration levels (Fig. S3). Similarly, protonated cocaine ions were predominantly detected when the mixture was analyzed by non-contact nESI operated using 2 kV spray voltage in the absence of corona discharge (data not shown). Upon increasing the voltage from 2 to 6 kV, corona discharge was induced on the auxiliary Ag electrode, expecting the ionization of both polar and non-polar compounds delivered by the spray plume. The corresponding non-contact nESI/nAPCI positive-ion mass spectrum is shown in Figure 2b, which confirms the presence of all five analytes. Compounds 1, 2, and 4 were observed as protonated (M+H)+ ions at m/z 131, 195, and 304, respectively. Like conventional APCI experiment,27 dehydration reactions involving (pseudo) molecular ions were also observed with β-estradiol (MW 272) registering as [M+H−H2O]+ species at m/z 255. Radical species M•+ and (M−H2O)•+ were also detected for vitamin D2 (MW 397) at m/z 397 and 379, respectively. Other nonpolar compounds (carminic acid, thymol, surfynol 61, phenol), which could not be detected by conventional nESI at 1 ppm level, were also successfully characterized by non-contact nESI/nAPCI tandem MS (Table 1). These results establish the integrated non-contact nESI/nAPCI MS platform as efficient method to simultaneously ionize both polar and nonpolar compounds simply by increasing spray voltage from 2 to 6 kV.</p><!><p>These encouraging results motivated us to explore direct biofluid analysis. Here, 3 μL of ethyl acetate was first placed in the sharp tip of the disposable glass capillary. The organic ethyl acetate solvent was chosen because it is immiscible in water, and prior studies6,23 have shown it to have high solubilizing power for a wide range of organic compounds and it is suitable for electrospray. A small volume (5 μL) of the biofluid sample spiked with a selected analyte (Table 2) was then introduced on top of the ethyl acetate solvent (Fig. 3a), followed by a short shake (e.g., 1 – 3 strokes) to initiate liquid-liquid extraction in the capillary as well as to remove air bubbles that may be present at the capillary tip. Ion signal for extraction analytes was not significantly affected by different number (n>3) of strokes (data not shown). Note that the three strokes of shaking employed here form part of the regular nESI MS analysis, and do not add extra steps to the analytical process. Often, the shaking process resulted in the disintegration of the biofluid into smaller compartments (Fig. 3b), which facilitated efficient extraction via increased interfacial contact with the extracting organic solvent. The high buoyancy of the less viscous ethyl acetate solvent (density 0.9 g/mL) draws the clean extract to the sharp tip of the glass capillary for easy analysis by non-contact nESI/nAPCI MS. Moreover, since the Ag electrode is not in direct contact with sample/solvent, extraction equilibrium is not disturbed; a contact mode experiment where the electrode is pushed through the biofluid will reintroduce contaminants into the extract, which may cause matrix effects during analysis. The pure extract present at the tip of the glass capillary typically offered a stable 1 min spray time (Fig. 3c), which is sufficient for complete MS analysis, including tandem MS (MS/MS). Unless otherwise stated, internal standard (IS) used for analyte quantification was added to the biofluid. The optimal amount of extraction solvent (3 μL) was used to compromise between spray time and signal intensity. For instance, applying 3 μL versus 5 μL of ethyl acetate increased analyte to internal standard (A/IS) signal ratio for cocaine extracted from serum by a factor of 10 (here, IS was spiked into the extraction solvent to keep the amount of IS constant, Fig. S4). This increase in signal might be due to a number of related factors, which include: i) pre-concentration effects, arising from the large K (partitioning coefficient = 794 in chloroform, which is similar by polarity to ethyl acetate) value for cocaine between water and the organic solvent. In absolute terms, the amount of cocaine extracted into 3 μL solvent is expected to be ~2X more than that extracted into 5 μL μL; ii) it was observed that the volume of clean extract pooled to the emitter tip (Fig. 3b), after mixing with 5 μL biofluid, was comparable regardless of the initial volume of organic solvent used. For higher volumes (e.g., 5 μL), significant amount of the excess extract was found trapped at the back of the biofluid that cannot be accessed during electrospray. This effect decreases the concentration of analyte in the extract residing at the capillary tip for larger solvent volumes and hence reduced A/IS ratio; iii) effective saturation of organic solvent with water in smaller volumes. We determined that ethyl acetate saturated with water (2% maximum) is a much better electrospray solvent than the pure organic solvent (see SI, Fig. S5, for details). Therefore, we expect the doping of water from the biofluid into 3 μL volume to be more efficient during the liquid/liquid extraction process, which can contribute to the enhanced signal observed. Initial volumes of solvent lower than 3 μL result in decreased amount of extract pooled to the tip of the glass capillary, which in turn reduced spray times (< 1 min).</p><p>Representative product ion spectrum for 50 pg/mL cocaine spiked in undiluted blood (5 μL) is shown as insert in Figure 3d, which registered the diagnostic fragment ion at m/z 182 in high abundance. Figure 3d shows a calibration curve derived from comparing the product ion (m/z 182) intensity at different concentrations of cocaine analyte (50 – 1000 pg/mL) to that of internal standard (IS, cocaine-d3, 500 pg/mL) spiked into the blood sample. Excellent linearity (R2 = 0.999) and limit of detection (LOD) of 12 pg/mL were achieved. LODs for other analytes are shown in Table 2, which include 0.5 pg/mL sensitivity for cocaine in serum. Aside from high extraction efficiency (controlled by pre-concentration effect), high ionization efficiency, and minimal matrix effects, additional enhancing effect may arise from the smaller initial droplets of ethyl acetate expected from the low flowrate (~10 nL/min) non-contact mode nESI experiment. (Comparable tip size of μμm yields the typical ~30 nL/min flowrate in traditional nESI,28 Figs. S6 – S8).</p><p>The smaller flowrates derived from the highly volatile ethyl acetate spray solvent facilitated analyte introduction into the gas phase for efficient APCI ionization. On the contrary, we observed a markedly reduced ionization efficiency for aqueous-based samples (Fig. S9) presumably because of limited solvent evaporation from the aqueous droplets, which reduced the efficiency of ion formation via chemical ionization. Flowrates (Fig. S7) for non-contact nESI/nAPCI at high voltages (≥4 kV; Fig. S2) were surprisingly similar to those for non-contact nESI recorded at low voltages (≤2 kV). Although this requires further investigation, it might be related to differences in the shape of the electric field at the emitter tip, when comparing the presence (high voltage) and absence (low voltage) of corona discharge from the auxiliary electrode. Another factor influencing ionization efficiency, and hence sensitivity, is our ability to generate different ion types simply by using higher spray voltages. For example, the weakly polar and high eluent strength (0.58) properties of ethyl acetate is expected to result in high extraction efficiency for steroid analytes such as β-estradiol (MW 272 Da). However, analysis by contact mode nESI MS often yields low sensitivity due to low proton affinity. Derivatization reactions are typically used to overcome this limitation.6,29 A 10 ng/mL LOD (Table 2) was observed for β-estradiol in whole human blood by utilizing an optimized spray voltage of 6 kV, which enable the detection of (M+H–H2O)+ ion in tandem MS (m/z 225 → 199; see Fig. S10a at the limit of detection) mode without derivatization reactions. For caffeine (MW 194 Da) in whole blood, the measured LOD was 15 ng/mL (see Table 2 and Fig. S10b) in MS/MS analysis of (M+H)+ species (i.e., m/z 195 → 177). Vitamin D2 and phenol were detected as molecular ions (M•+; Table 1).</p><p>The fact that the non-contact dual non-contact nESI/nAPCI source is operated without the assistance of nebulizer gases, and in the presence of limited solvent molecules under the nL/mL flowrate conditions suggests highly reactive ionic species [e.g., H+(H2O)n; where n = 1 or 2]30,31 might be involved in the ionization process compared with the conventional APCI experiment, which employs N2 gas and high solvent flowrates (μL/mL). Evidence for the presence of low molecular weight water clusters is seen in the fragmentation of protonated species, which could be a result of the large difference in proton affinities (PA) between analytes (e.g., β-estradiol; PA 814 kJ/mol)32,33 and the reagent ion, H3O+ Aside from gas-phase chemical reactions, we believe the present configuration of the non-contact nESI/nPACI setup (co-axial) may allow surface-assisted ionization processes. This is exemplified by the detection of [M+H2+H]+ species from methanol solution of carminic acid (Table 1), which was recently characterized to involve the hydrogenation of C═O bond for analytes adsorbed at the surface of the corona discharge electrode.34 The presence of surface-assisted reactions has profound implications on the mechanism of (M+H)+ and M•+ ions formation, which may include field ionization and field-induced proton transfer reactions (M•+(surf) + H2O → [M+H]+ + HO•). Another unique feature of the proposed setup is that biosamples can be reanalyzed by repeated cycles of in-capillary extraction and ionization. Here, similar ratios of cocaine signal to that of IS were observed in serum for seven cycles of analyses performed in the same day (Fig. S11), and for four cycles of analyses conducted over a period of one month for a sample storage in freezer (Fig. S12). The in-capillary extraction methodology adapted here does not only provide a simple means to effectively process small volume of biofluids, but it also affords a way to contain the biofluid in the enclosed environment of the glass capillary preventing possible unexpected exposure of biofluid to analyst, something that is difficult to achieve with other direct MS analysis plat-forms such as desorption electrospray ionization and paper spray. The use of ethyl acetate spray solvent further minimizes safety risks due to its antimicrobial activity.35 In addition, ethyl acetate is immiscible in aqueous samples and thus extracts only the organic components, leaving behind the bulk of the blood matrix in the glass capillary for safe disposal. Like all other spray-based atmospheric pressure ion sources, additional protective casing can be used to restrain the charged aerosols.</p><!><p>The last integrated application examined for the new ion source was electrophoretic desalting and detection of proteins in concentrated salt solutions. As demonstrated by others,36,37 we also employed polarity-reversing on our non-contact nESI/nAPCI platform where a step potential was used starting from negative to positive high voltage polarities. A unique capability provided by our experimental setup is the fact that large step voltage differences (e.g., from −5 kV to +2 kV) can be used without damaging the disposable glass tip due to reduced Joule heating. Figure 4 shows real-time separation of cytochrome c in 1X phosphate-buffered saline solution (PBS, 137 mM NaCl, 2.7 mM KCl, 10 mM Na2HPO4 and 1.8 mM KH2PO4) obtained after applying −5 kV for 10 s followed by the application of +2 kV (see insert of Fig 4a; 0.1% of formic acid was added to the buffered protein solution). Figures 4a-c show three distinct time domains during the mass analysis at +2 kV: highly charged protein species were detected first (between 10 – 31 s). After this initial spray time, a broad range of protein charge states emerged within 31 – 88 s. All the denatured bulky proteins were exhausted after 88 s of continuous spray at which point only low charge state proteins were detected for the remaining 3.7 min spray time. Overall, the solution with depleted salt lasted for about 5 min, which is sufficient for complete MS analysis. Similar desalting effect was observed for ubiquitin using −5 kV to +2 kV step voltage conditions with 2.5 min of total spray time (Figs. S13). The differences in protein charge state distribution observed at different spray times may be related to a number of factors, including changes in protein conformation during the time of ionization, which could be caused by changes in solvent composition. Protein unfolding can also occur during the 10 s electrophoretic separation period, followed by refolding events in the ionization process. It is also possible that different conformations of protein may exist in solution prior to the electrophoretic experiment, which can be separated based on differences in electrophoretic mobilities between the bulky unfolded versus folded proteins. Note: without polarity reversing, proteins could not be detected in 1X PBS buffer employing either our setup or the regular contact mode nESI source. With polarity reversing, our setup offered acceptable separation in real-time not only for the temporal desalting of biomolecules but also the spatial separation of different conformations of a single protein. The later effect has not been reported before in all other polarity-reversing experiments. In some cases, the separation can be achieved without adding acid (i.e., ubiquitin, Fig. S13).</p><!><p>In conclusion, a new non-contact nESI/nAPCI ion source is described that can be activated to operate in electrospray, atmospheric pressure chemical ionization, and/or electrophoretic separation modes. The integrated dual ionization capabilities are achieved simultaneously on a single device allowing the detection of polar and non-polar organic compounds as well as large biomolecules. The technique is highly sensitive, reliable and fast enabling the direct analyses of (i) various biofluids via an in-capillary liquid/liquid extraction process and (ii) different proteins in high concentrated salt solutions through an online electrophoretic separation method. The robustness, simplicity, and ease of operation make this method very attractive for ultra-small complex mixture analysis, with vast implications in translational and biomedical research.</p>

PubMed Author Manuscript

The Fate of Benzene Oxide

Metabolism is a prerequisite for the development of benzene-mediated myelotoxicity. Benzene is initially metabolized via cytochromes P450 (primarily CYP2E1 in liver) to benzene oxide, which subsequently gives rise to a number of secondary products. Benzene oxide equilibrates spontaneously with the corresponding oxepine valence tautomer, which can ring open to yield a reactive \xce\xb1-\xce\xb2-unsaturated aldehyde, trans-trans-muconaldehyde (MCA). Further reduction or oxidation of MCA gives rise to either 6-hydroxy-trans-trans-2,4-hexadienal or 6-hydroxy-trans-trans-2,4-hexadienoic acid. Both MCA and the hexadienal metabolite are myelotoxic in animal models. Alternatively, benzene oxide can undergo conjugation with glutathione (GSH), resulting in the eventual formation and urinary excretion of S-phenylmercapturic acid. Benzene oxide is also a substrate for epoxide hydrolase, which catalyzes the formation of benzene dihydrodiol, itself a substrate for dihydrodiol dehydrogenase, producing catechol. Finally, benzene oxide spontaneously rearranges to phenol, which subsequently undergoes either conjugation (glucuronic acid or sulfate) or oxidation. The latter reaction, catalyzed by cytochromes P450, gives rise to hydroquinone (HQ) and 1,2,4-benzene triol. Coadministration of phenol and HQ reproduces the myelotoxic effects of benzene in animal models. The two diphenolic metabolites of benzene, catechol and HQ undergo further oxidation to the corresponding ortho-(1,2-), or para-(1,4-)benzoquinones (BQ), respectively. Trapping of 1,4-BQ with GSH gives rise to a variety of HQ-GSH conjugates, several of which are hematotoxic when administered to rats. Thus, benzene oxide gives rise to a cascade of metabolites that exhibit biological reactivity, and that provide a plausible metabolic basis for benzene-mediated myelotoxicity. Benzene oxide itself is remarkably stable, and certainly capable of translocating from its primary site of formation in the liver to the bone marrow. However, therein lies the challenge, for although there exists a plethora of information on the metabolism of benzene, and the fate of benzene oxide, there is a paucity of data on the presence, concentration, and persistence of benzene metabolites in bone marrow. The major metabolites in bone marrow of mice exposed to 50 ppm [3H]benzene are muconic acid, and glucuronide and/or sulfate conjugates of phenol, HQ, and catechol. Studies with [14C/13C]benzene revealed the presence in bone marrow of protein adducts of benzene oxide, 1,4-BQ, and 1,4-BQ, the relative abundance of which was both dose and species dependent. In particular, histones are bone marrow targets of [14C]benzene, although the identity of the reactive metabolite(s) giving rise to these adducts remain unknown. Finally, hematotoxic HQ-GSH conjugates are present in the bone marrow of rats receiving the HQ/phenol combination. In summary, although the fate of benzene oxide is known in remarkable detail, coupling this information to the site, and mechanism of action, remains to be established.

the_fate_of_benzene_oxide

1. One Hundred and Forty Years, and Counting<!>2. Benzene-oxide<!>3. trans-trans-Muconaldehyde (MCA)<!>4. Phenol and Hydroquinone<!>6. Hydroquinone-Thiol Conjugates<!>7. Benzene Metabolites in Bone Marrow<!>8. Some Speculation on the Potential Involvement of HQ-GSH conjugates in Benzene-mediated Hematotoxicity

<p>The metabolic conversion of benzene to phenol was demonstrated by Schultzen and Naunyn in 1867 [1], and the conjugation of phenol with sulfate was established in 1876 by Baumann [2]. At that time, the mechanism of this metabolic conversion was not known, but following reports that anthracene [3], naphthalene [4], and phenanthrene [5] were all metabolized to dihydrodiols, Boyland [6] suggested that epoxide formation was the requisite intermediary in this metabolic reaction, simultaneously providing an explanation for the formation of other aromatic hydrocarbon metabolites (phenols, ring-opening). Boyland's suggestion occurred at about the same time (1949) that Dennis Parke joined R.T. Williams' group at St. Mary's Hospital Medical School, where he embarked on a study of all the known pathways of benzene metabolism, focusing particularly on whether benzene did indeed form an epoxide and/or a dihydrodiol, and on the isomers of muconic acid. The metabolism of the simplest of aromatic hydrocarbons has subsequently been revealed to be comparatively complex, giving rise to a myriad of metabolites.</p><!><p>The first step in the metabolism of benzene is the cytochrome P450 catalyzed formation of benzene-oxide (Figure 1). Epoxides such as benzene-oxide can be somewhat ambiguous in their nature. On the one hand, they are sufficiently electrophilic that they can react and covalently bind to nucleophilic sites within proteins and DNA. In contrast, epoxides can be relatively stable, in chemical terms, with half-lives in biological milieu ranging from seconds to minutes. For example, bromobenzene-3,4-oxide is capable of diffusing out of hepatocytes and being trapped extracellularly by glutathione (GSH) and the requisite GSH transferase (GST) [7]. Indeed, bromobenzene-3,4-oxide is sufficiently stable that it can be detected in venous blood of rats receiving bromobenzene via intra-peritoneal injections, with a half-life of ∼13.5 seconds [8], which is more than sufficient time to reach extrahepatic targets. Benzene-oxide has also been directly identified in the blood of rats administered benzene (400 mg/kg) and has a half-life of ∼8 mins when added to rat blood ex vivo [9]. Since the mean circuit time of blood in rats (blood volume/cardiac output) is about five to ten seconds, the concentration of benzene-oxide will not decrease greatly in a single pass between organs. Thus, nearly all of the benzene-oxide leaving the liver (and any other organ in which it is formed) will enter the lung, and the concentration of the epoxide leaving the lung will be essentially identical to its concentration in arterial blood entering the other organs/tissues of the body, including the bone marrow. The extent to which benzene-oxide in arterial blood contributes to its covalent binding in extrahepatic tissue, in particular bone marrow, will depend on the ability of the various tissues to convert the epoxide into downstream metabolites, and on its non-enzymatic rearrangement to phenol and ring-opened products. However, each particular non-enzymatic rearrangement reaction should occur at the same rate in all tissues. It will therefore be the relative distribution of the enzymes that catalyze the conversion of benzene-oxide into the dihydrodiol (epoxide hydrolase) and S-phenylpremercapturic acids (GST) that limit the exposure of extrahepatic tissues to blood-borne benzene-oxide.</p><p>Although liver-derived benzene-oxide is clearly capable of reaching bone marrow, is it sufficiently biologically reactive to play a major direct role in benzene-mediated hematotoxicity? The half-life of benzene-oxide in aqueous medium (95:5 [v/v] phosphate buffer in D2O with [CD3]2SO) at 25°C (pD/pH 7) was ∼34 mins [10]. Moreover, the half-life did not change in the presence of GSH (2-15 mM) or with a combination of GSH (2 mM) and GST [10]. In fact, the major product observed under these experimental conditions was phenol. These findings are consistent with the fact that S-phenylmercapturic acid is a relatively minor urinary metabolite of benzene. Perhaps the role of benzene-oxide in benzene hematotoxicity is to assist in the delivery of phenol and/or the ring-opened products (see below) to the bone marrow, where they may undergo further metabolic transformations into more reactive metabolites.</p><p>In contrast to the apparent stability of benzene-oxide, and it's inefficient reaction with GSH, S-pheny-L-cysteine, presumed to arise from the interaction of benzene-oxide with cysteine residues within proteins, has been isolated from hemoglobin and bone marrow proteins [11]. It would seem likely that the protein microenvironment increases the efficiency with which these cysteinyl thiols react with benzene-oxide. Consistent with this view, increasing the pH from 7.0 to 8.5 dramatically increases (20-fold) the reaction of GSH with benzene-oxide [10]. Nonetheless, the major protein adducts identified in mouse marrow after benzene administration were derived from 1,4-benzoquinone (see Section 4) rather than benzene-oxide (<2%) [11]. Interestingly, the metabolites representing the major source of benzene-derived protein adducts in bone marrow remain to be identified. A more in-depth analysis of this issue can be found in the following presentations.</p><!><p>Benzene oxide can equilibrate spontaneously with the corresponding oxepin valence tautomer, which can ring open to initially yield a reactive α-β-unsaturated aldehyde, cis-cis-muconaldehyde, which subsequently isomerizes to the cis-trans, and ultimately to the trans-trans-muconaldehyde (MCA) isomer. The mechanism of ring-opening remains debatable, but has been proposed to occur via cytochrome P450-mediated metabolism of the oxepin to the oxepinoxide [12] Further reduction or oxidation of MCA gives rise to a variety of products. In vitro model systems indicate that cytosolic aldehyde dehydrogenases may oxidize MCA to 6-oxo-trans-trans-hexadienoic acid, a mixed-aldehyde acid whose aldehyde functional group can undergo (i) oxidation to form trans-trans-muconic acid or (ii) reduction to 6-hydroxy-2,4- trans-trans-hexadienoic acid. Muconic acid was the first ring-opened urinary metabolite to be identified from benzene treated animals, by Jaffe in 1909 [13]. However, at that time the prevailing view was that the cis-cis isomer was the primary product, and it was not until 1952 that Parke and Williams unequivocally confirmed the structure as trans-trans-muconic acid, following [14C]-benzene administration to rabbits [14]). Alternatively, cytosolic alcohol dehydrogenases can reduce (reversible reaction) MCA to 6-hydroxy-2,4-trans-trans-hexadienal, a mixed-aldehyde alcohol whose aldehyde functional group can undergo oxidation to form 6-hydroxy-2,4-3 trans-trans-hexadienoic acid [15]. It is important to note, however, that while [3H]-trans-trans-muconic 14 acid and [14C]-6-hydroxy-2,4-trans-trans-hexadienoic acid have been isolated from the urine, blood, and/or bone marrow of rodents treated with [3H] or [14C]benzene, the reactive aldehydes, MCA and 6-hydroxy-2,4-trans-trans-hexadienal, have yet to be isolated in vivo. Thus, while MCA is produced in liver microsomal incubations containing benzene [16], and is undoubtedly an in vivo metabolite of benzene, albeit transient, the reactivity of this metabolite (6 sec in the presence of GSH) suggests that hepatic-derived MCA is unlikely to reach the bone marrow in sufficient quantities to cause toxicity [17]. This again emphasizes the critical balance between the chemical stability and the biological reactivity of benzene metabolites with respect to their potential role in benzene-mediated hematotoxicity. Both MCA and the hexadienal metabolite are myelotoxic in animal models [18]. Of current interest is the ability of benzene metabolites to block gap junction intercellular communication, in particular hydroquinone, MCA, and 6-hydroxy-2,4-trans-trans-hexadienal [19]. MCA decreases connexin 43 levels, inhibition of which has been linked to disruptions in hematopoesis. More recently, MCA induced cross-linking of connexin 43 likely underlies the inhibition of gap junction intercellular communication [20], further elaboration of which will provided later in this session. However, information on the presence of ring-opened metabolites in bone marrow following exposure to benzene remains limited (see Section 5).</p><!><p>Benzene-oxide spontaneously rearranges to phenol, the majority of which undergoes conjugation with either glucuronic acd or sulfate. That fraction that escapes conjugation can be further oxidized, with hydroquinone (HQ) as one of the products. Phenol itself is not hematotoxic, but synergistic effects of phenol and HQ, both of which accumulate in bone marrow, has become an established model of benzene hematotoxicity [21]. It should be emphasized that although animal models can be an appropriate approximation for human benzene metabolism, they are a less than ideal model for the human myelotoxic effects of benzene; perhaps highlighting the fact that metabolism may be necessary, but not sufficient for benzene hematotoxicity. The effectiveness of the HQ/phenol combination appears to be related, in part, to a pharmacokinetic interaction between HQ and phenol [22]. Phenol likely competes with HQ for conjugative enzymes and depletes the liver of UDPGA and PAPS, resulting in greater fractions of HQ and phenol available for delivery to bone marrow [23]. Bone marrow suppression may then result from phenol-stimulated peroxidase and/or phenoxy radical-mediated oxidation of HQ, which (in theory) initiates redox cycling and leads to the formation of the reactive intermediates, 1,4-benzo-semiquinone and 1,4-BQ. However, the current model of benzene hematotoxicity may need to be extended to include a role for thioether metabolites of HQ (see below). While depletion of cofactors for glucuronidation and sulfation increases the amount of "free" HQ and phenol in bone marrow, it will also increase the fraction available for oxidation and GSH conjugation. Such increases could be critical since both BGHQ and TGHQ are very potent hematotoxicants in vivo, inhibiting [59Fe] incorporation to the same degree as benzene at significantly lower doses (see Lau et al., this issue). Based on the (re)activity of HQ-GSH conjugates, it is possible that many of the hematotoxic effects attributed to HQ (or 1,4-BQ) may, in fact, be mediated by their thiol conjugates. Indeed, although semiquinone formation from a variety of quinones leads to O2 consumption, no O2 consumption occurs in reactions in which the 1,4-benzosemiquinone free radical is formed enzymatically [24], because 1,4-benzosemiquinone is so electron affinic that its rate of reduction by O2•- [25] is over four orders of magnitude faster than the reverse reaction (reduction of O2 to O2•-) [26, 27], which is usually responsible for O2 consumption via redox cycling. Thus, although redox cycling of semi-quinone radicals resulting in the generation of reactive oxygen species is proposed to be of major importance in the toxicity of many quinones, this mechanism is clearly excluded in the case of 1,4-BQ [28] (see also Figure 2). Indeed, although 1,4-BQ is cytotoxic to hepatocytes, it causes rapid depletion of cellular thiols without oxidative stress [29].</p><!><p>HQ is readily oxidized to 1,4-BQ, which is then efficiently scavenged by GSH. However, this reaction does not represent a true detoxication reaction, since the initial conjugate. 2-(glutathion-S-yl)hydroquinone (GS-HQ) is also readily oxidized to the corresponding GS-1,4-BQ. This cycle of the reductive addition of GSH to the quinone and subsequent cross-oxidation continues, resulting in the eventual formation of all possible addition products, GS-HQ, 2,3-(GS)-HQ, 2,5-(GS)-HQ, 2,6-(GS)-HQ, 2,3,5-(GS)-HQ, and 2,3,4,5-(GS)-HQ [30]. The in vivo relevance of these reactions was demonstrated following the administration of HQ (1.8 mmol/kg, i.p.) to AT-125 (an inhibitor of γ-glutamyl transpeptidase [γ-GT]) pretreated male Sprague Dawley rats. Five S-conjugates of HQ were identified in bile and one S-conjugate in urine [31]. The major biliary S-conjugate identified was GS-HQ. Additional biliary metabolites were 2,5-(GS)-HQ, 2,6-(GS)-HQ, and 2,3,5-(GS)-HQ. 2-(N-Acetylcystein-S-yl)HQ was the only urinary thioether metabolite identified. The quantity of S-conjugates excreted in urine and bile within 4 hrs of HQ administration was sufficient to propose a role for such metabolites in HQ mediated nephrotoxicity and nephrocarcinogenicity. Whether the conjugates are formed in sufficient amounts following either HQ/phenol or benzene administration to contribute to benzene-mediated hematotoxicity remains to be determined, but they can be identified in the bone marrow of HQ/phenol or benzene treated animals (see Section 7).</p><!><p>The gap in our knowledge with respect to benzene metabolism is the extent to which each of the metabolites exist/persist in bone marrow. The glucuronide conjugates of both phenol and HQ were identified in the bone marrow of B6C3F1 mice, but not in the bone marrow of F344 rats exposed for 6 hr to 50 ppm benzene [32] an intriguing finding especially given the fact that such conjugates are readily recoverable in the urine of these animals. More intriguing is the observation that HQ conjugates could not be detected in blood, liver, or lung of F344 rats, whereas they are readily detected in these tissues in B6C3F1 mice. Whether phenol is generated within bone marrow, or delivered there either as the conjugate or in the free form is not known. In addition to HQ and phenol conjugates, trans-trans-muconic acid and phenyl or catechol sulfate were also identified in the bone marrow of B6C3F1 mice and F344 rats, with the latter representing the highest fraction of benzene metabolites in both species. The presence of benzene metabolites in target tissue in humans is complicated by the potential contribution of dietary, over-the-counter medicinals, and perhaps other environmental sources (cigarette smoke) of such metabolites. For example, phenol, HQ, catechol, and trans-trans-muconic acid may all be derived from dietary sources [33,34], in addition to environmental/occupational exposures to benzene.</p><p>More recently, following co-administration of HQ/phenol (2.0 mmol/kg, ip) to rats, GS-HQ, 2,5-(GS)-HQ, 2,6-(GS)-HQ, and 2,3,5-(GS)-HQ were all detected in bone marrow [35]. The γ-GT catalyzed metabolite of GS-HQ, 2-(Cys-Gly)HQ, the dipeptidase metabolite, 2-(Cys)HQ, and the mercapturic acid metabolite, 2-(NAC)HQ were also all identified in bone marrow. Moreover, 2-(Cys)HQ and 2-(NAC)HQ appeared to persist in bone marrow. A similar metabolic profile was observed in mice, with the exception that 2,5-(GS)-HQ, 2,6-(GS)-HQ, and 2,3,5-(GS)-HQ were only sporadically detected. The latter conjugates were only observed in mice exhibiting high concentrations of GS-HQ in marrow. Nonetheless, concentrations of quinol thioethers in bone marrow were higher in mice than in rats of HQ/phenol-treated animals, which correlates with the relative sensitivity of these two species to benzene-induced hematotoxicity [36]. Similar results were obtained in benzene-treated animals. Following twice daily administration of benzene (11.2 mmol/kg) for 2 days, GS-HQ, (Cys-Gly)-HQ, (Cys)-HQ, and (NAC)-HQ were all detected in the bone marrow of rats and mice. The presence of a functional mercapturic acid pathway in bone marrow of both species was confirmed by in vitro studies. Concentrations of (Cys)-HQ were relatively high and persisted in bone marrow. The availability of bone marrow γ-GT and dipeptidases for processing of GSH conjugates is important, because Cys-Gly and Cys conjugates of HQ are more chemically (re)active than their corresponding GSH conjugates [37]. Consequently, such metabolites will be more potent arylators and redox cyclers. The half-wave oxidation potentials of the GSH conjugates of hydroquinone are also considerably lower than the half-wave oxidation potential of HQ [31]. In addition, the conjugates are also more readily oxidized by cytochrome(s) P450 to the corresponding quinones than HQ and generate more superoxide anions (Figure 2).</p><p>Because the liver produces significant amounts of the multi-substituted GSH conjugates following administration of HQ (1.8 mmol/kg) [31], it seems plausible that the conjugates identified in the bone marrow might be transported there via the circulation. However, an intravenous dose of GS-HQ (100 μmol/kg) yields quinol thioether concentrations in bone marrow far below those detected following coadministration of HQ/phenol (2.0 mmol/kg, ip), suggesting that the major fraction of the HQ-GSH conjugates present in bone marrow is formed in situ. Myeloperoxidase- and prostaglandin H synthase-mediated oxidation of HQ to 1,4-BQ, in the presence of GSH, has been shown to yield GS-HQ in vitro [38]. This reaction should therefore occur in bone marrow which contains both enzymes [39]. If concentrations of GS-HQ saturate bone marrow γ-GT (km = 68 μM), subsequent oxidation and GSH substitution will lead to the formation of multi-substituted GSH conjugates of HQ. Since estimates of quinol thioether concentrations in bone marrow exceed 180 μM, this metabolic pathway should occur readily in vivo. This view is supported by the identification of 2,5-(GS)-HQ, 2,6-(GS)-HQ, and 2,3,5-(GS)-HQ in the bone marrow of rats following co-administration of HQ/phenol.</p><!><p>GSH conjugates of HQ possess unique structural features that confer upon them the ability to interact with proteins that utilize GSH/cysteine or GSH/cysteine containing molecules as substrates or cofactors. The fact that such proteins play important roles in hematopoesis provides a basis for these specific benzene metabolites to interfere with this process, and to mimic the actions of benzene. For example; (i) HQ-GSH conjugates inhibit γ-GT [40], and acivicin, a potent inhibitor of γ-GT, is hematotoxic [41]. γ-GT possesses a unique thiol that is not required for catalysis, but is present in the active site of the enzyme on the light subunit [42]. This thiol may be a target for HQ-GSH conjugates. The major function of γ-GT is to regulate the transport of amino acids into cells, but its relative role in maintaining intracellular GSH concentrations is cell-type specific. γ-GT expression in the kidney is so high that even when 95% of the enzyme is inhibited, there remains more total activity than in most other tissues [43]. Consequently, renal GSH levels are only minimally decreased following treatment with acivicin. Tissues expressing very low levels of γ-GT usually possess a very active cystathionase pathway, in which cystathionine is deaminated and cleaved to form free cysteine and α-ketobutyrate. Therefore, even in the presence of acivicin, these tissues maintain high levels of GSH [44]. However, γ-GT activity in bone marrow is relatively low [43], and the more immature, undifferentiated cells within the marrow (targets of benzene) express almost no cystathionase [45]. Thus, inhibition of γ-GT in hematopoietic tissue dramatically reduces intracellular GSH levels [46]. (ii) Benzene- and HQ-treated mice exhibit increased granulopoiesis in bone marrow [47, 48]. Increases in G-CSF and GM-CSF can stimulate granulocytic differentiation [49] and HQ synergizes with GM-CSF to increase the number of myeloid progenitor cells in isolated mouse bone marrow [50]. HQ also mimicks the action of leukotriene D4 (LTD4) a downstream mediator of G-CSF, to initiate terminal differentiation in IL-3-dependent murine myeloblasts [51]. Interestingly, the ability of 1,4-BQ to induce granulocytic differentiation is prevented by inhibitors of γ-GT and by LTD4 receptor antagonists [52]. Subsequently, 1,4-BQ was shown to induce granulocytic differentiation by activating the LTD4 receptor, mimicking LTD4 [53]. How does the inhibition of γ-GT prevent 1,4-BQ induced granulopoesis? If the effects of 1,4-BQ are mediated by the corresponding GSH conjugates, then inhibition of γ-GT will prevent processing of the conjugates to the cysteinylglycine conjugate and interaction with the LTD4 receptor. LTD4 is a cysteinylglycine conjugate, and we predict that HQ-thioether conjugates will possess greater selectivity for the LTD4 receptor than HQ/1,4-BQ. (iii) Finally, functional roles for ATP-binding cassette (ABC) transporter proteins in hematopoietic stem cell function have recently been described [54, 55]. Structurally related catechol-GSH conjugates are potent inhibitors of MRP-1 mediated LTC4 transport [56] suggesting that GSH-HQ conjugates have the potential to interfere with ABC transporter function within hematopoietic stem cells. Moreover, ABC transporter expression/conformation/function are modulated by ROS, which induce defects in hematopoietic stem cell homeostasis [55].</p>

PubMed Author Manuscript

Concentration dependence of the sol-gel phase behavior of agarose-water system observed by the optical bubble pressure tensiometry

concentration dependence of the sol-gel phase behavior of agarose-water system observed by the optical bubble pressure tensiometry nobuyuki ichinose * & Hodaka UraWe have studied an expansion behavior of pressurized bubbles at the orifice of a capillary inserted in gelator-solvent (agarose-water) mixtures as a function of the gelator concentration in which the phase transition points are included. the pressure (P) -dependence of the radius of the curvature (R) of the bubbles monitored by laser beam has shown a discontinuous decrease in the exponent (m) of the experimental power law R = KΔP −m (K: constant) from 1 to 1/2 and a discontinuous increase in the average surface tension (γ ave ) obtained from the work-area plots of the mixtures exceeding that of pure water (72.6 mN/m) at 0.02 < [agarose] < 0.03 wt%, which is attributed to the disappearance of the fluidity. The apparent surface tension (γ app = ΔP/2 R) of the system in the concentration range of 0.03-0.20 wt% has been analyzed by a modified Shuttleworth equation γ app = σ 0 + τln(A/A 0 ), where σ 0 is an isotropic constant component and the second term is a surface area (A) -dependent elastic component, in which τ is the coefficient and A 0 is the area of the orifice. The analysis has indicated that σ 0 coincides with the γ app value of the mixture of 0.02 wt% and the second term at >0.02 wt% is the dominant component. from the appearance of the elastic component and concentration dependence of τ, the plateau of τ for the agarose-water mixtures at 0.03-0.10 wt% (Region II) has been explained to the phase separation giving two-phase mixtures of 0.02 wt% sol and 0.10 wt% gel and the upward inflection of τ at 0.10 wt% has been assigned to an increase in the elasticity of the gel with the increase of the agarose concentration in the range of >0.10 wt% (Region III). On considering the concentration dependence of the surface tension of agarose-water mixtures, the discontinuous and inflection points were assigned to the 1st-and 2nd-order phase transition concentrations of the agarose gel, respectively. Given the results with our tensiometry based on the optical bubble pressure method, distinct gelation points for other systems could be determined both mechanically and thermodynamically which will provide a diagnostic criterion of sol-gel transitions.Gel is a versatile state of substances widely seen in nature, industrial products, and foods due to its solid-fluid dualism, which is owing to the high holding content of solvent in the 3D network which reveals its viscoelastic nature 1,2 . Natural and synthetic water-soluble polymers often form hydrogels where more than 90% of water is contained in weight. Polymer gels are formed by the introduction of crosslinking bonds (chemical gels) or by the mutual aggregation through an increase in the concentration or cooling of the sol (physical gel) 2 . Agarose (Fig. 1) is a polysaccharide taken from a seaweed family (Geridiaceae) [3][4][5][6] , whose molecular weight has been reported to be M w = 0.8-1.4 × 10 5 g/mol 7 and is a typical substance showing physical gel formation, which has been reported to be above the concentration of 0.13 wt% at 20.0 °C8 . However, there are several reports on the minimum gelation concentration (MGC) and sol-gel transition temperature inconsistent each other.Sol-gel transition of physical gels has been studied intensively from mechanistic viewpoints, and several approaches by mechanical, thermal and spectroscopic measurement techniques have been conducted through the

concentration_dependence_of_the_sol-gel_phase_behavior_of_agarose-water_system_observed_by_the_optic

<!>Discussion<!>Material<!>Method

<p>dynamic methods 1,2,[9][10][11] . However, it remains difficult to determine the transition temperature or concentration through the dynamic rheological 8 , spectroscopic 9 , or thermal 10 measurements owing to the variation in the definition of gel, which depends on the measurement technique employed 2 . Similarly, it is not feasible to determine MGC for physical gels as a critical phase transition concentration because of the difficulty in observing precisely the point of disappearance of fluidity and the point of the appearance of a gel state.</p><p>The mechanistic study on the gelation of polymers also has a long history of experimental and theoretical chemistry and physics 1,2,12 . Although the main mechanism of the gelation has focused on cross-linking and fibril formation, their order is not clear unless the polymer chains are not intended to be cross-linked in the preparation process of a gel. Although fibrils of linear polymers and even small molecules have been observed by transmission electron microscopy (TEM) and atomic force microscopy (AFM) techniques for flash-freeze-dried agarose gels 2 , evaporation or freezing of the solvent from the systems may induce the fibril formation, and it is not clear whether the fibril formation is essential for gelation. Although the fibril formation affords the "elasticity" or the "plasticity" to gels, which would ensure measurement of mechanical properties of the soft materials, it is not easy to assign the fibril formation as a phase transition phenomenon in a "quasi-static" thermodynamic sense. Thermal hysteresis observed during gelation also makes the analysis complicated. Although agarose gel, for example, melts at 80-90 °C upon heating, the gel is formed at 35-40 °C upon cooling 13 . Since the hysteresis could arise from the existence of several species such as aggregates with single-and double-helices and free polymer in the mixture owing to the extremely slow diffusion and relaxation processes, we planned the surface tension measurement at a constant temperature by means of the bubble pressure method, a tensiometry (surface tension measurement) used for liquids 13 , to detect the deformation of the surface of soft materials including liquids. Since the surface tension (γ) is defined as a ratio of a mechanical work (dW) needed to create a new surface area of an infinitesimal amount (dA), γ = dW/dA and this work is identical to the change in the Helmholtz energy (F) at constant volume and temperature or to the change in the Gibbs energy (G) at constant pressure (p) and temperature (T). Furthermore, since change in the surface tension is also related to the sum of the chemical potentials (μ J ) of the chemical components (J) at the surface area (the Gibbs isotherm) at an equilibrium with the bulk, the mechanical measurement of the surface of a sol or a gel containing J's as a solvent and a gelator can be linked directly with the thermodynamic functions of the system.</p><p>According to the Ehrenfest classification of phase transition 14,15 , continuity of 1st order derivatives of the chemical potential (μ) as intensive variables are supposed to be the key diagnostic criteria (see also Supporting Information). For one-component systems, μ is a function of T and p, μ = μ(p,T). Since μ requires another variable X J , the mole fraction of component J (J = 1, 2) for two-component systems, phase transition also occurs as a function of X J at a finite condition of temperature and pressure. Since the surface tension is a 1st order derivative of the Gibbs energy, dG/dA for unit mole could be considered as the 1st order derivative of the chemical potential, dμ/dA which is equal to γ for unit mole and unit area. Therefore, some thermodynamic functions such as molar entropy can be derived from the surface tension in a quantity per unit area. For this reason, the surface tension can be a diagnostic parameter for phase transition phenomena on extending the Ehrenfest classification. For example, reported surface tension values of metals below and above the melting point are quite different each other 16 . However, the surface tension is less employed to study phase transition phenomena 17 because of the difficulty in the measurement of the surface tension with the same tensiometry for two different phases, except for liquid-like monolayers at the liquid-air or liquid-liquid interfaces 18,19 where the same tensiometry can be applied and changes in the surface tension can be treated with the Gibbs isotherm. The purpose of our study is to elucidate the sol-gel phase transition through a tensiometry. In other words, changes in an intensive variable γ will indicate the changes in other intensive variables, i.e. the chemical potentials upon the phase transition. However, the surface tension of gels has not been known and the Young-Laplace relationship is not assured in the gel phase. We now report the volume expansion behavior of pressurized bubbles at the orifice of a capillary inserted in gelator-solvent (agarose-water) mixtures as well as the surface tension of the mixtures as a function of the gelator concentration in which the transition point can be determined. We chose agarose as a sample of gelator because of the plentiful reference data in the literature [3][4][5][6][7][8][9][10] . In this article, the pressure-dependence of the radius of the curvature of the bubbles monitored by laser beam has been studied as a function of gelator concentration to examine the Young-Laplace relationship [20][21][22] in the sol and gel states, and to establish a measurement method of the surface tension of gels. We also have studied the discontinuity and the inflection in the surface tension induced by the increase of the concentration in the mixture to examine the 1st-and 2nd-order phase transitions, respectively 14,15 . www.nature.com/scientificreports www.nature.com/scientificreports/ a manometer to the bubble and the static pressure at the depth of the orifice (ΔP). The bubbles in the aqueous solutions of agarose with the concentrations of ≤0.02 wt% (Region I) showed a decrease of the radius upon the increase of ΔP obeying the Young-Laplace relationship with a constant surface tension. The surface tension was obtained by a nonlinear least square curve fitting showing a gradual decrease from that of pure water (72.6 mN/m) to 63.4 mN/m at 0.02 wt% with the agarose concentration as observed for amphiphilic polymers such as polyethylene oxide. The decrease of the radius of the meniscus for the mixtures of ≥0.03 wt% by the applied pressure became smaller compared to that for the dilute solutions, although the appearance of the mixture was solution-like. The double logarithmic plot of R versus ΔP indicated a slope of −1 for the solutions and a slope of −1/2 for the mixture of ≥0.05 wt%, and mixtures of 0.03 and 0.04 wt% showed an intermediate value of the slope of ≈−2/3. This abrupt change in the R-ΔP power law behavior of the bubble strongly suggests that the bubble surface of the mixtures of ≥0.03 wt% is no longer liquid-like (Fig. 4).</p><p>To compare the surface property of the mixture of ≥0.03 wt% with that of solutions of ≤0.02 wt%, we estimated the surface tension of the mixture without using the Young-Laplace relationship throughout the R-ΔP sets at the agarose concentrations. Assuming the spherical surface of the meniscus, the radius of the curvature (R) can been converted to the surface area (A) and the volume (V) of the meniscus, and the R-ΔP curve has been translated into a ΔP-V curve, then accumulation of the ΔP-V curve gives a course of to the Gibbs energy change (ΔG) as the meniscus has done the work (ΔW) for the volume expansion against the surface tension of the agarose-water mixture. The plot of ΔG vs A shows straight lines for all the mixtures (Fig. 5). Then, the slopes of the lines indicates the average surface tension of the system (γ ave ) as defined thermodynamically, Applied pressure / Pa www.nature.com/scientificreports www.nature.com/scientificreports/ γ = (∂G/∂A) T,p . The error in the surface tension of the solutions with that obtained by the nonlinear least square method is less than 1% (±0.1 mN/m). The surface tension estimated from the Gibbs energy change has a slightly larger error (±0.2 mN/m) than that obtained by the non-linear curve fitting due to the trapezium approximation of 5 sets of the data points in the integration of the ΔP-V curve. Nevertheless, the surface tension values obtained by the two calculations for the solution with concentrations of ≤0.02 wt% agreed within 0.2%. The average surface tension plotted vs. agarose concentration clearly indicates a jump from 63.4 to 75.0 mN/m between 0.02 and 0.03 wt% (Fig. 6). The surface tension of the 0.03 wt% mixture exceeds that of pure water (72.6 mN/m measured with our system) ruling out the exuding of pure water from the 3D network. The surface tension of the mixture is almost constant up to 0.10 wt%. However, it increases for the mixtures of >0.10 wt% until 154.7 mN/m for the 0.20 wt% mixture, but our measurement system was not able to measure the surface tension of mixtures of >0.20 wt%. These results clearly indicate that the mixtures undergo the sol-gel phase transition and the expansion behavior of the bubbles in the gels is network-limited.</p><!><p>As the Young-Laplace relationship has been derived from the equation concerning the work for volume expansion and that for the increase of the surface area of a bubble with a small change in the radius (dR), it is necessary to consider the work for the volume expansion for the agarose-water mixtures of ≥0.03 wt%. On considering the gel as a solid, we cannot apply the equation with a constant surface tension, dW = dAγ because the presence of the surface strain due to the elasticity. However, the Young-Laplace equation must hold at a given ΔP. We have calculated the surface tension for each R-ΔP data set as apparent surface tension (γ app ) and reconsider the surface tension with the equation dW = d(Aσ) = σdA + Adσ, where σ is the thermodynamic surface tension. Differentiation of this equation with respect to A gives an equation γ mec = σ + dσ/dlnA according to the treatment of the surface tension of solids 16,17 . As shown in Fig. 7, γ mec increases linearly with the increase of ln(A/A 0 ) and tends to converge to the γ mec value for the 0.02 wt% solution (γ mec = 63.4 mN/m) upon extrapolation of γ mec values for various mixtures at zero expansion (A = A 0 ). The intercepts are almost independent of the agarose concentration, while the slope is dependent on the agarose concentration. Therefore, we can rewrite the equation for γ mec as γ mec = σ 0 + dσ/dln(A/A 0 ), where σ 0 = 63.4 mN/m. The average surface tension γ ave can be considered as a representative value of γ mec at the average surface area (A = Ᾱ), which is dependent on the applied pressure within the fracture limit of the bubble at an agarose concentration.</p><p>We now return to the R-ΔP relationship. The experimental power law observed for the radius of the curvature to the applied pressure showed an exponent of −1/2, i.e. R = K(ΔP) −1/2 , where K is the constant. This relationship could be obtained from the Young-Laplace equation R = ΔP/2γ mec and the equation for γ mec at the corresponding surface area A = π(R 0 2 + z c 2 ). Although we did not obtain the analytical solution for R = R(ΔP), the second term (compressibility modulus) in γ mec originating from the surface strain appears and dominates with the increase of the agarose concentration above the minimum gel concentration and this will cause the change in the exponent. We also obtained a linear relationship between K and γ mec in their double logarithmic plot (Fig. S5). As a result of the introduction of the surface strain term for the analysis of the R-ΔP data, we could conclude that the observed change in the exponent m is phenomenal although detailed numerical analysis might reveal the R-ΔP relationship.</p><p>We have introduced the surface strain term dσ/dln(A/A 0 ) in the mechanical surface tension of agarose-water mixture with the concentration of >0.02 wt% and now show a plot of γ mec -concentration (Fig. 8). The introduction of the surface strain term indicates the abrupt increase of γ mec at 0.02 < [agarose] < 0.03 wt% and an inflection at 0.10 wt% more clearly as compared to the γ ave -concentration plot (Fig. 6). As shown in Fig. 8 (and Fig. S6), a plateau is seen in the concentration range of 0.03-0.10 wt% (Region I). According to Herring, γ mec contains a scalar www.nature.com/scientificreports www.nature.com/scientificreports/ and a tensoric terms, the latter is corresponding to the surface strain term. Rusanov 23,24 explained nonequivalence of the mechanical (γ mec ) and thermo-dynamic (σ) surface tension for wetting of an isotropic solid surface co-existing mobile components by relating σ to γ mec and the sum of the chemical potentials of the mobile (I) and immobile (J) components dσ = − s (J) dT + (γ mec − σ)dlnA − ΣΓ I(J) μ I , where s (J) is the entropy surface density and Γ I(J) is the surface excess of the mobile component at the surface of the immobile component. For the present case, I = water and J = agarose, ΣΓ I(J) μ I can be considered to be independent of A and the experiment has conducted at the constant temperature. Therefore, we obtain γ mec = σ + dσ/dlnA = σ 0 + dσ/dln(A/A 0 ) again on considering γ mec = σ 0 at A = A 0 . This means the mechanical surface tension is consisted of scalar term as the surface tension of σ 0 value for the 0.02 wt% solution and the strain term dσ/dlnA due to the strain of the agarose gel as a solid.</p><p>The strain term is also related to the chemical potentials of the immobile components. However, there is no fast diffusion and no equilibrium between the surface and the bulk phase meaning the inequality of the chemical potentials and there is an area-dependent strain due to the measurement. For this reason, we cannot treat the strain term directly as a parameter for the phase behavior of the solid. Fortunately, the elasticity of polymer gels originates from the entropy of the polymer chain, i.e. ΔF = −TΔS 25 . Now we can treat the strain term as the entropy surface density of the immobile component, i.e. the gel network. With a small isotropic deformation of the surface area, www.nature.com/scientificreports www.nature.com/scientificreports/ experimental value of the strain term is given by (dσ/dlnA)ln(A/A 0 ) ≈ 2ετ, where ε is the isotropic tensor and τ = dσ/dlnA. Therefore, the mechanical surface tension γ mec can be a thermodynamic quantity of gels and inflection of the strain term is attributable to a change in the elasticity due to a structural change of the gel. We have concluded that the surface tension can be a criterion for sol-gel transition: the observed discontinuity explained by the appearance of the strain term is assigned to the sol-gel phase transition as the 1st-order phase transition is, and the inflection of τ reflects the change in the elasticity of the solid gel phase as the 2nd-order phase transition.</p><p>Finally, we will enter the phase behavior of the agarose-water mixture. The surface tension of the mixture of ≤0.02 wt% (Region I) is the liquid with no doubt because of the Young-Laplace behavior and γ mec = σ throughout the applied ΔP range. The observation of the plateau in γ mec in the range of 0.03-0.10 wt% (Region II) indicates that the strain term of the gel is almost constant. This strongly suggests that the chemical composition of the gel is almost identical despite of the change in the net concentration. As judged from the above discussion on σ 0 , the chemical composition of the sol is also constant. Therefore, only quantities of both phases seem to vary with the concentration. This has invoked the idea of phase separation and the lever rule to explain the slight increase of γ mec in Region II. Spinodal decomposition of the mixture followed by gelation and concomitant phase diagrams have been reported for the mixture of atactic polystyrene in cyclohexane as judged by the test tube tilting and ball-drop method 26 . Indei has reported network formation of poly(vinyl alcohol)-borax by percolation as judged by the micro-rheology 27 . Similar mechanisms for gelation of the agarose-water system at lower concentrations have been proposed on the basis of the rheological 8 and dynamic light scattering measurements 28 . Spinodal decomposition of agarose-water at 0.02 < [agarose] < 0.03 wt% gives 0.02 wt% solution and 0.10 wt% gel whose amounts are determined by the lever rule as an origin of σ 0 and weak concentration-dependence of γ mec in Region II. This leads us to a conclusion that the gel in Region II is two-phase gel with a percolated or a fractal-like structure because of the freedom of the system for intensive variables (f) is 0 for the sake of the phase rule (f = c -p + 2, where c and p are the numbers of components and phases, respectively. Note that two of f are occupied by the temperature and pressure of the experimental condition.). The slight increase in γ mec can be explained by the number of the cross-linking points which increases with the increase of the amount of the gel domain in the mixture.</p><p>As Rees predicted that agarose molecules form double helix aggregates 5 , which have been observed by X-ray diffraction and other micro-imaging techniques with ≥0.1 wt% mixtures, Liu et al. proposed a "primary fiber" as an intermediate in the gelation of agarose 6 . The primary fiber could be an aggregate formed by twisting of two agarose chains which act as a substrate for self-epitaxial nucleation to form the 3D network. Our observation with the mixtures in Region II and the >0.10 wt% concentration region (Region III) strongly suggests that the expansion of the network requires some energy to unwind the primary double helix, i.e. the energy to break the hydrogen and the hydrophobic bindings and the energy to unwind the multiple helix requires increases by the increase of the concentration. Tieleman et al. have measured the area-dependence of the surface tension of a lipid monolayer, which undergoes phase transition by area expansion 29 . They have reported the compressibility moduli (corresponding to τ = dσ/dlnA) of dipalmitoyl-phosphatidylcholine (DPPC) to be 1400 ± 20 mN/m for liquid condensed phase at a molecular area (A L ) of A L = 0.475 nm 2 , 200 ± 12 mN/m for liquid condensed and liquid expanded coexisting phase at A L = 0.58 nm 2 , and 100 ± 20 mN/m for liquid expanded phase at A L = 0.620 nm 2 , respectively 29 . They have also observed inflection points in the γ-A plot 29 . On the other hand, our observation indicated the values of dσ/dlnA = 300-500 mN/m in Region II and 500-3500 mN/m in Region III together with the inflection point at 0.10 wt% (Fig. S6). These data clearly show that the two-phase gel in Region II has a weak binding nature as compared to the monolayer of alkyl chains. Upon increasing the agarose concentration, dσ/dlnA values of the gel increase drastically by the increase of the density of cross-linking points 12 . This means that the structure of the gel changes with the concentration and the freedom f = 1 corresponding one-phase gel (p = 1) in Region III. Although melting of the agarose gels of >0.10 wt% (Region III) is well known to occur at 35-50 °C5,6 , the systems below this concentration (Region II) have been reported to be a suspension of micro-gel 8,28 , and its phase transition behavior upon the concentration change has been mentioned less frequently. Although intensive study including temperature dependence is needed to draw an exact phase diagram, we show a schematic phase diagram which explain the observed behavior of the surface tension as a function of agarose concentration (Fig. 9).</p><!><p>Agarose (Agarose-S tablets, Nippon Gene, for electrophoresis use, sulfate (−SO 4 ) content ≤0.1%) was used as received. Its molecular weight of Agarose-S as given by the manufacturer was 2-3 × 10 5 g/mol determined by liquid chromatography. Mother solution of agarose (1.0 wt%) was prepared by heating a mixture of an agarose tablet (≈0.50 g) and distilled deionized water (500 mL) was heated to 90-95 °C in a beaker with a microwave oven. The solution was diluted at ≈40 °C (above the gelation point of 37-39 °C for 1.5 wt% solution).</p><!><p>We have modified the bubble pressure method where the experimental apparatus possesses a Teflon cell holding a capillary and an optical window which enable the pressurization of the sample and monitoring of the radius of the curvature (R) of the meniscus (bubble) optically with a laser beam passing through the capillary (pressurizing optical probe) 30,31 . The meniscus formed in the solution or gel acts as a concave lens for the laser beam. The focal length of the meniscus is determined by an external optics and a photodetector system. The radius of the curvature was translated into the surface tension of the mixture, as a first derivative of Gibbs energy (G) with respect to the surface area (A), through the translation into the surface area and volume (V) of the meniscus as a function of the pressure applied (P). The gelation has been judged by the apparent surface tension exceeding that of pure water. The details are described in the Information.</p><p>www.nature.com/scientificreports www.nature.com/scientificreports/ conclusion</p><p>In conclusion, we have established a tensiometry for weak gels using the optical bubble pressure method through the demonstration of the sol-gel transition behavior of the agarose-water mixtures as consecutive occurrences of the loss of the fluidity and the increase of the surface tension upon the increase of the gelator concentration, which can be attributed to the 1st-and 2nd-order phase transitions through the analyses of the Gibbs energy change of the meniscus for its expansion and the mechanical surface tension for solids as introduced by Shuttleworth 32 and Herring 33 , and with the aid of the concept of entropic elasticity. Therefore, the increase in the surface tension upon the gelation can be attributed to the changes in the mechanical properties of the polymer network as a solid with entropic elasticity and can be interpreted into thermodynamic phase behaviors. The present results has indicated that the surface tension measurement will provide a reliable criterion for the sol-gel transition of other polymeric systems and also for that of low molecular-mass organic gelator systems, which are being increasingly studied 34,35 .</p>

Scientific Reports - Nature

Chloromethyl-modified Ru(<scp>ii</scp>) complexes enabling large pH jumps at low concentrations through photoinduced hydrolysis

Photoacid generators (PAGs) are finding increasing applications in spatial and temporal modulation of biological events in vitro and in vivo. In these applications, large pH jumps at low PAG concentrations are of great importance to achieve maximal expected manipulation but minimal unwanted interference. To this end, both high photoacid quantum yield and capacity are essential, where the capacity refers to the proton number that a PAG molecule can release. Up to now, most PAGs only produce one proton for each molecule. In this work, the hydrolysis reaction of benzyl chlorides was successfully leveraged to develop a novel type of PAG. Upon visible light irradiation, Ru(II) polypyridyl complexes modified with chloromethyl groups can undergo full hydrolysis with photoacid quantum yield as high as 0.6.Depending on the number of the chloromethyl groups, the examined Ru(II) complexes can release multiple protons per molecule, leading to large pH jumps at very low PAG concentrations, a feature particularly favorable for bio-related applications.

chloromethyl-modified_ru(<scp>ii</scp>)_complexes_enabling_large_ph_jumps_at_low_concentrations_thro

Introduction<!>Theoretical calculations<!>Synthesis and characterization<!>pH changes<!>Photoacid quantum yields<!>Photoacid mechanism<!>Biological applications<!>Conclusions<!>Conflicts of interest

<p>Photoacid generators (PAGs) are molecules which can generate acids under light irradiation, 1 and have been widely used as photo-initiators of cationic polymerization in photolithography, photocuring, and three-dimensional (3D) printing. 2 The great success that PAGs have achieved in the elds of the microelectronic industry and microfabrication relies on their spatial and temporal control of proton concentrations. Protons also play pivotal roles in a variety of biological events. As a result, PAGs have found ever-increasing applications in biochemical, biological, and biomedical areas in recent years, such as photodynamic therapy, 3 drug delivery, 4 adenosine triphosphate (ATP) biosynthesis, 5 photocontrol of enzyme activity, 6 and protein conformations, 7 as well as proton transfer in biomolecules. 8 For these bio-related applications, PAGs may have some specic characters, 9 including proper solubility in aqueous solutions, light response in the visible or near-infrared region for deep tissue penetration and minimal damage to biomolecules, 10 and large pH jumps to get remarkable bio-effects. To reach instant large pH jumps, high photoacid quantum yields were vigorously pursued. However, mM levels of PAGs, concentrations not easy to realize under in vitro and in vivo conditions, were usually needed because each PAG molecule can theoretically generate one proton only. 4a,5-7 High concentrations of PAGs and their corresponding photoproducts may also interfere with the examined biological processes, hampering their application severely. Though efforts have been made to address this issue by integrating two PAG moieties into a single molecule, the resultant PAGs showed poor photoacid quantum yields. 1c,3 New strategies that can endow PAGs with both high photoacid capacity (i.e. the proton number a PAG molecule can release) and high photoacid quantum yield are in urgent demand to boost their bio-related utilization.</p><p>Thermal hydrolysis of haloalkanes may release HX (X ¼ Cl, Br, I) and the acid capacity depends strictly on their halogenation levels. 11 To the best of our knowledge, such a classic reaction has not been capitalized on in PAGs. What attracts our attention is the hydrolysis of benzyl chloride, the kinetics of which can be effectively modulated by electron donating/ withdrawing groups. 12 Electron donating groups will accelerate the process while an opposite effect may be observed for electron withdrawing groups. As reported by A. Fry and S. Yamabe, para-methoxy benzyl chloride has a hydrolysis rate of 4 orders of magnitude higher than that of the para-NO 2 counterpart. 12 Inspired by these facts, we herein synthesized a series of 4,4 0 -bis(chloromethyl)-2,2 0 -bipyridine (bcm-bpy) coordinated Ru(II) complexes (1-3, Scheme 1) to explore their PAG capability. The rationale behind our design is as follows. (1) The bcm-bpy ligand is expected to be inactive in thermal hydrolysis due to the electron-decient feature of the pyridine ring, which may be consolidated further upon coordination to the Ru(II) center. 13 A good stability in the dark is a prerequisite for a desired PAG. (2) The highest occupied molecular orbital (HOMO) of Ru(II) polypyridyl complexes is generally Ru(II) centered, 14 while the lowest unoccupied molecular orbital (LUMO) of complexes 1-3 should localize on the bcm-bpy ligand due to the electronegativity of Cl atoms. 15 Thus, the bcm-bpy related metal-to-ligand charge transfer (MLCT) state will be accessed preferentially upon light irradiation, from which an efficient hydrolysis of the chloromethyl groups is also anticipated due to the greatly enhanced electron density on the bcm-bpy ligand. This is indeed what we observed in our experiments. Complexes 1-3 can undergo efficient hydrolysis upon visible light irradiation in aqueous solutions, and release 2, 4, and 6 equivalents of HCl, respectively, with photoacid quantum yields as high as 0.6. Both high photoacid capacity and high photoacid quantum yield of complex 3 make large pH jumps feasible at low concentrations. Complex 3 is not only the rst PAG which can release six protons per molecule, but also the rst type of PAG that makes use of hydrolysis reaction of benzyl chloride groups. As a conceptual demonstration, complex 3 was successfully used to switch on the activity of acid phosphatase upon visible light irradiation at a concentration as low as 10 mM. 6</p><!><p>Theoretical calculations based on the Gaussian 09 program package 16 (see ESI †) were carried out before experiments to examine our ideas. As expected, the calculated HOMO and LUMO of 1-3 are mainly Ru(II) and bcm-bpy based, respectively (Fig. 1 and S1-S2 †). Thus MLCT Ru/bcm-bpy excitation means pumping one electron from the Ru(II) center to the bcm-bpy ligand, which will greatly enhance the electron density of bcmbpy. Such an expectation is further convinced by comparison of Mulliken charges of the ground state (GS) and T1 state. Taking complex 1 as an example (Table S1 †), the selected Mulliken charges of the atoms on the bcm-bpy ligand in the T1 state are all negatively shied compared with that in the GS. In addition, the length of C-Cl bonds stretches from 1.83 Å (GS) to 1.85 Å (T1) (Fig. S3-S5 †). All these results are favorable for photoinduced hydrolysis of the chloromethyl groups.</p><!><p>The synthesis and characterization of complexes 1-3 were reported in our previous work. 13 The aqueous solutions of all the complexes are quite stable in the dark, as evidenced by the negligible changes in absorption spectra, emission spectra, and 1 H NMR (nuclear magnetic resonance) spectra, as well as pH values of the solutions (Fig. S7 †). Upon irradiation at 520 nm, rapid changes in these spectra were observed. Taking complex 3 as an example, the absorbance in the regions of 315-375 nm and 425-500 nm decreased gradually, along with a slight blue-shi of the ligand-based transition peak centered at 291 nm (Fig. 2). The emission intensity also increased quickly upon irradiation (Fig. 2). The absorption and emission spectra did not change any more aer irradiation for only 3 min. At that time, the solution exhibited nearly the same absorption and emission as that of [Ru(bhm-bpy) 3 ] 2+ (20 mM, bhm-bpy ¼ 4,4 0bis(hydroxymethyl)-2,2 0 -bipyridine), suggesting full hydrolysis of the six chloromethyl groups. The excited [Ru(bhm-bpy) 3 ] 2+ without competing photolysis decay should have a longer excitation lifetime and also a higher luminescence quantum yield compared with complex 3, and thus a turn-on luminescence was observed. The known PAGs usually show absorption changes upon light irradiation. Our complexes represent the rst class of PAGs that exhibit remarkable luminescence turn-on along with proton release. By virtue of the higher sensitivity of uorescence detection as well as the wide application of uorescence confocal microscopy, this feature may facilitate biological studies. In addition, the transformation of 3 into [Ru(bhmbpy) 3 ] 2+ is also conrmed by high resolution electron spray ionization mass spectra (HR ESI-MS) (Fig. S16 †) and 1 H NMR spectra (Fig. 3). Only an m/z peak of 375.0861 which can be ascribed to the product with six hydroxymethyl groups was observed aer irradiation in H 2 O. A similar result was obtained Scheme 1 Chemical structures of complexes 1-3.</p><!><p>The pH changes of the aqueous solutions of complexes 1-3 (10 mM) were monitored with a pH meter (Fig. 4). The initial pH values of the solutions were about 6.6. The weak acidity may be attributable to the dissolved CO 2 absorbed from ambient air. 17 The pH values of the solutions kept unchanged in the dark for 24 h (Fig. S7 †). Upon LED (light-emitting diode) irradiation at 520 nm, the pH values decreased quickly to 4.9 for 1, 4.6 for 2 and 4.3 for 3 in 3 min (the theoretically calculated results are 4.7, 4.4 and 4.2 for complexes 1-3, respectively). A large pH jump of 2.3 units was obtained for complex 3 at a concentration as low as 10 mM, which obviously prots from its ability to generate 6 protons for each molecule. To reach a similar pH jump, much larger concentrations, usually several hundreds of micromolar, were generally needed for the reported systems, 6,17,18 unfavorable for biological applications.</p><p>In bio-related applications, bioactive molecules, such as glutathione (GSH), may serve as strong nucleophilic agents to impair the dark stability of these chloromethyl-modied Ru complexes. 19 Thus, the effect of GSH was also studied. The initial pH value of a GSH (1 mM) aqueous solution was measured to be 3.4. Addition of 1-3 (10 mM) did not cause any pH changes in 24 h without irradiation (Fig. S20 †). The negligible pH changes may also be the result of the strong buffering ability of GSH. To rule out this possibility, ethanethiol (1 mM) was used instead of GSH. Similar results were obtained again at either room temperature or 37 C (Fig. S21 and S24 †), con-rming the good dark stability of complexes 1-3 even in the presence of strong nucleophilic agents. Upon 520 nm irradiation, the pH values of 1-3 (10 mM) solutions containing ethanethiol decreased quickly (Fig. S22-S24 †), showing their desirable anti-interference ability.</p><!><p>The photoacid quantum yields of 1-3 (50 mM) in H 2 O were measured using potassium ferrioxalate as a chemical actinometer (ESI). 20 The light of 520 nm was not suitable for potassium ferrioxalate due to its quite small extinction coefficient at 520 nm, and thus a 470 nm LED was selected. The obtained quantum yields at 470 nm are 0.62 AE 0.01 for complex 1, 0.65 AE 0.02 for complex 2 and 0.61 AE 0.01 for complex 3, which are among the highest reported for PAGs. 1c,d,2a,b</p><!><p>The high photoacid quantum yields and photoacid capacities of 1-3, along with their good dark stability, make the hydrolysis mechanism of the chloromethyl groups anchored on these complexes particularly interesting. Generally, a thermal hydrolysis of a haloalkane may undergo through either an S N 1 (unimolecular) or an S N 2 (bimolecular) pathway. 12 For benzyl chlorides, the mechanism may change from S N 1 to S N 2 depending on the electron donating/withdrawing abilities of substituents. According to the reports of A. Fry and S. Yamabe, the hydrolysis of para-methoxy benzyl chloride proceeds through an S N 1 way, while S N 2 is found for the para-NO 2 compound. For complexes 1-3, MLCT excitation enhances the electron density of the bcm-bpy ligand, which is favorable for S N 1 rather than S N 2 Based on the theoretical and experimental results, a possible photoacid mechanism is schematically illustrated in Scheme 2, taking complex 1 as an example. Furthermore, we also examined the photolysis of the bcm-bpy ligand in H 2 O. The absorption of the free bcm-bpy ligand appears below 310 nm. Under direct UV light (254 nm) irradiation, HCl was also generated, most probably through a different mechanism involving homolysis of a carbon-chorine bond as reported by R. Sinta et al. in the studying of photoacid properties of 4,6-bis(trichloromethyl)-1,3,5-triazines. 21 Compared with the clean transformation of 1-3 into their corresponding hydroxymethyl products, the photolysis of bcm-bpy To the best of our knowledge, this is the rst time that the hydrolysis reaction was applied to develop a novel type of PAG. Since their discovery four decades ago, PAGs have been dominated by iodonium and sulfonium salts. Only over the past decade, new structures along with their unique properties have emerged in the PAG arsenal, such as photochromic triangle terarylenes, 1e,2a the open form of spiropyrans, 17,18 and N-oxyimidesulfonates. 22 Bearing in mind that the benzyl chloride groups are easy to integrate into a variety of photosensitizers, such as porphyrins, phthalocyanines, and boron dipyrromethene (BODIPY) dyes, more diverse PAG structures and their broader applications are expected.</p><!><p>As a conceptual demonstration of its possible biological applications, complex 3 at 10 mM was used to modulate the activity of acid phosphatase (ACP). 6 ACPs are widely distributed in the human body. An abnormal activity of ACPs is related to many diseases, including prostate cancer, kidney disease, multiple myeloma, Gaucher disease, etc. 23 Therefore, controlling the ACP activity may nd potential applications in disease treatment and drug development. S. Kohse and co-workers successfully utilized a photoacid (2-nitrobenzaldehyde) to realize efficient tuning of the ACP activity. 6 Similarly, 4-methylumbelliferyl phosphate (MUP) was selected as the substrate of ACP. The transformation of MUP into 4-methylumbelliferone (MU) was monitored by absorption spectra. The initial pH of the solution of MUP and acid phosphatase was kept at 8.0, where the activity of the enzyme was faint as negligible spectral changes were observed. Upon addition of complex 3 (10 mM) and irradiation for 90 s with an LED at 520 nm, the activity of the enzyme was successfully switched on, and elevated to a level of 0.05 mmol (min mg enzyme ) À1 (Fig. 5 and S27 †), which is consistent with the reported results. 6 However, a PAG concentration as high as 500 mM was implemented in their experiments. Light irradiation in the control experiment without the enzyme caused no effects on MUP (Fig. S28 †).</p><!><p>In conclusion, we designed and synthesized three bcm-bpy based Ru(II) complexes as novel PAGs, which can work in aqueous solution and be excited by green light with photoacid quantum yields of about 0.6. Complex 3 can release six protons per molecule, leading to large pH jumps at low concentrations. Theoretically, more protons may be released provided more chloromethyl groups are anchored on the ligand. Our results may open new avenues for developing novel PAGs with both high photoacid quantum yield and capacity to meet more challenging demands particularly in bio-related areas.</p><!><p>There are no conicts to declare.</p>

Royal Society of Chemistry (RSC)

Dispersion of TiO2 nanoparticles improves burn wound healing and tissue regeneration through specific interaction with blood serum proteins

Burn wounds are one of the most important causes of mortality and especially morbidity around the world. Burn wound healing and skin tissue regeneration remain thus one of the most important challenges facing the mankind. In the present study we have addressed this challenge, applying a solution-stabilized dispersion TiO 2 nanoparticles, hypothesizing that their ability to adsorb proteins will render them a strong capacity in inducing body fluid coagulation and create a protective hybrid material coating. The in vitro study of interaction between human blood and titania resulted at enhanced TiO 2 concentrations in formation of rather dense gel composite materials and even at lower content revealed specific adsorption pattern initiating the cascade response, promising to facilitate the regrowth of the skin. The subsequent in vivo study of the healing of burn wounds in rats demonstrated formation of a strongly adherent crust of a nanocomposite, preventing infection and inflammation with quicker reduction of wound area compared to untreated control. The most important result in applying the TiO 2 dispersion was the apparently improved regeneration of damaged tissues with appreciable decrease in scar formation and skin color anomalies.Accelerated and less painful healing of wounds caused by burn or mechanical injuries and skin and muscle tissue engineering for decreased scar formation and minimization of permanent damage belong to most prominent challenges in modern surgery 1 . Application of nanostructured materials for improved tissue regeneration has become a well-developed and accepted practice in the application of metal bone implants, where a thin layer of nanostructured titanium dioxide is deposited on the top of the implant surface 2,3 . Nano titania is rapidly getting coated with proteins when immersed into the biological fluids due to its well-recognized ability to adsorb and coordinate proteins 4 and phospholipids 5 on its surface. Adsorption of phospholipids can be considered as one of the factors guiding the attachment of cells and grafting on the growing tissue on an implant 6 . In the domain of skin regeneration a strong effort so far has been set on application of stem cells. They have been applied in different approaches, in particular, including sprays 7 . Use of nanomaterials for wound treatment and skin repair has also been intensively investigated ranging from silicone based artificial skin layers 8 , to the development of new materials for wound dressing with delayed and prolonged release of medicines 9,10 and even to direct application of nanoparticle dispersions either possessing themselves antibacterial effects, such as silver and gold nanoparticles [11][12][13] , or loaded with painkillers and antibiotics. The latter have been realized with metal oxide nanoparticles

dispersion_of_tio2_nanoparticles_improves_burn_wound_healing_and_tissue_regeneration_through_specifi

<!>Results and Discussion<!>Figure 5. (A)<!>Conclusions<!>Methods<!>Particle characterization.<!>Thrombin time test.<!>Blood sampling.<!>Skin sampling.<!>Contact activation complexes.<!>Histological analysis.

<p>which the FDA approved for intravenous application, namely with Al 2 O 3 14 or with Fe 3 O 4 15 as carriers. The use of medicine-loaded sol-gel alumina and iron oxide resulted in appreciable reduction of the scar tissue sizes with wound healing times not principally different, however, from those when only a solution of the selected medicines was applied directly to the wound 14,15 . Considerable improvement in the size and structure of the scar could be observed when anti-inflammatory natural medicine curcumin loaded siloxane gels 16 , while using silver nanoparticle-graphene-polymer nanocomposites an appreciable acceleration of wound healing was achieved 17 . Nano TiO 2 has also drawn a considerable attention in the recent years. In the view of broad application of nano titania a lot of effort were set on, in the first hand, investigation of its potential toxicity in different forms 18 . It has been demonstrated in numerous studies reviewed, analyzed and reproduced in 19 that in the dark no or negligible toxicity could be associated with essentially any form of nano titania. The apparent toxicity, in particular DNA damage, could be observed for both human cells 19 and for microorganisms, such as, for example, micro algae 20 in the presence of highly crystalline larger (over 25 nm and in most apparent cases -about 100 nm 20 ) particles, obtained by spray pyrolysis, subjected to irradiation by UV and visible 20 and in some cases even IR light 21 . This effect was unequivocally related to the photocatalytic properties of the applied titania and was more pronounced for the more catalytically active anatase phase 20 . The application of nano TiO 2 in wound healing was also very much focused on the exploitation of the photocatalytic effect for production of reactive oxygen species to target bacteria in wound infections 21,22 . To enhance the bactericidal action of titania via photochemical effects, quite a number of applications have been developed for grafting titania onto either fabrics 23,24 or porous (hydrophilic) polymer nano composites 25,26 . In some cases such composites were used, however, primarily for controlled drug delivery to the wounds 9,10 .</p><p>In the present work we have chosen a completely different approach in use of nano titania for wound healing. The material applied here was a dispersion of small (less than 10 nm) anatase particles produced by sol-gel method in solution. The colloid was stabilized by grafting of (protonated) antioxidant ligand triethanolamine on the surface of the particles, making them positively charged in the originally applied media and photochemically inactive 27 . The small anatase particles stabilized by antioxidant ligands have been demonstrated to be biocompatible for both human cells 28 and for bacteria 27 and micro algae 29 . The applied dispersion did not contain any additional bioactive substances or medicines. The aim was set to investigate possible effects originating solely via surface interactions (adsorption) of the nano titania, possessing large active surface area, with body fluids.</p><!><p>Nanoparticles of titania, independently of their phase composition and in spite of their broadly recognized adsorbent properties towards biomolecules, are commonly considered as inert in relation to living systems 18 . Titania as anatase powder is broadly used in food and hygiene industry, referred to as E171 food colorant. Commercially available Degussa P25 titania nano powder is considered as broadly accepted negative standard in the in vitro acute toxicity studies 18 . The dispersion of sol-gel produced anatase nanoparticles stabilized by charging via surface complexation with triethanolammonium ligands applied in this work has been characterized in several earlier publications and was proved to be biocompatible in contact with both human and plant cells up to rather high concentrations reaching 100 μg/mL 30,31 . Recent investigation of the Degussa P25 nanoparticles at very low concentration of below 50 ng/mL has demonstrated them to be capable to induce activation of the contact system eliciting thromboinflammation 32 .</p><p>This latter observation was considered to be associated with potential health risk from titania nanoparticles if they emerge in the body fluids. On the contrary, we hypothesized that a dispersion of TiO 2 nanoparticles should be applied on the skin to cause enhanced blood coagulation, which is an important first step in initialization of the wound healing processes. The nanoparticles applied in this work were produced by hydrolytic route from titanium ethoxide modified by triethanolamine ligands, following the procedure adopted from 27 with some minor adjustments (Please, see the experimental part and Supplementary). They belong to the anatase phase as is clearly indicated by the distances between fringes for the aligned {101} planes of 0.354 nm in the high resolution TEM images (see Fig. 1A and Figs S1, S3 and S4). The size of the particles is rather uniform and is well in agreement with the observed hydrodynamic size in both the initial alcohol-based dispersion and in the dispersion obtained by its 10 times dilution by de-ionized water (Figs 1B and S2). In contrast, the dilution by 0.9 wt% NaCl (physiological solution), while not leading to precipitation is associated with extended aggregation and apparent increase in the hydrodynamic size of the produced aggregates with distribution between about 100 nm and several micrometers (Fig. 1C).</p><p>Spectrophotometric measurements of the clot formation at quite enhanced final TiO 2 concentrations in human blood plasma (1 mg/mL and 10 mg/mL respectively) showed that it was strongly accelerated compared to normal clotting in air. The process was completed in just over 30 s at 1 mg/mL and in about 10 s for 10 mg/mL compared to over 1 min with untreated blood serum (see Fig. 2).</p><p>In order to bring insight into the interaction of blood with the applied titania dispersion, we produced clots by addition of a droplet (0.02 mL) of the dispersion (or PBS solution for reference) to 0.10 mL of whole blood drop placed onto an optical glass slide. Quick clotting in case of TiO 2 dispersion led to formation of a composite in a form of dark brown, almost black, brittle solid. The SEM analysis revealed formation of a dense solid TiO 2 film on the outer surface of solidified droplet with thickness of 30-40 μm, covering a complex structure of iron-rich protein composite with only 3-4 wt% content of titania as calculated from EDS analysis (see Fig. 3B,C and Supplementary Fig. S3, and Table S1).</p><p>The structure of the protein clot is apparently different for pure blood or the blood with added PBS solution on one hand, and the material produced on interaction with titania dispersion on the other hand: the microstructure in the latter case is larger and more smooth, indicating stronger interactions within the clot. It indicates that the addition of titania is resulting in stronger interactions within the forming solid. We have applied rather high concentration of titania aiming to produce a dense nanocomposite material with perspective to form such composite coatings on the wounds as protective patches instead of using polymer patches as recommended earlier 9,[23][24][25][26] . Interaction of the dispersion with freshly removed human epidermis (outer skin layer) was also investigated, showing that the complex skin structure becomes coated with a uniform dense layer of solid titania film (Fig. 3D-F) with thickness dependent of the concentration and amount of added dispersion, but typically thicker than 10 μm. The thickness of a film obtained by single deposition is apparently quite high, which leads to formation of a uniform system of surface cracks originating from gel densification on the evaporation of solvent. The coating has apparently good adhesion to the skin surface and is not removed mechanically when dry. Washing off with mechanical brushing in a water flow removes major part of it with residues remaining persistently in the skin micro folds. These features in combination indicated that treatment of the wounds with the applied concentrated titania dispersion were going to create on their surface quite dense mechanically tough and strongly adhering hybrid coatings. Such coatings could be potentially capable to serve as protective patches on their own, eliminating the need to cover a wound with some additional protective material/bandage.</p><p>Platelet activation was measured as the reduction in platelet numbers in the blood after incubation with the TiO 2 nanoparticle-coated surface (TiO 2 ), polystyrene surface (PS) and Corline heparin surfaces (CHS) compared to the initial blood samples (i.e., not exposed to the chambers). The TiO 2 surface induced a clear reduction in platelets as only 25.8 ± 6.4% (mean ± SEM) of the platelets remained in the blood after the incubation, while the blood incubated with the plain PS surface still had 66 ± 9.4% of the platelets and the in the blood from the CHS surface more than 90% (93.2 ± 2.1%) of the platelets were remaining after the same incubation time (See Fig. 4A).</p><p>Activation of the coagulation system was further analyzed by measuring the generation of TAT complexes after blood exposure to the TiO 2 nanoparticle containing surface, the PS surface and the CHS surface. As expected the TiO 2 -nanoparticle coated surface resulted in a large increase in TAT levels compared to the PS surface that just showed a small increase in TAT levels, and the CHS control surface that gave no significant rise in TAT concentration compared to the initial blood samples (See Fig. 4B).</p><p>The intrinsic pathway of coagulation is triggered when FXII come in contact with foreign materials, which initiates the clot formation and thereby also the wound healing process. Hence, it would be of importance to see if the TiO 2 nanoparticle-coated surfaces induce this type of contact activation. In a previous publication we investigated protein adsorption and contact system activation induced by TiO 2 nanoparticles incubated in Figure 1. TEM image of the dried dispersion of the applied TiO 2 nanoparticles produced using modified methodology from ref. 18 (A). Hydrodynamic size of the particles in water (B) and in isotonic salt solution ((C), 0.5 ml of dispersion diluted by 10 ml isotonic NaCl), both by DLS. Natural blood clot and its EDS analysis (A), enlarged structure of the inner part of the blood clot forming on interaction with the TiO 2 dispersion and its EDS analysis (B), TiO2 crust on the surface of a treated blood clot and its EDS analysis (C). The integrations of EDS spectra are presented in Table S1 (Supplementary). Untreated skin sample (D) coating on the surface of skin sample (E), enlarged view of the titania film on a skin sample (F). human EDTA-plasma and whole blood without anticoagulantia, respectively 32 . The formed protein corona was abundant in most contact activation proteins; five out of the ten protein identified with highest score identified by MALDI-TOF belonged to the contact system. High amounts of contact system activation complexes were generated reflecting this binding 32 . In the present study, the generation of FXIIa-AT and FXIIa-C1INH complexes was measured in the blood after incubation with the TiO 2 coated surface, with and without the FXII-specific inhibitor CTI. The result showed that both FXIIa-AT and FXIIa-C1INH complexes were formed in the plasma after contact with the TiO 2 nanoparticle coated surface, but the addition of CTI inhibited the formation of these complexes with ca. 80%, thus confirming an FXIIa-dependent complex formation (data not shown).</p><p>In the view of the observed strong effects on blood coagulation potentially attractive for wound healing, it was decided to evaluate the use of titania sol in a spray application on burn wounds in vivo in rats that were treated with a pre-heated copper disc, causing burns of second (Groups 1 and 2, untreated and treated respectively) and fourth (Groups 3 and 4, untreated and treated respectively) degree. The rats with untreated wounds were used as controls for both types of incurred damages. Following the healing processes it was possible to note that while the duration of the healing processes in total did not differ appreciably, the dynamics of wound surface reduction was clearly and for more severe wounds even dramatically different. The treatment with a titania sol was apparently resulting in quicker decrease of the exposed wound area, reaching for 4 th degree burns as much as 30% reduction in the middle of the healing period (see Fig. 5A,B).</p><p>Application of the titania colloid onto the wounds led also to an apparent reduction of the forming scar tissue (see Fig. 6) and its less abnormal appearance. Histological analysis of the healed wound tissues turned to be fully in line with the visual observations. In case of Group 1, i.e. the healing of untreated wounds, the epidermis was not changed, with normal keratinization. In the papillary layer the fibers were thickened and lied more tightly than normal. The number of glands was greatly reduced, in particular, the sebaceous were reduced in size, the sweat epithelium was flattened (with a reduced height of the cells), and hair follicles were isolated (no more than three in sight). The reticular layer was showing more pronounced fibrosis with the activation of fibroblasts, overproduction of the basic substance, and hypervascular focal perivascular leukocyte infiltration (see Fig. 6A,B)).</p><!><p>Wound surface reduction: Group 1 -healing of untreated 2 nd degree burns, Group 2 -second degree burns treated with nano titania, Group 3 -untreated 4 th degree burns, Group 4-4 th degree burns treated with nano titania. Each value is an average of 3 rats/group (for details, please, see Supplementary Table Histological analysis of the results from Group 2, i.e. healing of second degree wounds treated daily with a titania colloid, gave very exciting and encouraging results. The healed area demonstrated unchanged normal skin structure without any skin structure alterations (see Fig. 6C,D).</p><p>For the Group 3, where 4 th degree burns were healing without treatment, the epithelium was thinned, sometimes missing, exposing the dense connective tissue that replaced rarely thinned dermis through the entire thickness. Dermal papilla also were smoothed and skin appendages absent. Vascularization of all layers as compared with the sample of Group 4 was reduced. In the deep were observed poorly developed networks of thin-walled vessels with weak perivascular inflammatory infiltration. More fibrosis hypodermis with complete replacement of the connective tissue layer of fat and muscle fibers could be seen (see Fig. 6E,F). Tripe corresponded to 4 th degree burns.</p><p>For the Group 4, where the 4 th degree burns were treated daily with titania, the epithelium was a thickened prickly layer (acanthosis). Dermal papillae were completely smoothed out (unavailable). Dermal thickness was reduced, fibrosis was more pronounced with increasing abundance of fibers and increase in the number of fibroblasts and the base material. Skin appendages -all glands and hair follicles -were completely absent. The reticular layer was a more developed network of young newly formed blood vessels, arranged vertically. Among them -a moderate diffuse leukocyte infiltration, focal hemosiderin deposits and accumulation of hemosiderophages occurred. Severe fibrosis of hypodermis with almost complete replacement of adipose tissue and muscle fiber atrophy took place. Tripe corresponded to the 3rd degree of the burn (Fig. 6G,H).</p><p>The main message of the work presented here is the ability of stabilized anatase TiO 2 NPs (applied as a solution) to promote burn wound healing tested in an in vivo rat model. This was evident by the formation of a firm crust of hybrid protein-titania nanocomposite with significantly higher anti-bacterial and anti-inflammatory properties compared to that of untreated controls. The rats did not reveal any abnormal behavior or apparent pathologies after completion of the healing process. To get an insight into possible side effects in treatment with nano titania we have carried out a thorough investigation of the tissues (liver, kidney, spleen, and brain) of treated animals (and untreated ones as reference) with respect to possible retention of titanium (for details, please, see the description below in Methods). It was clearly demonstrated that the content of titanium did not increase in any of the vital organs of the treated rats, staying at the same level as in the control animals (see Fig. 7 and Supplementary S5 and S6). This result appears quite logical in the view that the applied small TiO 2 particles possess, as revealed, very strong affinity to proteins. They are apparently either fixed on the surface inside the blood clot or adsorbed directly on the walls if they come into contact with damaged body fluid vessels. It has to be mentioned that the starting level of titanium content in vital organs was in all the studied cases quite low and did not show any statistically appreciable difference between the test and the control samples. The only case, where the difference on the first glance appeared considerable was for the starting spleen samples. It was in this case actually the control that displayed higher titanium content. The reason of the latter might be that the 4 studied rats in this selection ( 1 / 3 of the starting 12 animals sacrificed on the first day of the experiment) have by accident eaten some titanium-containing material (paper, straw or sand). The difference would most probably not be statistically significant if a bigger group of test animals could be investigated, but this would be not ethically acceptable. Multiple reports are found in the literature describing activities of TiO 2 immobilized to different matrices, e.g., antibacterial activity in vitro [33][34][35] , as well as anti-inflammatory and accelerated wound healing activity 36 but to our knowledge, this is the first report of these activities induced by topical application of a solution of TiO 2 NPs.</p><p>Previously, we have studied innate immunity activation by low concentrations of TiO 2 NPs in whole human blood. We found dose dependent platelet activation, monitored as TAT complex formation, release of thrombospondin-1 (a platelet β-granule protein), and platelet loss. There was substantial activation of the contact/kallikrein system, reflected as generation of FXIIa-AT and FXIIa-C1-INH complexes, and concomitant production and release of the pro-inflammatory chemokines Interleukin (IL-8), Monocyte chemoattractant protein (MCP)-1, Macrophage inflammatory protein (MIP)-1α and MIP-1β (detected using a multiplex analytical panel). All these parameters, except production of MIP-1α and MIP-1β were inhibited by the specific contact system inhibitor CTI 32 . That study confirmed earlier results from our group where we found substantial platelet activation (TAT, platelet loss, release of beta-thromboglobulin [another platelet β-granule protein], generation of FXIIa-AT/ C1-INH, and release of the platelet derived growth factor [PDGF]) induced by planar Ti and TiN surfaces 37 . It should be noted that PDGF was not included in multiplex panel used in 22 and therefore was not detected in that study.</p><p>The link between contact system activation and the release of the same chemokines (IL-8, MCP-1, MIP-1-α, MIP-1β, and PDGF) in addition to vascular endothelial growth factor (VEGF) was also evident in a previous study where we utilized a number of polymers as a tool to investigate these interactions 38 .</p><p>In the present work we observe that TiO 2 NPs greatly accelerates blood clotting in vitro in two different models, first by turbidimetry when added to human citrate-plasma, and secondly when evaporated onto polystyrene surfaces which were then incubated with whole human blood. In the latter case, the readouts were TAT generation, platelet loss, and generation of FXIIa-AT and FXIIa-C1-INH complexes, both of which decreased in the presence of CTI.</p><p>Cutaneous wound healing is a multistep process where coagulation-induced inflammation is a critical first event 39 . During this initial phase, a protective fibrin clot is formed, platelets are activated to contribute to clot formation, but also to release chemokines and growth factors, which recruit and activate neutrophiles and monocytes. Chemokines, which are essential to promote wound healing include PDGF (chemoattractant for neutrophils, monocytes and fibroblast), IL-8 (the major attract and activator for neutrophils), MCP-1 and MIP-1-α, (which, in conjunction promote macrophage response), MIP-1-β (mixed leukocyte recruiter), and VEGF (which promotes angiogenesis at a late stage in the healing process).</p><p>Since the production and secretion of PDGF, IL-8, MCP-1 and MIP-1-α, MIP-1-β, and VEGF all have been shown to be induced by TiO 2 , in conjunction to contact system (FXII) activation, we conclude that this, at least to a certain extent, explains why the administration of TiO 2 nanopaticles accelerates wound healing.</p><!><p>Colloidal solution of pH-neutral stabilized titania displayed clear trends to enhanced and accelerated blood clotting. This process was not hindered by addition of common anti-coagulants such as heparin. Interaction of titania dispersion with both blood and skin samples resulted in formation of dense films on the surface with uniform micro cracks caused apparently by contraction of the gel on evaporation of the solvent. The biochemical analysis indicated clearly that this was associated, on one hand, with apparent strong blood clotting ability, and, on the other hand, with activation of the contact system resulting in enhanced wound healing effect. Using the colloidal titania for treatment of burn wounds in vivo resulted in apparently quicker reduction of the exposed wound area, while the duration until skin total recovery was comparable with untreated wounds. The most striking effect in application of titania was its logicrent ability to promote restoring of the normal skin structure resulting in the absence of the scar tissue after healing of the 2 nd degree burns and improvement of the scar tissue to the appearance typical of a 3 rd degree burns in the cases of the 4 th degree burn damage.</p><!><p>Preparation of sol-gel titania. The synthesis of the stable size-uniform titania colloids used in this work was made following the earlier described technique 27 . For producing the initial precursor solution Ti(OEt) 4 (5 mL) was dissolved in anhydrous ethanol (5 mL) and then 1.5 mL of triethanolamine were added on continuous stirring. Hydrolyzing solution (1 mL) was produced by mixing 0.5 M nitric acid, HNO 3 (0.5 mL), with ethanol, EtOH (2.0 mL). The resulting clear transparent yellowish solution contained 120 mg/mL TiO 2 according to TGA measurements. The details of particle characterization are provided in the Supplementary.</p><!><p>The size of the initial particles in the aqueous sols was measured by dynamic light scattering (Microtrac instrument). FTIR spectra of sols and gels were recorded with a Perkin-Elmer Spectrum 100 instrument without dilution in a cell fitted with CaF 2 windows. The morphology of the xerogels was studied with a Hitachi TM-1000-μ-DeX 15 kV scanning electron microscope (SEM), and the agglomerate size and crystallinity were studied with a Topcon EM-002 B ultrahigh-resolution analytical electron microscope (TEM). UV/Vis spectra were recorded using a Hitachi U-2001 spectrophotometer.</p><!><p>Lyophilized citrate human plasma and human thrombin (150 NIH units/mg) were obtained from «Kvik» LTD Company, Russia. Thrombin time was measured as a period for clot formation from human citrate plasma with known concentration of plasminogen and fibrinogen. With this aim, 10 mg of lyophilized human plasma was disssolved in 1 mL of triple distilled water (giving final plasminogen concentration-102 μg/mL, fibrinogen concentration -2.8 mg/mL) and then 0.1 mL of the plasma solution was mixed with 1 mL of 0.9% NaCl solution (isotonic). Thrombin solution was prepared by solving 1 mg of thrombin in 1.5 mL of NaCl solution (isotonic). Clotting mixture was prepared in 1 × 0.5 mm plastic cuvette by mixing 1.1 mL plasma solution and 0.1 mL thrombin solution respectively. Turbidity at 315 nm was immediately monitored during 175 sec. For the tests with titania sol, before addition of thrombin solution 10 or 100 µl titania sol has been added (corresponding to final TiO 2 content in the mixture of 1 mg/mL and ca. 10 mg/mL respectively) and compared with the samples diluted with the same volumes of isotonic NaCl.</p><!><p>Fresh human blood samples were obtained from healthy volunteers who had not received any medication for at least 10 days prior to donation. Blood samples were collected in an open system with no soluble anticoagulant. In this system, any material that comes into contact with blood is furnished with the Corline heparin surface (Corline Systems AB, Uppsala, Sweden) to prevent material-induced contact activation. Preparation followed the manufacturer´s recommendations.</p><p>Ethical approval was obtained from the regional ethics committee (Uppsala University Hospital). All methods were carried out in accordance with relevant guidelines and regulations, in particular, complying with the rules summarized by the Swedish Research Council for treatment of human tissue samples summarized at http://www. codex.vr.se/en/manniska4.shtml.</p><!><p>The skin samples 1.5-2 mm in diameter were donated by the corresponding author (VGK) and cut by a scalpel from the finger tips.</p><p>Written informed consent was obtained from all patients involved in the study.</p><p>The whole blood model. To investigate the influence of the TiO 2 -particles on the blood coagulation cascade in human whole blood a slide chamber model was used, which has been described previously (by Hong et al. 40 ), containing two circular wells with an inner diameter of 17 mm. The test surfaces with TiO 2 -nanoparticles were prepared by adding 0.5 mL of TiO 2 particle suspension (120 mg/mL in ethanol) to polystyrene (PS) microscope slides followed by evaporation overnight. As a reference PS slides were treated the same way, but without the TiO 2 -nanoparticles. The chambers, the control surface (PVC) and the tubes, tips and tubing to be used in contact with the blood were pre-coated with heparin (Corline Systems AB). Blood was drawn from healthy volunteers, who not had received any medication at least 10 days prior to blood donation. The wells were filled with 1.5 mL freshly drawn blood containing 0.5 IU/mL heparin (Leo Pharma) and the test surface was attached with two clips, thereby constituting a lid over the two chamber wells. These devices were then incubated under constant rotation at 30 rpm for 60 min. at 37 °C. After incubation the blood was mixed with EDTA at a final concentration of 10 mM to inhibit further activation of the blood cascade systems. Before centrifugation platelet counts were performed. The blood samples were then centrifuged at 2500 g for 15 min., the plasma was collected and stored at −70 °C for further analysis of coagulation markers. The experiment was repeated four times (different blood donors each time) in duplicates. To one series of experiments 3.5 μM Corn Trypsin Inhibitor (CTI; Enzyme Research Laboratories), which is a specific FXIIa inhibitor, was added to the blood prior to incubation with the surfaces.</p><p>Platelet count. The number of platelets was analyzed in the blood samples before and after incubation with the test surfaces using a Sysmex XP-300 Hematology Analyzer (Sysmex Corp.). Platelet count was calculated as the remaining amount as compared to the initial sample (before incubation in the chambers) and was expressed as mean percent of initial ± SEM.</p><p>Thrombin-Antithrombin complexes (TAT) ELISA. Plasma levels of TAT were analyzed by a conventional sandwich ELISA. The plasma samples were diluted in normal citrate-phosphate-dextrose plasma. The TAT complexes were captured by an anti-human thrombin antibody (Enzyme Research Laboratories) and detected with an HRP-conjugated anti-human AT antibody (Enzyme Research Laboratories). As standard pooled human serum diluted in in normal citrate-phosphate-dextrose plasma was used. All values were given in μg/L.</p><!><p>For the detection of FXIIa-antithrombin (AT) and FXIIa-C1-inhibitor (C1INH) complexes in the plasma samples a standard sandwich ELISA described by Sanchez et al. 41 was used. Microtiter plates were coated with anti-human FXIIa antibodies (Enzyme Research Laboratories) and captured complexes were subsequently detected with either biotinylated anti-human AT (Dako) or biotinylated anti-human C1INH (Enzyme Research Laboratories) followed by HRP-conjugated streptavidin (GE Healthcare). Standard solutions were diluted in normal plasma. All measured values are given in nmol/L.</p><p>In vivo Investigation of burn wound healing properties. Male Hooded rats (body weight range 200-250 g) were used for the study. Animals were acclimatized under standard animal laboratory condition for 7 days prior to the experiment. All experiments were approved by institutional animal ethical committee (Ivanovo State Medical Academy, Russia, Protocol No. 2 from 06.04.2015) and are in agreement with the guidelines for the proper use of animals for biomedical research 42 . Animals were divided into 4 groups, each consisting of 3 rats: I group -rats with 10 sec treatment with heated disc; II group -rats with 10 sec treatment with heated disc and healing titania; III group -rats with 20 sec treatment with heated disc; IV group -rats with 20 sec treatment with heated disc and healing titania. All animals survived and did not suffer weight loss within standard deviation until the last day of the experiment (day 19).</p><p>Animals were anesthetized with ketamine (dose 60 mg/kg), acting as both sedative and long-term pain-killer agent 43 , by intraperitoneal injection, the dorsal hair was shaved and disinfected. Burns were made 1 cm diameter copper disc preliminary heated up to 300 °C. For groups II and IV the materials were applied on excised burns. The burns were treated daily with 0.1 mL of prepared titania solution. Wound sizes were measured daily until the healing is complete. The wound outline was transferred to transparent films and scanned with an Epson Perfection 2480 scanner. The wound area was calculated with ImageJ 1.30 v. software. The percentage wound reduction was calculated according to the following formula: where C n is the percentage of wound size reduction, S o is initial wound size, S n is wound size on respective day. The rats were kept in individual cages 20 × 30 cm 2 area and had free ability to motion and access to both food and water. As the wounds were located in the dorsal area, there was no risk that the animals should lick or bite their own wounds.</p><!><p>Fragments of skin with scar excised and completely fixed in 10% formalin solution during 24 hours. After routine gynecological wiring samples were poured into paraffin. 20 slices with 5 μm thickness were prepared and stained with hematoxylin and eosin from each paraffin block.</p><p>Chemical analysis of tissue samples. Analysis of the content of titania in the organs was carried out according the following procedure. Animals (in the experiment, 12 rats were used) on 1, 14 and 28 days respectively after wound healing with titania were anesthetized with isoflurane and were killed by cervical dislocation and organs (liver, kidney, spleen and lung) were collected and weighed immediately after killing of the animals. Dissolution of organs was carried out with a mixture of concentrated sulfuric and nitric acids. Completeness of dissolution was achieved by organs heating in heat-resistant glasses. 1 ml of concentrated HNO 3 and 3 ml of concentrated HCl were added to the resulting syrup-like transparent solutions with following transferring to 25 ml volumetric flasks using distilled water. The titanium concentration was determined by atomic-absorption spectroscopy with inductively coupled plasma. The studies were carried out on HORIBA Jobin Yvon ULTIMA 2. Three rats without titania treatment were used as a control sample for organs.</p><p>The in vivo experiment was repeated for another group of animals and in this case liver, and brain tissues were removed from one representative animal per group, cut into pieces about 0,2 g that were weighed and then dissolved in 3 ml of aqua regia. The pH after dissolution was adjusted to 3.0 by addition of 1.0 M NH 3 solution. The produced liquids were analyzed with ICP-AES Spectro Cirros CCD Instrument, Kleve, Germany.</p><p>Two of the authors, GAS and VGK declare their involvement in the activities of the CaptiGel AB company, Sweden, developing metal oxide colloids for environmental and biomedical applications.</p>

Scientific Reports - Nature

Modular Synthesis of Trifunctional Peptide-oligonucleotide Conjugates via Native Chemical Ligation

Cell penetrating peptides (CPPs) are being increasingly used as efficient vectors for intracellular delivery of biologically active agents, such as therapeutic antisense oligonucleotides (ASOs). Unfortunately, ASOs have poor cell membrane permeability. The conjugation of ASOs to CPPs have been shown to significantly improve their cellular permeability and therapeutic efficacy. CPPs are often covalently conjugated to ASOs through a variety of chemical linkages. Most of the reported approaches for ligation of CPPs to ASOs relies on methodologies that forms non-native bond due to incompatibility with in-solution phase conjugation. These approaches have low efficiency and poor yields. Therefore, in this study, we have exploited native chemical ligation (NCL) as an efficient strategy for synthesizing CPP-ASO conjugates. A previously characterized CPP [ApoE(133–150)] was used to conjugate to a peptide nucleic acid (PNA) sequence targeting human survival motor neuron-2 (SMN2) mRNA which has been approved by the FDA for the treatment of spinal muscular atrophy. The synthesis of ApoE(133–150)-PNA conjugate using chemo-selective NCL was highly efficient and the conjugate was obtained in high yield. Toward synthesizing trifunctional CPP-ASO conjugates, we subsequently conjugated different functional moieties including a phosphorodiamidate morpholino oligonucleotide (PMO), an additional functional peptide or a fluorescent dye (Cy5) to the thiol that was generated after NCL. The in vitro analysis of the bifunctional CPP-PNA and trifunctional CPP-(PMO)-PNA, CPP-(peptide)-PNA and CPP-(Cy5)-PNA showed that all conjugates are cell-permeable and biologically active. Here we demonstrated chemo-selective NCL as a highly efficient and superior conjugation strategy to previously published methods for facile solution-phase synthesis of bi-/trifunctional CPP-ASO conjugates.

modular_synthesis_of_trifunctional_peptide-oligonucleotide_conjugates_via_native_chemical_ligation

Introduction<!><!>Introduction<!><!>Materials<!>Peptide Synthesis and Purification<!>PNA Synthesis and Purification<!>Conjugation of the Peptide-hydrazide to PNA<!>Synthesis of Trifunctional Peptide-PNA Conjugates<!>Mass Spectral Characterization of Peptides and Peptide-PNA Conjugates<!>Cell Culture, Transfection and Confocal Imaging<!>RNA Extraction and RT-qPCR Analysis<!>Results and Discussion<!><!>Results and Discussion<!><!>Results and Discussion<!>Conclusion<!>Data Availability Statement<!>Author Contributions<!>Funding<!>Conflict of Interest<!>Supplementary Material<!>

<p>CPPs are relatively short cationic, amphipathic peptides (Kurrikoff et al., 2020) and are being widely used for intracellular drug delivery for a wide range of cell-impermeable cargoes, such as therapeutic antisense oligonucleotides (ASOs). The conjugation of CPPs to ASOs has been shown to enhance their cell permeability, intracellular bioavailability, and subsequently, an increase in their therapeutic efficacy (Tajik-Ahmadabad et al., 2017; Klein et al., 2019; Nikan et al., 2020). CPPs have also been utilized to cross the blood-brain barrier for delivery of therapeutic ASOs into the central nervous system (CNS) (Hammond et al., 2016; Shabanpoor et al., 2017). The potential of CPPs as safe and effective delivery vectors is also being recognized and embraced by pharmaceutical industry. Sarepta Therapeutics has shown significant improvement in the pharmacokinetics of PMOs and is developing CPP-PMO conjugates as the next-generation of PMO-based therapy for Duchenne muscular dystrophy (Roberts et al., 2020).</p><p>One of the key challenges for developing CPP-ASO conjugates as therapeutics is the large and frequent doses required to achieve the desired pharmacological effect. There is limited evidence of ASO-associated toxicity but CPPs can be toxic even at very low doses. To address this issue, we previously developed trifunctional CPP-ASO conjugates by coupling two ASOs to a single CPP, to mitigate cytotoxicity (Shabanpoor et al., 2015). A key consideration when designing CPP-ASO conjugates is their efficient synthesis. They are often assembled separately via solid-phase synthesis and then conjugated using a variety of chemical linkages, such as disulfide (Turner et al., 2005; Saleh et al., 2010), amide (Bruick et al., 1996; Betts et al., 2012; Shabanpoor and Gait, 2013), thioether (Patil et al., 2019), triazole, oxime, hydrazone, and thiazole bonds (Lu et al., 2010; Nikan et al., 2020). Despite the wide range of ligation reactions available for the preparation of CPP-ASO conjugates, there are still some limitations associated with their use, such as poor in vivo stability (e.g., disulfide bonds). Furthermore, linkages with steric bulk impart restricted conformational movement that can potentially affect functional activity. We have previously found that the formation of thiazoles and in particular triazoles (Patil et al., 2019) are slow, and reduce the solubility of the CPP-ASO conjugate.</p><p>Ligation reactions that generate native linkages such as amide bonds are more desirable due to their biocompatibility and minimal impact on the physicochemical properties of CPP-ASO conjugates. This method has been routinely used by our group and others to prepare peptide-ASO conjugates (Betts et al., 2012; Shabanpoor and Gait, 2013; Shabanpoor et al., 2015; Klein et al., 2019). The formation of amide linkages is typically achieved via a C-terminal active ester reacting directly with an amine. Unfortunately, this strategy is not compatible with peptide sequences containing lysine, histidine, glutamic acid, and aspartic acid. The aforementioned studies using this conjugation strategy have reported low conjugation yields in the range of 20–40%. Therefore, an improved method for amide bond formation is needed for a more efficient and scalable synthesis of CPP-ASO conjugates.</p><p>In this study, we present a simple and efficient chemo-selective method for the preparation of CPP-ASO conjugates, which utilizes NCL to form a stable amide linkage between a CPP and a cysteine-bearing ASO. The highly reactive thiol artifact from the ligation enables a second chemo-selective conjugation with an additional maleimide- or haloacetyl-bearing ASO, bioactive peptide or fluorescent label (Scheme 1). The advantage of using NCL is its tolerability to side-chain functionalities of the unprotected peptides. Besides, it can be performed in mild conditions and does not require special amino acids.</p><!><p>Peptide and ASO conjugation using native chemical ligation and subsequent conjugation of a second moiety such as ASO, peptide or fluorophore to the thiol group.</p><!><p>We chose an FDA-approved oligonucleotide sequence (Table 1) targeting survival motor neuron-2 (SMN2) mRNA splice correction as a model ASO (in PNA chemistry) for conjugation to a previously characterized CPP derived from the receptor-binding domain of Apolipoprotein E [ApoE(133–150)] (Meloni et al., 2020). In order to synthesize trifunctional conjugates, three different functional moieties were selected for conjugation to the side-chain of cystine that was generated during NCL of CPP-ASO. These functional moieties include the FDA-approved oligonucleotide sequence as above but incorporated into the PMO scaffold, and a peptide (HA2) derived from the hemagglutinin protein of influenza virus, which enhances endosomal release of conjugated cargoes by disrupting the endosomal membrane (Midoux et al., 1998). It is worth noting that both peptides (ApoE and HA2) either alone and as conjugate to ASOs have been previously characterized by our group and others for cell uptake and also shown to have no significant cytotoxic effect in a concentration range of 20–30 µm (Neundorf et al., 2009; Shabanpoor et al., 2017). The other functional moiety was a fluorophore (Cy5), to visualize the uptake of the CPP-ASO conjugate. To our knowledge, this is the first report on the use of hydrazide-based native chemical ligation for the synthesis of bifunctional CPP-ASO conjugates and subsequent conjugation of a second functional moiety to generate a trifunctional CPP-ASO conjugate.</p><!><p>Sequences, molecular weight and % yield of ASOs, peptides, peptide-PNA conjugate and trifunctional conjugates. Bold indicates where the modifications have been made to different sequences.</p><!><p>TentaGel® XV resin (0.25 mmol/g) was obtained from Rapp Polymere (Tuebingen, Germany). 2-chlorotrityl chloride polystyrene resin (0.77 mmol/g) was purchased from ChemPep. The 20-mer PMO (5′-ATT​CAC​TTT​CAT​AAT​GCT​GG-3′) was purchased from Gene Tools LLC (Philomath, United States). 9-Fluorenylmethoxycarbonyl (Fmoc) protected L-α-amino acids, 1-[Bis(dimethylamino)methylene]-1H-1,2,3-triazolo [4,5-b]pyridinium-3-oxide hexafluorophosphate (HATU) and aminohexanoic acid (Ahx, X) were purchased from GL Biochem (Shanghai, China). Fmoc-PNA(Bhoc)-OH monomers were purchased from PANAGENE (Daejeon, South Korea). Ethyl cyano (hydroxyimino)acetate (Oxyma Pure) was obtained from Mimotopes (Melbourne, Australia). (1-Cyano-2-ethoxy-2-oxoethylidenaminooxy)dimethylamino-morpholino-carbenium hexafluorophosphate (COMU) was purchased from Chem Impex (United States). Dimethylformamide (DMF), diisoproplyethylamine (DIEA), piperidine and acetonitrile were obtained from Merck (Melbourne, Australia). 8-(9-Fluorenylmethyloxycarbonyl-amino)-3,6-dioxaoctanoic acid (Fmoc-miniPEG-OH) was purchased from IRIS Biotech GMBH (Marktredwitz, Germany). N,N′-Diisopropylcarbodiimide (DIC), triisopropylsilane (TIS), 3,6-dioxa-1,8-octanedithiol (DODT), 4-methylmorpholine (NMP), chloroacetic acid, acetic anhydride, 4-mercaptophenylacetic acid (MPAA), tris(2-carboxyethyl)phosphine hydrochloride (TCEP·HCl) and guanidine hydrochloride (Gn·HCl) were obtained from Sigma-Aldrich (Castle Hill, Australia). Trifluoroacetic acid (TFA) was sourced from Auspep (Melbourne, Australia). SsoAdvanced Universal SYBR Green Supermix and iScript reverse transcription supermix were purchased from BioRad and qPCR primers from Sigma-Aldrich (Melbourne, Australia). Sulfo-cyanine5 maleimide was obtained from Lumiprobe life science solution (Hunt Valley, Maryland). SMA patient-derived fibroblast cells (Type-I, GM03813) were purchased from Coriell Cell Repositories (NJ, United States). µ-Slide 4 Well Ph + Glass Bottom were purchased from ibidi GmbH (Gräfelfing, Germany).</p><!><p>The peptide sequence (HA2): ClAc-X-GLFHAIAHFIHGGWH (X: aminohexanoic acid, ClAc: chloroacetyl) was synthesized as a C-terminal amide on TentaGel XV RAM resin (100–200, 0.25 mmol/g) via Fmoc solid-phase peptide synthesis (SPPS) on a CEM LibertyTM microwave peptide synthesizer. It was synthesized at a 0.1 mmol scale using a 4-fold excess of Fmoc-protected amino acids, which were activated using DIC (4-fold excess) in the presence of OxymaPure (4-fold excess). The removal of Fmoc protecting groups was achieved using piperidine (20% v/v) in DMF. Coupling was carried out once at 90°C for 5 min, except for histidine which was coupled at 50°C for 10 min.</p><p>The ApoE peptide hydrazide was assembled manually via Fmoc solid-phase peptide synthesis on a 0.15 mmol scale using 2-chlorotrityl chloride resin as the solid support. After treating 0.3 mmol of resin with a solution of 5% hydrazine and 2% DIEA in DMF, a 50/50 mixture of Fmoc- and Boc-protected glycine was coupled to the resin via COMU/DIEA activation at 50°C (2 eq. of amino acid). This reduced the effective resin loading by approximately 50%, to minimize aggregation on the solid support during peptide chain elongation. The remaining residues were coupled via COMU/DIEA (2 eq.) at 50°C, followed by N-terminal acetylation via acetic anhydride/DIEA in DMF.</p><p>The resin-bound peptide was cleaved from the solid support by treatment with a cocktail of TFA:DODT:H2O:TIPS (94%:2.5%:2.5%:1%), 10 ml for 2 h at room temperature. The ApoE peptide was cleaved without DODT using TFA:H2O:TIPS (94%:2.5%:2.5%). Excess TFA was evaporated off and the cleaved peptide was precipitated by addition of ice-cold diethyl ether and centrifuged at 1,500 rpm for 4 min. The peptide pellet was washed in ice-cold diethyl ether twice more. Crude peptides were analyzed and purified by RP-HPLC on Phenomenex Jupiter columns (4.6 × 250 mm, C18, 5 µm) and (21.2 × 250 mm, C18, 10 µm) respectively. 0.1% trifluoroacetic acid in water was used as solvent A and acetonitrile containing 0.1% TFA as solvent B. A gradient of 20–50% B over 30 min was used at a flow rate of 1.5 ml/min for the analytical and 10 ml/min for the preparative column on a WATERS HPLC with a 996 photodiode array detector. Both peptides were purified to greater than 95% purity as determined by analytical RP-HPLC.</p><!><p>A 20-mer PNA antisense sequence for human survival motor neuron-2 (SMN2) (ATT​CAC​TTT​CAT​AAT​GCT​GG) was synthesized using Fmoc/Bhoc chemistry. The PNA was synthesized on TentaGel XV RAM resin (100–200, 0.25 mmol/g) using the Tribute automated peptide synthesiser at a 20 µmol scale. The Fmoc deprotection was carried out twice using Piperidine 20% in DMF at room temperature for 5 min. The coupling of PNA monomers was achieved using a 4-fold excess of Fmoc-protected PNA(Bhoc)-OH monomers dissolved in NMP activated with HATU (4 eq) and DIEA (8 eq). The coupling was carried out once at room temperature for 1 h. Following final PNA monomer coupling and deprotection, a miniPEG spacer was coupled, followed by incorporation of a cysteine residue at the N-terminus (5′-end). The PNA were cleaved from solid-support by treatment with a cocktail of TFA:H2O:TIPS (95:2.5:2.5, v/v/v) for 2 h. The excess TFA was evaporated and the PNAs were precipitated in ice-cold diethyl ether; the PNA pellet was then washed twice with more ether. The crude PNA was analyzed and purifed to greater than 95% purity as described above. The gradient for analysis and purification was 5–35% buffer B over 30 min.</p><!><p>The synthesis of the ApoE-PNA conjugate was achieved using native chemical ligation. ApoE-NHNH2 (1.5 µmol) was dissolved in 0.2 M sodium phosphate solution containing 6 M Gn.HCl (pH 3.0–3.3). The solution was cooled in an ice-salt bath to at least −15°C. To oxidize hydrazide to azide, NaNO2 (10 eq.) was added to the peptide solution on ice and stirred for 15 min. In a separate Eppendorf tube, Cys-PNA (1 µmol) and MPAA (75 µmol) were dissolved in 0.2 M sodium phosphate solution containing 6 M Gn.HCl (pH 3.0–3.3). The reaction mixture was removed from ice and the pH was adjusted to 6.9 with a solution of NaOH (1M). The ligation reaction was monitored by removing 5 µl and quenching it by adding 50 µl of 0.2 M sodium phosphate solution containing 6 M Gn.HCl (pH 3.0–3.3). The reaction was reduced by the addition of 20 µl of 0.1 M TCEP before analyzing with MALDI-TOF and RP-HPLC.</p><!><p>The trifunctional ApoE-PNA conjugates were synthesized by conjugating the maleimide-functionalized sulfo-Cy5 and PMO or chloroacetylated HA2 (Cl-HA2) to the thiol of cysteine generated during native chemical ligation of ApoE to PNA. ApoE-PNA and Maleimide-Sulfo-Cy5, Maleimide-PMO or Cl-HA2 (1.5 eq) were dissolved in 0.1 M phosphate buffer (pH 7.5). The two solutions were mixed and the pH readjusted to 7.5 with NH4HCO3, to offset the presence of TFA counterions. The reaction was monitored by MALDI-TOF mass spectrometry and the absence of peptide-PNA indicated completion of reaction. In the case of Cl-HA2, due to the lack of conjugate formation between the peptide-PNA and Cl-HA2, the pH of the reaction was increased, but unfortunately a precipitate started to appear, despite the addition of acetonitrile and heating. However, once the pH was lowered to <7, the precipitate disappeared, and the reaction solution was clear. The formation of the trifunctional ApoE-(HA2)-PNA conjugate was monitored with MALDI-TOF and a significant amount of unreacted ApoE-PNA and Cl-HA2 was observed. This was confirmed with HPLC analysis. As the reaction between chloroacetyl and thiol groups require a pH of >7.5 and Cl-HA2 is poorly soluble at this pH, the resultant yield was low for this conjugate. The concentration of the peptide-PNA and trifunctional conjugates was determined by measuring the molar absorption of the conjugates at 265 nm in 0.1 N HCl solution.</p><!><p>Matrix-assisted laser desorption ionization time-of-flight mass spectrometry (MALDI-8020, Shimadzu) was used to characterize the peptides and peptide-PNA conjugates with sinapinic acid as the matrix.</p><!><p>Spinal muscular atrophy (SMA)-patient-derived fibroblasts were used to test the antisense activity and Cy5-fluorescent uptake of the peptide-PNA and trifunctional peptide-PNA conjugates. The cells were maintained in DMEM with 10% fetal bovine serum, 1% L-glutamine and 1% penicillin-streptomycin at 37°C with 5% CO2. Fibroblasts were plated at a density of 3 × 105 cells per 2 ml per well in 6 well plates. The peptide-PNA and trifunctional peptide-PNA conjugate concentrations (1 and 2 µm) were made up in serum-free Opti-MEM and added to each well and incubated for 4 h at 37°C. The transfection medium was then replaced with growth medium and cells incubated for a further 20 h at 37°C. Cells were washed with PBS once and stored at −80°C until ready for RNA extraction.</p><p>For fluorescent uptake studies, SMA fibroblast cells were plated at a density of 60,000 cells/well of a µ-Slide 4 Well Ph+ Glass Bottom chamber slides precoated with poly-L-ornithine. The cells reached 80% confluency after 24 h. They were treated with Cy5-labelled peptide-PNA conjugates made up at a concentration of 1 and 2 µm in OptiMEM without serum for 1 h at 37°C. At 15 min to the end of incubation time, Hoechst (1:1,000 dilution) was added to each well. Cells were washed with PBS and images of cells were taken using the Leica SP8 Resonant Scanning microscope with 63×/1.4 oil objective. The images were acquired at the Cy5-specific excitation of 633 nm and emission of 638–779 nm and Hoechst excitation and emission of 405 nm and 410–585 nm, respectively.</p><!><p>Total RNA was extracted using the ISOLATE II RNA Mini Kit (Bioline, Australia) as per the manufacturer's protocol. The purified RNA (400 ng) was subsequently reverse transcribed to single-stranded complementary DNA (cDNA). The transcription was performed using the Veriti Thermal Cycler (ThermoFisher Scientific) following the thermal cycling conditions: 5 min at 25°C for priming, 20 min at 46°C for reverse transcription and finally 1 min at 95°C for reverse transcription inactivation, as stated in the manufacturer's protocol (BioRad, Australia).</p><p>Quantitative PCR was subsequently carried out using 20 ng of cDNA per well of 96-well plate in triplicates for each treatment, using 2X Fast SYBR Green Mastermix (ThermoFisher Scientific). The CFX96 Touch Real-Time PCR Detection Cycler, from BioRad, was used to perform the qPCR under the following conditions: 95°C for 2 min, followed by 39 cycles of amplifications, with 95 °C for 5 s and 60°C for 30 s, then 95°C for 5 s and finally to determine the melt curve, the temperature was increased from 65°C to 95°C using increments of 0.5°C. The amplification was carried out using SMN2-specific primers (Fw:5′-GCTTTGGGAAGTATGTTAATTTCA-3′, Rv: 5′-CTA​TGC​CAG​CAT​TTC​TCC​TTA​ATT-3′) for detecting full-length SMN2 mRNA transcripts and the human HPRT1 (Fw: 5′-GAC​CAG​TCA​ACA​GGG​GAC​AT-3′, Rv: 5′-CCT​GAC​CAA​GGA​AAG​CAA​AG-3′) as the reference gene. The cycle thresholds (Ct) of all triplicates were averaged and analyzed using the ∆∆Ct method, corrected against the house-keeping gene's Ct values. The resulting values were further normalized to the untreated control values, which were set to 1. Data were analyzed using GraphPad Prism (v 8.4.3, San Diego, CA, United States) and expressed as mean ± standard error of the mean (SEM) from at least three independent experiments. The statistical significance of the data was determined using One-way ANOVA with post-hoc Bonferroni and a p value of <0.05 was considered statistically significant.</p><!><p>The PNA was synthesized using previously established Fmoc/Bhoc solid-phase synthesis protocols (Tailhades et al., 2017) and functionalized with cysteine at the N-terminus (5′-end). A polyethylene glycol spacer (miniPEG) was introduced between the PNA and the cysteine. Cys-miniPEG-PNA was purified via reversed-phase HPLC and obtained in an excellent yield (54%) (Table.1, ESI Supplementary Figures S1C, S2C). The CPP [ApoE (133–150)] was synthesized as a C-terminal hydrazide and a miniPEG spacer was placed between the C-terminal glycyl hydrazide and the peptide sequence. The HA2 peptide was N-chloroacetylated to enable conjugation to the thiol side-chain of cysteine in the peptide-PNA conjugate. Both peptides were purified to greater than 95% purity as determined by analytical HPLC and obtained in a high yield (Table 1, ESI Supplementary Figures S1, S2). In order to synthesize the trifunctional peptide-(PMO)-PNA conjugate, a 20-mer PMO with the same sequence as the PNA (Table 1) was functionalized at its 3′-end with maleimidopropionic acid, as previously described (Tajik-Ahmadabad et al., 2017). The purified Maleimide-PMO was obtained in excellent yield (78%) (ESI Supplementary Figures S1D, S2D).</p><p>Conjugation of Cys-PNA with the peptide hydrazide was carried out by in situ oxidation of the C-terminal hydrazide with NaNO2 to the corresponding acyl azide, followed by thiolysis to obtain the required thioester (Fang et al., 2011; Zheng et al., 2013). There are three other reported methods describing the conjugation of peptides to ASOs via NCL (Stetsenko and Gait, 2000; Stetsenko and Gait, 2001; Diezmann et al., 2010; Jang et al., 2020). The first published method requires isolation of a thiol-functionalized ASO and a peptide N-terminally capped with a thioester (Stetsenko and Gait, 2000; Stetsenko and Gait 2001). This approach, however, requires the preparation of specialized precursors and/or additional deprotection steps. Isolation of peptide thioesters can sometimes be challenging due to the risk of hydrolysis during subsequent handling and purification. The other two approaches rely on modified reactive nucleobases such as Oxanine for incorporation into the ASO sequence during solid-phase assembly. A drawback of such approaches is opening of the nucelobase ring structure during ligation, which can potentially cause steric hindrance due to its close proximity to the thiol group (Jang et al., 2020). However, the in-situ approach used in this study, does not require any modifications or special reactive nucleobases such as Oxanine.</p><p>First, we synthesized the CPP-PNA conjugate (C1) by treating the CPP hydrazide with NaNO2 at pH 3.3 in a −15°C ice/salt bath for 20 min 4-mercaptophenylacetic acid (MPAA), followed by Cys-PNA, were then added to the reaction mixture and the pH was adjusted to 6.9. The reaction was completed in 30 min, as indicated by the disappearance of the signal corresponding to the Cys-PNA species in the MALDI-TOF and RP-HPLC spectra (Figure 1A). PNA with a cysteine deletion co-eluted during purification but nevertheless, the CPP-PNA conjugate (C1) was isolated and obtained in good yield (75%) based on the mass of starting Cys-PNA (ESI Figure 2A, Supplementary Figures S3A, S4A, Table.1). This is significantly higher than yields of 10–60% obtained using previously reported approaches for the solution-phase conjugation of peptide-oligonucleotides via amide bond formation (Diezmann et al., 2010; Shabanpoor and Gait, 2013).</p><!><p>Reverse-Phase HPLC analysis of (A) CPP-PNA conjugate (C1) formed using native chemical ligation (B–D) trifunctional conjugates C2, C3, and C4 respectively. O PNA without N-terminal Cysteine, # Unreacted peptide thioester, + Malimide-Sulfo-Cy5, ^ Mix of C1 conjugate and modified maleimide-PMO, *Unreacted Cl-HA2 peptide.</p><p>(A) RT-qPCR analysis of antisense activity (change in the level of full-length SMN2 mRNA) in SMA patient-derived fibroblasts treated with PNA (1 and 5 µm) and ApoE-PNA conjugate (C1-blue) and trifunctional conjugates C3 (green) and C4 (orange) (1 and 2 µm). N = 3, *p < 0.01, **p < 0.001 cf. untreated and PNA. #p < 0.05 cf. 1 µm C1, ##p < 0.001 cf. 2 µm C1. (B) Confocal microscopy images of fibroblasts treated with Cy5-labelled ApoE-PNA conjugate (C2) at 1 µm and 2 µm for 1 h at 37°C. Scale bar: 75 µm.</p><!><p>Subsequently, the CPP-PNA conjugate was further functionalized via the free thiol. The first trifunctional conjugate (C2) was prepared by forming a thioether between the free thiol group and a maleimide-functionalized fluorophore, Cy5, which was complete in 15 min (Figure 1B). This trifunctional conjugate was obtained in a high yield of 90% based on the amount of starting CPP-PNA material (C1) (Supplementary Figures S3B, S4B). The formation of the trifunctional conjugate C3 (CPP-(PMO)-PNA) (ESI Supplementary Figures S3C, S4C, Table 1) was also achieved using the thiol-maleimide "click" reaction to form a thioether linkage. Progress of the reaction was analyzed at 15 min (Figure 1C). MALDI-TOF analysis of the two peaks showed the peak eluting at 24.1 min to be the desired trifunctional conjugate. The peak at 22.9 min showed a mixture of unreacted CPP-PNA and an unidentified maleimide-PMO adduct (+80 m/z). An isolated yield of 45% was obtained for the C3 conjugate. The synthesis of trifunctional conjugate C4 (Figure 3) was achieved using the thiol-halide SN2 reaction instead of the thiol-maleimide Michael addition. The HA2 peptide was functionalized at the N-terminus with chloroacetyl via the corresponding carboxylic acid and carbodiimide activation. The SN2 reaction was initially performed at pH 7.5, but the expected product was not observed, therefore, the pH of the reaction mixture was increased to 8. This resulted in formation of desired trifunctional C4 conjugate (Figure 1D). However, we observed precipitate formation in the reaction mixture, due to insolubility of the HA2 peptide at the elevated pH, which is close to its iso-electric point. The poor solubility of HA2 under these reaction conditions led to a lower than expected yield (19.2%). This is intrinsic to this peptide and we believe that peptides with good solubility at pH 7-8 will have higher conjugation efficiency and yields. Nevertheless, as proof-of-concept, we have shown that the thiol group that is generated during NCL of CPP-PNA conjugates can be utilized to ligate a variety of functional moieties, to generate trifunctional CPP-ASO conjugates.</p><!><p>Synthesis of CPP-PNA conjugate (C1) using native chemical ligation. Initial step of peptide-hydrazide activation using NaNO2 (10 eq) at pH 3–3.5 and −15°C for 20 min followed by in situ thiolysis to generate thioester using MPAA (50 eq) which reacts with the N-terminal cysteine to form an amide linkage. Subsequent synthesis of trifunctional conjugates (C2, C3, C4) by coupling a variety of second functional moieties to the side-chain of cysteine through thioether linkages via thiol-halide SN2 and thiol-maleimide Michael addition reactions using 1.5-fold excess of maleimide- and haloacetyl-functionalized moieties at pH 7.5–8.</p><!><p>The activity of the peptide-PNA and trifunctional conjugates were evaluated in an SMN2 exon-inclusion assay using spinal muscular atrophy (SMA) patient fibroblasts. This assay measures the level of full-length SMN2 mRNA which indicates the efficacy of the peptide (ApoE) in delivering the conjugated PNA and PMO into the cells. The conjugates were tested at 1 and 2 µm concentrations to enable comparison of CPP efficiency in delivering the conjugated PNA, PMO and fluorophore into the cells. The maximum concentration for CPP-ASO conjugates was set at 2 µm, as higher concentrations would results in saturation of response based on our previous studies on cellular uptake of ApoE and similar peptides (Shabanpoor et al., 2017; Tajik-Ahmadabad et al., 2017). All conjugates significantly increased the level of full-length SMN2 compared to untreated cells, as measured by RT-qPCR (Figure 2A). The trifunctional C3 conjugate showed higher activity compared to the peptide-PNA (C1) conjugate due to the higher amount of ASO being delivered into the cells coupled to a single CPP. Confocal microscopy imaging of the Cy5-labelled trifunctional (C2) conjugate showed a concentration-dependent increase in uptake in fibroblasts (Figure 2B). The higher level of C2 uptake at higher concentration correlates with the level of activity obtained for C1, which is the unlabelled analogue of C2. Given that the trifunctional C4 conjugate has an endosomal disrupting peptide, HA2, it was expected to have higher antisense efficacy compared to the peptide-PNA conjugate (C1). Although C4 showed a concentration dependent increase in the activity, which is higher than PNA alone, the antisense activity is not statistically significant compared with activity of C1. This can be due to the partial solubility of the C4 in the cell culture media which prevents its efficient cellular uptake.</p><!><p>In summary, we have demonstrated for the first time the use of hydrazide-based NCL as an efficient and site-specific approach for the preparation of trifunctional peptide-ASO conjugates. This methodology utilizes the thiol artifact from the NCL as an additional biorthogonal handle for further functionalization. We have demonstrated that incorporation of a second functional moiety does not hamper CPP-PNA uptake or activity. Here we have also shown that two ASO sequences conjugated to a single CPP can be simultaneously delivered into cells. In principle, this method is applicable to the preparation of any trifunctional CPP-ASO conjugate.</p><!><p>The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding author.</p><!><p>MD, JK, and FS performed chemical syntheses, experimental design and drafted the manuscript; AA and BT performed confocal imaging experiment and the bioassays. All authors worked on the manuscript.</p><!><p>This work was supported by National Health and Medical Research Council (Project Grants 1138033 to FS and BT). AA is a recipient of Bethlehem Griffith Research Foundation scholarship (1808). FS is a recipient of FightMND Mid-career Fellowship (04MCR). BT is a recipient of NHMRC-ARC Dementia Research Leadership Fellowship (1137024) and Stafford Fox Medical Research Foundation Grant.</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>The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fchem.2021.627329/full#supplementary-material.</p><!><p>Click here for additional data file.</p>

PubMed Open Access

Binding of the Microbial Cyclic Tetrapeptide Trapoxin A to the Class I Histone Deacetylase HDAC8

Trapoxin A is a microbial cyclic tetrapeptide that is an essentially irreversible inhibitor of class I histone deacetylases (HDACs). The inhibitory warhead is the \xce\xb1,\xce\xb2-epoxyketone side-chain of (2S,9S)-2-amino-8-oxo-9,10-epoxydecanoic acid (L-Aoe), which mimics the side-chain of the HDAC substrate acetyl-L-lysine. We now report the crystal structure of the HDAC8\xe2\x80\x93trapoxin A complex at 1.24 \xc3\x85 resolution, revealing that the ketone moiety of L-Aoe undergoes nucleophilic attack to form a zinc-bound tetrahedral gem-diolate that mimics the tetrahedral intermediate and its flanking transition states in catalysis. Mass spectrometry, activity measurements, and isothermal titration calorimetry confirm that trapoxin A binds tightly (Kd = 3 \xc2\xb1 1 nM) and does not covalently modify the enzyme, so the epoxide moiety of L-Aoe remains intact. Comparison of the HDAC8\xe2\x80\x93trapoxin A complex with the HDAC6-HC toxin complex provides new insight regarding the inhibitory potency of L-Aoe-containing natural products against class I and class II HDACs.

binding_of_the_microbial_cyclic_tetrapeptide_trapoxin_a_to_the_class_i_histone_deacetylase_hdac8

<p>Ever since the discovery of histone acetylation in transcriptional regulation more than 50 years ago,1 thousands of histone and non-histone proteins have been identified as lysine acetylation targets in the mammalian acetylome.2-4 The reversible acetylation of L-lysine side chains is a critical molecular strategy for the regulation of protein function in vivo and rivals phosphorylation in the regulation of diverse biological processes, including the cell cycle and central carbon metabolism.5,6 The acetylation of a specific L-lysine residue on a target protein is catalyzed by a histone acetyl transferase (HAT) using acetyl CoA as a co-substrate;7,8 the hydrolysis of acetyl-L-lysine to form free L-lysine and acetate is catalyzed by a histone deacetylase (HDAC).9,10</p><p>Upregulated HDAC activity is associated with tumorigenesis, neurodegenerative diseases, and immune disorders, so this enzyme family represents an attractive target for therapeutic intervention.11,12 Phylogenetic analysis indicates four classes of deacetylases: class I HDACs (1, 2, 3, and 8), class IIa (4, 5, 7, and 9) and class IIb (6 and 10) HDACs, class III HDACs (sirtuins 1-7), and the single class IV enzyme HDAC11.13 Class I, II, and IV HDACs require Zn2+ or Fe2+ for optimal catalytic activity.14 The first crystal structure of a metal-dependent deacetylase15 revealed an α/β fold identical to that first observed in arginase.16,17 This fold is distinct from that adopted by sirtuins, which employ a different catalytic mechanism that requires the cofactor NAD+.18 Among the metal-dependent HDACs, HDAC8 was the first to yield a crystal structure.19,20 Important catalytic residues in the HDAC8 active site include tandem histidines (electrostatic catalyst H142 and general base-general acid H143)21,22 and a conformationally-flexible23 tyrosine residue that assists the Zn2+ ion in the polarization of the substrate carbonyl group.24,25</p><p>The tightest-binding HDAC inhibitors are those capable of coordinating to the active site Zn2+ ion and also forming hydrogen bonds with catalytically-important active site residues. Currently, four HDAC inhibitors are approved for clinical use in cancer chemotherapy, one of which is the cyclic depsipeptide romidepsin.26,27 Macrocyclic HDAC inhibitors such as romidepsin or the related depsipeptide largazole28,29 comprise rigid scaffolds to which a Zn2+-coordinating functional group is attached through a linker that is the approximate length of a lysine side chain.</p><p>Trapoxin A, first isolated from the microbial parasite Helicoma ambiens, is a macrocyclic tetrapeptide HDAC inhibitor with the amino acid sequence cyclo-[L-Phe–L-Phe–D-hPro–L-Aoe] (hPro = homoproline, also known as pipecolic acid; L-Aoe = (2S,9S)-2-amino-8-oxo-9,10-epoxydecanoic acid).30 The α,β-epoxyketone moiety of the L-Aoe side chain serves as the Zn2+-coordinating group and its ketone carbonyl group is isosteric with the scissile carbonyl of the HDAC substrate acetyl-L-lysine (Figure 1). Trapoxin A is proposed to act as an irreversible inhibitor of class I HDACs31,32 and was used for the first isolation of HDAC1 from the nuclear extracts of human Jurkat T cells.33 The epoxide moiety of the L-Aoe side chain was thought to react with the nucleophilic side chain of an active site residue; however, a covalent enzyme-inhibitor complex was not observable by SDS-PAGE/autoradiography.32 Curiously, although trapoxin A was found to irreversibly inhibit HDAC1, a class I enzyme, it was found to reversibly inhibit the class II enzyme HDAC6.34 The molecular basis of these activity differences has remained unclear in the absence of structural data.</p><p>Here, we report the first X-ray crystal structure of trapoxin A complexed with a class I histone deacetylase, HDAC8, at 1.24 Å resolution (Figure 2a). Experimental procedures, including the preparation of a new HDAC8 construct for crystallographic studies, are outlined in the Supporting Information (data collection and refinement statistics are recorded in Supporting Information Table 1). The structure of the HDAC8–trapoxin A complex reveals that the conformation of the tetrapeptide backbone of trapoxin A is identical to that observed in the crystal structure of the uncomplexed inhibitor,30 with root-mean-square (rms) deviations of 0.21–0.24 Å for 16 main chain atoms of the cyclic peptide backbone; the L-Aoe side chains adopt different conformations (Supporting Information Figure S1). All peptide linkages of trapoxin A adopt the trans configuration except for the Phe-hPro linkage, which forms a cis-peptide. All three backbone NH groups donate hydrogen bonds to the side-chain carboxylate group of D101. The hydrogen bonds between D101 and the backbone NH groups of L-Aoe and the adjacent L-Phe residue are similar to those observed between D101 and linear tetrapeptide substrates bound to inactivated HDAC8 (Supporting Information Figure S2).24,25 Apart from interactions described below for the L-Aoe side chain, no other direct enzyme-inhibitor hydrogen bonds are observed. Although there are no intramolecular hydrogen bonds in trapoxin A, the side chains of its two L-Phe residues make a favorable quadrupole-quadrupole interaction with edge-to-face geometry.</p><p>Conformational changes are required in the enzyme active site to accommodate the steric bulk of the rigid peptide macrocycle in comparison with substrate binding, primarily in the L2 loop (residues G86-I108, of which D87-I94 are disordered). Relative to the structure of H143A HDAC8 complexed with a tetrapeptide substrate,25 the greatest change is observed for Y100, which undergoes a 116° change in side chain torsion angle χ1 (Supporting Information Figure S2). Similar conformational flexibility of Y100 accommodates the binding of the macrocyclic depsipeptide inhibitor largazole.29 In the HDAC8–trapoxin A complex, the conformational change of Y100 is triggered by one of the inhibitor L-Phe residues, with which one of two observed conformers makes a favorable edge-to-face interaction.</p><p>Surprisingly, the ketone carbonyl of the L-Aoe side chain undergoes nucleophilic attack by water upon binding to HDAC8, such that the inhibitory species is a gem-diolate (or perhaps a gem-diol) stabilized by Zn2+ coordination and three hydrogen bonds. Thus, trapoxin A mimics the binding of the tetrahedral intermediate and its flanking transition states in catalysis; the origins of high affinity are undoubtedly rooted in the fact that trapoxin A binds as an analogue of the postulated transition state. The Zn2+–O1 and Zn2+–O2 distances for the gem-diolate are 2.5 Å and 1.9 Å, respectively. The O1 hydroxyl group also forms hydrogen bonds with H142 and H143 (O–O separations = 2.7 Å each), and the O2 oxyanion accepts a hydrogen bond from Y306 (O–O separation = 2.6 Å). While it is unusual to see an unactivated ketone binding as a tetrahedral gem-diolate, which exists to less than 0.2% in aqueous solution (based on the hydration of the unactivated ketone carbonyl of acetone in aqueous solution35), it is notable that the L-Aoe side chain of the cyclic tetrapeptide inhibitor HC toxin (Figure 1) similarly undergoes nucleophilic attack in the recently-determined structure of its complex with catalytic domain 2 of HDAC6.36 This behavior is also reminiscent of the binding of unactivated aldehyde and ketone substrate analogues to carboxypeptidase A, which similarly bind as tetrahedral gem-diolate transition state analogues.37,38</p><p>In the HDAC8–trapoxin A complex, the epoxide ring of the L-Aoe side chain is clearly intact and makes no hydrogen bond interactions with any enzyme residues or water molecules (Figure 2b). The closest side chains to the epoxide moiety are those of W141, H142, C153, and Y306 with interatomic separations of 3.3–3.7 Å. The epoxide moiety is believed to be required for essentially irreversible inhibitory activity against class I HDACs, based on the lack of irreversible inhibitory activity for the cyclic tetrapeptide inhibitor apicidin (Figure 1), which lacks an epoxide moiety.30 Thus, it is curious that the epoxide moiety of trapoxin A does not react with the enzyme. However, the crystal structure reveals that although the epoxide moiety binds in the vicinity of catalytic general base H143 and highly conserved C153, neither of these potential nucleophiles is positioned or oriented for nucleophilic attack at the epoxide (Figure 2b).</p><p>We confirmed the irreversibility of trapoxin A inhibition by assaying HDAC8 activity following multiple rounds of dialysis of the enzyme-inhibitor complex. Regeneration of activity was not observed for HDAC8 preincubated with a 10-fold molar excess of trapoxin A, but was observed under the same conditions using apicidin (Figure 3). This result confirms that the epoxide moiety is required for essentially irreversible inhibition. However, this result does not prove that a covalent enzyme-inhibitor complex is formed.</p><p>To study the covalent modification of HDAC8 by trapoxin A in solution, we employed liquid chromatography-tandem mass spectrometry (LC-MS/MS) on HDAC8 preincubated for 18 hours with 10 molar equivalents of trapoxin A and subsequently digested with trypsin. Prior to digestion, mass peaks corresponding to HDAC8 covalently modified with one or two trapoxin A molecules were observed for this sample by MALDI mass spectrometry (Supporting Information Figure S3). Following trypsin digestion, LC-MS/MS analysis, and sequence analysis for residue modifications, a mass shift corresponding to the molecular weight of trapoxin A was sporadically observed for various cysteine and histidine residues located on the surface of HDAC8 (Supporting Information Figure S4 and Table S2). Only nonspecific labeling of the enzyme was observed in the presence of excess inhibitor, with no particular preference for covalent modification in the active site. Additionally, incubation of trapoxin A for 1 hour in the presence and absence of HDAC8 followed by LC-MS analysis indicated the presence of only intact trapoxin A (i.e., with an intact epoxide ring; Supporting Information Figure S5). These results strongly suggest that trapoxin A is simply an exceptionally tight-binding, noncovalent transition state analogue inhibitor of HDAC8. As noted by Schramm and colleagues,39 transition state analogues with essentially irreversible binding behavior are feasible for enzymes that exhibit catalytic rate enhancements of 1010 or greater, which is likely the case for an amide hydrolase such as HDAC8 based on uncatalyzed amide bond hydrolysis half-lives measured in centuries by Radzicka and Wolfenden.40</p><p>Isothermal titration calorimetry (ITC) measurements yield HDAC8 dissociation constants Kd = 3 ± 1 nM for trapoxin A and Kd = 250 ± 70 nM for apicidin (Supporting Information Figure S6), indicating 83-fold enhanced binding affinity for trapoxin A relative to apicidin. Given the structural similarity between the two cyclic tetrapeptides, and the fact that the macrocycle makes very few intermolecular interactions in the HDAC8–trapoxin A complex, it is clear that the epoxide moiety of trapoxin A is indeed responsible for tight binding to HDAC8.</p><p>ITC measurements of HDAC8 active site variants indicate that H142 is critical for trapoxin A binding just as it is important for catalysis, since the H142A mutation results in significantly weaker affinity with Kd = 17 ± 5 μM – H142A HDAC8 exhibits a 233-fold reduced kcat/KM value relative to wild-type HDAC8.22 This residue functions in catalysis with a positively charged imidazolium group that serves as an electrostatic catalyst.22 Electron density is observed for the Nε-H proton of H142 in the 1.24 Å electron density map (Supporting Information Figure S7), so the H142 imidazolium group must similarly stabilize the zinc-bound gem-diolate.</p><p>Although the L-Aoe side chains of trapoxin A and HC toxin bind to HDAC8 and HDAC6, respectively, as the gem-diolate, there are notable differences between the structures of each enzyme-inhibitor complex that may explain the tighter binding of these inhibitors to class I HDACs.34 In the class IIb HDAC6-HC toxin complex (Ki = 350 nM),36 the C–O bond of the epoxide adopts an energetically unfavorable eclipsed conformation with respect to the C–O bond of the hydroxyl group of the zinc-bound gem-diolate (Figure 4a); in the class I HDAC8–trapoxin A complex (Kd = 3 nM), the C–O bond of the epoxide adopts an energetically favorable staggered conformation (Figure 4b). The energetically unfavorable eclipsed conformation in the HDAC6-HC toxin complex appears to be caused by the bulky P571 residue in the L3 loop, conserved in all class II HDACs – if the epoxide adopted an energetically favorable staggered conformation in the HDAC6 active site, the epoxide methylene group would clash with P571. In class I HDACs, P571 is not conserved and the active site is more open. Thus, a more favorable binding conformation is accessible to the epoxyketone moiety only in the active site of a class I HDAC.</p><p>Although the α,β-epoxyketone epoxide moiety of trapoxin A remains intact in the crystal structure of its complex with HDAC8, it is notable that this novel functionality is chemically reactive in the binding of inhibitors to other enzymes. For example, the proteasome inhibitor carfilzomib contains an α,β-epoxyketone that forms a covalent adduct to block proteasome function.41 The crystal structure of a similar natural product, epoxomicin, complexed with the yeast 20S proteasome indicates a multistep cyclization sequence leading to the formation of a morpholino ring between the former α,β-epoxyketone of the inhibitor and the reactive N-terminal threonine residue of the proteasome subunit.42 A two-step mechanism for inhibitor binding is initiated by nucleophilic attack of the threonine hydroxyl group at the epoxyketone carbonyl followed by a 6 Exo-Tet ring closure reaction between the α-amino group of threonine and the epoxide to generate the morpholino product.43 The chemistry of inhibitor binding in this system indicates that the carbonyl group appears to be more reactive than the epoxide of the α,β-epoxyketone moiety. This is consistent with reactivity trends observed in organic synthesis for various α,β-epoxycarbonyl derivatives, where the carbonyl group preferentially undergoes nucleophilic addition while leaving the epoxide moiety intact.44-46 With respect to trapoxin A, it is notable that an additional peak is observed in mass spectra consistent with gem-diol formation even in the absence of enzyme (Supporting Information Figure S5).</p><p>Finally, it is interesting to compare the structures of zinc coordination polyhedra in different HDAC8-inhibitor complexes (Supporting Information Figure S8). Only one oxygen of the gem-diolate of trapoxin A is sufficiently close for inner-sphere coordination, so the overall zinc coordination geometry is best described as 4-coordinate distorted tetrahedral. In this regard, zinc coordination geometry approaches that observed in the HDAC8–largazole complex, which exhibits nearly perfect tetrahedral coordination geometry through the binding of the largazole thiolate group.29 In contrast with these examples, the hydroxamate group of trichostatin A coordinates to zinc in bidentate fashion, so that the overall coordination geometry is 5-coordinate square pyramidal.20</p><p>In conclusion, the current work demonstrates that trapoxin A is an essentially irreversible noncovalent inhibitor of HDAC8: the α,β-epoxyketone side chain of the inhibitor undergoes nucleophilic attack by zinc-bound water to bind as a tetrahedral gem-diolate transition state analogue. Along with a favorable staggered conformation of the intact epoxide moiety relative to the zinc-bound gem-diolate, these structural features contribute to an exceptionally tight enzyme-inhibitor complex effectively locked into the enzyme active site.</p>

PubMed Author Manuscript

Controllable stereoinversion in DNA-catalyzed olefin cyclopropanation <i>via</i> cofactor modification

The assembly of DNA with metal-complex cofactors can form promising biocatalysts for asymmetric reactions, although catalytic performance is typically limited by low enantioselectivities and stereocontrol remains a challenge. Here, we engineer G-quadruplex-based DNA biocatalysts for an asymmetric cyclopropanation reaction, achieving enantiomeric excess (ee trans ) values of up to +91% with controllable stereoinversion, where the enantioselectivity switches to À72% ee trans through modification of the Fe-porphyrin cofactor. Complementary circular dichroism, nuclear magnetic resonance, and fluorescence titration experiments show that the porphyrin ligand of the cofactor participates in the regulation of the catalytic enantioselectivity via a synergetic effect with DNA residues at the active site. These findings underline the important role of cofactor modification in DNA catalysis and thus pave the way for the rational engineering of DNA-based biocatalysts.

controllable_stereoinversion_in_dna-catalyzed_olefin_cyclopropanation_<i>via</i>_cofactor_modificati

Introduction<!>Results and discussion<!>Conclusions

<p>Cyclopropane motifs feature in many natural products and medicinal agents 1 which constitute versatile intermediates for the total synthesis of therapeutic compounds. 2 As these compounds are in high demand, signicant effort has been devoted to the development of cyclopropane synthesis, in particular via the use of hemoprotein enzymes engineered via directed evolution. Cytochrome P450 (ref. 3 and 4 ) and myoglobin [5][6][7][8] enzymes have been evolved to catalyze asymmetric olen cyclopropanations with excellent performance. [9][10][11][12][13][14] This method has been applied to the synthesis of drug molecules 15,16 and natural product scaffolds. 17,18 Despite this progress, the developed biocatalytic protocols are generally restricted to protein enzyme engineering. 19 The discovery of the catalytic functions of nucleic acids 20,21 has expanded the breadth of biocatalytic protocols to include RNA and DNA catalysts, initiating the pursuit of nucleic acidbased enzymes. DNA possesses inherent advantages as a catalyst. A catalytic sequence can be entirely identied from a random sequence population and folds into its practical tertiary structure spontaneously. A number of asymmetric reactions, especially Lewis acid-catalyzed reactions, have been successfully realized using DNA-based biocatalysts resulting in remarkable performances. [22][23][24][25][26][27][28][29][30][31] Recently, the Roelfes 32 and Sen 33 groups expanded the scope of reactions catalyzed by DNA-based biocatalysts to include olen cyclopropanation. Although the enantioselectivities achieved were moderate, this has paved the way for DNA-catalyzed carbene transfer reactions. Therefore, designing a DNA-based biocatalyst that catalyzes cyclopropanation in high enantiomeric excess (ee) remains a challenge, especially to achieve an ee greater than 90%. 34 DNA catalysis is opening a promising avenue for biosynthesis, but the catalytic scope and performance of DNA catalysts still need improvement in comparison with catalysts based on protein enzymes. Considering the great role played by the cofactor ligand in the rst-coordination-sphere in catalytic reactions, [35][36][37][38][39][40] cofactor modication, which is as powerful as directed evolution [41][42][43][44] but seriously disregarded in the eld of DNA catalysis, is introduced for the development of DNA-based biocatalysts. Here, we report a cyclopropanation reaction catalyzed by a G-quadruplex (G4)-Fe-porphyrin biocatalyst that results in enantioselectivity as high as 91%. By tuning the N-methyl position of Fe-porphyrin from the para-to the ortho-position, the catalytic enantioselectivity of the reconstituted G4-Fe-porphyrin biocatalyst reverses to À72% ee trans . Complementary spectral, nuclear magnetic and isothermal titration characterization studies reveal that the stereo-divergence of the product mainly arises from the participation of the porphyrin ligand in the regulation of the enantioselectivity. This work succeeds in diversifying the functionality of DNA-based biocatalysts via cofactor modication, highlighting the great potential for cofactor modication in the eld of DNA-based biocatalyst engineering.</p><!><p>Fig. 1a illustrates the design of the G4-based biocatalysts. Three Fe-meso-tetra-(N-methylpyridyl)porphyrins with para-(FeTM-PyP4), meta-(FeTMPyP3), and ortho-(FeTMPyP2) N-methyl substituents were chosen as parallel cofactors to investigate the effect of the rst-coordination-sphere on the catalytic performance. The non-covalent binding of the cofactors with mA9A G4 (d[G 2 T 2 G 2 TGAG 2 T 2 G 2 A]), a thrombin binding aptamer (TBA) variant, formed the G4-based biocatalysts. The assembled biocatalysts were then tested in a reaction between styrene and ethyl diazoacetate (EDA), which results in a chiral cyclopropane product (Fig. 1b).</p><p>Table 1 lists the results obtained using the Fe-porphyrins and their corresponding G4 biocatalysts. The use of the free FeTMPyPn (n ¼ 4, 3, 2) cofactor as the catalyst led to relatively low activities and no chiral induction (Table 1, entries 1-3). The biocatalysts (mA9A-FeTMPyPn, n ¼ 4, 3, 2) assembled from FeTMPyPn and mA9A G4 signicantly improved the catalytic activities with turnover frequencies (TOF) increased about 10fold when compared to those of the free FeTMPyPns. More surprisingly, mA9A-FeTMPyP2 induces the inversion of the enantioselectivity relative to that of mA9A-FeTMPyP4 from +74% to À46% (Table 1, entries 4-6). To further understand the complementary chiral induction mechanism, circular dichroism (CD), nuclear magnetic resonance (NMR), uorescence titration and other characterization methods were then performed.</p><p>The CD spectra in Fig. 2a show that the antiparallel G4 signatures of mA9A, featuring two positive peaks at 245 nm and 295 nm, and one negative peak at 265 nm, are still present in the mA9A-FeTMPyPn (n ¼ 4, 3, 2) catalysts. But of note is that the CD spectrum of mA9A-FeTMPyP2 shows a specically induced CD signal (ICD) at 420 nm. Only when an asymmetric conformation is formed can an ICD signal be induced. DNA does not absorb at 420 nm, but FeTMPyP2 does (Fig. 2b). Therefore, the ICD signal can be attributed to FeTMPyP2, whose planar symmetry is broken due to the interaction with mA9A.</p><p>The ultraviolet-visible (UV) absorption spectra (Fig. 2b) of mA9A-FeTMPyPn further support the CD results. The spectra of FeTMPyPn (n ¼ 4, 3, 2) exhibit broad Soret absorption bands at around 420 nm. As the N-methyl group varies from the para-to the ortho-position, the Soret band shows a blue shi, indicating that FeTMPyP2 has decreased electronic conjugation relative to the other porphyrins. Due to the steric hindrance of the 2-Nmethyl group, the bond between the pyridine ring and the porphyrin ring of FeTMPyP2 has a tendency to rotate, causing the pyridine ring and the porphyrin ring to be non-co-planar (Fig. 2b, right inset), which explains the generation of an ICD signal in mA9A-FeTMPyP2. When well-folded mA9A is added to FeTMPyPn, the Soret bands rst fall to a minimum and then slightly rise for all cofactors. This suggests that FeTMPyPn (n ¼ 4, 3, 2) have a similar external stacking mode on mA9A, rather than intercalating between two G-quartets. [45][46][47] Moreover, binding with FeTMPyP4 and FeTMPyP3 makes mA9A more stable according to UV melting experiments (Fig. S1 †), while binding with FeTMPyP2 does not.</p><p>To obtain further structural information on the mA9A-FeTMPyPn catalysts, we conducted an NMR characterization of the systems. The NMR spectrum of mA9A shows seven discrete peaks over the range of 11.5-12.5 ppm (Fig. 2c), which can be assigned to the eight guanine imino protons (H1) of the Gquartet. These signals indicate the formation of a two-layer antiparallel G4 structure in the potassium phosphate buffer (Fig. 2d), like that of TBA. 48,49 NMR titration experiments (Fig. 2c) where mA9A is added to the three FeTMPyPn (n ¼ 4, 3, 2) cofactors show similar line broadening and reductions in peak intensity. However, there are two adjacent peaks around 12.0 ppm (peaks labelled with *) with different relative declines, indicating that the coordination modes between mA9A and the different Fe-porphyrin cofactors vary slightly. Moreover, the addition of FeTMPyP2 causes an obvious upeld peak-shi, which is attributed to the strong electronic shielding effect arising from the porphyrin ligand stacking upon the Gquartet. 50 Therefore, in combination with the CD and UV-vis results, we nd that FeTMPyP2 adjusts to an optimal conformation through bond rotation and distortion, thereby forming a tight p-p stacking mode with mA9A and creating a strong electronic shielding effect on the imino protons of the Gquartet.</p><p>A uorescence-based binding assay was then implemented to locate the catalytic sites. 51 Since UV titration had determined the external stacking mode of the FeTMPyPn (n ¼ 4, 3, 2) cofactors on mA9A, the 5 0 end (5 0 FAM-mA9A) and loop 1 (int 0 FAM-mA9A) of the two G-quartets of mA9A were labelled separately with the uorophore 5-carboxyuorescein (FAM). The uorescence of FAM can be suppressed by the presence of FeTMPyPn (Fig. 3a and S2 †). By adding FeTMPyPn dropwise to the FAM-labelled mA9A, we obtained the titration quenching curves (Fig. S3 †). The tted apparent equilibrium dissociation constants (Kd app ) at the different FAM-labelled sites, shown in Fig. 3b, indicate that all the FeTMPyPn (n ¼ 4, 3, 2) cofactors prefer binding at the 5 , 3 0 end of the G-quartet. Considering that the ratio of mA9A and FeTMPyPn is greater than 1 : 1 when assembling the DNA-based biocatalyst and that the ee trans values for the catalytic cyclopropanation can be maintained at the highest level, the preferential binding site of FeTMPyPn with mA9A is regarded as the active center for chiral regulation. Therefore, according to the uorescence-based binding assay, FeTMPyPn binding at the 5 0 , 3 0 -end of the G-quartet of mA9A constructs the active catalytic site of mA9A-FeTMPyPn. The Kd app of FeTMPyP2 is the lowest at 38 nM, compared to 70 nM for FeTMPyP3 and 52 nM for FeTMPyP4, indicating that FeTMPyP2 binds the strongest with mA9A.</p><p>Isothermal titration calorimetry (ITC) provides information on intermolecular interactions by recording the heat discharged or consumed during a bimolecular reaction. Fig. 3c shows that there are several distinct exothermic processes during the titration of mA9A with FeTMPyP4 and FeTMPyP3, indicating complicated multiple binding behaviours between FeTMPyP4, FeTMPyP3 and mA9A. As explained in the above section, although mA9A enables the binding of multiple iron porphyrins, the strongest binding sites were the catalytic sites for the chiral regulation of the cyclopropanation reaction. The titration curves were tted using a sequential binding sites model to calculate the binding parameters. The highest affinities between FeTMPyPn and mA9A were quantied to be 33 nM (n ¼ 4), 50 nM (n ¼ 3) and 14 nM (n ¼ 2), coinciding with the trend of Kd app as measured using uorescence titration. This further supports the conclusion that the 5 0 , 3 0 -end of the G-quartet of mA9A constructs the active catalytic sites. Given that FeTMPyP2 has exible peripheral groups (according to UV spectra), the tight assembly with mA9A can be attributed to the conformational transition of the TMPyP2 ligand, which is also conrmed by the CD spectra.</p><p>NMR, UV and uorescence titration studies indicate that ne-tuning the N-methyl position of the cofactor does not change the preference of FeTMPyPn for binding at the 5 0 , 3 0 -end of the G-quartet of mA9A. This suggests that a similar active DNA pocket for cyclopropanation catalysis is provided by all the mA9A-FeTMPyPn (n ¼ 4, 3, 2) catalysts (Fig. 4a). The ligands TMPyP4 and TMPyP3 are planar symmetric molecules that are unable to induce chirality in the catalytic process of the mA9A-FeTMPyP4 and mA9A-FeTMPyP3 catalysts. It is the deoxynucleotide residues at the 5 0 , 3 0 -end of the G-quartet of mA9A that hold the iron porphyrin carbene (IPC) intermediate in a certain orientation and dene the conguration of the product (Fig. 4b). Nevertheless, mA9A-FeTMPyP3 shows a similar but lower enantioselectivity than mA9A-FeTMPyP4. Considering the similar binding strengths of FeTMPyP3 with the two G-quartets of mA9A (Fig. 3b), the reduction in enantioselectivity can be reasonably attributed to the multi-site binding behaviour of FeTMPyP3. In contrast to mA9A-FeTMPyP4 and mA9A-FeTMPyP3, the symmetry breaking of the porphyrin ring in TMPyP2 allows it to participate in chiral regulation. This, in synergy with the deoxynucleotide residues, induces the formation of the cyclopropane product with the opposite conguration to that catalyzed by mA9A-FeTMPyP4 (Fig. 4b).</p><p>To extend the substrate scope of the mA9A-FeTMPyPn catalyzed cyclopropanation reaction, a series of olens and diazoesters were investigated (Table 2). All three mA9A-FeTMPyPn (n ¼ 4, 3, 2) catalysts show obvious substrate specicities. The catalytic enantioselectivities of mA9A-FeTMPyP4 and mA9A-FeTMPyP3 toward olen substrates with substituents on the phenyl group are reduced (entries 1 vs. 2-6). The enlargement of the diazoester functional group from ethyl (Et) to -CCH 3 (i-Pr) 2 (i-Pr is the abbreviation for isopropyl) enables a signicant enhancement in the ee trans values to 91% (entries 1 vs. 8-10). However, the use of a -CH(Cy) 2 (Cy is the abbreviation for cyclohexyl, entry 11) group does not result in an enhancement.</p><p>Several studies have reported that the modication of diazoester substituents causes a signicant impact on the activities and selectivities of carbene transfer reactions. 6,52 The G4-Feporphyrin-catalyzed cyclopropanation has been characterized to proceed through a catalytic IPC intermediate. 34 Diazoester reagents attack the active [Fe] center to form an IPC intermediate and release one molecule of N 2 . An appropriate substituent on the diazoester reagent can coordinate with the deoxynucleotide residues (especially dA9) through hydrophobic interactions to directly determine the conformation and properties of the IPC intermediate. Therefore, a -CCH 3 (i-Pr) 2 substituent promotes IPC convergence to a single well-dened orientation, and the steric hindrance of the deoxynucleotide residues allows one face of the IPC to be exposed to the olen, while keeping the other inaccessible, resulting in high enantioselectivity. For mA9A-FeTMPyP2, variation in the substituents on the phenyl group of the olen increases the enantioselectivity to À72% ee trans (entry 6), but the catalytic cyclopropanation activities are generally lower than those of the other two biocatalysts. Although the three mA9A-FeTMPyPn (n ¼ 4, 3, 2) catalysts show different responses to substrate variation, they are all trans product-selective with trans/cis ratios of more than 86 : 14, and almost nonselective toward the R1substituted olen substrate (entry 7).</p><!><p>In conclusion, stereo-divergence of G4 biocatalyst-catalyzed olen cyclopropanation was achieved via cofactor modication. By tuning the N-methyl substituent of the porphyrin ligand in the cofactor from the para-to the ortho-position, the selfassembled G4-Fe-porphyrin biocatalysts are able to switch the enantioselectivity of the reaction from +91% to À72% ee trans . CD, NMR, ITC, and other characterization studies reveal that the porphyrin ligand cooperating with the deoxynucleotide residues gives the IPC intermediate a single well-dened conguration and results in a specic enantiopreference. This nding is down to the rational design of DNA-based biocatalysts through cofactor modication, a method which serves as an effective way to regulate the catalytic performance of DNA-based biocatalysts.</p>

Royal Society of Chemistry (RSC)

High-Accuracy Determination of Potassium and Selenium in Human Serum by Two-Step Isotope Dilution ICPMS

A high-accuracy measurement technique for determining potassium and selenium in human serum was developed by using two-step Isotope Dilution Inductively Coupled Plasma-Mass Spectrometry (ICPMS) in this research. A more accessible method of the quadrupole ICPMS was employed in this research to achieve an equally high accuracy which had been achieved by a much more expensive method, namely, high-resolution sector field ICPMS (SF-ICPMS), with a comparatively easy and simple operation. In addition, we have evaluated the uncertainty of this method. The results showed that the determination limits of potassium and selenium in serum were 0.8 mg/kg and 2.7 μg/kg, respectively, and the precision for both measurements was lower than 0.2% and 0.7%. The measurement, when employed to measure potassium and selenium in standard materials NIST956D, NIST909C, and GBW09152, had caused a maximum deviation of less than 0.9%, within the stated uncertainty range of standard materials. The RELA international inter-laboratory comparisons of potassium in serum in 2018 conducted by our laboratory also yielded a satisfactory result.

high-accuracy_determination_of_potassium_and_selenium_in_human_serum_by_two-step_isotope_dilution_ic

1. Introduction<!>2.1. Materials and Reagents<!>2.2. Instrument Parameters<!>3.1. Mass Spectrometry Procedures<!>3.2. Preparation of the Solution<!>4.1. The Selection of Methane Reagent Gas Flow<!>4.2. The Impact of the Concentration of Nitric Acid in the Solution on the Value of 80Se/78Se<!>4.3. Interference Verification and the Calculation of Correction Coefficient<!>4.4. The Process Blanks (LoB) and Detection Limits (LoD)<!>4.5. The Precision of the Method<!>4.6. The Trueness of the Method<!>4.7. Result of RELA 2018<!>4.8. Evaluation of Uncertainty in Measurements<!>5. Conclusions

<p>Isotope Dilution ICPMS, with an edge over many conventional methods, has been recognized by CCQM as one of the top five methods with absolute measurement properties. It is also the only authoritative method to measure the values of trace elements and the ultra-trace elements [1]. Isotope Dilution ICPMS is achieved through the practice of balance weighing and the mass spectrometry measurement of isotope abundance, converting the chemical composition analysis into mass spectrometry measurement. It has absolute measurement properties, and the values obtained can be traced directly to the international basic unit of the mass, namely, mole. In conventional detection methods, factors such as the loss of tested elements during the pretreatment process of the sample, the matrix effects, and the instrument noises may affect the accuracy of the tests negatively, which can be eliminated to a great extent by the isotope dilution method [2–5]. In Isotope Dilution ICPMS, the isotope abundance ratio rather than the concentration of the sample is measured. When the isotope spike is added to the tested sample to reach a sufficient balance, the isotope abundance ratio, without the contamination by foreign isotope, will become a certain value [6–10]. In the subsequent separation and concentration processes, the change of the abundance ratio, even in case of a loss of the sample, will not be affected. Consequently, the operation procedures can be simplified and the errors caused by the preprocessing of the sample can be eliminated as well. Originated from the British Government Chemist's Laboratory (LGC) [11], the two-step Isotope Dilution ICPMS has the following advantages over traditional Isotope Dilution ICPMS methods: there is no need to calibrate the concentration of isotope spike in advance for the determination. The concentration of isotope spike can only be determined by using the expensive high-resolution sector field ICPMS (SF-ICPMS). Also, all that is needed is to get the value of a given isotope abundance ratio. The chemical components in the serum can be determined, with the requirements for the mass spectrometer consequently lowered [12–15].</p><p>In the area of medical laboratory traceability, Isotope Dilution ICPMS has also been universally acknowledged as reference method of the best accuracy for measurement. The Joint Committee for Traceability in Laboratory Medicine (JCTLM) is responsible for evaluating candidate reference materials, reference measurement procedures, and reference laboratories and will list the evaluated reference materials, reference measurement procedures, and reference measurement laboratories on the website of its secretariat. Until recently, there is no reference method for measuring serum selenium by using the Isotope Dilution ICPMS in the list published by JCTLM, with only the existence of measuring potassium in serum by Isotope Dilution ICPMS [16]. The concentrations of potassium and selenium in people's body, which are related to many diseases, are important indicators of their health condition. Potassium is the major cation to maintain the physiological activities of cells, and it is also an important electrolyte for intracellular fluid [17, 18]. The potassium ions in people's body can help maintain a dynamic balance in the exchange process between the cells and body fluids. Therefore, the determination of the concentration of potassium ions in the extracellular fluid can, to a certain extent, reflect the concentration of potassium in the cells indirectly, which helps the diagnosis of electrolytes balance or the acid-base dysequilibrium [19–21]. Selenium can help enhance immunity, and it also has the function of anti-age and anti-cancer. Deficiency or excess of selenium can cause such diseases as autoimmune thyroiditis and type 2 diabetes [22–26].</p><p>Two-step Isotope Dilution ICPMS has been established in our laboratory to achieve high-accuracy determination of potassium and selenium in human serum. The reliability of this method has been verified by its employment to the analysis of standard materials and through the RELA comparison. Up to now, there are few reports about the reference method of selenium and potassium in serum by two-step isotope dilution mass spectrometry. This method can be recommended as a traceable reference method for calibrating detection systems or assigning values to certified reference materials.</p><!><p>The laboratory environment in this research was a class 100,000 cleanroom. The water used was provided by Water Purification System: Milli-Q Advantage (Millipore, USA). The nitric acid (Ultrapure-BVIII) and hydrogen peroxide (30% H2O2, Ultrapure-BVIII) were produced by Beijing Institute of Chemical Reagents, China. The microanalytical balance used for sample weighing in this experiment was Mettler Toledo XS205 (Switzerland); the microwave digestion instrument for sample preparation was MARS (Pynn, USA), and the sample analysis was conducted in ICP mass spectrometer: Elan DRC-e (PerkinElmer, USA). The standard materials used for method traceability were selenium (Se) standard solution SRM3149 (NIST USA), potassium (K) standard solution SRM3141a (NIST, USA); isotope spikes used in the method were 41K (Assay: 95%, ISOFLEX, USA) and 78Se (Assay: 98.8%, Oak Ridge, USA). The accuracy verification was achieved by selenium in human serum SRM909c (NIST, USA), electrolytes in frozen human serum SRM 956d (NIST, USA), and inorganic components in frozen human serum GBW09152 (National Institute of Metrology, China).</p><!><p>The instrument parameters for this research are listed in Table 1.</p><!><p>The concentrations of potassium and selenium in the serum sample were calculated according to concentration formula (1) in the two-step Isotope Dilution ICPMS [27].(1)Cx=Cz·mZcmYc·mYmX·RY−κ·RBκ·RB−RZ·RZ−RBcRBc−RY −CB ,where Cz is the concentration of the standard solution. In the mixture of the serum sample and the enriched isotope, mY is the mass of the enriched isotope, while mX is the mass of the serum sample. In the mixture of the standard solution and the enriched isotope, mYc is the mass of the enriched isotope, while mzc is the mass of the standard solution. Rz is the isotope abundance ratio of 39 K/41K or 80Se/78Se in the standard solution. RY is the abundance ratio of 39 K/41K or 80Se/78Se in the enriched isotope. RB is the abundance ratio of 39 K/41K or 80Se/78Se in the mixture of the serum sample and the enriched isotope. RBc is the abundance ratio of 39 K/41K or80Se/78Se in the mixture of the standard solution and the enriched isotope. k is the correction coefficient of the reference sample and the serum sample. CB is the blank of the measurement process.</p><p>In order to obtain the best accuracy in the isotope ratio measurement, an effort should be made to make RB ≈ RBc≈1 [28]. In the experiment, it had been found that the fluctuation of Rz and RY barely had effects on the final results, whereas the fluctuation of RB and RBc did affect the final results. Therefore, RB and RBc had to be measured alternately, and at the same time, the error caused by the measurement drift of the instrument could be eliminated. Each mixture of the serum sample containing selenium and the enriched isotope had been measured alternately 3 times, and each mixture of the serum sample containing potassium and the enriched isotope had been measured alternately 6 times.</p><!><p>The weighing method was used to perform a secondary dilution for NIST SRM3141a at a ratio of approximately 1 : 400 and a tertiary dilution for NIST SRM3149 at a ratio of approximately 1 : 8000. The final dilution must be made on the exact day of the experiment.</p><p>Enriched isotopes 41K as KCL and 78Se selenium powder were separately dissolved by using BVIII grade nitric acid and then diluted with ultrapure water to an appropriate concentration. A secondary dilution was carried out on the exact day of the experiment.</p><p>The weighing method was used to make the mixed solution of SRM3141a and 41K spike solution and the mixed solution of serum sample and 41K spike solution. An effort was made to guarantee the isotope ratio of the two mixed solutions (39 K/41K) close to 1, and at the same time, the values of the cps signal intensity of 39K and 41K in the two mixed solutions were close.</p><p>1 g~1.3 g of serum was accurately weighed and put in a microwave digestion tank, with an appropriate amount of 78Se spike solution added. We made sure that mY of the added 78Se spike solution was accurately weighed to make the value of 80Se/78Se close to 1 so as to achieve a better accuracy. After 10 ml of HNO3+1mLH2O2 was added, the digestion tank was digested in a microwave digestion instrument by way of gradient heating with the maximum temperature of the digestion being 180°C for a time span of 40 minutes. After the microwave digestion stood still overnight after the digestion, it was moved to ISOLSB PP beaker at 150°C and steamed until the solution volume was reduced to about 2 ml, and then H2O2 (10%) was added to drive the acid repeatedly 6 times until the sample solution was colorless. Then, a value of 30 ml was set and its isotope ratio was measured on the exact day of the experiment. A mixed solution of isotope spikes of SRM3149 and 78Se was made to make sure the value of 80Se/78Se was close to 1, and at the same time, the values of the cps signal intensity of 80Se and 78Se in the two mixed solutions were close.</p><p>On the exact day of the experiment, a potassium solution and a selenium solution with appropriate concentrations were separately made to make the signal intensity of 39K (80Se) in the solutions consistent with that of the mixed solution mentioned previously.</p><p>On the exact day of the experiment, a 41K solution and a 78Se solution with appropriate concentrations were prepared separately to make the signal intensity of 41K (78Se) in the solutions consistent with that of the mixed solution mentioned previously.</p><!><p>In this experiment, the dynamic reaction cell of ICPMS was used to eliminate the interference. When selenium was measured, there was a serious interference of Ar2 due to the fact that the interference of Ar2 to 80Se is more serious than its interference to 78Se [29]. Therefore, an appropriate flow rate of the reactant gas CH4 and argon should be selected so as to obtain the most reliable results for the measurement of 80Se. In this research, the most appropriate flow rate of the reactant gas CH4, namely, 0.8 mL/min of methane and 0.45 mL/min of argon, was selected by using the dynamic reaction cell of ICPMS. When potassium was measured, there was a serious interference of ArH, CaH, and MgO, especially to 41K. Therefore, an appropriate flow rate of the reactant gas CH4 and argon should be selected so as to obtain the most reliable results for the measurement of 41K. The most appropriate flow rate of the reactant gas CH4, namely, 1.5 mL/min of methane and 0.8 mL/min of argon, was selected by using the dynamic reaction cell of ICPMS.</p><!><p>Solutions with different concentrations of nitric acid and the same amount of Se (5.0 μg/L) were made and then a measurement of the signal intensities (cps) of 78Se and 80Se was conducted under the same experimental conditions. The results showed that the cps response signal of 78Se and 80Se decreased with the increase of acidity (Figure 1), but the value of 80Se/78Se hardly changed (Figure 2). Therefore, it can be concluded that the difference in acidity can be ignored when preparing the reference solution and the sample solution, but the analysis sensitivity is higher when the acidity is relatively low.</p><!><p>The signal intensities of 39K, 41 K, 80Se, and 78Se in ultrapure water have been tested under the established experimental conditions. The results show that the ratio of the signal response intensity of ultrapure water to the signal intensity of the corresponding isotope in the serum sample is less than 0.1%. Therefore, it is safely believed that the interference of Ar2 or ArH has been eliminated by the way of dynamic reaction cell, or to be more specific, their interference on ultrapure water can be ignored. Then, the ratio of the isotope abundance ratio (RZ) of 39 K/41K (80Se/78Se) in the standard solution to the isotope abundance ratio (RS) of 39 K/41K (80Se/78Se) in the serum sample was tested. In the case that the ratio RZ to RS ranges from (100.0–0.1) % to (100.0 + 0.1) %, it can be concluded that the matrix interference of serum samples can be ignored, for the measurement precision of the instrument is of the same error level. In this experiment, 39 K/41K in the potassium standard solution is consistent with 39 K/41K in the serum sample, while 80Se/78Se in the selenium standard solution is greater than 80Se/78Se in the serum sample. After the microwave digestion of the serum selenium sample, the interference of the serum matrix on 80Se/78Se ratio has been reduced; nevertheless, it has been proved that the ratios of 80Se/78Se at different concentrations levels are consistent. According to calculation formula (2) of the correction coefficient, the correction coefficient of potassium was 1 and that of selenium was about 1.027, and we made sure that the correction coefficient of selenium was determined simultaneously on the exact day of the experiment.(2)k=RZRS.</p><!><p>While preparing the mixed solution of serum sample and enriched isotope, we added an appropriate amount of 78Se (or 41K) enriched isotope into the blank sample tube as the process blank, where the value of 80Se/78Se (or 39 K/41K) was about 2. The process blank, together with the sample, went through the entire preprocessing procedure. In accordance with the requirements of International Union of Pure and Applied Chemistry (IUPAC), the sample blank should be tested 20 times repeatedly according to the formula LoD = LoB + k∗Sb1 (k = 2, and Sb1 is the standard deviation of the blank sample). The results showed that the process blanks of potassium and selenium in serum were −0.3 mg/kg~0.3 mg/kg and 1.0 μg/kg~1.50 μg/kg, respectively, and when the confidence interval of 95% was determined, the detection limits of potassium and selenium in serum were 0.8 mg/kg and 2.7 μg/kg, respectively.</p><!><p>Assigned value experiments were conducted on the candidate standard materials of potassium and the candidate standard materials of selenium in human serum in our laboratory. The experiments were carried out in two days, with 6 bottles per day. The results are shown in Table 2. The measurement precision for potassium and selenium in serum was lower than 0.2% and 0.7%, respectively.</p><!><p>The measurement was employed to determine potassium and selenium in standard materials NIST956D, NIST909C, and GBW09152. The results are shown in Table 3, where the maximum deviation was less than 0.9%, within the stated uncertainty range of standard materials.</p><!><p>Our laboratory had joined the RELA international inter-laboratory comparisons of potassium in serum in 2018. In Figure 3 and Table 4, lab code 115 showed the results obtained by our laboratory. The result is in the center of the acceptable data range. We evaluated the combined standard uncertainty of the results of RELA. The uncertainty caused by the sample weighing, the balance itself, the single measurement, and the uncertainty caused by the vial to vial difference, the traceability standard material itself, and by the process blanks are all taken into consideration in the process of evaluation. The combined standard uncertainty for potassium in sample A and sample B was 0.04 mmol/L and 0.03 mmol/L, respectively (k = 2).</p><!><p>In this research, the uncertainty caused by factors such as the experimental reagents, the samples, the laboratory environments, the solution preparation, the instrument measurement, and the data processing has been evaluated as the source of uncertainty in the measurement process. It can be concluded that by evaluating the uncertainty of each parameter in formula (1), the uncertainty caused by each factor in the measurement process can be fully included. Each parameter in formula (1) is an independent parameter, and the uncertainty  uc(y) related to measurement is calculated as follows:(3)ucy=∑i=1N∂f∂xi2u2xi.</p><p>The formula of the sensitivity coefficient ((∂f/∂xi)) of each parameter in formula (1) is as follows:(4)∂Cx∂Cz=mZcmYc·mYmX·RY−K·RBK·RB−RZ·RZ−RBcRBc−RY,∂Cx∂mY=Cz·mZcmYc·1mX·RY−K·RBK·RB−RZ·RZ−RBcRBc−RY,∂Cx∂mYc=Cz·−mZcmYc2·mYmX·RY−K·RBK·RB−RZ·RZ−RBcRBc−RY,∂Cx∂mx=Cz·mZcmYc·−mYmX2·RY−K·RBK·RB−RZ·RZ−RBcRBc−RY,∂Cx∂mZc=Cz·1mYc·mYmX·RY−K·RBK·RB−RZ·RZ−RBcRBc−RY,∂Cx∂Rz=Cz·mZcmYc·mYmX·RY−K·RBK·RB−RZ2RZ−RBcRBc−RY+RY−K·RBK·RB−RZ·1RBc−RY,∂Cx∂RY=Cz·mZcmYc·mYmX·RZ−RBcK·RB−RZ·RBc−RY+RY−K·RBRBc−RY2,∂Cx∂RB=Cz·mZcmYc·mYmX·RZ−RBcRBc−RY·−KK·RB−RZ−KRY−K·RBK·RB−RZ2,∂Cx∂RBc=Cz·mZcmYc·mYmX·RY−K·RBK·RB−RZ·−RBc−RY−RZ−RBcRBc−RY2,∂Cx∂K=Cz·mZcmYc·mYmX·RZ−RBcRBc−RY·−RBK·RB−RZ−RBRY−K·RBK·RB−RZ2,∂Cx∂CB=−1.</p><p>Therefore,(5)ucy=∂Cx∂Cz2·ucCz2+∂Cx∂mY2·ucmY2+…+∂Cx∂CB2·ucCB2.</p><p>This research has evaluated the uncertainty of measurement for the data in Table 2. Table 5 shows the source of uncertainty for measuring potassium in serum, and Table 6 shows the source of uncertainty for measuring the parameters of selenium in serum. The uncertainty of the measurement of type A is the experimental standard deviation of the 6 repeated instrument measurements, taking the worst result in the experiment as the evaluation data. The uncertainty of the measurement of type B is the uncertainty of solution preparation, which is synthesized from the uncertainty resulting from the electronic balance calibration and the uncertainty resulting from weighing.</p><!><p>The two-step Isotope Dilution ICPMS established in our laboratory measured accurately potassium and selenium in human serum. The results show that this method, which is easy to operate, has a high accuracy and good repeatability. Since the concentration of potassium in serum is of ppm level, the sample was diluted more than 100 times, leaving the matrix effect on serum almost neglected, and thus there was no need for complicated preprocessing for the sample. At the same time, the concentration of selenium in serum is of ppb level. Thus, another sample was then diluted at a lower ratio, leaving an obvious serum matrix effect. In addition, this sample was microwave digested for the elimination of the matrix interference. There is no need to measure the exact concentration of the enriched isotope in the two-step isotope dilution mass spectrometry process, lowering the requirements on the mass spectrometry equipment [30, 31]. This method has been proved to be a reliable reference method for assigning values of standard materials. The values obtained can be traced directly to international system of units (SI).</p>

PubMed Open Access

Comparison of rhenium–porphyrin dyads for CO₂ photoreduction: photocatalytic studies and charge separation dynamics studied by time-resolved IR spectroscopy

We report a study of the photocatalytic reduction of CO 2 to CO by zinc porphyrins covalently linked to [Re I (2,2 0 -bipyridine)(CO) 3 L] +/0 moieties with visible light of wavelength >520 nm. Dyad 1 contains an amide C 6 H 4 NHC(O) link from porphyrin to bipyridine (Bpy), Dyad 2 contains an additional methoxybenzamide within the bridge C 6 H 4 NHC(O)C 6 H 3 (OMe)NHC(O), while Dyad 3 has a saturated bridge C 6 H 4 NHC(O)CH 2 ; each dyad is studied with either L ¼ Br or 3-picoline. The syntheses, spectroscopic characterisation and cyclic voltammetry of Dyad 3 Br and [Dyad 3 pic]OTf are described. The photocatalytic performance of [Dyad 3 pic]OTf in DMF/triethanolamine (5 : 1) is approximately an order of magnitude better than [Dyad 1 pic]PF 6 or [Dyad 2 pic]OTf in turnover frequency and turnover number, reaching a turnover number of 360. The performance of the dyads with Re-Br units is very similar to that of the dyads with [Re-pic] + units in spite of the adverse free energy of electron transfer. The dyads undergo reactions during photocatalysis: hydrogenation of the porphyrin to form chlorin and isobacteriochlorin units is detected by visible absorption spectroscopy, while IR spectroscopy reveals replacement of the axial ligand by a triethanolaminato group and insertion of CO 2 into the latter to form a carbonate. Time-resolved IR spectra of [Dyad 2 pic]OTf and [Dyad 3 pic]OTf (560 nm excitation in CH 2 Cl 2 ) demonstrated electron transfer from porphyrin to Re(Bpy) units resulting in a shift of n(CO) bands to low wavenumbers. The rise time of the charge-separated species for [Dyad 3 pic]OTf is longest at 8 (AE1) ps and its lifetime is also the longest at 320 (AE15) ps. The TRIR spectra of Dyad 1 Br and Dyad 2 Br are quite different showing a mixture of 3 MLCT, IL and charge-separated excited states. In the case of Dyad 3 Br, the charge-separated state is absent altogether. The TRIR spectra emphasize the very different excited states of the bromide complexes and the picoline complexes. Thus, the similarity of the photocatalytic data for bromide and picoline dyads suggests that they share common intermediates. Most likely, these involve hydrogenation of the porphyrin and substitution of the axial ligand at rhenium.

comparison_of_rhenium–porphyrin_dyads_for_co₂_photoreduction:_photocatalytic_studies_and_

Introduction<!>General procedures<!>Synthesis<!>Synthetic methodology<!>Emission spectroscopy<!>Photocatalysis<!>Photocatalytic intermediates<!>Picosecond time-resolved infrared spectroscopybromide complexes<!>Energetics of electron transfer<!>Emission spectroscopy<!>Time resolved infrared spectroscopy<!>Mechanism<!>Conclusions

<p>Much of the world's energy need is satised by the combustion of fossil fuels. The processes associated with generating energy in this way release gigatons of CO 2 into the atmosphere every year, contributing to climate change. 1 In addition to environmental unsustainability, the fossil fuels are essentially nite as they require geological timescales to form. The sun provides a clean source of energy that can satisfy our energy demands now and in the future. 2 It is critical therefore, that we develop systems that can harvest visible light and store the energy as chemical fuel: systems that perform articial photosynthesis.</p><p>Supramolecular assemblies containing components capable of light harvesting and catalysis can in principle perform arti-cial photosynthesis. There are several examples of this type of system for water oxidation, 3,4 proton reduction, [5][6][7][8][9] and CO 2 reduction. [10][11][12] For supramolecular assemblies to be active for photocatalytic redox reactions, they must be designed such that photoinduced electron transfer is favourable and such that charge separation lifetimes are sufficiently long for the catalytic reaction to occur prior to recombination.</p><p>Photocatalytic CO 2 reduction to CO is an attractive choice because CO 2 is consumed and CO can subsequently be converted into energy-dense hydrocarbon fuels. [13][14][15][16] CO is also an industrial feedstock and a fuel in its own right. 17 Diimine complexes of rhenium have received much attention since the discovery, reported in 1983, that they are active and selective photo-and electro-catalysts for CO 2 reduction to CO. 18 In the context of solar fuels, the rhenium complexes are limited because they cannot utilize much of the solar spectrum and turnover numbers of CO (TON CO ) are low due to catalyst instability. 19 Introduction of a sensitizer molecule can improve visible light absorption. The use of lower energy radiation and transferring the role of light absorption to another molecular unit will remove pathways of photo-degradation for the rhenium complex and increase stability. Indeed high TON CO have been reported for dyads consisting of rhenium catalysts covalently linked to ruthenium bipyridyl units. 10,[20][21][22][23][24][25][26][27][28][29][30][31] Sensitizing dyes have also been used in association with Re catalysts supported on TiO 2 . 32 Zinc porphyrins are good candidates for sensitization for several reasons. 33 They show intense absorption in the visible spectrum, in particular the Q bands centred around 560 nm. 11 The excited state redox potential of zinc porphyrin can be tuned to be negative with respect to the ground state of rhenium diimine complexes. 34 The porphyrin motif is closely related to chlorophylls 35 that are utilized in natural photosynthesis for light harvesting and charge separation. 36 The visible light absorption and photoinduced electron-transfer ability of zinc porphyrins has led to high efficiencies in dye-sensitized solar cells. 37 We and others recently demonstrated that zinc porphyrins can sensitize rhenium diimine complexes for CO 2 reduction to CO with long-wavelength visible light. 11,38,39 Rhenium bipyridine tricarbonyl complexes have been used extensively for photocatalytic and electrocatalytic CO 2 reduction. 18,[40][41][42][43][44][45][46][47][48][49][50] There have been important recent developments in understanding the mechanism of such reactions. Kubiak has tracked reduced intermediates and their reactivity toward CO 2 . 45,46,51,52 Ishitani has shown that the usual sacricial reducing agent, triethanolamine (TEOA), coordinates to rhenium by deprotonation to form a rhenium alkoxide of the type ReOCH 2 CH 2 N(CH 2 CH 2 OH) 2 which can insert CO 2 to form a rhenium carbonate derivative. 53 Inoue et al. have used mass spectrometry to study reduction of ReCl(4,4 0 dimethyl-2,2 0bipyridine)(CO) 3 with triethylamine. 54 They demonstrate that CO 2 displaces a solvent molecule in the one-electron reduced complex to form a Re-CO 2 radical which is then protonated to form a Re-COOH radical cation. Thus there is good evidence of direct CO 2 coordination in the absence of TEOA and a complete cycle has been postulated for the electrochemical reaction. 52 For the photochemical reaction with TEOA, the new evidence indicates CO 2 insertion into the alkoxide complex, but the subsequent steps remain undened.</p><p>Closely related zinc porphyrins bound to rhenium carbonyls have been investigated for photo-induced charge separation. 55,56 Iron porphyrins have also been used successfully as electrocatalysts for CO 2 reduction. [57][58][59] There are several photophysical investigations into porphyrins linked to metal carbonyl complexes, 38,[60][61][62][63][64][65][66] but investigations connecting photophysical data and photocatalytic activity across a range of catalyst structures are scarce. 10,20 Pump-probe time resolved infrared spectroscopy (TRIR) is an invaluable technique for measuring excited state dynamics in this kind of assembly. 34,[67][68][69][70][71][72][73][74][75][76][77][78][79][80] Metal carbonyl n(CO) stretches can be observed with high intensity in a region of the infrared where few other vibrational bands are present. Crucially, they are very sensitive to the electron density on the metal centre and can be used to monitor charge transfer.</p><p>In our previous investigations of long-wavelength (l > 520 nm) photocatalytic CO 2 reduction with Re complexes covalently linked to zinc porphyrins, we investigated [Dyad 1 pic]PF 6 (Fig. 1) with a C 6 H 4 NHCO bridge. 11 To increase catalytic activity we sought to reduce the rate of charge recombination by increasing the separation between donor and acceptor 74,75 by inclusion of a methoxybenzamide molecular spacer ([Dyad 2 pic]OTf), and this dyad indeed displayed higher catalytic activity. 11 We now report the synthesis and catalytic activity of a new dyad with a C 6 H 4 NHCOCH 2 saturated molecular spacer [Dyad 3 pic]OTf (Fig. 1). We also compare the catalytic performance of these three cationic dyads to those of the corresponding neutral bromide complexes Dyad 1 Br, Dyad 2 Br and Dyad 3 Br. To our surprise the catalytic performance of each of the bromide complexes is very similar to that of the corresponding cationic dyads. This is intriguing as the bromide dyads do not undergo photoinduced reaction with intermolecular electron donors and their reduction potentials are signicantly more negative than those of the cationic complexes. 81,82 In our previous investigations of [Dyad 1 pic]OTf we showed by TRIR spectroscopy that charge separation occurs within a few ps and the lifetime of the charge-separated state is of the order of tens of ps. We now report on the TRIR spectroscopy of [Dyad 2 pic]OTf, [Dyad 3 pic]OTF and that of all three bromide complexes. We also show by TRIR spectroscopy that the excited state behaviour of the neutral bromide complexes is very different from that of the cationic picoline complexes. We propose mechanisms that can reconcile the different excited state and electrochemical behaviour with the similar photocatalysis.</p><!><p>Chemicals were obtained from the following suppliers: diisopropylamine, 2.5 M n-butyl lithium in hexanes (Acros); EDTA, AgOTf, 4,4 0 -dimethyl-2,2 0 -bipyridine, sodium sulde, celite 512 medium, Et 3 N, 2-chloro-methylpyridinium iodide, triethanolamine, anhydrous DMF, methyl chloroformate, copper(II) acetate (Aldrich); 3-picoline (BDH Chemicals); CO 2 -CP-grade with 5% CH 4 (or 1% CH 4 ) (BOC); Na 2 SO 4 , Na 2 CO 3 , NaHCO 3 , NaOH, HCl, KOH, ammonium hydroxide (Fisher); Zn(OAc) 2 $H 2 O, CH 3 CO 2 Na (Fisons).</p><p>Solvents for general use were obtained from Fisher. Solvents were dried by reuxing over sodium wire (C 6 H 6 , THF, toluene) or over CaH 2 (CH 2 Cl 2 ). DMF was dried using a Pure Solv 400-3-MD (Innovative Technology). For TRIR experiments, CH 2 Cl 2 (99.9%, Merck) was distilled under an inert atmosphere of Ar from calcium hydride and anhydrous THF ($99.9%, inhibitorfree, Sigma Aldrich) was used as supplied and stored in a glove box.</p><p>CD 2 Cl 2 , CD 3 OD, DMSO-d 6 and CDCl 3 were used as obtained (Aldrich) and THF-d 8 was dried over potassium. Diisopropylamine was distilled from sodium hydroxide. Methyl chloroformate was distilled prior to use. n-BuLi was titrated against nbenzylbenzamide prior to use. Routine separation of porphyrins by ash chromatography was performed on a CombiFlash Rf system using 24 g RediSep Rf silica columns (Teledyne Isco), and dry-loading the samples on silica (Fluka).</p><p>NMR spectroscopy. NMR spectra were run on a Bruker AV500 ( 1 H at 500 MHz) spectrometer or Bruker ECS400 (400 MHz). 1 83 IR and UV/vis absorption and emission. IR spectra were recorded on a Mattson RS FTIR instrument, averaging 64 scans at resolution 2 cm À1 . ATR-IR spectra were an average of 32 scans. UV/visible absorption spectra were measured using an Agilent 8453 spectrometer. Steady state emission spectra were measured using a Hitachi F-4500 uorimeter. The uorescence was taken against a ZnTPP reference for the bromide complexes and against the individual dyad ligand ZnTPP-link-Bpy for the picoline complexes. Time-resolved emission was measured with an Edinburgh Instruments FLS980 equipped with a 560 nm pulsed LED (EPLED 560, pulsewidth 1.5 ns) and a red PMT detector. All samples were either degassed by three freezepump-thaw cycles or de-aerated by purging the sample with Ar. Correction was applied for instrument response. All absorption and emission measurements were made in 10 Â 10 mm quartz cuvettes.</p><p>Mass spectrometry. ESI mass spectra were recorded on a Bruker micrOTOF instrument with a sample ow rate of 0.2 mL min À1 , nebuliser gas pressure of 1.5 bar, dry gas ow of 8 L min À1 and a dry gas temperature of 180 C. EI mass spectra were run on a Waters GCT premier with a source temperature of 180 C, electron energy of 70 eV and a trap current of 200 mA. Some compounds and the reaction mixture ESI mass spectra were run on a Bruker Esquire 6000 via direct infusion using a syringe pump at 240 mL min À1 . Nebuliser gas and dry gas ows and temperatures were optimised for each individual sample along with the spray voltage. m/z values are quoted for 64 Zn, 185 Re and 79 Br.</p><p>Electrochemistry. Cyclic voltammetry was performed in CH 2 Cl 2 with 0.1 M [Bu 4 N][PF 6 ] (TBAP) electrolyte. The setup comprised reference electrode (Ag/AgCl, 3 M NaCl), working electrode (platinum disc) and counter electrode (platinum wire). Ferrocene was used as internal standard. All scans were made at 100 mV s À1 . Cyclic voltammetric experiments used a BASi Epsilon potentiostat with C3 cell stand.</p><p>X-ray diffraction. X-ray diffraction data for [Dyad 1 pic]PF 6 and 5-[4-[(2-methoxy-4-nitro-phenylcarbonyl)-amino]phenyl]-10,15,20-triphenyl porphyrin were collected at 110 K on an Agilent SuperNova diffractometer with MoKa radiation (l ¼ 0.71073 Å). Data collection, unit cell determination and frame integration were carried out with "CrysalisPro". Absorption corrections were applied using crystal face-indexing and the ABSPACK absorption correction soware within CrysalisPro. Structures were solved and rened using Olex2 implementing SHELX algorithms. [Dyad 1 pic]PF 6 was solved using SUPER-FLIP 84 whereas 5-[4-[(2-methoxy-4-nitro-phenylcarbonyl)-amino] phenyl]-10,15,20-triphenyl porphyrin was solved using direct methods within the SHELXS algorithm. Structures were rened by full-matrix least squares using SHELXL-97. All non-hydrogen atoms were rened anisotropically. Carbon-bound hydrogen atoms were placed at calculated positions and rened using a "riding model".</p><p>For [Dyad 1 pic]PF 6 , one of the phenyl groups on the porphyrin ring was disordered and modelled in two positions with rened occupancies of 0.817 : 0.183 (12). The ADP of equivalent carbons in the disordered phenyl were constrained to be equal, e.g. C51 & C51A. The hexauorophosphate was disordered over two sites. For one of these, the phosphorus was centred on a special position and for the other, the occupancy was 50% with a dichloromethane of crystallisation occupying the site at other times.</p><p>In addition to the ordered dichloromethanes of crystallisation, the crystal also contained some disordered solvent, believed to be a mix of hexane and dichloromethane for which a suitable discrete model could not be obtained. This was accounted for using a solvent mask; this space had a volume of 213 Å3 and predicted to contain ca. 17 electrons. The large residual density peaks are believed to provide evidence for twinning but a suitable method for modelling this was not found.</p><p>For 5-[4-[(2-methoxy-4-nitro-phenylcarbonyl)-amino]phenyl]-10,15,20-triphenyl porphyrin, the NH hydrogen was located by difference map. The crystal also contained dichloromethanes of crystallisation. One was partially occupied and was modelled with an occupancy of 0.1875; the carbon of this CH 2 Cl 2 was restrained to be approximately isotropic. The other was fully occupied but disordered and modelled with the carbon in two different positions with relative occupancies of 0.814 : 0.186 (12). Crystallographic parameters are listed in the ESI. † Ultrafast infrared experiments. Picosecond time-resolved infrared (TRIR) spectra were obtained using purpose-built equipment based on a pump-probe approach. Details of the equipment and methods used for the TRIR studies have been described previously, 85,86 a brief description of which is given here. The pump beam (560 nm, ca. 150 fs) and tunable probe beam (180 cm À1 spectral band width, ca. 150 fs) were generated from a commercial Ti:sapphire oscillator (MaiTai)/regenerative amplier system (Spitre Pro, Spectra Physics). The mid-IR probe was detected using a 128-element HgCdTe array detector (Infrared Associates) typically with a resolution of ca. 4 cm À1 . All the solutions for analysis were prepared under an inert atmosphere of Ar, degassed by three freeze-pump-thaw cycles and put under Ar. [Dyad 2 pic]OTf and [Dyad 3 pic]OTf were run in CH 2 Cl 2 at 1.5 mM and 1.0 mM respectively, with a path length of 0.5 mm. Dyad 1 Br, Dyad 2 Br and Dyad 3 Br were run in THF at 1 mM with a path length of 0.25 mm. A Harrick solution cell with CaF 2 windows was used and 20 mL of solution was continuously circulated during the measurements.</p><p>Photocatalysis. Photocatalysis was performed in a custommade cell 67 comprised of a 10 Â 10 mm quartz cuvette with a headspace of a minimum volume of 10 mL. Above the headspace was a ground glass joint, which was sealed with a size 21 septum. Samples were taken through this septum for GC analysis. The headspace had a sidearm, isolated by a Young's tap, joining it to a gas phase IR cell with CaF 2 windows. The IR cell was connected to a vacuum joint via a second Young's tap. The IR cell was put under vacuum. At the end of a catalytic run the headspace was opened to the IR cell and the gas produced from the reaction would be drawn through and could be monitored by IR spectroscopy.</p><p>The concentration of catalytic solution was typically 0.05 mM, making the absorbance of the porphyrin Q band at 560 nm, Q(1, 0), ca. 1 by UV/vis spectroscopy. A 10 mL stock solution of 0.25 mM catalyst in DMF would typically be made. These stock solutions allowed the catalysts to be weighed out in amounts greater than 1 mg. They were stored in a freezer at À25 C and could be used up to a month later without noticeable degradation in their catalytic performance, UV/vis spectrum or mass spectrometric analysis. The 0.05 mM catalytic solution was made from the stock by diluting 2 mL into 10 mL. To make a 10 mL solution in DMF : TEOA 5 : 1, 1.87 g TEOA was weighed into a 10 mL volumetric ask, approximately 2 mL of DMF was added so the catalytic stock was not being added to neat TEOA. Then 2 mL of stock was added, followed by DMF up to the 10 mL mark. The catalytic solutions were protected from light as much as possible and stored in the freezer. A sample (3 mL) of catalytic solution was added to the photoreaction cuvette and was bubbled with CO 2 /CH 4 95/5 for 10 min, protected from light throughout this time.</p><p>Irradiation of all samples was performed with an ILC 302 Xe arc lamp. Light from the lamp was directed through a water lter (10 cm) and a 660 nm short pass lter (<660 nm, Knight Optical) to remove heat, such that any sample directly in the beam was at a temperature of 33 C. A l > 520 nm optical lter was added (Schott).</p><p>The amount of CO produced was determined by GC analysis using a UnicamProGC+ (ThermoONIX) with a thermal conductivity detector. Air, CO, CH 4 and CO 2 were separated on a Restek ShinCarbonST 100/120 micropacked column (2 m, 1/16 00 OD, 1.0 mm ID) tted with "pigtails" of Restek intermediate-polarity deactivated guard column on either end (fused silica, 0.53 mm ID, 0.69 AE 0.05 mm OD). The carrier gas was ultra high purity He (N6.0, BOC gases) passed through a GC triple lter (Focus Technical) to remove trace impurities prior to the column. The GC method began with 1 min at 40 C followed by a 5 C min À1 gradient up to 120 C (16 min). Injections (200 mL) were made manually with a Hamilton gastight locking syringe (500 mL) at 220 C with a 30 mL min À1 split ow. The carrier gas was kept at constant pressure (165 kPa). The detector block and transfer temperatures were 200 and 190 C respectively, at a constant voltage of 10 V with makeup and reference ows of 29 and 30 mL min À1 respectively. The amount of CO was determined using a calibration plot. Known volumes of CO were mixed with a mimic experimental solution (3 mL DMF : TEOA 5 : 1 (v/v)), headspace and solution were purged with CO 2 : CH 4 (99 : 1 or 95 : 5) and sampled to GC. Quantication was by comparison of integrations of the CO peak against the CH 4 internal standard. Corrections were made for temperature and the change in headspace pressure at each injection.</p><!><p>Re(CO) 5 Br, 87 4 0 -Methyl-2,2 0 -bipyridine-4-acetic acid (AABpy). Procedure 1: a modication of that by Ciana. 89 A 250 cm 3 round-bottomed ask was ame dried and ushed with Ar. THF (3 cm 3 ) and diisopropylamine (2.1 cm 3 ) were added and the mixture cooled to À78 C. 2.5 M butyl lithium in hexanes (6 cm 3 ) was added via syringe and the mixture was stirred for 0.75 h. A solution of 4,4 0dimethyl bipyridine (3 g) in THF (72 cm 3 ) was added, the solution turned black and was stirred for 2 h at À78 C. Dry CO 2 (g) was set bubbling through a ame dried round-bottomed ask charged with Et 2 O (30 cm 3 ) and cooled to À78 C. The black lithiated bipyridine solution was added to the Et 2 O/CO 2 mixture via cannula and a yellow precipitate soon appeared. The reaction was le under an atmosphere of CO 2 overnight and allowed to warm to RT. Et 2 O was added (30 cm 3 ) and the product extracted with 3 M NaOH (3 Â 30 cm 3 ). The alkaline layer was then acidied to pH 1 with concentrated HCl and cooling. The product was then extracted with Et 2 O (30 cm 3 ) and buffered to pH 5 with solid CH 3 CO 2 Na. A saturated aqueous solution of Cu(CH 3 CO 2 ) 2 was added causing precipitation of a blue Cu complex. The solid was ltered off with a microber lter paper and washed with water, ethanol and ether and then air-dried. The product was suspended in water (60 cm 3 ) and H 2 S bubbled through for 20 min resulting in a dark brown colour. The product was ltered through celite, concentrated to 9 cm 3 and ltered again. The solution was evaporated to dryness under reduced pressure to yield a yellow oil. Recrystallisation twice from ethanol/hexane yielded pure AABpy (569 mg, 2.682 mmol, 16%). Analysis was in agreement with the literature. 89 Procedure 2: a modication of that by Tomioka. 90 To a ame dried 100 mL round-bottomed ask was added THF (5 mL) and freshly distilled diisopropylamine. The mixture was cooled to À78 C and freshly titrated n-butyl lithium (1.1 eq.) was added. Dimethylbipyridine (1 g, 5.43 mmol) was dissolved in THF (20 mL) and added by cannula. The mixture was stirred at À78 C for 2 h and then freshly distilled methyl chloroformate (0.6 mL) in THF (2 mL) was added by syringe. The reaction was stirred at À78 C for 1 h and then at RT for 2 h. The mixture was then washed with saturated NaHCO 3 solution and extracted into ethyl acetate. The extracts were washed with brine and dried over Na 2 SO 4 . The product was puried on Si-60 eluting with 2% Et 3 N in pentane and 0-10% EtOAc. The second fraction was collected and the solvent removed (257 mg, 0.858 mmol, 22%). The methyl ester was hydrolysed to produce the free acid. A 50 mL round-bottomed ask was charged with the methyl ester (284 mg), which was dissolved in the minimum amount of methanol. KOH (131 mg) was added. The reaction was stirred at 35 C for 2 h. The solvent was removed and the solid taken up in H 2 O and titrated to pH 7 with a 10% solution of HCl. The H 2 O was removed and the product used without purication. (15 cm 3 ) was added, followed by Et 3 N dropwise. The mixture was stirred at 0 C for 5 min and then warmed to RT. Aer stirring at RT for 0.5 h TLC showed negligible quantities of starting porphyrin and so the reaction was stopped. The reaction was quenched with 10% HCl (50 cm 3 ) and the porphyrin extracted with CH 2 Cl 2 . The extract was washed with saturated NaHCO 3 followed by brine and then dried over MgSO 4 . The product was puried with column chromatography on Si-60 eluting with CH 2 Cl 2 and CH 3 OH (0% to 3%). The second fraction was collected and the solvent removed to yield the desired product (156 mg, 0.186 mmol, 94%). A 100 cm 3 round-bottomed ask was charged with CH 2 Bpy-H 2 TPP (152 mg, 181 mmol), Zn(OAc) 2 (179 mg, 815 mmol), CH 3 OH (5 cm 3 ) and CHCl 3 (25 cm 3 ). The mixture was heated to reux for 1 h and the reaction was followed by UV/vis spectroscopy. The reaction mixture was allowed to cool, pumped to dryness and redissolved in 100 cm 3 CHCl 2 and 20 cm 3 CHCl 3. This was washed with EDTA solution (2 g in 200 cm 3 10% Na 2 CO 3 solution), water (3 Â 200 cm 3 ), dried (MgSO 4 ) and the solvent removed to yield the desired compound (160 mg, 178 mmol, 98%). 1 5-{4-[Rhenium(I)tricarbonyl(bromide)-4-methyl-2,2 0 -bipyridine-4 0 -methylene carboxyamidyl]phenyl}-10,15,20-triphenylporphyrinatozinc(II) (Dyad 3 Br). A two-neck 50 cm 3 round-bottomed ask was tted with a reux condenser and gas valve. The setup was ame dried. Under Ar CH 2 Bpy-ZnTPP (200 mg, 221 mmol) was added, followed by ReBr(CO) 5 (90 mg, 221 mmol). Dry benzene was added (30 cm 3 ) by syringe. The mixture was heated to 65 C and the reaction was followed by IR spectroscopy and judged to be complete aer 22 h. The reaction mixture was ltered to leave a solid product and used without further purication (263 mg, 210 mmol, 95%). 5-{4-[Rhenium(I)tricarbonyl(3-picoline)-4-methyl-2,2 0 -bipyridine-4 0 -methylene carboxyamidyl]phenyl}-10,15,20-triphenylporphyrinatozinc(II) triuoromethanesulfonate ([Dyad 3 pic] OTf). A two-neck 50 cm 3 round-bottomed ask was tted with a gas valve and ame dried. It was taken into a glovebox and AgOTf was added (206 mg, 800 mmol). A condenser tted with a single-neck round-bottomed ask and gas valve was ame dried, then the round-bottomed ask was removed under Ar and the condenser and reaction ask were brought together. THF and 3-picoline (1.09 mL, 11.2 mmol) were added and nally Dyad 3 Br (200 mg, 160 mmol). The mixture was heated to reux for 2 h and checked for completion by IR spectroscopy. The mixture was allowed to cool, ltered to remove AgBr and dried under vacuum for 72 h. The oil was re-dissolved in THF and applied to Sephadex LH20 eluting with THF. The THF was removed and the solid washed with an ethanol/petrol 20/ 80 mixture. The solid was dried to yield [Dyad 3 pic]OTf (95 mg, 67.10 mmol, 42%). 1</p><!><p>The synthetic methods for the preparation of [Dyad 1 pic]OTf, Dyad 1 Br, [Dyad 2 pic]OTf and Dyad 2 Br have been reported previously. 11,34 The synthetic strategy for [Dyad 3 pic]OTf is shown in Fig. 2. AABpy 89 and NH 2 -H 2 TPP 88 were prepared by literature procedures. The two were coupled using 2-chloromethylpyridinium iodide in excellent yield (94%). 91 Zinc was inserted into the porphyrin and Re(CO) 3 Br was complexed to the Bpy as reported previously for [Dyad 1 pic]OTf. 34 Bromide was substituted for 3-picoline using AgOTf in THF. The product was puried using size exclusion chromatography (Sephadex LH20) eluting with THF. (Fig. S13 †). Two reversible oxidation waves were observed on scanning to anodic potentials, corresponding to the rst and second oxidation of the porphyrin. 92 In the cathodic direction a quasi-reversible reduction wave was observed, corresponding to the rst reduction of the rhenium unit. The rst oxidation of the porphyrin is at similar potential to that of [Dyad 2 pic]OTf and the Re-free porphyrin CH 2 -Bpy-ZnTPP, but the reduction of the rhenium is at a more negative potential than those of either [Dyad 1 pic]PF 6 or [Dyad 2 pic]OTf (Table 1). 11 [(2-methoxy-4-nitro-phenylcarbonyl)-amino]phenyl]-10,15,20triphenyl porphyrin (Fig. S16 †). The structure conrms that the hydrogen atom on the porphyrin amide forms a hydrogen bond with the oxygen of the methoxy group, consistent with the low eld at which the amide proton resonates in the 1 H NMR spectrum. 11</p><!><p>Comparison of the singlet pp* uorescence from the porphyrin moiety of a dyad, with that of a suitable rhenium-free analogue provides valuable information on the quenching ability of the rhenium unit. We have previously reported emission quenching determined in this way for the zinc porphyrin unit in [Dyad 1 pic]OTf (>95% in PrCN) and for Dyad 1 Br (50% in THF) compared with the emission of the rhenium-free ZnTPP-link-Bpy analogue. 34,81 Similar emission measurements were performed with [Dyad 2 pic]OTf and [Dyad 3 pic]OTf (Table 2) showing 55% and 23% emission quenching, respectively (Fig. S17 and S18 †). The emission lifetimes (Table 2) show a similar trend to the steady state measurements. [Dyad 1 pic]PF 6 has the shortest lifetime, followed by [Dyad 2 pic]OTf, which also shows a longer component. [Dyad 3 pic]OTf shows the least shortening of the uorescence lifetime (Fig. S19 †). The emission quenching is attributed principally to electron transfer from excited state zinc porphyrin to the rhenium with only minor heavy atom effects. The large variation in quenching is suggestive of corresponding variations in electron transfer rates.</p><p>Emission quenching in Dyad 1 Br, Dyad 2 Br and Dyad 3 Br was measured relative to zinc tetraphenylporphyrin in THF (Fig. S20 †). In agreement with previous reports, 81 Dyad 1 Br displays 41% emission quenching relative to a simple zinc porphyrin while Dyad 2 Br and Dyad 3 Br show 11% and 0% emission quenching, respectively. A very similar trend is observed in the emission lifetimes (Table 2, Fig. S21 †). We also checked the emission yield of CH 2 -Bpy-ZnTPP (Fig. 2) relative to unsubstituted ZnTPP and found that the emission of CH 2 -Bpy-ZnTPP is 15% more intense than that of ZnTPP for samples of equal absorbance at the exciting wavelength. The minor quenching of the bromide complexes demonstrates that heavy atom effects are unimportant and that electron transfer plays a less signicant role than in the corresponding picoline complexes. We note also that ZnTPP and zinc tetraphenyl chlorin uorescence is not quenched by TEOA. 11</p><!><p>All six dyads were tested for CO 2 photoreduction to CO under irradiation with l > 520 nm in a solution of DMF : TEOA 5 : 1 at 0.05 mM (Fig. 4 and Table 3). Overall turnover frequencies (overall TOF) are calculated over the full period of irradiation, whereas maximum turnover frequencies</p><!><p>UV/vis spectra were taken at regular intervals during photocatalysis. We previously reported that signicant changes occur in the Q-bands of the porphyrins during CO 2 photoreduction catalysis by the porphyrin-rhenium dyads. 11 The Q-bands of porphyrins and their derivatives provide excellent spectroscopic handles in the visible region, providing a clear indication of structural changes. These changes were assigned to formation of chlorin, a reduction product of the porphyrin in which one CC bond of a pyrrole group is saturated.</p><p>The UV/vis spectra of [Dyad 3 pic]OTf during catalysis are shown in Fig. 5. At early photolysis times, the Q-bands of the porphyrin decrease in intensity and a product band grows at 625 nm with a shoulder at 610 nm, seen most clearly in the difference spectra (Fig. 5(b)). The relative intensities of the 610 and 625 nm bands change with time. The band at 625 nm may be assigned to zinc chlorin product, while the 610 nm band is assigned to the zinc isobacteriochlorin, the derivative in which two adjacent pyrrole groups are saturated. 94 This second hydrogenation product is formed in greater amounts for [Dyad 3 pic]OTf and persists longer. For all dyads the photocatalytic conditions eventually lead to complete bleaching of the Q-band region of the spectrum.</p><p>The exact chemical structure of the chlorin cannot be determined from UV/vis spectroscopy alone. It has been shown previously that triethylamine can add to the pyrrole to form both the simple hydrogenation product and a product in which a C-H bond has been formally added across the C]C bond. 95 A large-scale (50 mg) photolysis (l > 520 nm) was performed on ZnTPP in DMF : TEOA 5 : 1 under Ar and the product was exhaustively extracted into ether aer addition of water. The ether was separated and the product dried under vacuum. The 1 H NMR spectrum of the product dissolved in CDCl 3 matches the spectrum of an authentic sample of zinc tetraphenylchlorin (Fig. S23 †), demonstrating that the major product is formed by simple hydrogenation (Fig. 6).</p><p>ESI-mass spectrometry measurements were made on samples from CO 2 photoreduction by [Re(Bpy)(CO) 3 (pic)][PF 6 ] and zinc tetraphenyl porphyrin (ZnTPP). Zinc possesses several isotopes of signicant abundance producing a pattern that spans several m/z units. As a result, the signals for the various hydrogenation products of zinc porphyrin overlap closely. The signals obtained centre around m/z ¼ 680 and match well with the calculated isotope pattern for a mixture of ZnTPP and the dihydrogenated (chlorin) and tetra-hydrogenated (isobacteriochlorin) products (Fig. S24 † and 6). (Fig. S27 †). 96 A similar experiment with Dyad 2 Br produced no change thermally, but a product with bands at the same wavenumbers appeared on photolysis with l > 520 nm under N 2 (Fig. S28 †). In an attempt to increase conversion and the signal of the substitution product, CO 2 was bubbled through Dyad 2 Br in DMF : TEOA 5 : 1 under l > 520 nm irradiation. This time, different signals were observed at 2015, 1911 and 1885 cm À1 that correspond closely to those reported 53 for the carbonato complex (Fig. 7 and S28 †).</p><p>Considering the excellent t to Ishitani's data for Re(OCH 2 -CH 2 NR 2 )(Bpy)(CO) 3 , the product from [Dyad 2 pic]OTf may be assigned as Dyad 2 OCH 2 CH 2 NR 2 . We are not able to show denitively whether the product from Dyad 2 Br is the same or the analogue where the porphyrin has been reduced to chlorin, since the timescale for hydrogenation is similar the timescale for reaction with TEOA. Nevertheless, the IR evidence supports CO 2 insertion into the metal-oxygen bond to form species containing the Re{OC(O)OCH 2 CH 2 NR 2 }(Bpy)(CO) 3 unit. 53 Picosecond time-resolved infrared spectroscopypicoline complexes</p><p>Extensive previous TRIR investigations of the excited states of Re(Bpy)(CO) 3 derivatives 34,68,80 show that formation of 3 MLCT excited states result in high frequency shis of the carbonyl vibrations, whereas charge transfer to the Re(CO) 3 results in substantial low frequency shis. 34 The photophysics and photochemistry of [Dyad 1 pic]OTf have previously been investigated using TRIR spectroscopy in PrCN. 34 Excitation of [Dyad 1 pic]OTf at 600 nm resulted in the initial formation of an excited state localised on the porphyrin moiety of the dyad, followed by subsequent electron transfer to the Re(diimine) ligand generating a charge-separated (CS) state. The CS state reached a maximum within 10 ps and decayed over 40 ps. Charge recombination back to the porphyrin moiety via a hot ground (HG) state regenerated the parent complex within 200 ps. In addition, a sharp peak in the TRIR spectra at 2026 cm À1 could be observed during the rst 5 ps, which was tentatively assigned to the formation of an intraligand (IL) pp* excited state. 97 The complexity of the transient spectroscopy was reconciled with a model which postulates that the dyad molecules adopt a range of conformations each with their own kinetics. We have performed 4.</p><!><p>The photophysics and photochemistry of Dyad 1 Br, Dyad 2 Br and Dyad 3 Br were monitored using TRIR spectroscopy following excitation at 560 nm in THF. All TRIR spectra were obtained in THF since the dyads are not sufficiently soluble in CH 2 Cl 2 . In general, the TRIR spectra of the bromide dyads are more complex than those of the picoline dyads, with multiple transient species observable in the spectra.</p><p>The TRIR spectra obtained following excitation of Dyad 1 Br are shown in Fig. 10. Three negative bands are observed corresponding to the parent complex at 2022, 1922 and 1900 cm À1 . At early time delays (<50 ps) a band at 2055 cm À1 and a broad band at ca. 1960 cm À1 can be observed, characteristic of the high frequency shi associated with the formation of a 3 MLCT state on the Re moiety of the dyad. 68,77,98,99 This 3 MLCT excited state is formed initially <10 ps aer excitation from vibrationally hot excited states and decays over the subsequent 250 ps (Fig. 10(c), blue squares). The formation of peaks at 1998 cm À1 and 2015 cm À1 can also be observed on a similar timescale to the decay of the 3 MLCT state. The peak at 1998 cm À1 is analogous to observations made on the picoline dyads (see above) and is assigned to the formation of a CS state. The corresponding lower energy bands associated with the CS species can be observed at ca. 1880 cm À1 , but due to their weak intensity, the exact band positions could not be determined. The band at 2015 cm À1 suggests the simultaneous formation of an IL pp* excited state, 77,97 similar to that observed following the photolysis of [Dyad 1 pic]OTf in PrCN. 34 The associated low energy bands of the IL pp* excited state cannot be observed as they are low intensity and fall in a similar region of the spectrum to the ground state bleach. Bleaching of the ground state does not reach a maximum negative signal until 15 ps, and it recovers over the subsequent 1000 ps (Fig. 10(c), red dots). The recovery of the signal at 2022 cm À1 occurs over two distinct timescales. The rst (0-100 ps) is mainly associated with deactivation of the 3 MLCT and the second (100-1000 ps) is principally due to the decay of the CS and pp* excited state. The kinetics of the CS state and the IL pp* excited state were not fully determined as the bands are weak and overlap with other bands in this region of the spectrum.</p><p>The TRIR spectra obtained following excitation of Dyad 2 Br are shown in Fig. 11. Parent bleaches at 2020, 1922 and 1900 cm À1 can be observed as well as the formation of two transient species (Fig. 11(b)). At all time delays, bands 2057 cm À1 and ca. 1975 cm À1 (broad) are visible, associated with the formation of a 3 MLCT excited state on the Re moiety. This 3 MLCT state is initially formed from vibrationally hot excited states at time delays <10 ps. In addition, bands at 1997, 1887 and 1871 cm À1 can be observed <500 ps aer excitation, which are assigned to the formation of a CS state. The CS species grows in on a timescale faster than 2 ps and decays over the subsequent 1000 ps (Fig. 11(c), black squares) as the parent bleach partially recovers (65%, Fig. 11(c), red dots). An IL pp* excited state was not observed at any time delay in this experiment. At 500 ps aer photolysis, the only bands visible in the TRIR spectrum are those originating from the 3 MLCT and these bands along with the parent bleaches do not change intensity signicantly on the timescale of this experiment (up to 1000 ps).</p><p>The TRIR spectra recorded aer ash photolysis of Dyad 3 Br are shown in Fig. 12. Bands associated with the formation of a 3 MLCT excited state at 2055 cm À1 and ca. 1975 cm À1 (broad) grow in over the rst 100 ps and do not deplete signicantly up to 1000 ps aer excitation. In addition, an IL pp* excited state band at 2014 cm À1 can be observed that grows in over the rst 30 ps and completely decays by 100 ps. The low energy bands of the IL pp* excited state cannot be observed as they are weak in intensity and overlap with the ground state bleaches. The 3 MLCT state is probably formed via energy transfer from the porphyrin pp* excited state. 97 This is energetically feasible as the higher energy emission maximum of Dyad 1 Br is at 606 nm, compared to the emission maximum for the 3 MLCT state of ReBr(Bpy)(CO) 3 at 620 nm. 81 In contrast to Dyad 1 Br and Dyad 2 Br, a CS state was not observed following the photolysis of Dyad 3 Br. The ground state bleach reaches a maximum at 30 ps and has recovered by 65% at 1000 ps aer excitation. Through a separate ns-TRIR experiment we determined that the 3 MLCT state decays with a lifetime of ca. 2 ns as the parent complex reforms. However, this experiment had to utilise a 532 nm excitation pulse which is not ideal as it falls at the edge of the porphyrin Q band absorption and led to relatively weak TRIR signals. We examined the possible quenching of the 3 MLCT Br. quenching of the 3 MLCT state is expected to be a small component of the decay because of the short excited state lifetime. However, no reductive quenching was observed. Given the low signal-to-noise of these measurements due to the unfavourable excitation wavelength, we can only state that if quenching occurs then it represents less than 1% of the 3 MLCT decay.</p><!><p>The change in free energy for the dyad picoline cations on intramolecular electron transfer from the excited state of the sensitizer to rhenium can be estimated using eqn (1) where E ox and E red are taken as the potentials for the rst oxidation of the sensitizer and rst reduction of the rhenium, respectively. The potentials were estimated from cyclic voltammograms measured in CH 2 Cl 2 . For E 00 , we used the highest energy emission maximum of the sensitizer, measured at room temperature. The potentials, emission maxima, driving forces and maximum TON CO are given in Table 5. We do not report quantum yields for CO production because there is signicant photoreaction at the porphyrin during catalysis.</p><p>For the bromide dyads, we expect an additional electrostatic contribution to the free energy of electron transfer, since the electron transfer generates a pair of charges. The edge-to-edge distance from porphyrin to Bpy may be regarded as the minimum distance for electron transfer and is measured at 8.0 Å in the crystal structure of [Dyad 1 pic]PF 6 . The electrostatic contribution in CH 2 Cl 2 is calculated as À0.20 eV and may be taken as an upper limiting value for Dyad 1 Br. The corresponding value for Dyad 2 Br would be signicantly less negative, while that for Dyad 3 Br may be more negative at ca. À0.27 eV because of its ability to fold about the CH 2 group. However, in DMF, the solvent used for CO 2 reduction, these values become of little importance because of the high dielectric constant of the solvent: À0.04 eV for Dyad 1 Br and À0.05 eV for Dyad 3 Br. 5 for [Dyad 3 pic]OTf and Dyad 1 Br are close to zero, while that for Dyad 3 Br is positive. The bromide dyads have completely different potentials from the picoline dyads yet their photocatalytic behaviour is very similar and sometimes superimposable. Furthermore, Dyad 3 Br is very active, yet the driving force in DMF is not favourable for electron transfer. Considering just the picoline complexes, the greater is the driving force for electron transfer, the lower is the observed maximum turnover number. We can deduce from these points that the porphyrin dyad bromides are not the active species in photocatalysis. However, we have previously shown that the driving force for electron transfer from the excited state of zinc tetraphenylchlorin to Re complexes is 150 meV more negative than that for the excited state of zinc tetraphenylporphyrin. 11 Thus, it is possible that bromide dyads become more active on reduction to the chlorin derivative (see Mechanism section of Discussion, below). The corresponding values of the reduction potentials of Re(OCH 2 CH 2 NR 2 )(Bpy)(CO) 3 (R ¼ CH 2 CH 2 OH) and related dyads are not known, but we would expect them to be close to those of the bromide complexes. The reduction potential of Re(OCOOCH 2 CH 2 NR 2 )(Bpy)(CO) 3 (R ¼ CH 2 CH 2 OH) is reported to be very similar to that of the simple bromide complex. 53</p><!><p>Emission data allow us to derive yields for uorescence from the singlet pp* state and estimates of the intramolecular quenching rate (Table 2). The quantum yields are based on the quantum yield of uorescence for ZnTPP in toluene of 3%. 93 Across both the picoline and bromide dyads, steady state emission quantum yields and emission lifetimes increase signicantly as photocatalytic activity increases. This lack of correlation may be resolved if some of the dyads are emissive but inactive with respect to charge separation, while others are non-emissive but undergo charge separation. Thus the uorescence data show that there are signicant excited state populations that do not directly lead to photocatalysis. This diversity of behaviour is attributed to conformers which are not predisposed to the required electron transfer process.</p><p>If there was no issue of multiple conformers, all the molecules would end up in the state with the shortest rise-time following formation of the porphyrin p-p* S 1 state.</p><!><p>There are some striking differences in the TRIR spectra and kinetics obtained between the different dyads. are not ideal for electron transfer and represent one conformer out of many that may be present in solution (Fig. 3). Comparison shows that [Dyad 3 pic]OTf exhibits the longest risetime for charge separation and the longest lived charge-separated state (Table 4). Only [Dyad 1 pic]OTf undergoes charge recombination via a hot ground state. 34,100 Although the lifetimes of the CS states correlate with photoactivity, they are extremely short if bimolecular reaction is to occur with any species other than either triethanolamine or DMF which are components of the solvent, even for [Dyad 3 pic]OTf.</p><p>The TRIR spectra of the bromide complexes are very different from those of the picoline complexes. In the bromide complexes, we observe a 3 MLCT state in all three dyads and the CS state can only be clearly observed in Dyad 1 Br and Dyad 2 Br. These observations are consistent with the driving force calculations above. The 3 MLCT states of Dyad 2 Br and Dyad 3 Br have lifetimes on the ns timescale. The risetimes of the 3 MLCT states are in the range of tens of picoseconds which is again incompatible with the rate of quenching of the pp* states. We suggest that these differences reect the presence of multiple conformers. The absence of the CS state of Dyad 3 Br appears remarkable considering its strong photocatalytic activity.</p><!><p>On photo-excitation, [Dyad 2 pic]OTf and [Dyad 3 pic]OTf form charge-separated states, as seen for [Dyad 1 pic]OTf. 34 The close match in the photocatalysis curves for [Dyad 2 pic]OTf and Dyad 2 Br (Fig. S22 †) suggests that these compounds share a catalytically active species. The other two pairs also exhibit strong similarity in the curves. The parallel photocatalytic behaviour contrasts with the very different excited states observed by TRIR spectroscopy (see above). We have presented evidence from steady-state spectroscopy that the photocatalysts undergo reaction with triethanolamine both at the porphyrin centre and the rhenium centre. Photoreaction at the porphyrin causes initial 2-electron hydrogenation to the chlorin and subsequently a further 2-electron hydrogenation to the isobacteriochlorin. Thermal reaction of [Dyad 2 pic]OTf and photoreaction of Dyad 1 Br yield evidence for formation of Re(OCH 2 CH 2 NR 2 )-(Bpy)(CO) 3 derivatives. In addition, triethanolamine and DMF are capable of coordinating to zinc, probably forming equilibrium mixtures. An electron can be supplied to the oxidised porphyrin by the h ligand on zinc. Taken together with the arguments presented above on energetics and TRIR spectra, these considerations indicate that the dyads act as pre-catalysts. Within the rst 30 min of irradiation, signicant reduction to chlorin and reaction with triethanolamine occurs, probably generating the true photocatalysts.</p><!><p>We have shown that a new design of zinc porphyrin-Re bipyridine tricarbonyl dyad with a methylene spacer is active for the photocatalytic reduction of CO 2 to CO with l > 520 nm irradiation. The TON is higher than those for previously reported dyads by a factor of ten and there is a major increase in TOF. These gures exceed those for the two component system ZnTPP + [ReBpy(CO) 3 (pic)][PF 6 ] by a factor of three. 11 The benet of using a saturated bridge between Bpy and the porphyrin is in accord with Ishitani's binuclear Ru-Re complexes which were also most effective with a saturated bridge. 10 The most likely reason for the enhanced activity is the exibility for the Re(Bpy) unit to adopt the best orientation and closest approach to the porphyrin unit. Furthermore, in these dyads the groups at the 4 and 4 0 positions of the Bpy are CH species, followed by formation of an isobacteriochlorin and eventually complete bleaching. Previous results indicate the chlorin intermediates are active catalysts. 11 It is likely that the isobacteriochlorin intermediates are active also. Complete bleaching renders the dyads unable to absorb visible light and is one route for deactivation. The dyads react with triethanolamine in DMF to form alkoxide complexes containing a Re(OCH 2 CH 2 NR 2 )Bpy(CO) 3 moiety which undergoes CO 2 insertion. The picoline complexes undergo this transformation thermally while the bromide complexes require irradiation.</p><p>[Dyad 3 pic]OTf undergoes charge separation in 8 ps and the charge-separated state has a lifetime of 320 ps. For comparison, [Dyad 2 pic]OTf undergoes faster charge separation but the majority of the CS photoproduct decays much faster (with time constant of 42 ps). The bridge in Dyad 3 has slowed down charge separation and charge recombination. The chargeseparated state is one order of magnitude longer-lived in Dyad 3 than in Dyad 2. The bromide complexes show very different photochemical behaviour on the ps timescale with combinations of either 3 MLCT and CS, or 3 MLCT and IL excited state products. Dyad 1 Br and Dyad 2 Br form some CS product, whereas Dyad 3 Br does not. The CS state is unlikely to be responsible for the activity of the bromide complexes since Dyad 3 Br is very active. All three bromide dyads display formation of a 3 MLCT state, the lifetimes of which decrease in the order Dyad 3 Br > Dyad 2 Br > Dyad 1 Br, in line with their photocatalytic activities. The 3 MLCT state may be responsible for the activity of the bromide dyads. As expected for such short lifetimes, ns-TRIR experiments on Dyad 3 Br showed little or no bimolecular reaction with TEOA but this cannot be ruled out. Thus the bromide complexes display very similar photocatalytic behaviour to the picoline complexes but totally different excited states.</p><p>Taken together, the data strongly suggest that the active photocatalyst is formed by a combination of reaction of triethanolamine at rhenium and photoreduction of the porphyrin. We have previously shown that zinc chlorin is more reducing than zinc porphyrin 11 and may allow for the formation of signicant amounts of charge-separated state in the bromide complexes as well as the picoline complexes. This hydrogenation can also explain why the picoline dyads are not de-activated on thermal substitution of picoline for the anionic alkoxide/ carbonato complexes, which would be expected to have similar reduction potentials to the bromides. TRIR experiments demonstrate that bimolecular reaction of the 3 MLCT state of Dyad 3 Br is minimal and thus support the chlorin theory.</p>

Royal Society of Chemistry (RSC)

Discovery and Optimization of Triazine Nitrile Inhibitors of Toxoplasma gondii Cathepsin L for the Potential Treatment of Chronic Toxoplasmosis in the CNS

With roughly 2 billion people infected, the neurotropic protozoan Toxoplasma gondii remains one of the most pervasive and infectious parasites. Toxoplasma infection is the 2nd leading cause of death due to foodborne illness in the US, causes severe disease in immunocompromised patients, and is correlated with several cognitive and neurological disorders. Currently, no therapies exist capable of eliminating the persistent infection in the central nervous system (CNS). In this study we report the identification of triazine nitrile inhibitors of Toxoplasma cathepsin L (TgCPL) from a high throughput screen, and their subsequent optimization. Through rational design, we improved inhibitor potency to as low as 5 nM, identified pharmacophore features that can be exploited for isoform selectivity (up to 7-fold for TgCPL versus human isoform), and improved metabolic stability (t1/2 > 60 min in mouse liver microsomes) guided by a metabolite ID study. We demonstrated this class of compounds is capable of crossing the blood-brain barrier in mice (1:1 Brain:Plasma at 2 hr). Importantly, we also show for the first time that treatment of T. gondii bradyzoite cysts in vitro with triazine nitrile inhibitors reduces parasite viability with efficacy equivalent to a TgCPL genetic knockout.

discovery_and_optimization_of_triazine_nitrile_inhibitors_of_toxoplasma_gondii_cathepsin_l_for_the_p

INTRODUCTION<!>Synthesis of TgCPL inhibitors<!>Docking of the triazine scaffold into TgCPL and HsCPL:<!>SAR of P3 Vectors<!>SAR Analysis of P2 Vectors:<!>P2 Vectors for Improved Metabolic Profile and Selectivity:<!>Pharmacokinetics:<!>In vitro efficacy studies:<!>CONCLUSIONS AND PERSPECTIVES<!>Chemistry General Information:<!>General Procedure A1: Reductive Amination<!>General Procedure A2: SNAr<!>General Procedure A3: SNAr<!>General Procedure A4: Cyanation<!>General Procedure B1: One-pot double SNAr<!>General Procedure B2: Cyanation<!>General Procedure B3: Alkylation with NaH and aryl/alkyl X/Ms/Ts<!>General Procedure B4: Boc-Deprotection<!>Computational Modeling:<!>TgCPL and HsCPL Inhibition Assay:<!>High throughput library screen:<!>Metabolic Stability in Mouse Liver Microsomes:<!>Pharmacokinetic Studies in Mice:<!>MDR1/MDCK Assay:<!>Bradyzoite Viability Assay:<!>HFF Cell Viability by MTS Assay:

<p>Estimates suggest over 30% of the global human population is currently infected with Toxoplasma gondii, with approximately 60 million infected in the United States alone.1–3 Capable of infecting nearly all warm-blooded mammals, this single-celled apicomplexan is the second leading cause of death due to foodborne illness in the US.4 As a member of the phylum Apicomplexa, Toxoplasma gondii exhibits a complex infection cycle in humans.5 Through consumption of contaminated food, or direct exposure to cat fecal matter, the parasite enters an initial tachyzoite infection stage, in which the parasite undergoes rapid asexual replication and disseminates throughout the host tissues, including the central nervous system (CNS).6 Upon activation of the host immune response, T. gondii transforms into its chronic bradyzoite phase, in which the parasite exists as lifelong tissue cysts within the muscle and CNS. If the host maintains a healthy immune system, this chronic infection generally presents few symptoms, although recently there is a growing body of evidence suggesting that this chronic stage infection may significantly contribute to a wide array of neuropsychiatric symptoms including schizophrenia, bipolar disorder, obsessive compulsive disorder, and various other mood alterations.7–10</p><p>A more obvious impact on human health from this parasite is revealed when the host immune system is compromised. The unchecked parasite cysts reactivate into the acute tachyzoite form, causing tissue inflammation and more serious complications such as: encephalitis, blindness, and death.11 Treatment options for toxoplasmosis reactivation are generally pyrimethamine in combination with sulfadiazine, or clindamycin.12 However, these options are prone to adverse side effects and only act upon the parasite in the peripheral system. Currently, no approved treatment options exist to clear the latent and chronic Toxoplasma infection in the CNS, therefore it is critical to develop new and alternative treatment strategies. A recent review by Deng et al13 summarizes current approaches to targeting enzymes critical to the life cycle of the parasite, including calcium-dependent protein kinase 1 (TgCDPK1), thymidylate synthase-dihydrofolate reductase, enoyl-acyl carrier protein reductase, and cathepsin L. TgCDPK1 inhibitors in particular have shown promise for treating the chronic form of T. gondii.14</p><p>Cathepsins L and B are the major cysteine proteases found in protozoans, and are emerging as viable targets for anti-parasitic agents due to their essential roles in parasite survival.15–17 For example, Rhodesain (T. brucei), cruzain (T. cruzi), and falcipain (P. falciparum) express cathepsin L-like proteases that have gained a considerable amount of attention for the potential treatment of African sleeping sickness, Chagas disease, and malaria respectively.18 Toxoplasma gondii expresses five members of the C1 family of cysteine proteases, including TgCPL, TgCPB, and TgCPC1–3.19–22 TgCPL is localized in the plant like-vacuole (PLV) vacuolar compartment (VAC, used hereafter), and is also responsible for the maturation and activation of TgCPB.20, 23 Although inhibition or disruption of TgCPL in the parasite has been shown to moderately impede parasite invasion and growth in the tachyzoite stage of infection, TgCPL is not critical to the acute stage.21, 23 Interestingly, parasites in the cystic bradyzoite stage express heightened levels of TgCPL and TgCPB, suggesting proteolysis in the VAC plays an important role in the chronic infection.24 Accordingly, bradyzoites deficient in TgCPL die after forming cysts in both culture and in the neuronal cells of infected mice. Upon restoration of catalytically active TgCPL, bradyzoite viability is re-established. In contrast, expression of a catalytically inactive TgCPL fails to restore cyst viability. Taken together, these strongly implicate TgCPL as a viable target for the treatment of this chronic stage infection. The parasite dependence on TgCPL has been further validated in vitro, as well as through the treatment of cysts with the covalent, irreversible cathepsin inhibitor LHVS (1, Figure 1).25 This disruption of the proteolytic activity in the lysosomal VAC ultimately results in parasite death.26 However, LHVS is not a viable pharmaceutical lead as it failed to reduce neural cysts in infected mice, due to its poor drug-like properties (e.g., fails to cross the blood-brain barrier).27</p><p>Recently, we reported the first lead optimization study for the inhibition of TgCPL with a series of dipeptide nitriles.28 While we were able to gain insight into the pharmacophore necessary for TgCPL inhibition and significantly improve the potency and selectivity over human isoforms, we were unable to carry forward the dipeptide series due to the inadequate pharmacokinetic (PK) properties of optimized lead 2. Despite being potent in vitro and demonstrably CNS penetrant, the drug was rapidly cleared in vivo. While the dipeptide-nitriles are synthetically convenient for investigating SAR in the cathepsin pockets, they have less than desirable physiochemical properties for gaining robust blood-brain barrier (BBB) permeability, such as several rotational bonds, extensive hydrogen bonding, and high topological polar surface area (TPSA).29 We thus began searching for alternative scaffolds that might exhibit better physiochemical properties to improve upon the in vivo PK, while maintaining the pharmacophoric elements necessary for potent TgCPL inhibition.</p><p>A high throughput screen for TgCPL inhibitors was conducted at the Center for Chemical Genomics at the University of Michigan on over 150,000 small molecules. From this we identified several inhibitor classes that demonstrated promising inhibitory activity against TgCPL, reasonable synthetic approachability, and physiochemical profiles optimal for use as a lead scaffold in a CNS targeted campaign. To minimize potential PK and off-target toxicity issues in later development, we chose to exclude compounds that were peptidic in structure, or irreversibly covalent. After our triage process had identified several potential lead molecules, we searched the literature to determine if any of these chemotypes had any previously demonstrated activity against cysteine proteases. An extensive amount of study has been done on inhibitors of human cathepsin L (HsCPL).30 Since Toxoplasma gondii cathepsin L shares high homology with HsCPL, we considered chemotypes with precedent for inhibiting HsCPL as good potential leads. The overall sequence similarity between the two is about 50 %. However, for residues within 4.5 A of the triazine, the similarity is about 72 %. Only 5 out of 18 residues in the binding site are different in the two structures (Figure 2). Among the several potential lead molecules discovered in our screen, we selected a modestly active triazine nitrile 3 (TgCPL IC50 = 2.5 μM, HsCPL IC50 = 2.3 μM) as a lead candidate (Figure 1), given its fragment-like size and solid literature precedent as a cathepsin inhibitor class. Several recent publications have reported triazine nitrile-based inhibitors of human cathepsins, and additionally demonstrated potent inhibition of rhodesain and falcipain.31–33 The activity reported against these analogous cysteine proteases from the related apicomplexan parasites P. falciparum and T. brucei further supports the potential of this chemotype as an antiparasitic agent. Additional triazine nitriles, pyrazolopyrimidine nitriles, and related heterocylic-nitrile motifs have also been reported as inhibitors of cysteine proteases.34–37</p><p>Although these recent studies have established the potential of this chemotype as a clinically relevant antiparasitic agent, Toxoplasma gondii, as well as other protozoan parasites such as T. brucei, establish a chronic infection in the CNS. Compounds intended for CNS targets have a somewhat constrained physiochemical profile as compared to peripheral drugs. Key among these are the minimization of both the number of hydrogen bond donors and rotatable bonds, which can increase the rate of passive BBB penetration. Additionally, CNS drugs exhibit relatively low MW (<450), high lipophilicity (logP 2–5), and exclusion of acidic moieties. The triazinyl-nitrile scaffold was particularly promising in that it could readily incorporate the structural motifs necessary for binding in the S2 and S3 pockets of TgCPL (Figure 2), which we elucidated in our previous dipeptide work. This scaffold also has a potentially improved pharmacokinetic profile and CNS permeability by reducing the rotational degrees of freedom, eliminating a hydrolyzable amide and removing two hydrogen bond donors (HBD). Molecular docking and energy minimization of the dipeptide nitrile (2) and proposed triazine nitrile (4) in the active site of TgCPL showed good structural agreement into the pharmacophore of TgCPL inhibition (Figure 3). Altogether this encouraged us to attempt a scaffold-hop from the optimized dipeptide to a triazine nitrile. In this paper we report our success at this transformation, and our initial SAR work at enhancing selectivity for the parasite enzyme vs human cathepsin L. Furthermore, we demonstrate for the first time the ability of this class of compounds to inhibit bradyzoite viability in vitro and to penetrate into the brain in vivo, two key milestones for developing an effective therapeutic for the chronic stage of T. gondii.</p><!><p>Under anhydrous conditions, the desired aryl aldehyde or acetophenone Int-1a was condensed with the desired alkylamine overnight, at which point the formed imine was reduced with sodium borohydride or sodium triacetoxyborohydride (Scheme 1). Next, the reaction between the secondary amine Int-2a and cyanuric chloride provided intermediates Int-3a. Subsequent SNAr with morpholine afforded intermediates Int-4a. In the case of 50, oxetanamine was used as the P1 solubilizing group in place of morpholine. Finally, cyanation of the scaffold with potassium cyanide in DMSO provided the desired final compounds.</p><p>To expedite rapid production of diverse analogs, a synthetic strategy analogous to that reported by Giroud et. al. for the synthesis of triazine-nitrile HsCPL inhibitors was employed (Scheme 2).31 First, cyanuric chloride underwent SNAr reactions with the desired benzyl or alkyl amines Int-1b to set the various S2 and S3 vectors for intermediates Int-2b. A second SNAr step was employed to install a morpholine as a P1 solubilizing group. Cyanation of the intermediate was performed with potassium cyanide in DMSO, to afford Int-3b. Subsequently, the anion of intermediates Int-3b was generated using sodium hydride, and the desired alkyl/benzyl halide was added to provide analogs Int-4b. In the case of analogs bearing a Boc-protected amine side chain, TFA:DCM deprotection was employed to provide the desired final compounds.</p><!><p>The IC50 for the triazine nitrile we identified in our HTS (3) was benchmarked at 3.5 μM and 2.3 μM against TgCPL and HsCPL, respectively (Table 1). Considering this lead bears an ethyl as the only potential P2/3 vector, it left obvious room for improvement. As previously discussed (Figure 3), we elected to first synthesize the triazine nitrile 4 analogous to an optimum dipeptide nitrile from our previous work. Figure 4 presents a comparison of models of 4 docked into the active sites of TgCPL and HsCPL. The triazine nitrile replaces the P1–P2 amide bond and retains the reversibly covalent nitrile interaction with the active site cysteine. In the P2 position, we previously found that leucine is the optimal dipeptide amino acid, so in the respective position on the triazine, which we hypothesized would project into P2, we placed an isoamyl amine to achieve the same length into the pocket. In P3, our most active compound previously bore a 4-fluorobenzyl, and therefore we placed the same motif in the hypothetical P3 position for the triazine nitrile scaffold. An additional feature found both in our HTS lead and known literature analogs was the P1 morpholine. While the S1 pocket in TgCPL is absent, the morpholine in the P1 position projects into solvent and has been shown to improve the solubility of this chemotype (Figure 4).</p><p>The installment of these two pendants in our first triazine analog 4 indeed provided over a 100-fold improvement in potency over the HTS lead 3, affording an IC50 of 34 nM and 32 nM for TgCPL and HsCPL, respectively (Table 1). These results supported our docking model, and validated our hypothesis that we could successfully scaffold-hop from the dipeptides to the triazine, suggesting that much of our dipeptide SAR should translate directly to the triazine series. The active site of TgCPL contains four residues (Figure 4A) that are either non-conserved across the human cathepsin isoforms A-X, or unique to TgCPL. Notably, the S3 pocket of TgCPL tends to be significantly more polar as compared to HsCPL, with Gln69, Asp78, and Glu75 making up the key pocket residues. The human cathepsin L (Figure 4B) is somewhat more lipophilic with the S3 pocket residues Glu63, Tyr72 and Leu69, respectivly. While this presents potential for differences in ligand selectivity, the S3 pocket is shallow and somewhat promiscuous with its preferred residues. We anticipated that the majority of our potential selectivity would come from the S2 pocket, which tends to be deeper and somewhat more defined. Importantly, the parasitic cathepsin bears an Asp218 in the S2 pocket, while the human isoform contains an Ala214 in the corresponding position. We therefore expected that selectivity for TgCPL could be achieved by exploiting the differences in the overall topology and unique residues. As such, we chose to explore various P2 and P3 vectors to probe these diffences and elucidate the pharmacophore for selective and potent inhibition of TgCPL.</p><!><p>We began our SAR investigation by retaining the isoamyl P2 vector and varying the P3 position (Table 1). A scan of fluorine mono-substitution of the P3 benzyl (5 and 6) showed retention of activity, but no significant improvement in either overall potency or selectivity. A similar outcome was seen with the P3 di-fluoro analogs 7, 8, and 9. The non-fluorinated P3 benzyl pendant 10 exhibited roughly a 3-fold loss of potency versus the 4-fluoro lead 4. Homologation of the linker from methyl to ethyl (11) or propyl (12), while better tolerated for the human isoform, resulted in the same reduced potency for TgCPL. An even greater loss of potency for the parasitic enzyme was seen when large bulky pendants like the quinoline or naphthyl of 13 and 14 were installed. Taken together, this demonstrates that the parasitic isoform does not productively accomodate large lipophilic vectors in the P3 position, which is consistent with the topologically more polar nature of the S3 pocket defined by Asp78 and Glu75 (Figure 4A).</p><p>With this in mind, we tried the 2-, 3-, and 4-pyridyl pendants (15, 16, and 17) in P3, with the objective of improving our selectivity for the parasite, as well as potentially improving our metabolic profile by decreasing the clogP. Of these, the 3-pyridyl (16) retained potency, but unfortunately did not offer any improvement in selectivity over the human isoform. Anilines 18, 19, and 20 were also evaluated. We had hoped to gain a meaningful interaction with Asp78, Gln69, or Glu75 in TgCPL (Figure 4A), but none of the anilines were very potent for TgCPL. However, analog 20 did afford a small measure of selectivity toward TgCPL (HsCPL IC50/TgCPL IC50 = 2.2). Analog 21 was tested to determine if a basic amine might afford greater TgCPL selectivity, but resulted in nearly a 9-fold loss in potency versus parent compound 4.</p><p>As a comparison to our 4-fluorobenzyl pendant, we tested the 4-chlorobenzyl analog 22 to determine the impact of a larger halogen. While equally effective against TgCPL to the 4-fluoro, this change afforded an order of magnitude improvement in potency for HsCPL. Analog 23 was synthesized to assess the effects of a moderately electron withdrawing acetophenone, but afforded a mild reduction in potency against TgCPL. A methyl scan was performed to probe the space around the ring and determine if any of the substitution patterns might lead toward some selectivity for TgCPL. While 24 and 25 did not offer any improvement in potency, 26 was nearly twice as potent (TgCPL IC50 = 19 nM) as the parent scaffold 4, but more strongly improved HsCPL inhibition (HsCPL IC50= 5 nM), similar to what we observed with the 4-chloro analog 23. It is worth noting that the o-methyl benzyl analog 24, while less potent than the 4-flurobenzyl, showed mild selectivity for TgCPL. Dimethyl analogs 27 and 28 provided more potent inhibitors and were consistent with the trend that an ortho substitution (28) may favor selectivity for TgCPL. Undesirably, these also increased overall molecular weight and introduced potential metabolic hotspots, and therefore were not further pursued. The electron withdrawing 3-trifluoromethyl 29 provided one of the more potent TgCPL inhibitors (IC50 = 17 nM), but exhibited selectivity in favor of HsCPL. The increased MW versus 4 was also undesirable in trying to remain within the calculated properties for BBB permeability. The corresponding 3-methoxy analog 30 was also tested to assess the effects of electron donation into the ring; however, this did not seem to offer any significant improvement in potency or selectivity as compared to 29. Interestingly, a nitro scan in analogs 31, 32, and 33 revealed that the o-nitrobenzyl P3 vector of 31 exhibited a 3-fold selectivity over the human isoform (TgCPL IC50 = 27 nM, HsCPL IC50 = 96 nM), consistent with the modest selectivity we observed with ortho-methyl analogs 24 and 28. As this was the most TgCPL-selective analog to date, additional ortho-substituted analogs 34, 35, 36, and 37 were synthesized in an effort to better explain the observed selectivity. The nitrile (34), trifluoromethyl (35), and methyl ester (35) vectors were tested as isosteres of the nitro. Unfortunately, none of these substituents reproduced the selectivity we saw with the o-nitro vector. The o-methoxy analog 37 was also synthesized to determine if the observed effects were due to the electron withdrawing nature of the nitro, or a conformational change due to sterics. While tolerated in HsCPL, the potency significantly decreased against TgCPL.</p><p>Based on observations in our modeling (Figure 5), we hypothesize that the P3 pendant may have the ability to bind outside of the S3 pocket and interact with the backbone of Asp166, giving rise to a higher tolerance for vectors as compared to that of S2. This concept is supported by the overall lack of robust SAR trends for the S3 pocket and the reltively high substrate tolerance in the P3 position observed in analogs 4–47. Furthermore, a recent publication observed a similar trend in triazine nitrile inhibitors of human cathepsin L.31</p><!><p>We then turned our attention to evaluating the SAR among the postulated S2 vectors. In general, the papain family of proteases achieve their substrate specificity primarily from interactions in the S2 pocket.38 Based on our previous SAR studies with the dipeptide nitriles, we had determined that leucine was the optimal amino acid for the P2 position. As already noted with 4, the isoamyl sidechain in the respective position for the triazine nitriles afforded a huge improvement in potency to HTS lead 3 (TgCPL IC50 = 34 nM). We wanted to first determine if the previous dipeptide SAR trends we found tracked in a similar fashion with this new chemotype. Initially, we evaluated the effects of shorter, aliphatic vectors. The n-propyl, n-butyl, and n-butenyl (38, 39, and 40) were tolerated in HsCPL, but all decreased in potency for TgCPL. Consistent with our previous SAR, the shorter isobutyl S2 vector in 42, which closely mimics a valine sidechain, slightly decreased activity for TgCPL to 67 nM, but enhanced HsCPL inhibition (IC50 = 2 nM). Interestingly, the dehydroleucine analog 41 retained nearly equivalent potency (TgCPL IC50 = 46 nM) to 4, a trend that was not observed in the previous dipeptide series. This indicated to us that, while similar, the SAR between the dipeptide and triazine chemotypes was not identical and that binding of the triazine nitrile series may be somewhat better accomodated by the enzyme. Extension of the S2 sidechain of 4 by one carbon in 43 resulted in a significant loss of potency. We believe this may be due to a clash in the back of the S2 pocket, consistant with the S2 size constraints observed previously.28 Compound 44 was made to determine if the S2 pocket could tolerate a less lipophilic group thereby reducing overall clogP and improving compound solubility. Unfortunately, this resulted in a large decrease in potency for both enzymes. Given that TgCPL bears a unique Asp218 versus the Ala214 in HsCPL (Figure 4), we rationalized that inclusion of a basic group in the S2 position might afford some selectivity, as well as improve compound solubility. As such, compounds 45-48 were synthesized. We had hoped that the stepwise increase in side chain length might elucidate the ideal distance required to gain a productive interaction with Asp218. Unfortunately, these changes were not well tolerated and resulted in a drastic drop in potency down to the low micromolar range for both enzymes, indicating that perhaps these compounds are not binding as predicted, or that Asp218 is not oriented the way it appears in our model.</p><!><p>The previously developed dipeptide nitrile 2 exhibited a fair level of stability toward mouse liver microsomes (MLM) with a half-life of t1/2 = 22 min. While the scaffold hop from dipeptide 2 to the triazine nitrile 4 resulted in an improvement in both potency and predicted CNS profile, it also imparted a significant drop in metabolic stability (MLM t1/2 = 7 min). The metabolic liabilities predicted by SmartCyp (Figure 6A) led us to install the α-methyl and exchange the P1 solublizing group with 3-amino-oxetane in analog 50 (Table 3). Unfortunately, this did not offer any improvement to the microsomal stability profile (MLM t1/2 = 5 min). Therefore, a metabolite ID study was performed to experimentally determine sites of oxidative metabolism. As shown in Figure 6B oxidation of 4 occurs predominantly on the P2 isoamyl vector and the P3 benzylic methylene. Similar oxidation of S2 isoamyl sidechains has been reported in the development of odanacatib for the inhibition of cathepsin K, and it was overcome by replacement of the sidechain with fluoroleucine.39 With this guidance in mind, analogs were designed to retain the physiochemical profile of a CNS drug and simultaneously circumvent these potential metabolic liabilities (Table 3).</p><p>Analog 49 was prepared first to evaluate the impact of the alpha-methylbenzyl on potency since we had previously observed a loss of potency with oxetane analog 50. In fact this single change was well tolerated and provided a small improvement in overall potency; however, the potency improvement was 2-fold greater for HsCPL. Presumably due to the added MW and lipophilicity, compound solubility significantly decreased versus the parent compound (thermodynamic aqueous solubility: 4 = 42 μM, 49 = 16 μM). In an effort to resolve this, we tried replacing the S1 morpholine group with oxetanamine in analog 50. As noted previously, the potency decreased to 92 nM for TgCPL, but solubility improved to 42 μM. Analogs 51, 53, and 54 were all synthesized with the intent of blocking oxidation of the isoamyl CH.39–40 The cyclopropyl leucine mimic (51) lost an order of magnitude in potency (TgCPL IC50 = 310 nM) compared to 4, and the t-butyl analog, 53, while 2-fold selective for TgCPL, decreased to 68 nM against TgCPL. It should be noted, however, that these respective changes in the dipeptide nitrile series were not tolerated nearly as well, giving us some confidence that the triazine nitrile may be a superior scaffold with regards to potential for improving PK properties. Furthermore, this data suggests these two chemotypes are not binding in precisely the same manner, consistent with what we had observed earlier in our SAR. Compound 52 was synthesized to determine if the cyclopropyl and α-methylbenzyl combination in P2 and P3 respectively would be sufficient to slow metabolism. Unfortunately, this actually decreased the MLM t1/2 to 4.4 min, indicating a likely switch to another metabolic site, perhaps N-dealkylation of the P2 sidechain or oxidation of the P1-morpholine. Compound 54, bearing a fluoroleucine in S2 (a vector also not tolerated in the dipeptide series), demonstrated excellent potency for TgCPL (IC50 = 17 nM) and HsCPL (IC50 = 26 nM), although MLM stability was decreased to t1/2 = 3.5 min, further indicating that blocking the primary metabolic site alone is insufficient for improving stability. The ethyl-cyclohexyl vector of 55 resulted in a significant decrease in potency for both cathepsins, likely due to the same steric clash in the S2 pocket observed for compound 43. Interestingly, shortening the pendant by one carbon to the methyl-cyclohexyl (56) both retained potency for TgCPL (IC50 = 26 nM) and achieved nearly 4-fold selectivity for the parasite isoform over human. Furthermore, it doubled the MLM t1/2 of lead 4 to 15 min. The P2 cyclopentyl in 59 is comparable to the cyclohexyl in terms of potency, but not as selective. The smaller ring sizes in analogs 57 and 58 significantly decreased potency for TgCPL and were not further pursued.</p><p>Given the favorable TgCPL selectivity imparted by the P2 cyclohexyl ring for 56, we also explored incorporation of a basic amine with the objective of enhancing this selectivity through a favorable ionic interaction with Asp218. As such, S2 piperidinylmethyl compounds 62 and 63 were synthesized. Interestingly, the 4-piperidine of 63 resulted in sharply reduced activity against both enzymes, while the 3-piperidine 62 demonstrated good selectivity (7-fold) for TgCPL. Despite a decrease in overall potency to 0.180 μM, 62 exhibited a much improved MLM stability with t1/2 > 60 minutes. The drastic improvement in stability seen for compound 62 is likely due to the added charge of the basic amine impeding CYP binding. The addition of an extra carbon in the sidechain for analog 60 decreased potency to 1.9 μM for TgCPL, consistent with the poor activity of the homologated cyclohexane analog 55. The ethyl 4-piperidine P2 vector 61 showed a slightly improved potency as compared to analog 60, however it was still significantly lower than parent analog 56. The respective tetrahydropyran analogs 64 and 65 were made to determine if the basic amine was important for this observed selectivity. These displayed a similar potency and selectivity profile, with the heteroatom being better tolerated at the 3-position (65, TgCPL IC50 = 49 nM) than the 4-position (64, TgCPL IC50 = 400 nM). Unfortunately, compound 65 was not stable to MLM. Reduction of the piperidine ring size to a pyrrolidine in analog 66 abolished the selectivity and potency profile. Finally, we attempted to combine our most selective P2 (3-piperidinylmethyl) and P3 (o-nitro benzyl) vectors with analog 67. Unfortunately, these effects were not synergistic and resulted in a loss of both selectivity and potency, strongly suggesting that the two congeners are not binding the same way.</p><!><p>Compound 4 was evaluated in a mouse PK study to benchmark the CNS profile of the triazine nitrile chemotype and compare it to that of the dipeptide nitrile 2 (Table 4). We were pleased to find that 4 does in fact access the CNS with a brain AUC/plasma AUC = 0.61. However, the compound levels decreased rather quickly, likely due to a combination of high volume of distribution and metabolic instability. Nonetheless, this was a promising result indicating that the triazine nitrile scaffold has the physiochemical profile needed to cross the blood-brain barrier.</p><p>We hoped that by increasing the metabolic stability of the scaffold, we might gain a better pharmacokinetic profile. We attempted to evaluate compound 56, as this analog exhibited an improved half-life in vitro, good potency, and demonstrated some selectivity for TgCPL, but unfortunately its poor solubility precluded in vivo dosing. As mentioned above, the replacement of the P1 morpholine of 49 with a 3-amino-oxetane in compound 50 offered an improvement in solubility (16 μM to 42 μM respectively). We decided to perform a PK study on 50 to determine if the additional hydrogen bond donor of the oxetanamine would impede our CNS penetrance, or if this could be a viable strategy to improve the solubility of future compounds (Table 4). We were very pleased to find that in addition to retaining CNS access, the overall exposure levels were significantly increased compared to 4. Given that the 3-piperidine P2 vector in 62 significantly improved the metabolic stability, we also evaluated its pharmacokinetic profile. While the exposure levels were improved in plasma versus 4, this compound had markedly reduced BBB penetrance. Basic amines are often associated with efflux by Pgp, so the difluoro-3-piperidine compound 68 was synthesized with the intent of attenuating the basicity of this sidechain (Table 3). Unfortunately, this change was not well tolerated and potency was reduced to low micromolar for this compound (TgCPL IC50 = 1.6 μM).</p><p>Because of their markedly divergent brain:plasma ratios observed in the PK studies, compounds 50 and 62 were assessed in vitro to determine whether they are potential substrates of P-glycoprotein (P-gp). Evaluation in an MDR1/MDCK permeability assay indicated that compound 50 has high passive permeability and an efflux ratio <1, and thus does not appear to be a Pgp substrate (Table 5). This would be expected to translate to good CNS penetrance, which is in agreement with our PK results. 62 also demonstrated high passive permeability, but has an efflux ratio of over 17, indicating this compound is likely a P-gp substrate, explaining its poor brain exposure in vivo. As noted above, this is very likely to be due to the presence of the basic amine in 62.</p><!><p>Based on their potencies vs TgCPL, five compounds were selected for evaluation of their effects on plaque formation in a T. gondii bradyzoite cyst qPCR/plaque viability assay in vitro (Figure 7). This assay involves isolating bradyzoites after treatment and quantifying viability by applying them to a new monolayer of host cells. Viable bradyzoites invade the monolayer, differentiate to tachyzoites, and form a visible plaque after 12 days of growth. Viability is calculated as the number of plaques/1000 genomes of input bradyzoites (measure by qPCR) and expressed as a percentage of solvent (DMSO) treated bradyzoite cysts. Positive controls in the assay included a TgCPL knockout strain of T. gondii (Δcpl), as well as the irreversible TgCPL inhibitor LHVS (1)25. We were pleased to find that the triazine nitriles 4 and 50 exhibited strong efficacy at 5 μM that was equivalent to the genetic knockout and irreversible inhibitor LHVS at 1 μM, with zero plaques formed in any biological replicate. To our knowledge, this is the first example of a reversible inhibitor of TgCPL achieving in vitro efficacy against the chronic (bradyzoite) stage of T. gondii. Triazine nitrile analogs 56, 65, and 62, on the other hand, showed variable efficacy. 56 and 65 were slightly less efficacious than 4 and 50 at the concentration tested, reducing cyst viability by ~80% across three biological replicates. Only one biological replicate of compound 62 reduced viability (<20%), suggesting that it was comparatively ineffective. Dipeptide nitrile lead (2) demonstrated only modest activity against the bradyzoite cysts, exemplifying the improved potential of the triazine nitrile as a therapeutic.</p><p>To confirm that the observed reduction in bradyzoite viability was not due to human foreskin fibroblast (HFF) host cell viability being indirectly compromised, MTS assays were also conducted to assess cell cytotoxicity (CC). The relatively high CC50 values indicate that none of the compounds would be expected to cause significant host cell cytotoxicity at the concentration used in the qPCR/plaque bradyzoite cyst viability assays (5 μM).</p><!><p>Throughout this project, significant advances have been made in developing small molecule inhibitors of TgCPL. We have elucidated key SAR for the inhibition of TgCPL, developed potent and CNS penetrant inhibitors, and demonstrated efficacy against bradyzoite cysts in vitro. We started by conducting a high throughput screen to identify new chemotypes capable of the inhibition of Toxoplasma gondii cathepsin L as potential new therapeutics for toxoplasmosis. The low μM triazine nitrile scaffold (3) was selected based on the previously established literature precedent of this chemotype for inhibition of human cathepsins, its low molecular weight, and physical properties consistent with CNS-active compounds. Translation of the pharmacophore previously developed with our dipeptide nitrile series to the triazine nitrile scaffold resulted in an immediate 100-fold gain in potency (TgCPL IC50 = 34 nM), but reduced metabolic stability relative to the corresponding dipeptide. A TgCPL covalent docking model was developed using the crystal structure of TgCPL and the crystal structure(s) of human cathepsin (HsCPL) in complex with various dipeptide and triazine nitrile inhibitors. Using this we were able to identify key distinctions between the parasite and human isoforms of cathepsin L. In particular, the smaller S2 pocket and the four unique residues (Asp218, Glu75, Asp78, and Gln69) were targeted with the goal of gaining selectivity for TgCPL. 68 analogs in this series were synthesized with varying S2 and S3 vectors to optimize metabolic stability, CNS permeability, and selectivity over human isoforms. In general, we found that variations in the P3 position had little impact on potency or selectivity. As expected, the bulk of the potency and selectivity seems to be imparted by efficient binding in the S2 position. The S2 pocket across the human cathepsins tends to be more uniformly lipophilic than TgCPL, and we predicted that inclusion of a basic residue in the P2 position might interact favorably with non-conserved Asp218 in TgCPL. It is worth noting that out of the nine analogs bearing an amine in P2, all exhibited IC50s for HsCPL of >0.5 μM, supporting that the human isoform does not tolerate a basic amine in S2. TgCPL, conversely, appears to be able to better accommodate either a nitrogen or oxygen in S2, likely due to the polar Asp218 residue in the back of the S2 pocket, and this may represent an avenue for further TgCPL selectivity enhancement. Work in our laboratory is underway to exploit these features and further explore the SAR trends with various basic and polar groups in P2.</p><p>Overall, we improved potency against TgCPL to as low as 5 nM and identified features that can be exploited to gain selectivity vs HsCPL (up to 5–7 fold), and significantly improved metabolic stability (up to t1/2 >60 min in MLM). Importantly, we were able to demonstrate significant brain penetration with several analogs in vivo in mice. Although our most Tg-selective analog 62 bearing a basic amine in the P2 pendant proved to be highly susceptible to efflux in MDR1-MDCK cells in vitro and demonstrated low brain exposure in vivo, an analog lacking this basic amine (50) was shown not to be a P-gp substrate and achieved good exposure in the brains of mice after IP dosing, suggesting the triazine nitrile chemotype has promise as an in vivo probe for chronic T. gondii infection studies. Synthesis of new analogs to further improve PK and isoform selectivity are underway. Finally, we demonstrated for the first time that in vitro treatment of T. gondii bradyzoite cysts with non-peptidic triazine nitrile inhibitors reduces parasite viability with efficacy equivalent to a TgCPL genetic knockout, in contrast to one of our previously developed dipeptide nitriles, highlighting the increased therapeutic potential of the triazine template. This is the first example of a CNS penetrant, reversibly covalent TgCPL inhibitor showing efficacy in an in vitro model of bradyzoite stage parasites. Current studies are underway to advance the triazine nitrile series into an in vivo model of latent Toxoplasma infection.</p><!><p>All reagents were used without further purification as received from commercial sources unless noted otherwise. 1H NMR spectra were taken in DMSO-d6, MeOD, or CDCl3 at room temperature on Varian Inova 400 MHz or Varian Inova 500 MHz instruments. Reported chemical shifts for the 1H NMR and 13C NMR spectra were recorded in parts per million (ppm) on the δ scale from an internal standard of residual tetramethylsilane (0 ppm). Mass spectrometry data were obtained on either a Micromass LCT or Agilent Q-TOF. An Agilent 1100 series HPLC with an Agilent Zorbax Eclipse Plus–C18 column was used to determine purity of biologically tested compounds. Unless otherwise noted, all tested compounds were determined to be >95% pure using a 6 minute gradient of 10–90% acetonitrile in water followed by a 2 minute hold at 90% acetonitrile with detection at 254 nm. Flash chromatographic purifications were performed using a Teledyne ISCO Combiflash RF with Redisep Gold RF columns.</p><!><p>To a dry round bottom flask with DCM was added aryl aldehyde or aryl ketone Int-1a (1 eq), followed by the addition of the desired primary amine (1 eq), final reaction concentration of 0.1–0.3 mM. The vessel was then stirred under a nitrogen atmosphere at room temperature overnight. Sodium triacetoxyborohydride (3 eq) was then added and reaction was stirred at room temperature for 1–3 hr. The reaction was quenched slowly with water then poured into water and extracted 3x with DCM. The combined organic layer was dried over sodium sulfate and concentrated in-vacuo. The crude product was further purified by column chromatography (0–10% MeOH in DCM + 0.1% TEA) to afford the desired secondary amine Int-2a.</p><!><p>Under a dry nitrogen atmosphere, cyanuric chloride (1 eq) was added a roundbottom flask along with DCM (final conc. = 0.1–0.3 mM). The vessel was cooled to −10°C (Ice/Brine) and the desired secondary amine (1 eq) and DIPEA (1 eq) were then added. Reaction was stirred at −10°C for 1 h. The reaction was poured into water and extracted 3x with DCM. Combined organic layers were dried over MgSO4 and concentrated in vacuo. The crude product was further purified by column chromatography (0–100% EtOAc in Hexanes gradient) to afford the desired intermediate Int-3a.</p><!><p>Under a dry nitrogen atmosphere, dichloro triazine intermediate (1 eq) was added a roundbottom flask along with DCM to a final reaction concentration of 0.1 mM. The vessel was cooled to −10°C (Ice/Brine) and DIPEA (1 eq) and morpholine (1 eq) were then added. Reaction was stirred and allowed to warm to room temperature over 6–12 h. The reaction was poured into water and extracted 3x with DCM. Combined organic layers were dried over MgSO4 and concentrated in vacuo. The crude product was further purified by column chromatography (0–100% EtOAc in Hexanes gradient) to afford the desired intermediate Int-4a.</p><!><p>A dry pressure vessel was charged with the appropriate triazinyl chloride (1.0 eq), and suspended in DMSO/H2O 9:1. KCN (1.1 eq.) and 1,4-diazabicyclo[2.2.2]octane (DABCO, 2.0 eq.) were added and the reaction was heated to 80°C. Reaction was left to stir at this temperature until LC/MS showed completion of the reaction (6–12 h). Reaction was cooled to room temperature, diluted with EtOAc, and washed thoroughly with brine (3–5 x). The organic layer was separated, dried over MgSO4, filtered, and evaporated. The crude product was further purified by column chromatography (0–100% EtOAc in Hexanes gradient) to afford the desired product.</p><!><p>Under a dry nitrogen atmosphere, cyanuric chloride (1 eq) was added a roundbottom flask along with DCM (final conc. = 0.1–0.3 mM). The vessel was cooled to −10°C (Ice/Brine) and DIPEA (1 eq) followed by the desired primary amine Int-1b (1eq) were then added. Reaction was stirred at for 1–3 hr, until TLC/HPLC indicated substitution was complete. Next, DIPEA (1 eq) was added followed by morpholine (1 eq) and reaction was allowed to stir and warm to rt over 4–12h until TLC/HPLC indicated substitution was complete. The reaction was poured into water and extracted 3x with DCM. Combined organic layers were dried over MgSO4 and concentrated in vacuo. The crude product was further purified by column chromatography (0–100% EtOAc in Hexanes gradient) to afford the desired intermediate Int-2b.</p><!><p>A dry pressure vessel was charged with the appropriate triazinyl chloride (1.0 eq), and suspended in DMSO/H2O 9:1. KCN (1.1 eq) and 1,4-diazabicyclo[2.2.2]octane (DABCO, 2.0 eq) were added and the reaction was heated to 80°C. Reaction was left to stir at this temperature until LC/MS showed completion of the reaction (6–12 h). Reaction was cooled to room temperature, diluted with EtOAc and washed thoroughly with brine (3–5 x). The organic layer was separated, dried over MgSO4, filtered, and evaporated. The crude product was further purified by column chromatography (0–100% EtOAc in Hexanes gradient) to afford the desired secondary amine Int-3b.</p><!><p>The triazinyl-nitrile (1 eq) was dissolved in DMF and cooled to −10°C (Ice/Brine bath). 60% NaH in mineral oil (1–1.5 eq) was added and reaction was stirred for 30 min. The appropriate alkyl or aryl Br/Cl/Ms/Ts (1–2 eq) was then added. The reaction was allowed to slowly warm to room temperature and left to stir (2–12 hr) until HPLC indicated reaction was complete. The solution was poured into EtOAc, washed 3x with brine, dried over NaSO4, and concentrated. Crude residue was purified by flash chromatography EtOAc:Hexanes 0–100% gradient to give the pure tertiary amine Int-4b.</p><!><p>The Boc-protected amine (1 eq) was dissolved into a 1:2 mixture of TFA:DCM (~0.1–0.3 mM) and allowed to stir at room temperature for 0.5–2h, until HPLC or TLC indicated reaction was complete. Solvent was removed in vacuo and compound was purified by reverse phase chromatography (C18, 10–100% ACN in water + 0.1% TFA gradient) and concentrated to afford desired product as the TFA salt.</p><!><p>To construct a model for TgCatL-triazine complex, the X-ray structure of TgCPL in complex with its propeptide (PDB: 3F75) was superposed with the X-ray structure of HsCPL (PDB: 5MAJ) using The Molecular Operating Environment (MOE), version 2008.10, Chemical Computing Group Inc., Montreal, Quebec, Canada. Then, the HsCPL enzyme from the 5MAJ and the propeptide of TgCPL from the 3F75 were removed and the TgCPL enzyme and the triazine compound are saved as a model for TgCPL-triazine complex. All the water molecules from the 3F75 and water molecules outside the binding site from 5MAJ were removed. Covalent docking of the triazine compound in the TgCPL was performed with induced fit option instead of rigid receptor for refinement. The covalent docking protocol in MOE was developed by creating a reaction profile for this particular virtual reaction and saving in the MarvinSketch 17.14.0 (http://www.chemaxon.com).</p><!><p>Compound potency and selectivity was evaluated in vitro for both TgCPL and HsCPL activity in a fluorescence-based assay by monitoring the hydrolysis of Cbz-Leu-Arg-aminomethylcoumarin (Z-L-R-AMC). Protein was expressed and purified as described previously.25 The substrate hydrolysis results in the release of fluorescent 7-amino-4-methylcoumarin (AMC) that can be monitored spectrophotometrically with linear kinetics for up to 60 min. Inhibitors were serial diluted in DMSO in a 1-to-3 dilution, spanning at least 10 concentration points in either duplicate or triplicate. The enzyme was pre-incubated with the inhibitory compounds for 5 min at 23°C, followed by the addition of the AMC substrate. The assay final conditions had a total volume of 200 μL, consisting of 90 μL enzyme (conc.=50 ng/mL), 100 μL Z-L-R-AMC substrate (conc.= 80 μM), and 10 μL inhibitor or DMSO. The relative fluorescence of the AMC generation is measured over the course of 5 min (Excitation: 380 nm, Emission: 460 nm). LHVS (1) and DMSO were used as positive and negative controls, respectively. All dose response data was obtained with at least three independent replicates (n=3). Graphpad Prism software was used to visualize inhibition curves and calculate IC50 values from the reaction mean ν. Assay final concentrations in each well: 40 μM ZLR-AMC, TgCPL or HsCPL (final conc= 0.0225 ng/μL). Assay Buffer: 100 mM NaAc, 2 mM EDTA, 900 mM NaCl, 50 mM DTT. Substrate: Cbz-Leu-Arg-AMC (Bachem, purchased as HCl salt).</p><!><p>We utilized compound libraries provided by the University of Michigan Center for Chemical Genomics (CCG) to screen for inhibitors against TgCPL. The HTS primary screen was conducted on ~150,000 candidates, representing a wide diversity of structures and sources. The libraries were composed of small molecule compounds and clinically tested bioactive drugs obtained from a variety of sources include ChemBridge, ChemDiv, Maybridge, MicroSource, TimTec, NCI, and the NIH Clinical Collection. Reactions of the aforementioned Z-Leu-Arg-AMC assay were conducted in a 384-well plate at the scale of 10 μL per well. Each reaction was initiated by the addition of 50 nL of 2 mM stocks of the compounds to each well using a Caliper Sciclone ALH3000 with a V&P pintool, resulting in a final concentration of 10 μM for each compound. A Thermo Multidrop Combi Liquid Dispenser was used to add 5 μL of recombinant TgCPL (0.5 μg/mL) and 5 μL of Z-Leu-Arg-AMC (80 μM), both diluted in activity assay buffer. After 10 min incubation at RT, 10 μL of E64 (100 μM) diluted in in assay buffer was added with the multidrop reagent dispenser. A Perkin Elmer Envision Multimode Plate Reader measured the fluorescence after 30 min at RT with excitation at 355 and the emission at 460 nm. Active compounds were triaged using computational and experimental results to eliminate compounds likely to interfere with the assay or to present development issues. Computational evaluation of primary screen hits was first conducted in light of toxicity, cysteine reactivity, promiscuity, and inhibitor level. Primary screen hits available after this attribute triage were then subjected to confirmation screen to compose the experimental evaluation. Assay conditions from the primary screen were repeated, but conducted in triplicate for each candidate (using a TTP Labtech Mosquito X1 for compound addition). Confirmed hits were retested to establish a concentration response curve of activity in a secondary screen. Assay conditions were repeated on the 384-well plates, and conducted in duplicate for each compound. Compounds found active after the concentration response secondary screen were prioritized based on chemical properties, and given appropriate rankings by consulting medicinal chemists. High and medium ranked hits were prioritized, and fresh powders of the library stocks were obtained from Sigma Aldrich. The available compounds were reconstituted in DMSO to 10 μM. All assay conditions of the concentration response screen were repeated for the repurchased compounds against both TgCPL and HsCPL.</p><!><p>The metabolic stability was assessed using CD-1 mouse liver microsomes. One micromolar of each compound was incubated with 0.5 mg/mL microsomes and 1.7 mM cofactor β-NADPH in 0.1 M phosphate buffer (pH = 7.4) containing 3.3 mM MgCl2 at 37°C. The DMSO concentration was less than 0.1% in the final incubation system. At 0, 5, 10, 15, 30, 45, and 60 min of incubation, 40 μL of reaction mixture were taken out, and the reaction is quenched by adding 3-fold excess of cold acetonitrile containing 100 ng/mL of internal standard for quantification. The collected fractions were centrifuged at 15000 rpm for 10 min to collect the supernatant for LC–MS/ MS analysis, from which the amount of compound remaining was determined. The natural log of the amount of compound remaining was plotted against time to determine the disappearance rate and the half-life of tested compounds.</p><!><p>All animal experiments in this study were approved by the University of Michigan Committee on Use and Care of Animals and Unit for Laboratory Animal Medicine. The abbreviated pharmacokinetics for compounds was determined in female CD-1 mice following intraperitoneal (ip) injection at 10 mg/kg. Compound was dissolved in the vehicle containing 15% (v/v) DMSO, 15–20% (v/v) PEG-400, and 70% (v/ v) PBS. Four blood samples (50 μL) were collected over 7 h (at 0.5h, 2h, 4h, and 7h), centrifuged at 3500 rpm for 10 min, and plasma was frozen at −80°C for later analysis. Plasma concentrations of the compounds were determined by the LC–MS/MS method developed and validated for this study. The LC–MS/MS method consisted of a Shimadzu HPLC system, and chromatographic separation of tested compound which was achieved using a Waters Xbridge-C18 column (5 cm × 2.1 mm, 3.5 μm). An AB Sciex QTrap 4500 mass spectrometer equipped with an electrospray ionization source (ABI-Sciex, Toronto, Canada) in the positive-ion multiple reaction monitoring (MRM) mode was used for detection. All pharmacokinetic parameters were calculated by non-compartmental methods using WinNonlin, version 3.2 (Pharsight Corporation, Mountain View, CA, USA).</p><!><p>12-well transwell plates were seeded with MDCKII-MDR1 cells (0.5 million/well) and cultured for 24h. Cells were washed with DMEM 3 times (both sides). 0.5 ml of 1μM test compound in DMEM was added to apical side (for A to B measurement) or basolateral side (for B to A) (donor side) and 0.5ml of DMEM+0.1% DMSO to the receiving side. Cells were incubated for 4h and 2 × 200μL was sampled from the receiving side and stored at −20°C for future use. For calibration standard, compound standards were dissolved in DMSO then further diluted in acetonitrile to a concentration of 10 μg/mL. Calibration standards were prepared from this stock with internal standard (5 nM CE302). Samples were prepared by protein precipitation from media followed by addition of internal standard. Samples were analyzed by LC–MS/MS. The LC–MS/MS method consisted of a Shimadzu HPLC system, and chromatographic separation of tested compound which was achieved using a Waters Xbridge-C18 column (5 cm × 2.1 mm, 3.5 μm). An AB Sciex QTrap 4500 mass spectrometer equipped with an electrospray ionization source (ABI-Sciex, Toronto, Canada) in the positive-ion multiple reaction monitoring (MRM) mode was used for detection.</p><!><p>Bradyzoite viability was assessed by combining plaque assay and quantitative polymerase chain reaction (qPCR) analysis of genome number, as previously performed by Di Cristina et al, 2017.26 Briefly, human foreskin fibroblast (HFF) cells were infected with T. gondii tachyzoites in six-well plates. Following host cell invasion by parasites, tachyzoites underwent differentiation to bradyzoites by being maintained at 37°C/0% CO2 in alkaline media (RPMI 1540 w/o NaHCO3, 50 mM HEPES, 3% FBS, Pen/Strep, pH 8.2), resulting in the generation of in vitro tissue cysts. Differentiation was carried out over the course of seven days, replacing the alkaline media daily. Samples were subsequently treated with the test compounds for seven days, replacing them daily. A DMSO-treated vehicle control sample was also incorporated into each assay. Following the treatment period, the culture media in each well was replaced with 2 mL HBSS and cysts were liberated from the infected HFF monolayers by mechanical extrusion, by lifting cells with a cell scraper and syringing several times through 26G needles. Then, 2 mL of pre-warmed 2× pepsin solution (0.026% pepsin in 170 mM NaCl and 60 mM HCl, final concentration) was added to each sample and sample were left to incubate at 37°C for 30 min. Reactions were stopped by adding 94 mM Na2CO3, removing the supernatant after centrifugation at 1,500g for 10 min at room temperature and re-suspending pepsin-treated parasites in 1 mL of DMEM without serum. Parasites were enumerated and 1000 parasites per well were added to six-well plates containing confluent monolayers of HFFs in D10 media, in triplicate. To allow for the formation of plaques, parasites were left to grow undisturbed for 12 days. After this period, the number of plaques in each well was determined by counting plaques with the use of a light microscope. Five hundred microlitres of the initial 1 mL of pepsin-treated parasites was used for genomic DNA purification, performed using the DNeasy Blood & Tissue Kit (Qiagen). Genomic DNA was eluted in a final volume of 100 μL. To determine the number of parasite genomes per microliter, 10 μL of each gDNA sample was analysed by qPCR, in duplicate, using the tubulin primers TUB2.RT.F and TUB2.RT.R. Quantitative PCR analyses were performed using Brilliant II SYBR Green QPCR Master Mix (Agilent) and a Stratagene Mx3000PQ-PCR machine. The number of plaques that formed per genome was then calculated and expressed as a percentage of control (DMSO) treated parasites.</p><!><p>MTS assays were performed to evaluate the cytotoxicity of test compounds on HFF cells in vitro. HFF cells were plated onto 96-well plates and allowed to become fully confluent at 37°C/5% CO2, prior to the addition of the test compounds. Test compounds were diluted to 100 μM in conversion media (the same culture media used in bradyzoite viability assays) and applied to HFF cells in triplicate. Subsequently, test compounds underwent two-fold serial dilution across the plate to a concentration of 0.195 μM. Plates were incubated for a further 7 days at 37°C/5% CO2, replacing the media daily to apply fresh compounds. Following the 7-day treatment, 20 μl of CellTiter 96 Aqueous One Solution reagent (Promega) was added to each well and left to incubate for 2 hours at 37°C. Absorbance was then measured at 490nm using a Synergy H1 Hybrid Multi-Mode Reader (BioTek). MTS assays were performed in triplicate. Untreated HFF cells were used as a negative control. HFF cell viability was calculated as a percentage by dividing the measured A490 of test samples by the mean A490 of the untreated control samples. The 50% toxicity dosage (TD50) of each test compound was established by plotting the % HFF cell viability against compound concentration and reading the concentration at 50% HFF cell viability.</p>

PubMed Author Manuscript

Advances in the Development of Shape Similarity Methods and Their Application in Drug Discovery

Molecular similarity is a key concept in drug discovery. It is based on the assumption that structurally similar molecules frequently have similar properties. Assessment of similarity between small molecules has been highly effective in the discovery and development of various drugs. Especially, two-dimensional (2D) similarity approaches have been quite popular due to their simplicity, accuracy and efficiency. Recently, the focus has been shifted toward the development of methods involving the representation and comparison of three-dimensional (3D) conformation of small molecules. Among the 3D similarity methods, evaluation of shape similarity is now gaining attention for its application not only in virtual screening but also in molecular target prediction, drug repurposing and scaffold hopping. A wide range of methods have been developed to describe molecular shape and to determine the shape similarity between small molecules. The most widely used methods include atom distance-based methods, surface-based approaches such as spherical harmonics and 3D Zernike descriptors, atom-centered Gaussian overlay based representations. Several of these methods demonstrated excellent virtual screening performance not only retrospectively but also prospectively. In addition to methods assessing the similarity between small molecules, shape similarity approaches have been developed to compare shapes of protein structures and binding pockets. Additionally, shape comparisons between atomic models and 3D density maps allowed the fitting of atomic models into cryo-electron microscopy maps. This review aims to summarize the methodological advances in shape similarity assessment highlighting advantages, disadvantages and their application in drug discovery.

advances_in_the_development_of_shape_similarity_methods_and_their_application_in_drug_discovery

Introduction<!>3D shape similarity methods<!><!>Atomic distance-based descriptors<!><!>Atomic distance-based descriptors<!><!>Atomic distance-based descriptors<!>Atom-centered gaussian-based shape similarity methods<!><!>Atom-centered gaussian-based shape similarity methods<!>Surface based 3D shape similarity comparison methods<!><!>Surface based 3D shape similarity comparison methods<!><!>Surface based 3D shape similarity comparison methods<!>Other shape similarity approaches<!>Application in virtual screening<!>Applications in protein structure comparison<!>Applications in fitting of atomic models into cryo-electron microscopy maps<!>Conclusion and future directions<!>Author contributions<!>Conflict of interest statement

<p>Molecular similarity is a key concept in drug discovery and has been routinely used in the discovery and design of new molecules. It is based on the notion that two molecules often share similar physical properties and biological function if they are structurally similar. This similarity principle has been widely utilized in early phases of drug development to discover new molecules. Virtual screening has been used to filter large databases of compounds to a smaller number based on this similarity principle. Molecular similarity has been also employed to optimize the potency and pharmacokinetic properties of lead compounds based on their structure–activity relationships.</p><p>There are two components of molecular similarity analysis (1) structural representations and (2) quantitative measurements of similarity between two structural representations. Many types of structural representations have been suggested to measure the similarity between two molecules. These include physiochemical properties, topological indices, molecular graphs, pharmacophore features, molecular shapes, molecular fields etc. Further, there are various methods to quantify the similarity between two structural representations, e.g., Tanimoto coefficient, Dice index, cosine coefficient, Euclidean distance, Tversky index etc. Among these, Tanimoto coefficient (Rogers and Tanimoto, 1960) is the most popular and widely used similarity measure. Based on the structural representation, molecular similarity approaches can be broadly classified into 2D or 3D similarity methods. The 2D similarity methods rely only on the 2D structural information and are among the fastest, efficient and most popular similarity search methods. Moreover, they do not rely on structural alignments for estimating the similarity between two molecules. These methods include substructure search, fingerprint similarity search and 2D descriptor-based methods. However, most of these methods are limited in their ability to enable scaffold hopping and provide no structural and mechanistic insights. To deal with the limitations associated with 2D similarity methods, several approaches were developed that account for 3D conformations of a molecule while performing similarity search. These methods include pharmacophore modeling, shape similarity, molecular field-based methods, 3D fingerprints among others. In recent years, ligand 3D shape-based similarity analysis has become a method of choice in increasing number of virtual screening campaigns. Several successful applications of shape similarity to discover new molecules have been published in the literature. The major advantage with shape-based virtual screening methods is that scaffold hopping can be conveniently accomplished and scaffolds other than the query can be identified.</p><p>In this review, we will summarize the development and application of various 3D shape similarity methods and will comment on their utility in drug discovery. We will first outline the classification and various types of 3D shape similarity methods highlighting their advantages and disadvantages. Later, we will describe various applications of 3D shape similarity methods in drug discovery.</p><!><p>The 3D shape has been widely recognized as a key determinant for the activity of small molecules and other biomolecules (Zauhar et al., 2003; Rush et al., 2005; Schnecke and Boström, 2006; Kortagere et al., 2009). The shape complementarity between ligand and receptor is necessary for bringing the receptor and ligand sufficiently close to each other so they can form critical interactions necessary for binding. Two molecules with similar shape are likely to fit the same binding pocket and thereby exhibiting similar biological activity. Shape comparison methods could be broadly classified as (1) Alignment-free or non-superposition methods and (2) Alignment or superposition-based methods. Both of these methods have their own advantages and disadvantages. Alignment-free methods are independent of the position and orientations of molecules. As such, they are much faster and could be used to screen large compound databases. Alignment-based methods rely on finding the optimal superposition between the compounds. Alignment-based methods are highly effective in identifying shape similarities among the molecular structures but they are computationally expensive. These methods enable comparison of the surface properties such as hydrophobicity and polarity. Visualization is one of the advantages with the alignment-based methods and the similarity between two molecules can be displayed. This information is useful in the design of new molecules and to guide further optimization. However, a subpar molecular alignment may lead to errors in comparing two molecules. Apart from this broad classification, shape similarity methods could be classified based on the underlying representation of molecular shape. The similarity between these shape representations is evaluated by employing various similarity metrics. A schematic overview of the similarity calculation between a query and database molecules is given in Figure 1. In the following paragraphs, we will outline commonly utilized shape representations with their advantages and disadvantages. As this review is targeted toward a broader readership, we will only provide an overview of the methods. For algorithmic details and mathematics behind each method, original publications may be referred.</p><!><p>A schematic overview of similarity calculation between a query and database molecules.</p><!><p>These methods are based on the assumption that the shape of a molecule can be described by the relative positions of its atoms. The similarity between molecules can be then calculated by comparing the corresponding distributions of atomic distances. As these descriptors only require the computation of interatomic distances in compounds, these methods are faster compared to other shape comparison methodologies. Additionally, these methods do not require the alignment between two molecules for shape comparison. An overview of various atomic distance-based methods is given in Table 1 highlighting their availability as well as their advantages and disadvantages. One of the earlier atomic distance-based shape comparison method was based on atom triplet distances (Bemis and Kuntz, 1992). This method considered each molecule as a collection of three atom sub-molecules. The atom triplet triangle perimeters were used to generate shape histograms which were then utilized to compare the shape of molecules. This method however has a few limitations. It is difficult to select bin size suitable for all molecules. Each molecule typically generates 300–500 atom triplets and storing them require large space especially when comparing a large database of molecules. To deal with this limitation, another atom triplet based molecular shape comparison method was developed where a 2,048 bits long single condensed triplet shape signature was employed to represent the entire set of triplets in each molecule (Nilakantan et al., 1993). A signature of the query molecule is first compared with the already stored signatures of database molecules. Then only the compounds with adequately similar signatures are compared in detail by generating all triplets. Although this method was efficient but there was a risk of missing similar compounds due to the use of highly reduced signature representation. Another group developed molecular descriptors based on atom triplet triangles, angular information from surface point normal and local curvature to facilitate shape comparisons (Good et al., 1995). However, these descriptors have limited discriminating power and require large disk space for storage.</p><!><p>Atomic distance based shape comparison methods.</p><!><p>Ultrafast shape recognition (USR) (Ballester and Richards, 2007a,b; Ballester, 2011) is possibly the most popular atomic distance-based method developed to overcome alignment and speed problems associated with shape similarity methods. This method also uses the relative positions of atoms to describe the shape of a molecule. The schematic overview of USR method is given in Figure 2 along with an example of the shape similarity evaluation. USR calculates the distribution of all atom distances from four reference positions: the molecular centroid (ctd), the closest atom to molecular centroid (cst), the farthest atom from molecular centroid (fct) and the atom farthest away from fct (ftf). Consecutively, the first three statistical moments (mean, variance, and skewness of distribution) are calculated from each of these distributions. Hence, each molecule has a vector of twelve descriptors to describe its 3D shape. Finally, the similarity between shapes of two molecules is calculated through an inverse of the Manhattan distance of these 12 values:</p><p>where Mq and Mi are vectors of shape descriptors for query and ith molecule, respectively. The performance of USR was retrospectively compared with EigenSpectrum Shape Fingerprints (EShape3D) where better mean enrichment for USR was observed (Ballester et al., 2009). A retrospective comparison with three state-of-the-art shape similarity methods: EShape3D, shape signatures and ROCS revealed that USR is 1,546, 2,038, and 14,238 times faster than each one of them respectively (Ballester and Richards, 2007a). A web implementation of USR (USR-VS) is an extremely fast way of carrying out shape similarity calculations (Li et al., 2016). USR-VS is capable of screening 55 million 3D conformers per second and can calculate similarity scores for 94 million 3D conformers in about 2 s. This extremely fast speed is achieved as the features for all 3D conformers are preloaded into the memory. Moreover, the multi-threaded design of the webserver and alignment-free nature of USR method also contributed to such a high computational efficiency. A hardware implementation of USR has been shown to achieve two-fold speed gains over standard CPU based implementation of USR (Morro et al., 2018). In this implementation, a computing technique, Spiking Neural Networks, has been adapted utilizing Field-Programmable Gate arrays to allow highly parallelized implementation of USR. Prospective application of USR in the identification of arylamine N-acetyltransferases, protein arginine deiminase 4 (PAD4), falcipain 2, phosphatases of regenerating liver (PRL-3), p53-MDM2 inhibitors and for phenotypic targets such as colon cancer cell lines established the real-world applicability of USR (Li et al., 2009; Ballester et al., 2010, 2012; Teo et al., 2013; Hoeger et al., 2014; Patil et al., 2014). As USR is an ultrafast, purely shape-based similarity method, several methods augmenting the original USR capabilities were developed. These include a method where USR was combined with MACCS key encoding the topological information of small molecules (Cannon et al., 2008). To clearly distinguish between enantiomers, methods complementing USR with optical isomerism descriptors were developed (Armstrong et al., 2009; Zhou et al., 2010). Electroshape, a USR variant appended partial charge and atomic lipophilicity (alogP) as additional molecular properties to account for electrostatics and lipophilicity along with shape recognition (Armstrong et al., 2010, 2011). A web implementation of Electroshape is available at SwissSimilarity (Zoete et al., 2016). AutoCorrelation of Partial Charges (ACPC) also utilized partial charges with atomic distances to measure similarity between two molecules (Berenger et al., 2014). The method uses an autocorrelation function and a point charge model to encode all atoms of a molecule into two vectors that are rotation translation invariant. Another implementation of USR method is Ultrafast Shape Recognition with Atom Types (UFSRAT) which introduced pharmacophoric constraints to USR by incorporating atom type information (Shave, 2010; Lim et al., 2011; Shave et al., 2015). UFSRAT is capable of very fast comparison of query molecule with small molecule libraries from several major chemical vendors via its webserver (Table 1). Application of UFSRAT method in the discovery of MDM2, PRL-3, FK506-Binding Protein 12, kynurenine 3-monooxygenase and 11β-hydroxysteroid dehydrogenase type 1 (11βHSD1) inhibitors demonstrated its utility in key areas of drug discovery such as cancer, Alzheimer's disease, inflammation and type-II diabetes. (Hoeger et al., 2014; Houston et al., 2015; Shave et al., 2015, 2018). Another similar implementation, USRCAT utilized CREDO atom types to encode pharmacophoric information to USR (Schreyer and Blundell, 2009, 2012). USRCAT not only retained USR abilities to retrieve hits with low structural similarity but also demonstrated improved performance over the original USR implementation.</p><!><p>(A) An overview of USR shape representation. In USR approach, the shape of a molecule is described by the distribution of atomic distance to four reference points. (B) An example of shape similarity calculation between two small molecules utilizing the USR approach.</p><!><p>Atomic distance or descriptor-based methods are widely used due to their ability to quickly compare the shapes of query molecules with large small molecule libraries. A fast comparison of a wide range of chemical space increases the chances of finding novel hits. These methods are not only computationally efficient but also have produced excellent hit rates as revealed from several successful prospective studies against a wide range of molecular and non-molecular targets. Moreover, they are also capable of retrieving chemical scaffolds which are different from the query molecule, thus allowing scaffold hopping. As atomic distance-based shape similarity approaches are alignment-free, the visual inspection of shape similarity may be sometimes challenging especially for molecules which have low structural similarity. Selection of the right query compound is a key component of atomic distance-based shape similarity methods and their performance depends on optimal query selection. Hit rate can be improved by employing multiple queries and increasing the diversity of selected hits. Moreover, clustering based on shape similarity could be utilized to understand how different chemotypes arrange in binding pockets and thereby generating consensus queries (Pérez-Nueno et al., 2008; Pérez-Nueno and Ritchie, 2011) to improve virtual screening performance and reducing false positives.</p><!><p>Among many methods of describing the molecular shape of a molecule, hard sphere (Connolly, 1985; Masek et al., 1993) and Gaussian sphere (Grant and Pickup, 1995; Grant et al., 1996) are two most widely adopted models. Both of these models describe the shape in terms of the volume of a molecule. Two molecules will possess similar shape if they have similar volume. Hard sphere model represents a molecule by a set of merged spheres where each sphere serves as an atom with its van der Waals radius. The volume of a molecule can be calculated by a formula that describes the union of a number of sets and their intersection. Although the analytical expression of the volume and its derivatives is reported in the original publication (Masek et al., 1993), it is not easy to implement as the formulas become very complicated with increasing number of intersections. Gaussian sphere model (Grant and Pickup, 1995, 1997; Grant et al., 1996) represents a molecule using a set of overlapping Gaussian spheres and measures the integral volume over all overlapping Gaussians. In this model, each intersection is expressed as the integral of a set of overlapping atom-centered Gaussian spheres and the volume of a molecule is described based on the inclusion-exclusion principle. Analytical expression for the volume calculation is given in the original publication which describes highly accurate volume calculation up to sixth order intersections (Grant and Pickup, 1995). The authors also proposed comparing shapes of two molecules by numerically optimizing the overlap between two molecules (Grant et al., 1996).</p><p>Several methods based on Gaussian overlays were developed to measure the shape similarity between two molecules. An overview of these methods is presented in Table 2. Among these, Rapid Overlay of Chemical Structures (ROCS) is undoubtedly the most widely used method that utilizes Gaussian functions to measure the shape similarity between two molecules (Rush et al., 2005; Hawkins et al., 2007). ROCS algorithm is based on the original Gaussian overlay approach that finds and quantifies the maximum volume overlap between two molecules (Grant and Pickup, 1995; Grant et al., 1996). An overview of ROCS shape similarity calculation is given in Figure 3. However, to improve the efficiency of volume overlap calculations, it incorporated several modifications to the original implementation. ROCS ignores hydrogens for the volume calculations and uses equal radii for all heavy atoms. Furthermore, ROCS utilizes only the first order terms of shape density function. ROCS employs Tanimoto (Rogers and Tanimoto, 1960) and Tversky (Tversky, 1977) correlation coefficients as similarity metrics to calculate the overlap between two molecules which are defined as:</p><p>where Oa, b is the volume overlap between molecules a and b, Oa is the volume of molecule a and Ob is the volume of molecule b. α and β are parameters for Tversky index. ROCS also considers chemical complementarity by including the chemical features to improve shape-based superposition. ROCS has been successfully employed in various drug discovery campaigns such as in the identification of small molecules inhibitors (Kumar et al., 2014b), to scaffold hop from one chemical class to another (Kumar et al., 2016), to rescore docking generated poses (Kumar and Zhang, 2016a) and to predict binding poses and ranking of inhibitors (Kumar and Zhang, 2016b,c). ROCS can routinely perform shape and chemical feature comparisons of about 600–800 conformers per second on a modern CPU. Although this speed is reasonable for alignment-based shape similarity methods, it takes several hours to screen a moderately sized virtual screening library. To facilitate large scale shape comparison, e.g., to screen large small molecule libraries within minutes, FastROCS (https://www.eyesopen.com/molecular-modeling-fastrocs), a GPU implementation of ROCS has been developed that increased the shape comparison speed by about three orders of magnitude over its CPU implementation. FastROCS is capable of processing up to a million conformers per second on a single NVIDIA Tesla K20 GPU (https://docs.eyesopen.com/toolkits/python/fastrocstk/architecture.html). PAPER, an open source GPU implementation of ROCS algorithm, also demonstrated speed acceleration up to two orders of magnitude on an NVIDIA GeForce GTX 280 GPU over its open source CPU implementation on a Intel Xeon E5345 CPU (Haque and Pande, 2010). MolShaCS is another method that engages Gaussian description of shape to evaluate molecular similarity between two molecules (Vaz de Lima and Nascimento, 2013). In addition to shape, MolShaCS utilizes Gaussian description of charge distribution to optimize overlays and similarity computations using Hodgkin's index (Hodgkin and Richards, 1987; Good et al., 1992). It was able to process 21 compounds per second, which seems to be a quite impressive speed for computers of that time. As Gaussian overlay based methods require precise alignment for the calculation of shape similarity, several groups employed approaches such as pharmacophore and field based methods to generate initial alignment. SHAFTS (SHApe-FeaTure Similarity) (Liu et al., 2011) adopted pharmacophoric point triplets and least square fitting to generate initial alignment. A weighted sum of pharmacophoric fit and volume overlap was then used to assess shape similarities. Phase Shape (Sastry et al., 2011) also employed the same concept of atom distribution triplets to generate initial alignments which were then refined by maximizing the volume overlap. Phase Shape is capable of performing shape comparisons of about 500 conformers per second. Reminiscent of Shape and Electrostatic Potential (ShaEP) (Vainio et al., 2009) also resembles SHAFTS and Phase Shape as it utilizes a hybrid approach that combined field-based methods with volumetric methods to estimate molecular similarity. ShaEP borrowed a graph matching algorithm to generate initial superposition. Molecular graphs represented shape and electrostatic potential at points close to molecular surface. The method then optimized the initial alignment by maximizing the volume overlap calculated through Gaussian functions. Another similar method, SimG (Cai et al., 2013), adopted downhill simplex method (Nelder and Mead, 1965) to evaluate the similarity in shape and chemical features of a molecule and a binding pocket or ligand. SimG shape similarity method possessed advantage over other methods described here in the sense that it is capable of performing shape similarity evaluations between a ligand and a binding pocket. SABRE method (Hamza et al., 2012, 2013) introduced two modifications to the original Gaussian overlay based shape similarity implementation. First, it utilized reduced chemical structures by removing the functional group not present in query to generate initial alignments. Reduced chemical structures were subsequently replaced by full structures and the initial alignments were refined by rigid-body translation and rotation using steepest descent to produce shape density overlap with the query. Secondly, to avoid bias for large sized ligands when using Tanimoto similarity metric, a new scoring function Hamza–Wei–Zhan (HWZ) score was developed. An extension to SABRE method enabled its utility in chemogenomics area (Wei and Hamza, 2014). Shapelets (Proschak et al., 2008) is unlike any other Gaussian overlay based shape comparison method. It describes the shape of a molecule by decomposing its surface into discrete patches. This 3D graph representation can then be used for either full or partial shape similarity evaluations.</p><!><p>An overview of commonly used Gaussian overlay based shape comparison methods.</p><p>An overview of the shape similarity calculation by ROCS program.</p><!><p>In most Gaussian function based overlay methods shape density of a molecule is described as the sum of shapes of individual atoms which sometimes results in the overestimation of the volume, for example, in molecules where some atoms highly overlap with others in the vicinity. Weighted Gaussian algorithm (WEGA) method (Yan et al., 2013) puts forward a modification where a weight factor is introduced for every atom. This weight factor reflects the crowdedness of an atom with its neighbors. The shape density of a molecule is represented by the linear combination of weighted atomic Gaussian functions. Utilizing this modification, WEGA method demonstrated improved shape similarity and virtual screening performance. The speed of WEGA shape similarity calculations varies with the size of query and database compounds. For an average drug-like query, WEGA can process 1,000–1,500 conformations per second (Yan et al., 2013). A GPU implementation of this method (gWEGA) has also been developed that reported a virtual screening speed increase by two orders of magnitude on one NVIDIA Tesla C2050 GPU over its CPU implementation on a quad-core Intel Xeon X3520 CPU (Yan et al., 2014). Another WEGA derivative, HybridSim proposed a hybrid metric combining 2D fingerprints with WEGA shape similarity and demonstrated improved virtual screening performance over standalone 2D fingerprint and shape similarity methods (Shang et al., 2017).</p><p>Overall, atom-centered Gaussian-based shape similarity methods present many advantages over other shape similarity methods. Although not as fast as distance based methods, these methods are fast enough for large scale virtual screenings. The major advantage with atom-centered Gaussian-based shape similarity methods is the visualization. The visualization of shape similarity between two molecules is immensely helpful in deriving the structure activity relationship for the optimization and for scaffold hopping. A majority of these methods address the problem of ligand flexibility by utilizing conformational ensemble. However, in some cases it may not be trivial to sample all possible conformations, e.g., natural products. Moreover, several top performing conformational generation methods face difficulty in modeling the correct conformation of some molecules, e.g., macrocycles, peptidomimetics etc. Another limitation with these methods is that their performance highly depends upon the query molecule and choosing the right query is a critical component of a shape-based virtual screening campaign (Kirchmair et al., 2009). Despite these limitations, atom-centered Gaussian overlay based methods are the most widely used shape similarity methods. They have provided many successful examples demonstrating their utility in various areas of drug discovery which will be discussed later in this manuscript.</p><!><p>Molecular surface is another way of depicting the shape of a molecule. Comparison of molecular surfaces based on their shapes can reveal similarity in their physical and biological properties. There are many ways to describe the surface of a molecule. Precise definitions such as surface based on quantum mechanical wave functions are not practical especially for large molecules (Mezey, 2007). Surface definitions such as solvent-accessible surface (Lee and Richards, 1971; Connolly, 1983) and van der Waals surface are more practical and much easier to calculate. Some studies employed alpha shapes (Edelsbrunner et al., 1983; Edelsbrunner and Mücke, 1994; Edelsbrunner, 1995) which is a coarse representation of Connolly surface (Connolly, 1983) to describe the shape of a molecule (Wilson et al., 2009). Alpha shapes of a set of points "S" are generalization of convex hull and utilize a parameter, α to describe the shape with varying levels of details. For large α values, the alpha shape is equivalent to convex hull and shape feature details such as concavities and voids started to appear with decrease in α value. The alpha shape method has been applied to represent and compare shapes of 3D molecules (Wilson et al., 2009).</p><p>Shape signatures or shape histograms offer another representation of molecular shape that can be used to explore 3D volume of a molecule confined by the solvent accessible surface (Zauhar et al., 2003; Meek et al., 2006). Shape signatures are probability distribution histograms borrowed from a computer graphics technique, ray-tracing. In this method, a ray is initiated within a molecule bound by its solvent accessible surface. Propagation of a ray trace inside of the triangulated solvent accessible surface is recorded as probability distribution histograms. The histograms for query and any other molecule can be easily compared using the following metrics:</p><p>where 1D represents the probability distribution of ray-trace lengths only while 2D represents ray-trace lengths in combination with additional molecular property such as electrostatic potential. Shape signature encodes shape, molecular size and surface charge distribution of a molecule and can be utilized to compare the histogram of a query molecule with the pre-generated histograms of small molecule libraries. The utility of shape signatures as a virtual screening approach has been demonstrated in several studies (Nagarajan et al., 2005; Wang et al., 2006; Hartman et al., 2009; Ai et al., 2014; Werner et al., 2014). As shape signature based similarity comparisons are fast and do not require the alignment of molecules, they are capable of screening millions of molecules in a short time. In addition to shape similarity, shape signatures also allow shape complementarity comparisons against a receptor binding pocket. Although shape similarity calculations with shape signature have been effectively used in many inhibitor discovery efforts, the high number of false positives is a concern especially for large and complex queries. To cope with these drawbacks, a few modifications to the original methods were reported. These include fragment-based shape signature (FBSS) (Zauhar et al., 2013) and inner distance shape signature (IDSS) (Liu et al., 2009, 2012). FBSS involves the generation and comparison of shape signatures for fragments in the molecules. IDSS utilizes inner distance which is the shortest path between landmark points within the molecular shape. IDSS has been shown to be especially useful in case of flexible molecules as it is insensitive to shape deformation of flexible molecules.</p><p>Several methods employed local surface shape similarity to align and estimate the similarity between molecules. One such method applied subgraph isomorphism to molecular surface comparison (Cosgrove et al., 2000). In this method, molecular surface was represented by patches of the same shape. Alignment between two molecules was obtained by using a clique-detection algorithm to obtain overlapping patches. Quadratic shape descriptors (Goldman and Wipke, 2000) exploited a similar concept where molecular surface was divided into a series of patches. Each patch was represented by geometrically invariant descriptors such as the normal, the shape index and the principle curvatures which were then used to identify similar patches. SURFCOMP (Hofbauer et al., 2004) further applied several filters such as surrounding shape and physicochemical properties to identify corresponding patches on surfaces of two molecules (Table 3).</p><!><p>An overview and availability of a few surface-based shape comparison methods.</p><!><p>Spherical harmonics (SH) based representations which are expansion of SH functions also allow quantitative description of molecular shapes (Max and Getzoff, 1988). In this representation, shapes are expressed as functions on a unit sphere. Each point on a unit sphere surface is described by its spherical coordinates (r,θ,ϕ) and setting f (θ,ϕ) = r, where r is a radial function encoding the distance of surface points from a chosen origin. This function can be determined by deriving an expansion of SH basis function given by:</p><p>where Ylm(θ, ϕ) is the SH basis function for degree l and order m. cl, m are coefficients of SH function. L is the chosen limit to get desired resolution of the surface. The number of terms in the function depends upon this limit as a value of L, which yields (L+1)2 terms. In general, SH are not rotation translation invariant as magnitude of cl, m change based on the rotation of r(θ, ϕ). Hence, prior alignment is necessary before comparing the shape of molecules. Efforts were also made to make SH rotation translation invariant (Kazhdan et al., 2003; Mak et al., 2008), however, these modifications increase the number of terms thereby increasing the complexity of SH.</p><p>About two decades ago, it was shown that SH functions could be applied to estimate the 3D molecular similarity between two macromolecules (Ritchie and Kemp, 1999). Since then, it has been successfully applied in virtual screening (Cai et al., 2002; Mavridis et al., 2007), protein structure comparisons (Tao et al., 2005; Gramada and Bourne, 2006), protein-ligand docking (Ritchie and Kemp, 2000; Lin and Clark, 2005; Yamagishi et al., 2006), binding pocket similarity comparison (Morris et al., 2005) etc. Additionally, several groups utilized variations of SH to compare the shapes of small molecules. The first implementation of SH to compare shapes of small molecules opened the way for many applications ranging from virtual screening to quantitative structure-activity relationship (QSAR) model building (Lin and Clark, 2005). SpotLight program utilizes SH to superpose and classify small molecules (Mavridis et al., 2007). To enable high throughput virtual screening, the vector interpretation of SH coefficients was used to construct rotation translation invariant fingerprints (RIFs) which were compared using a distance score (Mavridis et al., 2007). In this method, rotational invariance was gained by binning together the SH coefficients of the same order. This method was later developed as ParaFit (http://www.ceposinsilico.de) (Table 3). In another study, SH based molecular surface was decomposed and the norm of decomposition coefficients were used to describe the molecular shape (Wang et al., 2011). Norms of decomposition coefficients are partially rotation translation invariant enabling large scale comparison. The performance of this method was retrospectively demonstrated and was also prospectively applied in the discovery of cyclooxygenase-1 and cyclooxygenase-2 inhibitors. SHeMS method utilizes genetic algorithm to optimize the weights of SH expansion coefficients for a reference set (Cai et al., 2012). Through optimization of weights, SHeMS demonstrated improved performance over original SH implementation and USR method. To facilitate measurement of similarity between sets of compounds, many shape similarity methods were complemented with physicochemical properties. Harmonic pharma chemistry coefficient (HPCC) method combined SH shape representation with pharmacophoric features (Karaboga et al., 2013). In HPCC method, SH surfaces are discretized as triangle meshes which are assigned pharmacophoric features. Tanimoto similarity for both shape and pharmacophore features is calculated separately between query and test molecules. A combo score is finally calculated by adding Tanimoto scores for shape and chemical overlay. HPCC method demonstrated improved performance for the combo approach over utilizing the shape alone.</p><p>In several studies, 3D-Zernike descriptors (3DZD) (Novotni and Klein, 2003), which are the extension of SH were employed to compare the shapes of molecules and cryoEM maps (Figure 4 and Table 3). 3DZD differs from SH in terms of their mathematical description. 3DZD can model molecular shape precisely as compared to SH which can only model single valued or star-shape surfaces. They are rotation translation invariant, whereas SH depends on the orientation of the molecule. Although rotation translation invariant SH descriptors have been developed (Kazhdan et al., 2003), the number of terms are much higher in SH descriptors. 3DZD is also suitable to represent other properties on molecular surfaces such as hydrophobicity and electrostatic potential (Sael et al., 2008a). In the drug discovery area, 3DZD was initially applied to compare shapes of protein molecules (Sael et al., 2008b; Figure 4A). Later, the concept was extended to measuring shape similarity and small molecules (Venkatraman et al., 2009a) and between binding pockets (Kihara et al., 2009; Venkatraman et al., 2009b; Figures 4B,C). In 3DZD method, 3D Zernike function is described as:</p><p>where Ylm(θ, ϕ) is the SH basis function while Rnl(r) is the radial function. Zernike moments are calculated using the following equation:</p><p>As Zernike moments are not rotationally invariant, so to make them rotation translation invariant, they are expressed as norm Fnlm which is known as 3DZD. Shape similarity between two molecules based on 3DZD is compared using the following metrics:</p><p>Ligand 3D shape similarity comparison using 3DZD is fast and rotation translation invariant. As no alignment step is required for comparison, it can be utilized as a virtual screening tool to filter a database of compounds based on shape similarity with a query molecule.</p><!><p>Application of 3D Zernike descriptors in (A) protein protein similarity (B) small molecule similarity (C) protein ligand complementarity and (D) comparison of cryoEM maps.</p><!><p>Overall, surface-based shape similarity methods present attractive options for comparing the shapes of small molecules and macromolecules. They were quite successful in estimating the global and local similarities between macromolecules. However, most of these methods are still in infancy as far as small molecule shape comparison is concerned. Several reasons may have contributed to the lack of interest from researchers in accepting these methods as small molecule shape comparison tools. Surface-based methods such as SH and 3DZD are mathematically complex and involve inclusion of many terms to fully capture the shape of a molecule. Moreover, they are slow in comparison to atomic distance-based shape description and comparison methods while their accuracy in retrieving compounds similar in shape to a query does not match Gaussian overlay-based shape similarity methods. Further, while these methods capture very well the global shape of a molecule, the local shape similarity is not represented comprehensively which is very critical in comparing the shapes of small molecules. However, these methods present several new areas of shape comparison such as comparing shape of ligands with that of binding pockets which may be of immense utility for structure-based design.</p><!><p>There are many other approaches of shape representation and methods of similarity measurement in addition to these described above. Another way of representing molecular shape is to use molecular descriptors. Several shape-based descriptors have been traditionally used to compare small molecules and develop QSAR models. These descriptors mostly represent shape implicitly with other properties such as size, symmetry and atom distribution. These include Weighted Holistic Invariant Molecular (WHIM) descriptors of shape (Gramatica, 2006), shape indices, descriptors for moments of the distribution of molecular volume (Mansfield et al., 2002). Most of molecular descriptors are alignment independent, however, some such as moments of the distribution of molecular volume require superposition of molecules. Comparative Molecular-Field Analysis (CoMFA) (Cramer et al., 1988) is a widely used technique to develop QSAR models and understand SAR for a series of compounds. CoMFA compares a set of molecules by placing them on a grid and calculating potential energy fields. The differences and similarities between molecules are then correlated with differences and similarities in their biological activities. As CoMFA requires molecules to be pre-aligned, the 3D shape similarity of molecules can be obtained based on potential energy fields. A modification of CoMFA approach, Comparative Molecular Moment Analysis (CoMMA) calculates geometric moments from the center of mass, center of charge and center of dipole of a molecule (Silverman and Platt, 1996). However, superposition of molecules is not required in this approach. Shape of the molecules can also be inferred from structural descriptors such as molecular quantum numbers (MQNs) (Nguyen et al., 2009; van Deursen et al., 2010). The MQN represents counts for 42 structural features such as atom, ring and bond types, polar groups and topology. MQN system has been used to effectively classify and visualize large libraries of organic molecules such as ZINC, GDB, and PubChem.</p><p>Volumetric aligned molecular shapes (VAMS) method (Koes and Camacho, 2014) uses data structures to represent and compare shapes of 3D molecules. It applies inclusive and exclusive shape constraints to estimate the similarity in shapes of 3D molecules. In VAMS method, the shape of a molecule is represented by solvent-excluded volume calculated from its heavy atoms using a water probe of radius 1.4 Å. Volume is discretized on a grid of 0.5 Å resolution where each point on the grid represents a Voxel or 3D pixel. An oct-tree data structure is used to store voxelized volume. This method requires all the shapes to be pre-aligned to a standard reference coordinates. The conformations of the molecule are aligned using the moment of inertia of heavy atoms. Voxelized shapes are compared using Tanimoto similarity (Rogers and Tanimoto, 1960) where the ratio of number of voxels common in two shapes and number of voxels present in either of the shapes is measured. The performance of VAMS method as a standalone virtual screening tool is not better than many other shape similarity methods, e.g., ROCS, however, VAMS is reasonably fast and could perform a million shape comparisons in about 10 s. Hence, it may be used as a pre-filtering tool for other shape similarity methods. Fragment oriented molecular shape (FOMS) is the extension of VAMS method, where shapes are aligned using fragments (Hain et al., 2016).</p><!><p>Shape similarity attempts to quantify the resemblance between two molecules utilizing several descriptions of molecular shape as described previously. This approach has been successfully utilized as a virtual screening tool to identify molecules similar to a given query from the library of chemicals. Several retrospective studies have been published demonstrating the utility of shape based similarity methods over 2D and other 3D similarity methods (Nagarajan et al., 2005; Renner and Schneider, 2006; Ballester et al., 2009; Giganti et al., 2010; Venkatraman et al., 2010; Ballester, 2011; Hu et al., 2012, 2016). Several studies also presented computational approaches to improve the performance and efficiency of shape comparison methods. One study recommended the selection of a suitable query and incorporation of chemical information such as pharmacophoric features of the query molecule to improve the performance of shape-based virtual screening (Kirchmair et al., 2009). Another study demonstrated that the application of a machine learning method, Support Vector Machine (SVM), to shape comparisons can significantly improve virtual screening efficiency (Sato et al., 2012). The need of automation was further suggested specially to carry out multiple query searches which ensure a diverse hit list (Kalászi et al., 2014).</p><p>Apart from retrospective tests, many prospective applications of shape similarity have been published in the literature. In numerous studies, it was employed as the only virtual screening approach to filter and prioritize compounds from a large library to a number small enough for biological testing (Rush et al., 2005; Boström et al., 2007; Freitas et al., 2008; Ballester et al., 2010, 2012; Kumar et al., 2012; Vasudevan et al., 2012; Sun et al., 2013; Hoeger et al., 2014; Patil et al., 2014; Temml et al., 2014; Chen et al., 2016; Bassetto et al., 2017). Among these studies, the shape based identification of a compound active on colon cancer cell line is quite interesting (Patil et al., 2014). This study employed USR to screen a database of approved drugs. The top virtual screening hit displayed dose dependent inhibition of a colon cancer cell line. This study not only repurposed a known drug but also demonstrated the applicability of shape similarity methods for phenotypic screens, e.g., anti-bacterial or anti-fungal drug discovery where molecular target is often unknown. This is especially important considering the fact that most approved drugs come from phenotypic screens (Swinney and Anthony, 2011). In other investigations, it was combined with other ligand-based virtual screening methods or structure based approaches such as molecular docking. Among ligand-based approaches, shape similarity was frequently used in combination with electrostatic similarity. As electrostatic comparison between two small molecules requires precise alignment between them, shape matching was first performed and then followed by the electrostatic potential similarity calculations. This hierarchical combination was utilized to discover a wide variety of binders including enzyme inhibitors (Hevener et al., 2011), mRNA binders (Kaoud et al., 2012), chemical probes (Naylor et al., 2009), protein-protein interaction inhibitors (Boström et al., 2013), SUMO activating enzyme 1 inhibitors (Kumar et al., 2016), and Aurora kinase A inhibitors (Kong et al., 2018).</p><p>Although shape-based approaches demonstrated considerable success in ligand-based virtual screening studies, the true potential of the method was realized when it was combined with structure based methods in a hierarchical manner or in a parallel manner. To effectively use shape based virtual screening, several groups employed hierarchical virtual screening (Kumar and Zhang, 2015) where it was coupled with molecular docking. As shape matching calculations are comparatively faster than structure based virtual screening methods, it is generally used during initials steps of a hierarchical virtual screening protocol. This hierarchical combination of shape similarity with molecular docking has been successfully employed in the discovery of type II dehydroquinase inhibitors (Ballester et al., 2012) and that of MDM2 inhibitors (Houston et al., 2015), 11β-hydroxysteroid dehydrogenase 1 inhibitors (Xia et al., 2011), PPARγ partial agonists (Vidović et al., 2011), inhibitors of chemokine receptor 5 (CCR5)-N terminus binding to gp120 protein (Acharya et al., 2011), Grb7-based antitumor agents (Ambaye et al., 2013), fungal trihydroxynaphthalene reductase inhibitors (Brunskole Švegelj et al., 2011), non-steroidal FXR ligands (Fu et al., 2012; Wang et al., 2015), novel SIRT3 scaffolds (Salo et al., 2013), protein kinase CK2 inhibitors (Sun et al., 2013), SUMO conjugating enzyme inhibitors (Kumar et al., 2014a), and chemokine receptor type 4 inhibitors (Das et al., 2015). Combination of shape similarity methods with structure-based methods such as docking provide several advantages. Ultrafast shape comparison methods such as USR can very quickly filter large libraries for compounds that are similarly shaped as known inhibitors. Hence, the time required for docking could be drastically reduced by eliminating compounds that doesn't fit in the binding pocket. Moreover, in case of some proteins the inhibitor activity is driven by key moieties in compounds, e.g., metal binding groups in case of metalloproteins, reactive functional groups in cysteine proteases, hinge binding groups in kinases etc. In these scenarios, docking will help in the prioritization of compounds based on the interactions they make with the binding pocket. Sometimes the difference in shape similarity scores for compounds is very small and it is challenging to cherry pick for biological assay. Here, docking of shape similarity hits could also help in the prioritization of compounds for purchase or chemical synthesis. However, the combination of shape similarity with molecular docking is not always advantageous especially for proteins with highly flexible binding pockets, multiple pocket conformations or homology models where accurate docking is challenging. A virtual screening scheme where USR hits were re-ranked using Autodock-Vina score produced no active hits as docking was performed in a quite different pocket conformation (Hoeger et al., 2014). In another study, shape-based virtual screening alone produced better hit rates than hierarchical combination of shape similarity and docking methods (Ballester et al., 2012). In numerous studies, shape similarity calculations along with molecular docking were complemented with other approaches such as 2D similarity search, pharmacophore modeling, electrostatic potential matching, machine learning and MM-PBSA method (Mochalkin et al., 2009; Alcaro et al., 2013; Poongavanam and Kongsted, 2013; Wiggers et al., 2013; Hamza et al., 2014a; Kumar et al., 2014b; Pala et al., 2014; Feng et al., 2015; Corso et al., 2016; Mangiatordi et al., 2017; Xia et al., 2017). The use of different virtual screening approaches in parallel has been previously suggested as different methods tend to identify different set of compounds and virtual screening hit rates could be improved by employing them in parallel manner (Sheridan and Kearsley, 2002). In parallel virtual screening, several methods are run independently and the top hits from each method is selected. Parallel combination of various ligand and structure based methods with shape similarity approaches was found to be productive especially in case of challenging targets (Swann et al., 2011; Langdon et al., 2013; Hoeger et al., 2014). A parallel virtual screening to identify inhibitors of PRL-3 employing several ligand and structure-based methods against the same screening library produced contrasting hit rates for different approaches (Hoeger et al., 2014). Many prospective applications suggest the utility of hierarchical or parallel combination of shape similarity approaches with other ligand and structure-based methods. However, no benchmark study demonstrating their utility has been published. A systematic study will help researchers to identify areas where the combination of several approaches will be better than employing shape based virtual screening methods alone.</p><p>One application of shape similarity methods is to hop from one chemical scaffold to another in order to improve the potency, selectivity, physicochemical properties and to create novel intellectual property positions (Hu et al., 2017). Shape similarity methods are capable of identifying several scaffolds which are structurally different from the query compounds and each scaffold may be pursued separately. Scaffold hopping is highly effective in rescuing the problematic leads that cannot be pursued further due to problems in selectivity, pharmacology and pharmacokinetics. Both atomic distance-based and Gaussian-overlay shape similarity methods can effectively perform scaffold hopping as exemplified from several prospective studies. Among the first prospective application of shape similarity based methods in scaffold hopping, small molecule inhibitors of ZipA-FtsZ protein-protein interaction were identified (Rush et al., 2005). Some recent scaffold hopping applications include the identification of inhibitors of arylamine N-acetyltransferases (Ballester et al., 2010), type II dehydroquinase inhibitors (Ballester et al., 2012) sumoylation enzymes (Kumar et al., 2014b, 2016), anti-tubercular agents (Hamza et al., 2014b; Wavhale et al., 2017), anti-tumor agents (Ge et al., 2014), 11βHSD1 inhibitors (Shave et al., 2015), leucine zipper kinase inhibitors (Patel et al., 2015), kynurenine 3-monooxygenase inhibitors (Shave et al., 2018), and partial agonist of inositol trisphosphate receptor (Vasudevan et al., 2014). In addition to prospective application, rigorous benchmarking of shape similarity methods for their scaffold hopping capabilities is important. However, systematic benchmarking is challenging due to disagreement on the definition of scaffold. In one retrospective study, the scaffold hopping potential of atomic distance-based shape similarity method USRCAT has been demonstrated utilizing DUD-E dataset (Schreyer and Blundell, 2012). For the tested benchmark dataset, USRCAT was capable of identifying structurally dissimilar active hits that could not be retrieved by utilizing topological similarities. Shape similarity was also used to repurpose existing drugs for previously unknown activity (Vasudevan et al., 2012). Another application is in silico target fishing or the identification of protein targets of orphan chemical compounds. In one recent research, the target of anti-fungal macrocycle amidinoureas was identified following a shape similarity screening (Maccari et al., 2017). The representative structure from a series of macrocycle amidinoureas was used as a query to obtain most similar crystallographic ligand from all solved crystal structures. A prioritized list of targets based on similarity score and subsequent docking and enzymatic assay revealed Trichoderma viride chitinase as target of this class of compounds. Along the same line, retrospective studies showed that the combination of molecular shape and chemical structure similarity can reliably achieve biological target prediction (Abdulhameed et al., 2012; Gfeller et al., 2013). Additionally, shape similarity comparison based on spherical harmonics surface representation has been demonstrated that it can be used to predict drug promiscuity (Perez-Nueno et al., 2011). Furthermore, shape similarity comparisons could also be used to predict subtype selectivity of ligands (Kuang et al., 2016).</p><p>One important application of shape similarity methods in drug discovery is the clustering of known inhibitors of a protein target. As the performance of most shape-based methods highly depend on the selection of right query for the virtual screening (Kirchmair et al., 2009), special attention was paid toward the development of methods dealing with this problem. It has been reported that clustering of known inhibitors based on their shapes could help the identification of optimal query for virtual screening (Pérez-Nueno and Ritchie, 2011). Clustering of spherical harmonics-based consensus shapes assisted in the identification of ligands that bind to different regions in the binding pocket of some protein targets such as CCR5 (Pérez-Nueno et al., 2008). Further, the clustering of molecular shapes also helped in the identification of promiscuous protein targets and ligands (Pérez-Nueno and Ritchie, 2011). Selection and use of high quality compound libraries is an important aspect of high throughput screening (HTS). However, testing a large number of compounds is not economically viable. In silico, mostly 2D similarity based, methods are commonly employed to generate a subset or focused set for HTS (Huggins et al., 2011; Dandapani et al., 2012). The limitation with 2D similarity methods is that they ignore inherent property such as the shape of a molecule. Use of shape-based clustering of large compound libraries for creating quality HTS library present several advantages. Clustering of molecular libraries based on atomic distance-based methods such as USR can achieve similar or significantly better computational efficiency as 2D fingerprint-based methods. Moreover, it will ensure maximum diversity with less number of compounds in HTS library.</p><p>Apart from employing ligand 3D shape similarity as a virtual screening method, several groups adopted it to improve the performance of other virtual screening methods. Molecular docking is one such method widely used in drug discovery. Although there has been significant progress in the development of molecular docking methods, challenges still remain both in sampling and scoring of binding poses within protein binding pockets. In the last few years, several methods were developed that utilized ligand 3D shape similarity to improve both sampling and scoring performance of molecular docking. The shape overlap with known crystallographic ligands for the target protein was utilized to guide ligand conformational sampling toward critical regions of protein binding site (Wu and Vieth, 2004). Other methods used shape similarity based alignment for the selection of reliable poses among many docking generated poses (Fukunishi and Nakamura, 2008, 2012; Anighoro and Bajorath, 2016; Kumar and Zhang, 2016a). Ligand 3D shape similarity was also a key component of many pose prediction methods where shape similarity with existing ligand bound crystal structures was utilized to predict binding poses of unknown ligands (Kelley et al., 2015; Huang et al., 2016; Kumar and Zhang, 2016b,c). Several of these methods demonstrated excellent retrospective and prospective performance. Moreover, shape similarity also facilitated the improvement in scoring and rank-ordering performance of a docking method. Several methods have reported improved virtual screening performance of a docking method when shape overlap with crystallographic ligands was employed to select the best binding pose of ligands in a screening library (Roy et al., 2015; Anighoro and Bajorath, 2016). Consideration of protein flexibility in molecular docking is a challenging problem and several methods have been developed to tackle it (B-Rao et al., 2009). Among these, receptor ensemble based methods demonstrated reasonable performance (Bottegoni et al., 2011) where the receptor ensemble is selected either from many crystallographic structures or from those generated by in silico methods such as molecular dynamics simulation. It has been shown previously that the selection of receptor ensemble based on binding pocket shape similarity is an effective way of considering receptor flexibility in molecular docking (Osguthorpe et al., 2012). Further, one method suggested utilizing a single suitable receptor for each ligand in a screening library instead of docking all compounds to multiple receptor structures (Kumar and Zhang, 2018). It was also shown that single suitable receptor selection based on ligand 3D shape similarity is superior to 2D similarity based selection.</p><!><p>Evaluation of structural similarity between protein structures has many applications including but not limited to classification of protein structures, evolutionary relationship between protein structures, identification of templates for homology modeling, functional annotation, protein-protein interactions etc. Conventional methods for protein structure comparison are based on the alignment of protein atoms or residues. These methods require extensive rotational and translational sampling thereby limiting their utility for large scale protein structure comparisons. Several methods have been developed that utilize shape similarity to detect global or local similarity between protein structures. Classification of these methods also follows the previously described classification including Gaussian overlay based methods, surface-based methods using spherical harmonic descriptors, 3D Zernike descriptors etc. Among these, surface-based methods were developed previously to measure similarity between protein structures. Only later they were applied to the small molecule area. Several methods of protein structure comparison employed SH to represent shapes of protein structures (Tao et al., 2005; Gramada and Bourne, 2006; Konarev et al., 2016). Like SH, 3D Zernike based moments are also suitable to compare shapes of protein structures (Sael et al., 2008b; Figure 4A). Not only they were suitable to estimate the similarity between two proteins but also their rotation-translation invariant nature allows fast real-time search of similar proteins in structural databases such as PDB (La et al., 2009; Kihara et al., 2011; Xiong et al., 2014). A Gaussian mixture model based protein shape similarity method (Kawabata, 2008) also allows large scale comparisons of proteins with data from PDB and EMDB. This method has been implemented as Omokage search in PDB Japan (Suzuki et al., 2016; Kinjo et al., 2017). The server compares global shapes of proteins and results are obtained reasonably fast within 1 min after submission of a query. Large scale comparison of protein structures based on shape is useful in functional annotation, selection of templates for comparative modeling etc. An application of shape comparison method to protein classification has also been reported (Daras et al., 2006).</p><p>One important application of shape matching is the evaluation of similarity between protein binding pockets. This field is especially interesting as sequence and structural alignments are often not useful when comparing binding pockets of proteins with different folds. As protein binding pockets are much more conserved than protein structures (Gao and Skolnick, 2013), a reliable comparison between protein binding pockets is crucial for predicting protein functions, polypharmacology of ligands and for drug repurposing. Numerous methods based on distinct structural representations as described previously were developed in the last decade. One such method employed spherical harmonics to represent and compare the shapes of protein binding pockets (Morris et al., 2005). This method was later extended to compare the shape of protein binding pockets with that of binding ligands (Kahraman et al., 2007). PocketMatch compares two binding pockets based on the sorted list of distances that captured chemical nature and 3D shape of the binding pocket (Yeturu and Chandra, 2008). Another method based on property-encoded shape distributions (PESD) combines the concept of shape distributions with the chemical environment of the binding pocket surface to effectively capture binding pocket similarities (Das et al., 2009). Pocket-Surfer utilizes pseudo-Zernike descriptors and 3D Zernike descriptors to represent and compare properties and 3D shapes of binding pockets (Chikhi et al., 2010). An extension of this method, Patch-Surfer searches local similarity by representing a binding pocket as amalgamation of segmented surface patches which are described by properties such as shape, electrostatic potential, concaveness and hydrophobicity (Sael and Kihara, 2012). Similarity between protein cavities was also measured by representing the pockets by pharmacophoric grid points and aligning them by optimizing their volume overlap (Desaphy et al., 2012).</p><p>Concept of pocket similarity was also extended to complementarity between binding pockets and ligands. This gave rise to a new virtual screening methodology based on shape complementarity between binding pockets and ligands. PL-Patch-Surfer2 program evaluates the compatibility between ligand and binding pocket by measuring the complementarity between ligand surface and local surface patches in the binding pocket (Shin et al., 2016a,b; Figure 4C). The program utilizes 3DZD to represent molecular shape while physicochemical properties are also mapped onto the surface. The method was evaluated on benchmark datasets and revealed better performance than two docking programs. Spherical harmonics expansion coefficients have also been employed in the approximation and comparison of binding pockets and ligand surfaces (Cai et al., 2002). The complementarity was demonstrated utilizing 35 protein-ligand complexes. Elekit adopted shape and electrostatic complementarity concept to discover small molecule inhibitors of protein-protein interactions (Voet et al., 2013). Elekit assesses the similarity between small molecules and protein ligands of a receptor protein based on the electrostatic potential values stored on a 3D grid.</p><!><p>Recent developments in cryo-electron microscopy (cryo-EM) has helped researchers to overcome resolution barrier and provide structural and mechanistic insights into structures of difficult proteins and large protein assemblies. Most of these improvements came from the advances in sample preparation, electron detector technologies, improved microscope and computational data processing. Computational methods played an important part in particle picking, particle reconstruction, building and fitting of structures into cryo-EM maps. In recent years, several methods were developed to improve building, fitting and refinement of protein structures in cryo-EM maps (Esquivel-Rodríguez and Kihara, 2013). Among these methods, a few methods employed shape similarity to fit atomic structures of protein subunits into the cryo-EM maps of multi-subunit proteins. One method, Gaussian Mixture macromolecule FITting (gmfit), utilizes Gaussian mixture models (GMM) to represent the shape of cryo-EM maps and atomic models (Kawabata, 2008). GMMs are probability distribution functions obtained by joining many 3D Gaussian functions. Initially, both the cryo-EM map and atomic models are first converted into GMM followed by the fitting of a single subunit GMM into the GMM of protein complex using random and gradient based local search. Finally, the fit between atomic models and cryo-EM map is obtained based on the position and orientation of GMM. This method is reasonably fast and can fit multiple subunits with reasonable accuracy. PDB Japan (https://pdbj.org) has implemented this method in its EM navigator utility to provide shape based structural similarity search against protein databases (Kinjo et al., 2017). Another method adopted a surface-based approach where 3DZD was used to represent and compare isosurface derived from low resolution cryo-EM maps of protein structures (Sael and Kihara, 2010; Figure 4D). It was demonstrated that 3DZD can distinguish proteins of different folds even at low resolution of 15 Å. A web-based platform for comparing cryo-EM maps was also developed by the same group (Esquivel-Rodríguez et al., 2015; Han et al., 2017). A similar method utilized 3D Zernike moments to search a database of protein structures for matching protein structures to a cryo-EM map (Yin and Dokholyan, 2011). EMLZerD method also utilized 3DZD to fit multiple structures in a cryo-EM map (Esquivel-Rodríguez and Kihara, 2012). The method generates hundreds of putative configurations of subunit arrangement using a protein-protein docking method. These configurations were later compared with a cryo-EM map using 3DZD and Euclidean distance. The biggest advantage of 3D Zernike moments methods is that they are rotation translation invariant and no computational expensive step of rigid body or flexible structural alignment is required. Moreover, these methods enable screening of proteins from structural databases such as PDB to find out models that can fit into a cryo-EM map.</p><!><p>3D shape similarity methods have contributed immensely to the overall acceptance of the computational virtual screening methods in drug discovery. Most shape similarity methods for shape comparison of small molecules and macromolecules took inspiration from the approaches developed to compare the shapes of 3D objects in computational geometry field. Several approaches were developed ranging from extremely fast atom distance-based methods to comparatively slower mathematically complex methods such as SH and 3DZD. Among all the 3D shape comparison methods, atomic distance-based and Gaussian overlay-based methods are the most widely used. These approaches possess several advantages over surface-based methods. Atomic distance-based methods present an extremely fast way of quickly comparing the shapes of small molecules. This has facilitated the screening of very large libraries of millions of compounds within a few seconds. Moreover, screening large libraries increased the probability of finding novel chemical scaffolds. Furthermore, as most of these methods depend on shape rather than the underlying chemical structure, scaffold hopping can be conveniently achieved. Another possible application of these fast shape similarity evaluation methods would be the clustering of large chemical space to generate quality shape diverse HTS screening libraries. Although Gaussian overlay-based methods are slower than atomic-distance based methods, they are fast enough to allow high throughput virtual screening. GPU implementations of these methods is not very difficult as exemplified by the development of several GPU compatible programs such as FastROCS, PAPER, gWEGA etc. resulting in further increase in the processing speeds. Another advantage with Gaussian-based methods is that they allow visualization as they require alignment of molecules prior to shape similarity calculations. Visualization is helpful in understanding the features responsible for biological activity and critical for the optimization of a molecule especially for the molecules with low structural similarity with query compound. However, a suboptimal alignment can lead to errors in volume overlap calculations and thereby affecting similarity scores and visualization. As alignment is the key component of Gaussian overlay methods, efforts should be focused toward improving molecular alignment. Some of these methods employ chemical features to refine global overlays. As alignment is global optimization problem, molecular alignment could also be improved by employing fast local optimization methods. Both atomic distance-based and Gaussian overlay-based shape similarity methods handle ligand flexibility by employing the conformational ensemble. The performance thus indirectly depends upon conformation generation methods. Current state-of-the-art conformation generation methods still struggle to generate near-native conformations of ligands such as peptidomimetics, macrocycles etc. Development of novel conformation generation approaches utilizing knowledge from experimental databases such as CSD and PDB will steer improvement in performance of shape-based virtual screening approaches. Surface based methods such as SH expansion coefficients and 3DZD are suitable for comparing macromolecules and atomic models with electron density maps, however, comparatively less efforts have been made toward utilizing them in small molecule area. One advantage with surface-based methods is that the protein ligand complementarity search is possible by comparing enclosed shapes of binding pockets and ligands. This will be handy in cases where ligand-based virtual screening methods could not be used due to the lack of active compounds. Finally, shape-based similarity could be used in combination with other ligand and structure-based approaches either in hierarchical or parallel manner to improve hit rate especially for difficult targets.</p><!><p>All authors listed have made a substantial, direct and intellectual contribution to the work, and approved it for publication.</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. The reviewer XL and handling Editor declared their shared affiliation.</p>

PubMed Open Access

Targeting Cancer Stem Cells for Chemoprevention of Pancreatic Cancer

Pancreatic ductal adenocarcinoma is one of the deadliest cancers worldwide and the fourth leading cause of cancer-related deaths in United States. Regardless of the advances in molecular pathogenesis and consequential efforts to suppress the disease, this cancer remains a major health problem in United States. By 2030, the projection is that pancreatic cancer will be climb up to be the second leading cause of cancer-related deaths in the United States. Pancreatic cancer is a rapidly invasive and highly metastatic cancer, and does not respond to standard therapies. Emerging evidence supports that the presence of a unique population of cells called cancer stem cells (CSCs) as potential cancer inducing cells and efforts are underway to develop therapeutic strategies targeting these cells. CSCs are rare quiescent cells, and with the capacity to self-renew through asymmetric/symmetric cell division, as well as differentiate into various lineages of cells in the cancer. Studies have been shown that CSCs are highly resistant to standard therapy and also responsible for drug resistance, cancer recurrence and metastasis. To overcome this problem, we need novel preventive agents that target these CSCs. Natural compounds or phytochemicals have ability to target these CSCs and their signaling pathways. Therefore, in the present review article, we summarize our current understanding of pancreatic CSCs and their signaling pathways, and the phytochemicals that target these cells including curcumin, resveratrol, tea polyphenol EGCG (epigallocatechin-3-gallate), crocetinic acid, sulforaphane, genistein, indole-3-carbinol, vitamin E \xce\xb4-tocotrienol, Plumbagin, quercetin, triptolide, Licofelene and Quinomycin. These natural compounds or phytochemicals, which inhibit cancer stem cells may prove to be promising agents for the prevention and treatment of pancreatic cancers.

targeting_cancer_stem_cells_for_chemoprevention_of_pancreatic_cancer

INTRODUCTION<!>PANCREATIC CANCER STEM CELLS<!>CANCER STEM CELL SIGNALING PATHWAYS AND CHEMOPREVENTION<!>WNT SIGNALING<!>HEDGEHOG SIGNALING<!>NOTCH SIGNALING<!>HIPPO SIGNALING<!>JAK-STAT PATHWAY<!>PI3K/Akt/mTOR SIGNALING<!>MAPK-ERK PATHWAY<!>Chemo-preventive Agents and Pancreatic Cancer Stem Cells<!>CONCLUSION AND FUTURE DIRECTION

<p>Pancreatic ductal adenocarcinoma continues to one of the deadliest cancers in the world. It is the fourth leading cause of cancer-related deaths in United States with the highest mortality rate among all cancers. With 5-year survival rate is <6%. The American Cancer Society has estimated that 43,090 Americans (22,300 men and 20,790 women) would die of pancreatic cancer in 2017 and 53,670 new cases for 2017 (27, 970 men and 25,700 women)[1]. In both men and woman, lifetime risk for developing pancreatic cancer is about 1 in 65 (1.5%). By 2030, it is predicted that pancreatic cancer will be the second leading cause of cancer-related deaths in United States [2]. Regardless of advances in molecular pathogenesis, pancreatic cancer has major unsolved health problem in the United States because of its drug resistance and susceptibility to metastasis [3, 4]. Moreover, delineation of several germline or acquired genetic mutations and the most common being K-Ras (90%), CDK2NA (90%), TP53 (75%–90%), SMAD4/DPC4 (50%), along with genomic and epigenetic alterations, also played an important role in poor prognosis of this disease. These mutations can direct us to focus on the precision medicine. In addition, tumor microenvironment, the chemo-resistant cancer stem cells, and the desmoplastic stroma have been the target for recent promising clinical investigations [5]. Previously, Gemcitabine was the treatment option for metastatic pancreatic cancer. Currently, two combination regimens for metastatic disease have been used for gold standard: 5-fluorouracil (5-FU)/leucovorin with irinotecan and oxaliplatin (FOLFIRINOX)[6, 7] and nabpaclitaxel with gemcitabine [8]. With these approaches, response rates range between 23% and 31%, progression-free survival rates are 5.5–6.6 months, and overall survival is between 8.5 and 11 months [8]. At this time, FOLFIRINOX and gemcitabine/nab-paclitaxel is being used in studies for metastatic disease, in both adjuvant and neoadjuvant setting, and also for the treatment of locally advanced but inoperable pancreatic cancers.</p><p>However, the response rate to current chemotherapy is below 31% [8]. The extent of this problem mandates the need for novel preventive/therapeutic agents. Studies have suggested that CSCs may significantly influence drug resistance as well as metastasis [9]. The purpose of this review is to summarize our current understanding of cancer stem cells, highlighting recent advances and analyzing the preventing/therapeutic potential of targeting pancreatic cancer.</p><!><p>Emerging evidence suggests that CSCs characterize a subset of cancer cells with distinct stemness features that permit them to drive tumorigenesis and metastasis [10, 11]. Moreover, studies have proven that CSCs have resistant to the current chemotherapy and radiation, which renders them a primary source for tumor recurrences after or even during treatment. Furthermore, primary tumors containing more number of CSC stem cell signature resulted in poor patient outcomes and higher rates of metastases [9].</p><p>Several proteins have developed as potential markers for the identification of pancreatic cancer CSCs (Fig. 1). One subpopulation of cells, marked by CD44+CD24+ESA+ (epithelial specific antigen) and represents between 0.2–0.8% of pancreatic cancer cells has a 100-fold increased tumorigenic potential compared with the rest of the cancer cells [12]. Also, CD133+[13, 14], c-MET+/CD44+ [15], increased 26S proteasome activity [16] and ALDH1 [17] have been reported to encode potential pancreatic CSC markers. Furthermore, Ischenko et al. demonstrated that a population of cells with surface markers expression of EPCAM+ CD24+CD44+CD133-Sca-bears CSC properties and metastatic potential [18]. Moreover, we have identified the expression of Doublecortin calmodulin-like kinase 1 (DCLK1) protein in a small proportion of cells in pancreatic cancer [19]. In addition, DCLK1 is found to be marked with a distinct subpopulation of cells in pre-invasive pancreatic cancer with characteristics of stem cells [20]. Furthermore, recent studies have also demonstrated that DCLK1+ cells initiate K-Ras mutant pancreatic tumors in the circumstance of pancreatitis and K-Ras and have shown that DCLK1 are candidates for the origin of pancreatic cancer [21–23].</p><!><p>Multiple pathways have been identified to differentially activate in stem-like cells (Fig. 2). In this manuscript, we have focused on the key pathways and targeting them for prevention.</p><!><p>Aberrant Wnt/β-catenin signaling is one of the concerns in several cancers including pancreatic cancers [24, 25]. Around 65% of pancreatic adenocarcinomas shown to have active Wnt/β-catenin, but β-catenin gene mutations are also seen independently in most of the tumors [25]. Wnt/β-catenin signaling is mainly responsible for developmental process that regulates cell proliferation, differentiation, migration, polarity and asymmetric cell division [26]. β-catenin is an intracellular protein that is localized in cell membrane, cytoplasm and nucleus, an important molecule in this pathway. Wnt ligand binds to its receptors inhibits phosphorylation of β-catenin in the N-terminal region and prevent the protein from degradation which leads to accumulation of the protein in the cytoplasm, and subsequent translocation to the nucleus. Once β-catenin gets localized to the nucleus, it binds to target gene promoters interacting with T-cell factor/lymphoid enhancer factor (TCF/LEF) family members of transcription factors and induces their expression [27].</p><p>In pancreatic cancers, more than 65% of the tumors exhibit an increase in total β-catenin, which are enhanced membranous, cytoplasmic, and nuclear localization of which two have showed CTNNB1 mutation [25]. In addition, gene array analysis demonstrated that canonical arm of the Wnt pathway upregulated in pancreatic cancers [28]. Targeting the Wnt/β-catenin signaling pathway have shown to enhance the sensitivity of chemotherapeutic agents in pancreatic cancer> However, to completely understand the mechanism of action one would have to look at the various pathways affected by Wnt/β-catenin signaling including angiogenesis, cell cycle regulation, apoptosis and maintaining of highly resistant CSCs [29].</p><!><p>Abnormal hedgehog signaling has been shown in many types of human cancers including pancreatic cancers. Three different types of hedgehog genes reported so far are desert hedgehog (DHH), Indian hedgehog (IHH) and sonic hedgehog (SHH). These genes function as ligands for the 12-pass transmembrane receptor, patched (PTCH1) [30]. Hedgehog signaling plays a dual role, it can act as mitogen or can promote differentiation. Increased hedgehog signaling has been shown to alter the behavior of the tumor microenvironment and stroma in pancreatic carcinogenesis. Therefore, hedgehog signaling pathway can be an important target to treat pancreatic cancer [31]. Once hedgehog ligands bind its receptor PTCH1, it results in the internalization and degradation and release of Smoothened (SMO), a G-protein coupled receptor (GPCR) and subsequent dissociation of the suppressor of fused (SUFU)-GLI complex. GLI1 and GLI2 transcription factors translocate to the nucleus and induce the transcription of target genes. GLI3, however, acts as repressor in a normal situation but is degraded during the transcription function[32]. Furthermore, recently it has been shown that mutant K-Ras are involved in the development of pancreatic intraepithelial neoplasia and also in the maintenance and progression of pancreatic cancer in mouse models. Deletion of a single allele of GLI1 resulted in characteristic inflammatory response and improper remodeling of stroma associated with pancreatic cancer in this mouse model. Studies have shown that loss of Gli1 has been identified in cytokines IL6, IL8, monocyte chemoattractant protein-1 (MCP1), and macrophage colony-stimulating factor MCSF which are Gli1 target genes. These studies have also shown that canonical hedgehog signaling are essential for pancreatic recovery. Thus, both GLI1 and hedgehog signaling are critical for regulation of the pancreatic microenvironment [33]. Hedgehog signaling also plays a role in regulation of cellular proliferation, stemness, cell fate determination, and cellular survival [34]. Its downstream target Smo and GLI transcription factors are involved in noncanonical activation of hedgehog signaling [35]. In noncanonical activation the GLI proteins can dodge the inhibition of Smo result in lesser efficacy and also leads to resistance of Smo inhibitors. The family of transcription factors GLI genes transcriptionally regulate downstream targets in hedgehog-dependent survival. Gli1 as a transcriptional target of GLI2, is a primary activator of hedgehog signaling pathway and this may be involved in amplification of hedgehog-induced target gene expression [36]. GLI1 and GLI2 increase transcription of overlapping and distinct sets of target genes [36, 37]. GANT61 is a small molecule inhibitors GLI1-mediated transcription, was identified by cell-based screening. It blocks GLI function in the nucleus, thereby reducing the GLI1- and GLI2-mediated transcription, and inhibits GLI1-DNA binding[38, 39].</p><p>Cyclopamine, a natural compound found in the plant Veratrum californicum, commonly called the corn lily which was the first phytochemical identified to inhibit the hedgehog pathway [40]. Cyclopamine inhibits the activation of Smo, which is the target of the sonic hedgehog[40]. In addition, treatment with cyclopamine significantly reduced GLI1 expression in pancreatic cancer cells [41]. Furthermore, Cyclopamine effectively targets pancreatic CSCs[42]. In addition, studies have demonstrated that pancreatic CSCs are effectively eliminated by hedgehog/GLI inhibitor GANT61 in combination with mTOR inhibition [43]. In pancreatic cancer, CSCs have elevated expression of Sonic hedgehog protein [12], which are believed to be mediators of pancreatic tumor invasion and metastasis [13]. Cyclopamine and gemcitabine combination therapy resulted in inhibition of metastatic spread and reduced primary tumor burden in pancreatic orthotopic xenografts [44]. Moreover, curcumin, an active ingredient of the spice turmeric has been shown to inhibit Sonic hedgehog-GLI1 signaling pathway by downregulating Sonic hedgehog protein and its downstream targets GLI1 and Ptch1 [45]. Furthermore, curcumin, can reverse the epithelial-mesenchymal transition in pancreatic cancer by suppressing the Hedgehog signaling pathway [46]. In addition, curcumin treatment reduces hypoxia-induced pancreatic cancer metastasis, thereby inhibiting the hedgehog signaling pathway [47]. Studies have shown that resveratrol can also reduce proliferation and induce apoptosis through the inhibition of GLI1 and Ptch1 [48, 49]. Sulforaphane also inhibits self-renewal of pancreatic CSCs by suppressing the Sonic hedgehog-GLI pathway [50]. Similarly, the combination of (−)-epigallocatechin-3-gallate and Quercetin can synergistically inhibit the self-renewal capacity of CSCs through reduction of TCF/LEF and GLI activities. Together, these studies suggest that targeting CSCs and sonic hedgehog pathway may improve the outcomes of patients with pancreatic cancer [51]. Recent studies demonstrated that Crocetinic acid isolated from saffron inhibits hedgehog signaling to inhibit pancreatic cancer stem cells [52]. These studies suggest that plant polyphenols or natural compounds target CSCs self-renewal properties can highlight a potential for cancer prevention.</p><!><p>Notch signaling pathway plays an important role in the differentiation and maintenance of stem cells [53]. It has shown that aberrant activation of Notch signaling is linked to the development of many cancers including pancreatic cancers [54]. There are four transmembrane notch receptors (notch 1–4) which upon binding with its ligands (JAG1, JAG2, delta-like 1–4) undergo cleavage, releasing notch intracellular domain that further translocates to nucleus and interacts with its target genes, including Hes-1 and Hey1 [55], cyclin D1[56], p21CIP1[57], NF-κB[58] and c-myc[59]. γ-secretase is a multiprotein intramembrane-cleaving proteases which has four components presenilin, nicastrin, Pen2, and Aph1 and they are all thought to be essential for activity [60]. Presenelin is responsible for the catalytic activity and nicastrin plays critical role in substrate recognition.</p><p>Curcumin has been shown to inhibit Notch signaling pathway in several cancers [61–65]. Similarly, resveratrol a compound found in grapes, berries and peanuts has been shown to exhibit anti-cancer properties by affecting the Notch pathway [66]. In addition, genistein has been shown to inhibit cells growth and induce apoptosis in pancreatic cancer cells by down-regulating Notch-1 [67]. By inhibiting Notch and CXCR4 activities, genistein has been shown to reduce the number of pancreatic CSCs [68]. Furthermore, sulforaphane from broccoli gained attentions of researchers for developing a combination therapy for targeting pancreatic cancer stem cells. a combination treatment of Sulforaphane and gemcitabine inhibits ALDH1 activity because of suppression of Notch-1 and c-Rel expression in pan-creatic cancer [69]. Recent studies have also demon-strated that the Quinoxaline antibiotic Quinomycin A inhibits pancreatic CSCs by inhibiting Notch signaling pathway [70].</p><!><p>The Hippo signaling pathway is essential to maintain tissue homeostasis, organ size, and tumorigenesis. There are two sets of core kinases Mst1/2 and Lats1/2, whose functions are controlled through Sav and Mob. When Hippo signaling is active, Yes associated protein 1 (YAP1) phosphorylate or TAZ transcriptional coactivators by Lats1/2which leads to cytoplasmic sequestration and degradation. On the other hand, inactivation of Hippo signaling can facilitate unphosphorylated YAP/TAZ to enter into the nucleus and thereby one of the four TEAD family members activates transcription. Verteporfin, is a porphyrin molecule which blocks YAP1-TEAD complex formation by binding to YAP1 and result in conformational change of the complex [71]. Dysregulation of the pathway is implicated in many types of cancers. YAP1/TAZ and TEAD are upregulated in a several human tumors and the mechanisms involved through gene amplification and silencing of upstream components of hippo pathway. TAZ is a transducer of the Hippo pathway which has shown to induce epithelial-mesenchymal transition and thereby promote progression and development of pancreatic cancer [72]. It is reported that both YAP1 and TAZ can control the direct activation of JAK-STAT3 signaling pathway, thereby initiating pancreatic cancer in mouse models [73]. YAP has also shown as a critical oncogenic K-Ras effector and a promising therapeutic target for pancreatic cancer [74]. Recent study demonstrated that 14-3-3σ can interact with YAP1, thereby inducing gemcitabine resistance along with promoting ribonucleotide reductase expression [75]. Furthermore, YAP overexpression promotes the epithelial-mesenchymal transition and chemo-resistance in pancreatic cancer cells [76]. Therefore, Hippo signaling protein YAP/TAZ is important target to develop a novel compounds for prevention and treatment of pancreatic cancer.</p><!><p>The Janus kinase (JAK) and signal transducer and activator of transcription (STAT) pathway are involved in various cytokines and growth factors signaling pathways and affects various cellular functions, including cell proliferation, angiogenesis, metastasis, and immune response. JAK-STAT pathways are known to be upregulated in various cancers, including pancreatic cancer [77–79] constitutively activate STAT3 expression, have shown in human pancreatic cancer specimens but the activation is relatively low in normal pancreatic tissues [80]. Recent study demonstrated that gemcitabine treatment enhances the CD24+ and CD133+ cells ratio and also expression of stemness-associated genes Bmi1, Nanog, and Sox2, suggesting that gemcitabine promotes pancreatic cancer stemness by activating Nox/ROS/NF-κB/STAT3 signaling cascade [81]. Therefore, simultaneous targeting of Notch and JAK2/STAT3 signaling pathways may be better than either one alone [82]. In addition, combination of Notch inhibitor GSI IX and JAK2/STAT3 inhibitor AG-490 have shown to be potential as a therapeutic modality for pancreatic cancer [82]. Indole-3-carbinol (I3C) and genistein combination treatment significantly inhibits constitutive activated STAT3 expression in pancreatic cancer cells [80]. Curcumin downregulate the expression of survivin/BIRC5 gene and inhibits constitutive STAT3 phosphorylation in human pancreatic cancer cell lines [83]. Similarly, resveratrol inhibits STAT3 phosphorylation in pancreatic cancer cells in vitro [84].</p><!><p>PI3K/Akt and mTOR signaling pathways are important for many physiological and pathological conditions, such as cell proliferation, angiogenesis, metabolism, differentiation and survival[85]. Most importantly, this pathway acts as a master regulator of cancer. During tumorigenesis, it plays a major role in growth, proliferation, motility, survival and angiogenesis [86, 87]. A recent study demonstrated that the combined inhibition of PI3K/Akt/mTOR and Shh pathways resulted in reduced human pancreatic cancer stem cell characteristics and tumor growth [88]. Shin-Kang et al have looked at the effects of the Vitamin E δ-tocotrienol in pancreatic cancer cells. They observed that Vitamin E δ-tocotrienol inhibits the activation of Akt, ERK/MAP kinase and also its downstream mediator RSK (ribosomal protein S6 kinase) [89]. γ-tocotrienol showed to suppressed the activation of AKT resulted in downregulation of p-GSK-3β and upregulation along with nuclear translocation of FoxO3. Moreover, vitamin E δ-tocotrienol have shown to induce apoptosis and also suppress cell survival and proliferative pathways such as PI3-kinase/AKT and ERK/MAP kinases, which occurred in part by suppressing Her2/ErbB2 expression [89]. Similarly, plumbagin promotes cell cycle arrest and autophagy in pancreatic cancer cells. However, more importantly, the compound suppresses epithelial to mesenchymal transition by inhibiting PI3K/AKT/mTOR and p38 MAPK mediated pathways, and activation of 5′-AMP-dependent kinase [90]. It would be fascinating to evaluate effects of these compounds on cancer stem cells, and also to study understanding mechanism of action for prevention and treatment of pancreatic cancer.</p><!><p>MAPK pathway plays an important role in regulating wide variety of signals which leads to numerous cellular responses such as inflammation, growth, differentiation and cell death. In pancreatic cancer, K-Ras transduces MAPK signaling, which regulates cell proliferation, differentiation, and apoptosis [91]. K-Ras mutation constitutively hyperactivate downstream signaling pathways, including extracellular signalregulated kinase (ERK), PI3K, and the Ral guanine nucleotide exchange factor [92], which subsequently leads to cell transformation and tumorigenesis [93]. MAPK signaling activation results in resistance to TGF-β–induced apoptosis in CD133+ cells as compared to CD133− cells [94]. In addition, CD133+ CSCs specifically show increased activation of ERK1/2 as a result of increased MAPK signaling. Moreover, the CCL21/CCR7 axis is involved in the increased metastatic properties of CD133+ pancreatic cancer stem-like cells thorough EMT and Erk/NF-κB pathways [95]. Furthermore, Chai et al. demonstrated that P70S6K phosphorylation was associated with ERK1/2 phosphorylation, and that metformin is able to suppress this activation, thereby inhibiting proliferation of CD133+ cells proliferation in pancreatic cancer [96].</p><!><p>Recent studies demonstrated that cyclooxygenase-2 (COX-2) and 5-lipoxygenase (5-LOX) inhibitor licofelone, significantly decreased DCLK1+ cells, as well as inflammation and proliferation in a mouse model of pancreatic cancer [97]. In addition, simultaneous targeting of 5-LOX-COX by licofelone and EGFR by gefitinib significantly reduced DCLK1+ CSCs, thereby blocking progression of pancreatic ductal adenocarcinoma [98]. Furthermore, metformin significantly reduced pancreatic cancer stem cell marker proteins CD44, CD133, ALDH1 and EPCAM expression and mTOR signaling pathway, thereby preventing the progression of pancreatic cancer [97]. It has been shown that sulforaphane inhibits self-renewal of pancreatic cancer stem cells through inhibition of Sonic hedgehog-GLI pathway [50]. PanC-1 sphere cells have observed to be more resistant to conventional chemotherapy which shows the presence of significantly increased level of CSC markers. Metformin and curcumin target these pancreatic CSCs [99]. Recent study demonstrated that Triptolide inhibits pancreatic CSCs by reducing the spheroid formation, and ALDH1 expression (100). In addition, it also reduced proliferation and mesenchymal cells along with upregulation of markers for apoptosis and epithelial cells [100]. It has been shown that Minnelide can reduce CD133+ side population and effectively eliminates pancreatic CSCs [101]. Moreover, crocetinic acid, which was purified from Saffron, significantly reduced the expression of pancreatic cancer stem cell markers DCLK1 and CD133 and spheroid formation[52]. Furthermore, the Quinoxaline antibiotic Quinomycin also inhibited pancreatopshere formation, and the number of DCLK1+ cells, as well as suppressing the levels of CSC markers DCLK1, CD44, CD24 and EPCAM in pancreatic cancer cells. In addition, this compound reduced expression for DCLK1, CD44, CD24 and EPCAM in pancreatic cancer tumor xenografts. These data show that Quinomycin may be a potent suppressor of pancreatic cancer that targets the stem cells by inhibiting the Notch signaling pathway [70]. Chemopreventive and natural compounds that inhibit pancreatic cancer stem cells are shown in Fig. (1).</p><!><p>It has become increasingly clear that pancreatic cancer remains one of the deadliest cancers worldwide and the fourth leading cause of cancer-related deaths in the United States. By 2030, it is expected that pancreatic cancer will be the second leading cause of cancerrelated deaths in the United States [2]. Pancreatic tumors have distinct type of cells, which are resistant to conventional therapies and lead to metastasis. One of the small subsets of these cells is CSC in the tumor which is responsible for the growth and metastasis. CSCs might have signaling pathways that are potentially unique to them, such as hedgehog, Wnt/β-catenin, and Notch signaling. If we can identify natural compounds that can specifically target cancer stem cells, then these could be the scaffolds for developing new generations of anti-pancreatic cancer drugs. However, one of the biggest questions in the field is identifying the CSC. Multiple markers have been identified some of which include three cell surface proteins, such and CD44, CD24 and EpCAM [12]. We and others have clearly demonstrated that DCLK1 marks a rare group of cells in the tumor and that these cells have the ability to develop new tumors [19]. Other markers that have been presented are CD133 and ALDH1A1 [13, 17]. We are currently looking for both natural compounds and synthetic compounds that can target DCLK1. Once such compounds are identified, these compounds either alone or in combination with conventional therapy would be a significant step in clinical trials, and could provide a novel approach for prevention and treatment of pancreatic cancers.</p>

PubMed Author Manuscript

Drug repurposing of approved drugs Elbasvir, Ledipasvir, Paritaprevir, Velpatasvir, Antrafenine and Ergotamine for combating COVID19

Aims: Pneumonia of unknown cause detected in Wuhan, China was first reported to the WHO Country Office in China on 31 December 2019. The outbreak was declared a Public Health Emergency of International Concern on 30 January 2020. Currently, there is no Vaccine against COVID-19 pandemic and infection is spreading worldwide very rapidly. The present study aimed to meet the exigent requirement of practicable COVID19 drug treatment with a computational multitarget drug repurposing approach.Main methods: Many reports are available with in silico drug repurposing. However, the majority of them engrossed on a single target. In the present study, 1735 FDA approved drugs screened with molecular docking approach against Covid19 protein and extracts the drug combination targeting COVID19 proteins comprehensively.

drug_repurposing_of_approved_drugs_elbasvir,_ledipasvir,_paritaprevir,_velpatasvir,_antrafenine_and_

Introduction<!>Target preparation<!>Data Analysis<!>Results<!>Comparative binding interaction of drugs against HCV and COVID19<!>Discussion<!>Conclusion

<p>The World Health Organization announced in February 2020 that COVID-19 is the official name of the disease. World Health Organization chief Tedros Adhanom Ghebreyesus explained that CO stands for corona, VI for virus and D for disease, while 19 is for the year that the outbreak was first identified; 31 December 2019 [1]. Coronavirus disease 2019 (COVID-19) is a contagious disease caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) previously referred to as the 2019 novel coronavirus (2019-nCoV) [2]. In 2019 at Wuhan, the capital of Hubei, China, the disease was first reported and then it spread worldwide, resulting in the 2019-20 coronavirus pandemic [3,4].</p><p>At present, there are no clinically proven vaccines and medicines for COVID-19 prevention and treatment as per U.S. Food and Drug Administration (FDA), the Centers for Disease Control and Prevention (CDC) and World Health Organization (WHO) [5]. On an interactive web-based dashboard to track COVID-19 in real-time as on April 19, 2020, in the entire world more than 200 countries/territories are having 2.3 Million confirmed cases and 1,61,283 deaths worldwide and more than 50000 increase daily reported since March 26, 2020 [6,7].</p><p>Recovery observed in patients of Covid19 treated with the mixture of Anti-HIV drugs like Libonavir & Ritonavir, Anti-SARS drugs Oseltamivir and Anti-malarial drug Chloroquine in India (Wadhawan, 2020). In South Korea, human MERS-CoV successfully revokes the viral clearance using a combination of Lopinavir/Ritonavir (LPV/RTV) (Anti-HIV drugs) pegylated interferon and ribavirin [8]. Still, the treatment with anti-HIV drugs is a mystery for the Patients and Researchers as well [9].</p><p>As per data available on https://www.excelra.com/covid-19-drug-repurposing-database/data currently, 125 different drug molecules are reported under trial.</p><p>Recently, diverse computational approaches explored for screening the drug molecules for COVID19 treatment. Molecular docking of Lopinavir, Darunavir and Ritonavir reported against the modelled structure of COVID19 proteases, coronavirus endopeptidase C30 (CEP_C30) and papain-like viral protease (PLVP) [10]. MM-PBSA-WSAS (Molecular dynamics simulations followed by binding free energy calculations using an endpoint method) employed for Fast Identification of Possible Drug against COVID-19 protease [11]. Anti-HCV drugs, Sofosbuvir, IDX-184, Ribavirin, and Remidisvir reported promising drug candidates with a docking approach against modelled COVID-19 RNA dependent RNA polymerase (RdRp) [12]. Hirokawa et al., (2020) identified one hundred and several dozen potentially candidate drugs for 3CL protease inhibitors, which are already approved as antiviral, HIV protease inhibitors, antibacterial or antineoplastic agents with in silico docking-based screening approach, which combines molecular docking with a protein-ligand interaction fingerprint (PLIF) scoring method.</p><p>Many other reports are also available with in silico drug repurposing but the majority of them engrossed on a single target. The present study designed for docking FDA approved Drugbank compounds with molecular weight less than 700 against all COVID-19 experimentally and computationally generated protein structure.</p><!><p>Crystal structure of COVID-19 main protease in complex with an inhibitor N3 (PDB ID: 6LU7) [13], SARS-Coronavirus NSP12 bound to NSP7 and NSP8 co-factors(6NUR) [14] and Pre-fusion 2019-nCoV spike glycoprotein with a single receptor-binding domain (6VSB) [15] retrieved from the Protein Data Bank available at https://www.rcsb.org/. Additionally, computationally predicted 24 modelled structures based on protein sequences translated from the complete genome of Severe acute respiratory syndrome coronavirus 2 isolate Wuhan-Hu1 (Genbank Accession: MN908947.3) available on I-TASSAR online server also included in the present study (Table 1) [16]. Total 27 receptors threedimensional structures were subjected addition of Hydrogen for pH 7.0 and Gasteiger charges were using Open Babel [17]. The resulting structures converted to PDBQT format using python script -prepare_receptor.py‖ from AutoDock Tools [18].</p><!><p>The Binding energy for each drug ligand exported for analysis from the Autodock Vina v4.2 software and need-based analyses performed manually using python script and excel software. Drug target interaction analysed with Ligplot + [22].</p><!><p>Drug repurposing study for FDA approved 1735 drug molecules < 700 MW included in the present study. The molecular docking accomplished against a total of 24 modeled proteins and 3 PDB structures followed by seven-stage screening to determine the best combination for COVID19 treatment.</p><p>In the first stage, a total of 4,68,650 docking solutions obtained from Molecular docking of approved 1735 drug molecules against 27 COVID19 proteins. From Autodock Vina output, one best pose was extracted from out of 10 poses in the second stage and 46,865 pose data utilized for further analysis (Table 2). In stage three screening step, the top 20 drug molecules based on binding energy were separated for each COVID19 Protein. Which lead to 540 drugs with a potential range of docking scores (Table 1 & 2, Figure 1). The redundant drugs removed with the merger of the top 20 dock score for all COVID19 protein together yields 133 unique drugs from the 540 drugs (Supplementary material). All the 133 drugs were manually inspected for pharmacological activity reported in Drugbank. Consequently, 18 anticancer drugs, 6 anti-inflammatory, 7 anti-HIV, 8 anti-HCV, 6 drugs for lung disease, 3 drugs for anti-parasitic activity, 2 anti-migraine and few other activities reported.</p><p>The anti-cancer drugs and drugs with clinically major side effects were omitted. The remaining 35 drugs were selected for further comparative analysis. This included anti-HIV, anti-HCV, anti-Inflammatory, Lung disease, anti-migraine activity, and anti-parasitic activity. Along with nine clinically reported promising anti-COVD19 drugs also included for further analysis. The binding energy of selected 35 compounds graded in five grades displaying dark green, light green, yellow, orange and red for easy comparison. Concerning anti-COVID19 activity against single and multiple target analysis, the best results obtained for 8 drugs from anti-HCV followed by 7 anti-HIV and 3 currently clinically used drugs. In addition to that, 1 drug from an anti-inflammatory and anti-migraine group also had a good docking score. The overall analysis presented the comparatively better scoring of the selected drug in this study over currently clinically applied drugs for COVID19 treatment (Figure 1) To narrow down the number of drug combinations and removing drug acting on the same target binding energy replaced with docking rank. Based on dock rank 10 drugs selected for further analysis (Figure 2). To reduce the drug for treatment and avoid duplication, the drug acting on the same target was removed. Consequently, the combination of Elbasvir, Ledipasvir, Paritaprevir, Velpatasvir, Antrafenine Ergotamine drug figured out as a potential cocktail for COVID19 treatment (Figure 3).</p><!><p>To analyse the comparative binding interaction with HCV targets for anti-COVID19 receptors, three targets of HCV NS3 Helicase, NS5B RNA-dependent RNA polymerase and NS3/4A S168A protease docked against 26 potential drugs along with known inhibitors. The result indicated, Chloroquine antiplasmodial drug computationally significant for three HCV receptors but the same is less significant for COVID19 (Figure 4). Similarly, anti-HIV drugs Dolutegravir and Maraviroc had considerable binding against HCV targets but less potential with COVID19 receptors. Conversely, Elbasvir, Ledipasvir reacted with related potential against Helicase, RdRp and Protease from both virus HCV and COVID19. This projected the potential of anti-HCV drugs for inhibition of the COVID19 virus.</p><p>While ergotamine used in the present study proposes significant binding against all three COVID19 proteins but no promising binding evident with HCV. This suggests the presence of a specific binding site present in COVID19 which might be absent in HCV.</p><!><p>The earlier report of drug repurposing mainly focused on a single target option we represented the holistic approach of targeting all the viral proteins for effective anti-COVID19 activity. The combination suggested will effective on the nonstructural proteins, structural proteins of COVID19 and also included drug the anti-inflammatory with synergic anti-COVID19 activity. Comparative computational analysis of the proposed drug is superior to currently clinically used drug combinations. Which, reflects the possibility of the promising effect of the proposed drug combination in further clinical applications.</p><p>Ledipasvir is an inhibitor of the Hepatitis C Virus (HCV) Non-Structural Protein 5A (NS5A) [23].</p><p>This protein is crucial for viral RNA replication and assembly of HCV virions. Ledipasvir and Velpatasvir were reported promising in one of the recent studies in virtual screening [24]. Clinical trials against anti-HCV suggest the Ledipasvir treatment significantly improves the patients within one to twelve weeks [25,26].</p><p>Similarly, Velpatasvir acts as a defective substrate for NS5A (Non-Structural Protein 5A) sharing a similar function as Ledipasvir. The preclinical study shown the patient having an infection of HCV genotype 1 to 6 can be treated with Velpatasvir [27].</p><p>Elbasvir was primarily used for the treatment of HCV. However, some trials are also carried out for using Elbasvir in the treatment of COVID19 [28]. Paritaprevir reported as an inhibitor of COVID19</p><p>Protease [29]. No reports are available for anti-COVD19 effectively of Antrafenine and Ergotamine.</p><!><p>There is an urgent need for anti-COVID-19 drugs to address the global medical emergency. Several drugs are currently employed with empirical clinical knowledge along with some contradictions.</p><p>Many drug repurposing reports are available but those are mainly concentrated on a single target. The present study directed with a holistic approach of targeting multiple COVID19 proteins with a mixture of FDA approved drugs. The proposed blend drugs include Elbasvir, Ledipasvir, Paritaprevir, and Velpatasvir which currently used for HCV treatment. Inclusion anti-inflammatory Antrafenine and anti-migraine Ergotamine drugs can be effective for dual action of inflammation reduction and COVID19 inhibition. The anticipated combination of drugs acting on both non-structural and structural proteins therefore, can be able to reduce the COVID19 infection process and also reduce viral multiplication. The present study can be immediacy explore further by medical, pharm and research experts effectively to find out the best strategy for anti-COVID19 treatment.</p><p>The Green colored cell contains the lowest binding energy for each row followed by the greenish-yellow, yellow, orange and red color. The lowest binding energy represents the firm binding of ligand and protein molecules. The Comparative analysis establishes the drugs used for the treatment of Hepatitis C represented a more number of greenish cells. This indicates these drugs stay on top throughout the library used in the present study. The Elbasvir, Ledipasvir, Paritaprevir, Velpatasvir from Anti-HCV followed by Anti-HIV drugs Saquinavir, Ritonavir, Dolutegravir, Maraviroc, Tipranavir, Etravirine and Rilpivirine having good binding energy one or more targets. Only one drug, Antrafenine used for anti-inflammatory and Ergotamine used for anti-migraine activity shown significant activity against the different targets. However, the drugs used worldwide for clinical purposes except for Danoprevir and Lopinavir have shown poor binding with receptor molecules used in the study. The drugs from the list removed having overlapping target applicability. The removal identified 6 unique drugs acting on the multiple targets. The combination of these drugs overcomes the lower affinity of the drug molecules to the specific receptors targeting all the proteins available in the present study. Ergotamine shown good binding against 18 receptors, followed by Lepidasvir, Elbasivir and Pariptapevir respectively against 16, 12 and 9 receptors. While velpatasvir and anterafenine showed binding against 5 which might be very useful in a multi-drug combinatorial approach. The results graded on five color green for lowest binding energy followed by greenish-yellow, yellow, orange and red for the highest binding energy for each receptor.</p><p>Figure 4: Comparison of HCV targets as a reference against COVID19.</p>

ChemRxiv

Merging of the photocatalysis and copper catalysis in metal–organic frameworks for oxidative C–C bond formation

The direct formation of new C-C bonds through photocatalytic oxidative coupling from low reactive sp 3 C-H bonds using environmentally benign and cheap oxygen as oxidant is an important area in sustainable chemistry. By incorporating the photoredox catalyst [SiW 11 O 39 Ru(H 2 O)] 5À into the pores of Cu-based metal-organic frameworks, a new approach for merging Cu-catalysis/Ru-photocatalysis within one single MOF was achieved. The direct Cu II -O-W(Ru) bridges made the two metal catalyses being synergetic, enabling the application on the catalysis of the oxidative coupling C-C bond formation from acetophenones and N-phenyl-tetrahydroisoquinoline with excellent conversion and sizeselectivity. The method takes advantage of visible light photoredox catalysis to generate iminium ion intermediate from N-phenyl-tetrahydroisoquinoline under mild conditions and the easy combination with Cu-catalyzed activation of nucleophiles. Control catalytic experiments using similar Cu-based sheets but with the photoredox catalytic anions embedded was also investigated for comparison.

merging_of_the_photocatalysis_and_copper_catalysis_in_metal–organic_frameworks_for_oxidative_c–c_bon

Introduction<!>Synthesis and characterizations of CR-BPY1<!>Catalysis details of CR-BPY1<!>Synthesis and catalytic characterizations of CR-BPY2<!>Conclusions<!>General methods and materials<!>Synthesis of CR-BPY2<!>X-ray crystallography

<p>The direct formation of new C-C bonds through oxidative coupling reactions from the lower active sp 3 C-H bonds using oxygen as oxidant is an important area in sustainable chemistry. 1 Among the reported promising examples, the oxidative activation of the C-H bonds adjacent to nitrogen atom in tertiary amines represents a powerful strategy, giving valuable, highly reactive iminium ion intermediates for further functionalization. 2 Recent investigations also revealed that the visible light photoredox catalysis was a promising approach to such reaction sequences 3 with respect to the development of new sustainable and green synthetic methods. It was also postulated that the combination of the photocatalysis and the metal catalysis within a dual catalytic transformation is attractive to circumvent the potential side reactions relative to the highly active intermediates that exist in the photocatalysis. 4 The hurdles that need to be overcome include the careful adaptation and the ne tuning of the reaction rates of the two catalytic cycles, 5 beside the appropriate choice of the metal catalysis and photocatalysis. 6 Metal-organic frameworks are hybrid solids with innite network structures built from organic bridging ligands and inorganic connecting nodes. Besides the potential applications in many diverse areas, 7 MOFs are ideally suited for catalytic conversions, since they can impose size and shape selective restriction through readily ne-tuned channels or pores, 8 providing precise knowledge about the pore structure, the nature and distribution of catalytically active sites. 9 In comparison to the heterogeneous catalytic systems that have been examined earlier, the design exibility and framework tunability resulting from the huge variations of metal nodes and organic linkers allow the introduction of more than two independent catalyses in one single MOF. 10 The combination of photocatalysis with the metal ions or organocatalysis was expected to be a promising approach to create synergistic catalysts. 11 By incorporating a ruthenium(III) substituted polyoxometalate [SiW 11 O 39 Ru(H 2 O)] 5À within the pores of copper(II)-bipyridine MOFs, herein, we reported a new approach to merge the visible light photocatalytic aerobic oxidation and copper(II) catalytic coupling reaction within one MOF (Scheme 1). We envisioned that the ruthenium-containing fragments possibly worked as oxidative photocatalyst to generate the iminium ion from N-phenyl-tetrahydroisoquinolines, 12 whereas the Cu-based MOF potentially activated the nucleophiles, as it was shown in the oxidative C-C bond coupling. 13</p><!><p>Solvothermal reaction of 4,4 0 -bipyridine (BPY), Cu(NO 3 ) 2 $3H 2 O and K 5 [SiW 11 O 39 Ru(H 2 O)]$10H 2 O gave CR-BPY1 in a yield of 52%. Elemental analyses and powder X-ray analysis indicated the pure phase of its bulk sample. Single-crystal structural analysis revealed that CR-BPY1 crystallized in a space group P42 1 m. Two crystallographically independent copper(II) ions are connected by BPY ligands and m 2 -water bridges alternatively to produce 2D wavy-like Cu-BPY sheets (Fig. S5, ESI †). The Cu(2) atom adopted a six-coordinate octahedral geometry with four nitrogen atoms from four BPY ligands positioned in the equatorial plane and two water molecules occupied the axial positions. The Cu(1) atom displayed a ve-coordinate square pyramidal geometry with two m 2 -water groups and two nitrogen atoms of BPY ligands positioned in the basal plane, and a terminal oxygen atom of the depronated [SiW 11 O 39 Ru(H 2 O)] 5À polyoxoanion occupied the vertex position. The ruthenium atom disordered in the twelve equivalent positions within a depronated [SiW 11 O 39 Ru(H 2 O)] 5À . 14 The availability of vacant dorbitals on the metal atoms adjacent to the heteroatom allows the polyoxometalate matrix to function as a p-acceptor ligand. 15 While these copper atoms were connected by the BPY ligands to form two-dimension square grid at rst, adjacent sheets were connected together using the deprotonated [SiW 11 O 39 Ru(H 2 O)] 5À polyoxoanion by Cu II -O-W(Ru) bridges to generate a 3D framework. Two symmetric-related frameworks further interpenetrated each other perpendicularly to consolidate the robust structure (Fig. 1), in which the opening of the pores was reduced to 10.0 Å Â 5.3 Å. To the best of our knowledge, CR-BPY1 represents the rst example of MOFs which are comprised of ruthenium substituted polyoxometalate [SiW 11 O 39 Ru(H 2 O)] 5À . As the noble metal substituted polyoxometalates exhibited excellent photoreactivity in various catalytic oxidation processes of organic substrates, 16 such kinds of MOFs potentially allow the combination of photocatalysis and MOF-based heterogeneous catalysis to achieve synthetically useful organic transformations. Moreover, the directly bridging of the copper and ruthenium by Cu II -O-W(Ru) provided a promising way to achieve the synergistic catalysis between photocatalyst and metal catalyst.</p><p>Confocal uorescence microscopy has attracted much attention in biological imaging. It may provide a way to analyse relatively thick porous materials, because it offers the advantage of increased penetration depth (>500 mm). 17 The assessment of guest-accessible volume in MOFs can be reliably done by using confocal uorescence microscopy with a tool-box of dyes with a wide range of sizes. It would be applicable to any porous materials, whose single-crystal structures are not available, or non-crystalline materials. 18 Dye uptake investigation was carried out by soaking CR-BPY1 in a methanol solution of 2 0 ,7 0 -dichlorouorescein. It gave the quantum uptake equivalent to 5% of the MOF weight (Fig. S11, ESI †). 19 The confocal laser scanning microscopy exhibited strong green uorescence (l ex ¼ 488 nm) assignable to the emission of the uorescein dye (Fig. 2), conrming the successful uptake of the dye molecules inside the crystals of the MOF. 20 Furthermore, the rather uniform distribution of the dye molecules throughout the crystal suggested that the dyes penetrated deeply into the crystal rather than staying on the external surface. Without guest water molecules, the effective free volume of CR-BPY1 was estimated to be 29.0% by PLATON soware. 21 CR-BPY1 exhibited an absorption band centered at 398 nm in the solid state UV-vis absorption spectrum (Fig. S1, ESI †), assignable to the transitions of [SiW 11 O 39 Ru(H 2 O)] 5À . 22 Upon excitation at this band, CR-BPY1 did not exhibit any obvious emission, however, progressive addition of the N-phenyl-tetrahydroisoquinoline into the dichloromethane suspension of CR-BPY1 up to 0.50 mM caused the appearance of the Ru IIrelative emission band at about 422 nm (Fig. 3a). 23 The results suggested that CR-BPY1 oxidized N-phenyl-tetrahydroiso-quinoline to form the Ru II species and the iminium intermediate. 24 Electrospray ionization mass spectrometry of the CH 2 Cl 2 suspension containing N-phenyl-tetrahydroisoquinoline and CR-BPY1 aer 3 hours light irradiation exhibited an intense peak at m/z ¼ 208. This peak was assignable to the relative imine ion, conrming that CR-BPY1 oxidized N-phenyl-tetrahydroisoquinoline to form the Ru II species and the iminium intermediate (Fig. S13, ESI †). The electron paramagnetic resonance (EPR) of CR-BPY1 exhibited the characteristic signal of Cu II with g ¼ 2.14 (Fig. 3c). Solid state electrochemical measurements (Fig. 3d) exhibited a broad redox band centred at À186 mV (vs. SCE) relative to the overlap of the Cu II /Cu I and Ru III /Ru II redox couples. The potentials were comparable to these Cu II and Ru III -containing catalysts, 25 and enabled CR-BPY1 to prompt the oxidative coupling of N-phenyl-tetrahydroisoquinoline with nucleophiles under light. 26 It seems that CR-BPY1 adsorbed the N-phenyl-tetrahydroisoquinoline in its pores and activated the substrate to form the iminium intermediate.</p><!><p>The catalysis was examined initially using N-phenyl-tetrahydroisoquinoline and nitromethane as the coupling partners, along with a common uorescent lamp (18 W) as the light source. The resulting reaction gave a yield of 90% aer 24 hours irradiation. The removal of CR-BPY1 by ltration aer 18 hours shut down the reaction, and the ltrate afforded only 12% additional conversion for another 18 hours at the same reaction conditions. The observation suggested that CR-BPY1 was a true heterogeneous catalyst. 27 Solids of CR-BPY1 could be isolated from the reaction suspension by simple ltration alone and reused at least three times with moderate loss of activity (from 90% to 82% of yield aer three cycles). The index of XRD patterns of CR-BPY1 ltrated off from the reaction mixture suggested the maintenance of the crystallinity (Fig. S14, ESI †). With the size of the microcrystals reduced to 2 mm by grinding CR-BPY1 crystals for 20 min, the time of the reaction giving the same conversion to that of the as-synthesized materials was reduced by about 10% (Fig. S15, ESI †). It seems that the MOFbased particles having well-dened size were really helpful for the catalytic reactions, but the size of the crystals did not dominate the catalysis directly.</p><p>Control experiments for the C-C coupling reaction of N-phenyl-tetrahydroisoquinoline and nitromethane were carried out and summarized in Table 1. Almost no conversion was observed when the reaction was conducted in the dark (entry 7), while a very slow background reaction was observed in the absence of catalyst (entry 6), which demonstrated that both the light and the photocatalyst are required for efficient conversion to the coupling products. In addition, using the same equiv. of copper(II) salts or/and K 5 [SiW 11 O 39 Ru(H 2 O)] as catalysts, respectively gives conversions of 39%, 25% and 42% in homogeneous fashion (entry 3-5). These results suggested that the direct connection of copper(II) ions to [SiW 11 O 40 Ru] 7À anions not only enabled the dual catalysts to individually activate N-phenyl-tetrahydroisoquinoline and nitromethane, but also enforced the proximity between the potential intermediates i.e. the iminium ion and nucleophile, avoiding the unwanted side reactions or reverse reactions. 28 Although several examples of photocatalysts and metal copper catalysts have been reported to prompt the oxidative coupling C-C bond formation, CR-BPY1 represents a new example of a heterogeneous bimetal catalyst that merges the copper catalyst and the ruthenium(III) substituted polyoxometalate catalyst within one single material. The high The reactions were carried out in the presence of a common used secondary amine, L-proline, as an organic co-catalyst to activate the ketones. 30 In the case of the acetophenone as reactant with a uorescent lamp (18 W) as the light source; the catalytic reaction gave a yield of 72%. Control experiments demonstrated that the use of K 5 [SiW 11 O 39 Ru(H 2 O)] or copper(II) salts as catalysts, only gave less than 25% of the conversions, respectively. The results indicated the signicant contribution of cooperative effects of the individual parts within one single MOF. From the mechanistic point of view, the ruthenium(III) of the polyoxometalate [SiW 11 O 39 Ru(H 2 O)] 5À interacted with Nphenyl-tetrahydroisoquinoline to form iminium ions, whereas the copper atoms coordinated to the acetophenones weakly to form the enol intermediate that worked as active nucleophile for the oxidative coupling C-C bond formation. At the same time, the presence of copper ions could enhance the activation of N-phenyl-tetrahydroisoquinoline, beneting the synergistic catalysis between photocatalyst and metal catalyst. Importantly, in contrast to the smooth reactions of substrates 1-3, the C-C coupling reaction in the presence of bulky ketone (1-(3 0 ,5 0 -ditert-butyl [1,1 0 -biphenyl]-4-yl)-ethanone) 4, gave less than 10% conversion under the same reaction conditions (Table 2, entry 4). The negligible adsorption by immersing CR-BPY1 into a dioxane solution of substrate 4, coupled with the fact that the size of substrate 4 was larger than that of the channels, 31 revealed that 4 was too large to be adsorbed in the channels. Furthermore, it is suggested that the synergistic catalytic coupling reaction indeed occurred in the channels of the MOF, not on the external surface.</p><!><p>To further investigate the synergistic interactions between the inorganic copper and [SiW 11 O 39 Ru(H 2 O)] 5À anion, a reference compound CR-BPY2 was assembled using the same starting components but different synthetic conditions (hierarchical diffusion). CR-BPY2 was synthesized by a diffusion method in a test tube by laying a solution of 4,4 0 -bipyridine in acetonitrile onto the solution of K 5 [SiW 11 O 39 Ru(H 2 O)]$10H 2 O and Cu(NO 3 ) 2 $3H 2 O in water for several days in a yield of 59%. Elemental analyses and powder X-ray analysis indicated the pure phase of its bulk sample. Single-crystal structural analysis revealed that CR-BPY2 crystallized in the orthorhombic lattice with a space group Pccn. Two crystallographically independent copper(II) ions connected four BPY bridges alternatively to produce a 2D sheet (Fig. 5), which were further stacked paralleled along the crystallographic a axis to form the 3D structure with embedded [SiW 11 O 39 Ru(H 2 O)] 5À (Fig. S8, ESI †). The copper(II) ions resided in an octahedral geometries with the equatorial plane which was dened by four nitrogen atoms of BPY ligands, and the axial positions were occupied by two water molecules (Fig. S7, ESI †). Without guest water molecules, the effective free volume of CR-BPY2 was also estimated to be 33.9% by PLATON soware, which is quite larger than that of CR-BPY1. These results suggested that the pore of CR-BPY2 is larger enough to adsorb the substrates. Since [SiW 11 O 39 Ru(H 2 O)] 5À polyoxoanions were embedded in the channels, it is thus an excellent reference for investigating the catalytic activity on the same coupling reaction. a Reaction conditions: N-phenyl-tetrahydroisoquinoline (0.25 mmol), 1 mol% catalyst, 2.0 mL nitromethane, 18 W uorescent lamp at room temperature. b The conversions aer 24 hour irradiation were determined by 1 H NMR of crude products. CR-BPY2 also exhibited an absorption band centered at 398 nm in the solid state UV-vis absorption spectrum. Upon excitation at this band, CR-BPY2 did not exhibit obvious emission, however, progressive addition of the N-phenyl-tetrahydroisoquinoline into the dichloromethane suspension of CR-BPY2 up to 0.50 mM caused the appearance of the Ru II -relative emission band at about 422 nm, suggesting that CR-BPY2 oxidized N-phenyl-tetrahydroisoquinoline to form the Ru II species. The EPR of CR-BPY2 exhibited the characteristic signal of Cu II (g ¼ 2.13). 32 The sharper peak shape compared to that of CR-BPY1 might be one of the indicator of isolated Cu II ions in CR-BPY2. No metal-metal interactions were found corresponding to the Cu II ions in CR-BPY2. Solid state electrochemical measurements exhibited two redox peaks corresponding to the Cu II /Cu I and Ru III /Ru II redox couples, with the redox potential calculated at 75 mV and 84 mV (vs. SCE). The potentials enabled CR-BPY2 to prompt the oxidative coupling of N-phenyl-tetrahydroisoquinoline with nucleophiles under light. However, the redox peaks also suggested that these Cu I and Ru III ions did not interacted directly. It seems that CR-BPY2 adsorbed the N-phenyl-tetrahydroisoquinoline in its pores and was a convincing reference to investigate the synergistic action between Cu II -O-W(Ru) bimetal of CR-BPY1.</p><p>The catalytic activities of CR-BPY2 in the C-C coupling reactions were examined under the same conditions using nitromethane and N-phenyl-tetrahydroisoquinoline as the reactants. About 1 mol% loading amount of the catalyst gave rise to a 46% conversion, which was superior to the case when copper(II) salts and the K 5 [SiW 11 O 39 Ru(H 2 O)] were employed as catalysts, indicating the signicance of the two constitute parts for CR-BPY2 as a photocatalyst. However, the catalytic activities of CR-BPY2 were signicantly weaker than that of CR-BPY1 (Fig. 3b). It should be concluded that the direct bridging of the copper and ruthenium by Cu II -O-W(Ru) provided a promising way to achieve the synergistic catalysis between photocatalyst and metal catalyst, and the high reaction efficiency in the reactions was dominated by the spacious environment of the channels, like those of CR-BPY1.</p><!><p>In a summary, we reported the new example of copper MOFs containing the ruthenium substituted polyoxometalate with the aim of merging the synergistic Cu-catalysis/Ru-photocatalysis in a single MOF. CR-BPY1 exhibited perpendicularly inter-penetrated structure and the catalytic sites positioned in the robust pores of MOFs. Luminescence titration and IR spectra of the MOF-based material revealed the adsorbance and activation of N-phenyl-tetrahydroisoquinoline and acetophenone, by the ruthenium center and copper ions, respectively. The direct connection of copper(II) ions to [SiW 11 O 40 Ru(H 2 O)] 5À not only provided the possibility of the dual catalysts to individually activate the substrates, but also enforced the proximity between the intermediates, avoiding the unwanted side reactions or reverse reactions. CR-BPY1 exhibited high activity for the photocatalytic oxidative coupling C-C bond formation with excellent size-selectivity, suggesting the catalytic reactions occurred in the channels of the MOF, and not on the external surface.</p><!><p>All chemicals were of reagent grade quality obtained from commercial sources and used without further purication. 1 H NMR was measured on a Varian INOVA-400 spectrometer with</p><!><p>The CR-BPY2 was synthesized by a diffusion method in a test tube. A mixture of acetonitrile and water (1 : 1, 10.</p><!><p>Data of CR-BPY1 and CR-BPY2 were collected on a Bruker Smart APEX CCD diffractometer with graphite-monochromated Mo-Ka (l ¼ 0.71073 Å) using the SMART and SAINT programs. 34 Their structures were determined and the heavy atoms were found by direct methods using the SHELXTL-97 program package. 35 Crystallographic data for them are summarized in Table 3. Except some partly occupied solvent water molecules, the other non-hydrogen atoms were rened anisotropically. Hydrogen atoms within the ligand backbones were xed geometrically at their positions and allowed to ride on the parent atoms. In both of the two structures, the ruthenium atoms were disordered in the equivalent positions of tungsten atoms. For CR-BPY2, several bond distances constraints were used to help the renement on the BPY moiety, and thermal parameters on adjacent oxygen atoms of the polyoxometalate anion were restrained to be similar.</p>

Royal Society of Chemistry (RSC)

Nucleophilic vinylic substitution in bicyclic methyleneaziridines: S_NV_π or S_NV_σ?

A stereodefined monodeuterated methyleneaziridine is shown to be prepared via coordinated reductive ring-opening of an alkynyl epoxide and diastereoselective tethered allene aziridination. Ring-opening of this aziridine with copper-based organometallics follows a pathway that results in stereoretentive substitution, replacing the exo-C-N bond with a corresponding C-C bond; this stereochemical outcome supports either an overall S N V p mechanism or a C-N insertion/reductive coupling process.Scheme 1 The formation and S N V ring-opening of fused 1,3-oxazolidin-2-one methyleneaziridines.

nucleophilic_vinylic_substitution_in_bicyclic_methyleneaziridines:_s_nv_π_or_s<s

Introduction<!>Results and discussion<!>Conclusions<!>General information<!>Trans-2-(deuterioethynyl)-3-isopropyloxirane (7)

<p>In 2010 both we 1 and Blakey 2 reported the first examples of intramolecular allene aziridination with sulfamate substrates, with the major products being derived in most cases via 2-amidoallylcation intermediates. 3 Our group followed this up with the first report 4 of analogous reactions of carbamate substrates 1 (Scheme 1) and, in that work, somewhat unstable bicyclic 1,3-oxazolidin-2-one methyleneaziridines 2 were obtained following Lebel's modification 5 of the Du Bois protocol 6 for Rh(II)-nitrenoid generation. Soon afterwards, Schomaker's group took on the area and developed it extensively, optimising the conditions for generating the methyleneaziridines, engineering the substrates for synthetic tractability (non-terminal allenes, formation of 1,3-oxazinan-2-ones), and elaborating the products into a variety of hydroxy/amino stereotriads and -tetrads and rearranged heterocycles. 7 In our original publication we noted that the methyleneaziridines were constrained by the ring-fusion such that only the exocyclic aziridine C-N bond is electronically activated in the ground state through hyperconjugation with the carbamate carbonyl p-system. This suggested the possibility of effecting direct substitution/ ring-opening at the sp 2 -carbon, in contrast to the prevailing reactivity of unconstrained methyleneaziridines in which ring-opening occurs preferentially at the sp 3 -carbon. 8 At the time, the only sp 2 -C-N bondcleaving processes involved either transition metal-mediated processes 9 or stepwise radical addition/b-scission. 10 In the event, treatment of methyleneaziridine 2 (R = i-Pr) with lithium diphenylcuprate, or various Grignard reagents in the presence of CuI, led to moderate to good yields of the products 3 of nucleophilic vinylic substitution (S N V). 11 That publication concluded with an intention to clarify the stereochemical details of the S N V reaction; the current paper describes studies to that end.</p><!><p>A stereochemically defined monodeuterated analogue 4 (Scheme 2) of methyleneaziridine 2 (R = i-Pr) was targeted that would allow the stereochemistry of the S N V process to be probed without presenting any steric or electronic bias compared with the original methyleneaziridine. At the outset of this study, a dissociative mechanism for the substitution reaction was ruled out on the basis of the aprotic, low-temperature conditions for the process and the relative instability of a vinylic cation. An out-of-plane (relative to the cleaving C-N bond) stepwise p-addition/elimination process, proceeding via a short-lived formal carbanion located on the terminal methylene carbon, or an equivalent concerted mechanism, would proceed with retention of configuration (S N V p pathway, -5). An in-plane concerted process, akin to an S N 2 reaction in aliphatic substrates, would lead to inversion of configuration (S N V s pathway, -6). The operation of either of these reaction modes would then be revealed in the relative disposition of the newlyformed C-C bond and the adjacent H/D atoms, as shown.</p><p>In the absence of any literature precedent for the synthesis of a stereodefined terminally monodeuterated buta-2,3dienol, 12 a synthesis of methyleneaziridine 4 was proposed based upon diastereoselective coordinated delivery of hydride 13 to deuterated alkynyl epoxide 7 (Scheme 3) and the known stereochemical course of the intramolecular aziridination. Following this proposal, trans-2-ethynyl-3-isopropyloxirane 14 was stirred with an excess of D 2 O under basic conditions 15 to yield the deuterated alkyne 7 (94% deuterium incorporation). Alkyne 7 was treated with DIBAL in dichloromethane as a non-coordinating solvent that would support epoxide chelation with the aluminium centre, and allene 8 was isolated apparently as one predominant stereoisomer, 16 depicted as that expected, and confirmed retrospectively from the NMR data for methyleneaziridine 4. A slightly modified variant of Lebel's protocol for nitrenoid formation afforded consistent yields (B25%) of methyleneaziridine 4 from N-tosyloxy carbamate 9; lower yields were obtained from carbamate 10 with a range of Rh(II) catalysts including Rh 2 (OAc) 4 , Rh 2 (esp) 2 , 17 and Rh 2 (TPA) 4 . 18 The stereochemistry in methyleneaziridine 4 was confirmed by comparisons with the NMR data for non-deuterated methyleneaziridine 2, 4 and the NOE correlations shown in Fig. 1. In the 1 H NMR spectrum of methyleneaziridine 4, the adjacent methine protons at d 1.81 and 4.35 show 3 J HH = 9.5 Hz, indicating a dominating trans-antiperiplanar disposition that places one of the diastereotopic methyls more regularly in close proximity to the CHN and =CHD protons, as seen in the NOE spectra. A simple dihedral drive calculation supports this view (ESI †). 19 Two variants of the S N V reaction were carried out, both of which converted methyleneaziridine 4 into products with reasonable overall efficiency (Scheme 4). In the first, addition of lithium dimethylcuprate gave a 77% isolated yield of 4-isopropenyl oxazolidinone 11, in which the methyl group was found (see below) to be cis-to the deuterium atom. In the second, a copper-catalysed Grignard reaction with vinylmagnesium bromide gave 4-(buta-1,3-dien-2-yl) oxazolidine 12 as the major product, again with the new C-C bond formed cis-to the deuterium atom. The azirine 13 was also isolated in this work; its formation may be explained by competing addition at the carbonyl followed by 1,4-vinylation of the so-formed a,b-unsaturated ester. 20 A combination of NMR experiments, including NOE (Fig. 2) provided support for the stereochemical assignments in S N V products 11 and 12. Notably, in 11 no NOE correlation was observed between the vinyl methyl protons andQCHD; similarly, in compound 12, there were no significant correlations between the vinyl protons and QCHD.</p><p>An invertive S N V s reaction appears to be stereoelectronically accessible in methyleneaziridines 2 and 4, and the microscopic reverse of such a process is supported in the NaNH 2 -mediated formation of simple methyleneaziridines from 2-bromoallylic amines. 21 Despite this, our results clearly rule out the S N V s mode of ring-opening, the stereochemical outcome being consistent with a (retentive) S N V p mode of reaction. Setting aside the extent of the involvement of the metal counterions in this process, at one simplistic mechanistic extreme, as the delivery of the methyl or vinyl ligand to the methylene group initiates and charge begins to build on the terminal carbon, the sp 2 -C-N bond weakens, with progression along this pathway</p><!><p>To the best of our knowledge, the direct nucleophilic sp 2 C-N bond cleavage reactions that we reported in 2010 remain the only examples in methyleneaziridine chemistry. In this work, we have demonstrated that the substitution is stereoretentive, ruling out an S N V s pathway, but the detailed mechanism of these reactions remains open to speculation and further work is intended to close this particular chapter of methyleneaziridine reactivity. 23</p><!><p>All solvents for anhydrous reactions were obtained dry from Grubbs solvent dispenser units after being passed through an activated alumina column under argon. THF was additionally distilled from sodium/benzophenone ketyl under argon. Commercially available reagents were used as supplied unless otherwise specified. Triethylamine was distilled from CaH 2 and stored over KOH pellets under argon. 'Petrol' refers to the fraction of light petroleum ether boiling between 30 and 40 1C; 'ether' refers to diethyl ether. All reactions were carried out in oven-dried glassware and under an atmosphere of argon unless otherwise specified. Thin layer chromatography (TLC) was carried out using Merck aluminium backed DC60 F254 0.2 mm precoated plates. Spots were then visualised by the quenching of ultraviolet light fluorescence (l max 254 nm) and then stained and heated with either anisaldehyde or KMnO 4 solutions as appropriate. Retention factors (R f ) are reported along with the solvent system used in parentheses. Flash column chromatography was performed using Merck 60 silica gel (particle size 40-63 mm) and the solvent system used is reported in parentheses. Infrared spectra were recorded using a Bruker Tensor 27 FT-IR fitted with a diamond ATR module. Absorption maxima (n max ) are reported in wavenumbers (cm À1 ) and are described as strong (s), medium (m), weak (w) or broad (br). Proton ( 1 H) and carbon-13 ( 13 C) spectra were recorded on Bruker AVIII HD 500, AVII 500, or AVIII HD 400 spectrometers. Chemical shifts (d H or d C ) are reported in parts per million (p.p.m.) downfield of tetramethylsilane, internally referenced (in MestReNova) to the appropriate solvent peak: CDCl 3 , 7.26/77.16; acetone-d 6 , 2.05/29.84. Peak multiplicities are described as singlet (s), doublet (d), triplet (t), quartet (q), septet (sept), octet (oct), multiplet (m), and broad (br) or a combination thereof. Coupling constants (J) are rounded to the nearest 0.5 Hz. Assignments are made on the basis of chemical shifts, integrations, and coupling constants, using COSY, HSQC and nOe experiments where appropriate. High Resolution Mass Spectra (HRMS) were recorded by the staff at the Chemistry Research Laboratory (University of Oxford) using a Waters GC-TOF spectrometer (EI/FI). Melting points were recorded on a Griffin melting point apparatus and are uncorrected.</p><!><p>Trans-2-ethynyl-3-isopropyloxirane (2.03 g, 18.4 mmol) was added to a stirring solution of K 2 CO 3 (3.76 g, 27.2 mmol) in acetonitrile (42 mL). After 30 min, D 2 O (20 mL) was added and stirring was continued for 5 h. The product was extracted from the reaction mixture into petrol (5 Â 100 mL). The combined Scheme 4 Reagents and conditions: (i) Me 2 CuLi (1.0 eq.), THF, À20 1C -RT, 30 min (11, 77%); (ii) vinyl-MgBr (2.0 eq), CuI (5 mol%), THF, À50 1C -0 1C, 1 h (12, 31%; 13, 23%). extracts were dried (MgSO 4 ) and the solvent was removed in vacuo [CARE: the product is volatile] to afford the title compound as a pale yellow oil (1.50 g, 73%, 94% deuterium incorporation). R f 0.58 (petrol/ether, 3 : 1); n max /cm À1 (thin film): 2966m, 2589m, 1980w, 1469m; d H (400 MHz, CDCl 3 ) 0.97 (3H, d, J = 7.0 Hz), 0.99 (3H, d, J = 7.0 Hz), 1.52 (1H, oct, J = 7.0 Hz), 2.29 (0.1H, d, J = 1.5 Hz, residual RCH), 2.89 (1H, dd, J = 7.0, 2.0 Hz), 3.11 (1H, d, J = 2.0 Hz); d C (100 MHz, CDCl 3 ) 18.1, 18.7, 30.4, 43.9, 65.4, 71.5 (t, J = 38.5 Hz), 80.3 (t, J = 7.5 Hz).</p><p>(3R*,5R*)-6-Deuterio-2-methylhexa-4,5-dien-3-ol (8)</p><p>A solution of epoxyalkyne 7 (1.50 g, 13.5 mmol) in dichloromethane (100 mL) was added dropwise to a stirred solution of DIBAL (33.7 mL, 1.0 M in hexane, 33.7 mmol) in dichloromethane (100 mL) at 0 1C. The reaction mixture was stirred for 1 h then quenched by careful addition of water. A satd. aq. solution of Rochelle's salt (200 mL) was added dropwise and the mixture was stirred overnight to allow the solvent layers to separate completely. The mixture was extracted with dichloromethane (5 Â 50 mL), the organic layers were combined and washed with brine (50 mL), then dried (MgSO 4 ) and the solvent removed in vacuo. Purification by flash chromatography (petrol/ ether, 8 : 1) afforded the title compound as a pale yellow oil (810 mg, 53%). R f 0.23 (petrol/ether 3 : 1); n max /cm À1 (thin film) 3409br, 2957m, 2924s, 2854m, 1953w, 1464w, 1379w, 1261w, 1024w; DNNaO 3 , 195.0850; found, 195.0846. Recrystallised p-TsCl (619 mg, 3.25 mmol) was added to a stirred solution of the hydroxycarbamate (554 mg, 3.22 mmol) in dry ether (30 mL) at 0 1C. Triethylamine (0.45 mL, 3.23 mmol) was then added dropwise and stirring was continued for 18 h. The mixture was then diluted with ether (30 mL), washed with brine (2 Â 20 mL), dried (Na 2 SO 4 ), and the solvent removed in vacuo. The crude product was purified by flash chromatography (petrol/ether, 5 : 1 pure ether) to afford the title compound as a pale yellow oil (893 mg, 85%). R f 0.50 (petrol/ether, 1 : 1); n max /cm À1 (thin film) 3284br, 2967m, 1770s, 1737s, 1598m, 1467m, 1379s, 1192s, 1179s, 1019m, 742m; d H (400 MHz, CDCl 3 ) 0.82 (3H, d, J = 7.0 Hz), 0.84 (3H, d, J = 7.0 Hz), 1.</p>

Royal Society of Chemistry (RSC)

18O Kinetic Isotope Effects Reveal an Associative Transition State for Phosphite Dehydrogenase Catalyzed Phosphoryl Transfer

Phosphite dehydrogenase (PTDH) catalyzes an unusual phosphoryl transfer reaction in which water displaces a hydride leaving group. Despite extensive effort, it remains unclear whether PTDH catalysis proceeds via an associative or dissociative mechanism. Here, primary 2H and secondary 18O kinetic isotope effects (KIEs) were determined and used together with computation to characterize the transition state (TS) catalyzed by a thermostable PTDH (17X-PTDH). The large, normal 18O KIEs suggest an associative mechanism. Various transition state structures were computed within a model of the enzyme active site and 2H and 18O KIEs were predicted to evaluate the accuracy of each TS. This analysis revealed that 17X-PTDH catalyzes an associative process with little leaving group displacement and extensive nucleophilic participation. This tight TS is likely a consequence of the extremely poor leaving group requiring significant P-O bond formation to expel the hydride. This finding contrasts with the dissociative TS in most phosphoryl transfer reactions from phosphate mono- and diesters.

18o_kinetic_isotope_effects_reveal_an_associative_transition_state_for_phosphite_dehydrogenase_catal

<p>Phosphite dehydrogenase (PTDH) catalyzes the oxidation of phosphite (PT) to phosphate with the concomitant reduction of NAD+ to NADH.1 The enzyme allows microorganisms to use PT as their sole phosphorus source.2 The favorable thermodynamics (ΔGo = −15 kcal/mol) and the low cost of PT have attracted attention to PTDH as a cofactor regeneration system.3,4 This application spurred generation of PTDH mutants with increased activity,5 thermostability6 and decreased cofactor specificity to enable regeneration of both NADH and NADPH.7 These PTDH variants have been used in biocatalytic systems, both as individual regeneration systems and as fusions with monooxygenases within self-sufficient biocatalysts.8–11 Because PTDH endows organisms with the unique ability to grow on PT, the enzyme has also received attention for biocontainment.12–15</p><p>Extensive efforts have been made to understand how PTDH displaces an extremely poor leaving group (hydride) with a water nucleophile. Much of this work has focused on a thermostable variant termed 17X-PTDH.16 All available data suggests that 17X-PTDH and wild type PTDH operate via the same catalytic mechanism.16,17 His292 is the putative base that activates the water nucleophile and Arg237 is involved in orienting the substrate for catalysis.1 Other catalytically important residues (Figure 1) were identified by site-directed mutagenesis,18 crystallography19 and computation.20 Pre-steady state kinetics and kinetic isotope effects (KIEs) with deuterated PT have demonstrated that chemistry is entirely rate limiting.17 While a reasonable mechanism for this process can be postulated from this information (Figure 1), the relative timing of bond-making and bond-breaking and the protonation state of the substrate remain unclear.</p><p>Phosphate monoester hydrolyses generally proceed via loose transition states (TSs).21 However, the chemistry catalyzed by 17X-PTDH is significantly different from a typical phosphoryl transfer. As 17X-PTDH does not accept alternative substrates or nucleophiles,22 linear free energy relationships (LFERs) cannot be used to probe the TS. Instead, we turned to secondary 18O KIEs (18(V/K)). This tool is well-developed for the study of phosphoryl transfer reactions,21,23 where 18(V/K) > 1 is expected for a tight TS (significant P-O bond formation, little P-H bond cleavage) and 18(V/K) < 1 for a loose TS (little bond formation, significant bond cleavage). This interpretation assumes that the dominant contribution to 18(V/K) is the difference in P-O bond stretching vibrations in the ground state and the TS (Scheme S1). This framework is an oversimplification and other techniques including computational approaches are required to interpret secondary 18O KIEs confidently and to fully characterize the TS.</p><p>Steady-state kinetic parameters of 17X-PTDH were determined from the initial rates of NADH production at varying concentrations of NAD+ and PT (Table 1; Figure S1). The primary H/D KIEs on Vmax (DV) and kcat/Km,PT (D(V/K)) were determined by direct comparison of the values obtained with protiated and deuterated PT (Figure S2). The dependence of kcat/KM,PT on pH and the independence of DV with varying acidity agree with previously reported data.16</p><p>Values of 18(V/K) were determined using NMR spectroscopy as described by Bennet.24 18O-labeled PT was prepared by hydrolyzing PCl3 with H 18O.25 The effect per 18O on the 31P chemical shift (Δδ ~ 0.03 ppm) allowed quantification of the isotopic composition by spectral deconvolution (Figure 2a). The initial isotopic composition (R0) of a mixture of 18O3-PT and 16O3-PT was determined from the relative peak areas of each isotopologue. After the addition of 17X-PTDH, 31P NMR spectra were recorded over time. The ratio of 18O3-PT to 16O3-PT (R) was determined as a function of the fractional conversion of the light isotopologue (F1) and fit to Equation 1.26 A subset of spectra from one experiment and the corresponding fit to Equation 1 are shown in Figure 2. (eq 1)RR0=(1−F1)(1KIE)−1</p><p>18(V/K) was determined at different acidities (Figures S3-S5). In all cases, 18(V/K) was large and normal (Table 2). The variation with pH might suggest changes in the TS but this is inconsistent with the constant DV across the pH range. As the protonation state of PT (pK 2 = 6.62) changes across this pH range, an equilibrium isotope effect (EIE) on the pKa of the labeled PT likely contributes to the observed 18(V/K). We determined this EIE by monitoring the chemical shift of each isotopologue as a function of pH (Figure S6).27 The obtained EIE of 1.016 ± 0.001 was in excellent agreement with reported EIEs on the second pKa of phosphate.28</p><p>The EIE enriches the dianionic PT in 16O. If 17X-PTDH binds dianionic PT (Scheme S2), the "active" substrate pool would be enriched in the light isotopologue, resulting in a normal contribution to the observed 18(V/K). Dividing the KIEs observed at each acidity by the appropriate portion of the EIE (as determined by the percent of PT present as the dianion at each pH; see Supporting Information)29 results in a pH independent 18O KIE of 1.017 ± 0.001 (Table 2). Conversely, if 17X-PTDH binds monoanionic PT (Scheme S2), the active substrate would be enriched in 18O and the EIE would inversely contribute to the observed 18(V/K). Multiplying each observed KIE by the appropriate portion of the EIE gives a pH independent 18O KIE of 1.031 ± 0.001 for monoanionic PT (Table 2). Regardless of which substrate protonation state is correct, these values (1.017 and 1.031) are both large, normal KIEs.</p><p>To facilitate interpretation of the observed KIEs, TSs for the 17X-PTDH-catalyzed reaction were computed within a model of the active site. Such analysis typically requires determination of the intrinsic KIEs.30–33 However, it has been demonstrated that chemistry is entirely rate-limiting for 17X-PTDH and the observed DV is the intrinsic KIE (Dk).17 Since DV and D(V/K) are essentially identical (Table 1), a significant forward commitment factor (cf) is unlikely and the (V/K) will reflect the intrinsic KIE (k; eq 2)34. (eq 2)(V/K)18=k18+cf1+cf</p><p>A TS for the reaction of monoanionic PT was located using the M06–2X/6–31G* level of theory as implemented in Gaussian 09.35 Solvation effects were considered with the polarizable continuum model.36,37 This structure exhibited significant P-Onuc bond formation and little P-H bond cleavage. Next, P-Onuc (d1; Figure 3a) and P-H (d2; Figure 3a) bond distances were altered by 0.05 Å intervals and additional TSs were identified with these distances fixed. KIEs were computed for each structure using ISOEFF07 (Figure 3b).38 Structures with d2 = 1.580 Å gave 2H KIEs in good agreement with the observed value. With d1 = 1.841 Å, the calculated 18O KIE matched the expected value for monoanionic PT. No TSs could be found that predicted 18O KIEs as large as 1.031. TSs located with dianionic PT predicted inverse 18O KIEs. Collectively, these results suggest a mechanism in which dianionic PT binds to 17X-PTDH with His292 protonated. Then, proton transfer occurs in the ternary complex before phosphoryl transfer from monoanionic PT (Scheme 1) that proceeds through a tight TS (Figure 3a) with bond orders of 0.75 for the P-O bond and 0.70 for the P-H bond. Chemistry occurring on the monoanionic substrate is satisfying since it resembles a phosphate diester from which phosphoryl transfers proceed through less dissociative mechanisms than phosphate monoesters.21 Calculations used to arrive at this structure were relatively insensitive to the basis set used (Tables S6 – S10).</p><p>The protonation of dianionic PT by His292 upon binding could be part of a binding isotope effect (BIE) that would inversely contribute to 18(V/K).28,39 That contribution may or may not be counteracted by other interactions in the ternary complex. While BIEs can be normal or inverse, 17X-PTDH would likely constrain the vibrational state of the oxygen atoms of PT.40 Since this potential inverse contribution of substrate binding is not accounted for by our model, we posit that 18(V/K) of 1.017 reflects the lower limit of 18k suggesting a P-O bond length of < 1.89 Å in the TS (Figure 3b). As DV accurately reflects Dk, the P-H bond length in the TS will be approximately 1.59 Å (for discussion of potential tunneling, see the Supporting Information).</p><p>Additional calculations were performed to evaluate the influence of the arginine mimics on the TS geometry and KIEs. Removal of the guanidinium ions did not significantly alter the tightness of the TS (Table S5). This finding is consistent with the observation that positive charges common to the active site of phosphotransferases do not promote tightening of the TS.21,41–43 In the presence of the arginine mimics, the predicted 18O KIEs are smaller (Table S5). This decrease results from the nonbridging oxygen atoms becoming more vibrationally restricted when interacting with the guanidinium ions. As both Arg237 and Arg301 are known to be important for 17X-PTDH catalysis and are in the active site,16,18 the model that includes two guanidinium ions better reflects the enzymatic reaction. TSs located within this model suggest the arginine residues orient the substrate for catalysis and activate the nucleophilic water for deprotonation. These calculations also predict a significantly larger activation barrier in the absence of the arginine mimics (Table S5), consistent with the drastic reduction in activity upon mutation of either catalytic arginine.16,18</p><p>Compared to the loose TSs observed for phosphate monoester hydrolyses, the TS proposed for 17X-PTDH catalysis is consistent with an expected anti-Hammond effect due to the significantly worse leaving group.21,44 On a More-O'Ferrall-Jencks plot45,46 for P-O bond formation and cleavage of the bond to a generic leaving group, the change from alkoxide to hydride as leaving group will raise the energy of the metaphosphate corner (Figure S7). Consequently, the TS will shift towards the phosphorane corner, leading to more P-O bond formation and less bond cleavage to the leaving group, which we observe. The upper-right corner of this plot would also increase in energy with a hydride leaving group. The resulting Hammond effect would shift the TS towards the products. If both effects were equal in importance, the net effect would predict increased P-O bonding to the incoming nucleophile and little difference in the extent of leaving group departure (Figure S7). Our proposed TS shows little evidence for this Hammond effect, which could reflect the extremely exergonic nature of the 17X-PTDH reaction or an imbalance in the magnitude of the Hammond and anti-Hammond effects (Figure S7).</p><p>The large, normal 18(V/K) reported here for 17X-PTDH catalysis is similar to observations for phosphate triester hydrolyses,21,47 which proceed via tight TSs. Supplementing the data with computation provides a TS structure that is consistent with the observed H/D and 18O KIEs. This structure illustrates a tighter TS than would be expected for the hydrolysis of phosphate mono- and diesters and provides the first insights into the TS for the unusual phosphoryl transfer catalyzed by PTDH.</p>

PubMed Author Manuscript

Optimizing P,N-Bidentate Ligands for Oxidative Gold Catalysis: Highly Efficient Intermolecular Trapping of \xce\xb1-Oxo Gold Carbenes by Carboxylic Acids

Steric Bulk or Conformation Control? Optimization of P,N-bidentate ligands reveals the importance of conformation control in the development of highly efficient intermolecular trapping of reactive \xce\xb1-oxo gold carbene intermediates. While a pendant piperidine ring offers suitable steric bulk, fixing its conformation to provide better shielding to the highly electrophilic carbene center turned out to be crucial for the excellent reaction efficiency. A generally highly efficient and broadly applicable synthesis of carboxymethyl ketones from readily available carboxylic acids and terminal alkynes is developed under exceptionally mild reaction conditions.

optimizing_p,n-bidentate_ligands_for_oxidative_gold_catalysis:_highly_efficient_intermolecular_trapp

<!>General procedure for gold-catalyzed synthesis of \xce\xb1-carboxymethyl ketones

<p>A few years ago we reported[1] that gold-catalyzed[2] intermolecular oxidation of alkynes offered an expedient access to synthetically highly versatile α-oxo gold carbene intermediates (Scheme 1A). This strategy circumvents the use of hazardous and potentially explosive α-diazo ketone precursors[3] and has led to the development of various efficient synthetic methods by us[1, 4] and others[5] based on their intramolecular trapping. The intermolecular counterpart, likely of exceptional synthetic utilities, however, proves to be very challenging due to the highly electrophilic nature of the carbene center, and is often plagued by over oxidation, reactions with solvents,[4b, 4g] and intractable side reactions. Earlier we reported for the first time that the reactivity of the gold carbene could be attenuated by using bidentate phosphine ligands so that it reacted with carboxamides efficiently, leading to a one-step synthesis of 2,4-disubstituted oxazoles.[6] Mor-DalPhos (see Scheme 1B),[7] a bulky P,N-bidentate ligand, was found as a uniquely effective ligand, and its role is proposed to enable the formation of a tris-coordinated gold carbene (e.g., A) instead of the typical bis-coordinated one (e.g., A'), leading to attenuated electrophilicity at the carbene center (Scheme 1B).</p><p>To further develop alkynes as surrogates of hazardous α-diazo ketones in gold catalysis especially in the context of synthetically important intermolecular trapping, we embarked on expanding the scope of suitable external nucleophiles beyond carboxamide. Our first target was carboxylic acid (Scheme 1A), a weaker nucleophile in its neutral form than carboxamide. Notably, the reaction, if developed, would offer a novel and rapid access to synthetically versatile α-carboxymethyl ketones (i.e, 3, Scheme 1) from readily available terminal alkynes and carboxylic acids.</p><p>At the outset, we used 1-dodecyne and benzoic acid (1.2 equiv) as the reacting partners and 8-methylquinoline N-oxide[4d] as the oxidant, and the results of reaction optimization are shown in Table 1. Consistent with our previous study,[6] cationic gold complexes derived from typical ligands such as Ph3P, IPr, and BrettPhos were largely ineffective, resulting in complex mixtures with little desired product (entry 1). On the contrary, Mor-DalPhos[7] again proved to be an effective ligand, and the oxidative gold catalysis led to the desired α-benzoxymethyl ketone 3a in a fairly good yield (entry 3). An identical yield was observed with the ligand L1[7] (Figure 1A), which differs from Mor-DalPhos by having a piperidine ring instead of a morpholine ring (entry 4). However, the smaller Me-DalPhos (Figure 1A)[7] was much inferior (entry 2), suggesting that the steric size of the pendant secondary amine might be critical.</p><p>Consequently, we modified the piperidine ring of L1 with different substituents. While the installation of a methyl group at its 4-position was inconsequential (entry 5, see L2 in Figure 1B), a much bigger t-butyl was detrimental (entry 6, see L3 in Figure 1B). However, to our delight, a 3-methyl group, as in L4 (Figure 1B), led to a notable increase of the reaction yield (entry 7); moreover, the use of cis-3,5-dimethylpiperidine (as in L5) as the pendant amine group resulted in a higher 84% yield of 3a (entry 8). On the contrary, the ligand L6 with a trans-3,5-dimethylpiperidine ring led to a much less efficient reaction (entry 9), which could be attributed to deleterious steric congestion caused by an inevitable axially oriented methyl group. The piperidine ring in L5 could be replaced with a morpholine ring, as in L7, with only small, yet adverse impact on the reaction outcome (entry 10). Interestingly, replacing the methyl groups in L7 with bigger phenyl groups as in L8 resulted in a yield even lower than that by Mor-DalPhos (entry 11). This result, along with that by L3, suggesting that there is an optimal steric bulk for the pendant six-membered ring in this reaction. When benzoic acid was the limiting reagent, L5 was a significantly better ligand than L7 for the gold catalysis (comparing entries 12 and 13). Moreover, the reaction yield in the former case is nearly quantitative, which is impressive considering the difficulties previously encountered in trapping these reactive gold carbene species and really showcased the opportunities in method development based on ligand design/development. Other solvents such as DCE (entry 14) and toluene (entry 15) were suitable for this reaction albeit not nearly as good as PhCl. Although the oxidant, 8-methylquinoline N-oxide had to be introduced to the reaction slowly via a syringe pump in order to avoid over oxidation, the reaction proceeded smoothly at ambient temperature.</p><p>With the optimized reaction conditions revealed in Table 1, entry 13, the scope of the transformation was first examined with various carboxylic acids. As shown in Table 2, entries 1–6, various substituted benzoic acids were allowed, yielding the desired products in mostly excellent yields. Even a Bpin group was tolerated, and the relatively low isolated yield was due to the co-elution of the boronated product 3f with 8-methylquinoline (entry 5). The reaction also worked well with other conjugated acids such as thiophene-2-carboxylic acid (entry 7), trans-cinnamic acid (entry 8), trans-3-(2-furyl)acrylic acid (entry 9), and trans, trans-hexa-2,4-dienoic acid (entry 10), delivering the corresponding products in ≥90% yields. Acetic acids with the α-carbon functionalized by a 1-methylindol-3-yl (entry 11), a trimethylsilyl (entry 12), a chloro (entry 13), and a phenoxy (entry 14), were all suitable substrates, and functionalized α-carboxymethyl ketones were again isolated in good to excellent yields; the reaction with N-Boc protected proline, however, only resulted in a serviceable yield of the corresponding product (entry 15). Other carboxylic acids such as cyclopropanecarboxylic acid (entry 16) and adamantane-1-carboxyclic acid (entry 17) were also allowed. While these reactions were run in 0.2 mmol scale, a 3 mmol-scale reaction of entry 2 proceeded well even with only 2 mol % of the catalyst, affording 3c in a respectful 82% yield.</p><p>The reaction also proceeded smoothly with various terminal alkynes. As shown in Table 3, phenylacetylene (entry 1) and 1-ethynylcyclohexene (entry 4) were excellent substrates, and so are the acetylenes substituted by cyclic alkyl groups (entries 2 and 3) or remotely functionalized linear ones (entries 5–7). The reaction yields were invariably excellent. Even with heteroatoms placed close to the C-C triple bond and hence to the in-situ formed electrophilic gold carbene center, this intramolecular reaction still led to a good (entry 8) or a serviceable yield (entry 9).</p><p>The ligand optimization via modification of the pendant piperidine/morpholine ring is worthy of commenting and further probing. In our previous carboxamide trapping chemistry,[6] the smaller Me-DalPhos was found as an equally effective bidentate ligand as Mor-DalPhos; On the contrary, in this chemistry, the steric size of the pendant amino group is apparently crucial for the reaction outcome (See Table 1). We attributed the difference to the decreased nucleophilicity of the carbonyl oxygen in carboxylic acid[8] in comparison to that in carboxamide. It is reasonable to suspect that steric shielding around the gold carbene center would facilitate, to a certain extent, slower reactions with small nucleophiles by minimizing sterically demanding side reactions. With regard to the ligand L1 and Mor-DalPhos, the enhanced efficiency with L5 could be attributed a priori to the bulkier piperidine ring; however, two cis methyl groups could at the same time impose conformation rigidity to the N-heterocycle. Scheme 2A outlines two chair conformers of the gold carbene intermediate with L1 as its ligand. It is noteworthy that the X-ray diffraction studies of Mor-DalPhos Pd complexes by Stradiotto[7a] reveal that the morpholine ring can adopt the chair conformation in B' and twist boat conformations. The conformer B', though understandably less stable and hence less populated, does not provide sufficient steric protection to the carbene center; consequently, the reaction could be improved if the conformers including B' and other less shielding ones could be minimized. We surmise that the two cis methyl groups in L5 might play the role of locking the piperidine in a conformation identical to that in B. As such, the carbene center is constantly shielded and hence would react preferably with smaller nucleophiles such as carboxylic acids over bigger ones including the oxidant. X-ray diffraction study of L5AuCl confirmed the preferred shielding chair conformation for the cis-3,5-dimethylpiperidine ring (Scheme 2B). In the case of L8, the phenyl groups are likely too bulky to hinder the carbene formation and/or the productive approach by a carboxylic acid. (1) </p><p>To offer more insight into the importance of conformation control, we prepared the ligand L9 (Scheme 2C), where an ortho methyl group on the benzene ring is installed to essentially prohibit the piperidine ring from adopting the chair conformation in L9-ax and likely minimizing the contribution of other non-chair conformers. Indeed, the gold catalyst afforded 3a in 83% yield (Eq. 1), much better than that with the ligand L1 (where the ortho-Me is absent) and virtually idential to that by L5 (see Table 1, entry 8). When the ligand L10 with an ortho-methyl group and a pendant cis-3,5-dimethylpiperidine was used for the gold catalysis, the reaction efficiency remained the same (Eq. 1), suggesting that the role of the piperidine methyl groups in the ligand L5 is fixing the desired chair conformation of the N-heterocycle instead of providing additional steric bulk. (4) </p><p>This one-step, generally applicable, and efficient synthesis of functionalized carboxylmethyl ketones permits rapid access to various useful cyclic structures. For example, 2-alkenyloxazole 4 was isolated in 80% yield by the combination of the gold catalysis and a subsequent BF3•Et2O-promoted cyclodehydration[9] in a one-pot process (Eq. 2); moreover, the phenone intermediate 3ab, isolated in 93% yield, has been previously converted into the γ-lactone 5 in 72% yield (Eq. 3).[10] Another approach to cyclized products are shown in Eq. 4, where DBU effectively promoted sequential intramolecular aldol reaction and dehydration of the ketones 3ac and 3ad, which in turn were obtained in excellent yields via the oxidative gold catalysis.</p><p>In conclusion, we have developed a generally highly efficient and broadly applicable synthesis of carboxymethyl ketones from readily available carboxylic acids and terminal alkynes under exceptionally mild conditions. In this oxidative gold catalysis, the highly electrophilic α-oxo gold carbene intermediate is most likely generated, and its challenging intermolecular trapping by weakly nucleophilic carboxylic acids is achieved upon extensive optimization of the P,N-bidentate ligands to the metal gold. While the steric bulk of the pendant amino group in these ligands is beneficial to a certain extent, controlling the conformation of the sterically suitable piperidine ring to provide better shielding to the carbene center appeared to be important to achieve the generally high efficiencies. Importantly, the reaction products can be rapidly converted into synthetically versatile cyclic structures. Further studies employing other types of nucleophiles including so-far unyielding enol ethers are currently underway.</p><!><p>To a 3 dram vial containing 2 mL of chlorobenzene were added sequentially a carboxylic acid (0.2 mmol), an alkyne (0.26 mmol), L5AuCl (0.01 mmol) and NaBArF4 (0.02 mmol). The resulting mixture was stirred at room temperature. To this vial a solution of N-oxide (47.7 mg, 0.3 mmol) in 4 mL of chlorobenzene was then added via a syringe pump in 12 h. Upon completion, the reaction mixture was concentrated under vacuum. The residue was purified by chromatography on silica gel (eluent: hexanes/ethyl acetate) to afford the desired product 3.</p>

PubMed Author Manuscript

Structural analysis and biological activities of BXL0124, a gemini analog of vitamin D

Gemini analogs of calcitriol, characterized by the extension of the C21-methyl group of calcitriol with a second chain, act as agonists of the vitamin D receptor (VDR). This second side chain of gemini is accommodated in a new cavity inside the VDR created by the structural rearrangement of the protein core. The resulting conformational change preserves the active state of the receptor and bestows gemini compounds with biological activities that exceed those of calcitriol. Of particular interest are gemini\xe2\x80\x99s anti-cancer properties, and in this study we demonstrate anti-proliferative and tumor-reducing abilities of BXL0124 and BXL0097, differing only by the presence or absence, respectively, of the methylene group on the A ring. BXL0124 acts as a more potent VDR agonist than its 19-nor counterpart by activating VDR-mediated transcription at lower concentrations. In a similar manner, BXL0124 is more active than BXL0097 in growth inhibition of breast cancer cells and reduction of tumor volume. Structural comparisons of BXL0097 and BXL0124, as their VDR complexes, explain the elevated activity of the latter.

structural_analysis_and_biological_activities_of_bxl0124,_a_gemini_analog_of_vitamin_d

Introduction<!>Compounds<!>Crystallization and structure determination<!>Cell culture<!>Transient Transfection and Luciferase Reporter Gene Assays<!>Cell proliferation assay<!>Xenograft tumor studies<!>BXL0124 shows transcriptional activity higher than its 19nor counterpart<!>MCF10DCIS and MCF7 cell growth inhibition by BXL0124 and BXL0097<!>BXL0124 inhibits MCF10DCIS human breast cancer xenograft tumor growth in SCID mice<!>Structural analysis of the BXL0124-VDR LBD complex<!>Conclusion<!>

<p>Breast cancer is a complex progressive disease with multiple subtypes and varying clinical outcomes. Ductal carcinoma in situ (DCIS) is an early, non-malignant lesion of the breast and recognized as a precursor of invasive breast cancer. A cancerous stem-cell-like population has the capacity to drive malignant progression to invasive ductal carcinoma (IDC), manifest by transgression of cancerous growth through the ductal lining and initiating metastasis and resistance to conventional therapies (Morrison et al. 2008). Calcitriol (1α,25-dihydroxy vitamin D3 or 1α,25(OH)2D3) and its analogs, particularly gemini compounds, can act as inhibitors of breast cancer progression by retarding or preventing the transition of DCIS to IDC (Wahler et al. 2014). Indeed, gemini reveal multifunctional activities that include inhibition of different types of breast carcinogenesis without displaying hypercalcemic toxicity and with potencies that are 10 to 100 fold higher than those of calcitriol (Lee et al. 2008; Pazos et al. 2014). Initial synthesis procedure of gemini analogs was described in (Norman et al. 2000). A novel synthetic methodology for gemini compounds was recently described (Pazos et al. 2016). Among the gemini analogs investigated in breast cancer models, BXL0124 (Figure 1) has been shown to be a potent agent for the prevention of different types of human breast cancer; it modifies cancer stem cell subpopulations into less stem-like differentiated cells, it inhibits mammosphere formation and suppresses mammary tumor growth in animals (Lee et al. 2006; So et al. 2011; Lee et al. 2010; Wahler et al. 2014; So et al. 2015).</p><p>Similarly to calcitriol, gemini compounds act in conjunction with the vitamin D nuclear receptor (VDR). We previously reported the crystal structures of the VDR ligand-binding domain (LBD) in complex with the parental gemini bearing two identical side chains and several of its derivatives (Ciesielski, Rochel, and Moras 2007; Huet et al. 2011; Maehr et al. 2013). A common feature of gemini compounds is their ability to induce a conformational change of the protein core by creating a new cavity that accommodates the second, saturated side chain obtained by C21 elongation. Most significantly, this structural rearrangement preserves the active state of the agonist-bound LBD.</p><p>The present study gains insights into the structure-activity relationships of BXL0124 and BXL0097 (Figure 1). Both compounds are closely related to cholecalciferol wherein the side chain of cholecalciferol was modified by a 23-triple bond and the six hydrogen atoms at positions 26 and 27 were replaced with fluorine atoms, and the position 21 was extended with a (3-hydroxy-3-trideuteromethyl-4,4,4-trideuterobutyl) group, thus maintaining the 20(R) configuration of cholecalciferol (Maehr et al. 2009). These chemical alterations were selected to prevent or retard biological degradation initiated by 24-hydroxylation thus extending the half-life of the compound. In contrast to BXL0097 which has a 19-nor A-ring, BXL0124 maintains the methylene group at position 19 of calcitriol. The 19-nor analogs have the advantage of enhanced chemical stability due to the absence of the triene function. Interestingly, binding of 19-nor-calcitriol to the VDR has been shown to be only 30% of that of calcitriol while the effects on HL60 differentiation were similar and calcemic effects were reduced (Bouillon et al. 1993; Perlman et al. 1990). Although numerous 19-nor-1,25(OH)2D3 analogs have been synthesized and their crystal structures solved [reviewed in (Belorusova and Rochel 2014)], there is no structural comparison of agonists differing only by the presence or absence of the C19 methylene group.</p><p>In this study we examine and compare the biological activities of BXL0124 and BXL0097. We show that both ligands are more potent in VDR-mediated transcriptional activation and in inhibition of breast cancer cell proliferation than the calcitriol, with BXL0124 displaying slightly higher activity than BXL0097. Importantly, BXL0124 is more effective in the suppression of mammary tumor growth than BXL0097. We further report the crystal structure of the zebrafish zVDR LBD in complex with BXL0124 explaining elevated potency of this compound.</p><!><p>Calcitriol was purchased from Sigma, and all gemini compounds (>95% purity) (Figure 1) were prepared as described (Maehr et al. 2009). All compounds were used in ethanol solutions.</p><!><p>cDNA encoding zVDR LBD (156–453 AA) was cloned into the pET28b vector to generate N-terminal His-tag fusion protein. Purification was carried out as previously described, including metal affinity chromatography and gel filtration (Ciesielski, Rochel, and Moras 2007). The protein was concentrated using Amicon ultra-30 (Millipore) to 3–7 mg/ml and incubated with a two-fold excess of ligand and a three-fold excess of the coactivator SRC-2 peptide (686-KHKILHRLLQDSS-698). Crystals were obtained in a solution containing 50 mM Bis–Tris pH 6.5, 1.6 M lithium sulfate and 50 mM magnesium sulfate. Protein crystals were mounted in a fiber loop and flash-cooled under a nitrogen flux after cryo-protection with 20% glycerol. Data collection from a single frozen crystal was performed at 100 K on the ID30 beamline at ESRF (France). The raw data were processed and scaled with the HKL2000 program suite (Otwinowski and Minor 1997). The crystals belong to the space group P6522, with one LBD complex per asymmetric unit. The structure was solved and refined using BUSTER (version 2.11.2. Cambridge, United Kingdom: Global Phasing Ltd (Bricogne et al. 2016)), Phenix (Adams et al. 2010) and iterative model building using COOT (Emsley and Cowtan 2004). Crystallographic refinement statistics are presented in Supplementary Table 1. All structural figures were prepared using PyMOL (www.pymol.org).</p><!><p>The MCF10DCIS.com line was provided by Dr. Fred Miller at the Barbara Ann Karmanos Cancer Institute (Detroit, MI). MCF10DCIS.com cells were maintained in DMEM/F12 medium supplemented with 5% horse serum, 1% penicillin/streptomycin, and 1% HEPES solution at 37°C and 5% CO2. The cells were passed every 3 to 4 days. MCF-7 cells were acquired from ATCC. MCF-7 cells were maintained in DMEM/F12 supplemented with 10% fetal bovine serum and 1% penicillin/streptomycin at 37°C and 5% CO2. 17-β Estradiol supplementation for MCF-7 cells was purchased from Sigma Aldrich (St. Louis, MO) and dissolved in ethanol. HEK293 EBNA cells were maintained in DMEM medium supplemented with 5% fetal bovine serum, 40 µg/mL gentamycin and 1 mg/mL geneticin G418 at 37°C and 5% CO2.</p><!><p>HEK293 EBNA cells plated into 24-well plates at 105 cells per well were cotransfected with 150 ng of the expression plasmid pSG5-hVDR, 150 ng of the reporter plasmid pLuc-MCS (Stratagene, La Jolla, USA) containing the proximal promoter region (-414 to -64) of the human CYP24A1 gene, 3 ng of the pRL plasmid (Promega, Madison, USA) containing the Renilla luciferase gene (transfection and cell viability control), and 697 ng of the carrier plasmid pBluescript (Stratagene). Transfection was performed with jetPEI (PolyPlus Transfection, Illkrich, France) according to the manufacturer's instructions. Six hours after transfection, test-compounds were added. Cells were harvested after eighteen hours of incubation with ligands. The amounts of reporter gene product (firefly luciferase) and constitutively expressed Renilla luciferase produced in the cells were measured using Dual-Luciferase® Reporter Assay System (Promega) on a luminometer plate reader LB 96P (Berthold Technologies). Luminescence of firefly luciferase values were normalized to the Renilla luciferase activity. Luciferase activities were expressed as relative units of light intensity.</p><!><p>The assay protocol of cell proliferation is reported previously (Lee et al. 2008). For cell proliferation assay, MCF10DCIS cells were incubated with compound-containing DMEM/F12 medium supplemented with 5% horse serum medium for three days. MCF-7 cells were incubated with phenol-red free RPMI medium supplemented with 5% charcoal stripped fetal bovine serum and 17β-estradiol (100 pM) for three days. One microCi of [3H]-thymidine was added to each well three hours before the harvest. Incorporation of [3H]-thymidine into the cells was analyzed with a liquid scintillation spectrometer (Beckman Coulter).</p><!><p>The detailed protocol of xenograft tumor studies is reported previously (So et al. 2011). In brief, MCF10DCIS.com cells (one million cells per mouse) were injected in the mammary fat pad of severe combined immunodecifiency (SCID) mice. Vehicle control (0.1 mL) and gemini vitamin D compounds BXL0097 or BXL0124 (0.1 µg/kg body weight in 0.1 mL vehicle) were injected intraperitoneally daily from day three until the termination of the experiment. For in vivo animal experiments, compounds were diluted in cremophore/PBS (1:8, v/v). At autopsy, mammary tumors were measured and weighed. All animal studies were done in accordance with an institutionally approved protocol.</p><!><p>Transcriptional activities of BXL0097 and BXL0124 were evaluated and compared to the activity induced by calcitriol in HEK 293 EBNA cells transiently transfected with an expression vector encoding the full-length hVDR and a luciferase reporter plasmid encompassing the promoter region of the VDR target gene hCYP24A1. We found that VDR transcriptional activity is induced by the ligand BXL0124 at lower doses as compared to BXL0097 and calcitriol (Figure 2): this compound was able to activate transcription of the hCYP24A1 promoter at 10−12 M, which is one order of magnitude higher in comparison with BXL0097.</p><!><p>We determined the growth inhibitory effects of BXL0124 and BXL0097 on the cell proliferation of breast cancer cells, the ER-positive breast cancer cell line, MCF7 cells, and MCF10DCIS.com cells, a xenograft model of ER-negative mammary tumors that correspond to a basal-like breast tumor subtype. Both BXL0097 and BXL0124 are superior inhibitors of cell growth when compared to calcitriol, with BXL0124 being more active than BXL0097 (Figure 3). Both BXL0097 and BXL0124 showed subnanomolar ranges of IC50 values in the growth inhibition of the two breast cancer cell lines, which were 10-fold or better than that of 1,25(OH)2D3.</p><!><p>Gemini analogs have been shown to prevent estrogen-receptor positive and negative mammary tumorigenesis without displaying hypercalcemic toxicity and to suppress mammary tumor growth in mice (Wahler et al. 2014; So et al. 2011). We now have compared the effects of BXL0124 and BXL0097 on MCF10DCIS human breast cancer xenograft tumor growth in SCID mice. Animals treated with BXL0124 are normocalcemic as previously shown for BXL0097 (Lee et al. 2008) and BXL0124 (Lee et al. 2010), and showed a significant reduction in tumor volume, with the effect larger than for the animals treated with BXL0097 (Table 1), suggesting that BXL0124 is more active in tumor reduction than its 19-nor analog.</p><!><p>To characterize the molecular mechanisms underlying the enhanced activity of BXL0124, the crystal structure of zebrafish zVDR LBD in complex with BXL0124 and SRC-1 coactivator peptide was solved at 2.5 Å resolution. Data collection and refinement statistics are summarized in Supplementary Table 1. The complex preserves the general fold and canonical active agonist; the Cα atoms of the complex have a rmsd of 0.5 Å when compared to the zVDR LBD-calcitriol structure [PDB ID: 2HC4] and of 0.3 Å when compared to the zVDR LBD-BXL0097 [PDB ID: 3O1E].</p><p>Positioning of BXL0124 ligand within the LBP is identical to that of BXL0097 (Figure 4A), although the trideuteromethyl containing side arm is more loosely positioned as seen from the weak electron density map (Supplementary Figure 1) of some atoms of this side chain. The side-chain containing the trifluoromethyl groups adopts the orientation of the natural chain of calcitriol and the trideuteromethyl-containing side chain occupies the additional pocket, created by the reorientation of the side chain at Leu337, similar to the zVDR LBD-gemini complexes (Ciesielski, Rochel, and Moras 2007; Huet et al. 2011). Thus, the previously observed propensity of the LBD to accommodate the unsaturated side chain containing the trifluoromethyl groups in the parental pocket, regardless of the gemini configuration at C20, is sustained.</p><p>The hydroxyl groups of BXL0124 form similar hydrogen bonds as the zVDR LBD bound to BXL0097. More specifically, 1-OH interacts with Ser265 and Arg302, 3-OH with Tyr175 and Ser306 and the hydroxyl group of the trifluoromethyl-containing side chain with His333 or His423 (Figure 4C). Similarly to BXL0097, the trifluoromethyl groups form additional interactions explaining the increased biological activities of these gemini derivatives compared to the parental compound. These additional interactions involve Leu255 (H3), Val262 (H3), Tyr427 (H11), Leu430 (H11), Leu440 (H12), Val444 (H12) and Phe448 (H12). Although the two trideuteromethyl groups of BXL0124 adopt slightly different orientation than those of BXL0097, they engage similar contacts. The A-, seco B-, C-, and D-rings present conformations which are similar to those observed in the presence of the natural ligand. The main observed difference between BXL0124 and BXL0097 structures is the C19 methylene group which is present in BXL0124 and absent in BXL0097. The C19 atom in BXL0124 interacts with residues of helix H3, namely Leu261 (3.7 Å) and Ser265 (3.1 Å), and these interactions are not observed in the BXL0097 complex (Figure 4B). Analysis of the average temperature factors on the overall structure of the zVDR LBD in complex with BXL0124 and its comparison with the BXL0097 complex reveal stabilization of helix H3, but also additional stabilization of helix H12 and of the coactivator peptide in the presence of BXL0124 (Figure 5). While the additional contacts between fluorine atoms and hydrophobic residues of H3, H11 and H12 stabilize in a significant manner the agonist conformation of VDR and explain the superagonist potency of these compounds, the 19-methylene group on the A-ring provides additional stabilization of the agonist conformation of the complex, thus explaining the enhanced activity of this ligand.</p><!><p>In this structure-function study, we have compared two related gemini compounds, BXL0124 bearing the 19-methylene group and its 19-nor counterpart BXL0097. We have shown that both analogs are potent superagonist ligands. Both compounds contain trifluoromethyl groups introduced at the terminus of one of the side chain, and these fluorine atoms are involved in the specific additional interactions stabilizing VDR H12 and contribute to the increased transactivation properties and pro-differentiating action in cancer cells. Moreover, BXL0124 is more active than BXL0097 in the different functional assays we performed. The methylene group at position 19 on the A-ring in BXL0124 forms additional interaction that stabilize helices H3 and H12 thus further stabilizing the agonist conformation of the complex.</p><!><p>This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.</p><p>Data deposition: The atomic coordinates, structure factors, and crystallography have been deposited in the Protein Data Bank, www.pdb.org (PDB ID code: 5LGA)</p><p>Note: The atomic numbering system used refers to the conventional steroidal nomenclature.</p>

PubMed Author Manuscript

Detection of glucosamine as a marker for Aspergillus niger: a potential screening method for fungal infections

Graphical abstractSeveral species of fungus from the genus Aspergillus are implicated in pulmonary infections in immunocompromised patients. Broad screening methods for fungal infections are desirable, as cultures require a considerable amount of time to provide results. Herein, we developed degradation and detection methods to produce and detect D-glucosamine (GlcN) from Aspergillus niger, a species of filamentous fungus. Ultimately, these techniques hold the potential to contribute to the diagnosis of pulmonary fungal infections in immunocompromised patients. In the following studies, we produced GlcN from fungal-derived chitin to serve as a marker for Aspergillus niger. To accomplish this, A. niger cells were lysed and subjected to a hydrochloric acid degradation protocol. Products were isolated, reconstituted in aqueous solutions, and analyzed using hydrophilic interaction liquid chromatography (HILIC) in tandem with electrospray ionization time-of-flight mass spectrometry. Our results indicated that GlcN was produced from A. niger. To validate these results, products obtained via fungal degradation were compared to products obtained from the degradation of two chitin polymers. The observed retention times and mass spectral extractions provided a two-step validation confirming that GlcN was produced from fungal-derived chitin. Our studies qualitatively illustrate that GlcN can be produced from A. niger; applying these methods to a more diverse range of fungi offers the potential to render a broad screening method for fungal detection pertinent to diagnosis of fungal infections.Supplementary InformationThe online version contains supplementary material available at 10.1007/s00216-021-03225-7.

detection_of_glucosamine_as_a_marker_for_aspergillus_niger:_a_potential_screening_method_for_fungal_

Introduction<!><!>Introduction<!>Materials<!>Sample preparation<!>LC-MS methods<!><!>Data analysis<!><!>Analysis of GlcN standards and chitin polymers<!><!>Analysis of degradation products from A. niger<!>Conclusions<!><!>Funding<!>Conflict of interest<!>Ethics approval<!>Source of biological material<!>Statement on animal welfare

<p>Aspergillus niger is a ubiquitous species of filamentous fungi [1, 2]. Colony formation can be visualized after a few days of growth by characteristic black coloration, which is produced by conidial spores [3]. A. niger is a species of filamentous fungus which produces branching hyphae [4] and is widely studied for its ability to produce citric acid, along with industrially significant enzymes such as glucoamylase [5–7]. In addition, it is a well-known plant pathogen [8], and is responsible for black mold formation which can lead to food spoilage in human food sources such as onions and grapes [5]. Occasionally, A. niger can be an opportunistic human pathogen [9]. Inhalation of spores can lead to pulmonary aspergillosis, which usually occurs in immunocompromised patients [10–12]. Diagnosis of respiratory aspergillosis remains problematic, leading to delays in treatment and high mortality rates [13]. One of the primary structural constituents of filamentous fungi is chitin, a β-1-4-linked homopolymer of GlcNAc residues [14]. In the following studies, D-glucosamine (GlcN) produced from structural chitin in A. niger and used as a molecular indicator for the presence of fungus. Ultimately, these methods hold the potential to screen for the presence of filamentous fungi, and play a role in fungal infection diagnostics.</p><p>Chitin is a β-1-4-linked homopolymer of GlcNAc residues [14]. It is the second most abundant natural polysaccharide in the world, behind cellulose [15, 16]. Chitin is a component of crustacean and arthropod exoskeletons and is found in fungal cell walls, where it confers rigidity to its parent organism [17]. It exists in nature as nearly straight microfibrils with average diameters of ~ 2.8 nm and of indeterminate lengths [18]. Its size has been reported from ~ 100 GlcNAc residues in yeast to 5–8000 residues in crab exoskeletons [19]. While most species of yeasts contain only 1–2% chitin, its abundance in filamentous fungi ranges from 10 to 30% [14]. Chitin is found in the cell walls and septa of all species of pathogenic fungi [20]. Some fungi produce chitin deacetylase enzymes that modify chitin to chitosan during biosynthesis; however, chitosan's abundance in most fungi has not been well-defined [14, 21, 22].</p><!><p>Simultaneous chemical modifications that occur upon the exposure of chitin to HCl. Polymers are simultaneously depolymerized and deacetylated, producing a low molecular weight chemical fingerprint that consists primarily of GlcN</p><!><p>Fewer than 1% of fungal species are potentially pathogenic [23]. Fungal spores are found ubiquitously in the environment, and most individuals are exposed to large numbers of airborne fungi with no notable health repercussions. However, a minority of fungal species are extremely effective opportunistic pathogens [24]. Worldwide, over 300 million cases of serious fungal infections occur annually [25]. Fungal infections can occur superficially (e.g., athlete's foot) or as systemic infections such as candidemia or aspergillosis [26–29]. Pulmonary fungal infections comprise a significant number of systemics fungal diseases, with global incidence of > 10,000,000 patients. 1,400,000 of these result in fatalities annually [25]. A variety of genera such as Aspergillus, Cryptococcus, and Pneumocystis are responsible for these infections [30]. Chronic pulmonary aspergillosis is listed by The Global Action Fund for Fungal Infections (GAFFI) as one of four "priority fungal infections." Disease is most often caused by Aspergillus fumigatus but can also be caused by A. niger or A. flavus [30]. Chronic pulmonary aspergillosis (CPA) is most frequently seen in patients that have at one time had lung disease, such as tuberculosis, chronic obstructive pulmonary disease, or lung cancer [31].</p><p>CPA progresses rapidly in immunocompromised patients, resulting in high mortality rates and necessitating prompt treatment [32, 33]. We propose that liquid chromatography-mass spectrometry (LC-MS) can be used as a broad screening method to assist in diagnosing pulmonary fungal infections (e.g., CPA) via the detection of chitin contained within fungal cell walls. Sputum or lavage samples from patients with pulmonary fungal infections contain fungi [34, 35], facilitating noninvasive sample acquisition. Chitin may be detected via the separation and detection of GlcN monomers and oligomers that are produced during the controlled degradation of chitin. Given this, detecting polymeric chitin has the potential to serve as a screening method for pulmonary fungal infections in clinical settings. Our results are significant as they corroborate methods of chemical modification and instrumental detection which can be used in tandem to detect fungal-derived GlcN. Our instrument configuration was chosen to provide a straightforward qualitative analysis that fungal-derived GlcN can be produced from the chemical modification of A. niger and can subsequently be detected using LC-MS. Having established the efficacy of our methods, we suggest that future experiments approaching clinical applicability test our methods using an LC-MS configuration common in clinical laboratories, such as an LC-MS/MS. Eventually, we hope that these methods will be implemented clinically as part of fungal infection diagnostics.</p><!><p>A. niger fungi were obtained from domestic cat hair samples and cultured on an agar medium. D-Glucosamine hydrochloride was obtained from MP Biomedicals (Santa Ana, CA). Chitin polymer (100% acetylated) and N-acetyl-D-glucosamine (> 98.0%) were obtained from TCI (Portland, OR). Chitin polymer (20–30% deacetylated) was obtained from Alfa Aesar. ACS-grade acetone and Optima LC-MS grade formic acid were obtained from Fisher Chemical (Waltham, MA). Hydrochloric acid was obtained from Ward's Science (Rochester, NY). LC-MS grade water, LC-MS grade acetonitrile, low molecular weight chitosan (96% deacetylated), and ammonium acetate (> 98.0%) were obtained were obtained from EMD Millipore (Burlington, MA). 0.2-μm Captiva Econofilters were obtained from Agilent (Palo Alto, CA). C18 Macro spin columns were obtained from Harvard Apparatus (Holliston, MA).</p><!><p>A. niger was grown and maintained on Sabouraud dextrose agar at 22 °C until harvesting. The fungal culture was originally obtained from domestic cat hair sample and cultured on a dermatophyte test media (DTM) agar plate via standard toothbrush method. The identification of the fungi was performed using colony morphology and Internal Transcribed Spacer 2 (ITS-2) region sequencing. To obtain ITS-2 sequencing results, fungus was harvested for DNA extraction and used for conventional PCR analysis. PCR products were sent for sequencing at Psomagen and sequences were compared to entries in GenBank. A. niger cells were transferred into 15-mL centrifuge tubes in a biosafety hood. Five milliliters chilled acetone (− 20 °C) was added prior to removal from the hood. Samples were incubated for 60 min at − 20 °C. Following incubation, samples were vortexed for 30 s then centrifuged for 10 min at 15 k×g. The supernatant was removed, and samples were allowed to stand for 30 min to encourage the evaporation of residual acetone. Liquid nitrogen was added directly to the dried pellets to lyse cells. After 5 min, 5 mL of − 20 °C acetone was added and the pellet vortexed again for 30 s. The supernatant was decanted, and the pellet was allowed to evaporate for 30 min. Four milliliters 10 M HCl was added to the dried pellet. The pellet was vortexed, and the resulting suspension was transferred to a round-bottom flask. The flask was heated in a water bath held at 90 °C and the acid evaporated over ~ 6 h. Following evaporation, the dried material in the flask was reconstituted using Milli-Q grade water (18.2 MΩ•cm). Aliquots were drawn from the flask and passed through 0.2-μm filters. The eluent was added to C18 spin columns and centrifuged for 4 min at 2 k×g. The filtrate was collected and analyzed via LC-MS.</p><!><p>LC-MS analyses were carried out on an Agilent 6224 time-of-flight mass spectrometer coupled to an Agilent 1260 binary liquid chromatograph (Agilent, Palo Alto, CA). Separations were performed using a Thermo Fisher Acclaim HILIC-10 column with dimensions of 4.6 × 150 mm and 5-μm particle sizes. The mobile phase was composed of LC-MS grade water and acetonitrile (ACN). Each mobile phase component contained 10 mM ammonium acetate and 0.05% formic acid with a final pH 4.0. The total length of chromatography runs was 50 min. Solvent flow rate was set to 0.350 mL/min for the duration of the separation. A gradient elution was used with initial solvent proportions of 90:10 ACN:H2O. The solvent ratio was adjusted to 80:20 ACN:H2O from 0 to 30 min. From 30 to 31 min, the solvent ratio was returned to 90:10 ACN:H2O. From 31 to 50 min, the column was allowed to re-equilibrate and the baseline stabilize. The ion source used was a dual electrospray ionization source operating in positive ionization mode. Ion source conditions were as follows: 3500 V capillary voltage, 120 V fragmentor voltage, 60 V skimmer voltage, 250 V octupole voltage, 10 L min−1 gas flow (N2) at 300 °C, and 45 psig nebulizer pressure. The detection range was set to 95–3200 m/z.</p><!><p>Polymers used for degradation studies</p><p>aAlfa Aesar certificate of analysis; bTCI certificate of analysis</p><!><p>All data were analyzed using Agilent MassHunter Qualitative Analysis B.07.00 software. ESI optimization was performed prior to these experiments and a library of potential degradation products was generated to enable extracted ion chromatogram (EIC) scans for data deconvolution. Degradation experiments performed on A. niger samples were compared to the LC-MS results for chitin polymer degradations. Retention times and accurate mass measurements were both accounted for to provide a two-step validation confirming the identity of degradation products.</p><!><p>Ions observed in MS extractions from LC-MS of GlcN standards</p><!><p>While it was expected that the most prevalent ion would be a protonated molecule at 180 m/z, the most abundant ion in mass spectra was shown at 381 m/z. To the best of our knowledge, dimerization has not been previously reported in ESI analysis of GlcN; however, [2 M + H]+, [2 M + Na]+, and variations of these dimers have been observed in other compounds [36, 37]. Dimerized ESI ions are proposed to result from noncovalent interactions between residues in solution, as Coulombic barriers preclude dimerization following electrospray ionization [38]. One hundred sixty-two, 180, and 202 m/z are commonly seen ions in ESI analysis of GlcN. 162 m/z occurs following the dehydration and protonation of GlcN and has previously been observed in the ESI analysis of GlcN [39]. 180 m/z is a protonated GlcN molecule for which formation is promoted by the acidic pH in the mobile phase used. 202 m/z represents a sodiated adduct.</p><!><p>Generic structure of chitin (> 50% GlcNAc) showing GlcN (x) and GlcNAc (y) subunits</p><p>Representative chromatogram of chitin polymer 1 degradation products. Three peaks contained m/z that indicated the presence of the degradation products of chitin. Unlabeled peaks did not contain m/z that correlated to chitin degradation products</p><p>Summary of peaks and the ions contained within these</p><p>Chromatogram of chitin polymer 2 degradation products. Two peaks contained m/z that indicated the presence of chitin degradation products. Remaining peaks did not contain m/z representative of chitin degradation products</p><p>Summary of peaks and the ions contained within these</p><p>Chromatogram of A. niger degradation products. Analytes of interest eluted in a broad peak at ~ 24 min. The inset shows an extracted ion chromatogram for the monoisotopic mass of GlcN, 180.08</p><p>Overlay of EICs from degradation experiments. The value for EIC scans was set to 180.08 to highlight the elution of protonated GlcN. The black trace represents GlcN produced from chitin polymer 1. The blue trace represents GlcN produced from chitin polymer 2. The green trace represents GlcN produced from A. niger fungi</p><p>MS overlays from LC-MS analysis of polymer and fungal degradations. MS extractions from chitin polymer 1 are shown in black. MS extractions from chitin polymer 2 are shown in blue. MS extractions from A. niger are shown in green</p><!><p>Extractions of mass spectra from chromatographic peaks at ~ 24 min all displayed m/z at 180, indicative of ions with formulas matching those of protonated GlcN molecules, [C6H13NO5 + H]+. The degradation of both chitin polymers displayed a m/z at 202, indicative of ions with formulas matching those of sodiated GlcN adducts, [C6H13NO5 + Na]+. The mass spectrum of A. niger contained a m/z at 202; however, its abundance was very low. The degradation products from chitin polymers also displayed a m/z at 162, indicative of ions with formulas matching those of protonated dehydrated GlcN, [C6H13NO5 - H2O + H]+. While the mass spectrum of A. niger contained a m/z at 162, its monoisotopic mass varied from that expected from a protonated dehydrated GlcN ion. Closer examination revealed that the peak at 162.1109 m/z was split, suggesting that two molecules with similar but not identical m/z ratios were present in mass spectra. Greater resolving power would be required necessary to distinguish between these peaks.</p><p>Further studies to differentiate between endogenous and exogenous GlcN derived from chitin will be necessary prior to implementation of these methods in clinical facilities. Separating GlcN from these sources could be done using several approaches: by aqueous liquid extraction, by sub-micron filtration, or by dialysis. In addition, centrifugation has been shown to be an effective means of concentrating fungal spores [40]. Regardless of the approach chosen, a crucial aspect of separating endogenous and exogenous GlcN is that endogenous GlcN is removed prior to the production of GlcN from chitin, thereby reducing the likelihood of false positives. Use of an aqueous extraction step prior to the degradation of chitin would remove endogenous GlcN, given its high solubility. Alternatively, isolating fungi immediately following sample acquisition could be accomplished by sub-micron filtration, dialysis, or centrifugation and recovery of spores and other insoluble species prior to their exposure to HCl. The detection of GlcN by LC-MS/MS systems has been performed in multiple studies, with limits of quantitation at or below 10 ng/mL [41–43]. For our qualitative application, these values indicate limits of detection for GlcN in biologically relevant matrices as low as ~ 3 ng/mL for clinically relevant LC-MS systems. Ultimately, we hope that the methods presented herein may serve as a key component in the rapid detection of fungi.</p><!><p>In our studies, we used LC-MS to detect GlcN from fungal-derived chitin. To accomplish this, A. niger fungi were obtained and exposed these to cell lysis and sample cleanup steps. Lysed cells were exposed to previously developed degradation protocols using HCl. The acidic solution was evaporated, and water-soluble analytes were resuspended into an aqueous suspension. These were filtered and analyzed using HILIC-ESI-MS. LC-MS analysis indicated the presence of a chromatographic peak with a retention time matching that of GlcN produced from the degradation of chitin. Mass spectral extractions of this peak provided secondary confirmation of its identity, showing a m/z at 180.08, matching that of a protonated GlcN molecule. Chitin degradation methods were developed to degrade fungal-derived chitin and to produce GlcN. The analyte produced from fungal degradations matched those generated during the analysis and comparison of degradation protocols using several variations of chitin and chitosan. Retention times and accurate mass measurements were compared to validate that chitosan polymers, chitin polymers, as well as A. niger produce GlcN following exposure to HCl. Cumulatively, the novelty of our results lies in our combination of chemical modification methods and analytical detection methods (HILIC-ESI-MS) to produce and detect GlcN from A. niger. By applying reproducible methods to detect a species of fungus implicated in pulmonary fungal infections, our studies address a clinical problem using sound analytical chemistry. Ultimately, we hope these methods will be implemented into clinical labs for purposes of fungal diagnostics.</p><!><p>(DOCX 818 kb)</p><p>Publisher's note</p><p>Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.</p><!><p>C.A. was supported by the National Institutes of Health (NIH 5R01HL140301-02). A.M was supported by Biomedical Research Training for Veterinarians (NIH 9-T32 OD0-10437-18), NIH/NCATS Colorado CTSA (TL1 TR002533), and EveryCat Health Foundation (W20-032).</p><!><p>The authors declare no conflict of interest.</p><!><p>Not applicable; no human/animal studies were performed during the course of our experiments.</p><!><p>The source of cells used is covered in Section 2 Materials and methods under Materials.</p><!><p>Hair collection from shelter-owned cats has been approved by Colorado State University's Clinical Review Board and has been granted an Institutional Animal Care and Use Committee (IACUC) waiver. The veterinarians at the collaborating shelters have given consent for hair sample collection from their shelter-owned cats and performed the collections themselves.</p>

PubMed Open Access

Non-iterative Method for Constructing Valence Antibonding Molecular Orbitals and a Molecule-adapted Minimum Basis

While bonding molecular orbitals exhibit constructive interference relative to atomic orbitals, antibonding orbitals show destructive interference. When full localization of occupied orbitals into bonds is possible, bonding and antibonding orbitals exist in 1:1 correspondence with each other. Antibonding orbitals play an important role in chemistry because they are frontier orbitals that determine orbital interactions, as well as much of the response of the bonding orbital to perturbations. In this work, we present an efficient method to construct antibonding orbitals by finding the orbital that yields the maximum opposite spin pair correlation amplitude in second order perturbation theory (AB2) and compare it with other techniques with increasing the size of the basis set. We conclude the AB2 antibonding orbitals are a more robust alternative to the Sano orbitals as initial guesses for valence bond calculations, due to having a useful basis set limit. The AB2 orbitals are also useful for efficiently constructing an active space, and work as good initial guesses for valence excited states. In addition, when combined with the localized occupied orbitals, and relocalized, the result is a set of 1 molecule-adapted minimal basis functions that is built without any reference to atomic orbitals of the free atom. As examples, they are applied to population analysis of halogenated methane derivatives, H-Be-Cl, and SF 6 where they show some advantages relative to good alternative methods.

non-iterative_method_for_constructing_valence_antibonding_molecular_orbitals_and_a_molecule-adapted_

Introduction<!>Defining the set of antibonding orbitals<!>Population analysis using the effective minimal basis<!>Implementation details<!>Computational details<!>Results and discussion<!>Orbitals, orbital Energy, and orbital variance<!>CCVB iterations<!>CAS methods<!>Excited States<!>Population Analysis<!>Conclusion

<p>Virtual orbitals are important in chemistry as they play a central role in molecular orbital theory. From a computational standpoint, orbital mixing between occupieds and virtuals determines the optimal occupied orbitals in mean-field Hartree-Fock theory [1][2][3] and Kohn-Sham density functional theory. [4][5][6][7] In wavefunction theory, electron correlation is typically described by amplitudes such as the pair correlations describing the simultaneous promotion of two electrons from occupied to virtual orbitals. The virtual orbitals span the unoccupied space, and the choice of representation is important. Canonical virtual orbitals are delocalized levels that are appropriate for electron attachment. Localized virtuals, such as the redundant non-orthogonal basis of atomic orbitals projected into the virtual space, 8,9 permit development of efficient local correlation methods, because the amplitude tensors describing correlation become sparse. 10 Other prescriptions for localized orthogonal virtuals exist, [11][12][13] as well as proposals to form sets of virtuals that are specifically optimized for correlations that involve a given occupied level, as will be discussed below.</p><p>The virtual orbitals span the entire unoccupied space, which can be contrasted with the intuitive notion of antibonding orbitals that exist in 1:1 correspondence with bonding orbitals. The 1:1 correspondence is evident from constructive and destructive interference of a pair of 1s-type functions on two hydrogen atoms in H 2 :</p><p>Antibonding orbitals themselves play a central role in describing chemical reactivity [14][15][16][17][18][19] of one molecule with another through donor-acceptor interactions between a high-lying occupied of one species with a low-lying antibonding orbital of the other. Frontier orbital theory is constructed on these ideas. Antibonding orbitals also play an important role in describing strong electron correlations. A simple example is the stretching of the H-H bond which leads, in a minimal basis, to a strong increases in the amplitude for σσ → σ σ excitation which breaks the bond.</p><p>While the antibonding orbitals are intuitive, [16][17][18]20 it is nonetheless not routine to extract them from modern quantum chemistry calculations performed in extended basis sets, which return canonical orbitals. By contrast, in a minimal basis description of hydrocarbons, the space of antibonding orbitals is naturally spanned by the canonical virtual orbitals. In larger basis sets however, different methods have been developed to extract the antibonding orbitals, often by relying on projection back onto some chosen minimal basis, 13,[21][22][23][24][25] typically a tabulated one for a specific free-atom Hartree-Fock energy eigenstate. For example, Schmidt et. al. found antibonding orbitals by performing an SVD of the overlap between the virtual orbitals and a minimal basis to produce valence virtual orbitals. 13 Some methods have been developed to produce a minimal basis specifically adapted to a molecular environment, 26,27 but those are non-linear optimization procedures that are often iterative and costly. One famous method that does not rely on a reference minimal basis is the Natural Bond Orbital (NBO) procedure, 14,15 where the density matrix coupling between multiple atom-tagged orbitals is utilized to produce bonding and anti-bonding orbitals. However, atom tagging of basis functions plays a critical role in the NBO procedure -in fact, the standard NBO method is specific to atom-centered orbital (AO) basis calculations. Few methods cut the umbilical cord to the minimal basis in producing antibonding orbitals.</p><p>Aside from the Sano antibonding orbitals 28 (discussed below), Foster and Boys 29 suggested oscillator orbitals which are virtual orbitals with the maximum dipole from localized occupied orbitals.</p><p>Local correlation has been intensively studied, 8,9,25,[30][31][32][33][34][35][36][37][38] leading to the conclusion that dynamic correlation can be well approximated using domains of localized virtual orbitals that are in the same spatial region as a localized occupied orbital. 8,9 This reduces the 4th rank tensor of pair correlation amplitudes to an asymptotically linear number of significant elements. Nevertheless, all virtual orbitals are required for post-SCF methods such as coupled cluster theory that recover dynamic correlation, rather than just the much smaller set of valence virtual orbitals. By contrast, static or strong correlation, resides mostly in the valence virtuals (i.e. the antibonding orbitals). Thus complete active space (CAS) methods that seek to describe strong correlation require only a description of the valence virtuals. Methods in this class include CASSCF, 2,39-42 spin-coupled valence bond (VB), [43][44][45][46] and approximations such as generalized valence bond (GVB), 47 coupled cluster valence bond (CCVB), [48][49][50] etc.</p><p>CAS, GVB and CCVB methods thus need an initial guess for the antibonding orbitals.</p><p>We do note that the orbitals associated with key amplitudes for strong correlation are not necessarily spatially localized. [51][52][53][54] One method used to obtain initial guess antibonding orbitals is the so-called Sano procedure. 28 In brief, after localizing a set of occupied orbitals using standard methods, 12,[55][56][57] the Sano procedure finds the virtual orbital that has maximum exchange interaction with each given localized occupied orbital. The idea of maximizing exchange is very old 58,59 and comes from its predecessor, the modified virtual orbitals 60,61 (note that modified virtual orbitals have been since used to refer to any non-canonical set of virtual orbitals 62 ). The resulting orbitals are symmetrically orthogonalized to yield a set of valence antibonding orbitals.</p><p>This method has worked quite well for GVB-PP and CCVB calculations in moderately sized basis sets. [49][50][51]63,64 In this work we will show that the Sano procedure shows undesirable behavior with increasing the size of the AO basis set. This motivates the need for a better behaved alternative. We suggest that finding the antibonding orbital which gives the largest first order perturbation amplitude for exciting an electron pair from a given bonding orbital is a suitable alternative. A range of numerical results confirm this to be the case. These antibonding orbitals can be viewed as a specific instance of orbital specific virtuals. [30][31][32] 2 Theory</p><!><p>Solving the mean field Hartree-Fock (HF) equation self consistently gives the lowest energy single Slater determinant electronic wave function. To solve the many-body problem, one needs to include the missing correlation energy. 65 Second order Møller-Plesset (MP2) perturbation theory 66,67 offers a useful and computationally inexpensive approximation to treat the correlation yielding the following expression in the case of restricted HF orbitals:</p><p>where</p><p>This expression folds together contributions from the correlation of two electrons of opposite spin (OS), with amplitudes:</p><p>together with the contribution of correlations of electrons with the same spin. The twoelectron repulsion integrals (ERIs) over spatial orbitals describing the interaction of each occupied with each virtual are:</p><p>Let us collect the ERIs associated with occupied orbital i into the symmetric matrix K i , where:</p><p>K i is positive semi-definite, and thus the eigenvector belonging to its largest eigenvalue will correspond to the virtual level with the strongest exchange interaction with occupied level i. That is the Sano prescription 28 for finding the antibonding orbital associated with i.</p><p>We can likewise define a matrix of second order pair correlation amplitudes, T i , associated with a given occupied orbital:</p><p>This matrix is negative semi-definite since the denominators are negative for the ground state determinant. We can therefore find the largest OS pair-correlation amplitude as the lowest eigenvalue, t i max of T i , and the associated virtual orbital, |i * = a |a c ai * is the eigenvector, with expansion coefficients c ai * in the original virtual basis:</p><p>Upon repeating for each occupied level, most naturally in a localized representation, and using similar arguments to Kapuy's zeroth in the Fock and 2nd order in correlation approximation, 68,69 we suggest that this is an appropriate non-iterative way to find a set of antibonding orbitals in 1:1 correspondence with the bonding orbitals. This approach may be contrasted with Sano's suggestion to obtain the virtual orbital with maximum repulsion from the bonding orbital by solving the eigenvalue problem for each orbital using K i rather than T i . Inclusion of orbital denominators in Eq. 9 provides a clear physical meaning of the antibonding orbital as having strongest pair correlation amplitude with its parent bonding orbital. As will be demonstrated numerically later, this property also dramatically improves basis set convergence relative to the Sano definition.</p><p>We will refer to these virtual orbitals as "second order antibonding" (AB2) MOs to emphasize their second order origins, and their 1:1 correspondence with bonding MOs. In terms of existing literature, the AB2s are directly related to the "orbital-specific virtual" (OSV) orbitals 30,31 that are sometimes used to evaluate the correlation energy. Each AB2 orbital is the most important OSV for a given localized bonding orbital. Of course the reason for selecting the amplitudes associated with MP2 is computational efficiency. The exact limit of this procedure would be to diagonalize the corresponding exact (i.e. from Full CI) doubles amplitudes; T i ab , via Eq. 9.</p><p>A closely related alternative that has some advantages over Eq. 9 above is to define the space of valence antibonding orbitals from the virtual-virtual block of the MP2 one-particle density matrix: 70,71</p><p>Upon diagonalizing, the (M − O) eigenvectors with largest occupation numbers span the valence antibonding orbital space, and, together with the occupied space, complete the span of a molecule-adapted minimal basis. Localization of these valence virtual orbitals will then yield an alternative to the localized virtuals above. The advantage of this approach is for cases where there is no simple 1:1 mapping between bonding and antibonding orbitals, as discussed more later.</p><p>The virtual orbitals obtained this way are the valence subset of the "frozen natural orbitals" (FNO), 70,71 and we emphasize that they are not generally localized in contrast to the AB2 MOs. They are close to the virtual natural orbitals associated with P MP2 as defined by the gradient of the MP2 energy, [72][73][74] with the caveat that only the virtual-virtual block is diagonalized.</p><!><p>Finding a suitable set of antibonding orbitals provides the missing part of the valence space not spanned by the occupied orbitals. Thus the union of the occupied space and the space of antibonding orbitals spans the space of an effective minimal basis. It is well accepted that full valence CASSCF wavefunction is spanned by an effective minimal basis within the molecule for this reason. 39,40 Accordingly, localizing the union of the occupied orbitals with the antibonding orbitals reveals a set of molecule-adapted atomic orbitals (MAOs): 75,76</p><p>For a given pair of well-localized bonding and antibonding orbitals (say σ and σ * ), this procedure amounts to inverting Eqs. 2 to discover the corresponding MAOs even though we may be using a very extended basis, or even a non-atom-centered basis, such as plane waves or a real-space grid, to perform the calculations.</p><p>The resulting MAOs, χ are thus expressed in terms of the AO's, ω, as χ = ωC MAO . The MAOs are orthogonal, and typically localize onto atoms. The MAOs exactly span the space of the occupied orbitals, and can be used for population analysis among other things. 26,40,[77][78][79][80][81][82][83] Let us denote p as an MAO label for χ p , which is centered at r p = χ p |r|χ p . Using A, B as atom labels, and given that the density matrix in the MAO basis is</p><p>one can make a population analysis as follows:</p><p>where Q A and Z A are the atomic charge and the nuclear charge, respectively. Such a population analysis has no dependence on atom-tagging of the underlying basis, and does not rely upon a reference minimal basis. Therefore it generalizes nicely to plane wave basis and real space methods. If the orbitals are unrestricted, we construct antibonding pairs and the MAOs for the alpha and beta spin spaces independently.</p><p>This approach to generating an MAO representation does have some limitations. First, it assumes that there is a 1:1 mapping between bonding and antibonding orbitals. One class of exceptions can be found in electron deficient molecules (e.g. LiH will not recover 2p-like orbitals on Li, and BH 3 will not recover a 2p z orbital on B). Such species can be said to have "virtual lone pairs", whose identification is a problem that we shall not address here.</p><p>A second class of exceptions lie in species such as cyclopentadiene anion, where there are 3 semi-localized π occupied orbitals, but the valence space only admits 2 antibonding orbitals.</p><p>Thirdly, in symmetric systems with multiple Lewis structures (e.g. C 6 H 6 ), the MAOs will derive from localized bond and antibonding orbitals corresponding to a single Lewis structure and may not reflect the indistinguishability of the atoms. Broadly, we can say that this MAO approach is readily applicable to neutral molecules with a single dominant Lewis structure.</p><!><p>Computational efficiency is very important for quantum chemistry in order treat molecules that are as large as possible for given computational resources (computer speed, memory size, etc). Our AB2 implementation uses exact 4-center integrals in a basis of Gaussian-type atomic orbitals (other alternatives such as using auxiliary basis expansions can also be readily implemented). Each step with its computational complexity is shown in Fig. 1. Note that for the figure and the discussion here we use O, V , and N for the number of bonding orbitals, virtual orbitals, and AO basis functions, respectively. We start by making a pseudo-density</p><p>To generate the two-electron integrals (µν|λσ), Q-Chem 84 only generates significant µν (i.e. AO basis) pairs to some target numerical cutoff, yielding a total that we term as</p><p>for small systems but approaches linear scaling (i.e. (N N ) cut ∝ N ) in the limit of large system size. The integrals are made and contracted on-the-fly with the bonding orbitals' pseudo-densities to make bonding-specific exchange integrals K i µν with compute effort scaling as O(O(N N ) 2 cut ). The K i µν matrices are then transformed into the virtual space as K i ab in Eq. 7 with compute cost scaling as O(OV N 2 + OV 2 N ). Asymptotically this is the dominant step in this method unless more careful thresholding is considered. 85 Then, we divide by the appropriate denominator to get T i ab in Eq. 8 (with O(OV 2 ) effort). Lastly, we diagonalize T i for each bonding orbital to get the AB2 antibonding orbitals as in Eq. 9 with O(OV 3 ) effort. Note that the last step can in principle be made O(OV 2 ) since we are only solving for the eigenvector with the largest amplitude in each matrix. We can contrast this procedure with the modified FNO approach which has a dominant computational step that scales as the 5 th power of molecule size: constructing P ab in Eq. 10 with complexity of O(O 2 V 3 ).</p><p>Figure 1: A chart illustrating the mathematical steps needed to construct AB2 orbitals with the appropriate computational complexity for each step indicated. Here, O, V , N , and (N N ) cut refer to the number of occupied orbitals, virtual orbitals, AO basis functions, and significant AO pairs, respectively.</p><p>One reason for the efficiency of the AB2 approach compared to FNO comes from focusing on the bonding orbitals one at a time rather than the whole occupied space at once. It is then important to start by localizing the occupied space, which is known to be a cubic scaling iterative procedure for e.g. the Boys and Pipek-Mezey localization measures. 86,87 Then, one must also distinguish between localized orbitals with different character: specifically core, bonding, and non-bonding, e.g. lone pairs. Our implementation uses an automatic bonding detection option that runs before AB2. The detection process is simply determined by Pipek's delocalization measure 88 on Mulliken charges, where measures amounting to 1 indicate an orbital localized on an atom (core or non-bonding) and measures around 2 correspond to orbitals split between two atoms.</p><!><p>All methods discussed here were implemented in a developer version of Q-Chem 5. 84 The geometries used for molecular calculations were optimized at the ωB97X-D/def2-TZVPD level of theory. All geometries are included in the Supplementary Material (SI).</p><!><p>We will compare different approaches to generating effective antibonding orbitals: in particular we are interested in whether the second order antibonding (AB2) MOs significantly improve upon the Sano antibonding orbitals, as measured by usage-relevant metrics obtained from a set of numerical experiments. We will first examine orbital plots, orbital energies, and orbital variances. We then test the applicability of Sano and AB2 MOs to several valence correlation methods: coupled cluster valence bond (CCVB), 49,50 complete active space configuration interaction (CASCI), [89][90][91][92] and complete active space self-consistent field (CASSCF). [39][40][41][42] Next, we look into their uses for describing valence excited states. For basis set, we are using the Dunning basis set family 93 and Ahlrichs. 94 These are available in Q-Chem 5.3 with an automated detection of bonding orbitals.</p><!><p>We start by looking at the σ * orbital of H 2 , as shown in Fig. 2, evaluated by the Sano procedure, the AB2 approach, and CAS(2,2) (performed as 1-pair perfect pairing). It is visually clear that the Sano σ * orbital is contracting as the basis set is improved. Fig. 3 displays the orbital energy (diagonal matrix element of the Fock operator) and the variance ( r 2 − r 2 ) of the σ bonding orbital, and the Sano and AB2 models of the antibonding orbital. The variance confirms that the size of the Sano σ * -orbital contracts with basis size, while its orbital energy increases (reflecting increasing electron confinement) unsatisfactorily.</p><p>By contrast the behavior of the AB2 orbital is very close to the bonding orbital, with pleasing stability in both energy and variance as the basis set is converged towards completeness.</p><p>The stark difference is due to Sano orbitals including high energy orbitals to maximize the exchange interaction whereas AB2 biases against those higher energy orbitals with the denominator penalty in Eq. 8. orbitals. We will therefore use Pipek-Mezey orbitals whenever we encounter π orbitals.</p><p>Inspecting the σ C-C bond in C 2 H 4 in Fig. 4 shows that the shape of the occupied Pipek-Mezey and converged CCVB bonding orbitals both do not change much upon increasing the size of the basis set. By contrast, when looking at σ * in Fig. 4 we see even poorer behaviour of the Sano C-C antibonding orbital as a function of basis set size than we did for H 2 . This is confirmed in Fig. 5 where we compare the orbital energy and the orbital variance of the bonding and the antibonding C-C σ orbital in C 2 H 4 . The Sano σ * orbital does not converge with the size of the basis set, with the variance decreasing, and the energy increasing. By contrast, the AB2 σ * orbital converges rapidly both in terms of energy and variance for similar reasons to before. In Fig. 5 we compare the orbital energy and the orbital variance of the bonding and the antibonding orbitals for the C-C π in C 2 H 4 . The shortcomings of Sano seem to be much less severe in π * orbitals. We believe this is due to the diffuse nature of the π orbitals making the maximum exchange, thus spatial locality, sufficient to describe the π * . However, we can still see that the orbital energy and variance do not converge for Sano while they do for AB2, and converge to drastically different orbital energy and orbital variance. The quantitative advantage of the AB2 antibonding orbitals relative to the Sano orbitals seen so far can also become qualitative advantages in systems with more complex electronic structure. One such example is Cu 2 , which, considering that the valence state of Cu can be taken as 3d 10 4s 1 , is isoelectronic to H 2 . The σ orbital (HOMO) of Cu 2 is shown in the upper panel of Fig. 6, along with the optimized correlating orbital from CCVB, as well as the Sano and AB2 antibonding orbitals. Maximizing exchange results in a Sano antibonding orbital that resembles an empty π-bond between the two metals. By contrast, the AB2 and CCVB orbitals look qualitatively identical.</p><p>Figure 6: Comparison of the shape of the orbitals in Cu 2 where the σ bond is used to produce Sano and AB2 antibonding orbitals. While the AB2 method produces very similar orbitals to CCVB, the Sano approach fails to give a qualitatively correct antibonding orbital.</p><!><p>The CCVB method is a simple low-scaling approximation [49][50][51] to exponentially scaling spincoupled valence bond theory that can separate a system of 2n electrons into fragments with spin purity, provided that UHF can also reach the dissociation limit. One price to be paid for these advantages is a challenging orbital optimization problem: the CCVB orbitals have no invariances to rotations within the active space, in contrast to CASSCF. Hence a good initial guess is very important. Sano orbitals 28 have been commonly as a starting guess for valence bond methods 51,95 such as CCVB due to their resemblance to antibonding orbitals. For simple alkanes, we examine how many iterations are needed to converge a CCVB calculation with Sano and compare with AB2 shown in Fig. 7 with increasing the molecule size and the basis set size (using the Dunning cc-pVXZ sequence of basis sets 93 ). Since the doublezeta basis set does not involve many high energy orbitals, both methods converge almost at the same speed. Upon increasing the size of the basis set, overly-contracted Sano orbitals deviate more from the optimal antibonding orbitals, and therefore require far more iterations to converge. For this reason, we recommend using AB2 orbitals as a starting guess for valence bond methods instead of the Sano orbitals.</p><p>Figure 7: Number of iterations needed to converge CCVB calculations on alkanes of increasing size, with increasing ζ of the basis set. This shows a relatively constant number of iterations needed for AB2 regardless of system size, while the number of iterations rise unfavorably for the Sano guess in large basis sets. Geometric direct minimization (GDM) 95,96 is used to determine the steps.</p><!><p>The relative fraction of correlation energy recovered using AB2, Sano, FNO or other choices for antibonding orbitals to complete an active space can help us discern which ones are most appropriate to use for configuration interaction with fixed orbitals, as well as for a CASSCF initial guess. As a simple example, we stretch the C-C bond in C 2 H 4 while keeping the geometry of the methylene groups fixed at those of the equilibrium ground state geometry of ethene. Looking at Fig. 8, we see that canonical virtual orbitals capture less and less correlation as the def2 basis set is improved from SVP to TZVPP to QZVPP. We also observe that the gap between Sano and AB2 orbitals increases with increasing the size of the basis set. Finally, we can see that FNO and AB2 orbitals perform almost identically and are the Since the AB2 and FNO orbitals seem to capture quite a lot of the static correlation, we sought to compare them to CASSCF orbitals. In Fig. 9 we are comparing the smallest singular value of the overlap matrix between the CASSCF orbitals and those of canonical, Sano, AB2, and FNO, at the optimized geometry of C 2 H 4 . Once again, the canonical orbitals become dramatically worse with increasing the basis set size. Sano and FNO both become very slightly worse with increasing the size of the basis set, namely by increasing zeta, while AB2 seems to be nearly basis set-independent.</p><p>Figure 9: The smallest singular value from the overlap of CASSCF (12e,12o) orbitals with those from Sano, AB2, FNO and canonical orbitals. Canonical orbitals with the lowest energy and FNOs with the highest occupancy were selected. Canonical orbitals are differ strongly from optimized CASSCF orbitals while AB2 orbitals have the highest agreement.</p><!><p>Since the AB2 orbitals seem to be good guesses for GVB methods, and yield orbitals close to converged CASSCF orbitals, this led us to believe that they could also provide a good description of valence excited states. State-specific methods, such as orbital-optimized DFT (OO-DFT) 97 need a suitable starting guess, as convergence is typically to the nearest stationary point., 98 so we used Sano and AB2 guesses for the π → π * excitation in methanal (H 2 CO). For our purposes we employed the square gradient minimization method 98 which looks for saddle points in the orbital Hilbert space to converge restricted open-shell Kohn-Sham (ROKS). 97,99,100 In Fig. 10 we compare the overlap of the π * orbital from converged singlet open shell HF calculations with Sano and AB2 orbitals. For this excitation, AB2 orbitals overlap the optimized orbital by at least 0.9, and vary minimally with the size of the basis set. We note here that aside from the double-zeta case, converging the excited state starting from the Sano orbital sometimes lands on a Rydberg excited state, while AB2 landed on the correct π * state in all cases.</p><p>Figure 10: The overlap of the converged ROKS-HF antibonding orbital with the Sano and AB2 initial guesses in H 2 CO for the π → π * excitation. The π * orbital is well described by AB2 regardless of basis set size.</p><!><p>Antibonding orbitals belong to the valence space, and contribute to making a minimal basis that can be used to gain insight into chemistry, for instance via population analysis to assign effective charges on each atom. The population analysis we present here is constructed from the union of the occupied space and the antibonding orbitals without dependence on the basis set used. To study our atomic charge predictions and compare it to some other methods in the literature, we look into fluoro-and chloro-substituted methanes which have been studied theoretically [101][102][103] and experimentally. 104,105 These simple systems are nonetheless interesting because they manifest the effect of substituting electron withdrawing halogen atoms of different sizes and electronegativities for hydrogen in methane. How consistent or inconsistent are different atomic population analysis schemes as descriptors of these chemical substitutions?</p><p>In Fig. 11 we examine the effect of progressive substitution of hydrogen by chlorine and fluorine in the methane molecule on the computed net charge at the C atom. We consider some commonly used methods, specifically charges on electrostatic potential grid (ChElPG), 106 iterative Hirshfeld (Iter-Hirsh), 107,108 intrinsic atomic orbitals (IAO), 109,110 and the method presented in this work, molecular atomic orbitals (MAO). Most obviously, the charge transferred upon halogen substitution will depend strongly on the electronegativity difference between X and H. Furthermore, while halogens are more electronegative than hydrogen (or carbon), the electron donating capacity of C is not unlimited, and so we expect the first halogen substituted to pull away a greater fraction of an electron from C compared to the next, and so forth. Such a change will also have some dependence on the X vs H electronegativity difference. With these preambles aside, atomic charges are not observables and therefore no single answer should be viewed as strictly correct. Nevertheless, we can examine the results of each population analysis for signs of incorrectness relative to physical intuition.</p><p>Figure 11: The charge on the carbon atom for successive chlorination and fluorination of methane predicted using four different population analysis methods (see text for the names). The triangle, square, hexagon, and octagon correspond to charges using def2-SV(P), def2-SVPD, def2-TZVPD, and def2-QZVPD, respectively.</p><p>For instance, while all methods agree that the C -H bonds of CH relative to the slight positive charge predicted by IAO.</p><p>Finally we examine an unusual linear molecule, which is the result of insertion of Be into HCl, yielding H -Be -Cl. 111,112 While H -Cl is polarized as H δ+ Cl δ− , Be has lower electronegativity than H, and so there will be substantial charge transfer. Indeed the ionic limit would be H -Be 2+ Cl -. What then, is the actual charge distribution when we consider the covalent character of the molecular orbitals? The calculated populations are shown in Fig. 12, and it is immediately evident that predicted charges on Be vary widely. The least polar picture comes from ChElPG and MAO, with q(Be)≈+0.5, while the IAO scheme suggests q(Be)≈+1. 35. How should we understand this dramatic difference and suggest which might be more correct?</p><p>Figure 12: The charges on each atom in the BeHCl molecule predicted by the four methods mentioned in the text. The triangle, square, hexagon, and octagon correspond to charges using def2-SV(P), def2-SVPD, def2-TZVPD, and def2-QZVPD, respectively.</p><p>From the MAO perspective, there are two σ bonds involving Be, one with H and one with Cl. Each is made from sp hybrid orbitals on Be, meaning that the p orbitals of Be are at play in this σ bonding, as shown in Fig. 13. These bonds are both polarized away from Be, as expected. The origin of the much larger IAO charge can now be understood. The IAO reference minimal basis set, known as 'MINAO', 109 does not include 2p orbitals for Be, and therefore we are instead seeing essentially only the Be(2s) charge via the IAO approach! Iterative Hirshfeld evidently struggles with a similar issue, leading to similar overestimation of Be charge. Overall, this case nicely illustrates the advantages of the MAO population scheme that is based entirely on the system at hand, rather than some reference atomic orbitals or states. Next we study the hypervalent molecule, SF 6 , which has O h symmetry, and whose chemical bonding has long been of interest. 113 While empty 3d functions on sulfur are needed to form 6 equivalent sp 3 d 2 hybrids, the energetic cost of promoting electrons to the 3d shell is too high for d-orbital participation in the bonding to be chemically important. 46,[114][115][116][117] Rather, the bonding may be thought of as resonance between Lewis structures with 4 covalent S-F bonds, and 2 F − anions, with a formal charge of +2 on S. 118 There are some limitations associated with reducing the union of the localized occupied orbitals and the AB2 antibonding orbitals to a set of MAOs that should be mentioned. First, some conjugated π systems, such as benzene and C 5 H 5 -, present a multiple minimum solution problem for orbital localization methods. Since our method relies on the localization procedure heavily, we expect there will be inconsistencies in these systems. For example, in benzene, there are different sets of solutions for the localized π orbitals, nominally corresponding to the two different Kekule structures. Using the Boys localized orbitals yields populations that reflect D 6H symmetry, while the Pipek-Mezey scheme gives alternating charges on successive carbons going around the ring. There is a second class of molecules that are inaccessible in our method. These are anions where the natural valence minimal basis is too small to provide an antibonding orbital for each bonding orbital. One such example is C 5 H 5 -, the cyclopentadienyl anion. Forming the set of AB2 valence orbitals and taking the union with the occupied space leads to a set of orbitals that cannot be localized to atoms. Broadly, we can say that neutral species with a single Lewis structure are wellhandled by the approach described here; as well as some more complex bonding situations like SF 6 discussed above.</p><!><p>We presented a relatively cheap, non-iterative procedure to produce a set of antibonding orbitals that vary minimally with the size of the atomic orbital basis set from which they are constructed. Specifically, antibonding second order (AB2) orbitals show far less variation with basis than the Sano orbitals which are sometimes used as valence antibonding orbitals.</p><p>We showed that use of AB2 rather than Sano orbitals as initial guesses provides improved convergence for valence bond methods (specifically CCVB), as well as for CASSCF. The AB2 orbitals were successfully used as guesses for state-specific ROKS calculations of excited states, where they better resemble the converged orbitals than does the corresponding Sano orbital guess. We have shown how these AB2 orbitals can be used with the localized occupied orbitals to construct an effective minimal basis that can be used for population analysis among other things. Population analysis on the substituted fluormethane and chloromethane sequence shows the method is stable and consistent with other common methods that accord with chemical intuition. For the insertion of Be into HCl, the resulting charges show some advantages. Overall, the AB2 antibonding orbitals are relatively efficient to compute and quite useful for a variety of applications in quantum chemistry.</p>

ChemRxiv

Identification of Antifungal H+-ATPase Inhibitors with Effect on Plasma Membrane Potential

ABSTRACTThe plasma membrane H+-ATPase (Pma1) is an essential fungal protein and a proposed target for new antifungal medications. The compounds in a small-molecule library containing ∼191,000 commercially available compounds were screened for their ability to inhibit Saccharomyces cerevisiae plasma membranes containing Pma1. The overall hit rate was 0.2%, corresponding to 407 compounds. These hit compounds were further evaluated for ATPase selectivity and broad-spectrum antifungal activity. Following this work, one Pma1 inhibitor series based on compound 14 and analogs was selected for further evaluation. This compound series was able to depolarize the membrane and inhibit extracellular acidification in intact fungal cells concomitantly with a significant increase in intracellular ATP levels. Collectively, we suggest that these effects may be a common feature of Pma1 inhibitors. Additionally, the work uncovered a dual mechanism for the previously identified cationic peptide BM2, revealing fungal membrane disruption, in addition to Pma1 inhibition. The methods presented here provide a solid platform for the evaluation of Pma1-specific inhibitors in a drug development setting. The present inhibitors could serve as a starting point for the development of new antifungal agents with a novel mode of action.

identification_of_antifungal_h+-atpase_inhibitors_with_effect_on_plasma_membrane_potential

INTRODUCTION<!>Screening and evaluation of 191,000 compounds for Pma1 inhibition.<!><!>Screening and evaluation of 191,000 compounds for Pma1 inhibition.<!><!>Evaluation of Pma1 inhibitors.<!><!>Evaluation of Pma1 inhibitors.<!><!>Evaluation of Pma1 inhibitors.<!><!>Evaluation of Pma1 inhibitors.<!><!>The membrane potential decreases upon Pma1 inhibition without changes in membrane integrity.<!><!>The membrane potential decreases upon Pma1 inhibition without changes in membrane integrity.<!>iATP levels increase upon Pma1 inhibition.<!><!>DISCUSSION<!>MATERIALS AND METHODS<!>Purification of ATPases.<!>Measurement of ATP hydrolysis.<!>Library screening.<!>Fungal growth inhibition assay.<!>Hep-G2 cell viability assay.<!>Medium acidification assay.<!>Membrane potential and integrity assay.<!>Intracellular ATP.<!>

<p>Fungal infections affect 25% of the population, with superficial infections of the skin and nails representing the most common type. In a survey of women (age 16 years and older, n = 6,000), up to 49% had been diagnosed with vulvovaginal candidiasis (VVC), depending upon their ethnic origin, and approximately 20% of those women experienced recurrent VVC in a 12-month follow-up period (1), with a pronounced impact on quality of life. Invasive fungal infections are less common but of much greater concern because they are associated with extremely high mortality rates (20 to 90%) (2). The most common invasive fungal infections are caused by the yeasts Candida and Cryptococcus spp., followed by the molds Aspergillus and Mucor spp. Key areas of concern in the treatment of invasive fungal infections with the current antifungal medications include delays in diagnosis and the identification of the specific pathogenic species, intrinsic and acquired drug resistance, inconvenient drug administration, safety, and tolerability issues with prolonged use. For these reasons, there is a major unmet need for new antifungal agents (3).</p><p>The fungal plasma membrane H+-ATPase has long been recognized to be a promising antifungal target (4–6). The proton pump is essential for fungal growth, as shown by knockout studies (7). The PMA1 gene encodes the H+-ATPase, and the pump is referred to as Pma1. The fungal cell is dependent on Pma1 creating an electrochemical gradient across the plasma membrane, which is used by other transporters to energize the uptake of ions and nutrients. Pma1 pumps protons from the cytosol to the exterior of the cell, energized by ATP hydrolysis. In this regard, fungal cells are fundamentally different from human cells, where the plasma membrane potential is created by the Na+,K+-ATPase (8).</p><p>Pma1 belongs to the type III family of P-type ATPases. The related human ATPases, Na+,K+-ATPase, Ca2+-ATPase (sarcoplasmic endoplasmic reticulum Ca2+-ATPase [SERCA]), and H+,K+-ATPase, belong to the type II family. All mammalian ATPases share less than 30% amino acid sequence identity with Pma1. In contrast, the fungal H+-ATPase appears to be relatively conserved across the fungal kingdom (the amino acid sequence similarity is generally 70 to 90%). The high level of conservation seen for Pma1 warrants efforts to identify a specific Pma1 inhibitor with broad-spectrum antifungal activity.</p><p>Within the last 35 years, a number of nonspecific compounds have been evaluated as Pma1 inhibitors. To date, only a few Pma1 inhibitors, such as ebselen and the peptide BM2, have been demonstrated to inhibit the growth of living fungal cells at concentrations in the low micromolar range (4, 5, 9–12). Omeprazole is an inhibitor of the human H+,K+-ATPase and has also been evaluated as an inhibitor of Pma1 (4). Studies have shown that single mutations in the proposed binding site in Pma1 greatly alter the growth-inhibitory effects of omeprazole (4). Omeprazole requires an acid activation step to inhibit Pma1, and fungal growth inhibition is pH dependent. For full growth inhibition of Saccharomyces cerevisiae, 430 μM omeprazole is required at pH 3 (4), making it a low-potency antifungal compound.</p><p>The classical P-type ATPase inhibitor, vanadate, does not inhibit living cells due to a lack of membrane penetrability. Ebselen is known to inhibit the mammalian H+,K+-ATPase (13) and the Ca2+-ATPase ATP6 of Plasmodium falciparum (PfATP6), in addition to Pma1 (14). Due to the reactivity of ebselen with protein thiols, it is believed to target several enzymes and modify a range of biological activities (15).</p><p>Natural products, such as tellimagrandin II (16) and the immunoprotein lactoferrin (17), inhibit Pma1. Tellimagrandin II potently inhibits the growth of S. cerevisiae but not that of Candida albicans. Lactoferrin exhibits antifungal activity against both S. cerevisiae and C. albicans at concentrations in the low micromolar range.</p><p>BM2 is a cationic peptide which potently inhibits Pma1 in an ATP hydrolysis assay (50% inhibitory concentration [IC50] = 0.5 μM) (10). It displays greater than 90% fungal growth inhibition at 0.63 μM and pH 7.3 and at 5 μM and pH 7.0. At concentrations above the MIC for S. cerevisiae, BM2 also inhibits the Na+,K+-ATPase and HEp-2 cell growth and causes blood cell lysis. BM2, ebselen, and vanadate have been selected as comparators to the compounds characterized in this work.</p><p>In the present study, selected compounds were screened for their ability to inhibit ATP hydrolysis by measuring the formation of free phosphate ions, corresponding to inhibition of Pma1 under the assay conditions used. The identified inhibitors were tested for their effects on the membrane potential, the intracellular ATP concentration, extracellular medium acidification, and fungal growth inhibition.</p><!><p>A large library of small-molecule compounds was employed with the aim of finding novel Pma1 inhibitors. In total, 191,000 library compounds were screened using the ATP hydrolysis assay. An overview of the screening cascade is presented in Fig. 1. Compounds (at a final concentration of 20 μM) inhibiting Pma1 enzymatic activity by greater than 50% were regarded as hits in this study. All 407 hits were then counterscreened for their activity against Na+,K+-ATPase and SERCA in order to distinguish Pma1-specific compounds from general P-type ATPase inhibitors. All hits were screened for potential antifungal activity against S. cerevisiae and C. albicans.</p><!><p>Overview of library screening. Hits were tested for selectivity for the two mammalian ATPases SERCA and Na+,K+-ATPase as well as antifungal activity against Candida albicans and Saccharomyces cerevisiae. From this, several promising Pma1 inhibitors were identified, with compound 14 being the most promising candidate.</p><!><p>After evaluating the 407 hits, a selection of compounds was repurchased and reevaluated for Pma1 inhibition and antifungal activity. These steps were undertaken to confirm the structural identity of the library compounds and to ensure that no compound degradation had occurred during the library storage period. The structures of a selected number of these hits are presented in Fig. 2.</p><!><p>Structures of selected Pma1 inhibitors identified in the library screening.</p><!><p>Several interesting compounds with Pma1-inhibitory activity (IC50 < 25 μM) were identified through the library screening (Table 1). However, a number of these compounds, compounds 1, 2, 3, 6, 9, 11, and 12, were poor antifungal compounds, with their MICs being above 150 μM for several of the Candida spp. tested (Table 1). Compounds 4 and 8 were also poor antifungal compounds, with their MICs being >38 μM, which was the highest concentration tested due to poor compound solubility in dimethyl sulfoxide (DMSO). The precipitation of several of these compounds observed in the fungal growth medium may partially explain their lack of antifungal activity (Table 1). Unfortunately, several of these compounds that lacked antifungal activity were the compounds that were selective for Pma1 rather than SERCA and Na+,K+-ATPase. Compounds 5, 7, 10, and 13 were all potent Pma1 inhibitors (IC50 < 25 μM) with antifungal activity but were generally more potent against the mammalian ATPases than against Pma1 or equally potent against both the mammalian ATPases and Pma1. The most promising hit from the screen was compound 14, a pyrido-thieno-pyrimidine, whose structure is shown in Fig. 3, together with the structures of selected analogs of compound 14. Compound 14 was the only compound more potent against Pma1 than against the mammalian ATPases and which displayed a broad spectrum of antifungal activity against both yeasts and molds (Table 1).</p><!><p>ATP hydrolysis data and growth inhibition for identified Pma1 inhibitors from the library screeninga</p><p>Structures are shown in Fig. 2. For IC50, data are from 2 to 3 experiments. For MICs, data are from 2 to 5 experiments. NA, not available.</p><p>The ATP hydrolysis assay was performed at pH 6.5 for Pma1 and pH 7.4 for Na+,K+-ATPase and SERCA. For all other compounds, the ATP hydrolysis assay was performed at pH 7 for all ATPases.</p><p>Compound precipitation was observed in the growth medium at concentrations down to 15 μM.</p><p>Compound precipitation was observed in the growth medium at concentrations down to 48 μM.</p><p>Compound precipitation was observed in the growth medium at concentrations down to 150 μM.</p><p>Compound precipitation was observed in the growth medium at concentrations down to 75 μM.</p><p>Structure of compound 14 and analogs.</p><!><p>Additionally, 48 commercially available analogs were purchased for this study, and all of these analogs shared the common motif shown in Fig. 3. Eight of these analogs had an IC50 of less than 20 μM for Pma1. Compound 14 and four of the most potent or selective commercial analogs were chosen for further characterization (Fig. 3; Tables 2 to 5) in parallel with BM2, vanadate, and ebselen. Compound 17 was chosen as a negative control because it had a scaffold similar to that of compound 14 but weak Pma1 inhibition, probably due to the lack of an important lipophilic interaction in the R1 position (ethyl versus n-butyl in compounds 17 and 14, respectively).</p><!><p>Effects of compound 14 and analogs on ATPase activitya</p><p>Experiments were performed at pH 7, and standard deviations (n = 2) are given. Rabbit Ca2+-ATPase (SERCA1a) and porcine kidney Na+,K+-ATPase were used.</p><!><p>The four most potent analogs had a low level of Pma1 selectivity (Table 2), with compound 18 being the most selective, exhibiting approximately 3- to 4-fold greater selectivity toward Pma1 than toward SERCA or Na+,K+-ATPase. Compounds 15, 16, and 19 were more potent than compound 14 but had no selectivity for Pma1. All of the compounds exhibited antifungal activity against both Candida and Aspergillus spp. (Table 3), as well as a reduction in their activity on the viability/proliferation of mammalian Hep-G2 cells (Table 4). In our hands, BM2 was 8-fold more selective for Pma1 over Na+,K+-ATPase inhibition but had no selectivity toward SERCA. Furthermore, BM2 inhibited the growth of S. cerevisiae and C. albicans but had no effect on mammalian Hep-G2 cells.</p><!><p>Antifungal activities of compound 14 and analogs against yeast and mold species</p><p>Effects of Pma1 inhibitors on Hep-G2 cells</p><!><p>To determine if the observed antifungal activity was caused by Pma1 inhibition or another mechanism, a range of additional assays was performed. Following glucose activation of Pma1, subsequent proton extrusion leads to acidification of the buffer surrounding the cells. Inhibitors which prevent this extracellular acidification likely act through Pma1 inhibition, although these observations alone cannot discriminate between direct or indirect Pma1 inhibition (e.g., via membrane disruption). The acidification of the external medium is measured within a few minutes after compound addition. Thus, compounds which cause fast plasma membrane disruption indirectly affect proton transport and appear to be inhibitors of the acidification process. However, amphotericin B, a marketed antifungal which forms pores in the fungal membrane, did not inhibit extracellular acidification in the assay time frame of 12 min (Table 5). Compound 14 and its analogs inhibited the acidification process at concentrations similar to the IC50 for Pma1 in the ATP hydrolysis assay (Tables 2 and 5), which supports the direct inhibition of Pma1. BM2 was slightly more potent in inhibiting the acidification process than inhibiting the purified Pma1 enzyme in the ATP hydrolysis assay (Tables 2 and 5). The IC50 of ebselen for C. albicans was found to be 11 μM, in agreement with the previously reported value of 14 μM (18).</p><!><p>Effects of Pma1 inhibitors on medium acidification in S. cerevisiae and C. albicansa</p><p>The IC50 was determined to be the concentration which resulted in 50% inhibition of medium acidification normalized to the response from glucose-activated versus non-glucose-activated cells. AMB, amphotericin B. IC50 with standard deviations (n = 2 to 3) are indicated.</p><!><p>A striking feature of the fungal cell membrane is the very high membrane potential generated by Pma1. It has been measured to be −175 mV within the fungal species Neurospora crassa (19). This value is significantly higher than that observed in mammalian cells, where the membrane potential is typically measured to be −65 to −85 mV. The N. crassa measurements were performed on the large hypha of this mold species with patch clamp electrodes, a setup which is technically difficult to reproduce on smaller cells, such as those of S. cerevisiae. To further demonstrate Pma1 inhibition in living fungal cells, image cytometry was employed to investigate membrane potential changes and membrane integrity upon compound exposure. Membrane depolarization was assessed with the fluorescent probe bis-(1,3-dibutylbarbituric acid) trimethine oxonol [DiBAC4(3)], which is only able to enter cells with a decreased membrane potential. Simultaneously, propidium iodide (PI) was used to distinguish between membrane depolarization and general membrane permeabilization, since PI only enters cells with permeable cell membranes.</p><p>S. cerevisiae cells exposed to compound 14 yielded an elevated DiBAC4(3) signal after 5 min of exposure, while only a minimal increase of the PI signal was observed (Fig. 4A to C and Fig. S1). After 15 or 30 min of exposure, the sizes of both DiBAC4(3)-positive cell populations and DiBAC4(3)- and PI-positive cell populations had increased. These observations suggest that cells were first depolarized prior to the membrane integrity becoming compromised, thus allowing PI to enter the cells. After 30 min of exposure to compound 14, only 16% of the cells remained both PI and DiBAC4(3) negative, 44% were DiBAC4(3) positive and PI negative, and 39% were both PI and DiBAC4(3) positive. Only 0.7% of the cells were PI positive but DiBAC4(3) negative. In the DMSO control sample, 92% of the cells were both PI and DiBAC4(3) negative.</p><!><p>Effects of inhibitors on membrane potential and membrane integrity in S. cerevisiae. (A) Bar chart of the cumulative percentage of cells that were DiBAC4(3) positive or negative and PI positive or negative, as defined by the quadrants in panel B. Cells were treated with compound 14, 17, 19, or BM2 or CCCP at 5, 15, or 150 μM for 5, 15, or 30 min (as indicated). Error bars indicating SEMs are shown in only one direction for clarity. (B) Scatter plot of cells treated with 1% DMSO for 30 min (control). (C) Scatter plot of cells treated with 15 μM compound 14 for 30 min. Representative scatter plots for all conditions are shown in Fig. S1 in the supplemental material. RFU, relative fluorescent units.</p><!><p>The most potent antifungal analog of compound 14 was compound 19, and this compound was also evaluated with imaging cytometry at concentrations of 5 and 15 μM. Similarly to compound 14, exposure of cells to compound 19 also resulted in two large populations of DiBAC4(3)-positive and PI-negative cells and DiBAC4(3)- and PI-positive cells after 30 min of exposure.</p><p>BM2 exposure led to a marked increase in the PI-positive population within 5 min, indicating rapid effects on membrane integrity (Fig. 4A and Fig. S1). However, we also observed an increase in the DiBAC4(3) signal after 15 or 30 min in BM2-treated cells. Given the effects on membrane integrity, it is difficult to attribute the increased DiBAC4(3) signal to direct Pma1 inhibition. The protonophore carbonyl cyanide m-chlorophenylhydrazone (CCCP) had an effect on the PI and DiBAC4(3) signals similar to that of compounds 14 and 19. Compound 17, which had an IC50 of 107 μM for purified S. cerevisiae Pma1, did not increase the DiBAC4(3) signal in S. cerevisiae cells, and the PI signal was not significantly increased.</p><!><p>Pma1 requires ATP to energize proton transport out of the fungal cell. A decrease in the intracellular ATP (iATP) levels indirectly affects the activity of Pma1. To rule out the possibility that the Pma1 inhibition observed was indirectly caused by decreased ATP levels, we determined iATP levels after compound exposure (Fig. 5A to C). Treatment of S. cerevisiae and C. albicans cells with compound 14 led to a significant increase in iATP levels (200% and 1,500%, respectively) compared to those for the untreated control. When S. cerevisiae cells were exposed to compound 14 in the presence of oligomycin, an ATP synthase inhibitor, iATP levels remained comparable to those for control cells. Compound 19 behaved in a manner similar to that for compound 14, leading to a significant increase in iATP levels in S. cerevisiae and C. albicans. Treatment with BM2 also led to a large increase in ATP levels (∼1,100%) compared to that for the control cells. When S. cerevisiae cells were treated with BM2 and oligomycin together, the increase in iATP levels was limited to 300%. However, the addition of oligomycin to BM2-treated C. albicans cells did not alter the iATP level compared to that in cells treated with BM2 alone. Treatment with oligomycin alone, CCCP, vanadate, and the low-potency Pma1 inhibitor compound 17 had minimal effects on iATP levels compared to those found after treatment with compounds 14, 19, and BM2. Ebselen could not be tested, as it inhibited the luciferase enzyme used in the iATP assay. The increase in iATP levels in C. albicans for compounds 14, 19, and BM2 was shown to be dose responsive, with all compounds producing a maximum iATP level of about 500 nM (per well) at the highest concentrations of inhibitors tested (Fig. 5C). The half-maximal (50%) effective concentrations (EC50) were calculated to 13, 7, and 2 μM for compounds 14, 19, and BM2, respectively. Interestingly, these values were fairly close to the IC50 for the Pma1 enzyme from C. albicans of 17, 9, and 0.9 μM, respectively.</p><!><p>Pma1 inhibitors increase iATP levels in S. cerevisiae (A) and C. albicans (B and C). The total cellular ATP concentrations were determined after 30 min of treatment. The mean data (n = 3) with the SD are indicated as the percent change compared to that for the untreated control sample (1% DMSO) (A and B) or as the concentration (in nanomolar) of iATP (C). olig., oligomycin A (at 40 μM in all panels). *, P < 0.05 compared to the untreated control by unpaired t test; **, P < 0.01 compared to the untreated control by unpaired t test.</p><!><p>In the initial screening process, we identified a number of selective and relatively potent Pma1 inhibitors; however, many lacked antifungal activity. One interesting compound series which exhibited both antifungal activity and an affinity for Pma1 over the mammalian ATPases was identified and further characterized. An initial understanding of the structure-activity relationship for this series was established on the basis of such relationships for commercially available compounds.</p><p>The compound 14 series, containing a pyrido-thieno-pyrimidine group, was considered especially interesting since small changes to the scaffold yielded significant changes in the inhibitory properties toward the three ATPases tested. Our data suggest a direct interaction with the ATPases rather than an interaction via compound aggregates, which is a common false-positive mechanism (20, 21). 4-Aminoalkylamino derivatives of pyrido[3′,2′:4,5] thieno[3,2-d]pyrimidine have been shown to interact with a number of molecular targets, including phosphodiesterase IV, and to have potential use in the treatment of asthma and chronic obstructive pulmonary disease (22; S. L. M. Pages and M. J. Taltavull, U.S. patent application, June 2006) and in the control of tumor necrosis factor alpha (TNF-α) release (23) and to have beta2 adrenoreceptor agonist activity (24). Additionally, diverse biological activities, such as anticonvulsant (25–27) and neurotropic (28, 29) activities, have been ascribed to compounds similar to those presented here. To our knowledge, this is the first description of pyrido-thieno-pyrimidines as inhibitors of P-type ATPases. The Pma1-inhibitory activity was confirmed both through the inhibition of acidification of the extracellular medium and through the demonstration of membrane depolarization but not an immediate loss of integrity by the DiBAC4(3) assay.</p><p>DiBAC4(3) and PI are frequently used as stains for dead or dying cells. However, DiBAC4(3) enters the cells due to a loss of membrane potential, whereas PI enters the cells due to a loss of membrane integrity. By using a short time frame (5 to 30 min) and observing individual cells, DiBAC4(3) can be used to estimate changes in membrane potential, while costaining with PI reveals when the membrane loses its integrity. Previously, DiBAC4(3) (30) and other fluorescent probes used to measure membrane potential, such as 3,3-dipropylthiacarbocyanide iodide) [diS-C3(3)] (31), have been used to estimate changes in membrane potential. To our knowledge, DMM-11, a competitive inhibitor of Mg2+-ATP, is the only other Pma1 inhibitor which has previously been shown to depolarize the fungal plasma membrane (12).</p><p>The protonophore CCCP was used as a positive control in the membrane potential study; however, it required high concentrations to overcome the proton excretion induced by Pma1 and depolarize the membrane. CCCP has previously been shown to affect the mitochondrial membrane and, thereby, the supply of ATP (32). In this study, treatment with CCCP resulted in limited changes to the iATP levels in S. cerevisiae and C. albicans (Fig. 5). Treatment with the ATP synthase inhibitor oligomycin did not result in a large significant change in iATP levels, in agreement with earlier observations by Andrés et al. (17). These data suggest that fungal cells can compensate for ATP synthase inhibition for short periods of time under these conditions.</p><p>Compounds 14, 19, and BM2 gave rise to a large significant increase in iATP levels, while compound 17 (a low-potency Pma1 inhibitor) gave rise to a substantially lower increase. Increases in iATP levels have previously been reported with exposure to omeprazole (36% increase) in S. cerevisiae (33) and lactoferrin (>600% increase) in C. albicans (17). Both omeprazole and lactoferrin are described to be Pma1 inhibitors, although omeprazole is not fully capable of inhibiting extracellular acidification (33). We suggest that an increased iATP level is attributable to Pma1 inhibition, considering that Pma1 is the major consumer of ATP in the cell and is thought to consume 20 to 50% of the cellular ATP (34). Therefore, unused ATP accumulates in the cell when Pma1 is inhibited. Increased iATP levels have also been observed in Pma1 mutants with partial defects in pumping activity (35), supporting the hypothesis that increased iATP levels are a consequence of Pma1 inhibition.</p><p>Ebselen was found to inhibit not only Pma1 but also the mammalian ATPases SERCA and Na+,K+-ATPase. Given that ebselen has previously been reported to inhibit PfATP6 and the H+,K+-ATPase (13, 14), this compound is likely an unspecific P-type ATPase inhibitor.</p><p>In our hands, BM2 exhibited less selectivity for S. cerevisiae (ATCC 9763) Pma1 over porcine Na+,K+-ATPase than the previously reported 50-fold selectivity of S. cerevisiae (T48) Pma1 over canine Na+,K+-ATPase, determined when the IC50 of S. cerevisiae (T48) Pma1 and canine Na+,K+-ATPase were compared (10). Furthermore, we observed no selectivity of BM2 for Pma1 over SERCA. BM2 affected the membrane potential, but given the rapid and severe effects on overall membrane integrity, it is unclear if the depolarization was directly caused by Pma1 inhibition. Increased cell permeation of rhodamine 6G (Rh6G) was also previously observed in an Rh6G efflux assay with >10 μM BM2, which further supports our observations that BM2 compromises membrane integrity (10). Despite the decreased membrane integrity, our results suggest that Pma1 was inhibited, as indicated by the observed large increase in iATP levels.</p><p>Oligomycin did not completely prevent the rise in iATP levels of S. cerevisiae cells treated with BM2. Furthermore, iATP levels in C. albicans cells were significantly changed in response to BM2 exposure, regardless of the presence of oligomycin. We therefore speculate that the inhibition of ATP synthase by oligomycin is slower than the accumulation of iATP resulting from the actions of BM2. We speculate that the increase in iATP levels due to Pma1 inhibition, together with BM2's additional membrane-disruptive effects, facilitates the killing of fungal cells.</p><p>Together, our data support the suggestion that BM2 operates via a dual mechanism where both the fungal membrane and the fungal proton ATPase are affected, leading to fungal growth inhibition and cell death. Greater selectivity toward SERCA and the Na+,K+-ATPase is likely required. Special attention should be paid to possible membrane interactions, as BM2 affects fungal membrane integrity and has previously been linked to low-level hemolysis and detectable toxicity in HEp-2 cells (10). On the basis of the membrane potential and integrity results, the compound 14 series has a mode of action different from that of BM2. The antifungal effect of compound 14 appears to be more a direct consequence of Pma1 inhibition as compared to BM2, as the latter affects both Pma1 activity and the membrane integrity. As expected for an inhibitor of the essential and conserved Pma1 protein, the compound 14 series has a broad spectrum of antifungal activity, inhibiting the growth of all Candida and Aspergillus spp. tested. This is in contrast to BM2, which has poor activity against Candida glabrata and the Aspergillus spp.</p><p>Some of the compounds identified during the library screen inhibited Pma1 on an enzymatic level; however, this did not always translate into antifungal activity. There are several plausible explanations for this lack of translation. Fungal cells contain a complex cell wall, which may serve as an impermeable barrier to compound entry. The compounds might not be able to cross the cell membrane due to aggregate formation, leading to false-positive hits in the enzymatic assay. False-positive hits may also result from unspecific protein inhibition activity in a homogeneous assay setup (20, 21). Finally, the chemical properties of the compounds must be taken into consideration, as the possibility that some of the compounds tested precipitated when approaching concentrations required for their effect cannot be ruled out.</p><p>In this study, novel compounds with Pma1-inhibitory properties were identified and direct inhibition of Pma1 in fungal cells was confirmed by a decrease in the plasma membrane potential, while membrane integrity was preserved. Encouragingly, compounds 14 and 18 displayed selectivity against the mammalian ATPases, with both showing 3-fold selectivity toward SERCA and compound 18 showing 4-fold selectivity toward the Na+,K+-ATPase. Compound 19 was interesting, as it displayed a 3-fold toxicity window between C. albicans and Hep-G2 cells, despite it being an unselective ATPase inhibitor. However, the fungal growth and Hep-G2 cell assay conditions differed significantly by the presence of 10% fetal bovine serum (FBS) in the Hep-G2 cell assay mixture. In C. albicans growth assays, a 10-fold increase in the MIC value of compound 19 occurred when 10% FBS was added to the assay mixtures (see Table S2 in the supplemental material). This suggests significant protein binding of compound 19. The reduced toxicity window observed when fungal and Hep-G2 cell assays were run under comparable conditions also highlights that selectivity for Pma1 over mammalian ATPases should have high priority in any future Pma1 drug development.</p><p>The collection of methods presented herein is especially useful in antifungal drug development targeting Pma1. They can reveal whether inhibition of Pma1 enzymatic activity can be translated into Pma1 inhibition in living cells, while simultaneously ruling out several indirect mechanisms. Future directions for study should include the synthesis of more analogs with the aim of optimizing the potency and specificity of the compounds, thereby improving the toxicity window toward Hep-G2 cells. In conclusion, there is a requirement for improvement of both the physical and chemical properties as well as the selectivity for Pma1 over mammalian ATPases before a new Pma1-targeting drug candidate can be developed.</p><!><p>All chemicals were purchased from Sigma-Aldrich (St. Louis, MO) unless stated otherwise. Saccharomyces cerevisiae RS72 was used for purification of the plasma membranes which were used in the library screening. S. cerevisiae (ATCC 9763) was also used for purification of plasma membranes and for all assays except the initial library screening.</p><!><p>Plasma membranes containing Pma1 were isolated from S. cerevisiae RS72 cells containing the full-length cDNA of the S. cerevisiae plasma membrane H+-ATPase isoform PMA1 under the control of the PMA1 promoter, as described in reference 16.</p><p>A 100-ml overnight culture in YPD medium [10 g/liter yeast extract (BD, Sparks, MD), 20 g/liter Bacto peptone (BD), 20 g/liter d-(+)-glucose] was transferred to 1 liter YPD medium and grown at room temperature and 150 rpm for 18 h for wild-type S. cerevisiae (ATCC 9763) and 7 h for C. albicans. The cells were harvested and washed in Milli-Q H2O by centrifugation at 3,360 × g. The cells were glucose activated by the addition of 10% d-(+)-glucose for 10 min while shaking. The cells were then centrifuged at 3,360 × g for 3 min and frozen. Cells (60 to 80 g, wet weight) were then disrupted by bead beating (BioSpec, Bartlesville, OK) and processed as described in reference 16 to prepare the plasma membranes. For C. albicans only, microsomes were prepared, as the sucrose gradient step did not provide additional purity. All batches were validated by ATPase hydrolysis activity, pH optimum, and orthovanadate sensitivity.</p><p>Rabbit sarcoplasmic reticulum membranes containing sarcoplasmic endoplasmic reticulum Ca2+-ATPase (SERCA) were kindly provided by Claus E. Olesen and Jesper V. Møller, Aarhus University, and prepared as described in reference 36. Porcine kidney Na+,K+-ATPase was kindly provided by Natalya Fedosova, Aarhus University, and prepared as described in reference 37.</p><!><p>The protein activity levels were determined by measuring the amount of free phosphate produced from the ATP hydrolysis reaction. The assay was performed as described in reference 16. Each well contained ATPase enzyme, 2 μl of inhibitor (usually in the range of 333 to 0.01 μM) dissolved in dimethyl sulfoxide (DMSO) (Duchefa Biochemie, Haarlem, The Netherlands), and an assay buffer dependent on the enzyme for a final volume of 60 μl. Possible compound precipitation in the assay buffer was monitored using a Nephelostar plate reader measuring turbidity, and the information was taken into account in the IC50 determination. Pma1 assay buffer consisted of 20 mM 3-(N-morpholino)propanesulfonic acid (MOPS)–NaOH, pH 6.5 or 7.0, 8 mM MgSO4, 50 mM KNO3 (a vacuolar ATPase inhibitor), 25 mM sodium azide (a mitochondrial ATPase inhibitor), and 250 μM sodium molybdate (an acid phosphatase inhibitor). SERCA buffer consisted of 9 mM MOPS-NaOH, pH 7, 2.7 mM MgCl2, 90 μM CaCl2, and 72 mM KCl. Na+,K+-ATPase buffer consisted of 30 mM MOPS-NaOH, pH 7, 40 mM NaCl, 4 mM MgCl2, and 20 mM KCl. The concentration of protein was adjusted to obtain a signal of the optical density at 860 nm (OD860) of between 0.5 and 1.0 in the untreated samples (for all Pma1 batches, ∼1 to 2 μg/well; for SERCA and Na+,K+-ATPase, ∼0.1 to 0.2 μg/well). BM2 (RRRFWWFRRR-NH2, d-form amino acids) was custom synthesized by Peptide 2.0 (Chantilly, VA).</p><!><p>Libraries of compounds in a 96-well plate format were screened for Pma1 inhibition with the purified S. cerevisiae RS72 plasma membranes using the ATP hydrolysis assay at pH 6.5. All compounds were diluted in DMSO. The libraries were commercially available from the following suppliers: Key Organics (13,681 compounds; Cornwall, UK), ChemBridge (63,042 compounds; San Diego, CA), ChemDiv Inc. (34,000 compounds; San Diego, CA), InterBioScreen (24,072 compounds; Chernogolovka, Russia), Specs (36,854 compounds; Zoetermeer, The Netherlands), Asinex (9,021 compounds; Rijswijk, The Netherlands), and ComGenex (10,000 compounds; Budapest, Hungary). The CAS numbers of selected compounds are as follows: compound 1, 375352-86-6; compound 2, 845990-59-2; compound 3, 496771-56-3; compound 4, 372174-76-0; compound 5, 488107-35-3; compound 6, 847044-59-1.</p><!><p>The following fungal isolates were used: Saccharomyces cerevisiae ATCC 9763, Candida albicans SC5314, Candida parapsilosis ATCC 22019, Candida tropicalis Ct016, Candida glabrata ATCC 90030, Aspergillus fumigatus ATCC 13073, and Aspergillus flavus ATCC MYA-1005. Frozen stocks of yeast isolates in glycerol (final concentration, 20%) were prepared by growing the cells to log phase in YPD medium at 30°C and 150 rpm. Frozen stocks of mold spores were prepared by harvesting spores from 7-day-old potato glucose agar plates in phosphate-buffered saline, 0.1% Tween 80 and aliquoting these spores in the presence of glycerol (final concentration, 20%). The cells were stored at −80°C. The cells or spores were thawed and diluted to a final concentration of 0.5 × 105 to 2.5 × 105 CFU/ml in sterile Milli-Q H2O. In a 96-well plate, the fungal suspension was added to an equal volume of 2× RPMI medium (20.8 g/liter RPMI 1640 medium [catalog number R6504; Sigma-Aldrich], 0.33 M MOPS, 36 g/liter glucose adjusted to pH 7.0 with KOH) to which an inhibitor compound had been added, thus giving a final DMSO concentration of 1.5% in 1× RPMI medium. The plate was incubated for 20 to 24 h at 34°C (72 h for molds), and the OD490 was measured on a Victor X5 (PerkinElmer) plate reader. Mold growth assay plates were also visually inspected. The MIC was defined as the lowest concentration inhibiting the visual growth of the microorganism. Standard errors were typically below 5%.</p><!><p>In a tissue culture-treated 96-well plate (catalog number 655180; Grenier), 10,000 human hepatocyte Hep-G2 cells (catalog number 85011430; Sigma-Aldrich) were plated in 200 μl growth medium (Eagle minimal essential medium; catalog number M2279; Sigma-Aldrich), 2 mM l-glutamine (catalog number 03-020-1B; Biological Industries), 1% nonessential amino acids (catalog number XC-E1154/100; Biosera), 10% fetal bovine serum (catalog number BI-04-007-1A; Biological Industries) and incubated overnight at 37°C in 5% CO2. On the next day, fresh growth medium plus 2 μl compound in DMSO was added. The plate was incubated for a further 24 h at 37°C in 5% CO2. The medium was replaced with 100 μl freshly prepared 0.5 mg/ml 2,3-bis(2-methoxy-4-nitro-5-sulfophenyl)-5-[(phenylamino)carbonyl]-2H-tetrazolium hydroxide, sodium salt, solution (catalog number X4251; Sigma-Aldrich) in RPMI 1640 medium (catalog number R7509; Sigma-Aldrich) with 3.83 μg/ml phenazine methosulfate (catalog number P9625; Sigma-Aldrich) and incubated for 2 to 3 h at 37°C in 5% CO2. The color reaction was measured on a Victor X5 plate reader (PerkinElmer) by determination of the OD450 and the half-maximal (50%) effective concentration (EC50) was calculated.</p><!><p>Frozen stocks of C. albicans and S. cerevisiae were transferred to YPD agar plates and incubated overnight at 34°C. On the next day, the cells were suspended and washed twice in 50 mM KCl, pH 6.7, after pelleting of the cells by centrifugation at 3,000 × g for 5 min. The cells were then starved overnight at 4°C in the KCl solution. The cell suspension (final OD600 of 0.19), an inhibitor in DMSO (1.5%), and 1.3 μg/ml dextran-fluorescein isothiocyanate (molecular weight, 40,000; catalog number FD40S; Sigma-Aldrich) were then mixed before initiation of the assay by the addition of 2% d-(+)-glucose to a final assay volume of 200 μl. The pH drop was then monitored by determination of the fluorescent signal with excitation at 485 nm and emission at 538 nm on a Fluoroskan Ascent plate reader (Thermo Fisher Scientific) for 12 min. The rate of medium acidification was calculated on the basis of the slope of the drop in fluorescence.</p><!><p>Cells were transferred from frozen stocks to 3 ml of YPD medium and grown overnight at 30°C and 150 rpm. The culture was diluted to an OD600 of 0.15 and grown for 3 to 4 h at 30°C and 150 rpm to an OD600 of 0.5 to 0.7. The cells were pelleted by centrifugation at 2,000 × g for 2 min and washed in buffer A (100 mM MOPS and 1 mM KCl adjusted to pH 7.0 with Trizma base). The washing procedure was repeated, and the cells were resuspended in buffer A to an OD600 of 0.2. Fifty microliters of buffer A containing 3.3 μg/ml propidium iodide (PI; catalog number P1304MP; Thermo Fisher Scientific) and 1 μg/ml DiBAC4(3) (catalog number D8189; Sigma-Aldrich) was mixed with 50 μl the S. cerevisiae cell suspension in buffer A and transferred to a 96-well plate with 1 μl of inhibitor in DMSO, and the components were mixed.</p><p>Approximately 30 μl sample was transferred to an A2 glass slide (ChemoMetec, Allerod, Denmark), and the slide was incubated at 30°C before the cells were counted on a NucleoCounter NC-3000 cytometer (ChemoMetec). Five thousand cells were counted per experiment with an exposure time of 1,000 ms. A dark field was used as the masking channel to select the yeast cells. The DiBAC4(3) channel (excitation, 530 nm; emission, 675/75 nm) and the PI channel (excitation, 630 nm; emission, 740/60 nm) were used to measure the membrane potential and membrane integrity, respectively. To account for the fluorescent spillover from the PI channel to the DiBAC4(3) channel, 15% compensation was applied. In all experiments, 1.9% compensation was used to compensate for the DiBAC4(3)-to-PI spillover. The results were analyzed using NucleoView software (ChemoMetec). Only cells with a pixel size of 10 to 40 were included in the analysis, thus avoiding analysis of noncell artifacts. Carbonyl cyanide m-chlorophenylhydrazone (CCCP) was used as a positive control for decreased membrane potential, and lysis buffer was used to validate lost membrane integrity.</p><!><p>S. cerevisiae and C. albicans cells were grown and washed as described above for the membrane potential and integrity assay and resuspended in SG medium (0.7% yeast nitrogen base without amino acids [BD], 50 mM succinic acid adjusted to pH 7 with Trizma base) to OD600 of 0.2 and 0.1, respectively. One hundred microliters of an S. cerevisiae or C. albicans cell suspension and 1 μl compound in DMSO were incubated for 30 min. Twenty-five microliters of the suspension was then transferred to a black 96-well plate containing 25 μl BacTiter-Glo reagent (Promega, Madison, WI) and incubated for 15 min in the dark. The luminescence was read on SpectraMax X5 (Molecular Devices, Sunnyvale, CA) plate reader with a 10 s of shaking and a 0.5 s integration time. Standard curves at 10, 100, and 1,000 nM ATP were performed with every experiment.</p><!><p>Supplemental material for this article may be found at https://doi.org/10.1128/AAC.00032-17.</p>

PubMed Open Access

\xce\xb2-Keto and \xce\xb2-hydroxyphosphonate analogs of biotin-5\xe2\x80\x99-AMP are inhibitors of holocarboxylase synthetase

Holocarboxylase synthetase (HLCS) catalyzes the covalent attachment of biotin to cytoplasmic and mitochondrial carboxylases, nuclear histones, and over a hundred human proteins. Nonhydrolyzable ketophosphonate (\xce\xb2-ketoP) and hydroxyphosphonate (\xce\xb2-hydroxyP) analogs of biotin-5\xe2\x80\xb2-AMP inhibit holocarboxylase synthetase (HLCS) with IC50 values of 39.7 \xce\xbcM and 203.7 \xce\xbcM. By comparison, an IC50 value of 7 \xce\xbcM was observed with the previously reported biotinol-5\'-AMP. The Ki values, 3.4 \xce\xbcM and 17.3 \xce\xbcM, respectively, are consistent with the IC50 results, and close to the Ki obtained for biotinol-5\'-AMP (7 \xce\xbcM). The \xce\xb2-ketoP and \xce\xb2-hydroxyP molecules are competitive inhibitors of HLCS while biotinol-5\'-AMP inhibited HLCS by a mixed mechanism.

\xce\xb2-keto_and_\xce\xb2-hydroxyphosphonate_analogs_of_biotin-5\xe2\x80\x99-amp_are_inhibitors_of_

<p>Holocarboxylase synthetase (HLCS) is the sole enzyme in the human proteome capable of catalyzing the covalent attachment of biotin to lysine residues.1 HLCS localizes in the cytoplasm, mitochondria, and cell nuclei.1,2 HLCS catalyzes biotinylation of five carboxylases in mitochondria and cytoplasm, which play key roles in gluconeogenesis, fatty acid metabolism, and leucine metabolism.3 In addition, a recent mass spectrometry screen identified 108 novel biotinylated proteins; heat shock proteins and enzymes from glycolysis are overrepresented among these proteins.4 HLCS orchestrates the assembly of a multiprotein gene repression complex in human chromatin, partially mediated through HLCS-dependent methylation of the histone methyltransferase EHMT1 and the nuclear receptor co-repressor N-CoR.5 Consistent with the importance of HLCS in intermediary metabolism and cell function, no living HLCS null person has ever been reported, and persons with HLCS mutations require lifelong treatment with pharmacological doses of biotin.6</p><p>HLCS-dependent biotinylation involves basically two steps (Fig. 1). In the first, activation of biotin by ATP generates the mixed anhydride biotin-5′-adenosine monophosphate (Bio-5′-AMP) (1). In the second step, the phosphate anhydride serves as an acylating agent for a target lysine in carboxylases (2), histones (3), and other proteins,3,4,7 covalently linking biotin to the substrate via an amide bond.</p><p>The objective of this study was to develop a new class of synthetic HLCS inhibitors which could potentially be targeted to distinct cellular structures. Such an inhibitor would be a useful analytical tool in studies of HLCS-dependent biotinylation events in the cytoplasm, mitochondria, and nucleus. We based these studies on analogs of biotin-5′-AMP, namely biotin β-ketophosphonate-5′-AMP (β-ketoP) and biotin β-hydroxyphosphonate-5′-AMP (β-hydroxyP), that substitute a hydrolytically stable phosphonate for the acyl phosphate found in biotin-5′-AMP (Figure 2a). The use of phosphonates as unreactive isosteres of phosphates is well established.8 For example, nonhydrolyzable aminoacyl analogs of aspartyl adenylate exhibit potent inhibitory activity against E. coli aspartyl-tRNA synthetase.9 There is precedence for the efficacy of structurally analogous compounds sulfamides, sulfonamides, and 1,2,3-triazoles in the inhibition of a microbial biotin protein ligase, the HLCS ortholog BirA (Fig. 2b).10-13 We also investigated biotinol-5′-AMP, a known phosphate analog of biotin-5′-AMP which replaces the carbonyl oxygen with a methylene (CH2).10,14</p><p>Inhibitor synthesis: The central element of the synthesis is the formation of a protected version of a biotin ketophosphonate (4a) via condensation of a biotin-derived ketophosphonic acid (3) with a protected adenosine (Scheme 1). The synthesis begins with biotin methyl ester (1), prepared via the acid-catalyzed esterification of biotin.15 Reaction with the carbanion derived from methyl phosphonate was anticipated to offer a convenient route to a precursor of the desired phosphonates. However, reaction of ester 1 with the lithiated methylphosphonate, generated using lithium bis(trimethylsilyl)amide (LiHMDS) or n-butyl lithium (n-BuLi), resulted in poor yields. Fortunately, reaction of the ester with a large excess of the lithiated phosphonate, followed by quenching with deionized water to minimize demethylation of the phosphonate diester product, produced a 69% yield of dimethyl β-ketophosphonate 2.16 Selective monodemethylation with lithium bromide provides a good yield of mono ester 3. The pyridinium salt of 3 underwent coupling with 2′, 3′-isopropylidineadenosine (i-PrA) in the presence of O-(benzotriazol-1-yl)-N,N,N′,N′-tetramethyluronium hexafluorophosphate (HBTU) to provide a mixed phosphonate diester (4a) in 79% yield for two steps.17,18 No coupling was observed if N,N'-dicyclohexylcarbodiimide (DCC) was substituted for HBTU. The acidic methylene of the ketone phosphonate was observed (NMR) to readily undergo H/D exchange upon dissolution in d4-MeOH.</p><p>Reduction of the ketone with NaBH4 resulted in formation of the corresponding alcohol (4b) as a mixture of diastereomers at the newly formed stereocenter. Stirring the ketone diester (4a) or the alcohol diester (4b) in pyridine/water resulted in selective demethylation to afford monoesters 5a or 5b, respectively. Removal of the acetonide protecting group from the sugar followed by neutralization with ammonium bicarbonate provided the target biotin β-ketophosphonate (β-ketoP) 6a in 58% yield for two steps. The same procedure, when applied to alcohol 5b, furnished biotin-β-hydroxyphosphonate (β-hydroxyP) 6b in 99% yield.</p><p>Inhibition of HLCS: Biotin β-ketophosphonate-5′-AMP (β - ketoP, 6a) inhibited HLCS in a dose-dependent manner. The polypeptide p67 is a substrate for biotinylation by HLCS.19 When recombinant p67 was incubated with recombinant HLCS and 20 μM biotin in the presence of 50 to 500 μM β-ketoBP, HLCS inhibition was maximal at the highest inhibitor concentration tested. For example, the inhibition at 500 μM β-ketoBP was 81.3±11.1% (P<0.01; n=4) compared with vehicle control (Fig. 3A). Data are presented as mean±SD of 4 replicates. Incubation of p67 in the absence of HLCS produced no detectable signal (lane 6 in Fig. 3B). Under the conditions used here, IC50 and Ki for β-ketoBP equaled 39.7±1.9 μM and 2.0 ±0.1 μM, respectively; see Supplementary Materials for details regarding assays and calculations. When tested under identical conditions, β-hydroxyP inhibited HLCS activity by 67.7 ±10.0% (P=0.0001; n=4) at concentrations of 500 μM inhibitor. The IC50 and Ki values calculated under these conditions were 203.7 ±3.7μM and 10.4±0.2μM, respectively. β-KetoP is a competitive inhibitor of HLCS, based on competition studies with biotin. In these studies, the concentration of the inhibitor was held constant at 250 μM while that of biotin was varied from 0-320 μM; a second curve was generated in the absence of inhibitor (Fig. 3C). The apparent Vmax was similar for incubations with and without inhibitor [31.5±1.6 vs. 22.3±1.51 pmol biotinylated p67/(nmol HLCS x s); n=4] whereas the apparent Km for biotin was increased by the addition of inhibitor (77.9±10.3 vs. 1.6±1.8 μM biotin; N=4). β-HydroxyP (6b) also acts as a competitive inhibitor as evidenced by similar apparent Vmax values with and without inhibitor [20.6±1.8 vs. 22.6±1.4 pmol biotinylated p67/nmol HLCS x s); n=4] whereas reactions incubated with inhibitor increased the apparent Km for biotin (82.5±20.4 vs. 1.9±1.8 μM biotin; n=4). As a negative control we conducted incubations with a biotin ketophosphonic acid (compound 3 in Scheme 1). This substrate incorporates an electrophilic carbonyl carbon beta to a charged phosphonate but lacks the adenosyl fragment hypothesized as essential for mediating HLCS inhibition. Consistent with this theory, the ketophosphonic acid compound did not inhibit HLCS (data not shown).</p><p>The results were compared against assays conducted with biotinol-AMP, a known phosphate analog of biotin-5′-AMP which has previously been employed for inhibition of BirA (biotin protein ligase).10,11 Biotinol-AMP reduces HLCS activity by 98.01±0.1% at concentrations of 500 μM and has an IC50 value and Ki of 8.8±3.6μM and 754±303nM, respectively. When reactions incubated with biotinol-AMP were challenged with increasing amounts of up to 320 μM biotin, the apparent Vmax decreased compared to reactions without inhibitor (22.6±1.4 vs. 5.9±1.4; n=4) and Km increased (1.9±1.8 vs. 146±79; n=4) indicating the biotinol-AMP most likely acts as a mixed inhibitor. There remains limited knowledge regarding the structure of HLCS and the mechanism of catalysis.11,13b,20 Similarly, little is known about the basis for selectivity between the classic carboxylase targets of HLCS and novel targets in chromatin and other proteins. The objective of this study was to develop a synthetic HLCS inhibitor capable of penetrating cell membranes and which could potentially be targeted to distinct cellular structures. Such an inhibitor would be a useful analytical tool in studies of carboxylase biotinylation in cytoplasm and mitochondria, studies of chromatin protein biotinylation and HLCS-dependent formation of multiprotein gene repression complexes in nuclei, and studies of newly discovered species of biotinylated proteins throughout the cell.</p><p>Consistent with the importance of HLCS in intermediary metabolism and epigenetics, no living HLCS null individual has ever been reported, suggesting embryonic lethality. HLCS knockdown in Drosophila melanogaster (~30% residual activity) produces phenotypes such as decreased life span and reduced heat resistance.21 Mutations and single nucleotide polymorphisms have been identified and characterized in the human HLCS gene; these mutations cause a substantial decrease in HLCS activity, aberrant gene regulation and metabolic abnormalities.6,22 Unless diagnosed and treated at an early stage, homozygous severe HLCS deficiency is characteristically fatal.23 Three independent cancer and patent databases correlate HLCS loss or mutation with an increase in detected tumors.24</p><p>Several classes of biotin-5′-AMP analogs have been applied to study the function of biotin protein ligases (BPLs), exemplified by HLCS as well as BirA, an enzyme catalyzing biotinylation of acyl carrier protein in prokaryotes.10,13,14 BirA from E. coli has 21% sequence similarity to HLCS.25 Biotinol-5′-AMP, a phosphate ester lacking the acyl carbonyl of biotin-5′-AMP, binds tightly to the Escherichia coli biotin repressor (KD = 1.5 ± 0.2 nM)11,20 and inhibits biotin transfer to the acceptor protein.10,14 Biotinol-5′-AMP also binds tightly to Staphylococcus aureus BPL (Ki = 0.03 ± 0.01 μM). The activity of this analog would appear to suggest the key role of the phosphate moiety and the relative lack of importance of hydrogen-bonding interactions to the acyl carbonyl of biotin-5′-AMP. However, other work has demonstrated that a sulfamoyl-containing bisubstrate analog, replacing the acyl phosphate with an acyl sulfonamide, strongly binds Mycobacteria tuberculosis BPL (MtBPL); a co-crystal revealing multiple hydrogen-bonding interactions between the protein and the acyl sulfonamide.12 More recently, a biotin-5′-AMP analog replacing the acyl phosphate with a 1,2,3-triazole was found to bind tightly to BPL and exhibit >1100-fold selectivity for the S. aureus BPL over the human homologue.13 This suggests the possibility of designing potent inhibitors of bacterial BPL. However, no similar approach has been used to study the function of HLCS or human BPL.</p><p>A model of the HLCS/biotin-5′-AMP complex as well as the crystal structure of biotin-5′-AMP with BPL from Pyrococcus horikoshii OT3 (pdb:1wqw) suggests the importance of hydrogen bonding involving the carbonyl and phosphonate oxygen (Figure S1).13b,26 The β-ketophosphonate and β-hydroxyphosphonate analogs introduced here maintain the natural charge state of biotin-AMP and place a basic oxygen atom beta to the phosphonate group. However, in contrast to the BirA inhibitors described above, the ketophosphonate (β-ketoP, 6) incorporates an electrophilic carbon at the location of the original acyl group in biotin-5′-AMP. Although the reduced activity of the new inhibitors compared with biotinol-AMP suggests that preservation of an electrophilic center (C=O) or hydrogen bonding group (CHOH) beta to phosphonate is of limited importance in inhibitor design, we note that the 1,2,3-triazole analogs completely lacking a carbonyl group show no inhibition toward human BPL. It is also possible that conformational differences between the acyl phosphate of biotin-5′-AMP and the phosphonate of 6a and 6b might also contribute to the reduced binding observed.</p><p>In conclusion, we have described a new class of inhibitors of holocarboxylase synthetase HLCS based upon replacement of the ester of biotin-5′-AMP with a ketone or a secondary alcohol. The analogs produce significant levels of inhibition with isolated enzyme. Efficacy of the new inhibitors in vivo has not been tested and further investigations are warranted.</p>

PubMed Author Manuscript

Catalytic Dynamic Kinetic Resolutions with N-Heterocyclic Carbenes: Asymmetric Synthesis of Highly Substituted \xce\xb2-Lactones

An N-heterocyclic carbene (NHC)-catalyzed dynamic kinetic resolution of racemic \xce\xb1-substituted \xce\xb2-keto esters has been developed. This method relies on the epimerization of an NHC-enol intermediate before subsequent aldol/acylation events. Highly substituted \xce\xb2-lactones are produced in good yield (50\xe2\x80\x9388%) with good to excellent diastereoselectivity (5:1 to 20:1 dr) and excellent enantioselectivity (up to 99% ee).

catalytic_dynamic_kinetic_resolutions_with_n-heterocyclic_carbenes:_asymmetric_synthesis_of_highly_s

<!>Experimental Section

<p>The conversion of a racemic substrate to enantioenriched products, commonly referred to as a kinetic resolution, is an established method with broad applications.[1],[2] Due to the intrinsic lack of efficiency with this strategy where the theoretical yield is only 50%, many creative dynamic kinetic resolutions (DKRs) have been developed in which >99% yields are possible.[3] For these processes, the rapid interconversion between each enantiomer of a starting material provides an opportunity for a selective catalyst system[4] to promote a desired reaction favoring only one enantiomer, assuming the interconversion is faster than the irreversible step. Both transition metal and enzyme-based DKR reactions have seen significant development over the past two decades.[5] Dynamic kinetic resolutions based on organocatalysis are emerging as powerful and complementary approaches for the conversion of racemic substrates to products with high enantioselectivity. Organocatalytic DKRs have employed a wide variety of activation strategies, including Lewis bases,[6] hydrogen bond donors,[7] chiral Brønsted acids,[8] enamine/iminium ions,[9] and peptide-based biaryl oxidations.[10] While N-heterocyclic carbene catalysis has been used in very few cases of traditional kinetic resolutions[11],[12] or parallel kinetic resolutions, NHC-DKRs should provide significant new opportunities.[13] In this communication, we report a new NHC-DKR with β-keto esters to generate highly substituted β-lactones[14] with excellent levels of stereoselectivity (Scheme 1). This unique process leverages the basic conditions necessary to generate the NHC catalyst from the azolium salt to promote racemization of the substrates.</p><p>For a successful NHC-catalyzed DKR process, the racemic starting material must undergo facile epimerization under basic conditions typical with carbene generation. Given our experience with NHC-catalyzed homoenolate[15] and enolate reactions[16] we predicted that a similar approach could be applied to racemic α-substituted β-keto esters (1). The potential epimerization of the α-position under the basic reaction conditions could favor only one NHC-enol intermediate that would undergo the aldol/acylation (vide infra). A significant challenge for this process is controlling the selectivity to favor one of four potential diastereomeric products (Scheme 1). This approach provides access to cyclopentane-fused β-lactones (2) or substituted cyclopentanes (3) with three contiguous stereogenic centers (Scheme 1, eq 1).</p><p>Our investigation began by treating racemic enals with an azolium salt (10 mol %) and a stoichiometric amount of Hünig's base (Table 1). Azolium precursors A–D furnished the product β-lactone with moderate diastereoselectivity but with little to no conversion (entries 1–4). Triazolium E gave complete conversion of starting material with 7:1 dr and 90% ee, but there was a significant amount of decarboxylation to the corresponding cyclopentene, with a 31% isolated yield of the product lactone (entry 5).[17],[18] In this NHC-DKR process, the mitigation of the decarboxylation pathway is essential because complete decarboxylation of a 5:1 diastereomeric mixture of (+)-2a and (−)-2b would provide a cyclopentene product with only a ~65% ee.</p><p>We explored azolium F which we have observed in previous carbene-catalyzed processes to provide high yields and selectivity.[17e] Gratifyingly, this NHC catalyst afforded a reproducible increase in enantioselectivity (99% ee) with a slight decrease in diastereoselectivity. More importantly, the use of azolium F provided a much-improved yield with trace amounts of decarboxylation (72% isolated yield, entry 6). In an attempt to improve the diastereoselectively of this process we examined other 1,3–dicarbonyl compounds (entries 6–10) with varying acidities at the alpha position with the idea that different electron withdrawing groups would perturb the acidity of the α-position and thus increase the rate of enantiomer interconversion. With a β-ketoamide, the resultant lactone was formed in a 1:1 mixture of diastereomers, presumably due to the A1,3 strain created when tautomerizing to the enol form (entry 7).[19] The β-keto thioester enal was not productive and only decomposed rapidly under the reaction conditions to unidentified products (entry 8). Fluorinated ester substrates were also investigated as a means to modulate the acidity of the α-proton. The trifluoroethyl ester gave an increase in diastereoslectivity (8:1 dr) with excellent enantioselectivity (99% ee) but with a decreased yield (64%) (entry 9). Increasing the acidity further by switching to the hexafluoroisopropyl ester gave incomplete conversion and unproductive products (entry 10). Other alkyl esters such as tert-butyl or benzyl gave similar or lower selectivities (not shown).</p><p>We also examined the effect of base and solvent on the ethyl ester substrate. Our goal was to increase the α-proton acidity through the use of a more polar solvent or stronger base. Coordinating polar aprotic solvents (i.e.–THF, DMSO) provided sluggish reactions with poor selectivity while stronger bases resulted in starting material decomposition or poor reactivity. After screening a large array of solvent combinations and bases (not shown), it was determined that the most favorable conditions were observed when running the reaction in 1,2-dichloroethane (DCE) with cesium carbonate (30 mol %) as the base. Under these optimized conditions, the desired lactones were isolated in high yield (86%) with good diastereoselectivity (6:1 dr) and excellent enantioselectivity (entry 11). With the optimized reaction conditions established, we explored the scope of this NHC-catalyzed DKR (Table 2). Electron-deficient aromatic ketones led to desired lactones 6–12 in good yield (64–88%), with good diastereoselectivity, and excellent enantioselectivity (97–99%). Notably, the 2-fluorophenyl substrate exhibited high diastereoselectivity (20:1) and enantioselectivity, but with a decreased yield. Para-phenyl substitution provided the lactone product (13) with excellent diastereoselectivity. Electron-donating substitution was tolerated in the meta-position furnishing lactone products 15 and 16 in high yield (73–77%) with good selectivities. Finally, the cyclopropyl ketone proceeded to furnish lactone 17 in moderate yield and enantioselectivity, but with excellent diastereoselectivity.[20]</p><p>In this organocatalytic DKR process, electron-rich aryl ketones proceeded directly to the cyclopentene product (Table 3). Presumably, more available electron density promotes a facile decarboxylation. Unfortunately, along with this process comes an inherent decrease in enantioselectivity and investigations are underway to increase the selectivity. The 2-ethoxy-phenyl ketone proceeded in high yield and enantioselectivity (90% ee). Other substrates, however, provided decreased yields and enantioselectivities of cyclopentene products 19–21.[21]</p><p>Given the highly selective nature of the method, we envisioned a stereodivergent parallel kinetic resolution (PKR) would be possible.[22] In contrast to a standard kinetic resolution, a PKR converts both enantiomers of a starting material into two distinct products. For this process we employed a racemic substrate that is not epimerizable under the reaction conditions (Scheme 2). The reaction proceeded as planned with α-methylated β-keto ester 22 and complete cyclization to diastereomeric β-lactone products 23 was achieved in excellent yield (50% theoretical yield for each) and enantioselectivity.</p><p>β-Lactones are highly useful building blocks for the synthesis of target compounds, especially in the area of natural product synthesis.[23] To demonstrate the potential utility of this DKR, we processed these β-lactones into diverse compounds (Scheme 3). The treatment of 5a with benzylamine gave amide 24 in 94% yield while heating 5a with SiO2 at 60 °C[24] followed by lithium aluminum hydride provided homoallylic alcohol 25 in 84% yield over 2 steps. A selective hydrogenation with Crabtree's catalyst[25] afforded trans-cyclopentane 26 (79%, 8:1 dr), while 10% Pd/C hydrogenation favored the cis-cyclopentane (91%, 2.2:1 dr, see supporting information). Following a similar reaction sequence, sulfonylamine 27 was prepared in 3 steps with a 60% overall yield. The exposure of 27 to Du Bois's conditions generated bicyclic aziridine 28 (91% yield).[26]</p><p>Our current understanding of this DKR process is depicted in Scheme 4. Addition of the NHC to enal (±)-1 induces the formation of the extended Breslow intermediate. The homoenolate undergoes β-protonation to form enols I and III. The mildly basic reaction conditions allow for these two intermediates to be in rapid equilibrium. Addition to the re–face of the enol is favored for both intermediates (I and III) due to a favorable hydrogen bonding interaction between the NHC-enol and ketone forming a 6-membered transition state (II and IV). The importance of the hydrogen bonding is reinforced by the poor conversion observed with solvents that disturb this hydrogen-bonding interaction (coordinating polar aprotic). The major diastereomer ((+)-2a) arises from enol II, in which the ester is in a pseudo axial orientation. This conformation is more favorable than IV, which yields minor diastereomer ((+)-2b), due to destabilization created by a gauche interaction of the ethyl ester and the aryl group. This destabilization is increased with ortho substituted substrates (favoring intermediate II) and as a result virtually none of the minor diastereomer ((+)-2b) is observed (lactone 10 and cyclopentene 18). The combination of a fast aldol addition with II compared to IV and the rapid equilibrium of I and III drives the reaction primarily in the direction of (+)-2a (97–99% ee). The rationale behind the absence of the corresponding enantiomer (−)-2a (which would arise from si–face attack of III) is twofold. First, there is an unfavorable aryl-ester syn destabilization, and second, there is no hydrogen bonding to promote the aldol/acylation. The minor diastereomer (+)-2b (arising from intermediate IV) was likewise formed with high enantioselectivity (92–95% ee) (Table 2, lactones 5 and 6).</p><p>In conclusion, a new NHC-catalyzed dynamic kinetic resolution has been developed. This process takes advantage of the conditions necessary to generate the active NHC catalyst to simultaneously promote the epimerization of a β-ketoester substrate. The present study provides efficient access to highly enantioenriched β-lactones and cyclopentenes in good yield with good to excellent diastereoselectivity. With this blueprint for a new type of organocatalytic DKR process, further studies to probe the general reactivity, expand the substrate scope and application of this new NHC-DKR process in total synthesis are underway.</p><!><p>In a nitrogen filled dry box a screw-capped vial equipped with a magnetic stirbar was charged with the corresponding enal β-ketoester 4 (0.400 mmol), azolium precatalyst F (0.07 equiv), and cesium carbonate (0.30 equiv). The vial was capped with a septum cap, removed from the drybox and put under positive N2 pressure. The heterogeneous mixture was then diluted with degassed 1,2-dichloroethane (12 mL, 0.033 M) and stirred for 12 hours under static nitrogen pressure. Upon consumption of the aldehyde (all reactions were completed within 12 hours) the reaction mixture was diluted with dichloromethane (5 mL) washed with brine (10 mL) and separated. The aqueous phase was back extracted with dichloromethane (5 mL). The combined organic layers were filtered through a Biotage ISOLUTE® phase separator, and the organic filtrate was concentrated. The material was purified by flash chromatography with EtOAc/hexanes to afford the corresponding β-lactone.</p>

PubMed Author Manuscript

Biofuel metabolic engineering with biosensors

Metabolic engineering offers the potential to renewably produce important classes of chemicals, particularly biofuels, at an industrial scale. DNA synthesis and editing techniques can generate large pathway libraries, yet identifying the best variants is slow and cumbersome. Traditionally, analytical methods like chromatography and mass spectrometry have been used to evaluate pathway variants, but such techniques cannot be performed with high throughput. Biosensors - genetically encoded components that actuate a cellular output in response to a change in metabolite concentration - are therefore a promising tool for rapid and high-throughput evaluation of candidate pathway variants. Applying biosensors can also dynamically tune pathways in response to metabolic changes, improving balance and productivity. Here, we describe the major classes of biosensors and briefly highlight recent progress in applying them to biofuel-related metabolic pathway engineering.

biofuel_metabolic_engineering_with_biosensors

Introduction<!>State-of-the-art in analytical metabolite detection<!>Classes of genetically encoded biosensors<!>Fluorescent protein biosensors<!>RNA biosensors<!>Cytosolic transcription factor (TF) biosensors<!>G-protein coupled receptor (GPCR) and Two-component biosensors<!>Biosensor Applications<!>Biosensors with phenotypic output<!>Biosensors for dynamic pathway regulation<!>Outlook<!>Input domains<!>Allostery<!>Output domains

<p>Metabolic engineering of microbes holds the promise of producing many classes of chemicals, including fuels, from renewable feedstocks [1]. However, to compete with established production methods, engineered organisms must be highly productive, efficient, and robust at industrial scales. Many factors, such as the enzymes employed, regulatory proteins and genetic regulatory elements, can affect these phenotypes, and so a fundamental aspect of pathway engineering is identifying the complex genetic alterations required to create an optimized strain. While there are numerous ways to engineer genetically diverse strain libraries - in both random and/or directed fashions [2,3] - there are few assays that scale with the bandwidth of modern genetics (Figure 1). As such, it is critical to develop novel detection technologies in order to bring the full power of genetics to bear on metabolism.</p><p>An effective screening tool must be specific, high throughput, and sensitive to relevant metabolite concentrations. Most metabolites, except for special cases or by the use of exogenous chemical dyes (reviewed in [4]), cannot be measured using rapid optical methods. Chromatography and mass spectrometry (MS) are thus the only analytical tools available for measuring most biofuel-related metabolites despite their low throughput. Biosensors, genetically encoded components that respond to an input signal (e.g. metabolite concentration) and transduce that signal into a detectable output (e.g. fluorescence or gene expression), are emerging as a high-throughput alternative for measuring metabolite concentrations in vivo. Often adapted from natural proteins or aptamers, biosensors can be specific, sensitive, and non-destructive.</p><p>Here, we provide an introduction to biosensors and their use in modern metabolic engineering. As a point of comparison, we start with recent advances in analytical chemistry (reviewed in greater detail in [5]) and contrast this with the commonly employed classes of biosensors. We then focus in detail on the applications of specific biosensors to biofuel-related metabolic engineering.</p><!><p>Analytical chemistry methods, including chromatography and MS, are the gold standard for measuring metabolism. These methods are label-free, sensitive, and can detect many (e.g. 100+) metabolites in a single measurement [6]. However, these methods require time- and labor-intensive metabolite extractions that result in destructive, bulk measurements that are generally low throughput (101–103 per day). Two emerging MS-based platforms that may aid in overcoming these limitations are the RapidFire high-throughput MS system from Agilent Technology, Inc. [7] and surface-based MS techniques, such as Nanostructure-Initiator MS (NIMS) [8].</p><p>RapidFire uses robotics to automate the metabolomics workflow. Samples in microtiter plates are purified by solid-phase extraction and directly injected into an MS instrument. The instrument can processes a single sample in less than 15 s, which is over 100x faster than traditional liquid-chromatography-MS measurements [7]. NIMS is a surface-assisted laser desorption/ionization technique that requires little sample preparation and uses a liquid "initiator," instead of a co-crystallization matrix, to produce spectra with high sensitivity and lower noise in the metabolite mass region. NIMS was recently used to screen >100 glycoside hydrolases (enzymes important for biomass hydrolysis) with a wide range of substrates and reaction conditions to generate more than 10,000 data points [9]. Although surface and automated MS techniques are not yet widely used, it is likely they will continue to increase in throughput and find applications in metabolic engineering.</p><!><p>Biosensors are genetically encoded components that convert an input signal (e.g. metabolite concentration) into a measurable output like fluorescence or gene expression (Figure 2). In the following sections, we introduce common classes of biosensors constructed from fluorescent proteins, RNA, cytosolic transcription factors (TFs), G-protein-coupled receptors (GPCRs) as well as two-component systems and discuss their inherent advantages and disadvantages.</p><!><p>Genetically encoded biosensors based on Förster resonance energy transfer (FRET) or single fluorescent proteins are promising tools for the analysis of metabolic pathways and their products. FRET biosensors consist of a ligand-binding domain (LBD) attached to a pair of fluorescent proteins that have overlap in their emission and excitation spectra, capable of FRET (Figure 2) [10]. Binding of a metabolite to the LBD alters the distance between the two fluorophores and changes the energy-transfer efficiency, measured as a ratio of fluorescence. While FRET biosensors have been developed for many different metabolites [11,12], they typically exhibit low dynamic ranges (e.g. tens of % change in signal) that significantly impede their use in screening applications. Single fluorescent protein biosensors (SFPBs) are fluorescent proteins that either directly detect input signals or are inserted within the primary sequence of a conformationally-labile LBD such that ligand binding affects fluorescence intensity (Figure 2) [13]. While genetically encoded SFPBs have high dynamic range (e.g. 10-fold) and are used in cell biology studies [14,15], they are not widely used in metabolic engineering [16••]. There are currently few available SFPBs due to the difficulty in engineering the coupling between an LBD and a fluorescent protein partner. Methods enabling rapid SFBP engineering may therefore be useful to increase the availability of this promising class of biosensors [17].</p><!><p>Riboswitches are naturally evolved ligand-responsive RNA elements that possess two components: a sensor (aptamer) domain that detects metabolite binding and a regulatory domain that converts binding-induced conformational changes into changes in gene expression (Figure 2) [18]. RNA-based biosensors also benefit from known techniques (e.g. SELEX) for generating aptamers against new metabolites [19] and have been adapted as biosensors for engineered pathways [20–22]. To date, however, use in metabolic engineering has been limited, likely from the challenges of recapitulating in vitro behavior within the cellular environment.</p><!><p>TF-based biosensors detect environmental changes, such as metabolite levels, and alter gene expression in response (Figure 2). The most widely used are bacterial TFs, which are composed of an LBD that controls the engagement of a cognate DNA-binding domain to promoter/operator sites associated with target genes. Depending on the TF, DNA binding may lead to gene repression or activation. These biosensors can offer high sensitivity and dynamic range; small changes in ligand concentration are amplified through gene expression into large changes in protein abundance.</p><p>An early implementation of TF–based biosensors was the development of whole-cell biosensors where expressed reporter genes (e.g. luciferase or β-galactosidase) were used to detect environmental pollutants [23]. Subsequently, TF-biosensors have been used in high-throughput strain evaluation by linking metabolite levels to fluorescence [24,25] and growth advantages such as antibiotic resistance [24,26,27•]. More recently, TF-biosensors have been linked to regulatory or pathway genes to provide dynamic feedback within engineered pathways [28–31••]. This modularity of input and output domains in TF-based biosensors makes them attractive for many metabolic engineering applications.</p><p>Despite the increasingly widespread adoption of TF-based biosensors for metabolite sensing, there are potential disadvantages. First, there is a large difference in the timescales of metabolite turnover (~1 sec) and those of transcription and translation (~1–10 min [32]), which makes real-time sensing impossible when using TF-based biosensors. Additionally, TF-based biosensors are not always robust; bacterial TFs may not be portable to eukaryotes due to fundamental differences in the transcriptional process. Finally, expression of a non-native TF may have unanticipated side-effects, including non-specific binding to DNA and interfering with transcription.</p><!><p>An alternative to cytosolic TFs is GPCR-based biosensors expressed on the cell surface [33–35]. For these biosensors, the binding of an extracellular metabolite to a GPCR results in signal transduction and, ultimately, changes in gene expression (Fig. 2b). As with TF-based biosensors, the modular nature of GPCR-based biosensors and the wide variety of molecular specificities [36] make them broadly useful for metabolite sensing. However, one potential caveat is that sensing only occurs extracellularly, which may limit applications. The analog of GPCRs for prokaryotes are the two-component regulatory systems in which one component acts as a transmembrane sensor and the second component acts as an intracellular response regulator. Studies demonstrating that the extracellular sensing domain of one transmembrane sensor could be fused with the intracellular domain of another to create a hybrid biosensor, as well as studies showing that the promoter for a particular response regulator could be used to control the expression of an arbitrary output of interest, have been met with excitement in the synthetic biology community [37,38]. However, the engineering of two-component systems has met with practical difficulties [39] and more studies will be needed to determine design principles of re-engineering ligand specificity [40].</p><!><p>In the following sections we highlight recent applications of biosensors to the i) isolation of improved mutants and ii) dynamic control of metabolic pathways (Figure 3a). TF-based biosensors predominate in these examples as they have been the most widely adopted, to date, in biofuel-related metabolic engineering.</p><!><p>Biosensors are often used to generate a phenotypic output that can be screened or selected (Figure 3a–b). The sensed ligand is generally an intermediate or the product of a desired pathway and the biosensor is used to isolate genotypes with higher titers and improved pathway flux. For example, malonyl-CoA production, the first committed step of fatty acid biosynthesis (Figure 3e), was targeted for improvement in S. cerevisiae using a TF-based biosensor [25]. FapR, a TF that represses expression in the absence of malonyl-CoA, was used in a FACS screen to isolate genes from a cDNA library of ~106 variants that improved malonyl-CoA production (Figure 3a). Similarly, an SFPB for the fundamental co-factor NADPH was used to isolate production strains with more biosynthetic potential. This biosensor allowed rapid micro-well plate quantification of 624 computationally designed synthetic-pathway variants for high NADPH titer in E. coli [16]. When combined with a terpenoid biosynthetic pathway, the improved NADPH pool increased production by nearly 2-fold (Figure 3e). The rapid screening achieved with these examples illustrates the advances that can be made when production strains are evaluated in high throughput.</p><p>Biosensors can also generate an output that imparts a growth advantage to cells, allowing for growth selection and increasing the potential number of testable designs by orders of magnitude (Figures 1 and 3b). Raman and coworkers optimized a TolC antibiotic-resistance output linked to a TF biosensor for the commodity chemical glucaric acid [26]. Selection of a 107 library after multiplex genome engineering led to a strain with 22-fold greater glucaric acid titer. Alternatively, an innovative biosensor design developed by Chou and Keasling to increase isopentenyl pyrophosphate (IPP, a terpene building block, Figure 3e) production had mutation rate as the output [27•]. Starting with a high mutation rate, an artificial TF decreased the expression of the mutD5 polymerase (mutator) as IPP concentration increased. This stabilized high IPP-producing genotypes, resulting in a ~17-fold improvement in lycopene (terpene) production over strains with no link between IPP concentration and mutD5. Coupling the appropriate biosensor with the right selection can therefore enrich immense libraries (>106 variants) to isolate improved strains.</p><!><p>Metabolic pathways are complex and tightly regulated in their native context. Maximizing pathway yields requires careful balancing of pathway flux to prevent bottlenecks and/or the accumulation of toxic intermediates. Dynamic pathway regulation, using TF-based biosensors, allows for flux to be altered in response to changing cellular and environmental conditions (Figure 3c). A groundbreaking study in 2000 by Farmer and Liao demonstrated the utility of this approach by improving lycopene yield in E. coli using a biosensor for acetyl-phosphate [28]. The concentration of acetyl-phosphate is a proxy for glucose availability and reduced growth (Figure 3e). Lycopene pathway genes, which normally inhibit growth, were gradually activated by the TF-biosensor as cells exited exponential phase. This strategy improved lycopene titer by almost 20-fold over static expression. Pathway feedback can be especially beneficial when intermediates are toxic, preventing dangerous buildups. For example, a prominent study engineered a fatty acid ethyl ester (FAEE) pathway using a fatty acyl-CoA TF biosensor (FadR, [30]). The introduced feedback both reduced toxic ethanol accumulation and prevented depletion of cellular fatty acids, thereby creating a strain with 3-fold greater FAEE yield.</p><p>As mentioned in the previous section, malonyl-CoA is a crucial intermediate for the production of biofuel-related molecules (Figure 3e). Because of this, it has been targeted in multiple efforts to implement biosensor-based regulation. Xu and Koffas used a single TF biosensor (FapR) to repress the acc operon (catalyzing acetyl-CoA carboxylation to malonyl-CoA) and, in the presence of malonyl-CoA, induce a fatty acid biosynthesis operon (malonyl-CoA conversion to fatty acid) [31••] (Figure 3d). This combined regulation led to a >3-fold increase in fatty acid yield and improved cell growth as compared to the un-regulated strain. Interestingly, while FapR naturally represses fatty acid biosynthesis in B. subtilis, this work serendipitously identified FapR activation behavior in the promoter region of gapA (glyceraldehyde-3-phosphate dehydrogenase A) within the E. coli host, providing both types of control in the engineered strain. To achieve the same positive control with FapR in E. coli, another group used a genetic invertor to regulate acc expression, thus avoiding the need to re-engineer the TF or operator [41,42]. The same bacterial TF was also recently employed as a biosensor in S. cerevisiae to provide negative control of mcr (malonyl-CoA conversion to 3-hydroxypropionic acid) [43]. Fatty acids and their intermediates are desirable end products and these successful biosensor engineering efforts point to exciting future possibilities. We expect further integration of new biosensors, such as those for NADPH [44], to improve yields.</p><!><p>In addition to the biosensor designs discussed above, there have been several recent innovative approaches to biosensor construction. For example, Feng and colleagues developed eukaryotic biosensors based on protein lifetime that could be applied to biofuel-related molecules [45]. In this approach, a library of mutagenized LBDs fused to a reporter are screened for high signal in the presence of ligand, due to conditional protein stabilization, and low signal in the absence of ligand, due to degradation by the proteasome. Enzyme-based biosensors are another alternative. Such biosensors enzymatically convert a metabolite of interest to a molecule that is directly detectable, or for which there is a pre-existing genetically encoded biosensor [46, 47•,48]. SensiPath, a recently reported web-based tool (http://sensipath.micalis.fr) that searches for enzymatic pathways that convert metabolites to detectable molecules, may be especially useful in trying to achieve these goals [49•].</p><p>Despite these promising approaches, we believe that there is significant potential for single-component allosteric biosensors. Single protein molecules that link input and output domains via allostery, such as SFPBs, are powerful tools due to their genetic portability and fast response rates. We have already highlighted the benefit of biosensor-mediated feedback for pathway balancing and improved strain performance. However, to date, only multi-component, TF-based biosensors have been used for this purpose. These biosensors respond on a timescale that is too slow for many metabolic applications, as metabolic fluctuations are 10–100 times faster than transcription and translation [32]. In addition, delayed negative feedback can lead to oscillations in the concentration of regulated species [31••,50]. Using a pathway enzyme as an output in a single-component biosensor, as is common in natural metabolism [51], would dramatically improve the speed and robustness of feedback control. In this section, we highlight some of the future challenges and opportunities in developing this type of biosensor by focusing on techniques for the identification of novel input domains, the engineering of allostery and the exploration of new functional outputs.</p><!><p>A significant barrier to the development of biosensors is identifying novel LBDs. One approach to overcoming this problem is to mine the large numbers of naturally occurring proteins that are being identified in genome and metagenome sequencing projects (Figure 4a). Vetting and coworkers recently highlighted this strategy by using high-throughput protein expression and differential scanning fluorimetry to screen 158 candidate LBDs against 189 ligands to identify 40 new ligand-LBD interactions [52••]. Alternatively, techniques such as substrate-induced gene expression screening (SIGEX), which involves inserting restriction enzyme-digested (meta)genome fragments upstream of a reporter gene, may also be adapted to identify previously uncharacterized LBDs [53].</p><p>An alternative method for expanding the number of LBDs is to employ computational design approaches (Figure 4a). Tinberg and colleagues reported the design of a ligand-binding protein for the steroid digoxigenin (DIG) by computationally designing the protein-ligand interface and intramolecular interactions using a protein scaffold of unknown function that did not originally bind the target molecule. Of 17 designs that were experimentally characterized, two were functional for DIG-binding, and optimization of one of these constructs using site-saturation mutagenesis coupled with selections resulted in a 75-fold improvement in binding affinity [54]. Similarly, Taylor and coworkers have reported the redesign of the lac repressor transcription factor for a number of novel inducers using a combination of computational protein design, mutagenesis and gene shuffling [55]. As computational design of binding improves, a potential application of such work would be in tuning the affinity of a biosensor, and thus the operational range, for various applications. The utility of such biosensor tuning was demonstrated by the Frommer lab by developing a family of glucose biosensors with a range of affinities to visualize the different responses to glucose perfusion in various areas of the plant [56]. While still very challenging, continued advances in this area using de novo protein design, the redesign of existing LBD scaffolds and directed evolution will expand the number of LBDs and consequently the number of metabolite biosensors.</p><!><p>Once an LBD is identified for a small molecule of interest, the greatest challenge in creating a single-component biosensor is often engineering the allosteric connection to the desired output domain. One proven method for doing so has been via domain insertion whereby one protein domain is inserted into another such that the functions of the two independent domains are coupled (Figure 4b). However, while this strategy has proven successful, for example, as demonstrated by Guntas and coworkers in the development of a maltose-dependent β–lactamase biosensor [57], it is plagued by its low-throughput nature due to the difficulties in reliably predicting the insertion sites for linking the associated domains. To overcome this, our group recently reported a strategy for the rapid construction of biosensors termed domain-insertion profiling with sequencing (DIP-seq) [17]. In this approach, we created diverse libraries of potential SFPBs using modified transposons and then used high-throughput activity assays to identify functional biosensors. While we have applied this tool to the rapid construction of SFPBs, it may also be applied to the generation of allosteric biosensors of any type and function [58].</p><p>Engineering allostery through the redesign of protein surfaces to include ligand-binding sites may be another route (Figure 4b). Work from the Ranganathan group has shown that there are networks of physically connected and coevolving amino acids that link protein active sites to spatially distinct surface sites [59]. With continued computational advances, it may one day be possible to reliably predict these surface sites and further engineer them to include ligand-binding sites for the development of allosterically regulated biosensors.</p><!><p>The majority of reported biosensors employ fluorescence as an output, which is effective for visualization and screening. However, given the availability of alternative functional output domains, there are likely many alternative novel biosensor applications (Figure 4c). As mentioned above, the use of selection over screening can enable testing of orders of magnitude more designs by linking metabolite binding to host fitness. For example, by allosterically coupling E. coli MBP to TEM1 β-lactamase, Guntas and coworkers developed maltose-dependent switches that conferred growth selection phenotypes in the presence of β–lactam antibiotics [57]. Similar strategies may also be adopted for biofuels by linking the detection of the desired molecule to the growth of the host.</p><p>Compared with most available biosensors that have only single detection channels, multiplexed biosensors could provide more information by increasing the number of available outputs that can be detected at the same time (Figure 4c). For biosensors that provide fluorescence readout, the most common form of multiplexing involves combining biosensors with different wavelengths but this approach is limited due to spectral overlap. However, combining intensity measurements with time-resolved measurements may provide a means to multiplex in both the wavelength and time domains. Employing such time-resolved fluorescence lifetime biosensors are advantageous because they are quantitative and independent of biosensor concentration. Recently, Mongeon and coworkers demonstrated that Peredox, a previously reported SFPB for NADH:NAD+ ratio, could also be used as a fluorescence lifetime biosensor because it showed a large change in fluorescence lifetime over its sensing range [60].</p><p>Along with identifying and designing new ligand-binding domains, engineering methods for allosteric enzyme design are long-term goals that will benefit the field. We expect that future improvements in the identification of ligand-binding domains and in the predictable engineering of allostery in fluorescent proteins and enzymes will provide a biotechnological backbone that vastly improves our capacity to design, screen, and select pathway variants for the biological production of fuels. Ultimately, these improvements may enable biocatalysis to compete with fossil fuels in reliability and economic terms.</p>

PubMed Author Manuscript

MicroRNA-138 suppresses epithelial\xe2\x80\x93mesenchymal transition in squamous cell carcinoma cell lines

Down-regulation of miR-138 (microRNA-138) has been frequently observed in various cancers, including HNSCC (head and neck squamous cell carcinoma). Our previous studies suggest that down-regulation of miR-138 is associated with mesenchymal-like cell morphology and enhanced cell migration and invasion. In the present study, we demonstrated that these miR-138-induced changes were accompanied by marked reduction in E-cad (E-cadherin) expression and enhanced Vim (vimentin) expression, characteristics of EMT (epithelial\xe2\x80\x93mesenchymal transition). On the basis of a combined experimental and bioinformatics analysis, we identified a number of miR-138 target genes that are associated with EMT, including VIM, ZEB2 (zinc finger E-box-binding homeobox 2) and EZH2 (enhancer of zeste homologue 2). Direct targeting of miR-138 to specific sequences located in the mRNAs of the VIM, ZEB2 and EZH2 genes was confirmed using luciferase reporter gene assays. Our functional analyses (knock-in and knock-down) demonstrated that miR-138 regulates the EMT via three distinct pathways: (i) direct targeting of VIM mRNA and controlling the expression of VIM at a post-transcriptional level, (ii) targeting the transcriptional repressors (ZEB2) which in turn regulating the transcription activity of the E-cad gene, and (iii) targeting the epigenetic regulator EZH2 which in turn modulates its gene silencing effects on the downstream genes including E-cad. These results, together with our previously observed miR-138 effects on cell migration and invasion through targeting RhoC (Rho-related GTP-binding protein C) and ROCK2 (Rho-associated, coiled-coil-containing protein kinase 2) concurrently, suggest that miR-138 is a multi-functional molecular regulator and plays major roles in EMT and in HNSCC progression.

microrna-138_suppresses_epithelial\xe2\x80\x93mesenchymal_transition_in_squamous_cell_carcinoma_cell

INTRODUCTION<!>Cell culture and transfection<!>Fluorescence immunocytochemical analysis<!>Western blot analysis<!>In vitro cell migration and invasion assays<!>qRT-PCR (quantitative real-time PCR) analysis<!>MicroRNA target prediction<!>Dual-luciferase reporter assay<!>Statistical analysis<!>miR-138 regulates EMT in HNSCC cell lines<!>miR-138 targets Vim in HNSCC cells<!>miR-138 down-regulates transcriptional repressors (ZEB2 and Snai2) and promotes E-cad expression in HNSCC cells<!>miR-138 targets epigenetic regulator EZH2 and regulates E-cad expression in HNSCC cells<!>DISCUSSION

<p>MicroRNAs are endogenous small non-coding RNAs that control the target gene expression at the post-transcriptional level. Several microRNAs have been functionally classified as protooncogenes or tumour suppressors and are aberrantly expressed in different cancer types [1,2]. Deregulation (e.g. overexpression or loss of expression) of these `cancerous' microRNAs can figure prominently in tumour initiation and progression by facilitating an inappropriate cellular programme that promotes uncontrolled proliferation, favours survival, induces EMT (epithelial–mesenchymal transition) and/or promotes invasive behaviour.</p><p>miR-138 (microRNA-138) has been thought to regulate a number of essential biological processes, including the development of mammary glands [3], regulating dendritic spine morphogenesis [4], modulating cardiac patterning during embryonic development [5] and thermotolerance acquisition [6]. The deregulation of miR-138 has been frequently observed in a number of cancer types, including thyroid cancer [7], lung cancer [8] and leukaemia [9]. The frequent down-regulation of miR-138 has also been observed in HNSCC (head and neck squamous cell carcinoma), including cases that originated from various anatomic sites such as the pharynx [10], the tongue [11–14] and other sites of the oral cavity [15]. Two miR-138 precursor genes, termed miR-138-1 and miR-138-2, were identified recently in the mouse genome [16], and their human homologues were mapped to chromosome 3p21.33 and 16q13 respectively. Interestingly, LOH (loss of heterozygosity) at both chromosome loci have been detected frequently in HNSCC [17–19].</p><p>Recent studies by us have demonstrated that the down-regulation of miR-138 in HNSCC cell lines enhances cell migration and invasion [10,12], and is associated with marked morphological changes (e.g. loss of polarity and cell–cell adhesion, and the acquisition of mesenchymal-like cell morphology) that are characteristics of EMT. However, the molecular mechanism(s) underlying the observed effect of miR-138 on EMT is poorly understood. The aim of the present study was to investigate the functional roles of miR-138 in EMT.</p><!><p>The human HNSCC cell lines used in the present study were maintained in DMEM/F12 (where DMEM is Dulbecco's modified Eagle's medium) supplemented with 10% FBS (foetal bovine serum), 100 units/ml penicillin and 100 μg/ml streptomycin (Gibco) at 37°C in a humidified incubator containing 5% CO2. For functional analysis, miR-138 mimics and non-targeting miRNA mimics (Dharmacon), LNA (locked nucleic acid) knock-down probe specific to miR-138 (anti-miR-138 LNA) and negative control LNA (Exiqon), and gene-specific siRNAs (small interfering RNAs; On-TargetPlus SMARTpool, Dharmacon) were transfected into cells using DharmaFECT Transfection Reagent 1 as described previously [10,12]. For the induction of EMT, cells were treated with 10 ng/ml TGFβ (transforming growth factor β) 1 as described previously [20,21].</p><!><p>Immunofluorescence analysis was performed as described previously [12]. In brief, cells were cultured on eight-chamber polystyrene vessel tissue-culture-treated glass slides (BD Biosciences) fixed with ice-cold methanol, permeabilized with 0.5% Triton X-100/PBS and blocked with 1% BSA in PBS. The slides were incubated with primary antibodies against E-cad (E-cadherin) (1:100), EZH2 (enhancer of zeste homologue 2; 1:100) (BD Biosciences), Vim (vimentin l) (1:200; Cell Signaling Technologies) or ZEB2 (zinc finger E-box-binding homeobox 2; 1:50; Sigma–Aldrich). The slides were then incubated with a FITC-conjugated anti-rabbit IgG antibody (1:50; Santa Cruz Biotechnology) and Alexa Fluor® 594-conjugated goat anti-mouse IgG antibody (1:400; Invitrogen). The slides were mounted with ProLong Gold antifade reagent containing DAPI (4′ ,6-diamidino-2-phenylindole; Invitrogen) following the manufacturer's protocol. The slides were then examined with a fluorescence microscope (Carl Zeiss).</p><!><p>Western blots were performed as described previously [12] using antibodies specific against E-cad, EZH2 (BD Biosciences), Vim, Snai2 (Snail homologue 2), Suz12 (suppressor of zeste 12; Cell Signaling Technologies), Eed (embryonic ectoderm development protein; Millipore), ZEB2 and β-actin (Sigma–Aldrich).</p><!><p>Transwell assays were performed to assess cell migration and invasion using BD BioCoat Control Cell Culture Inserts (containing an 8.0 μm PET Membrane without matrix) or BD BioCoat BD Matrigel™ Invasion Chamber (containing a layer of BD Matrigel™ Basement Membrane Matrix) respectively. In brief, cells were treated with appropriate microRNA and/or siRNA reagents and then seeded in the upper Boyden chambers of the cell culture inserts. After 24 h of incubation (for migration) or 48 h of incubation (for invasion), cells remaining in the upper chamber or on the upper membrane were carefully removed. Cells adhering to the lower membrane were stained with Diff-Quik stain (Polyscience), imaged and counted using an inverted microscope equipped with a digital camera (Jenco). Experiments were performed in triplicate.</p><!><p>To examine the expressional changes of the EMT-related genes, a RT2 Profiler PCR array for human EMT (Qiagen/SABiosciences) was used which consists of qRT-PCR assays for 84 EMT-related genes. Two additional qRT-PCR assays for RhoC (Rho-related GTP-binding protein C) and EZH2 (Ori-Gene) were also included in the experiments. The relative expression level was computed using the 2ΔΔCt analysis method, where actin was used as an internal reference [22]. For VIM, ZEB2 and EZH2, independent qRT-PCR assays were also performed using TaqMan Gene Expression Assays (Applied Biosystems). Experiments were performed in triplicate.</p><!><p>The candidate targets of miR-138 were identified using miRGen [23], an integrated online database which contains a collection of five bioinformatics tools, including 4-way PicTar, 5-way PicTar, TargetScanS, miRanda at http://www.microrna.org and miRanda at miRBase. In addition, TargetScanHuman 5.1 [24] was also used for predicting the miR-138 targets. As such, the miR-138 targets are predicted by three different methods (PicTar, TargetScan and miRanda) with two different versions for each method. For the present study, genes that were predicted by at least one method were defined as potential miR-138 targets.</p><!><p>The luciferase reporter gene constructs for Vim (pGL-VimE1, pGL-VimE2 and pGL-VimE3) were created by cloning a 62-bp fragment and a 60-bp fragment from the coding region (position 518–579 and position 861′920, GenBank® accession number NM_003380, containing the miRNA-138-binding sites E1 and E2 respectively), and a 56-bp fragment from the 3′-UTR (untranslated region; position 1815′1870, containing the miRNA-138-binding site E3) of the Vim respectively, into the XbaI site of the pGL3-Control firefly luciferase reporter vector (Promega). The corresponding mutant constructs (pGL-VimE1m, pGL-VimE2m and pGL-VimE3m) were created by replacing the seed regions (position 2′8) of the miR-138-binding sites with 5′-TTTTTTT-3′. The luciferase reporter gene constructs for ZEB2 (pGL-ZEBE1, pGL-ZEBE2 and pGL-ZEBE3) were created by cloning a 60-bp fragment and a 70-bp fragment from the coding region (position 1751′1810 and position 3341′3410, GenBank® accession number NM_014795, containing the miRNA-138-binding sites E1 and E2 respectively), and a 70-bp fragment from the 3′-UTR (position 4891′4960, GenBank® accession number NM_014795, containing the identified miR-138-binding site E3) of the ZEB2 gene into the XbaI site of the pGL3 firefly luciferase reporter vector. The mutant constructs for ZEB2 (pGL-ZEBE1m, pGL-ZEBE2m and pGL-ZEBE3m) were created by mutating the seed region (position 2′8) of the miR-138-binding sites to 5′-TTTTTTT-3′. For EZH2, a 70-bp fragment from the coding region (position 1111′1180, GenBank® accession number NM_004456, containing the conserved miRNA-138-binding site Ec1) and a 60-bp fragment from the 3′-UTR of the EZH2 gene (position 2561′2620, GenBank® accession number NM_004456, containing the conserved miRNA-138-binding site Ec2) were cloned into the XbaI site of the pGL3 reporter vector. The mutant constructs for EZH2 (pGL-Ec1m and pGL-Ec2m) were created by replacing the seed region (position 2′8) of the miR-138-binding site with 5′-TTTTTTT-3′. An additional pair of constructs were also created for EZH2 which contained sequences of a previously described poorly conserved miR-138-binding site [6] from the human EZH2 gene (position 2388′2450, GenBank® accession number NM_004456, named pGL-Ep-hsa) and from the chicken EZH2 gene (position 2414′2476, GenBank® accession number XM_418879, named pGL-Ep-gga). The constructs were then verified by sequencing. Cells were transfected with the reporter constructs containing the targeting sequence from the three genes using Lipofectamine™ 2000 (Invitrogen). The pRL-TK vector (Promega) was co-transfected as an internal control for normalization of the transfection efficiency. The luciferase activities were then determined as described previously [12] using a Glomax 20/20 luminometer (Promega). Experiments were performed in quadruplicate.</p><!><p>Statistical analysis was performed using Student's t test. P < 0.05 was considered statistically significant.</p><!><p>The cell lines 1386Ln and 686Tu are previously established HNSCC cell lines with the miR-138 level in 1386Ln significantly lower than in 686Tu [10]. As shown with immunofluorescence analysis in Figures 1(A) and 1(B), ectopic transfection of the miR-138 mimic to the 1386Ln cells led to a dramatic decrease in Vim expression and a significant increase in E-cad expression. When the 686Tu cells were treated with anti-miR-138 LNA, a decrease in E-cad expression and an increase in Vim expression were observed (Figures 1C and 1D). These changes in E-cad and Vim expression were also confirmed by Western blot analysis (Figure 1E). Similar results were also observed in UM1 cells that were treated with the miR-138 mimic, and SCC9 and SCC15 cells treated with anti-miR-138 LNA (Supplementary Figure S1 at http://www.BiochemJ.org/bj/440/bj4400023add.htm). As shown in Figures 1(F) and 1(G), whereas the increased miR-138 level in 1386Ln resulted in reduced cell migration and cell invasion, the reduced miR-138 level in 686Tu led to enhanced cell migration and invasion. This finding is in agreement with our previous observation [12], and is in agreement with the notion that co-ordinated regulation of EMT is essential to cell motility, invasion and metastasis.</p><p>As illustrated in Figure 2(A), treatment of 686Tu cells with TGF-β, a potent EMT inducer, down-regulated the expression of E-cad and enhanced the expression of Vim. TGF-β treatment also led to the down-regulation of miR-138 (Supplementary Figure S2 at http://www.BiochemJ.org/bj/440/bj4400023add.htm). The TGFβ-induced changes in E-cad and Vim expression were blocked by ectopic transfection of miR-138 mimic. As expected, TGFβ treatment led to enhanced cell migration and invasion (Figures 2B and 2C). The TGFβ-induced increase in cell migration and invasion were blocked by transfection of miR-138 mimic. As illustrated in Figure 2(D), ectopic transfection of the miR-138 mimic to the 1386Ln cells led to a reduced expression of Vim and enhanced expression of E-cad. These changes in E-cad and Vim expression were accompanied by reduced cell migration and invasion (Figures 2E and 2F). The miR-138-induced down-regulation of cell migration and invasion can be partially reversed when E-cad was knocked-down by siRNA.</p><p>To fully explore the roles of miR-138 in EMT, it is important to identify the functionally relevant targets (e.g. mRNA). We utilized an EMT-specific qRT-PCR array to determine the differential expression of 86 EMT-related genes on 1386Ln cells transfected with the miR-138 and negative control mimics. Of the 86 genes tested, 23 genes were altered after miR-138 treatment (>2-fold, P < 0.10). These include nine down-regulated genes and 14 up-regulated genes (Supplementary Table S1 at http://www.BiochemJ.org/bj/440/bj4400023add.htm). To identify the potential targets of miR-138, a bioinformatics-based analysis was carried out based on a combination of six different sequence-based microRNA target prediction algorithms. Our analysis revealed that the set of down-regulated genes was significantly enriched with the predicted targets of miR-138 (Supplementary Table S2 at http://www.BiochemJ.org/bj/440/bj4400023add.htm), whereas no significant enrichment of the predicted miR-138 targets was observed in the set of up-regulated transcripts. Of these nine down-regulated genes, four are potential direct targets of miR-138, including EZH2, ZEB2, RHOC and VIM (Table 1).</p><!><p>Based on the bioinformatics analysis, three miR-138-targeting sequences were identified in the VIM mRNA (Figure 3A). The first and second targeting sequences (E1 and E2) are located in the coding region and the third sequence (E3) is located in the 3′-UTR of the VIM mRNA. To confirm that miR-138 directly targets these sequences, dual-luciferase reporter assays were performed using the construct in which these targeting sites were cloned into the 3′-UTR of the reporter gene (pGL-VimE1, pGL-VimE2 and pGLVimE3). As illustrated in Figure 3(B), when cells were transfected with the miR-138 mimic, the luciferase activities were reduced significantly for both reporter constructs as compared with the cells transfected with negative control. When the seed regions of the targeting sites were mutated (pGL-VimE1m, pGL-VimE2m and pGL-VimE3m), the miR-138 effects on the luciferase activity were abolished. Furthermore, as shown by qRT-PCR (Figure 3C), ectopic transfection of miR-138 in 1386Ln cells led to significant down-regulation of VIM expression, and knock-down of miR-138 in 686Tu cells enhanced VIM expression. Similar results were observed in additional cell lines (Supplementary Figure S1). The miR-138-induced change in Vim expression was also confirmed by immunofluorescence analysis (Figures 3D–3K). When 1386Ln cells were treated with miR-138 mimic, a marked reduction in Vim staining was observed (Figures 3D and 3E). The reduction in Vim was accompanied by an increase in E-cad staining (Figures 3H and 3I). In contrast, when 686Tu cells were treated with anti-miR-138 LNA, an apparent increase in Vim staining was observed (Figures 3F and 3G). This increase in Vim expression was accompanied by a decrease in E-cad staining (Figures 3J and 3K).</p><!><p>Three miR-138-binding sites were predicted in the ZEB2 mRNA, two located in the coding region and one located in the 3′-UTR (Figure 4A). As shown in Figure 4(B), dual-luciferase reporter assays were performed using constructs in which these targeting sites were cloned into the 3′-UTR of the reporter gene (pGLZEBE1, pGL-ZEBE2 and pGL-ZEBE3). A dramatic reduction in luciferase activity was observed for pGL-ZEBE3, when the cells were transfected with miR-138 as compared with those of the cells transfected with negative control. Relatively small, but statistically significant, reductions in luciferase activity were observed for pGL-ZEBE1 and pG–ZEBE2 in cells transfected with miR-138 as compared with those transfected with negative control. When the seed regions of these targeting sites were mutated (pGL-ZEBE1m, pGL-ZEBE2m and pGL-ZEBE3m), the miR-138 effect on luciferase was abolished. Furthermore, as shown by Western blot analysis (Figure 4C), ectopic transfection of miR-138 or siRNA against ZEB2 in 1386Ln cells led to significant down-regulation of ZEB2 expression, and knock-down of miR-138 in 686Tu cells enhanced the ZEB2 expression. The changes in ZEB2 expression were inversely correlated with changes in E-cad expression. The miR-138-induced change in ZEB2 expression was also confirmed by qRT-PCR (Figure 4D). Similar results were observed in additional cell lines (Supplementary Figure S1). As shown in Figures 4(E) and 4(F), when 1386Ln cells were treated with the miR-138 mimic, a marked reduction in ZEB2 nuclear staining was observed. The reduction in ZEB2 staining was accompanied by an increase in E-cad staining (Figures 4G and 4H). In contrast, when 686Tu cells were treated with anti-miR-138 LNA, an apparent increase in ZEB2 staining was observed (Figures 4I and 4J). This increase in ZEB2 expression was accompanied by a decrease in E-cad staining (Figures 4K and 4L).</p><p>It is noteworthy that we also observed a significant reduction in Snai2 expression at both the mRNA level (Table 1) and the protein level (Figure 4C) in 1386Ln cells treated with the miR-138 mimic. Reduced Snai2 expression was also observed in UM1 cells that were treated with the miR-138 mimic (Supplementary Figure S1). Although no apparent expression of Snai2 was observed in 686Tu cells (Figure 4C), enhanced Snai2 expression was observed in SCC9 and SCC15 cells that were treated with anti-miR-138 LNA (Supplementary Figure S1). However, we failed to identify any miR-138-targeting site in the SNAI2 mRNA.</p><!><p>PcG (polycomb group) proteins such as EZH2 are important epigenetic regulators. The EZH2-mediated repression of E-cad expression is associated with EMT in several cancer types [25–27]. Our bioinformatics analysis identified two conserved miR-138 targeting sequences located in the coding region (Ec1) and in the 3′ -UTR (Ec2) of the EZH2 mRNA (Figure 5A). In addition, a poorly conserved targeting site (Ep) that overlaps with the junction of the coding region and 3′ -UTR has been reported previously in chickens [6]. However, in humans, two bases in the seed region (positions 4 and 6) of this Ep site are replaced with non-consensus bases. As shown in Figure 5(B), the luciferase activities of the reporter constructs containing the conserved sites (pGL-Ec1 and pGL-Ec2) were reduced significantly when cells were transfected with miR-138 mimic in comparison with those transfected with negative control. No apparent change was observed in the luciferase activities of the reporter constructs containing the mutated sites (pGL-Ec1m and pGL-Ec2m) when cells were treated with either the miR-138 or control mimic. For the reporter construct containing the chicken Ep site (pGL-Ep-gga), the luciferase activity was reduced significantly when the cells were transfected with the miR-138 mimic as compared with the cells transfected with negative control. No apparent change was observed in luciferase activity of the reporter construct containing the human Ep site (pGL-Ep-hsa) when cells were treated with either the miR-138 or control mimic. Furthermore, as shown by Western blot analysis (Figure 5C), ectopic transfection of miR-138 or siRNA specific against EZH2 in 1386Ln cells led to significant down-regulation of EZH2 expression. Knock-down of miR-138 in 686Tu cells enhanced the EZH2 expression. The changes in EZH2 expression were inversely correlated with changes in E-cad expression. Similar results were observed in additional cell lines (see Supplementary Figure S1). We also examined the expression of Eed and Suz12, two additional core components of the PRC2 (polycomb repressive complex 2). Although no change in Eed expression was observed, a slight decrease and an apparent increase in Suz12 levels were observed in 1386Ln cells treated with EZH2 siRNA and in 686Tu cells treated with anti-miR-138 LNA. The miR-138-induced change in EZH2 expression was also confirmed by qRT-PCR (Figure 5D) and fluorescence immunocytochemical analysis (Figures 5E–5L). As shown in Figures 5(E) and 5(F), when 1386Ln cells were treated with the miR-138 mimic, a marked reduction in EZH2 nuclear staining was observed. The reduction in EZH2 staining was accompanied by an increase in E-cad staining (Figures 5G and 5H). In contrast, when 686Tu cells were treated with anti-miR-138 LNA, an apparent increase in EZH2 staining was observed (Figures 5I and 5J). This increase in EZH2 expression was accompanied by a decrease in E-cad staining (Figures 5K and 5L).</p><!><p>Cancer cells can de-differentiate through activation of specific biological pathways associated with EMT, thereby gaining the ability to migrate and invade. Previous studies have suggested that deregulation of microRNAs (including miR-200 family members) is associated with EMT and the progression of a number of different cancers [28–34]. However, the involvement of microRNA in EMT of the HNSCC cells has not been investigated fully. The down-regulation of miR-138 is a frequent event in HNSCC and has been consistently observed by multiple laboratories [10–15]. Our preliminary results show that miR-138 regulates cell migration and invasion by targeting RhoC and ROCK2 (Rho-associated, coiled-coil-interacting protein kinase 2) concurrently [12]. These miR-138-induced changes in cell migration and invasion are accompanied by marked morphological changes (e.g. loss of polarity and cell–cell adhesion, increased motility, and the acquisition of mesenchymal phenotype) which are characteristics of EMT. In the present study, we test the effects of miR-138 on the expression of 86 EMT-related genes using a qRT-PCR array. We found nine down-regulated genes and 14 up-regulated genes. Our bioinformatics analysis revealed that four of these nine down-regulated genes are potential direct targets of miR-138, including EZH2, ZEB2, RHOC and VIM. It is noteworthy that our qRT-PCR-based experiments measure the differential expression at the mRNA level, and are only sensitive to the targets that are regulated by microRNA-mediated degradation, but not to the targets that are regulated by microRNA-mediated translational inhibition. We anticipate that a portion of true miR-138 targets will not be detected by our approach. Nevertheless, the present study identified a panel of EMT-related genes that are regulated by miR-138, including the experimentally confirmed miR-138 target gene RHOC [12]. In the present paper we demonstrated that miR-138 regulates EMT via three novel and distinct pathways (Figure 6): (i) direct targeting of VIM mRNA and controlling the expression of VIM at the post-transcriptional level, (ii) regulating the transcriptional repressors (ZEB2 and Snai2) which in turn regulate the transcription activity of the E-cad gene, (iii) targeting the epigenetic regulator EZH2 which in turn modulates its gene-silencing effects on the downstream genes including E-cad.</p><p>Vim is an essential structural cytoskeletal protein constituting IFs (intermediate filaments) of mesenchymal cells. During EMT, the cellular IF status changes from a keratin-rich network (with connections to adherens junctions and hemidesmosomes) to a Vim-rich network connecting to focal adhesions. The results of the present study demonstrated that Vim is a functional target of miR-138 and that the down-regulation of miR-138 in HNSCC is associated with enhanced Vim expression. It is noteworthy that miR-138 has also been thought to play important roles in developmental processes [3–6]. Given the apparent involvement of Vim in EMT, it is possible that the miR-138-mediated regulation of Vim may have functional relevance to developmental processes. Further studies are required to define the precise mechanisms through whichmiR-138 andVim regulate these diverse biological processes.</p><p>ZEB2 is a member of the zinc finger E-box-binding homeobox family of proteins that function as transcriptional repressors and interact with activated SMADs, the transducers of TGFβ signalling. ZEB2 is a well-established inducer of EMT and a potent repressor of E-cad expression [35]. The results of the present paper demonstrate a direct interaction between miR-138 and ZEB2 mRNA which leads to the post-transcriptional suppression of ZEB2 expression, which in turn regulates the expression of E-cad and EMT. It is noteworthy that the ZEB family repressors (both ZEB1 and ZEB2) are targeted by a number of different microRNAs, including the miR-200 family [36]. Interestingly, ZEB1 has been showed to suppress the expression of miR-200 family members indicating that miR-200 members and ZEB factors reciprocally control each other in a negative-feedback loop [32,37]. It is not clear whether miR-138 is also regulated by ZEB factors (or its other target genes) by similar feedback mechanism(s). Further studies are needed to explore this potential regulatory mechanism. In addition to ZEB2, our data suggested that miR-138 also down-regulates Snai2, a member of Snail family of zinc finger transcriptional repressors that play an important role in the regulation of E-cad expression and EMT [35]. However, we failed to identify any miR-138-targeting site in the SNAI2 mRNA. It is possible that miR-138 indirectly regulates SNAI2 by targeting factor(s) that control SNAI2 gene expression. Alternatively, Snai2 may be regulated by miR-138 through a non-canonical targeting site.</p><p>A third EMT-related gene that targeted by miR-138 is EZH2. EZH2 is a critical component of the PRC2 complex that includes non-catalytic subunits Suz12 and Eed. It catalyses trimethylation on Lys27 of histone 3 protein (H3K27Me3), which in turn leads to chromatin condensation and epigenetic silencing of the downstream genes [38]. One of the well-established downstream target genes of EZH2 is E-cad, and the EZH2-mediated repression of E-cad is associated with EMT in several cancer types [25–27]. A previous study reported that overexpression of EZH2 in cancer cells down-regulates the expression of E-cad through histone H3K27 trimethylation at the promoter of the gene [25]. Knockdown of EZH2 in vitro has been shown to restore E-cad expression [39,40]. We confirmed that EZH2 is required for the suppression of E-cad in our cell types. We further demonstrated that the miR-138 down-regulates the expression of the EZH2 gene by binding to a conserved targeting site located in the 3′ -UTR of the EZH2 mRNA. This miR-138-mediated EZH2 down-regulation is reversely correlated with E-cad expression and EMT in HNSCC cells. Interestingly, miR-138 has been showed to target EZH2 in chickens [6]. Two miR-138-targeting sites were identified in the chicken EZH2 mRNA: an evolutionarily conserved site located in the 3′ -UTR which is also present in the 3′ -UTR of the human EZH2 mRNA, and a poorly conserved site that is over-lapped with the translational stop codon in the chicken EZH2. In humans, the seed region of this poorly conserved site has two base substitutions (at positions 4 and 6), which makes it non-functional. Nevertheless, the fact that EZH2 expression is down-regulated by miR-138 in both chickens and humans suggested that the miR-138-mediated suppression of EZH2 is an evolutionarily conserved molecular event. We also observed an apparent increase in Suz12 levels in 686Tu cells treated with anti-miR-138 LNA. However, no miR-138-targeting site was identified in the SUZ12 mRNA, and currently, the biological significance of these observed changes in SUZ12 expression is not clear. It is noteworthy that miR-200 has been shown to directly target Suz12 and control the E-cad expression by regulating the PRC2 complex [41]. Additional studies will be needed to fully explore the potential concordant effect(s) of anti-EMT microRNAs (e.g. miR-138 and miR-200) on PRC2-mediated repression of E-cad.</p><p>In summary, the results of the present study demonstrated that miR-138 regulates EMT in HNSCC cells. This, together with our previous observation that miR-138 regulates cell migration and invasion by concurrently targeting RhoC and ROCK2 [12], suggested that miR-138 is a multi-functional molecular regulator and plays major roles in EMT. Further studies are required to explore its potential as a novel therapeutic target for cancer patients at risk of metastasis.</p>

PubMed Author Manuscript

Golgi Associated HIF1a Serves as a Reserve in Melanoma Cells

Hypoxia-inducible factor-1alpha (HIF1a) is a key transcriptional regulator that enables cellular metabolic adaptation to low levels of oxygen. Multiple mechanisms, including lysosomal degradation, control the levels of HIF1a protein. Here we show that HIF1a protein degradation is resistant to lysosomal inhibition and that HIF1a is associated with the Golgi compartment in melanoma cells. Although pharmacological inhibitors of prolyl hydroxylation, neddylation and the proteasome inhibited degradation of HIF1a, attenuation of lysosomal activity with chloroquine did not alter the levels of HIF1a or its association with Golgi. Pharmacological disruption of Golgi resulted in nuclear accumulation of HIF1a. However, blockade of ER-Golgi protein transport in hypoxia reduced the transcript levels of HIF1a target genes. These findings suggest a possible role for the oxygen-dependent protein folding process from the ER-Golgi compartment in fine-tuning HIF1a transcriptional output.

golgi_associated_hif1a_serves_as_a_reserve_in_melanoma_cells

<!>CHLOROQUINE INHIBITS HIF1a DEGRADATION IN EPITHELIAL TUMOR CELLS BUT NOT IN MELANOMA<!>HIF1a ASSOCIATES WITH GOLGI IN MELANOMA<!>GOLGI-ASSOCIATED HIF1a SERVES AS A RESERVE<!>DISCUSSION<!>CELL CULTURE<!>IMMUNOFLUORESCENCE<!>SDS-PAGE, WESTERN BLOTTING, AND ANTIBODIES<!>REAGENTS<!>QUANTITATIVE REAL-TIME PCR

<p>Low levels of oxygen dictate the metabolic adaptation of cells inmetazoans. The transcription factor HIF1a is central to this metabolic flexibility, and its protein stability is tightly coupled to the levels of oxygen. Physiological oxygen levels trigger prolyl hydroxylases that lead to hydroxylation of specific prolines (Pro402 and Pro564) of HIF1a [Schofield and Ratcliffe, 2004; Semenza, 2009]. This modification enables the recruitment of von Hippel-Lindau (VHL) complex followed by ubiquitination and rapid proteosomal degradation of the HIF1a protein [Ivan et al., 2001; Jaakkola et al., 2001]. Conversely, hypoxic conditions limit the prolyl hydroxylase activity, leading to HIF1a protein accumulation and translocation into the nucleus. This nuclear HIF1a activates transcription of genes such as PDK1, LDHA, BNIP3 and BNIP3L that in turn switch the cells from generating ATP from oxidative phosphorylation (OxPhos) to glycolysis [Semenza, 2013].</p><p>Rapid turnover of HIF1a protein is controlled by multiple mechanisms. Prolyl hydroxylation primes HIF1a protein for recognition by the VHL complex [Soucy et al., 2009]. VHL is a part of VCB-Cul2 protein degradation scaffold, which requires neddylation for its activation [Stebbins et al., 1999]. Pharmacological inhibition of prolyl hydroxylases or blockade of neddylation stabilizes HIF1a protein, suggesting that a complex coordinated cascade of post-translational modifications regulates HIF1a stability.</p><p>Autophagy is an evolutionarily conserved process by which proteins are directed to the lysosome for degradation. Based on the mechanism of substrate selection, autophagy is defined as either microautophagy, macroautophagy, or chaperon-mediated autophagy (CMA) [Orvedahl and Levine, 2009]. Typically, macroautophagy targets organelles, microautophagy works via direct capture of substrates into the lysosome, while CMA selectively eliminates soluble proteins. In each case, the substrates are delivered to the lysosome for degradation [Glick et al., 2010]. Macroautophagy is known to promote tumor cell growth in nutrient limited environments, and inhibition of macroautophagy with chloroquine as a possible anti-cancer strategy has been reported [Amaravadi et al., 2011; Yang et al., 2011]. However, metabolic plasticity of tumor cells poses challenges to targeting the lysosome. It was reported that HIF1a binds to CMA mediators HSC70 and LAMP2A for transport and consequent destabilization in the lysosome [Hubbi et al., 2013]. Furthermore, pharmacological approaches that increase or decrease the activity of lysosomal proteases modulated the degradation of HIF1a protein in a manner congruent with the lysosomal activity [Hubbi et al., 2013].</p><p>We previously reported that chloroquine promotes apoptosis in melanoma cells but not in epithelial tumor cells [Lakhter et al., 2013]. This led us to examine the effects of chloroquine on HIF1a protein stability in melanoma. Consistent with the previous report [Hubbi et al., 2013], chloroquine inhibits lysosomal destabilization of HIF1a in epithelial tumor cells; however, this was not the case in melanoma cells. We found that in contrast to epithelial tumor cells, HIF1a protein is associated with the Golgi compartment in melanoma and disruption of this association reduces its transcriptional activity in hypoxia.</p><!><p>To test whether chloroquine impacts HIF1a protein stability in melanoma similar to previously reported effects on non-melanoma tumor cells [Hubbi et al., 2013], MEL526, RPMI8322 and MEL2664 melanoma cells and PC3 prostate cancer cells and HT1080 fibrosarcoma line were exposed to chloroquine or deferoxamine, and the levels of HIF1a were assessed. Regardless of the cell type, deferoxamine, a commonly used hypoxia mimetic, stabilized the HIF1a protein. In agreement with a previous report [Hubbi et al., 2013], chloroquine exposure of PC3 and HT1080 cells showed accumulation of HIF1a and the lysosomal substrate LC3II (Fig. 1A, B). Although chloroquine inhibited lysosomal degradation of LC3II in melanoma cells, it had minimal to no effect on HIF1a protein levels (Fig. 1C–E). These results prompted us to test whether known modulators of HIF1a protein stability differ in melanoma. To test this, MEL526 or PC3 cells were exposed to prolyl hydroxylase inhibitor DMOG, deferoxamine, the neddylation inhibitor MLN4924, or the proteosomal inhibitor MG132, and HIF1a protein levels were determined. Inhibition of prolyl hydroxylation, neddylation, or ubiquitination stabilized HIF1a in non-melanoma and melanoma cells alike, however, chloroquine inhibited HIF1a degradation only in non-melanoma cells (Fig. 1F, G).</p><!><p>The lack of HIF1a stabilization by lysosomal inhibition led us to a hypothesis that HIF1a localization in melanoma may differ from epithelial tumor cells. To test this, MEL526 and PC3 cells were treated with either vehicle or chloroquine, and localization of HIF1a was determined by immunofluorescence. HIF1a immunostaining from vehicle treated cells revealed a diffuse cytoplasmic localization in PC3 cells and a discrete pattern of localization in MEL526, reminiscent of the Golgi apparatus (Fig. 2A, B). Co-staining with the antibodies against the Golgi marker GM130 showed co-localization with HIF1a in melanoma (Fig. 2B). Upon exposure to chloroquine, PC3 cells showed increased HIF1a immunostaining in the nucleus whereas the Golgi association observed in melanoma cells was insensitive to chloroquine exposure. However, inhibition of prolyl hydroxylases with DMOG led to nuclear accumulation of HIF1a in both melanoma and non-melanoma cell types.</p><!><p>Brefeldin A (BFA) is an antibiotic that blocks protein transport from the endoplasmic reticulum (ER) to Golgi. To test whether inhibition of ER-Golgi transport alters HIF1a localization, MEL526 cells were treated with vehicle or BFA and HIF1a localization was determined. As evident from BFA-mediated loss of discrete GM130 localization suggesting the disruption of Golgi structure, BFA treatment induced HIF1a protein accumulation in the nucleus in MEL526 cells (Fig. 3A). We hypothesized that Golgi may play an important role in transcriptional activity of HIF1a in melanoma cells. MEL526 and PC3 cells were treated with vehicle or BFA in 1% oxygen and HIF1a target gene expression was determined. Although CA9 expression was induced in hypoxia, BFA did not change the transcript levels of HIF1a targets CA9, LDHA, LDHB and PDK4 even after six hours of hypoxia in PC3 cells (Fig. 3B). However, induction of these transcripts was significantly reduced in the presence of BFA under hypoxic conditions in MEL526 cells. These results suggest that an optimal transcriptional activity of HIF1a requires ER-Golgi protein transport.</p><!><p>The present study demonstrates that unlike in epithelial tumor cells, HIF1a protein is associated with the Golgi compartment in melanoma. Pharmacological disruption of ER-Golgi transport led to the nuclear localization of HIF1a. This suggests that melanoma cells harbor a novel mechanism to target a portion of HIF1a protein to the nucleus via the Golgi compartment. We found that disruption of ER-Golgi transport reduced HIF1a transcriptional activity in melanoma cells. Collectively, these results suggest that Golgi plays an important role in regulation of HIF1a activity in melanoma.</p><p>Nearly one third of newly synthesized proteins pass through the Golgi apparatus before reaching their destination. Therefore, the Golgi apparatus has been viewed as a protein sorting station [Guo et al., 2014]. In addition to its role in sorting proteins to various destinations, Golgi has been recognized for its role in storage, ensuring proper folding of proteins as well as post-translational modifications [Baumann, 2014]. Therefore, questions arise as to why melanoma cells direct HIF1a protein to Golgi and whether a specific HIF1a modification takes place at the Golgi or if it merely acts as immediate source of HIF1a protein supply in order to rapidly respond to a metabolic change. However, it is clear that melanoma cells exposed to BFA promote HIF1a localization to the nucleus suggesting that the Golgi-associated HIF1a can be mobilized to the nucleus similar to the effects of prolyl hydroxylase inhibition. These results invoke the possible existence of several distinct pools of HIF1a in cells that undergo different modes of degradation. The majority of HIF proteins undergo conventional prolyl hydroxylation and subsequent proteosomal degradation, while a pool of HIF1a protein in melanoma cells is directed to the ER-Golgi pathway. Disruption of HIF1a localization to Golgi in DMOG-treated cells suggests that prolyl hydroxylation likely precedes HIF1a accumulation at the Golgi compartment. Analysis of HIF1a transcriptional activity under hypoxic conditions reveals that intervention of ER-Golgi protein transport reduces the HIF1a transcriptional response. Although HIF1a protein lacks ER retention sequence, a substantial amount of HIF localization to the ER is mediated by its interaction with VHL [Schoenfeld et al., 2001]. A 3D two-photon confocal laser microscopy coupled co-localization study revealed that HIF protein localized to the ER was regulated by the local Fenton reaction at the ER [Liu et al., 2004]. A commonly used Golgi marker GM130 is a peripheral cytoplasmic protein that binds tightly to the Golgi membrane as a part of larger oligomeric complex [Nakamura et al., 1995]. Based on the observed co-localization of HIF with GM130, it is likely that HIF bound to the cytosolic side of Golgi in melanoma cells. Several proteins are known to move from the ER-Golgi compartment to the nucleus. For example, SOK1 translocates from the cytoplasmic side of Golgi to the nucleus and induces cell death in response to chemical anoxia [Nogueira et al., 2008]. Studies from yeast and mammalian cells suggest that proteins such as ATF6, PKA, SREBP2 and PIK1 are mobilized from the cytoplasmic side of Golgi to the nucleus [Constantinescu et al., 1999; Horton et al., 2002; Demmel et al., 2008; Guan et al., 2011]. Consistent with these studies, our observations suggest that melanoma presents a case where Golgi plays a critical role in regulation of HIF1a transcriptional activity.</p><p>An important consideration in hypoxia is the limited availability of oxygen, which is crucial for several biochemical processes in the cell. For instance, protein folding requires disulfide bond formation. In the presence of molecular oxygen, protein disulfide isomerase cooperates with endoplasmic reticulum oxidoreductase-1 (Ero1) to ensure optimal protein folding [Tu et al., 2000]. Although it is not known whether Ero1-La directly acts on HIF1a, it has been reported that Ero1-La cooperates with HIF1a in hypoxic adaptation [May et al., 2005]. It was recently reported that disufide bond formation in the ER requires molecular oxygen [Koritzinsky et al., 2013]. Importantly, there are 11 Cysteine residues in the dimerization domain of HIF1a that may require oxygen for disulfide bond formation to enable optional folding of the PAS domains. Given that suppression of the ER-Golgi transport resulted in reduced HIF1a transcriptional output in melanoma cells, it is plausible that differential folding, determined by the levels of oxygen in the ER, may dictate the dimerization capacity and promoter selection by the HIF1a transcription complex. It is conceivable that the structural variations in folding of the dimerization domain may fine tune transcriptional functions of HIF1a. Failure in protein folding results in unfolded protein response and if it exceeds a certain threshold, cells undergo programmed cell death [Sano and Reed, 2013]. Therefore, it is tempting to speculate that melanoma cells utilize HIF1a as a molecular sensor to monitor not only oxygen levels, but also the ER protein folding capacity, in order to avoid cell death. It is likely that this is a safeguard mechanism that ensures adaptation to hypoxic conditions, while avoiding deleterious effects of hypoxia, by synchronizing protein-folding capacity to low oxygen levels during the metabolic shift to glycolysis.</p><p>In light of our previous report demonstrating that a recurrent glioma mutation found in dimerization domain of HIF1a suppresses mitochondrial respiration [Lakhter et al., 2014], and given the potential role of the ER oxygen levels in oxidative folding of dimerization domain, a better understanding of the this domain in the context of ER-Golgi localization may have implications in patho-physiology of cancer, neurodegenerative diseases and diabetes.</p><!><p>Cell lines were maintained at 37 °C at 5% CO2 in RPMI 1640 or DMEM culture media (Sigma) supplemented with 10% fetal bovine serum (Sigma), 50 Uml−1 penicillin and 50 µgml−1 streptomycin (Life Technologies). MEL526, PC3 cell lines in RPMI 1640, and HT108, RPMI 8322, MEL2664 cell lines in DMEM maintained. Hypoxic conditions (1% O2) were achieved in a Ruskinn InvivO2 400 hypoxia chamber, by supplementing ambient air with balanced N2 and CO2.</p><!><p>Cells were grown on glass chamber slides (Millipore) and treated with indicated compounds. After treatment, cells were fixed with 4% paraformaldehyde for 10 min at room temperature, permeabilized with 0.05% Triton X-100 for 5 min, and blocked with 5% protease-free BSA for 1 h. Samples were incubated with primary antibody in 2% protease-free BSA for 1 h, and incubated with CF dye-conjugated secondary antibody (Biotium) for 1 h. Imaging was captured on Eclipse 80i fluorescent microscope (Nikon) with Retiga Exi (Qimaging) camera.</p><!><p>Whole-cell extracts were prepared in urea buffer (6 M urea, 100 mM sodium dihydrophosphate, 10 mM Tris pH 8). SDS-PAGE was performed using TGX gradient gels (Bio-Rad) and transferred onto PVDF membranes (Millipore) using TransBlot SD semi-dry transfer apparatus (Bio-Rad). The blots were probed with following antibodies: HIF1a (R&D Systems), LC3 (Novus), and tubulin (Sigma). Blot images were captured on ImageQuant LAS 4000 digital imaging system (GE Healthcare).</p><!><p>MLN4924 was purchased from BostonBiochem, chloroquine and DMOG from Sigma, all other chemicals were purchased from Cayman Chemical. HIF1a antibodies were purchased from R&D Systems, and GM130 from Cell Signaling Technology, LC3 from Novus, and tubulin from Sigma.</p><!><p>RNA was extracted using NucleoSpin II RNA extraction kit (Clontech) and followed by reverse transcription with RNA to cDNA EcoDry Premix (Clontech). Quantitative PCR was done on CFX 96 Real-Time PCR Detection System instrument (Bio-Rad) using SsoFast EvaGreen Supermix (Bio-Rad), and relative expression was calculated using ddCt method with target transcript normalized to that of RNPII and 18SRNA. Following oligonucleotide primer sequences were used: 18S RNA ACCCGTTGAACCCCATTCGTGA (forward) and GCCTCACTAAACCATCCAATCGG (reverse), CA9 TCCCTGCCGAGATCCACGTG (forward) and TTTCTTCCGGGCCCTCCTCC(reverse), LDHATTCTAAGGAAAAGGCTGCCA (forward) and ATGGCCTGTGCCATCAGTAT (reverse), LDHB TCCATGTATCCTCAATGCCCGG (forward) and TCTGCACTTTTCTTGAGCTGAGC (reverse), PDK4 TCTACTCGGATGCTGATGAACCA (forward) and ACCACTGCTACCACATCACAGT (reverse).</p>

PubMed Author Manuscript

Layered double hydroxide of Cd-Al/C for the Mineralization and De-coloration of Dyes in Solar and Visible Light Exposure

Cd-Al/C layered double hydroxide (Cd-Al/C-LDH) and Cd-Sb/C nanocatalyst are reported here for the decoloration and mineralization of organic dyes. These catalysts were largely characterized by FESEM, EDS, XRD, FTIR, XPS, PL and DRS. The diffuse reflectance data showed a band gap at 2.92 and 2.983 eV for Cd-Al/C-LDH and Cd-Sb/C respectively. The band gap suggested that both catalysts work well in visible range. The photoluminescence spectra indicated a peak at 623 nm for both the catalysts which further support the effectiveness of the respective catalyst in visible range. Both catalysts also showed good recyclability and durability till 4 th cycle. Five dyes, acridine orange (AO), malachite green (MG), crystal violet (CV), congo red (CR) and methyl orange (MO) were used in this experiment. Various parameters of different light intensity such as visible, ultraviolet, sunlight and dark condition are observed for the de-coloration of these dyes. The de-coloration phenomenon was proceeded through adsorption assisted phot-degradation. The low cost, abundant nature, good recyclability and better dye removal efficiency make these catalysts suitable candidates for the de-coloration and mineralization of organic dyes.Dyes industries play an important role in the progress and development of a country and make the human life beautiful. In pre-historical time people used various dyes to make their environment gorgeous. Most of the dyes stuff are categorized on the basis of its coloring properties, solubility, and chemical nature 1 . These dyes are used to color our clothes, food materials and beverages and even make some medicine colored. Literature survey revealed that there are approximately 10,000 commercially available dyes and 7, 00,000 tons are manufactured per annum worldwide [2][3][4] . During coloring practices most of the dyes were not intact and therefore, a large percentage of these remaining dyes were dumped into the stream. Approximately 10-15% of the dyes stuff are discharge into the environment which are esthetically not favorable 2 . Recently, dyes stuff are of pronounced environmental distress because of their carcinogenicity and mutagenicity 5 . More than 90% of approximately 4000 dyes were experienced in an ETAD survey (Ecological and Toxicological Association of the Dye stuffs) indicating more than 2 × 10 3 mg/ kg LD 50 values. The highest toxicity were found in diazo and basic dyes 6 . UK make an environmental policies in Sept. 1997 according to which zero synthetic chemical substances are to be released in the marine environment and ensure that the textile industries should treat their effluent before discharging into the water resources 6 . Developed countries and European community become more rigorous to control the dyes stuff from the industrial effluents 7 . Dyes industries contributed to the development of a country, but unfortunately most of the dyes stuff are discharge into our water resources without any treatment 8 . This led to the contamination of underground water resources by passing from soil to water beds and by owing its carcinogenic nature exposing human health and other organisms to high risk. Dyes stuff in minute quantity colorized the water and make a foam like layers on the surface of water 9 . These layers halted the penetration of sunlight and oxygen into water which finally led to the death of aquatic flora and fauna 10 . Therefore, it is very important to purify these effluent before discharging into the water resources.Dyes removal from water is one of main environmental problem due to its carcinogenic and mutagenic nature to aquatic life [11][12][13] . Previous methods such as physical, chemical and biochemical are not useful for the removal of dyes from the effluent 6 . Several methods have been reported in the literature for the de-coloration techniques such as activated carbon 7,14-18 however, the advanced oxidation processes (AOPs) effectively used for

layered_double_hydroxide_of_cd-al/c_for_the_mineralization_and_de-coloration_of_dyes_in_solar_and_vi

<!>Experimental<!>Synthesis of Cd-Al/C-LDH.<!>Synthesis of<!>Results and Discussion<!>Photocatalytic activity.<!>Selectivity of Dye.<!>Dark condition exposure.<!>Conclusion

<p>the degradation and mineralization of dyes into CO 2 and some inorganic ions. During the removal of dyes the AOPs generated highly reactive species known as ROS (reactive oxygen species i.e. O •2− , • OH, or HO •2 ), which possibly invade almost all contaminants. In all known AOPs, the heterogeneous photocatalytic process is the most effective method because of its high availability, low toxicity, inexpensive and diverse nature that might attacked and mineralize a large number of contaminants 13 .</p><p>A class of new materials with hydrotalcite like structure known as layered double hydroxide recently attract researchers attention due to their varied application and high anion exchange capacity which make them suitable candidates for catalysis 19 . Hydrotalcite is comparable to brucite like structure Mg(OH) 2 , in which divalent Mg +2 is octahedraly sandwich between hydroxyl groups 6,20 . LDH has positively charge cations on the surface intercalated negatively charge anion and water molecules 21 . The general formula for LDH is [M II 1−x M III x (OH) 2 ] z+ (A n− ) z/n •y H 2 O M II and M III are di-and trivalent cation, (Zn +2 , Cu +2 , Al +3 and Fe +3 etc.) while A n− is the counter anion for cation present in the interlayer of brucite like sheets while x is the (M II /M II + M III ) ratio.</p><p>A number of authors have been reported various metal-metal combinations for the synthesis of layered double hydroxide through various techniques for various application. For instance, Gaini et al. reported Mg-Al-CO 3 −2 for the de-coloration of indigo carmine 6 . Panda et al. reported Mg/Al-LDH and study its various factor on its growth such as various metal cation concentration, pH and aging time 22 . Similarly, Shao et al. investigated ZnTi-LDH by co-precipitation methods for the de-coloration of methylene blue dyes under the influence of visible light irradiation 23 . The ZnAl-LDH was synthesized by Beak et al. through hydrothermal methods 22 . Beside metal-metal combination the LDH was supported on various support to increase its catalytic performance. For instance, carbon nanotubes was used by Li et al. as a support for NiAl-LDH through solution method 24 . Similarly, reduced graphene oxide were used as a support for CoAl-LDH for the asymmetric electrochemical capacitors 25 . An efficient MgAl-LDH grown on multi-walled carbon nano tubes MWCNT for CO 2 adsorption 26 . Composite of NiCoAl-LDH coupled with Ni-Co-carbonate hydroxide supported on graphite paper were used for the asymmetric supercapacitors 27 . In the present work we synthesized Cd-Al and Cd-Sb grafted on activated carbon in order to increase the electronic conductivity through co-precipitation method in which Cd-Al/C were grown in layered double hydroxide morphologies. The synthesized materials were employed for the removal of five dyes AO, MG, CV, CR and MO through adsorption and adsorption assisted photo-degradation.</p><!><p>Chemicals and reagents. Salt of Al and cadmium nitrate and chloride of cadmium and antimony along with other reagents and dyes mentioned in this manuscript were obtained from Sigma-Aldrich, Ireland. Millipore-Q machine was used for double distilled water, present in chemistry department, King Abdulaziz University Saudi Arabia.</p><!><p>Salt of Cd and Al nitrate were well mixed in double distilled water and then mixed with activated carbon through co-precipitation method 19,28,29 . Briefly salt of Al(NO 3 ) 3 and Cd(NO 3 ) 2 were dissolved thoroughly in double distilled water in 1:3 molar ratio. To this reaction mixture, 10 wt% of activated carbon was added and well dispersed by continuous stirring with the help of magnetic stirrer. To this mixture freshly prepared 0.1 M NaOH solution are added and continuously monitored till pH 9. After this the reaction mixture were placed on a hot plate for 6 h at 60 °C with homogenous stirring. After completion of the reaction the surplus solution is removed and the precipitate was washed thrice with C 2 H 5 OH:H 2 O mixture (8:2). The resultant product was dried in an oven for overnight at 50 °C and store in clean tube for further characterization. Instrumental analysis and Characterization. FTIR (Thermo Scientific) for functional group analysis and powder X-ray diffractometer (PXRD) with a Kα radiations (λ = 0.154 nm) source were used for the purity and crystallinity of the catalyst. Field emission-scanning electron microscope (FESEM), JEOL (JSM-7600F, Japan) for surface morphology and average size of the particles, while, energy dispersive X-rays spectrometry (EDS) of oxford-EDS system was employed for the elemental composition of the catalyst. X-ray photoelectron spectroscopy (XPS) Thermo Scientific K-Alpha KA1066 spectrometer (Germany) in the range of 0 to 1350 eV was investigated for the elemental analysis as well as for the determination of binding energy in the respective catalyst. The photocatalytic reaction was monitored through Evolution 300 UV-visible spectrophotometer (Thermo scientific). The effect of visible and ultraviolet light on adsorption-degradation process of dyes were observed under visible lamp (OSRAM, 400 watt) and ultraviolet lamp (Smiec Shanghai China, 230 V, 11 watt) respectively. The solar light effect was studied under normal day sunlight in open atmosphere and dark effect in complete absence of light. Photoluminescence emission spectra were confirmed at 320 nm excitation wavelength (fluorescence spectrofluorophotometer), RF-5301 PC, Shimadzu, Japan. The UV-vis diffuse reflectance spectroscopy was recorded by PerkinElmer UV-vis diffuse reflectance spectrophotometer.</p><!><p>Procedure for dye removal. For this study 0.025% mmol effluent solution of acridine orange (AO), methyl orange (MO), congo red (CR), malachite green (MG) and crystal violet (CV) were prepared. The catalytic activity of Cd-Al/C-LDH and Cd-Sb/C were evaluated against the respective dyes under visible, solar, ultraviolet light, and dark condition. The adjusted dose 10 mg of the respective catalyst were added in 100 mL of 0.025 mmol concentration of dye solution and the gradual decrease in concentration of all the dyes were explored through UV-vis spectrophotometer. The % removal efficiency A.E. (%) of each catalyst was evaluated by using the following equation.</p><p>C 0 represents the original concentration of each dye solution at time = 0, C t is the concentration of dye solution by adding the catalyst after some time = t as indicated in equation (1). Similarly, A 0 designated the absorbance of the original concentration of the dye solution at time = 0 and A t is the absorbance of dye solution during reaction progress after passing some time = t. Each time 3 mL of the aliquot were taken after specified time and checked the reaction progress by using UV-vis. spectrophotometer.</p><!><p>Structural characterization of catalyst. The average size and morphology was scrutinize through FESEM. The FESEM images shows the sheet morphology of Cd-Al/C-LDH (Fig. 1a,b) and Cd-Sb/C (Fig. 1c,d) and the sheets are composed of nanoparticles with average particle size of 50 and 40 nm respectively. In both catalyst small particles are aggregated to form the sheet morphology, however the particle is in photoluminescence spectra of both catalyst further support the efficiency of the respective catalyst in visible range as presented in Fig. 6.</p><!><p>The photocatalytic activity of the Cd-Al/C-LDH and Cd-Sb/C was carried out for the de-coloration of five dyes AO, CR, MO, MG and CV under solar, visible, ultraviolet light and dark condition.</p><p>Visible light exposure. Prior to the effect of solar, ultraviolet and dark condition the effect of visible light was study on the de-coloration of cationic and anionic dyes. Both Cd-Al/C-LDH and Cd-Sb/C were evaluated against three different dyes cationic acridine orange (AO) and anionic congo red (CR) and methyl orange (MO) under visible of 400 watt.</p><p>Adjustment of catalytic dose. The catalytic dose was adjusted under visible light by selecting AO dye.</p><p>Initially, the catalytic doses were adjusted with AO by starting from 100 mg of the respective catalyst in 100 mL of 0.025 mmol of AO. The Cd-Al/C-LDH de-colorize 70% AO while Cd-Sb/C 53% in 1 h. The amount of both the catalyst were decreased to 40 mg in 100 mL of 0.025 mmol of AO in which 63% of AO is removed with Cd-Al/C-LDH and approximately 45% with Cd-Sb/C in 1 h. However, the removal efficiency of dyes is further increased as we increased the contact time. For instance, after 2.5 h under the same condition the % removal of AO with Cd-Al/C-LDH and Cd-Sb/C was approximately 80 and 63% respectively as shown in Fig. 7a,b. Further the amount of the catalysts were decreased to 10 mg in 100 mL of AO. This time the removal efficiency was dropped to 43 and 36% respectively with Cd-Al/C-LDH and Cd-Sb/C in 1 h as shown in Fig. 8a,b. However, by increasing the exposure time of reaction mixture with light the rate of dye removal is also increases and vice versa. During the optimization of the catalyst to dye solution the Cd-Al/C-LDH showed superior activity than Cd-Sb/C (Fig. 9a,b).</p><p>Using small amount of the catalyst to dye solution is eco-friendly and therefore, we selected 10 mg of the respective catalyst as an optimized amount for the further detail study. 10 mg of the respective catalyst were further used for the de-colorization of AO, CR and MO. After the first 15 min the % removal efficiency of Cd-Al/C-LDH against AO, CR and MO was 27.0, 13.7 and 4.8% respectively. However, at the same condition the Cd-Sb/C showed 30.7, 3.9 and 3.7% removal efficiency respectively. During the start of the reaction Cd-Sb/C showed slightly good response then Cd-Al/C-LDH. However, onward its activity was much lower than Cd-Al/C-LDH and this might be due to the LDH nature of Cd-Al/C-LDH. By increasing the contact time the % de-coloration of all dyes is increased and it was found that after 200 min, the Cd-Al/C-LDH showed strong response then Cd-Sb/C. For instance, after 200 min the % de-coloration of AO with Cd-Al/C-LDH is 69.4% as compared to Cd-Sb/C which was only 44.0%. It was found that both catalyst showed superior response against cationic dye (AO) as compared to anionic dye (CR and MO) and this is probably due to the large structure of these mentioned anionic dyes. It was also found that for all dyes Cd-Al/C-LDH showed superior performance than Cd-Sb/C. The reaction was also monitored without catalyst with almost no change in dye concentration under visible light, which shows that visible light itself has no role in the de-coloration phenomena. The decrease in the concentration and % removal efficiency of CR and MO are illustrated in Fig. 10a,b (CR), Fig. 11a,b (MO).</p><!><p>Figure 12 showing the selective removal of AO under visible light exposure as compared to CR and MO. Therefore, AO was selected for the further detail study with both catalyst under solar, ultraviolet light and dark condition.</p><p>Sunlight exposure. Under the same condition both the catalyst were evaluated for the de-coloration of AO in normal day sunlight exposure. The adjusted dose 10 mg of both the catalyst were added in 100 mL of 0.025 mmol of AO solution. The Cd-Al/C-LDH showed faster and better response then Cd-Sb/C in sunlight exposure. After the first 15 min of experiment the Cd-Al/C-LDH showed 13.5 while Cd-Sb/C showed 18.5% de-coloration of AO. Similarly, after 180 min the % de-coloration of AO with Cd-Al/C-LDH increased from 13.5 to 82.2% and Cd-Sb/C from 18.5 to 76.0%. This inferred the better performance of both catalyst with the passage of time as shown in the inset of Fig. 13a-d.</p><p>Under visible and solar light irradiation it was concluded that cationic dye AO is selectively removed with the corresponding catalyst. It was also confirmed that cationic dye AO is predominantly removed in solar light as The catalytic activity of the respective catalysts were excellent, better and good for AO, MG and CV respectively. Cd-Al/C-LDH showed stronger catalytic activity over Cd-Sb/C catalyst and selectively de-colorized AO over MG and CV as presented in Fig. 16. Therefore, we select AO for further study under ultraviolet light and dark condition.</p><p>Ultraviolet light exposure. UV lamp (230 volt, 11 watt) was used for the de-coloration of 0.025 mmol AO solution. Keeping the same condition as adjusted for visible and solar light, 10 mg of both the catalyst were used for the de-coloration of 100 mL of 0.025 mmol AO solution. After the first 15 min of exposure time the % removal efficiency of Cd-Al/C-LDH and Cd-Sb/C against AO was 8.5 and 7.5% respectively. However, after 200 min, the removal efficiency of AO (%) was increased to 36.3% with Cd-Al/C-LDH and 29.1% with Cd-Sb/C. Like visible and solar light the Cd-Al/C-LDH showing superior removal for AO as compared to Cd-Sb/C. The decrease in concentration and % removal efficiency of AO under ultraviolet light exposure are presented in the inset of Fig. 17a,b.</p><!><p>Prior to solar and ultraviolet light the effect of dark was studied for both catalyst to know the adsorption or degradation phenomena. The adjusted dose 10 mg of the respective catalyst were put in a beaker containing 100 mL of 0.025 mmol AO solution by providing complete dark condition. After the first 15 min of the reaction progress the Cd-Al/C-LDH displayed 11.0% and Cd-Sb/C 10.9% removal efficiency. However, after 200 min, the Cd-Al/C-LDH was showing 31.0% and Cd-Sb/C 26.0% removal of AO as indicated in Fig. 18a,b.</p><p>After the detailed study for the de-coloration of dyes (AO, CR, MO, MG, CV) under solar, visible, ultraviolet light and absence of light, it was concluded that cationic dyes removed preferentially then anionic dyes. Among the cationic dyes (AO, MG and CV) AO was predominantly removed by both catalyst in solar light. However, the removal efficiency of MG and CV are also good. It was inferred that AO adsorbed in the absences of light and adsorbed plus degraded in ultraviolet, visible and solar light. The adsorption process is triggered by the presence of activated carbon in the respective catalyst. The dyes is adsorbed on the catalyst and then degraded as the reaction proceeded.</p><p>Kinetic study of the reaction. The kinetics of the reaction was determined by applying pseudo first order kinetics (lnC t /C o ). This model was applied to compare the rate of Cd-Al/C-LDH and Cd-Sb/C in visible light by using different amount of the catalyst against the removal of AO. This equation also applied to investigate the rate of reaction in different parameters like dark, ultraviolet, visible and solar light. The rate of the reaction was determined by plotting lnC t /C o should be subscript verses time. The model showed the highest rate of Cd-Al/C-LDH then Cd-Sb/C in all conditions. The rate of reaction is directly related to the amount of catalyst. For instance, at 40 mg of Cd-Al/C-LDH the rate of reaction was 5.88 × 10 −3 mol L −1 min −1 as compared to 4.92 × 10 −3 mol L −1 min −1 obtained when 10 mg of the same catalyst was used under solar light. The same trends was observed in visible light. However, the rate of degradation is slow from solar light as inspected in Fig. 19a. In all other parameters Cd-Al/C-LDH showed higher rate of reaction then Cd-Sb/C as indicated in the inset of Fig. 19b,c.</p><p>Catalytic recyclability. The catalytic recyclability is critical issue while carrying catalysis. Most of the catalyst become de-activated after first or second cycle. During the recyclability of the catalyst 100 mg of the respective catalyst was added in 100 mL (0.025 mmol) of AO under visible light of 400 watt. After one hour, the reaction was stopped and 3 mL of the aliquot was taken through a clean syringe and examined in UV-vis. spectrophotometer. After this the catalyst was separated through filtration process and the filtrate was washed thrice with acetone. The recovered washed catalyst (obtained from first run) was used in the next cycle for 100 mL of 0.025 mmol AO solution without drying or heating. The reaction mixture was again placed for 1 h under visible light exposure. Similarly, after one hour the reaction was stopped for monitoring the decrease in concentration and catalytic activity through UV-vis spectrophotometer. The catalyst is likewise separated carefully and washed thrice with Structural feature of dyes and photocatalytic activity. The structural features of dyes play a significant role in dyes light. However, it is necessary to select specific materials according to the functional groups in dyes. Dyes degradation generally undergoes through breakage of various functional group i.e. cleavage of the aromatic, C-S bond breaking occurred between an aromatic ring and the sulphur of a sulphonyl group. Similarly, other functional group such as C-C, C-N and azo bond breakage will also happened during dyes degradation 31 . Both catalyst showed better performance for cationic dyes AO, MG and CV then anionic dyes CR and MO. Among the cationic dyes AO selectively removed as compared to MG and CV. The higher adsorption assisted photo-degradation of AO was probably due to the cleavage of C-C and C-N bond of aromatic ring and their non-bulky nature. While, lesser removal of CR and MO were due to the presence of azo group and large bulky groups which interfere with the charge transfer during de-coloration process. The LDH has layered double structure with upper cationic and inner anionic layers. The schematic phenomena of adsorption assisted photo-degradation of dyes is shown in Fig. 22.</p><!><p>The inexpensive and diverse morphology make layered double hydroxide a suitable candidate in the field of catalysis. Cd-Al/C-LDH and Cd-Sb/C were synthesized through co-precipitation method. Both the catalyst showed a narrow band gap which indicated its effectiveness in visible region. The PL also showed a peak at 623 nm for both catalyst which showed its efficiency in visible range. Both catalyst grown in nanosheet morphologies. The Cd-Al/C has LDH nature as confirmed from FTIR and XRD. The respective catalyst were used for the de-coloration and mineralization of organic dyes from the effluent under solar, ultraviolet, visible light and dark condition. The Cd-Al/C-LDH shown better catalytic activity in all conditions and this is probably due to its layered double hydroxide nature. The rate of reaction was determined from Langmuir isotherm indicating high rate of Cd-Al/C-LDH as compared to Cd-Sb/C. A predominate degradation and negligible adsorption was found in solar light and an equal percentage of degradation and adsorption were found in visible light. Similarly, adsorption was observed in dark condition and ultraviolet light. The Cd-Al/C-LDH and Cd-Sb/C showed good recyclability, durability and easy separation. These results showed the high activity and the ease in separation of the catalyst which are routinely encounter in nanocatalysis.</p>

Scientific Reports - Nature

Towards a Computational Ecotoxicity Assay

Thousands of anthropogenic chemicals are released into the environment each year, posing potential hazards to human and environmental health. Toxic chemicals may cause a variety of adverse health effects, triggering immediate symptoms or delayed effects over longer periods of time. It is thus crucial to develop methods that can rapidly screen and predict the toxicity of chemicals, to limit the potential harmful impacts of chemical pollutants. Computational methods are being increasingly used in toxicity predictions. Here, the method of molecular docking is assessed for screening potential toxicity of a variety of xenobiotic compounds, including pesticides, pharmaceuticals, pollutants and toxins deriving from the chemical industry. The method predicts the binding energy of the pollutants to a set of carefully selected receptors, under the assumption that toxicity in many cases is related to interference with biochemical pathways. The strength of the applied method lies in its rapid generation of interaction maps between potential toxins and the targeted enzymes, which could quickly yield molecularlevel information and insight into potential perturbation pathways, aiding in the prioritisation 1 of chemicals for further tests. Two scoring functions are compared, Autodock Vina and the machine-learning scoring function RF-Score-VS. The results are promising, though hampered by the accuracy of the scoring functions. The strengths and weaknesses of the docking protocol are discussed, as well as future directions for improving the accuracy for the purpose of toxicity predictions.

towards_a_computational_ecotoxicity_assay

Introduction<!>Methods<!>Receptors<!>Pollutants<!>Identification of Binding Sites<!>Molecular Docking<!>Re-docking co-crystallised ligands<!>Targeted Docking and Scoring<!>Binding affinity prediction<!>Re-docking co-crystallised ligands<!>Rescoring with RF-Score-VS<!>Identification of active chemicals<!>Chemical similarity of ligands<!>Hierarchical clustering of binding affinities<!>Assessment of known interactions<!>Assessment of Vina low scores<!>Potential refinement of the protocol

<p>Environmental pollution and ecotoxic stress through habitat-destruction, further exacerbated by climate change, have led to the onset of the sixth great mass-extinction event. 1 Pollution affects all organisms in the environment including humans, through a range of mechanisms, from chronic toxicity via the respiratory function and the gastrointestinal channel to dermatological uptake. This may trigger the detoxification pathways of the body in an attempt to eliminate toxic compounds from the system. 2,3 At continuous low-dose exposure, pollutants may also incur long-termed effects. 4,5 There is, however, limited or no available toxicity data for tens of thousands of compounds to which organisms in the environment, including humans, are exposed, due to the high cost and laborious nature of traditional toxicity testing. 6,7 This underscores the need for approaches that can rapidly screen and predict the toxicity of chemicals and prevent them from being released in the environment.</p><p>In 2007, the National Research Council, Committee on Toxicity Testing and Assessment of Environmental Agents (U.S.A.), proposed a new framework, emphasising the importance of integrating in vitro-based and computational methods for evaluating toxicity. 8 In response, a series of programs have been initiated in the last decade, aimed at identifying assays relevant to toxicity and screening the biological activity of large numbers of candidate pollutants in a cost-efficient and timely manner. In particular, the Toxicity Forecasting (ToxCast) project of the Environmental Protection Agency (U.S.A.) has been instrumental in setting the stage for in vitro high-throughput assays 9,10 In a significant effort, over 900 chemicals were employed in over 300 concentrationresponse assays to produce a large matrix with estimates of the chemical potency, in terms of the AC 50 parameter. 10 The launch of Phase III further broadened the chemical and assay space of ToxCast. Although the measured AC 50 values have no direct toxicological interpretation and estimated values may have large standard errors, 11 high-throughput screening (HTS) nevertheless offers an important benchmark and a means of comparing chemical activities. Other ongoing programs for predicting the hazardous character of pollutants rely on computational measurements and properties of the chemical structures under consideration. [12][13][14][15][16][17][18][19][20] The most common approach is the Quantitative Structure Activity Relationship (QSAR) method, which is based on the chemical similarity principle. Although these algorithms present good predictions for various compounds, they give no information on the mechanism of action of pollutants, which often requires structural and molecular biology methods of study.</p><p>Several factors affect the toxicity of a given chemical, including its dose and route, frequency and duration of exposure. Many modes of molecular toxicity may be generalised as the binding between a chemical and a biomolecular target. The ability of a chemical to interact with a protein has, for a long time, been modelled within the pharmaceutical industry using the molecular docking formalism. By yielding a prediction of the binding affinity of chemicals to targets, putatively related to toxicity, docking could provide molecular-level information and insight into relevant interactions. Since the targets culpable for the underlying adverse effects remain unknown for most environmental chemicals, docking could provide information on the most likely targets. This could, for example, aid in designing in vitro HTS bioassays. A multiple-target approach which considers all possible combinations of receptors and ligands could also yield a more complete picture of potential toxicity and the potential perturbation pathways. Furthermore, the docking output could also be combined with other computational approaches, for example to derive QSAR models to predict toxicity. [21][22][23] The main result of this paper is a computational ecotoxicity assay (CETOXA), containing 65 prepared protein targets belonging to various protein families. The methods for developing this computational study have been investigated in a previous paper published by our group on an ecotoxicological case study on vitamin B deficiency in moose. 24 Here, we explore the feasibility of applying a popular classical scoring function, Autodock Vina, and one of the most promising of the newer generation of machine-learning scoring functions, the Random Forest-based scoring function RF-Score-VS, for toxicity screening. The output from docking is a matrix with docking scores, representing an estimate of the binding energies, for all possible combinations of receptors and ligands. Of the two scoring functions considered, Vina shows promising results as it on average predicts higher affinities for active chemicals than for inactive ones. Furthermore, we also show that compound similarity is a strong predictor of binding affinity and that similar chemicals tend to show highest affinity for the same binding site of a given protein. Ultimately, the results may be integrated with existing toxicity and bioassay data to compare chemical profiles, to extract binding patterns and associations that may be key steps in triggering an adverse biological effect.</p><!><p>A flow-chart for the entire method of deriving the computational ecotoxicity assay is given in Fig. 1. The different steps in the flowchart are detailed below.</p><!><p>Based on the assays present in Phases I & II of the ToxCast program, 10 65 receptors were selected for which experimental structures are available in the Protein Data Bank (PDB). A complete list of the proteins can be found in Table S1 in the Supporting Information (SI), and the list includes kinases (22), phosphatases (5), proteases (6), G-protein-coupled receptors (9), nuclear receptors (11), and cytochromes P450 (4). These proteins represent a broad range of cellular functions that are critical for the survival and proliferation of cells.</p><p>PDB structures were chosen based on the availability of high-quality crystal structures. Only deposits with a resolution better than 2.85 Å were included. The protein structure quality was assessed using MolProbity. 25 Structures with an overall MolProbity score (representing a log- weighted combination of the clashscore, rotamer, and Ramachandran evaluations) lower than the crystallographic resolution were kept. In many cases, better scores were found for the corresponding structure in the PDB-REDO databank, which contains re-refined and rebuilt PDB entries. [26][27][28] Chains were removed as necessary, to ensure that each structure corresponds to the reported biologically-relevant assembly. Ligands, cofactors, and co-crystallised lysozyme molecules were also removed as needed (see Table S1). Zinc ions were kept, given their functional role in proteins, although they are not taken into account during docking. Common structural problems were corrected using Dock Prep 29 through the Chimera software. Missing non-terminal backbone residues were modelled using Modeller 30 and a local energy minimization was performed with the Amber ff14SB force field 31 through Chimera to relieve potential atomic clashes.</p><!><p>The ligands in the ToxCast chemical library can be broadly separated into four chemical use groups: phthalates and alternative plasticizers, pesticides, pharmaceuticals (both marketed and failed), and other consumer use chemicals (such as food additives, soaps, and shampoos), all together comprising a structurally diverse chemical space. Ligands in reference 10 are denoted by PubChem codes and were downloaded using a script. About 1/3 of the compounds were available only as 2D chemical coordinates and a semi-automated procedure to make 3D coordinates was performed, using the obgen tool in OpenBabel 32 followed by manual curation of all the structures.</p><p>Compounds with unsupported atoms, including metals and, e.g., boron were removed, leaving 957 environmentally relevant compounds.</p><p>Both ligands and receptors were treated by scripts from AutodockTools 33 to prepare them for docking.</p><!><p>Blind docking was employed to explore the protein surface for prospective binding sites. 34,35 For each potential chemical-target complex, a blind docking calculation was performed using QVina-W (an extension of QVina2, see below, specifically designed for blind docking), 36 with 64 independent runs per docking. Nine results were stored per complex, producing 9 x 957 bound ligand poses per target.</p><p>Common binding sites were identified by clustering the 9 x 957 centres of mass of the bound ligands using the OPTICS algorithm, 37 a density-based clustering algorithm similar to DBSCAN, available in the Python machine learning library scikit-learn. 38 To reduce errors associated with inappropriate choice of binding site, multiple binding sites were considered. For each receptor, the centres of mass of up to four of the most populated clusters were stored. In most cases, fewer than four clusters were found by the algorithm (see SI, Table S1). An example of the procedure is The receptors, with 1-4 binding sites each, define the computational ecotoxicity assay that can be used for in silico scoring of new and existing compounds.</p><!><p>AutoDock Vina 40 (or Vina) has been a popular choice for high-throughput screening, due to its efficiency and relatively high accuracy. Here, we use a revised version, Quick-Vina2 (or QVina2), which has improved on the local search algorithm, achieving significant speed-up in computation time without compromising accuracy. 41</p><!><p>In order to evaluate the docking protocol, preliminary docking simulations were performed in which each of the co-crystallised ligands was re-docked into the active site of its cognate protein target. The coordinates for the ligands were extracted from the PDB files, randomised using obconformer in OpenBabel, and prepared for docking using AutodockTools. The search box dimension was set as in ref. 40. Re-docking was performed with QVina2, with an exhaustiveness level of 8, producing up to nine possible poses for each run. The symmetry-corrected RMSD was calculated using DockRMSD. 42</p><!><p>Using the computational ecotoxicity assay, targeted docking of the pollutants was performed on the identified binding sites for each receptor. A total of 160,776 targeted dockings and scorings were performed, using the same settings as during re-docking. The lowest score, corresponding to the strongest binding, of the 1-4 binding sites for each complex was kept, producing a cross-docking matrix of binding free energy predictions for 62,205 complexes, for all possible combinations of receptors and ligands.</p><!><p>For our purposes, the ability to accurately predict the binding pose is not as important as the ability to accurately rank chemicals based on the binding affinity. Enhancing the accuracy of scoring functions for predicting binding affinities or biological activity remains a challenge. It has been noted that most classical scoring functions suffer from limitations and are unable to accurately predict biological activities. 43 In order to move beyond the limitations of docking codes solvent effects may have to be included, [44][45][46] at a substantial additional computational cost.</p><p>Among the classical scoring functions currently available, Vina has one of the best scoring powers. 47 In recent years, however, a new class of scoring functions have emerged that use a nonparametric machine learning approach instead of imposing a predetermined functional form. Such machine-learning (ML) scoring functions have been found to show improvements over classical scoring functions, in terms of ranking compounds by binding affinity. [48][49][50][51] It should be noted that while there is at least one ML scoring function specifically designed to perform well at experimental pose prediction, 52 other ML scoring functions do not necessarily improve success rates. 48,53 For the application of binding affinity predictions, however, the Random Forest-based scoring function RF-Score developed by Ballester et al. 54 has shown promising results, outperforming classical scoring functions such as the one used by Vina. 55 In particular, RF-Score-VS was adapted to virtual screening by training also on negative data (i.e., known inactive ligands), although the improvement in performance appears to be less substantial when applied to new targets not included in the training set. 51 Having generated an ensemble of viable docking poses with Vina, the top scored poses were rescored using the standalone version of RF-Score-VS. The Vina predicted binding affinities will be reported in terms of ∆G bind and the result of RF-Score-VS rescoring in terms of pKd.</p><!><p>In order to assess the quality and reliability of the docking protocol, the co-crystallised ligands were redocked and the resulting pose compared to the experimental one. Pose generation error is commonly assessed by measuring the root mean square deviation (RMSD) of the predicted pose from the experimental binding orientation. The cumulative frequency curve of the RMSD between the native and predicted conformation is shown in Fig. 3.</p><p>With 2 Å as a commonly accepted RMSD cutoff value, the success rate (i.e., fraction of predicted poses with an RMSD ≤ 2.0 Å) for the top pose of Vina is 50.0%. For comparison, a benchmark study by Wang et al. 47 found the average success rate for the top scored pose among academic docking programs to be 47.4%, with second best performance achieved by Vina (with a success rate of 49.0%). Our results are thus within the expected accuracy of the docking method.</p><p>Considering instead the best of the nine output poses (i.e., the pose with the lowest RMSD), the success rate increases to 67.2 % (with 80% falling below 3.0 Å, see Fig. 3). In other words, Vina manages to find a pose close to the experimental binding orientation, but may fail to rank it as the best-scoring pose, consistent with previous studies. 39,47,56 It should be noted that a large pose generation error does not by default imply inaccurate binding affinity predictions. A study by Li et al. found that there is a low correlation between pose generation error and binding affinity prediction error. 50 For our purposes, the latter is the more relevant quantity.</p><!><p>The nine output poses of Vina were rescored using RF-Score-VS, giving a new top pose. This leads to a drop in success rate to 30.1% (Fig. 3). Note that it is known that machine-learning scoring functions do not necessarily outperform classical scoring functions in regards to pose prediction. 48,53 Considering instead the reliability of the scoring functions in predicting binding affinities, a comparison of different docking programs using the PDBbind benchmark dataset 57,58 found that Vina had the best Pearson correlation coefficient and Spearman ranking coefficient between the docked scores and experimental binding affinities, with values of 0.564 and 0.580, respectively. 47 In comparison, using the DUD-E database, 59 Vina had an effective Pearson correlation of 0.18 while RF-Score-VS had a correlation of 0.56 when both training and test sets contained data from all targets, and a more modest correlation of 0.2 when the training and test data were created independently. 51 The performance of the scoring functions may also be judged by their ability to predict high scores for known binders. The crystal structure of the complexes were scored (i.e., without docking) using Vina and RF-Score-VS. For a direct comparison between the two functions, the Vina score was converted to pKd units by using pKd= − log(e)/RT • ∆G (where R is the gas constant and T = 298.15 K), 60 see Fig. 4. Values of pKd < 4 would suggest weakly bound compounds, while values above 10 would indicate tightly bound compounds. 61 As the activity cutoff, the value of pKd = 6 has previously been used to distinguish between active and inactive chemicals. 51 Vina predicts the majority (75%) of ligands to have a predicted binding affinity pKd > 5.2 (alt. < -7.1 kcal/mol), with a mean binding affinity of pKd = 6.2 (alt. -8.4 kcal/mol ). A number of the cocrystallised ligands are predicted to be weakly bound and some outliers even fall below pKd 4.</p><p>Among the latter, relatively poor binding affinities were predicted for a ligand involving halogenbonding interactions (3W2S), which is not accounted for in most classical scoring functions, and three ligands involving metal-coordination (1HFC, 4G9L, and 4H3X). In comparison, none of the known binders are predicted to be weakly binding by RF-Score-VS. Yet, none are predicted to be particularly potent either, generally falling below pKd 6.6.</p><p>Evaluating the linear and monotonic relationship of the two scoring functions, the Pearson's correlation coefficient and Spearman's ranking coefficient are 0.23 and 0.26, respectively. It can be noted that chemicals with high scores from RF-Score-VS also tend to have high Vina scores while no correlation is found for more weakly binding chemicals.</p><!><p>The pollutants from the ToxCast chemical library were docked to the 1-4 binding sites of each target, identified as shown in Fig. 2, and the Vina score for each complex was stored. The top poses were further rescored with RF-Score-VS.</p><p>For toxicity screening, the goal is to minimise the number of false negatives, meaning that no harmful chemicals should be classified as inactive. This poses a particular challenge since environmental chemicals can also elicit adverse effects from weak interactions.</p><p>To be able to make predictions on potential activity based solely on the scoring function, it would be necessary to identify a target-specific threshold score which could successfully delineate active chemicals from inactive ones while minimising the number of false negatives. Ideally, the active and inactive sets should be separated to such a degree to enable predictions by simply choosing an appropriate threshold value. Some promising results were found in one toxicity screening of environmentally pertinent chemicals, where two different molecular docking software (eHiTS and FRED) were compared and found to have the capacity to identify weakly active chemicals from inactive chemicals binding to rat estrogen receptors. 62 For the dataset investigated here, however, the two sets show significant overlap. As an example, Fig. 5 shows the sets of active and inactive chemicals binding to CYP 2C9, as determined by the recorded activities in the ToxCast HTS, as a function of the Vina docking score (Fig. 5 The application of molecular docking for toxicity screening is an intriguing method, however, the results indicate the limitations of using current scoring functions to predict chemical activity.</p><p>Possible routes for improving the performance could be to investigate complexation in a more detailed fashion through, e.g., molecular dynamics methods. Although implicit-solvent models remain popular, there is quite some evidence that they do not give sufficiently accurate results for predicting binding free energies. [63][64][65][66] Therefore, free energy calculations using explicit solvent will likely be required to get accurate estimates of binding energies. 67 In some cases, quantum-chemical estimates of binding-strength for toxin-receptor models 68,69 may yield additional insights as well.</p><p>Finally, whereas the purpose of drug discovery is to increase the enrichment factor, focusing only on the best ranking compounds, toxicity screening using molecular docking should also consider weakly binding chemicals, as these may still elicit adverse biological effects. This requires additional considerations and places new requirements on scoring functions. 70 Although a critical score cannot be set, Vina on average predicts better scores for active chem-icals, illustrated as follows. The average binding affinity for active and inactive chemicals to each target which had at least 5 active chemicals was calculated and the set values subtracted, ∆G bind,act − ∆G bind,inact . The difference is plotted in Fig. 6 and can be seen to be mostly negative. These are encouraging results as Vina would appear to somewhat consistently predict stronger binding affinities for active chemicals than for inactive ones. A similar trend can be seen for RF-Score-VS, although showing a poorer performance, at best producing a difference of -0.27 kcal/mol.</p><!><p>It is generally assumed that structurally similar molecules exhibit similar biological activities. 71 This notion is, for example, used in drug discovery to search chemical libraries for matching compounds 72,73 and to generate information about common receptors. 74 Toxicological data gaps may also be filled by read-across, in which chemicals with known toxicity are used to predict the toxicity of untested chemicals, based on their chemical similarity. [75][76][77] It should be noted, however, that structurally similar compounds may in some cases show large differences in potency, due to so-called activity cliffs. 78,79 In this context, reproducibility of docking results may be evaluated by testing whether the docking program can reproduce the same binding pattern for structurally similar chemicals (see SI for details). The Tanimoto coefficient for pairs of chemicals was computed and the generated similarity matrix hierarchically clustered, see Fig. 7. A value of zero implies no shared chemical fragments and a value of one represents identity. The dendrogram in Fig. 7 is labeled by the four chemical categories used in ToxCast. Some degree of clustering may be seen within the chemical class groups, notably among pesticides and pharmaceuticals. Overall, however, the chemical library is structurally diverse with only few chemicals classified as highly similar. Binding patterns were assessed for each pair of chemicals by computing the mean absolute deviation (MAD) in binding affinity to the various targets. A low MAD value would suggest similar binding behaviour to the different targets and vice versa. These were correlated to the Tanimoto coefficient by colour-encoding the clustered similarity matrix with the MAD values (Fig. 8). A qualitative comparison of the two matrices reveals that clusters along the diagonal in Fig.</p><p>MAD 7, corresponding to blocks of chemicals with shared fragments, also interact with similar binding affinities with the different protein targets, as indicated by their lower MAD values in Fig. 8. Low MAD values are also observed for pairs of highly dissimilar chemicals, notably for a group of consumer use chemicals and pesticides. These would appear to be chemicals which bind weakly to all targets, typically with an average Vina score > -6.0 kcal/mol.</p><p>A more rigorous approach was employed for a quantitative comparison, by looking at the spread in binding affinities among similar compounds. For each query compound in the pollutant library, a subset of N sim similar chemicals was stored, selected based on the corresponding Tan-imoto threshold (see SI). If N sim > 3, the standard deviation in binding affinity for this subset, std(∆G bind ), was calculated (Fig. 9). For comparison, a subset of least similar chemicals as well as randomly chosen chemicals was also stored. As reference, the target-based standard deviation of all chemicals and the overall standard deviation among all docked complexes (vertical line in Fig. 9) are shown. It can be seen that compound similarity is generally a strong predictor of binding affinity, with Vina predicting more tightly packed scores for the most similar compounds.</p><p>Conversely, the chemicals least similar to the query molecule also have a wider spread in affinities compared to the query. It should be noted that the latter subset may still contain chemicals that are similar to one another.</p><p>Having compared patterns in binding affinity, it is also of interest to investigate whether similar chemicals show highest affinity for the same binding site of a given protein. At the onset, each protein was assigned up to 4 docking sites. Using the same subsets as above, for each protein with at least 2 docking sites and each query molecule binding with the lowest Vina score to site A, the fraction pA = N sim,A /N sim of similar chemicals also binding best to A was calculated.</p><p>The value was normalized by the number of hits on each binding site, i.e., pAtot = N tot,A /N tot , to remove any potential bias for a more promiscuous binding site. The histogram of pA/pAtot is shown in Fig. 10, where a value greater than 1 indicates that the compounds are more likely to bind to the same site. It can be seen that Vina predicts structurally similar chemicals to preferentially bind to the same site compared to randomly chosen chemicals, with a distribution skewed towards larger pA/pAtot values, whereas the least similar ones, which may be significantly different in terms of size and functional groups present, have a distribution distinctly skewed towards zero, i.e., dissimilar chemicals prefer to bind to an alternative site. The fact that binding is in general largely correlated with structural similarity motivates the use of chemical similarity in predicting toxicity, as mentioned earlier. Based on available toxicity data, this would allow us to identify protein-ligand associations and predict a set of chemical fragments likely to contribute to the toxicity of a given molecule, which could be highly relevant in drug</p><p>design. An example of the implementation is shown in SI.</p><!><p>Hierarchical clustering was performed using the Seaborn package in Python, with Euclidean distance as the similarity metric and Ward variance minimization algorithm as the linkage method. Some prominent protein family clusters can be seen, particularly homogeneous clusters of NR, Kinase, and GPCR. Among the various protein families, CYP and GPCR are predicted to be the most affected, with over half the chemicals binding with an affinity < -8 kcal/mol to CYP 1A1, CYP 1A2, and the muscarinic acetylcholine receptors. Comparing chemicals in the different chemical use categories, consumer use chemicals and phthalates and alternative plasticisers tend to have low predicted binding affinities while pesticides and pharmaceuticals show a broader distribution of Vina scores. This can be seen more clearly in Fig. 12, which shows the binding free energy, averaged over all receptors, for pollutants within each chemical use group. A wide band of pesticides, pharmaceuticals, and consumer use products in Fig. 11, corresponding to 10% of all chemicals considered, have low binding affinities (∆G bind > -6 kcal/mol) to all protein targets. Among the chemicals that show specificity by binding to only a few targets with higher affinities are Mirex, a persistent organic pollutant, binding to the farnesoid X receptor α (∆G bind = -8.4 kcal/mol) and PI3K α (∆G bind = -7.7 kcal/mol); Pentachlorophenol, a pesticide and disinfectant, binding to CYP 1A1 and 1A2 (∆G bind = -7.6 and -7.7 kcal/mol, respectively); Lindane, an insecticide, pediculicide, and scabicide, binding to CYP 1A2 (∆G bind = -7.5 kcal/mol). Interestingly, Mirex has been found to inhibit particularly Adenosine triphosphatase, a protein which is central for liver function. 80 This enzyme shares structural motifs with Phosphoinositide 3-kinase, which binds Mirex as calculated by CETOXA with a free energy of -7.7 kcal/mol. The common features between these two proteins suggest that the docking procedure detects structural similarities of particular enzyme motifs, even though these two proteins pertain quite different 3D structures. The second important finding is the docking result associated with the Farnesoid receptor. Indeed, this receptor is critical for liver function and has a central role in the detoxification mechanism in hepatocytes. This protein is also related to carcinogenesis and its inhibition causes considerable liver and gallbladder complications. [80][81][82] The result that Mirex binds to the Farnesoid receptor is thus relevant for the biochemistry of this receptor, as it is for the case of Adenosine triphosphatase. Another interesting finding is that pentachlorophenol binds strongly to CYP1A1 and -1A2. Both these enzymes use O-deethylation to oxidize phenolic substances and both are expressed in the liver and perform detoxification in the liver microsomes of pyrimidinelike and phenolic xenobiotics such as caffeine, alpha-napthoflavone and 7-ethoxyocourmarine and phenacetin. 83 The results also indicate that lindane may be destined for detoxification by CYP1A2.</p><p>Interestingly, lindane is a hexa-chlorinated compound that binds particularly well to CYP1A2, as CYP1A2 has a 30% higher affinity towards halogenated compounds compared to CYP1A1. This indicates that the docking procedure in this case has the potential to distinguish enzymes with very similar specificity with however a 30% preference difference towards halogenated aromatics. 83 Among the plasticizers, one phthalate (butyl benzyl phthalate) and three alternative plasticizers (dipropylene glycol dibenzoate, pentane-1,5-diyl dibenzoate, and hexane-1,6-diyl dibenzoate)</p><p>showed promiscuous behaviour, targeting over 1/3 of the receptors with ∆G bind < -8 kcal/mol.</p><p>Among the consumer use chemicals identified as potentially highly promiscuous are the surfactants perfluorodecanoic acid and perfluoroundecanoic acid, as well as the widely used synthetic food dyes allura red and FD&C Blue no. 1. Among the pesticides targetting multiple targets with high binding affinities are Famoxadone, Novaluron, and Prosulfon. The interactions of Famoxadone with multiple protein targets has been confirmed in the literature. In a study by , 84 Famoxadone was found to inhibit both mitochondrial enzymes as well as the cytochrome system (particularly CYPBc1 85 ).</p><p>The heatmap notably contains a cluster of about 28 pharmaceuticals with high Vina scores, binding to multiple targets. These include failed or terminated drugs, such as the thiazolidinedione based and non-thiazolidinedione based antihyperglycemic agents Troglitazone and Farglitazar, which are specific ligands for peroxisome proliferator-activated receptors (PPAR). Troglitazone has been found to have hepatotoxic effects, 86 while Farglitazar did not reach past phase III clinical trials. Zamifenacin in this list was also identified to have highly promiscuous biological activity in the ToxCast HTS. 10 Although drugs are typically designed to interact with a specific target, unintended drug-target interactions are one of the major challenges in drug design, associated with side-effects and high failure rates. 87,88 Assessment of Vina high scores</p><p>The highest binding affinity per chemical in the heat map was extracted and the top 20 chemicaltarget pairs are shown in Table 1. the triple arrangement of their aromatic rings. They expose their lateral electrons for oxidation by detoxification enzymes and are then converted to their diol-epoxide forms which react with DNA. 92,93 These compounds are ranked as a top compound to target CYP1A2 (Table 1), which reflects well with empirical studies 94 showing a high preference of CYP1A2 towards ethoxyresorufin. Ethoxyresorufin is geometrically similar to these compounds, suggesting the docking system recognizes the conserved molecular volume and geometry. There are no studies on fluroanthene and benzo [b]fluoranthene which could confirm their interaction with CYP1A2, however, by their elongated geometries, it is highly feasible that CYP1A2 is the detoxification enzyme for these two compounds, as both are known to be genotoxic. 95,96 A further study showed that CYP1A2 is activated by benzo[b]fluroanthene, however the expression of CYP1A2 was not confirmed. 96 In the prediction, the docking procedure indicated that CYP1A2 was the preferred target for benzo[b]fluroanthene, which indicates that physiological responses may be different than cellular responses when considered on a cohort of individuals exposed to the compound via airparticulate matter.</p><p>The potentially genotoxic and carcinogenic 97 food dye C.I. Solvent yellow 14 is predicted to bind strongly to both P450 1A1 and 1A2, with an affinity of -11.7 kcal/mol and -12.2 kcal/mol, respectively. Although 1A1 and 1A2 are highly homologous, CYP1A1 is considered to be most efficient in metabolizing C.I. Solvent yellow 14, whereas other CYPs, including CYP1A2, were previously found to be almost ineffective. 98 The discontinued PPAR agonist Farglitazar is predicted to bind strongly to the beta-2 adrenergic receptor, although it was developed to treat type 2 diabetes. 99 It has a predicted binding affinity to PPAR α and PPAR γ of -9.4 kcal/mol and -11.3 kcal/mol, respectively. In other words, many of the Vina high scores can be correlated with known toxic compounds.</p><!><p>Based on literature data and as considered in the ToxCast publications, 9,10 we assessed whether Vina succeeds to predict high scores for chemicals that are known to interact with certain proteins.</p><p>Some of these known complexes and their predicted binding affinities are listed in Table 2.</p><p>Bisphenol A 100 and 2,2-bis(4-hydroxyphenyl)-1,1,1-trichloro ethane, 101 a metabolite of the pesticide methoxychlor, are known estrogen receptor agonists. Perfluorooctanoic acid (PFOA) and perfluorooctane sulfonic acid (PFOS) are chemicals with widespread use and with multiple reported toxicities, generally thought to be triggered by activating PPAR α. 102,103 In the HTS screening in ref. 9, both were active for PPAR γ, but only PFOA was active for PPAR α. In our case, the predicted binding affinity is relatively high also for PPAR γ. It should be noted that one in vivo study found that neither compound activates PPAR γ, 104 illustrating potential risks in extrapolat-ing in vitro or in silico results to in vivo. The pesticide lactofen 105 is also an expected PPAR activator found to be positive for PPAR assays in ToxCast. 9 The pharmaceuticals CP-471358 and CP-544439 are known inhibitors of matrix metalloproteinases MMP2, MMP9, and MMP13, 106,107 and these complexes all have high Vina scores. However, relatively low predicted binding affinities are found for the complexes involving the androgen receptor (AR). The pesticides linuron, prochloraz, and vinclozolin are known to cause toxicity by interacting with AR as antagonists. 108 Binding to the ligand binding domain of nuclear receptors is known to cause structural and dynamic changes in the protein. 109 For example, one study used computer simulations to follow the long-time scale conformational fluctuations of AR interacting with different ligands and found that agonists and antagonists induce distinct conformational changes. 110 Since the androgen receptor used for docking here corresponds to an agonist complex, 111 with an associated redocking score of -8.9 kcal/mol, the low binding affinities of the (antagonist) pesticides may be due to the shortcomings of neglecting protein flexibility. This issue can be addressed in the future by expanding the computational ecotoxicity assay to include multiple protein structures for the same receptor, with both agonist and antagonist co-crystallised conformations to account for multiple binding modes.</p><p>The HTS screening in ref. 10 in some cases missed known active compounds. Zamifenacin is a muscarinic antagonist 112 that did not show activity for the M1 assay. 5,5-diphenylhydantoin is a known substrate and inhibitor for many CYPs, 113 but was not found to inhibit any of the CYPs in the HTS screening. These chemicals show high predicted binding affinities to their respective targets. Additional chemical-target interactions that were considered, including estrogenic and nonestrogenic compounds within the EPA's Endocrine Disruptor Program, can be found in SI Tables S2 and S3.</p><!><p>Considering potentially missed interactions and weakly binding active chemicals, the 20 least promiscuous chemicals in the docking simulations were checked for measured activity in the HTS assays. Of these 1300 complexes with low predicted binding affinities, only 8 were reported as active in ToxCast. These assays include a number of nuclear receptors which, as noted above, may undergo conformational changes upon ligand binding. In one case, sodium nitrite in complex with monoamine oxidase A (MAO), the low predicted binding affinity may be due to the mechanism of inhibition, as nitrates would appear to inhibit MAO through oxidation of the SH-groups, 114 a mechanism not captured in current docking methods.</p><!><p>In this work, we assessed the strengths and weaknesses in employing molecular docking for screening potential toxicity of xenobiotic compounds. This approach allows the generation of interaction maps between potential toxins and targets linked to perturbation pathways in a a cost-efficient and timely manner, yielding molecular-level information. While neither the classical Vina scoring function nor the machine-learning scoring function RF-Score-VS considered here were capable of distinguishing active chemicals, Vina on average predicted higher affinities for active chemicals and succeeded in identifying several interactions that could be confirmed in the literature. To be able to achieve higher docking accuracy with fewer false negatives, the method can be further developed by considering, for example, protein side-chain conformational changes, covalent interactions, charge redistribution, and bio-transformation products. A further refinement could be to consider multiple structures per target to take into account multiple binding modes. More exact methods may be required to distinguish binders from non-binders and the obvious challenge there remains to balance computational cost and accuracy. Finally, current scoring functions are limited in their ability to predict biological activity, underscoring the need to consider other properties</p>

ChemRxiv

Ferroportin (SLC40A1) Q248H mutation is associated with lower circulating plasma tumor necrosis factor-\xce\xb1 and macrophage migration inhibitory factor concentrations in African children

Background Iron deficiency and the Q248H mutation in the gene, SLC40A1, that encodes for the cellular iron exporter, ferroportin, are both common in African children. The iron status of macrophages influences the pro-inflammatory response of these cells. We hypothesized that Q248H mutation may modify the inflammatory response by influencing iron levels within macrophages. Methods The Q248H mutation and circulating concentrations of ferritin, C-reactive protein and selected pro-inflammatory cytokines (interleukin-12, interferon-\xce\xb3, TNF-\xce\xb1, and macrophage migration inhibitory factor) and anti-inflammatory cytokines (interleukin-4 and interleukin-10) were measured in 69 pre-school children recruited from well-child clinics in Harare, Zimbabwe. Results In multivariate analysis, both ferroportin Q248H and ferritin <10 ug/L were associated with significantly lower circulating concentrations of tumor necrosis factor-\xce\xb1. Ferroportin Q248H but not low iron stores was associated with lower circulating macrophage migration inhibitory factor as well. Anti-inflammatory cytokine levels were not significantly associated with either ferroportin Q248H or iron status. Conclusions Ferroportin Q248H and low iron stores are both associated with lower circulating tumor necrosis factor-alpha, while only ferroportin Q248H is associated with lower circulating macrophage migration inhibitory factor. Whether the reduced production of tumor necrosis factor-\xce\xb1 observed in ferroportin Q248H heterozygotes may be of significance in anemia of chronic disease is yet to be determined.

ferroportin_(slc40a1)_q248h_mutation_is_associated_with_lower_circulating_plasma_tumor_necrosis_fact

Introduction<!>Study participants<!>Laboratory measurements<!>Definition of iron status<!>Statistical analysis<!>Circulating cytokine concentrations according to ferroportin Q248H<!>Circulating cytokine concentrations according to iron status<!>Independent relationship of circulating cytokine concentrations to ferroportin Q248H and iron status by multivariate analysis<!>Discussion<!>

<p>The SLC40A1 gene encodes a multiple trans-membrane domain protein, ferroportin, which is responsible for iron efflux from mature enterocytes of the duodenum and from macrophages of the spleen and bone marrow to plasma [1-3]. Duodenal enterocytes are responsible for absorption of iron from the diet and macrophages are responsible for recycling iron that is recovered from the catabolism of erythrocytes that they remove from the circulation [4]. Cellular export of iron by ferroportin is regulated by hepcidin, which is produced by hepatocytes in response to inflammatory cytokines or to increased iron stores [5-8]. Hepcidin directly interacts with ferroportin on the cell membrane causing internalization of ferroportin, subsequent degradation of the ferroportin by lysosomes, and reduced export of iron from cells [6,7]. The SCL40A1 mutations may be associated with predominantly parenchymal or predominantly macrophage iron-loading [9-11]. A number of disease-causing SCL40A1 gene mutations have been shown to render ferroportin resistant to hepcidin in model systems in vitro [12,13]. Such mutations would tend to be associated with increased iron absorption by enterocytes, increased iron-release by macrophages and parenchymal iron-loading.</p><p>The cDNA 744G>T substitution in exon 6 of the ferroportin gene (dbSNP rs11568350, www.ncbi.nlm.nih.gov), which results in the replacement of glutamine with histidine at position 248 (Q248H), is common in Africans and African Americans (prevalence of heterozygotes of 5% to 20%). Some studies have suggested association of the ferroportin Q248H with a tendency to increased iron stores in healthy adults and with possible protection from iron deficiency in children attending well child clinics [14-19].</p><p>Studies at the cellular level have shown that iron metabolism and immunity are interrelated [20]. Pro-inflammatory cytokines such as tumor necrosis factor-α (TNF-α), interleukin-1 (IL-1) or interleukin-6 (IL-6) induce increased synthesis and content of ferritin in mononuclear-phagocytic cells [21-23], whereas the anti-inflammatory cytokines such as IL-4 and IL-13 increase macrophage expression of transferrin receptors [24]. At the systemic level, studies in experimental animals and humans have shown that administration of TNF-α suppresses serum iron concentration [25-27]. Not only does inflammation affect iron metabolism, but iron status itself may modulate the production of inflammatory cytokines [28-30]. Iron deficiency anemia in infants was reported to increase lipopolysaccharide-induced production of tumor necrosis factor-α by peripheral blood mononuclear cells [29]. In contrast to this report, peripheral blood mononuclear cells from patients with hereditary hemochromatosis, a condition with paradoxically low iron concentration in such cells, released less TNF-alpha than peripheral blood mononuclear cells from controls [31], and iron supplementation of peripheral blood mononuclear cells from patients with iron deficiency anemia led to increased mRNA expression of TNF-α, IL-6 and IL-10 [32]. Several studies have found a relationship between reduced iron status and decreased NF-kB-mediated TNF-α production by rodent hepatic macrophages [33,34]. A peroxynitrite-mediated transient rise in intracellular labile iron may be a signal for endotoxin-induced IκB kinase (IKK) and NF-κB activation in rodent [35] and human macrophages [36] leading to TNF-α expression; this process is blocked by iron chelation.</p><p>Why the ferroportin Q248H allele has achieved a high frequency in African populations is not known, but one possibility is that it in some manner influences immune responses in a favorable manner through altering iron levels in macrophages and/or serum. In this study we investigated the effect of the ferroportin Q248H mutation on plasma cytokine concentrations.</p><!><p>Subjects were selected from a study of iron deficiency in African children that has been described previously [16]. Briefly, 208 apparently healthy children aged ≤60 months attending well-child clinics in Harare, Zimbabwe were enrolled. Five milliliters of peripheral blood were drawn from each child during the morning into two vacutainer tubes, one containing K3-EDTA and one with no anticoagulant. In the current study, only 69 children (37 females) had sufficient plasma to perform cytokine measurements. Ethical permission was granted by the Medical Research Council of Zimbabwe and the Howard University IRB and written permission was obtained from the mothers or guardians.</p><!><p>Complete blood counts were performed on an automated analyzer (Sysmex, Norderstedt, Germany). Serum ferritin and C-reactive protein (CRP) concentrations were measured using enzyme immunoassays with commercially available kits (Ramco Laboratories, Stafford, TX and ALPCO Diagnostics, Windham, NH, respectively). Plasma samples were assayed in duplicate and concentrations of IL-4, IL-10, IL-12, interferon-γ and TNF-α were measured using Human Cytokine/Chemokine Multiplex Immunoassay kits (LINCO Research, St. Charles, MO, USA). Plasma concentrations of macrophage migration inhibitory factor (MIF) were measured using Human Sepsis/Apoptosis Multiplex Immunoassay kits (LINCO Research, St. Charles, MO, USA).</p><p>The SLC40A1 Q248H mutation was identified as previously reported [16]. In brief, DNA was isolated from leukocytes obtained from whole blood using lymphocyte separation media (Media-Tech, Stirling, VA). Exon 6 of ferroportin gene was amplified using a set of primers (forward primer: 5′-CAT CGC CTG TGG CTT TAT TT-3′; reverse primer 5′-GCT CAC ATC AAG GAA GAG GG-3′) in a PTC- 100 thermocycler (MJ Research Inc. Waltham, MA) to make a 392bp product. The 392bp product was digested by PvuII enzyme (MBI Fermentas, Hanover, MD) and the resulting DNA fragments (252bp, 140bp) were separated on 3% agarose gel and detected with ethidium bromide.</p><!><p>Low iron stores was defined as serum ferritin concentration of <10 μg/L [37]. C-reactive protein concentrations >8.2 mg/L were taken to be elevated as provided by the manufacturer of the kit. Population reference values for C-reactive protein are not available for this population.</p><!><p>Statistical analysis was performed with SYSTAT software (version 11; SYSTAT Software, Inc, Point Richmond, CA). Clinical variables were compared according to ferroportin Q248H mutation status with the Kruskal-Wallis test for continuous variables and the Fisher exact test for proportions. Because age varied significantly according to iron status, continuous variables were compared according to iron status by analysis of variance with adjustment for age and categorical variables were compared by logistic regression with adjustment for age. To determine the independent associations of circulating cytokine concentrations with ferroportin Q248H and iron status, we constructed multivariate linear regression models. For parametric analyses, variables that followed a skewed distribution were log transformed.</p><!><p>Table 1 summarises the laboratory data of the study population. The median serum ferritin median concentration was 19 μg/L (interquartile range: 11-33 μg/L). The median TNF-α concentration was 16.02 pg/mL (interquartile range: 5.14-20.14 pg/mL) whereas the median MIF concentration was 13.68 ng/mL (interquartile range: 4.56-26.30ng/mL). Fourteen (22%) of 64 children were heterozygous for ferroportin Q248H. In five children ferroportin genotype was not determined due to poor quality of the DNA recovered. Clinical features and plasma cytokine concentrations are presented in Table 2 according to ferroportin Q248H status. Concentrations of tumor necrosis factor-α and macrophage migration inhibitory factor were significantly lower in the patients with ferroportin Q248H. Serum ferritin was <10 μg/L in 7% of the children with ferroportin Q248H versus 24% of those with ferroportin wildtype, but this difference did not reach statistical significance.</p><!><p>Thirteen (19%) of 69 children had serum ferritin concentration <10 μg/L Clinical features and plasma cytokine concentrations are presented in Table 3 according to serum ferritin concentration <10 μg/L versus ≥10 μg/L. Children with serum ferritin <10 μg/L were significantly younger than the children with higher serum ferritin concentration. Because of this, the clinical features and cytokine levels were compared according to iron status category in analyses that adjusted for age. The age-adjusted plasma TNF-α concentration were significantly lower in the children with lower ferritin concentration, but the macrophage migration inhibitory factor concentrations did not differ significantly according to iron status. The Q248H allele was present in 8% of the children with ferritin <10 μg/l compared to 25% of those with ferritin concentrations of 10 μg/L or higher.</p><!><p>We examined the relationships of both serum ferritin concentration <10 μg/L and ferroportin Q248H with log TNF-α in a single linear regression model that adjusted for age and log C-reactive protein concentration. The independent associations of both low iron status (P = 0.015) and the ferroportin Q248H allele (P = 0.007) with lower plasma TNF-α concentration were confirmed in this model as shown in Table 4. In a similar model, ferroportin Q248H but not low iron status was associated with lower plasma MIF concentration (geometric mean MIF concentration of 5.7 ng/mL with ferroportin Q248H versus 14.9 ng/mL with ferroportin wildtype; P = 0.013).</p><!><p>In this study with a limited sample size, we observed lower plasma concentrations of TNF-α in African pre-school children with the ferroportin Q248H mutation or with low iron status. We also observed lower concentrations of MIF in association with ferroportin Q248H but not with low iron status. Tumor necrosis factor-α is an important mediator of the inflammatory process and is produced predominantly by activated monocytes/macrophages [38,39]. Several mechanisms control the secretion of TNF-α. NF-kB plays a role in expression of proinflammatory genes including TNF-α gene [reviewed in 40], but lipopolysaccharide (LPS) and viruses have been shown to induce the TNF-α gene independently of NF-kB [41]. Macrophage migration inhibitory factor is a pro-inflammatory cytokine that is produced by activated macrophages and plays a role in the systemic inflammatory response by counter-regulating the inhibitory effects of glucocorticoids on TNF- α and IL-6 production [42]. The effect of ferroportin Q248H mutation on plasma concentrations of TNF-α and MIF in this study could not seem to be attributed to inflammation, for it persisted after adjustment for the C-reactive protein concentration.</p><p>Since monocytes and-macrophages are major sources of circulating TNF-α, the findings we report here are consistent with low iron stores in children limiting TNF-α production by macrophages. These results are also consistent with the possibility that ferroportin Q248H in children is associated with increased export of iron from macrophages, reduced intracellular labile iron concentration, and consequent decreased production of TNF-α. One in vitro study indicated that the Q248H allele impairs the egress of iron when expressed in Xenopus oocytes [43]. Other studies indicated that the Q248H allele retains the ability to export iron and respond to hepcidin when expressed in HEK 293T cells [12,13]. In fact, Drakesmith and colleagues found that ferroportin Q248H was as susceptible to 0.5 μM hepcidin as wildtype ferroportin in this experimental setting, but commented that this polymorphism may have a mild effect on ferroportin function that they could not detect and possibly lead "to disease in the presence of modifying factors" [12,13].</p><p>Thus, as postulated by Schimanski et al, Q248H mutation in SCL40A1 gene may result in gain of function by ferroportin leading to depletion of macrophage cellular iron [13]. This would lead to reduced macrophage cellular iron which in turn would lead to reduced activation of the transcription factor NF-kB and consequently reduced TNF-α production [33]. As a result, macrophage TNF-α production is reduced in persons with ferroportin Q248H. Conversely, increased iron content in Kupffer cells promotes activation of NF-kB and subsequent induction of TNF-α expression [35,44]. We have recently demonstrated that TNF-α release is enhanced in iron-laden macrophages derived from human blood monocytes due to accentuated intracellular labile iron [36]. However, to our knowledge no studies have demonstrated that macrophages obtained from ferroportin Q248H heteozygotes have lower intracellular iron content compared to macrophages obtained from ferroportin wildtype persons.</p><p>Low iron stores, plasma ferritin < 1μg/L, was associated with lower plasma TNF-α concentration. Similar findings were reported by Wang et al [45]. In experimental studies on Hfe knockout mice reduced intra-macrophage iron resulted in decreased TNF-α secretion [45]. In this study, ferroportin Q248H was associated with lower concentration of MIF compared to ferroportin wildtype. Secretion of MIF is mediated by several pathways. In one pathway, TNF-α induces MIF gene expression resulting in elevated levels of circulating plasma MIF [46]. This mechanism could explain the observed correlation of TNF-α plasma levels with MIF concentration. Thus, ferroportin Q248H which is associated with lower TNF-α concentration would have correspondingly lower MIF levels.</p><p>The biological significance of the ferroportin Q248H mutation which is unique and prevalent in African populations is yet to be determined. Iron deficiency is endemic in sub-Saharan Africa due to poor dietary iron sources and chronic hookworm infestations [47]. In the parent study, we proposed that the ferroportin Q248H mutation may be protective against iron deficiency in children exposed to repeated infections [16]. In recent studies in mice, TNF-α was implicated in the pathogenesis of anemia of chronic disease by enhancing iron sequestration in spleen macrophages, by reducing iron transfer from duodenal enterocytes [48] and by inducing hypoferraemia [27]. Furthermore, increased circulating TNF-α concentration induces macrophage iron accumulation through increased expression of divalent metal transporter-1 and down-regulation of ferroportin expression [23] probably due to TNF-α-enhanced hepatocellular production of hepcidin which binds to ferroportin causing its internalization and degradation leading to decreased release of recycled iron by macrophages. Conversely, reduced circulating TNF-α concentration would lead to reduced hepcidin production and hence decreased internalization of ferroportin [49]. However, in this study hepcidin concentration was not measured.</p><p>If findings of our study are corroborated in a study with a larger sample size, we hypothesize that ferroportin Q248H mutation may contribute to lower plasma TNF-α concentration. As a result iron recycling by macrophages would not be severely impaired in chronic inflammation. However, we were unable to measure serum iron concentration in this population due to insufficient sample volume. Whether the reduced production of TNF-α and subsequently decreased MIF concentration associated with ferroportin Q248H mutation may be of significance in anemia of chronic disease is yet to be determined.</p><!><p>Comparison by the non-parametric Kruskall-Wallis test for continuous variables and the Fisher exact test for proportions.</p><p>Comparison by ANOVA unless otherwise indicated. All analyses except age are adjusted for age.</p><p>Comparison by logistic regression.</p><p>Results are expressed as median (interquartile range) unless otherwise specified.</p>

PubMed Author Manuscript

Facile synthesis of AIEgens with wide color tunability for cellular imaging and therapy

Luminogens with aggregation-induced emission (AIE) characteristics are nowadays undergoing explosive development in the fields of imaging, process visualization, diagnosis and therapy. However, exploration of an AIE luminogen (AIEgen) system allowing for extremely wide color tunability remains challenging. In this contribution, the facile synthesis of triphenylamine (TPA)-thiophene building block-based AIEgens having tunable maximum emission wavelengths covering violet, blue, green, yellow, orange, red, deep red and NIR regions is reported. The obtained AIEgens can be utilized as extraordinary fluorescent probes for lipid droplet (LD)-specific cell imaging and cell fusion assessment, showing excellent image contrast to the cell background and high photostability, as well as satisfactory visualization outcomes.Interestingly, quantitative evaluation of the phototherapy effect demonstrates that one of these presented AIEgens, namely TTNIR, performs well as a photosensitizer for photodynamic ablation of cancer cells upon white light irradiation. This study thus provides useful insights into rational design of fluorescence systems for widely tuning emission colors with high brightness, and remarkably extends the applications of AIEgens.

facile_synthesis_of_aiegens_with_wide_color_tunability_for_cellular_imaging_and_therapy

Introduction<!>Synthesis and single crystal analysis<!>Photophysical properties<!>Theoretical calculations<!>Bio-imaging, visualization of cell fusion and photodynamic therapy<!>Conclusions<!>Materials and methods<!>Synthesis of compound TTV 7c,22<!>Cell imaging and confocal co-localization<!>Photostability<!>ROS generation and PDT study<!>Cell fusion<!>Conflicts of interest

<p>The exploration of uorescent materials and technologies has opened new avenues to scientic advancement, societal development and public health, 1 which is exemplied by the Nobel Prize successively awarded to uorescence-related research. As one of the most important branches of uorescent materials, uorescent bio-materials that offer researchers a powerful platform for analytical sensing and optical imaging have been proven to be extremely useful for biological visualization, clinical diagnosis and disease treatment by virtue of their noninvasiveness, in situ workability, excellent accuracy, superb sensitivity and simple operation. 2 Although many types of uorophores have been commercialized for biological applications, the current situation is still far from ideal, mainly due to some limitations: (1) inherent uorescence quenching upon aggregate formation due to intermolecular p-p stacking and other nonradiative pathways, which is notoriously known as aggregation-caused quenching (ACQ); 3 (2) the difficulty of widely tuning emission colors by simple modication of molecular structures; and (3) complicated and laborious syntheses of uorophores. 4 As an anti-ACQ phenomenon, aggregation-induced emission (AIE) was coined in 2001 by Tang's group, 5 which refers to a unique phenomenon that a novel class of uorophores are non-emissive or weakly emissive in the molecularly dissolved state but they emit intensively in aggregated states owing to the restriction of the intramolecular motions (RIMs). 6 Remarkably, the AIE principle has triggered state-of-the-art developments in an array of biological elds, ranging from bioimaging, biosensing, stimuli-responsive systems, and therapeutics to theranostics, mainly resulting from various impressive advantages of AIE luminogens (AIEgens), such as high photobleaching threshold, high signal-to-noise ratio for imaging, excellent tolerance for any concentrations, large Stokes shi, turn-on feature for detecting analytes, and efficient photosensitizing ability. 7 Although numerous AIEgens have been constructed on the basis of different structural motifs including tetraphenylethene, 8 hexaphenylsilole, 9 tetraphenylpyrazine 10 and distyrylanthracene, 11 to the best of our knowledge, there has been no single AIE system which allows arbitrarily tuning emissions ranging from each color of visible light to the nearinfrared (NIR) region. Considering the great signicance of tunable uorescent systems in the applications of multi-target sensing, optoelectronic devices and full-color bio-imaging, 12 the development of an AIE system exhibiting wide color tunability is highly desired and remains a challenging task.</p><p>Compared with inorganic complexes and quantum dots, organic uorophores are advantageous for bio-imaging, diagnosis and therapy, beneting from their good bio-compatibility, tunable molecular structures and chemical compositions, and scalable synthesis. 13 Evidently, the exploration of an organic uorophore system with both the AIE attribute and emission color tunability across a wide wavelength range would captivate much interest. Herein, we report for the rst time the design and synthesis of a series of AIEgens having widely tunable emissions covering violet, blue, green, yellow, orange, red, deep red and NIR regions (Fig. 1). Each AIEgen comprising the triphenylamine (TPA)-thiophene building block is facilely obtained through oneor two-step reaction, and the emission colors are tuned by simple alteration of the HOMO-LUMO energy level by the introduction of electron donor (D)-acceptor (A) substituents. 14 Moreover, these AIEgens can be successfully utilized as extraordinary lipid droplet (LD)-specic bioprobes in cell imaging, determination of cell fusion, and photodynamic cancer cell ablation.</p><!><p>As depicted in Scheme 1, the desired compounds were facilely prepared through one or two steps. TTV was synthesized through the Suzuki-Miyaura coupling reaction of 4-bromo-N,N-diphenylaniline with thiophen-2-ylboronic acid in the presence of the palladium catalyst using mixed THF/H 2 O as the solvent at 75 C. The same synthetic procedure was successfully conducted by employing substituted 4-bromo-N,N-diphenylaniline and modi-ed thiophen-2-ylboronic acid as starting materials, producing compounds TTG, TTY, TTO and TTR. The reactions between TTG/TTO and malononitrile proceeded smoothly, giving the corresponding products TTDR and TTNIR with moderate yields. In addition, TTB was obtained by the Suzuki-Miyaura coupling reaction of (4-(1,2,2-triphenylvinyl)phenyl)boronic acid with intermediate product 1, which was isolated through the Suzuki-Miyaura coupling reaction between (4-(diphenylamino)phenyl) boronic acid and 2,5-dibromothiophene.</p><p>All compounds are composed of sufficient moieties that can freely rotate in the single-molecule state leading to energy consumption of the excited state through non-radiative pathways, thus ensuring that these compounds are weakly emissive in solution. Aiming to further study and deciphering their optical properties in the aggregation state, single crystals of TTG, TTY and TTDR were grown in DCM-MeOH mixtures by slow solvent evaporation. As illustrated in Fig. 2, S1 and S2, † the twisted conformation of the TPA segment extends the intermolecular distance (>3.2 Å) between two parallel planes, remarkably reducing or avoiding the intermolecular p-p interactions, and essentially preventing emission quenching in its aggregation state. Moreover, the molecular conformation can be strongly rigidied by abundant intermolecular interactions (such as C-H/O, C-H/C, and S/C) resulting in the restriction of molecular motions, which is benecial for enhancing the solid state emission efficiency. On the basis of the above-mentioned XRD results, it is believed that these synthesized compounds are potentially AIE-active.</p><!><p>The UV-vis absorption spectra of TTV, TTB, TTG, TTY, TTO, TTR, TTDR and TTNIR were measured in acetonitrile (ACN). As shown in Fig. 3A and Table S1, † the solution of building block TTV displays a maximum absorption band at 348 nm, and the maximum absorption peaks of these modied compounds are located ranging from 383 nm to 512 nm. The gradually red-shied absorption wavelengths can be attributed to the orderly enhanced D-A effect from TTV to TTNIR. To investigate their AIE features, an ACN/H 2 O mixture with different H 2 O fractions was utilized as the solvent system. It was observed that compounds TTB, TTG, TTY, TTO, TTR, TTDR and TTNIR exhibit typical AIE features (Fig. 3C). Taking TTY as an example, there is almost no uorescence emission when the H 2 O fraction is below 60%. Aerwards, the PL intensity increases dramatically along with raising the fraction of water because of activation of RIM by molecular aggregation and reaches its maximum at 90% water fraction, which is 185-fold higher than that in ACN solution (Fig. 3B). Although the uorescence intensity of TTV is inversely proportional to the water fraction, its quantum yield in the solid state (27.5%) is higher than that in the solution state (18.6%), denitely demonstrating an aggregation-induced emission enhancement (AIEE) attribute. The gradually decreased uorescence intensity of TTV along with the increased water fraction could be attributed to its twisted intramolecular charge transfer (TICT) feature, 15 which was determined by both the red-shied emission wavelength and the declined emission efficiency accompanying the raised solvent polarity (Fig. S3 †). As one of the nonradiative pathways for the excited state to relax and 2019, 10, 3494-3501 | 3495 deactivate, the TICT effect is competitive with AIE properties in determining the PL intensity and efficiency using the ACN/ H 2 O solution system. In the case of TTV, the AIE feature is strongly depressed by the TICT effect in the nanoaggregation state. As illustrated in Fig. 3D and E and Table S1, † these TPAthiophene building block-based AIEgens emit efficiently in both nanoaggregation and solid states exhibiting relatively high quantum yields ranging from 3.11% to 40.79%. Each maximum emission wavelength accurately peaks in violet (402 nm), blue (482 nm), green (531 nm), yellow (580 nm), orange (612 nm), red (649 nm), deep red (667 nm) and NIR (724 nm) regions, respectively, suggesting the extremely wide emission color tunability, which is ascribed to both of their varied pconjugation and D-A effect. Additionally, the uorescence decay curves in the solid state show that their lifetimes range from 0.64 to 3.69 ns (Fig. 3F and Table S1 †).</p><!><p>To better understand the optical properties of these AIEgens, density functional theory (DFT) calculations were carried out at the B3LYP/6-31+G(d) level with molecular geometries optimized at the TD-B3LYP/6-31+G(d) level (Fig. 4). It was observed that, from TTV to TTNIR, the calculated HOMO-LUMO energy gaps generally decrease, and the results are in good accordance with the experimental data of emission maximums. The orderly declined energy gaps are realized through ingenious modication of the TPA-thiophene building block with diverse electron-donating (thienyl or methoxyl groups) and electron-accepting (aldehyde or cyano groups) units or the p-bridge. Except for TTB, the HOMOs of the remaining AIEgens are delocalized at the TPA moiety, whereas their LUMOs are distributed on the other side of the structures, demonstrating typical D-A structural features. It has been demonstrated that the separation of HOMO and LUMO distributions is essential to effectively reduce the singlet-triplet energy gap, which facilitates the generation of reactive oxygen species (ROS), 16 further endowing these AIEgens with prominent potential for photodynamic therapy (PDT) applications. 17 In contrast, TTB possesses an evenly distributed HOMO and LUMO, resulting from its both imperceptible D-A effect and long p-conjugation bridges.</p><!><p>In the preliminary bioimaging experiment, the cell imaging study was conducted by using HeLa cells as a cell model. Cells were incubated with 1 mM of TTNIR for 20 min; as illustrated in Fig. 5B, bright uorescence within cells can be observed showing excellent image contrast to the cell background. The co-localization study further proceeded upon incubating HeLa cells with TTNIR and BODIPY493/503 Green. The latter dye is a commercially available bioprobe for the LDs, which are ubiquitous lipid-rich spherical organelles and actively involved in various biofunctions, such as signal transduction, lipid , 2019, 10, 3494-3501 This journal is © The Royal Society of Chemistry 2019 metabolism, and protein degradation. The perfect overlap between TTNIR and BODIPY493/503 Green in cell imaging output indicates the excellent LD-specic targeting capability of TTNIR (Fig. 5B-D). Photostability is a key criterion for evaluating the overall stability of photosensitive substances. The continuous scanning method was then utilized to quantitatively study and compare the photostability of TTNIR and BODIPY493/503 Green. As shown in Fig. 5E-I, aer 15 minutes of laser irradiation, the uorescence intensity of BODIPY493/ 503 Green encounters an obvious decline, whereas TTNIR shows negligible photobleaching. Moreover, the photostability assessment was also conducted towards Nile Red, which is another commercially available dye for LD-staining (Fig. S13 †). It was observed that Nile Red suffered an obvious signal loss with remaining signal intensity around 60% aer 15 minutes of laser irradiation, strongly suggesting that the photostability of TTNIR is superior to that of commercially available bioprobes. To further prove its applicability, this staining and imaging strategy using TTNIR is exploited for other cell lines, including NCM460, DLD1, SW480, SW620 and COS-7 (Fig. S4 †). In each case aer incubation with TTNIR for 20 min, it shows strong and specic internalization into the LDs. Moreover, other AIEgens including TTV, TTB, TTG, TTY, TTO, TTR and TTDR were also investigated for cell imaging. It was observed that LDs can be clearly visualized with excellent image contrast to the cell background through respective incubation of cells with these presented AIEgens (Fig. 6). Pearson's correlation coefficients between AIEgens and commercially available LD-bioprobes were determined to be 90-95%, solidly demonstrating the high specicity of these AIEgens for staining LDs (Fig. S5-S12 †). Their excellent LD-staining specicity reasonably results from the lipophilic properties, which bring about efficient accumulation of them in the hydrophobic spherical LDs due to the "like-like" interactions. Evidently, these AIEgens possess various impressive features, such as high brightness, excellent targeting specicities to LDs, extraordinary photostabilities and widely tunable emission colors, making them remarkably important in visualization of biological structures and processes.</p><p>As a common phenomenon in nature, cell fusion is highly associated with many cellular processes, including fertilization, development of placental, regeneration of skeletal muscle, oncogenesis, aneuploidy, chromosomal instability and DNA damage. 18,19 In addition, a recent study shows that cell fusion could play a vital role in alternative therapies for restoring organ function through repairing cellular dysfunction. 19 Therefore, the development of effective methods for visualizing cell fusion is of great importance. Encouraged by the excellent cell imaging results and homology of the presented AIEgens, a straightforward method for visualization of the cell fusion outcome was conducted by using the combination of TTG and TTNIR as cell imaging agents, due to their minimal overlap of the emission range. In this experiment, two sets of cells were respectively stained with TTG and TTNIR, which were then mingled and treated with polyethylene glycol (PEG) to induce cell fusion. 20 As illustrated in Fig. 7, aer treatment with PEG, both green and red uorescence of lipid droplets were observed within one single cell, suggesting that cell fusion between TTG-and TTNIRstaining cells successfully proceeded. In addition, the cell fusion outcome was also solidly veried through a commercially available nuclei-staining agent Hoechst 33258. The appearance of two stained nuclei within one single cell (Fig. 7D) indicated that the visualization strategy of the cell fusion outcome by using two AIEgens with different emission ranges is denitely reliable. Evidently, the developed AIEgens having widely tunable emissions and high emission efficiencies are potentially useful in the fundamental study of cell fusion.</p><p>Intense uorescence in the near-infrared (NIR) region is highly desirable for many clinical processes, due to the salient advantages of deep tissue penetration, minimal photodamage to biological structures, and high image contrast to the physiological background. 21 Moreover, NIR emission is generally realized by intensifying the D-A effect of the structure, resulting For TTG, l ex : 405 nm (1% laser power), l em : 425-540 nm. For TTNIR, l ex : 560 nm (6.5% laser power), l em : 600-740 nm. For Hoechst 33258, l ex : 405 nm (3.5% laser power), l em : 425-540 nm. Concentrations: TTG (500 nM), TTNIR (2 mM), Hoechst 33258 (2.5 mM). Scale bar ¼ 20 mm.</p><p>in the separation of HOMO and LUMO distribution, as well as the decrease of the singlet-triplet energy gap, thus facilitating the generation efficiency of ROS. Therefore, the AIEgen TTNIR with both bright NIR emission and the strong D-A effect is potentially efficient for PDT, which is an extraordinary therapeutic modality, and has captivated much interest for treating various malignant and non-malignant diseases with minimal invasiveness and precise controllability. In the preliminary test, the ROS generation efficiency of TTNIR was investigated using H2DCF-DA as the indicator, which can emit uorescence at around 534 nm triggered by ROS. As shown in Fig. 8A, in the presence of TTNIR, the emission of H2DCF-DA was rapidly intensied with the increase of irradiation time using white light as the irradiation source, reaching 36-fold enhancement in 6 min compared with the original emission intensity. In contrast, the uorescence intensities of AIEgens or H2DCF-DA alone were very low and remained almost constant under the same irradiation conditions. These results reveal good photo-sensitizing properties for ROS generation. Quantitative evaluation of the phototherapy effect of TTNIR on HeLa cells was then explored through the standard 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay. The dose-dependent toxicity study shows that there is no obvious cytotoxicity observed for the HeLa cells treated with TTNIR in the dark, even with the TTNIR concentration reaching as high as 20 mM (Fig. 8B). Upon white light exposure, cell viability dropped gradually with raising the concentration of TTNIR. Only 7% of cell viability remained with utilizing 20 mM of TTNIR, demonstrating almost complete cell apoptosis. Apparently, TTNIR holds high effectiveness for cancer cell ablation by means of PDT.</p><!><p>To sum up, we report the rst series of AIEgens with widely tunable emissions covering the whole visible region extending to the NIR area. These TPA-thiophene building block-based AIEgens can be facilely prepared by extremely simple synthetic protocols, and show high uorescence quantum yields up to 40.79% in the solid state beneting from their intrinsic aggregation-induced emission nature. They have been successfully utilized for LD-specic cell imaging, showing excellent image contrast to the cell background and higher photostability than the commercial LD-staining uorophore. Additionally, the high brightness and homology of these AIEgens endow them with excellent performance for visualizing cell fusion. To the best of our knowledge, this would be the rst report on using AIEgens as uorescent probes for assessing cell fusion. Notably, upon exposure to white light irradiation, one of these presented AIEgens, namely TTNIR, displays high ROS generation efficiency, enabling its effective application for photodynamic ablation of cancer cells.</p><p>Our ndings in this study provide an ideal uorescence system for widely tuning emission colors with high brightness at will. This successful example would further facilitate the exploration of organic uorophores with AIE features for preclinical research and clinical applications.</p><!><p>Chemicals for synthesis were purchased from Sigma-Aldrich, MERYER or J&K, and used without further purication. All solvents were puried and dried following standard procedures. 1 H spectra were measured on Bruker ARX 400 NMR spectrometers using CD 2 Cl 2 or CDCl 3 as the deuterated solvent. Mass spectrometric measurements (HRMS) were performed on a Finnigan MAT TSQ 7000 mass spectrometer system operating in matrix-assisted laser desorption/ ionization time of ight mass spectrometry (MALDI-TOF) mode. UV-vis spectra were measured on a Milton Ray Spectronic 3000 array spectrophotometer. Steady-state photoluminescence (PL) spectra were recorded on a PerkinElmer LS 55 spectrophotometer. Fluorescence images of AIEgens in the solid state and aggregation state were collected on an Olympus BX 41 uorescence microscope. The cellular uorescence images were taken using a Zeiss laser scanning confocal microscope (LSM7 DUO) and analyzed using ZEN 2009 soware (Carl Zeiss).</p><!><p>A mixture of the bromide substituted triphenylamine moiety (1.2 mmol), thiophen-2-ylboronic acid moiety (1 mmol), THF (20 mL), K 2 CO 3 aqueous solution (2 M, 1.6 mL), and Pd(PPh 3 ) 4 (58 mg, 0.05 mmol) was degassed and charged with N 2 . The mixture was reuxed overnight. The reaction was quenched by the addition of water (30 mL) and extracted with CH 2 Cl 2 (3 Â 30 mL). The combined organic layer was dried over anhydrous Na 2 SO 4 and evaporated. The residue was puried by column chromatography over silica gel using petroleum ether to afford the desired product TTV with a yield of 78%. 1 H NMR (400 MHz, CD 2 Cl 2 ): 7.60 (d, J ¼ 6.8 Hz, 2H), 7.41 (d, J ¼ 8 Hz, 2H), 7.37-7.33 (m, 4H), 7.13-7.06 (m, 9H). 13 C NMR (100 MHz, CDCl 3 ): 147. 49, 147.20, 144.26, 129.27, 128.54, 127.95, 126.71, 124.42, 123.98, 123.75, 123.02, 122.21</p><!><p>In 35 mm glass-bottomed dishes, the cells (NCM460, DLD1, SW480, and SW620) were seeded and cultured at 37 C. Aer incubation with TTNIR (1 mM) for 20 min, the cells were washed with PBS three times and subjected to imaging analysis using a laser scanning confocal microscope (Zeiss Laser Scanning Confocal Microscope; LSM7 DUO). The excitation lter was 488 nm and the emission lter was 570-740 nm. For the costaining assay, the AIEgen loaded COS-7 cells were subjected to incubation with BODIPY 493/503 Green or Nile red for 20 min. Aerwards, the cells were washed with PBS and then observed with CLSM. The cells were imaged using appropriate excitation and emission lters for each dye. The co-localization efficiency was analyzed with Olympus FV10-ASW soware, in which the calculated Pearson's coefficient was above 0.90.</p><!><p>For the photostability test, the cells were imaged using a confocal microscope (Zeiss Laser Scanning Confocal Microscope; LSM7 DUO) and analyzed using ZEN 2009 soware (Carl Zeiss). Both TTNIR and BODIPY493/503 Green were excited at 488 nm for one-photon imaging (1% laser power). The scanning speed was 22.4 s per scan, and the repeated image scans were taken 40 times. The rst scan of both TTNIR and BODIPY493/ 503 Green was set to 100%, followed by which the pixel intensity values were averaged and plotted against the scan number. The resulting curve represents the bleaching rate.</p><!><p>H2DCF-DA was used as the ROS generation indicator. In the experiments, 10 mL of H2DCF-DA stock solution (1.0 mM) was added to 2 mL of TTNIR suspension, and white light (18 mW cm À2 ) was employed as the irradiation source. The emission of H2DCF-DA at 534 nm was recorded at various irradiation periods. HeLa cells were seeded in 96-well plates (Costar, IL, USA) at a density of 6000-8000 cells per well. Aer overnight cell culture, the medium in each well was replaced with 100 mL fresh medium containing different concentrations of TTNIR. Following 30 min incubation, the plates containing HeLa cells were exposed to white light (around 18 mW cm À2 ) for 30 min, and another array of plates with cells were kept in the dark as the control.</p><!><p>Two dishes of COS-7 cells were incubated with TTG and TTNIR for half an hour separately. Aer that the cells were washed with PBS 3 times, collected by adding trypsin, and centrifuged respectively. Then the cells were mixed together and incubated for 2 hours in another Petri dish with a cover glass. 10 g of polyethylene glycol 3400 was dissolved in 10 mL of Dulbecco's modied Eagle's medium (DMEM) without FBS. The mixed culture was overlaid for 5 min at 37 C with 2 mL PEG solution. Then the PEG solution was gradually diluted with DMEM in four steps at the interval of 2 min, by adding 0.5, 1, 2, and 4 mL DMEM, respectively, aer which the liquid was removed and replaced with DMEM.</p><!><p>There are no conicts to declare.</p>

Royal Society of Chemistry (RSC)

A Highly Stretchable and Robust Non-fluorinated Superhydrophobic Surface

Superhydrophobic surface simultaneously possessing exceptional stretchability, robustness, and non-fluorination is highly desirable in applications ranging from wearable devices to artificial skins. While conventional superhydrophobic surfaces typically feature stretchability, robustness, or non-fluorination individually, co-existence of all these features still remains a great challenge. Here we report a multi-performance superhydrophobic surface achieved through incorporating hydrophilic micro-sized particles with pre-stretched silicone elastomer. The commercial silicone elastomer (Ecoflex) endowed the resulting surface with high stretchability; the densely packed micro-sized particles in multi-layers contributed to the preservation of the large surface roughness even under large strains; and the physical encapsulation of the microparticles by silicone elastomer due to the capillary dragging effect and the chemical interaction between the hydrophilic silica and the elastomer gave rise to the robust and non-fluorinated superhydrophobicity. It was demonstrated that the as-prepared fluorine-free surface could preserve the superhydrophobicity under repeated stretching-relaxing cycles. Most importantly, the surface\xe2\x80\x99s superhydrophobicity can be well maintained after severe rubbing process, indicating wear-resistance. Our novel superhydrophobic surface integrating multiple key properties, i.e. stretchability, robustness, and non-fluorination, is expected to provide unique advantages for a wide range of applications in biomedicine, energy, and electronics.

a_highly_stretchable_and_robust_non-fluorinated_superhydrophobic_surface

Introduction<!>Results and Discussions<!>Conclusions<!>Preparation of silicone elastomer/silica microparticle composite surface<!>Water contact angle and sliding angel measurement<!>Material stretch apparatus<!>Surface structure characterization<!>Rheology of Ecoflex

<p>Superhydrophobic surfaces that show extremely strong water repellency have attracted wide interest due to their extensive applications, especially in the fields including self-cleaning1–2 and anti-fouling surfaces,3–5moisture-proof electronics,6 and drag-reduction for marine vessels,7–10 among others.11 While the existing superhydrophobic materials are generally realized by combinations of delicate microscale and nanoscale heterogeneous structures, the resulting superhydrophobicity is vulnerable to scraping or abrasion due to the destruction of the brittle structures.12, 13 A possible solution to the vulnerability is to use elastic substrates instead of rigid ones. As an example, micro-sized pyramid silicon arrays covered with gold nanoparticles were used as a template to fabricate superhydrophobic polydimethylsiloxane-based surfaces.14 However, despite all the efforts devoted in micro/nanofabrication of the molds, there remains a severe challenge related with faithful replication in nanoscale of the elastic materials, which tends to either break or deform at nanoscale during peeling off from the molds. Consequently, the failure in nanostructural replication has made the elastomer superhydrophobicity unsuccessful. On the other hand, reducing surface free energy has been used to achieve superhydrophobicity on elastomers. In this aspect, poly- or perfluoroalkyl surface treatments (also termed fluorination) are frequently implemented to substantially reduce the surface free energy.15–18 Nevertheless, the use of fluorination has raised some environmental concerns regarding their potential threats to human health.19, 20 For example, it has been reported that the exposure of human to these substances may increase the chances of many intractable diseases such as cancer, immune disorder, and hormonal disturbance.21 Consequently, there is a strong need to develop a stable superhydrophobic surface without fluorination.</p><p>While non-fluorinated and robust superhydrophobic surfaces are highly desirable, those integrating another remarkable feature, stretchability, are showing even greater promise in areas ranging from devices related to liquid motion manipulation22–24 to modern biomedical applications such as artificial skin25, 26 and wearable devices.27, 28 The true value of elastic superhydrophobic materials lies in that they preserve the superhydrophobicity when being stretched. To achieve this goal, researchers have devised a number of different strategies. For example, by enabling crumpling of graphene on an elastomer film of polyacrylate adhesive (VHB, 3M), the surface showed obvious superhydrophobicity with water contact angle as high as 152° in the crumpled state.29, 30 Unfortunately, the superhydrophobicity disappeared when the substrate was stretched (with contact angle decreased to 103° at ~80% strain), due to the sharp reduction of the surface roughness during graphene unfolding. More recently, another attempt has been made by spraying the solution of carbon nanofiber/paraffin/toluene and depositing the blend onto a natural rubber surface followed by solvent evaporation.31 The resulting superhydrophobicity persisted under high stretch, but this surface might be vulnerable against rubbing or abrasion because of the brittleness of individual carbon nanofibers that protruded out of the surface, and the lack of chemical bonding between carbon nanofiber and natural rubber.</p><p>Here we introduce a simple and environment-friendly non-fluorination method to prepare a highly stretchable and robust superhydrophobic surface by combining physical encapsulation and chemical bonding. The substrate material we chose was a commercial elastomer Ecoflex (Smooth-On, Inc.), which is a silicone rubber possessing exceptional stretchability and known non-toxicity.32 In particular, hydrophilic micro-sized silica particles were used to generate the superhydrophobic surface without the need of further fluorination. The as-prepared surfaces proved to be able to maintain superhydrophobicity after at least 1000 cycles of stretching-relaxing. Moreover, surface's superhydrophobicity could be well preserved after severe rubbing to the surfaces, which brings it exceptional value to practical applications.</p><!><p>Fig. 1 depicts the possible structural basis for the robustness of the superhydrophobicity on our prepared surface, contrasting with that on a traditional elastic superhydrophobic surface lack of robustness. For the latter one that is simply covered with a dense layer of hydrophobic micro/nano-scaled particles, the superhydrophobic property may be easily lost by either stretching or rubbing the surface (shown in Fig. 1a). By stretching the surface, the distance between the particles is enlarged; by rubbing the surface, the particles on the surface tend to be scratched off due to the lack of stable particle/rubber adhesion, further giving rise to enlarged particle distance. Both processes will result in the decrease of the surface roughness, and thus the loss of superhydrophobicity. For our superhydrophobic surface taking advantage of the combination of the physical encapsulation and chemical bonding, by contrast, the superphydrophobicity can be well-preserved under much larger strains or after severe rubbing (Fig. 1b) due to the ultra-high density, multi-layered micro-sized particles chemically bonded to the substrate. It is worth to mention that the strategy of fabricating particle-composited superhydrophobic surface has been reported before. For example, multiple-step chemical reactions were designed to functionalize nanoparticles with water-repellent fluorinated groups, followed by spraying or spin-coating organic solution of the nanoparticles on glass substrates to fabricate superhydrophobic surfaces.33, 34 The as-prepared superhydrophobic surface showed remarkable abrasion resistance. Unfortunately, the lack of deformable substrates prevented the superhydrophobic surfaces from being stretchable. Moreover, the sophisticated procedures for functionalization of the nanoparticles further increased complexity of the preparation.</p><p>Fig. 2a schematically illustrates the typical fabrication process of such a robust superhydrophobic surface. A completely cured silicone elastomer membrane was firstly biaxially stretched to typical strains in the range of 50–200%. The stretched membrane was then spin-coated with a thin layer of the oligomer (with a weight ratio of part A to part B at 1:1), followed by deposition of micro-sized hydrophilic silica particles using a dumping method. The excessive particles were removed by gently shaking the membrane and the resulting surface was placed in a 100°C oven for at least 10 h, allowing the oligomer to cure completely and silica microparticles fixation. After the entire procedure, a highly stretchable and robust superhydrophobic surface can be obtained. Of note, the pre-stretch of the membrane before spin-coating of the oligomer is crucial for the surface to preserve superhydrophobicity under subsequent high stretch. Also, the hydrophilicity of the silica microparticles and the heating process were both indispensable to the final superior performance of the surface.</p><p>A proposed mechanism is illustrated in Fig. 2b. Upon contact with the Ecoflex oligomer, the hydrophilic silica microparticles are rapidly encapsulated by the oligomer under capillary dragging, which is driven by the difference of the surface free energy between the silica microparticle and the silicone elastomer, Δγ. The viscosity η of the elastomer retards the encapsulation. By dimensional analysis, the relaxation time (τ) needed for the encapsulation process scales following the equation below, τ∼ηLΔγwhere L is the diameter of the silica microparticle. Specific to the experiment here, the silica microparticles used had a diameter distribution from 1 µm to 24 µm, with approximately 80% falling in the range of 2 µm to 5 µm (see Fig. S1). From the rheology measurement shown in Fig. S2, η was calculated to be around 5 Pa·s for freshly prepared silicone oligomer. Taking Δγ ~ 55 mN/m (surface free energy of the hydrophilic silica microparticles and the silicone oligomer is ~ 75 mN/m and 20 mN/m, respecitvely,35, 36) and L ~ 5 µm, it was derived that the relaxation time τ equaled to roughly 0.5 ms. Indeed, the result suggested that all silica microparticles were well encapsulated by the silicone oligomer almost immediately after deposition on the surface. Therefore, this physical encapsulation can firmly immobilize the microparticles on the surface of silicone elastomer. In addition, the high-temperature incubation during the elastomer curing process further facilitated the creation of chemical bonding between the silica microparticles and the silicone matrix. As shown in Fig. 1b, the reactive component in the silicone oligomer is polydimethylsiloxane. Upon heating to 100 °C, some Si-O groups in the backbone of polydimethylsiloxane break and bond with the hydroxyl groups hanging outside of silica microparticles, forming covalent bonding between the surface of the silica microparticle and silicone.37–39 In the meantime, the silicone oligomer is crosslinked by crosslinker through hydrosilylation reaction40 and firmly adheres to the pre-existing underlying elastomer substrate (see Fig. S3 for detailed reaction between silicone oligomer and silica microparticles, as well as crosslinking of silicone oligomer). The combined physical encapsulation and covalent chemical bonding of the silica microparticles with the elastomer matrix ensured the robustness of the fabricated stretchable superhydrophobic surface.</p><p>The encapsulation of the silica microparticles by the silicone elastomer was verified by the element distribution on the surface of silica microparticles before and after the wrapping process, using energy dissipative x-ray spectroscopy (EDX) element mapping method, as shown in Fig. 2c and 2d. While the silicon element could be detected on the silica microparticles in both situations (shown in blue), the carbon element only appeared on top of those after wrapping (shown in red). Since the elastomer layer that covered the outer surface of the silica microparticle was the only source of carbon signal gained during the survey, the presence of the carbon element on top of silica microparticles clearly indicated their wrapping by the elastomer. In addition, the encapsulation was also evidenced from appearance of the wrinkles on the otherwise smooth silica microparticles, as shown in the Fig. S4, where the red arrows denote the wrinkles formed during the curing process of the oligomer. This wrapping of elastomer on the surface of the silica microparticles simultaneously endowed the surface with roughness and low surface free energy, which led to the superhydrophobicity.</p><p>Fig. 3a shows the appearance of the pristine and superhydrophobic elastomer surfaces. The translucent membrane became totally opaque after silica microparticle deposition. This uniform opaque surface differed much from the surface prepared by depositing hydrophobic silica microparticles onto the same silicone oligomer, which could only result in scattered non-uniform islands due to the inability of the oligomer to wrap the hydrophobic silica microparticles that already have low surface free energy (Fig. S5a). Fig. 3b shows a scanning electron microscopy (SEM) image of the cross-section of the as-prepared elastomeric superhydrophobic material. An elastomer/silica microparticle composite layer was observed with a clear boundary to the underlying smooth elastomer substrate. The thickness of the composite layer was approximately 70 µm, which was formed under a spin-coating speed of 4500 rpm and silica microparticles were deposited onto the oligomer surface immediately after the spin-coating. Top-view SEM images in Fig. 3c and 3e also indicate the microscale structural changes before and after the superhydrophobic treatment. The pristine elastomer surface was smooth (Fig. 3c) and the water contact angle was approximately 99.6° (Fig. 3d), exhibiting intrinsic hydrophobicity. After superhydrophobic treatment, the surface became enormously rough at microscale, characterizing multi-layer hierarchical structures (Fig. 3e). Water contact angle on the resulting surface was 151.2° (Fig. 3f), much higher than that of 114.5° on the elastomer surface with hydrophobic silica microparticle deposition (Fig. S5b).</p><p>In fact, the thickness of the composite layer can be tuned by varying the spin-coating speed of the silicone oligomers (Fig. S6). The thicknesses of the composite layer corresponding to spin-coating speeds of 3000 rpm, 4500 rpm, and 6000 rpm, were 80 µm, 70 µm, and 60 µm, respectively. Despite of the difference in thickness, the resulting surfaces were all superhydrophobic, with the water contact angles on those surfaces being greater than 150°. However, the water sliding angles on these surfaces varied with the different thicknesses arisen from different spin-coating speeds. As shown in Fig. S6, the surface with a thickness of 70 µm prepared at 4500 rpm gave the lowest sliding angle of approximately 10°. It was believed that when the spin-coating speed was exceedingly slow (e.g. 3000 rpm), the thickness of the silicone oligomer layer was much larger than the diameter of the silica microparticles. During deposition, the silica microparticles (especially those small in size) were mostly immersed into the viscous liquid. As a result, the hierarchical surface roughness due to the different sizes of the silica microparticles could not be fully realized with most parts of the surface showing only bulk microscale roughness. Consequently, water droplet was prone to get pinned as it slid upon tilting. On the other hand, although for the surface obtained under a high spin-coating speed (e.g. 6000 rpm), the hierarchical structures could still be formed, since the thickness of the silicone oligomer layer is small, it was hard to achieve a homogeneous hierarchical structure on the surface (i.e., generation of clusters), leaving surface defected. In this situation, the water sliding angle was also increased. Surfaces prepared at 4500 rpm with a medium thickness of silicone oligomer showed superiority on both the generation of hierarchical structure and homogeneous structural distribution, resulting in the smallest water sliding angle.</p><p>In addition to the spin-coating speed, pre-curing time of the silicone oligomer prior to silica microparticle deposition is another parameter that would affect the ultimate performance of the surfaces. As shown in Fig. S7, with the pre-curing time varied from 0 min to 80 min (i.e. maximal time before the silicone Ecoflex oligomer completely cures) at room temperature, the resulting thickness of the composite layer monotonically decreased from approximately 70 µm to 10 µm. This reduction in thickness was believed to correlate with the increased viscosity of the oligomer and thus the wrapping hysteresis when pre-curing time was increased. Similar to the effect of spin-coating speed, the resulting change in the thickness did not significantly alter the water contact angles on these surfaces, with all being around 150° due to the large roughness and complete coverage of silica microparticles by silicone elastomer. Despite of the similarity of the water contact angles, sliding angles on these surfaces were quite distinct. From pre-curing time of 0 min to 20 min, the sliding angle increased from approximately 10° to nearly 30°, which was caused by the reduction of hierarchical assembly of the structures as abovementioned. Consequently, the spin-coating speed of 4500 rpm and 0 min pre-curing that gave rise to the largest water contact angle and smallest water sliding angle were determined to be the optimal conditions for preparing superhydrophobic elastomeric composite material, which were used for surface fabrication in following experiments unless otherwise noted.</p><p>As practical applications call for surface's superhydrophobicity to remain stable under large strains, we tested the stability of the superhydrophobicity of our surface in different stretching states. Fig. 4a and 4b show the water droplets sliding on the relaxed and stretched surfaces, respectively. When an 8-µL water droplet was dispensed onto the slightly tilted surface with zero strain, the droplet could slide off the surface with ease (Movie S1). After receiving a uniaxial strain of 200%, water droplet with the same volume could still slide off the surface at the same level of tilting (Movie S2). Interestingly, the sliding speed of water droplet on the stretched surface was even larger than that on the relaxed surface, with further moving distance within 133 ms on the superhydrophobic surface at 200% strain (approximate 17.6 mm) than that at 0% strain (approximate 10.8 mm). In fact, this superhydrophobicity could be maintained at even more stretched states of beyond 200% strain. As shown in Fig. 4c, the water contact angles maintained at around 150º at the strain < 500%, with inset images showing the uniaxial stretch under different strains. Beyond the strain of 500%, the superhydrophobic surface would break abruptly. However, the large water contact angle as well as the highly roughed microstructures were preserved on the broken surface after it was relaxed (Fig. S8). The large surface roughness stemming from deposition of silica microparticles on a pre-stretch surface contributed to this stable superhydrophobicity.</p><p>In addition, from Fig. 4c, we can also derive that the droplet moving speed on the surfaces with the same tilted angle of 13° but with different strains increased with increasing strains, which was consist with the results on the movement of water droplet in Fig. 4a and 4b. This increase of the moving speed with increasing strain could have arisen from the more obvious anisotropy of the surface due to the uniaxial stretch (Fig. S9), where anisotropic surfaces are more beneficial for directional movement of the droplets along the aligned direction comparing with isotropic surfaces.41, 42 Fig. 4d further demonstrated the stability of the superhydrophobicity under cyclic stretching between 0% and 200%. The superhydrophobicity of the surface could be well maintained after 1000 cycles. The inset images of water droplet on the surface before any stretch and after 1000 cycles of stretch-relax showed little difference with regard to the contact angle. Moreover, after 1000 cycles of stretching-relaxing at a strain of 200%, the superhydrophobic elastomer could still give a water sliding angle of 8°, similar to the value (9–10°) for that before stretching. This unchanged sliding angle was consistent with the well-preserved rough microstructures of the surface after cyclic stretching-relaxing, as shown in Fig. S10. These results indicated that the superhydrophobic surface was able to maintain stability in highly stretch states, which is promising in applications requiring superhydrophobicity under different strains. Of note, the method we show here can be extended to preparation of highly stretchable superhydrophobic surfaces for a broad range of elastomeric materials. It was also believed that with larger pre-stretch ratio of the elastomer substrate before the spin-coating process, the resulting surface should be able to maintain its superhydrophobicity at even larger strains. In other words, in our approach the stretch limit for maintain stable superhydrophobicity on a given elastomer surface is only limited by its own stretchability.</p><p>While stable superhydrophobicity of surfaces upon stretching are highly desired in certain applications, robustness of the superhydrophobicity under rubbing processes is another crucial advantage in applications related with surface wettability.42 To assess the robustness of our superhydrophobic surface, droplet adhesion or moving behavior on the superhydrophobic surface before and after receiving rubbing was recorded (Movie S3). As shown in Fig. 5a, the as prepared silicone elastomer-based superhydrophobic surface (white) was first wrapped around a finger. A water droplet of 8-µL staining red dye was dispensed on the superhydrophobic elastomeric surface. The water droplet slid down off the surface within 170 ms, as indicated by the yellow arrows showing the positions of the droplet at different time points. This non-adhesive down-moving behavior is typical for a superhydrophobic surface. After the initial test, the surface was rubbed with two fingers and then wrapped on the finger again, shown in Fig. 5b. Again, stained water droplets were deposited on the post-rubbing surface. As shown in Fig. 5c, the water droplets slid down off the surface fluently (170 ms) in the same manner with those on the surface before rubbing. Moreover, we also measured the water contact angle and the water sliding angle of the post-rubbing surface. As shown in Fig. S11, a contact angle of 148.5 ±0.4° and a sliding angle of approximately 11° could be achieved on the surface, suggesting the robustness of the surface superhydrophobicity.</p><p>In addition, to indirectly prove the formation of the covalent bonding between the hydrophilic silica particles and the silicone elastomer in our surface as well as its critical role to the robustness of the superhydrophobicity, we compared the droplet sliding behavior on a superhydrophobic surface with the same physical encapsulation but without the chemical bonding. The procedure for preparing such a surface was similar to the strategy used to prepare robust superhydrophobic surfaces with the difference that the curing process of the silicone oligomer after deposition of micro-sized silica particles was conducted at ambient temperature, making sure little or no chemical bonding was formed.37, 39 As shown in Fig. 5d–f, an 8-µL stained water droplet could slide down off the inclined superhydrophobic surface due to the presence of surface roughness; however, after a similar rubbing process, water droplets readily adhered to the surface, indicating damage to its superhydrophobicity. Indeed, changes of the microstructures of the surface before and after the rubbing process were observed (Fig. S12), showing loss of the silica microparticles from the surface. The distinct result from the comparison clearly shows that the chemical bonding between the silica particles and the silicone elastomer is critical in making the superhydrophobicity of the surface durable by preventing the silica microparticles from detaching the surface during harsh treatments such as rubbing.</p><p>Moreover, the robustness of the chemically bonded superhydrophobic surface was further verified by a sand paper abrasion test (Fig. S13), where the superhydrophobic surface was pressed against a 500-grit sandpaper surface by a 100-g weight placed on top. A horizontal force was used to draw the copper wire tethered to the weight, making the superhydrophobic surface move steadily on the sandpaper for 10 cm. Water droplets deposited on the resulting surface could still easily slide off the surface afterwards (Fig. S13a and 13b). Upon further observation of the fine structures of the superhydrophobic surface after the abrasion against sandpaper, there were only negligible differences comparing with that before abrasion treatment (Fig. S13c and 13d). Of note, the water contact angle and microstructure of the superhydrophobic elastomer surface were able to maintain even after 10 cycles of such test (Fig. S14). The preserved behavior of water droplets on the superhydrophobic surface with unchanged microstructures indicated robust superhydrophobicity against abrasion. Therefore, the combination of physical encapsulation and chemical bonding between silica microparticles and silicone elastomer should have accounted for the robustness of the superhydrophobicity.</p><!><p>In conclusion, we demonstrated a simple and environment-friendly method to fabricate superhydrophobic, non-fluorinated composite elastomer surfaces with multiple exceptional properties, such as stability under extensive and cyclic stretching, as well as robustness after severe rubbing and abrasion. The combination of the physical encapsulation and chemical bonding attributed to this superior performance. Difference between the relatively low surface tension of silicone oligomers and the high surface free energy of the silica microparticles was the driving force for the physical encapsulation; reaction between hydroxyl groups on the surface of silica microparticles and the Si-O groups at an elevated temperature resulted in the formation of covalent chemical bonding between silica and silicone elastomer. This highly stretchable and robust non-fluorinated superhydrophobic surface is expected to provide unique advantages for a wide range of applications in biomedicine, energy management, and electronics. In addition, the preparation method is simple and readily scalable, and can be applied as a general paradigm for fabricating composite elastomeric materials with versatile surface functions. Functional micro/nano-objects that are intrinsically rich of hydroxyl groups or can be easily treated to possess hydroxyl groups will be able to serve as the embedding materials for the fabrication of superhydrophobic surfaces while maintaining their own functionalities, such as magnetic ferroferric oxide particles.44</p><!><p>The composite surface was fabricated in two steps. First, commercial silicone elastomer Ecoflex sheets (with typical thickness of 1 mm) was prepared using a mixture of part A and part B at a weight ratio of 1:1 and allowed to cure completely at room temperature for 24 h. Second, the elastomer sheet was pre-stretched at a strain of 200% using a fixture, and silicone oligomers with same mixing ratio was deposited on the stretched silicone elastomer sheet by spin-coating. The spin speed was varied. After spin-coating, the substrate was maintained horizontally for a certain period of time (from 0 min to 80 min). Then a large amount of silica microparticles was directly deposited onto the substrate. After 1 min, excessive microparticles were physically shaken off. The surface was then incubated in an oven at 100 °C for 10 h.</p><!><p>The contact angle was measured and analyzed by OCA 20 (Dataphysics). A 5-µL water droplet was carefully deposited on the composite elastomer surface. A digital camera (Canon 7D) was used to take close-up photos in parallel to the elastomer surface. Optical fiber light source was applied from the background to provide a bright field. At least five points on a single surface were measured to obtain mean value of contact angle. For testing the sliding angle, the sample were fixed on a custom-made device. The device was gradually raised from horizontal plane to a certain tilting degree. The lowest tilting angle that allowed an 8-µL water droplet to slide from the surface was recorded as the slide angle.</p><!><p>The as-prepared composite elastomer was clamped at both ends on an Instron Materials testing machine (Instron, 5960 dual column). The material was then subjected to single stretch at different strains and cyclic stretch with strain = 200%. For the cyclic stretching, different samples were stretched from 100 to 1000 cycles at an increment of 100 cycles. Upon finishing the cyclic stretch test, the composite elastomer was released for contact angle measurement.</p><!><p>All the microstructures of the as-prepared composite surfaces and the silica microparticles were imaged on a Field-Emission SEM (FESEM, Zeiss ultra 55) after coated with 10 nm-thickness of Pt/Pd conductive layer on a sputter coater (EMS 300T D Dual Head Sputter Coater). The EDX mapping was conducted on the same SEM.</p><!><p>The viscosity of the silicone oligomers with prolonged time of curing was monitored using a Rheology (Anton-Paar MCR501) under constant shear mode.</p>

PubMed Author Manuscript

Fluorofoldamer-Based Salt-and Proton-Rejecting Artificial Water Channels for Ultrafast Water Transport

We report here the best artificial water channel ever reported in terms of structural robustness, facile synthesis and water transport property.Here, we report on a novel class of fluorofoldamer-based artificial water channels (AWCs) that combines excellent water over ion selectivity with extraordinarily high water transport efficiency and structural simplicity and robustness. These AWCs were produced by a facile one-pot copolymerization reaction under mild conditions. Among these channels, the best-performing channel (AWC 1) is a n-C8H17-decorated foldamer nanotube with an average channel length of 2.8 nm and a pore diameter of 5.2 Å. AWC 1 demonstrates an ultrafast water conduction rate of 1.4 × 10 10 H2O/s per channel, outperforming the archetypal biological water channel, aquaporin 1, by 27%, while excluding salts (i.e., NaCl and KCl) and protons. Unique to this class of channels, the inwardly facing C(sp2)-F moieties are proposed as being critical to enabling the ultrafast and superselective water transport properties observed.

fluorofoldamer-based_salt-and_proton-rejecting_artificial_water_channels_for_ultrafast_water_transpo

INTRODUCTION<!>Molecular design<!>AWC 1 exhibits the best water transport performance<!>Impact of channel length on water permeability<!>Impact of side chain type on water permeability<!>High salt and proton rejection capacity of AWC 1<!>Comparison with two high-performance AWCs<!>Critical roles played by fluorine atoms<!>Molecular dynamics simulation<!>CONCLUSIONS<!>Materials<!>Water transport study<!>The HPTS assay for anion transport<!>IF = [(Ft -F0)/(F1 -F0)]<!>Activation energy measurements<!>Ln(k) = Ln(A) -Ea/(RT)<!>Molecular dynamics simulations

<p>Scarcity of clean water is one of the critical grand challenges facing humanity that currently affects over 4 billion people worldwide (1,2). An important state-of-the-art technology for clean water production and wastewater reuse is reverse-osmosis (RO) membrane desalination (3). The key to RO desalination is precise control over transient or fixed sub-nanometer scale passages across the membrane that only allow water molecules to pass through while excluding other solutes like salt ions (4).</p><p>In Nature, living organisms regulate transmembrane water flow by membrane-embedded water channels, viz. aquaporins (AQPs). These proteinic channels facilitate superfast water translocation and at the same time completely reject salts and even protons (5,6). For instance, AqpZ, isolated from E. coli, features a water transport rate of ~ 6 × 10 9 H2O/s (7,8). The other type of AQPs, AQP1 that is present in specific human cells can transport ~ 1.1 × 10 10 H2O/s ( 9), yet with remarkably high water to monovalent ion selectivity over 10 9 . Integration of such water-permeating and salt-rejecting AQPs into polymer-based membranes represents an emerging approach for developing the next generation of water desalination and purification technology (10)(11)(12). Nevertheless, membrane proteins like AQPs usually suffer from high production costs, challenges with scalability, and questions about structural stability in abiotic environments (13), making them less ideal for large-scale industrial applications.</p><p>Motivated by the superior performance of natural AQPs, researchers have expanded extensive effort in developing artificial water channels (AWCs) with simpler structures yet comparable or even exceeding water transport capabilities (14)(15)(16). In 2007, Percec and co-workers reported the pioneering work in this field, wherein dendritic dipeptides were employed for the construction of AWCs in lipid membrane (17). Thereafter, various types of unimolecular or self-assembled AWCs have been designed and characterized, including imidazole-quartet (18)(19)(20)(21)(22)(23), pillar [n]arenes (24)(25)(26)(27), aromatic macrocycles (28), carbon nanotube porins (CNTPs) (29,30), porous organic cages (31), helically folded polymeric nanotubes (32)(33)(34), and hydrophilic hydroxyl assemblies (35). The collective conclusion states that water transport efficiency and selectivity highly depend on the geometry and surface chemistry of the channel interior lumen, in which channel-water and water-water interactions occurs, primarily via H-bonds (9,33,34). However, concurrently achieving high single-channel water permeability and high transport selectivity (e.g., rejection of salts and protons) in a single AWC still remains a daunting task to date that has been addressed in only a few studies (26,34).</p><p>Here we report on such a high-performance salt-and proton-rejecting AWC system that has a 5.2 Å-diameter cavity and transports water at a remarkable rate of ~1.3 times that of AQP1, outperforming by at least a factor of 4 all other hitherto known salt-rejecting AWCs , except for one very recent example (34).</p><!><p>Although the lone pair donation from fluorine, being the most electronegative element in the periodic table, is significantly suppressed, making it a poor H-bond acceptor (36), early studies have established the ability of C(sp2)-F to form weak intramolecular H-bonds in foldamer structures (37)(38)(39). Further, fluorine atoms may differ considerably from other H-bond-forming groups in determining the foldamer channel construct and guest binding behaviors (40) by altering the interior pore size and wall smoothness, channel backbone distortion, intermolecular host-guest H-bond interactions, etc. With these concepts in mind, we decided to explore fluorofoldamer-based polymeric hollow channels, having inward-facing fluorine atoms decorating the channel lumen, as possible AWCs.</p><!><p>Screening a matrix of such AWCs, combinatorically derived from different reaction conditions and monomer structures, culminated in a discovery of the best-performing AWC, water channel 1. AWC 1 is found to conduct water at an ultrafast rate of 1.4 × 10 10 H2O/s across lipid bilayer membrane, a value that is about 1.3 times higher than that of AQP1 and two times higher than its methoxy-containing analogous channel 1-OMe (see later discussions). Further, the readily synthesized 1 also demonstrates near-perfect salt (NaCl and KCl) and proton rejection, making it an excellent replacement of natural AQPs for possible industrial uses in fabricating next-generation of AWC-based RO membrane for seawater desalination or for use in therapeutics (26).</p><p>Synthesis of 1 was carried out by following a previously reported protocol (41). Briefly, a facile one-pot copolymerization reaction between diamine monomer A1 (with n-C8H17 side chains) and fluoro-containing diacid B using HBTU as the coupling reagent readily produced an off-white powdery product 1 with ~80% isolated yield (Fig. 1a). Apart from extensive π-π stacking, intramolecular H-bonds</p><p>are also expected to stabilize the polymeric product in a helically folded configuration (42). Molecular dynamics (MD) simulations of the pore scaffold shows that the optimized structure exhibits expected helical tubular shape with The inner pore diameter is ~5.2 Å after subtracting the van der Waals radii of the interior atoms. This pore diameter is larger than a water molecule (2.8 Å), but smaller than first-shell hydrated Na + or K + ions. The average molecular weight of 1 was measured to be 13.9 kDa using gel permeation chromatography (GPC). A NMR-based method was also applied to determine the molecular weight, in which a chiral group was introduced at the amine end of 1 as an internal NMR standard (Scheme S3). Based on the area integration ratio of specific 1 H signals (Fig. S1), molecular weight of 1 was determined as 15.4 kDa (Table S2), agreeing well with the GPC-derived value. Using the simulated pore structure as the guide (three AB units per helical turn, MW of AB unit = 616.7 Da; see Fig. 1b and Supplementary Table S1), 1 contains 25 AB units in average, measuring at 2.8 nm in average nanotubular length that is dimensionally comparable to the thickness of typical lipid bilayer membranes (e.g., 2.7 nm for DOPC) (43). In addition, 1 also displays a characteristic mass pattern with a repeating unit of 617 Da in the MALDI-TOF spectrum (Fig. S2).</p><p>The unique structural features of 1, including (i) appropriate inner pore diameter intermediate between a water molecule and hydrated ions (e.g., Na + ), (ii) membrane-spanning channel length of 2.8 nm, and (iii) special lumen surface chemistry with H-bond donors/acceptors and dipolar C(sp2)-F moieties, lays the structural basis for its water transport property. Stopped flow light-scattering method was employed to quantify the water transport efficiency, using large unilamellar vesicle (LUVs, 120 nm diameter, Fig. 1c) with channel 1 pre-inserted in the LUV wall (24,25). Under the shrinkage mode, LUVs were exposed to hypertonic buffer solution containing 200 mM sucrose, which induces water efflux and vesicle shrinking. The time-dependent variation of the light-scattering intensity was then captured and analyzed (Fig. 1d), from which water transport rate can be reliably determined. As shown in Fig. 1e, water permeability of 1 was largely independent of the lipid to channel molar ratio (mLCR), and the profile peaks at 12000:1, giving water permeability PW of (41.2 ± 2.1) × 10 -14 cm 3 /s. With a channel insertion efficiency of 87.9% at this mLCR (Table S3), water permeability translates into a single-channel water transport rate of (1.4 ± 0.07) × 10 10 H2O/s, which is 133% and 27% faster than the biological AqpZ and AQP1 water channels, respectively (44). The value of 1.4 × 10 10 H2O/s becomes 0.78 × 10 10 H2O/s using the new equation for Pf correction (44).</p><p>The water permeability of 1-reconstituted LUVs at different temperatures (6-25 °C) were measured, from which its activation energy Ea is calculated as 7.1 ± 1.2 Kcal mol -1 using the Arrhenius Equation (Fig. 1f). It is higher than that of the AQPs (~ 5 Kcal mol -1 ), but much lower than that from the blank DOPC LUV (12.3 ± 0.2 Kcal mol -1 ). In view of the superior water conduction rate of 1 compared to AQPs, we assume that low activation energy might not be a necessary feature for highly permeable AWCs, likely because the transport mechanisms differ from that of AQPs in Nature, as proposed before (33,34).</p><!><p>Following identical synthetic protocols (41), other amide coupling reagents (HATU, BOP and TBTU) produce the same A1B type channels with NMR-derived molecular weights of 20.1, 19.9 and 13.1 kDa (Table S1) that correspond to channel lengths of 4.0, 4.0 and 2.6 nm, respectively. As summarized in Fig. S3, their water transport rates were all found to be lower that of channel 1 (MW = 13.9 kDa, 2.8 nm). More specifically, at the mLCR of 12000:1 and compared to 1 of 2.8 nm (Pw = 41.2 × 10 -14 cm 3 /s), A1B type channels produced using HATU (4.0 nm), BOP (4.0 nm) and TBTU (2.6 nm) show much lower Pw values of 21.9 × 10 -14 , 22.0 × 10 -14 and 32.4 × 10 -14 cm 3 /s, respectively.</p><!><p>To examine the impact of channel side chains on water transport property, diamine monomers A2 and A3 (carrying n-(CH2CH2O)3CH3 and i C4H9 side chains respectively) were also employed in the HBTU-facilitated copolymerization reaction. Their corresponding products (A2B)n and (A3B)n were named channels 2 and 3, respectively. Monomers A1 and A2 were further pre-mixed in 1:1 ratio, and then stoichiometrically reacted with B to produce mixed copolymers (A1BA2B)n (e.g., channel 4). From their NMR-derived molecular weights (Supplementary Table 2), the channel tubular lengths can be estimated to be 3.1, 2.9 and 4.1 nm for 2, 3 and 4, respectively. At the mLCR of 12000:1, 2 -4 show much lower Pw values of 2.8 × 10 -14 , 20.2 × 10 -14 and 29.4 × 10 -14 cm 3 /s, respectively. The comparative data among 1-3 indicates the importance of channel side chain lipophilicity on water transport efficiency, and clearly the linear n-C8H17 represents the best performer for the fluorofoldamer-based AWC scaffold.</p><!><p>Besides ultrafast water conduction, the other major challenges for AWCs in mimicking AQP performance are to achieve complete rejection of salts and protons. To this end, we firstly compared the osmotic water permeability (Pf in cm/s) values of 1 under three hypertonic conditions (300 mM sucrose, 150 mM NaCl, or 150 mM KCl, Fig. 2a). Since large sucrose molecules are not able to permeate through channel 1, the reflection coefficient, defined as Pf(MCl)/Pf(sucrose) where M + = Na + or K + , was used to approximately gauge the transport of salt ions. The well-established dimeric cation-transporting channel gramicidin A (gA) was employed as the positive control, which shows expected reflection coefficients of 0.53 ± 0.02 and 0.07 ± 0.001 for NaCl and KCl, arising from its high permeability to both Na + and K + ions. In contrast, the reflection coefficients of 1 were calculated to be 1.02 ± 0.01 for Na + and 1.05 ± 0.01 for K + at 12000:1 mLCR, confirming the inability of 1 to transport either cation across the membrane and its near-perfect salt rejection property (19). The rejection of Na + and K + cations was further validated by the fluorescence-based HPTS assay, with pH-sensitive HPTS dye molecules entrapped in the LUVs (Fig. 2b). The intravesicular region is set pH 7, whereas the extravesicular environment is maintained at the same pH but with 200 mM M2SO4 (M = Na or K). Under this high salt gradient, H + /M + antiport will increase the pH of the intravesicular region and hence enhance the HPTS fluorescence intensity. As shown in Fig. 2b, 1 at 1 µM was found non-responsive towards Na + or K + gradient, affirming the impermeability of neither cation through the inner pore of 1. On the contrary, gA at the identical channel concentration (1 µM) could efficiently transport Na + (291%) and K + (343%) cations. Such observation is in excellent agreement with the reflection coefficient results described earlier, both confirming the inability of 1 to transport cations.</p><p>The anion transport ability of 1 was further examined by using Clˉ-sensitive SPQ dye molecules entrapped in LUVs (Fig. 2d). As expected, 1 at 1 µM displayed similar SPQ quenching as background (9%), whereas the well-established chloride transporter L8 (45) at 6 µM (corresponding to 1 µM channel concentration) displayed significant decrease (45%) in the SPQ fluorescence intensity. In another set of LUV-based experiments where intravesicular region has 100 mM NaCl at pH 7 and extravesicular region has 67 mM Na2SO4 at pH 8 (Fig. S5), gA (cation channel, 1 µM), FCCP (proton carrier, 1 µM) and L8 (anion channel, 1.3 µM) induce HPTS fluorescence increase of 56%, 22% and 138% respectively. In sharp contrast, 1 at 1 µM does not cause any fluorescence change.</p><p>Proton translocation was probed using the pH gradient set across the membrane (Fig. 2c). Serving such a purpose, the intra-LUV region contains 100 mM NaCl at pH 7, whereas the extravesicular region was set pH 8 with 100 mM NaCl or KCl. If 1 is able to transport protons, the proton efflux (coupled with passive diffusion of cations or anions for charge neutralization) will induce significant pH increase in the intravesicular region, and dramatic change in HPTS fluorescence intensity will emerge. Experimentally, no fluorescence change was observed at all after addition of 1 (1 µM), suggesting negligible transport of protons. Using a conservative approach (for details, see the Supplementary Section of "Estimation of Proton Transport Rate"), the proton transport rate of 1 is estimated to be less than 0.25 proton/s.</p><p>Stopped-flow fluorescence analysis was further applied to quantitatively measure chloride permeability through DOPC membrane in the absence and presence of 1 (Fig. S6) (34,46). Based on the determined single-channel Clpermeability PCl of (1.7 ± 07) × 10 -20 cm 3 /s, the water-to-Clpermselectivity (e.g., Pw/PCl) for 1 was calculated to be (2.5 ± 1.2) × 10 7 . Since NaCl permeability is limited by the Na + ions in actual desalination processes (46), 2.5 × 10 7 represents a conservative estimate of the water-to-NaCl permselectivity for 1. As compiled in Fig. 3a, this value exceeds the permeability-selective trade-off trendline of current desalination membranes (11,47) by a factor of ~10 2 , signifying good potential for developing novel AWC-based desalination membrane that incorporates or is made of 1.</p><!><p>As summarized in Fig. 3b, currently there are only two water-transporting systems having higher water conduction rates than both AQP1 (1.1 × 10 10 H2O/s) (9) and AWC 1 (1.4 × 10 10 H2O/s), i.e., the relatively low selectivity CNT porin (2.3 × 10 10 H2O/s) (30) and the highly selective AWC 4-LA (2.7 × 10 10 H2O/s) (34). But it is worth emphasizing that while the water-transporting CNT porin also conducts ions and protons (30), channel 4-LA requires additional lipid anchors (LA) installed at the helical ends to orient the channel's alignment to achieve the ultrafast water conduction (34). Without such LA modifications, it's water transport rate drastically drops by 75% to ~ 0.6 × 10 10 H2O/s (34), a value that is ~43% capacity of 1. Further, it is possible that the LA-enhanced water transport property might deteriorate over time or be altered by the complex environment of a water purification membrane. All these make both CNTP and 4-LA potentially less competitive for fabricating practical AWC-based biomimetic water purification membranes than 1 developed in this work (12,48).</p><!><p>To demonstrate the crucial role of C(sp2)-F moieties in determining the water transport property of channel 1, we compared it with the recently reported analogous channel denoted as 1-OMe of 3.0 nm in height, which differs from 1 in that 1-OMe contains methoxy groups in the positions of F-atoms of 1 (34). As a result of bulky hydrophobic methyl groups helically arranged around the pore interior of 1-OMe, its helical backbone is slightly less curved than that of 1 having its pore surface decorated by F-atoms. Consequently, the pore diameter of 1-OMe is enlarged to 6.5 Å across, which is larger than 1 (5.2 Å across). Under the identical conditions, the water transport rate of 1-OMe was determined at ~ 5 × 10 9 H2O/s (34), which is 36% that of 1. Furthermore, unlike 1 with excellent ion-rejection capability, 1-OMe was permeable to anions (41). Therefore, we speculate that the superior water transport properties of 1 should arise from a collection of influencing factors induced by the inward-facing C(sp2)-F moieties, including the smaller atomic size, weak H-bond acceptor ability, dipolar bond characteristics and good hydrophobicity. Further investigative efforts to decipher these factors are currently underway in our laboratory.</p><!><p>To provide a molecular level explanation of transmembrane water transport through the pore of 1 embedded in POPC lipid bilayer membrane, we performed 800 ns long all-atom molecular dynamics (MD) simulations, (Fig. 3c and Supplementary Video 1). To maintain QM-derived diameter of 1 at the MD level, we used the RMSD colvar module of NAMD during the course of MD simulation. Fig. S8 shows the RMSD of 1 as a function simulation time. As the simulation begins, water molecules rapidly start permeating across the lipid bilayer through 1 (Fig. 3d). A linear fit to the water permeation vs simulation time (excluding first 200 ns) yields a permeation rate of ~3 water molecules/ns, which is higher than 1.2 water molecules/ns for AQP1. At any given instant of time, a water cluster typically having 30-50 water molecules resides inside the channel, with a mean of 40.5 water molecules (Fig. 3e). Among them, 40.7% or 16.5 water molecules are considered as proton wire breakers (Fig. 3f), which were described and defined in our recent study (34). Interacting with the neighboring water molecules via zero or just one H-bond, or two H-bonds solely via only O-atoms or only H-atoms (Fig. 3g), these breakers prevent forming a continuously H-bonded channelspanning water chain through which protons hop via the Grotthuss mechanism. Interestingly, the breaker type involving the formation of two H-bonds with the adjacent water molecules using only H-atoms is also observed in the NPA motif of AQPs (49). The existence of these proton wire breakers accounts for low proton permeability of 1.</p><p>Due to the narrow pore, each water molecule forms 1.94 H-bonds with other water molecules inside the channel (Fig. S9a,b) and 0.79 H-bonds with the channel wall (Fig. S9c), leading to a total of 2.73 H-bonds per water molecule. Taking 4 H-bonds per water molecule (EH-bond = 5.1 Kcal/mol) in bulk water (50,51), the activation energy for water entry into 1 can be estimated to be 6.5 Kcal/mol, which is consistent with the experimentally determined value of 7.1 Kcal/mol (Fig. 1f). The fact that 1 has a higher activation energy but transports water faster than AQP1 can be largely attributed to its larger pore diameter of 5.2 Å vs ~ 2.8 Å opening in the central channel of AQP1 as well as the more than one water wire molecule occupying the pore lumen that differs from the single file transport seen in AQP1 (34).</p><!><p>In summary, we have demonstrated ultrahigh water transport efficiency and excellent selectivity of a novel class of fluorofoldamer-based artificial water channels. Produced by facile one-pot copolymerization reaction with good yields, the best-performing water channel 1 of 2.8 nm in average channel length shows a remarkable water conduction rate of 1.4 × 10 10 H2O/s and near-perfect rejection of salt ions (Na + , K + , Cl -) and protons. This work uncovers the positive effects of introducing C(sp2)-F moieties on the inner rim of foldamer-based water channel pores, providing a new dimension of channel design principles. This, we believe, will stimulate further development towards the next-generation of membrane technologies for water desalination, nano-filtration and medical dialysis applications.</p><!><p>All reagents were obtained from commercial suppliers and used as received unless otherwise noted. Aqueous solutions were prepared from MilliQ water. Egg yolk L-α-phosphatidylcholine (EYPC) and 1,2-dioleoyl-sn-glycero-3-phosphocholine lipid (DOPC) were obtained from Avanti Polar Lipids. HEPES, HPTS, SPQ and FCCP refer to 4-(2-hydroxyethyl)-1-piperazine-ethane sulfonic acid, 8-hydroxypyrene-1,3,6-trisulfonic acid, 6-methoxy-N-(3-sulfopropyl)quinolinium, carbonyl cyanide-p-trifluoromethoxyphenylhydrazone, respectively.</p><!><p>In a 2 mL microcentrifuge tube, 6 mg DOPC (0.24 mL, 25 mg/mL in CHCl3, Avanti Polar Lipids, USA) and water channel compound (dissolved in CHCl3) were mixed at different molar ratios (4000:1 to 15000:1). The solvent was slowly removed by N2 flow and the resulting thin film was dried under high vacuum overnight. 1 mL HEPES buffer (10 mM HEPES, 100 mM NaCl, pH = 7.0) was then added into each tube for lipid hydration. In order to maximize incorporation of channel molecules into the lipid bilayer, each microtube was vortexed for 30 s and sonicated for 150 s (37 kHz, power 100, 70 °C) for 10 cycles. A glass spatula was used to scratch down all the lipid residues from the microtube wall to minimize lipid loss and maximize channel incorporation whenever necessary. The lipid/channel mixture was then subjected to 10 freeze-thaw cycles (freezing in liquid N2 for 1 min and heating at 55 °C in water bath for 2.5 min). The mixture was then extruded through polycarbonate membrane (0.1 μm) at 80 °C for 15 times to give LUVs at 6 mg mL -1 lipid concentration. For stopped-flow experiments, this LUV solution was diluted to 1 mg mL -1 with buffer (10 mM HEPES, 100 mM NaCl, pH = 7.0). The LUVs were then exposed to a hypertonic solution (200 mM sucrose, 10 mM HEPES, 100 mM NaCl, pH = 7.0). During stopped-flow experiment, the abrupt decrease in vesicle size was expected due to transport of water to the extravesicular pool and this event leads to increase in the light scattering intensity of 90° angle according to the Rayleigh-Gans theory. The changes of light scattering intensity caused by vesicle shrinkage were recorded at a wavelength of 577 nm and all these plots were fitted in the following form of single exponential function to give rate constant (k) value using the equation shown below: y = Aexp(-kt) + y0 where y is change in the light scattering, k is the exponential coefficient of the change in the light scattering and t is time.</p><p>With the assumption that change in the light scattering intensity is proportional to the change in the vesicle volume (ΔV/V0) based on the Boyle-van't Hoff law, the osmotic permeability (Pf) in the unit of cm/s was commonly calculated as follow:</p><p>where k is the exponential coefficient of the change in the light scattering ; S and V0 are the initial surface area and volume of the vesicles, respectively; Vw is the molar volume of water, and Δosm is the osmolarity difference. The size of LUVs was determined by dynamic light scattering after 10 times dilution of the aforementioned LUVs solution (i.e., 1 mg mL -1 ) with buffer (10 mM HEPES, 100 mM NaCl, pH = 7.0).</p><p>Following the new approach proposed by Horner and co-workers (44), the water permeability can be alternatively calculated using the new equation shown below: Pf = k x (Cin,o + Cout)/ ((2Cout 2 ) x ((S/V0) x VW)) Where Cin,0 and Cout are the osmolytes concentration inside at t = 0 s and outside of vesicles, respectively.</p><p>To calculate the true water permeability (PW in the unit of cm 3 /s) of water channels, the Pf(blank) value of the blank vesicle without water channels needs to be deducted from Pf(channel), which was multiplied by the vesicle surface area (S) and divided by the number of water channels (N) incorporated in the liposome as shown below. Further taking into consideration of channel incorporation efficiency (CIE, Supplementary Table 3), the Pw values can be calculated by the following equation: PW = (Pf(blank) -Pf(channel)) x (S/(N x CIE))</p><p>The HPTS assay for cation transport EYPC (1 mL, 25 mg/mL in CHCl3, Avanti Polar Lipids, USA) was placed in a 10 mL round bottomed flask and solvent was evaporated by slowly purging N2. After drying the resulting thin film under high vacuum overnight at room temperature, the film was hydrated with a HEPES buffer solution (1 mL, 10 mM HEPES, pH = 7.0) containing pH-sensitive HPTS dye molecules (0.5 mM) at room temperature for 1 hour (with occasional vortexing after every 15 minutes) to give a milky suspension.</p><p>The mixture was then subjected to 10 freeze-thaw cycles (freezing in liquid N2 for 1 minute and heating at 55 o C in water bath for 2 minutes). The vesicle suspension was extruded through polycarbonate membrane (0.1 μm) to produce mostly monodispersed LUVs of about 120 nm in diameter with HPTS dyes encapsulated inside. The extravesicular HPTS dye was removed by using size exclusion chromatography (stationary phase: Sephadex G-50, GE Healthcare, USA; mobile phase: 10 mM HEPES buffer, pH = 7.0) and diluted with the mobile phase to yield 3.</p><!><p>The SPQ-containing LUV suspension (30 μL, 10 mM lipid, 200 mM NaNO3) was added to a NaCl solution (1.97 mL, 200 mM NaCl) to create a chloride concentration gradient for ion transport observation. A solution of 1 at 120 µM in DMSO was then injected into the suspension under gentle stirring. Upon channel addition, the SPQ dye emission was immediately monitored at 430 nm with excitation at 360 nm for 300 seconds using fluorescence spectrophotometer (Hitachi, Model F-7100, Japan), after which time an aqueous solution of Triton X-100 (20 μL, 20% v/v) was added to completely eliminate the chloride gradient. The final transport trace was obtained by normalizing the fluorescence intensity using equation shown below.</p><!><p>x 100 Where F0 = fluorescence intensity just before the channel addition (at t = 0 s), Ft = Fluorescence intensity at time t, and F1 = fluorescence intensity after addition of Triton-X100.</p><!><p>To determine activation energies for water transport, we conducted water permeability measurements at different temperatures at intervals of 5 °C between 20 and 40 °C. For these experiments, the solution reservoir and the measurement cell of the stopped-flow instrument were maintained at a set temperature by a recirculating heater/chiller (Polystat, Cole Parmer). Permeability rates through channels at varying temperatures were used to construct an Arrhenius plot.</p><!><p>where k is the exponential coefficient of the change in the light scattering; A is pre-exponential factor; Ea is activation energy; R is gas constant; T is absolute temperature in Kelvin.</p><!><p>All MD simulations were performed using the MD program NAMD2 with periodic boundary conditions and using the particle mesh Ewald (PME) method to calculate the long-range electrostatics. The Nose-Hoover Langevin piston and Langevin thermostat were used to maintain the constant pressure and temperature in the system. CHARMM36 force field parameters describe the bonded and non-bonded interactions of among, lipid bilayer membranes, water and ions. An 8-10-12 Å cutoff scheme was used to calculate van der Waals and short range electrostatics forces. All simulations were performed using a 2 fs time step to integrate the equation of motion. SETTLE algorithm was applied to keep water molecules rigid whereas RATTLE algorithm constrained all other covalent bonds involving hydrogen atoms. The coordinates of the system were saved at an interval of 19.2 ps. The analysis and post processing the simulation trajectories were performed using VMD and CPPTRAJ.</p><p>The initial structure of channel 1 having 25 AB repeating units was built using a fragment-assembly strategy. Specifically, a helical fragment containing 8 AB repeating units was built using Gaussview and optimized at the HF/6-31G(d) level. Based on the optimized structural parameters (bond angle/length, dihedral angle, etc) of this helical fragment, we then built longer channel 1. The topology and force field parameters for the monomeric unit of 1 were created using the CHARMM general force fields (CGenFF) webserver. Subsequently, the channel was embedded into a 10.5 x 10.5 nm 2 patch of pre-equilibrated POPC lipid bilayer membrane. The lipid patch was generated using the CHARMM-GUI membrane builder and pre-equilibrated for approximately 400 ns. Lipid molecules that overlapped with the channel were removed. The system was then solvated with water using the Solvate plugin of VMD. Sodium and chloride ions were added to 0.6 M concentration using the Autoionize plugin of VMD. The final assembled system measured 10.5 x 10.5 x 9.0 nm 3 and contained 100,482 atoms.</p><p>Following the assembly, the system underwent 1200 steps of energy minimization using the conjugate gradient method to remove steric clashes. After energy minimization, the system was subjected to a 48 ns equilibration at a constant number of atoms (N), pressure (P = 1 bar) and temperature (T = 300 K), the NPT ensemble, with harmonic restraints applied to all non-hydrogen atoms of channels that surrounded the transmembrane pore. The restraints were applied relative to the initial coordinates of the atoms, with spring constants at 1 kcal mol -1 Å -2 . After 48 ns, the harmonic restraints were removed, and the system was equilibrated while restraining the RMSD of the channel with respect to its QM optimized initial conformation using the colvar module of NAMD. For the corresponding references, see the supplementary information.</p>

ChemRxiv

A new, <i>substituted</i> palladacycle for ppm level Pd-catalyzed Suzuki–Miyaura cross couplings in water

A newly engineered palladacycle that contains substituents on the biphenyl rings along with the ligand HandaPhos is especially well-matched to an aqueous micellar medium, enabling valued Suzuki-Miyaura couplings to be run not only in water under mild conditions, but at 300 ppm of Pd catalyst. This general methodology has been applied to several targets in the pharmaceutical area. Multiple recyclings of the aqueous reaction mixture involving both the same as well as different coupling partners is demonstrated.Low temperature microscopy (cryo-TEM) indicates the nature and size of the particles acting as nanoreactors. Importantly, given the low loadings of Pd invested per reaction, ICP-MS analyses of residual palladium in the products shows levels to be expected that are well within FDA allowable limits.

a_new,_<i>substituted</i>_palladacycle_for_ppm_level_pd-catalyzed_suzuki–miyaura_cross_couplings_in_

Introduction<!>Results and discussion<!>Conclusions

<p>In a recent review by Alami and Messaoudi, 1 palladacycles were characterized as among the "most powerful" pre-catalysts to highly reactive, mono-ligated forms of Pd(0). 2 While they are stable, quite convenient, and broadly applicable, these features may not be sufficiently attractive for long term usage typically in the 1-5 mol% range. That is, not only is awareness of the endangered status of Pd gaining in appreciation, such reagents are also almost invariably used in environmentally egregious organic solvents and with limited levels of solvent and precious metal recycling. 3 Put another way, and notwithstanding Nobel Prize-level recognition bestowed in 2010 on Pd-catalyzed Suzuki-Miyaura (SM) cross-coupling chemistry developed several decades earlier, 4 such an approach to modern Pd-based catalysis as practiced today is both costly and not sustainable. Access to worldwide supplies, even under the best of geopolitical circumstances, is determined by current limits of technology that prevent access to metals that lie too deep within the Earth's surface. 5 One solution to this inevitable shortage calls for switching, in large measure, from a petroleum-to a water-based discipline, 6 akin to the role of water as the reaction medium in nature, in general, and biocatalysis in particular. 7 Along with this gradual transition comes myriad opportunities for developing new catalysts engineered to function both within an altered reaction medium and under newly unfolding rules for which analogies in organic solvent are non-existent. 8 Palladacycles present one such opportunity where, under traditional conditions (i.e., use in organic solvents), the focus has been on modications that include, e.g., the nature of the leaving group. Such design changes that feature steric and stereoelectronic effects, conformational biases, etc., have little to do with palladacycle solubility. In an aqueous medium containing micellarbased nanoreactors, 9 however, solubilization becomes a crucial parameter. Thus, one key to successful couplings in water involves inuencing the binding constant of a reagent to the micellar inner core: the greater the incentive to enter the site of reaction, the more catalytic activity is to be expected and the lower the catalyst loadings. Hence, the question arises: could the appropriate substitution pattern on the biaryl skeleton within a palladacycle pre-catalyst enhance its micellar entry, and thereby reduce the required level of otherwise precious metal, and associated (oentimes equally precious) ligand, in a cross-coupling reaction? Substituted biarylamine-based palladacycles are currently unknown, 10 since in organic solvent there is no reason to pursue such derivatization. In water, however, where new rules are operating, 8b the prospects for not only providing the convenience of palladacycles that lead to especially reactive catalysts as well as the potential for addressing the endangered nature of platinoids provides more than ample justication for investigating this nontraditional approach. In this report we describe such a newly adorned palladacycle pre-catalyst that, indeed, allows for a general and environmentally responsible process for SM couplings to be run, in most cases, in water under mild conditions and at the 300 ppm (0.03 mol%) level of Pd (Scheme 1).</p><!><p>Several newly substituted palladacycles were prepared initially containing the ligand rac-HandaPhos 11 (P1, P4, P5, P6 and P7), the structures of which are shown in Scheme 2. These included either one or two lipophilic t-butyl residues (as in P4 and P5, respectively), along with those bearing one (P6) or two isopropyl moieties (P7), as compared to the parent array (P1). Several additional ligands commonly used within palladacycles were also prepared and tested under identical conditions in the model reaction between bromide 1 and boronic acid 2 to arrive at biaryl 3. Clearly, the most effective catalyst, by far, is the diisopropyl-substituted HandaPhos palladacycle, P7. Surprisingly, even the di-t-butyl analog, P5, was not as effective. What may also be found striking at rst, but is perfectly in harmony with the "new rules" associated with this chemistry in water, 8 are the results observed for catalysts P9 and P10, where HandaPhos has been replaced by XPhos and SPhos, respectively. While both are excellent choices for SM couplings in traditional organic solvents (e.g., toluene or dioxane), 12 they are non-functional at 300 ppm in water. Likewise, other well-known ligands that make up catalysts P11 and P12 were not competitive under these conditions. Additional optimization studies regarding the choice of surfactant, base, as well as results using organic solvents are described in the associated ESI. † Several examples of SM couplings were carried out using 300 ppm of catalyst P7, as summarized in Table 1. While most cases were amenable to this very low loading of Pd, some could be conducted at levels even down to 25 (4) to 100 (5) ppm of P7. On the other hand, some cases (10, 12, 13, 18-21, 13 and 25) required up to 500 ppm, possibly reecting competition by the product for palladium. Both electron-donating and -releasing groups in the educt are tolerated. Heterocycles present in either the halide or boronic acid partner could be used. Partners with protecting groups are easily coupled (20 and 21). Polyaromatics such as 27 and 29 are easily fashioned. Alternatives to boronic acids, including Bpin, Molanderate BF 3 K salts, 14 and MIDA boronates 15 (see 22 and 23) appear to be compatible partners as well. Double SM couplings using the corresponding precursor dibromides proceeded smoothly to give the anticipated products (30, 31, and 32), using what is, formally, only 150 ppm of this Pd catalyst per bond formed.</p><p>The importance of organic co-solvents has also been addressed, as these additives can play a dramatic role in scaling up reactions under micellar conditions. 16 The co-solvent effect is responsible not only for increasing solubility of highly crystalline educts, but also enlarges micellar size, thereby expanding the interior volume available for reaction. The observed impact of three organic solvents used as 10 vol% in this aqueous surfactant system is shown in Scheme 3 involving coupling partners 33 and 34. Each solvent (THF, toluene, and acetone) was found to increase the rate of conversion. Analysis of the medium for 2 wt% TPGS-750-M/H 2 O vs. that with 10% THF by cryo-TEM (Scheme 3, bottom) revealed the enlargement of the former (ca. 50 nm) to ca. 200 nm due to the presence of THF, suggesting that larger nanoreactors may be responsible for enhancing the overall rates of these cross-couplings. The potential use of less reactive aryl chlorides was briey examined at the 500 ppm level of Pd catalysis (0.05 mol%). As the examples in Table 2 show, a variety of aromatic and heteroaromatic chlorides and boronic acids could be employed, arriving at the targeted biaryls in good isolated yields. Included in this study is the late stage derivatization of aryl chloride fenobrate 17 to analog 38.</p><p>Table 1 Substrate scope for couplings with ppm Pd pre-catalyst P7 in water a a Reaction conditions unless otherwise noted: 0.5 mmol aryl halide, 0.6 mmol aryl boronic acid, 1.0 mmol Et 3 N, 25-500 ppm P7 stirred at 45 C in TPGS-750-M/H 2 O (0.5 M); isolated yields are shown. Double SM couplings were carried out using 1.2 mmol of aryl boronic acid, 2.0 mmol of Et 3 N, and 10% THF as a co-solvent.</p><p>As an illustration of the opportunities to carry out multi-step processes given the commonality of reaction conditions (i.e., in aqueous nanoreactors at rt-45 C), the commonly used fungicide boscalid 18 could be prepared in three steps using a 1pot protocol (Scheme 4). Initially, biaryl 37 was constructed that, without isolation, was subjected to nitro group reduction using our previously described carbonyl iron powder. 19 The Scheme 4 Boscalid: 3-step, 1-pot synthesis in water. resulting aniline was then treated directly with 2-chloronicotynyl chloride. The nal product, boscalid, was ultimately isolated in 80% overall yield. The seemingly incompatible addition of this acid chloride to this aqueous medium, while counter-intuitive at rst, is yet another example of the "new rules" associated with chemistry in water. Reagents and/or reaction partners that are sensitive albeit insoluble in water simply do not hydrolyze or quench; rather, upon stirring they enter the hydrophobic inner micellar core where they react, usually as desired.</p><p>A multi-gram scale reaction between educts 45 and 46 was run in water using Pd catalyst P7 to document the prospects for scaling up these SM couplings (Scheme 5). Use of 24 mmol of 45 and 20 mmol of 46 in the presence of 40 mmol of Et 3 N were exposed to 300 ppm of P7. Stirring this heterogeneous mixture for 15 hours yielded 94% of the desired coupled product 47. In this case, the reaction was quite efficient in the absence of a cosolvent. That is, stirring was not an issue throughout the 15 h reaction period (see images (a-c)) The product 47 could be isolated as a white solid (image (d)), puried by simple ltration through silica gel.</p><p>Biologically active targets, such as precursors to (a) Merck's anacetrapib (48), 20a (b) sonidegib (49), 20b and (c) Novartis' valsartan (50), 20c could also be prepared efficiently under mild conditions using 300-500 ppm of catalyst P7 (Table 3). Additional representative examples of biaryls (51-53) en route to anticancer drugs are also to be found in this table. 21 Facile recycling of the aqueous TPGS-750-M solution is an important aspect to this environmentally responsible technology, leading to very low levels of aqueous waste streams. 22 By contrast, recycling of organic media typically requires fractional distillation to separate reaction and workup solvents for re-use. Scheme 6 illustrates just how effective a 2 wt% aqueous solution of TPGS-750-M can be, thereby dramatically minimizing aqueous waste streams. Recycling could be carried out using a different SM reaction with each of four recycles, following an initial coupling. Products were either separated via ltration or by decantation of the aqueous mixture; hence, individual extractions were not required prior to purication. cryo-TEM analysis of the aqueous mixture aer ve uses revealed that while the nanomicelles were of the same shape, they were unexpectedly larger (ca. 75 nm; Scheme 6).</p><p>From the perspective of the pharmaceutical industry, it is commonly assumed that under traditional SM cross-coupling conditions the amount of residual Pd in the product is going to be outside of the acceptable 10 ppm limit imposed by the US FDA. 23 Hence, additional processing is usually anticipated, potentially adding time and expense to the eventual API. But use of such low levels of Pd catalysts rarely exceed this limit. With catalyst P7 at the 300 and even 500 ppm loadings it was not surprising that, for the three cases randomly selected and examined by ICP-MS, no more than 6 ppm Pd was found for biaryls 12, 24, and 51 (Fig. 1). On the other hand, following traditional literature conditions used to make each of these biaryl products (e.g., 2 mol%, or 20 000 ppm Pd), residual levels of Pd were found to be orders of magnitude greater.</p><!><p>In summary, a new palladacycle has been uncovered that mediates Suzuki-Miyaura couplings in water at the 300 ppm level of precious metal. Key to this methodology is placement of an isopropyl group on each aromatic ring of the biaryl skeleton making up the palladacycle, a substitution pattern that could not have been predicted given the lack of precedent for such pre-catalysts. Likewise, screening of several monophosphines, including some of the most commonly used for such Pd-catalyzed cross-couplings, ultimately identifying HandaPhos as the preferred ligand (i.e., P7), requires further study to rationalize the effectiveness of this novel ligand/ palladacycle precursor combination. Applications to various targets within the pharma, agro, and materials domains have been demonstrated, along with the potential for large scale use, recycling of the aqueous reaction medium, and tandem 1pot processes. The nature of the nanomicelles involved has been determined via cryo-TEM measurements, both initially as well as aer use in the presence of added co-solvent. Residual levels of Pd in the products formed have been shown to be well within governmental limits for safety, further enhancing the attractiveness of this technology. The prognosis for use of the same pre-catalyst for other types of Pd-catalyzed crosscouplings (e.g., Stille, Sonogashira, and Heck couplings) looks encouraging, with the results from these ongoing studies to be reported in due course.</p>

Royal Society of Chemistry (RSC)

RexAB Promotes the Survival of Staphylococcus aureus Exposed to Multiple Classes of Antibiotics

ABSTRACTAntibiotics inhibit essential bacterial processes, resulting in arrest of growth and, in some cases, cell death. Many antibiotics are also reported to trigger endogenous production of reactive oxygen species (ROS), which damage DNA, leading to induction of the mutagenic SOS response associated with the emergence of drug resistance. However, the type of DNA damage that arises and how this triggers the SOS response are largely unclear. We found that several different classes of antibiotic triggered dose-dependent induction of the SOS response in Staphylococcus aureus, indicative of DNA damage, including some bacteriostatic drugs. The SOS response was heterogenous and varied in magnitude between strains and antibiotics. However, in many cases, full induction of the SOS response was dependent upon the RexAB helicase/nuclease complex, which processes DNA double-strand breaks to produce single-stranded DNA and facilitate RecA nucleoprotein filament formation. The importance of RexAB in repair of DNA was confirmed by measuring bacterial survival during antibiotic exposure, with most drugs having significantly greater bactericidal activity against rexB mutants than against wild-type strains. For some, but not all, antibiotics there was no difference in bactericidal activity between wild type and rexB mutant under anaerobic conditions, indicative of a role for reactive oxygen species in mediating DNA damage. Taken together, this work confirms previous observations that several classes of antibiotics cause DNA damage in S. aureus and extends them by showing that processing of DNA double-strand breaks by RexAB is a major trigger of the mutagenic SOS response and promotes bacterial survival.

rexab_promotes_the_survival_of_staphylococcus_aureus_exposed_to_multiple_classes_of_antibiotics

INTRODUCTION<!>Multiple classes of antibiotics cause DNA damage in S. aureus.<!><!>SOS induction is partly due to processing of DNA double-strand breaks by the RexAB helicase/nuclease complex.<!><!>SOS induction is partly due to processing of DNA double-strand breaks by the RexAB helicase/nuclease complex.<!>DNA DSB repair reduces bacterial susceptibility to several classes of antibiotics.<!><!>DNA DSB repair reduces bacterial susceptibility to several classes of antibiotics.<!>RexAB promotes staphylococcal tolerance of several classes of antibiotics.<!><!>RexAB promotes staphylococcal tolerance of several classes of antibiotics.<!>RexAB promotes staphylococcal survival during exposure to oxacillin and fosfomycin.<!><!>RexAB promotes staphylococcal survival during exposure to oxacillin and fosfomycin.<!>DISCUSSION<!>Bacterial strains and culture conditions.<!><!>recA-gfp fluorescent reporter assay.<!>Determination of MIC.<!>Antibiotic survival assay.<!>Endogenous ROS production.<!>Statistical analyses.

<p>Staphylococcus aureus is a common cause of both superficial and invasive infections (1). Many of these infections, such as infective endocarditis and osteomyelitis, can be difficult to treat, requiring lengthy courses of therapy (2–10). Staphylococcal infections are also associated with a high rate of relapse and/or the development of chronic infections, even when the bacteria causing the infection appear to be fully antibiotic susceptible (2–10).</p><p>There is, therefore, a pressing need to identify new approaches to enhance antibiotic efficacy. To do this, it is important to have a comprehensive understanding of the factors that influence bacterial susceptibility to antibiotics. For example, replication rate has been shown to correlate with susceptibility to several classes of antibiotic (11–13). However, recent evidence suggests that metabolic activity is a better indicator of susceptibility than the replication rate, indicating that metabolism contributes to the bactericidal activity of certain antibacterial drugs (14). This is because the inhibition of bacterial processes by bactericidal antibiotics leads to metabolic perturbations, which in turn result in the generation of reactive oxygen species (ROS) (15–22). These highly reactive molecules damage cellular molecules, including DNA, lipids, and proteins, and have been proposed to contribute to the lethality of bactericidal antibiotics (15, 16, 23–26). However, the magnitude of the damage caused by antibiotic-triggered ROS production and the degree to which these radicals contribute to bacterial killing are unclear (27–29).</p><p>DNA damage leads to induction of the SOS response, which involves the expression of genes that encode proteins involved in DNA repair (30–36). In S. aureus, the SOS response includes 16 genes, including RecA and LexA, which are the key regulators of the system (33). It also includes the error-prone polymerase UmuC, the expression of which increases the mutation rate, resulting in increased frequency of antibiotic resistance within populations exposed to SOS-inducing antibiotics and the emergence of the small-colony variant (SCV) phenotype associated with resistance to the oxidative burst of neutrophils and the establishment of chronic infection (33, 37, 38). However, what is not clear is the nature of the DNA damage that is caused by antibiotic-induced ROS or how this triggers the SOS response. This issue is worth resolving because a greater understanding of the mechanisms by which bacteria repair the damage caused by ROS may help to identify new therapeutics that enhance antibiotic activity and reduce the emergence of drug-resistant strains (39). For example, we have shown previously that the combination antibiotic cotrimoxazole (trimethoprim plus sulfamethoxazole) caused DNA double-strand breaks (DSB) and that processing of these by the RexAB nuclease/helicase complex was required for induction of the SOS DNA repair response (30). RexAB is a member of the AddAB family of ATP-dependent helicase/nucleases that process DNA DSBs to produce a 3′ single strand of DNA (34, 39–41). RecA binds to the single-stranded DNA, resulting in a nucleoprotein filament that triggers autocleavage of the LexA transcriptional repressor and induction of the SOS response (39, 41).</p><p>The generation of DNA DSBs by cotrimoxazole appeared to be oxygen-dependent, and these were lethal if not repaired, resulting in reduction in CFU counts of a mutant defective for DSB repair (rexB::Tn) 50- to 5,000-fold greater than the reduction in CFU counts of wild-type S. aureus (30). However, it was unclear whether DNA DSBs occurred with other antibiotics and if the repair of these by RexAB was a major contributor to induction of the SOS response. If DNA DSBs are a consistent occurrence with diverse antibiotics, then inhibition of RexAB may be an effective way of enhancing the bactericidal activity of antibiotics as well as reducing the emergence of drug-resistant and host-adapted SCV phenotypes.</p><p>To test whether our findings with cotrimoxazole were applicable to other antibacterial drugs, we undertook a comprehensive analysis of multiple classes of antibiotics. This revealed that most antibiotics cause DNA damage in S. aureus under aerobic conditions, which appeared to result in DNA DSBs, since mutants lacking DNA DSB repair complex RexAB were more susceptible to antibiotic killing and had reduced induction of the SOS response.</p><!><p>DNA damage in most bacteria, including S. aureus, triggers activation of the SOS response, which leads to the transcription of genes whose products contribute to DNA repair (30–34). These genes include recA, which encodes the RecA protein required for homologous recombination and, together with LexA, is a key regulator of the SOS response (30–36).</p><p>To determine whether antibiotics caused DNA damage in S. aureus, we used a well-characterized PrecA-gfp reporter system in two distinct genetic backgrounds: SH1000, a methicillin-sensitive S. aureus (MSSA) strain, and JE2, a community-associated methicillin-resistant S. aureus (CA-MRSA) strain of the USA300 lineage (30, 34, 42, 43). This system has been shown to produce a dose-dependent fluorescent response to DNA damage caused by the ROS generator paraquat, mitomycin C, and cotrimoxazole (30, 34).</p><p>These SOS reporter strains were then exposed to various classes of clinically relevant antibiotics across a range of concentrations that partially inhibited growth (Fig. S1). These included both bactericidal (cotrimoxazole, ciprofloxacin, nitrofurantoin, oxacillin, daptomycin, gentamicin) and bacteriostatic (chloramphenicol, linezolid) drugs. For most antibiotics, the concentrations ranged from 0.125× to 1× the MIC of the antibiotic. However, for cotrimoxazole, higher concentrations were needed to show growth inhibition (Fig. S1), most likely due to the inoculum effect since a higher concentration of bacteria was used in PrecA reporter assays than in MIC assays (44). We also used a higher range of concentrations of oxacillin for the USA300 strain because it is resistant to most β-lactams (42).</p><p>As expected, we found that cotrimoxazole, ciprofloxacin, nitrofurantoin, and oxacillin triggered SOS induction in both the SH1000 and JE2 strains, albeit to various degrees and with different temporal dynamics (Fig. 1A to D) (30, 34–36, 45). In all cases, however, there was evidence of dose-dependent induction of the SOS response (Table S1). DNA damage was also apparent during exposure to the bactericidal lipopeptide antibiotic daptomycin and the bacteriostatic drugs chloramphenicol and linezolid, again with differences in the size and time-dependence of the response between antibiotics and with some differences between the two strains (Fig. 1E to G) (Table S1). However, there was almost no induction of the SOS response during bacterial exposure to gentamicin at any of the concentrations used (Fig. 1H). Taken together, these data indicated that most clinically relevant classes of antibiotics, including bacteriostatic agents, caused DNA damage in S. aureus.</p><!><p>Induction of the SOS response in S. aureus SH1000 and JE2 by diverse classes of antibiotics. (A to H) Induction of SOS measured by GFP expression driven from a PrecA-gfp reporter construct upon exposure to a range of concentrations of various antibiotics. Concentrations were chosen based on their ability to cause growth inhibition and represent multiples of the MIC of the individual strain as indicated in the key above each graph. GFP fluorescence was normalized to OD600 to determine induction of SOS relative to cell density. Data represent the mean from 3 independent experiments (n = 3). Representative OD600 measurements alone are shown in Fig. S1. Error bars represent standard deviation of the mean.</p><!><p>We have shown previously that induction of the SOS response by cotrimoxazole is largely due to the processing of DNA double-strand breaks (DSBs) by the AddAB family RexAB nuclease/helicase complex and the resulting formation of a RecA nucleoprotein filament that leads to the autocatalytic cleavage of LexA (30, 31, 34, 37, 40, 41). Therefore, we determined whether SOS induction by other classes of antibiotics was also due to RexAB-mediated processing of DNA DSBs. As before, cotrimoxazole was included in these assays as a control.</p><p>To do this, we compared green fluorescent protein (GFP) fluorescence from wild-type S. aureus JE2 and a rexB::Tn mutant defective for RexAB, both of which contained the PrecA-gfp reporter system, during exposure to the same panel of antibiotics as that described for Figure 1 (30, 34, 38, 39). As expected from our previous work, we found that the lack of RexAB reduced recA induction relative to that of the wild type during exposure to cotrimoxazole (Fig. 2A and Table S2) (30). We also observed reduced recA expression in the rexB::Tn mutant relative to that in the wild type during exposure to the quinolone antibiotic ciprofloxacin, which is known to cause DNA DSBs (Fig. 2B and Table S2) (45).</p><!><p>RexAB is required for maximal induction of the SOS response during exposure to antibiotics. (A to H) Induction of SOS response of JE2 wild type and rexB mutant measured by GFP expression upon exposure to a range of sublethal concentrations of antibiotics. Concentrations of antibiotic are labeled by multiples of the MIC of the wild-type strain. GFP fluorescence was normalized by OD600 to determine induction of SOS relative to cell density (n = 3). Representative OD600 measurements alone are shown in Fig. S1. Error bars represent standard deviation of the mean.</p><!><p>For nitrofurantoin, oxacillin, daptomycin, chloramphenicol, and linezolid, we also observed lower levels of SOS induction in the rexB::Tn mutant relative to those in the wild type, although, while statistically significant, the difference between wild-type and mutant strains was not as large as that for cotrimoxazole and ciprofloxacin (Fig. 2C to H and Table S2). As expected from previous data (Fig. 1G), very little recA induction was observed from either wild type or rexB::Tn mutant during exposure to gentamicin (Fig. 2G). Therefore, as for cotrimoxazole, RexAB is required for maximal induction of SOS in response to DNA damage caused by several clinically relevant antibiotics, indicating that these drugs cause DNA DSBs in S. aureus.</p><!><p>The requirement of RexAB for maximal induction of the SOS response indicated that exposure to most antibiotics caused DNA DSBs (30, 34). Since DSBs are lethal if not repaired, we hypothesized that mutants defective for RexAB would be more susceptible than wild-type strains to those antibiotics that triggered the SOS response (34, 39, 40).</p><p>To test this, we determined the MIC of each antibiotic for wild-type S. aureus SH1000 and JE2 and associated rexB::Tn mutants (Table 1). The rexB mutants in both JE2 and SH1000 strains were ≥2-fold more susceptible to 7 of the 8 antibiotic-tested conditions (Table 1). Importantly, the absence of RexAB increased the susceptibility of the MRSA strain JE2 to both oxacillin and ciprofloxacin 4-fold, despite this strain being resistant to both antibiotics (42).</p><!><p>MIC values (μg ml−1) of S. aureus WT and rexB mutant in SH1000 and JE2 backgrounds for various antibiotics (n ≥ 3; median MIC is shown); the fold reduction in MIC of the rexB::Tn mutants relative to the wild type is also shown</p><!><p>The one exception was gentamicin, where the SH1000 rexB::Tn mutant was 2-fold more susceptible to the antibiotic, but the JE2 rexB::Tn mutant had the same MIC as the wild-type strain, in keeping with the fact that this antibiotic did not trigger the SOS response under the conditions tested (Table 1). Taken together, the MIC data provide additional evidence that most antibiotics cause DNA DSBs in S. aureus.</p><!><p>We have shown previously that DNA DSB repair by RexAB enables staphylococcal tolerance of the combination antibiotic cotrimoxazole (30). Since most of the other antibiotics we examined also appeared to cause DNA DSBs, leading to increased susceptibility of rexB::Tn mutants in MIC measurements, we hypothesized that RexAB would also contribute to bacterial survival during exposure to a supra-MIC of these other antibacterial drugs.</p><p>To test this, we exposed wild-type S. aureus SH1000 and JE2 and associated rexB::Tn mutants to 10× the MIC of the wild type of each of the antibiotics used in previous assays and measured survival after 8 h of incubation at 37°C in an aerobic atmosphere (Fig. 3). Similar to the MIC assays, 6 of 8 antibiotics tested were more active against the rexB::Tn mutant than against wild-type bacteria, resulting in lower survival of the DNA repair-defective strains (Fig. 3). The two antibiotics where there was no difference in survival between wild type and rexB::Tn mutants were linezolid and chloramphenicol, which are both bacteriostatic and did not reduce CFU counts of any of the strains (Fig. 3F and G). The remaining 6 antibiotics (cotrimoxazole, ciprofloxacin, oxacillin, nitrofurantoin, daptomycin, gentamicin), all of which are classified as bactericidal, caused significantly greater decreases in CFU counts of the rexB::Tn mutants than in CFU counts of wild-type bacteria (Fig. 3A, B, C, D, E, and H). The increased susceptibility of the rexB::Tn mutants resulted in reductions in CFU counts 5- to 500-fold greater than those of wild-type cells after 8-h exposure to the 6 bactericidal antibiotics (Fig. 3A, B, C, D, E, and H).</p><!><p>Lack of effective DNA repair increases the killing of S. aureus by bactericidal antibiotics under aerobic conditions. (A to H) Survival of S. aureus wild type (WT) and rexB::Tn mutant in SH1000 and JE2 backgrounds after 8 h of incubation at 37°C in TSB supplemented with 10× MIC. Survival was assessed under aerobic (blue) or anaerobic (green) conditions (n = 3). Data were analyzed by ordinary one-way ANOVA with Tukey's correction for multiple comparisons (*, P < 0.05 mutant versus wild type under the same atmospheric condition; NS, not significant) and presented as a box and whisker plot with error bars showing the full data range.</p><!><p>The observation that the rexB::Tn mutants of both SH1000 and JE2 were killed more efficiently than wild-type strains by gentamicin (Fig. 3H) was surprising given the lack of SOS response during exposure to the aminoglycoside antibiotic (Fig. 1H and 2H). To determine whether these findings were applicable to other aminoglycoside antibiotics, we measured the susceptibility of wild type and rexB::Tn mutants to kanamycin. Wild-type JE2 and SH1000, and both of the corresponding rexB::Tn mutants, had identical kanamycin MICs (1 μg ml−1). Furthermore, there were no differences in survival between wild type and rexB::Tn mutant after 8-h exposure to kanamycin at 10× the MIC (Fig. S2), although there were differences in susceptibility between the two strains. Therefore, while gentamicin was more bactericidal against rexB::Tn mutants than against wild-type strains, this does not appear to be the case for all aminoglycoside antibiotics.</p><p>To test whether DNA DSBs caused by bactericidal antibiotics were due to endogenous ROS production, we repeated bactericidal activity assays under anaerobic conditions. As reported previously, cotrimoxazole lost most of its bactericidal activity against wild-type bacteria in the absence of oxygen, as did ciprofloxacin and gentamicin, the latter due to the reduction in membrane potential in the absence of oxygen (30, 46, 47) (Fig. 3A, B, and H). The increased susceptibility of the rexB::Tn mutant to cotrimoxazole and oxacillin relative to that of wild type seen under aerobic conditions was also much less pronounced or absent under anaerobic conditions, suggesting that ROS may contribute to DNA DSBs in the presence of these antibiotics (Fig. 3A and C).</p><p>The bactericidal antibiotics nitrofurantoin and daptomycin retained bactericidal activity under anaerobic conditions (Fig. 3C to E). However, nitrofurantoin and daptomycin also retained bactericidal activity against the rexB::Tn mutants, suggesting that they caused DNA DSBs in an ROS-independent manner (Fig. 3D and E).</p><p>To explore the potential role of endogenous ROS in DNA DSB production further, we measured the production of ROS using a fluorescent dye (30) in S. aureus strains exposed to two antibiotics that were significantly more bactericidal against the rexB::Tn mutant under aerobic than under anaerobic conditions (cotrimoxazole and oxacillin) and compared them with two antibiotics that killed the rexB::Tn mutant efficiently under both aerobic and anaerobic conditions (nitrofurantoin and daptomycin). Antibiotic concentrations were the same as those used in PrecA-gfp reporter assays to enable comparison. As expected from previous work (30), exposure to cotrimoxazole caused dose-dependent ROS production in both S. aureus strains (Fig. S3A and Table S3). A similar ROS production profile was seen during oxacillin exposure, but nitrofurantoin and daptomycin did not cause dose-dependent ROS production (Fig. S3B to D and Table S3).</p><p>In summary, ROS appear to contribute to DNA DSBs caused by some antibiotics, as evidenced by increased susceptibility of the rexB::Tn mutant to several antibiotics relative to that of the wild type under aerobic conditions and dose-dependent ROS production. However, some antibiotics appear to cause DNA DSBs under aerobic conditions that may not be due to ROS, suggesting that additional mechanisms of DNA damage are possible. Furthermore, ROS were not required for the lethality of all antibiotics since some retained bactericidal activity under anaerobic conditions.</p><!><p>The finding that loss of RexAB resulted in increased killing of both the SH1000 MSSA and JE2 MRSA strains by the frontline antistaphylococcal β-lactam oxacillin was particularly noteworthy because this indicated a mechanism by which MRSA strains could be resensitized to the antibiotic. Therefore, we repeated this assay and included mutants complemented with the rexBA operon (34) (Fig. 4). As expected, the rexB::Tn mutants were more susceptible to killing by oxacillin than were wild-type bacteria under aerobic but not anaerobic conditions. Complementation of mutations with plasmids containing the rexBA operon, but not the plasmid alone, restored survival to wild-type levels, confirming the role of RexAB in staphylococcal tolerance of oxacillin (Fig. 4).</p><!><p>Complementation of the rexB::Tn mutant restores tolerance to oxacillin in S. aureus SH1000 and JE2. Survival of S. aureus WT and rexB mutant in SH1000 and JE2 backgrounds after 8 h of incubation at 37°C in TSB supplemented with 10× MIC. Survival was assessed under aerobic (blue) or anaerobic (green) conditions (n = 3). Data were analyzed by one-way ANOVA with Dunnett's correction for multiple comparisons (*, P < 0.05 versus wild type) and presented as a box and whisker plot with error bars showing the full data range.</p><!><p>To understand whether RexAB promoted staphylococcal survival during exposure to other β-lactams, we exposed wild type and rexB::Tn mutants to a panel of other β-lactams at 10× the MIC (Fig. S3 and Table S4). The rexB::Tn mutant of S. aureus JE2 was more susceptible to bactericidal effects of penicillin G and imipenem than was the wild type, but it was not more susceptible to the six other β-lactams tested (Fig. S3). In contrast, the rexB::Tn mutant of SH1000 was more susceptible than wild type to 6/8 β-lactams (Fig. S3). This difference may reflect the fact that JE2 is a MRSA strain while SH1000 is not. To further explore the susceptibility of the rexB::Tn mutant to cell wall-targeting antibiotics, we extended our analysis to vancomycin and fosfomycin. For both JE2 and SH1000 strains, the rexB::Tn mutants were more susceptible than wild type to killing by fosfomycin but not by vancomycin. Therefore, the absence of RexAB sensitizes S. aureus to oxacillin and fosfomycin, but these findings do not extend to all antibiotics that target the cell wall, with differences between strains and within antibiotic class.</p><!><p>Staphylococcal infections are associated with high rates of treatment failure, even in the case of apparently drug-susceptible strains (1–10). This, together with the threat posed by multidrug-resistant MRSA strains, necessitates a greater understanding of how antibiotics function and the identification of opportunities to improve their efficacy (9).</p><p>There is compelling evidence that diverse antibiotics trigger metabolic perturbations in bacteria that lead to endogenous ROS production under aerobic conditions (14–22). However, the consequences of this for bacterial viability remain a matter of debate (22, 27–29). This is important because if endogenous ROS production is a common property of antibiotics, then it could be exploited to enhance treatment outcomes, for example by designing inhibitors of bacterial processes that detoxify ROS or repair the damage it causes (30, 34, 39).</p><p>To understand whether antibiotics cause ROS-mediated damage in S. aureus, we focused on the degree to which antibiotic exposure resulted in bacterial DNA damage, since nucleic acids are frequently attacked by endogenous ROS, and the consequences of that damage for bacterial survival (23–26, 30, 34).</p><p>Using a PrecA-gfp reporter assay, we observed that the SOS response in S. aureus was triggered by several different classes of antibiotics, indicative of DNA damage. While this was expected for DNA-targeting antibiotics such as the fluoroquinolone ciprofloxacin (45), SOS induction also occurred with antibiotics that do not directly target bacterial DNA, such as oxacillin, daptomycin, and linezolid. It has been shown that certain β-lactam antibiotics induce the SOS response in Escherichia coli via the DpiBA two-component system rather than via DNA damage (48). Although this mechanism has been hypothesized for S. aureus, our data show that S. aureus rexB::Tn mutants in both the JE2 and SH1000 genetic backgrounds were more susceptible to killing by oxacillin, demonstrating that DNA damage does occur during exposure to this β-lactam antibiotic and that this is at least partially responsible for triggering the SOS response. Therefore, our findings are in keeping with work showing that β-lactam antibiotics trigger endogenous ROS production via elevated TCA cycle activity in response to cell wall damage, leading to increased mutation rate (49).</p><p>It is important to note that while JE2 is a MRSA strain, it is not highly resistant to β-lactam antibiotics as seen in the case of, e.g., strain COL (50). Therefore, it is not clear whether oxacillin would induce the SOS response in highly resistant MRSA strains. Furthermore, the rexB::Tn mutant of the JE2 strain was only more susceptible than wild type to a few β-lactam antibiotics, whereas the SH1000 rexB::Tn mutant was more susceptible than wild type to most of them. This may suggest that resistance to antibiotics provides S. aureus a measure of protection from antibiotic-induced ROS production and DNA damage. The differing sensitivity of the rexB::Tn mutant to each of the β-lactam antibiotics may also reflect differences in the antibiotics' affinity for penicillin binding protein 1, which is required for SOS induction (51).</p><p>While it is still a controversial topic, there is increasing evidence that many classes of antibiotics trigger the endogenous production of ROS. However, the degree to which these ROS contribute to bactericidal activity is less clear (22, 27–29). Our data provide evidence that many antibiotics cause DNA damage, in part via ROS but also apparently via ROS-independent mechanisms. However, this DNA damage appears to be largely tolerated by wild-type bacteria via RexAB-mediated processing of DSBs, which triggers the SOS response to facilitate repair via homologous recombination. Furthermore, while some antibiotics had greater bactericidal activity under anaerobic conditions, this was not the case for daptomycin or nitrofurantoin.</p><p>The production of ROS by bactericidal but not bacteriostatic antibiotics has been proposed to explain the antibiotics' differences in lethality. However, we observed SOS induction during exposure of S. aureus to the bacteriostatic antibiotics linezolid and chloramphenicol but not to the bactericidal antibiotic gentamicin.</p><p>The fact that linezolid and chloramphenicol appeared to trigger the SOS response but not DNA DSBs may be explained by differences in the type of DNA damage caused by each of the antibiotics. While several different types of DNA damage trigger SOS, only those leading to DSBs would be expected to promote susceptibility of the rexB mutant (41, 52, 53). As such, it is possible that bactericidal antibiotics trigger the potentially lethal DNA DSBs, while bacteriostatic antibiotics trigger nonlethal types of DNA damage. In keeping with this hypothesis, the absence of RexAB had only a small effect on SOS induction in S. aureus caused by linezolid or chloramphenicol.</p><p>It is unclear why gentamicin did not trigger SOS during antibiotic exposure since it appeared to cause DNA DSBs in the antibiotic tolerance assays, because we observed 5- to 10-fold increased susceptibility of RexAB-deficient strains to the aminoglycoside antibiotic in bactericidal killing assays. However, it may be the case that high concentrations of the antibiotic are needed for DNA damage.</p><p>Combined, our data indicate differences between antibiotics in the degree of DNA damage caused, as well as the time required to cause damage, and these differences may explain some of the debate around the contribution of ROS to antibiotic-mediated killing. However, the data strongly suggest that DNA DSBs are a common consequence of the exposure of S. aureus to several different classes of antibiotics and that an inability to repair those DSBs increases bacterial susceptibility to several antibacterial drugs. These findings are similar to those reported for E. coli, where mutants defective for DNA DSB repair (defective for recB or recC) were more susceptible than the wild type to at least 8 different antibiotics (54). Crucially, we found that disruption of DNA DSB repair restored quinolone susceptibility in an otherwise resistant strain of S. aureus, which is also similar to what has been seen in E. coli and Klebsiella pneumoniae (55). We also found that an inability to repair DSBs restored oxacillin susceptibility in the JE2 MRSA strain, although it remains to be seen if this finding is applicable to other MRSA strains, particularly those with high-level resistance to β-lactams.</p><p>The identification of RexAB as important for staphylococcal survival during exposure to several different antibiotics, and the fact that loss of RexAB resensitizes otherwise resistant strains to some antibiotics, makes this complex a potential target for novel therapeutics. Crucially, there is a lack of RexAB homologues in eukaryotes, reducing the likelihood of host toxicity (40, 41, 52, 53, 55, 56). Inhibitors of RexAB would be expected to enhance the bactericidal activity of several different classes of antibiotic, as well as reduce the induction of the mutagenic SOS response, which is associated with the emergence of antibiotic resistance and mutants that can resist host immune defenses (37, 38, 56, 57). We have also shown recently that DNA DSB repair is important for staphylococcal resistance to host immune defenses, in keeping with similar findings with several other bacterial pathogens, providing an additional potential benefit of targeting this complex (34, 58–61).</p><p>In summary, our data demonstrate that staphylococcal DNA is damaged by several classes of bactericidal antibiotics, which appears to result in DNA DSBs that are processed by RexAB and trigger the SOS response for repair. Therefore, RexAB promotes staphylococcal survival during exposure to multiple antibacterial drugs and is therefore a potential target for novel therapeutics that sensitize S. aureus to antibiotics.</p><!><p>The bacterial strains used in this study are listed in Table 2. S. aureus was cultured in tryptic soy broth (TSB) or Mueller-Hinton broth (MHB) to stationary phase (18 h) at 37°C, with shaking (180 rpm). Media were supplemented with antibiotics as required. For strains with the PrecA-gfp reporter plasmid, kanamycin (90 μg ml−1 was included), and for rexB::Tn mutants, erythromycin, was added to the medium (10 μg ml−1) (30, 34). The pitet plasmid integrates stably into the staphylococcal chromosome and did not require selection. To induce expression of the rexBA operon in complemented strains, the medium was supplemented with anhydrotetracycline (AHT) at 100 ng ml−1.</p><!><p>Bacterial strains used in this studya</p><p>Eryr, Camr, and Kanr indicate the presence of resistance markers for erythromycin, chloramphenicol, and kanamycin, respectively.</p><!><p>As described previously (30, 34), promoter-reporter gene constructs in JE2 and SH1000 backgrounds were used to quantify expression of recA. Antibiotic 2-fold dilutions were made in flat-bottomed, black-walled, 96-well plates containing TSB and inoculated with 1/10 dilution of a stationary-phase culture of the reporter strains. Plates were placed into an Infinite M200-PRO microplate reader (Tecan) where cultures were grown for 17 h at 37°C (700 rpm), and both absorbance at 600 nm (OD600) and GFP relative fluorescence units (RFU) were measured every 30 min. To account for differences in cell density, RFU values were normalized by OD600 data at each time point.</p><!><p>MICs were determined using a serial broth dilution protocol as described previously (30, 62). Bacteria were diluted to 1 × 105 CFU ml−1 and incubated in flat-bottomed 96-well plates with a range of antibiotic concentrations for 17 h at 37°C under static conditions (aerobic, anaerobic, or 5% CO2). Medium containing daptomycin was supplemented with 1.25 mM CaCl2. The MIC was defined as the lowest concentration at which no growth was observed.</p><!><p>Bacteria were adjusted to 108 CFU ml−1 in TSB (S. aureus) supplemented with antibiotics at 10× MIC. For aerobic incubation, 3 ml of medium was inoculated in 30-ml universal tubes and incubated with shaking at 180 rpm. For anaerobic conditions, 6 ml of prereduced medium in 7-ml bijou tubes was inoculated and incubated statically in an anaerobic cabinet. Cultures were incubated at 37°C, and bacterial viability was determined by CFU counts. Culture medium containing daptomycin was supplemented with 1.25 mM CaCl2. Survival was calculated as a percentage of the number of bacteria in the starting inoculum.</p><!><p>ROS production was detected and quantified using 2′,7′-dichlorodihydrofluorescein diacetate (H2DCFDA) using a 96-well-based assay as described previously (30). Bacteria (∼3.33 × 108 CFU ml−1) were incubated in TSB with shaking at 37°C. The growth medium was supplemented with 25 μM H2DCFDA and antibiotics at various concentrations, and OD600 and fluorescence (excitation, 495 nm; emission, 525 nm) were quantified every 1,000 s (∼17 min). Fluorescence data were normalized against OD600 values to account for differences in bacterial growth between antibiotic concentrations.</p><!><p>Data are represented as the mean or median from three or more independent experiments and analyzed by Student's t test or one-way or two-way analysis of variance (ANOVA) corrected for multiple comparisons, as described in the figure legends. For each experiment, "n" refers to the number of independent biological replicates. P < 0.05 was considered significant between data points. Statistical analyses and area under the curve values were calculated using GraphPad Prism 7 for Windows.</p>

PubMed Open Access

Candidate High-Resolution Mass Spectrometry-Based Reference Method for the Quantification of Procalcitonin in Human Serum Using a Characterized Recombinant Protein as a Primary Calibrator

Procalcitonin (PCT) is a widely used biomarker for rapid sepsis diagnosis and antibiotic stewardship. Variability of results in commercial assays has highlighted the need for standardization of PCT measurements. An antibody-free candidate reference measurement procedure (RMP) based on the isotope dilution mass spectrometry and protein calibration approach was developed and validated to quantify PCT in human serum. The method allows quantification of PCT from 0.25 to 13.74 μg/L (R > 0.998) with extension up to 132 μg/L after dilution of samples with PCT concentration above 13.74 μg/L. Intraday bias was between −3.3 and +5.7%, and interday bias was between −3.0 and −0.7%. Intraday precision was below 5.1%, and interday precision was below 4.0%. The candidate RMP was successfully applied to the absolute quantification of PCT in five frozen human serum pools. A recombinant PCT used as a primary calibrator was characterized by high-resolution mass spectrometry and amino acid analysis to establish traceability of the results to the SI units. This candidate RMP is fit to assign target values to secondary certified reference materials (CRMs) for further use in external quality assessment schemes to monitor the accuracy and comparability of the commercially available immunoassay results and to confirm the need for improving the harmonization of PCT assays. The candidate RMP will also be used to evaluate whether the correlation between the candidate RMP and immunoassays is sufficiently high. Overall, this candidate RMP will support reliable sepsis diagnosis and guide treatment decisions, patient monitoring, and outcomes.

candidate_high-resolution_mass_spectrometry-based_reference_method_for_the_quantification_of_procalc

<!>Chemicals and Reagents<!>Instrumentation<!>Sample Collection<!>Confirmation of Met-PCT [3–116] and Impurity Analysis<!>LC-MS Conditions for Intact Protein Analysis<!>LC-MS/MS Conditions for Top-Down Protein Analysis<!>Amino Acid Analysis (AAA)<!>Preparation of Calibration and QC Materials<!>Sample Preparation Procedure for PCT Quantification in Human Serum<!>LC-MS/MS Conditions<!>Method Validation<!>Uncertainty Evaluation of PCT Quantification in Human Serum<!>Results and Discussion<!>Confirmation of Met-PCT [3–116]<!><!>Impurity Analysis<!>Quantification of Primary Calibrator Stock Solution by AAA<!>Method Validation<!><!>Method Validation<!>Linearity<!><!>Trueness and Precision<!>Lower Limit of Quantification<!>Higher Limit of Quantification<!>Autosampler stability<!>Carryover<!>Application to the Measurement of Patient Samples<!>Evaluation of Measurement Uncertainty<!>Conclusions<!><!>Author Present Address<!>Author Contributions<!>

<p>Procalcitonin (PCT) is a recognized sepsis biomarker allowing patient stratification and antibiotic therapy management.1−3 Different clinical decision cut-offs were established (e.g., 0.5 μg/L for sepsis diagnosis and 0.25 μg/L for antibiotic initiation or discontinuation for a patient with moderate or mild illness outside ICU4). PCT measurement has been integrated into clinical guidelines and antimicrobial stewardship programs.4−6 Thus, reliable and accurate measurements of this biomarker are critical for sepsis diagnosis, guiding treatment decisions, and patient monitoring. Facing a growing demand for PCT testing, the number of commercialized assays based on different technical principles has increased considerably in recent years.7 Different studies underlined discrepancies of results provided by various commercially available PCT assays.8−11 These discrepancies may impact clinical decisions at cut-offs, leading to disease misclassification and inappropriate antibiotic treatment decision. However, the source of such variability remains unclear.12 A proposed route to improve comparability and accuracy of the results is developing reference calibration materials, which have been value-assigned with a higher-order reference measurement procedure (RMP).13−15 Such a higher-order reference measurement system is still missing for PCT. Some assays were harmonized through traceability to the Brahms PCT LIA assay, but this protocol was not adopted for all assays. Moreover, the traceability of the results to SI units has not yet been established. Having such a higher-order measurement system will pave the road toward the standardization/harmonization of PCT assays, which has been considered a high priority by the International Consortium for Harmonization of Clinical Laboratory Results.16 As a first step, an RMP would help confirm the need to improve PCT assay harmonization and evaluate if the correlation with the commercially available PCT immunoassays is suitable for standardization. In addition, an RMP will support the establishment of traceability of results to a higher-order reference, as required by ISO 17511:2020 and the European regulation 2017/746 for in vitro diagnostic devices.17,18</p><p>Thanks to their high selectivity and reproducibility, isotope dilution and mass spectrometry have been successfully implemented to develop RMPs for SI-traceable quantification of clinically relevant proteins.19−21 Three studies based on isotope dilution associated with liquid chromatography tandem mass spectrometry (ID-LC-MS/MS) were previously reported for PCT quantification in serum.22−24 Each relied on stable isotope labeled (SIL) peptides spiked in the sample after protein digestion. However, a SIL protein, spiked at the earliest stage of the sample preparation to overcome material loss or variability occurring during sample processing and digestion, is considered an ideal internal standard with the same behavior as the analyte of interest.25−28</p><p>Here, we described the development and validation of a candidate reference ID-LC-MS/MS method for the SI-traceable quantification of PCT in serum at clinically relevant concentrations using, for the first time, a recombinant protein as a primary calibrator and a SIL-recombinant protein as an internal standard (Figure 1). In addition, analytical performance in terms of trueness and precision was assessed, and the uncertainty of measurement results was evaluated. Finally, the present method was used to perform SI-traceable quantification of PCT in five pools of frozen human serum as a proof of concept for developing secondary certified reference materials (CRM).</p><p>Schematic analytical workflow for SI-traceable quantification of PCT in human serum and its uncertainty using protein-based matrix-matched calibration and labeled PCT recombinant protein as internal standard. Step 1: Preparation of calibrators and quality control (QC) materials in blank serum using the SI-traceable PCT primary calibrator after performing impurity-corrected amino acid analysis (AAA). Step 2: Preparation of patient samples by spiking labeled PCT. Step 3: Antibody-free sample preparation for calibrators, QC materials, and patient samples followed by LC-MS/MS analysis of final processed samples. Step 4: Establishment of a calibration curve using a linear regression model to determine PCT concentration measured per peptide SAL or FHT. Step 5: Determination of PCT concentration based on two selected peptides, and its associated uncertainty was estimated by combining all sources of uncertainty from steps 1 to 4 (ucal, usam, uprec, ulin).</p><!><p>Amino acid CRMs from NMIJ, chemicals, and reagents were described in a previous study24 and are detailed in document 1 of the Supporting Information.</p><p>The recombinant protein methionine-procalcitonin 3–116 (Met-PCT [3–116]) and the isotopically labeled protein methionine-procalcitonin 3–116 (SIL protein Met-PCT [3–116] labeled on arginine (R[13C6,15N4]) and lysine (K[13C4,15N]) residues) at a concentration of ∼1 g/L (Tris/NaCl buffer solution) were purchased from Promise Advanced Proteomics (Grenoble, France). The supplier purified Met-PCT [3–116] using three orthogonal techniques: ion exchange, reverse-phase, and size-exclusion chromatography.</p><!><p>Amino acid analyses, intact mass LC-MS measurements of the primary calibrator Met-PCT [3–116], and LC-MS/MS analyses of the digested serum samples were performed on a Thermo Scientific Dionex Ultimate 3000 ultraperformance liquid chromatography system coupled to a Thermo Scientific Q Exactive Focus hybrid Quadrupole-Orbitrap mass spectrometer (Thermo Scientific, Waltham, MA).</p><p>The top-down analysis of Met-PCT [3–116] was conducted on a Thermo Scientific Dionex RSLC Ultimate 3000 nano-LC system coupled to a Thermo Scientific Orbitrap Eclipse Tribrid mass spectrometer (Thermo Scientific).</p><!><p>University Hospital Montpellier (Montpellier, France) provided five pools of deidentified patient serum samples with different PCT concentrations. Each pool was produced by pooling 12 single frozen leftovers (collected in dry tubes) obtained from sepsis or septic shock patients. PCT concentration of these serum pools was determined at the clinical chemistry laboratory of Montpellier Hospital using the Brahms PCT sensitive Kryptor immunoassay (Compact Plus). The serum pools were then immediately stored at −80 °C ± 10 °C until analysis.</p><!><p>Met-PCT [3–116] protein and its impurities were characterized using two complementary approaches: high-resolution MS analyses of intact protein for impurity identification and top-down MS analyses using multiple fragmentation modes for protein characterization.</p><!><p>A Met-PCT [3–116] solution (∼0.1 g/L in H2O/ACN 95:5, v/v) was analyzed for potential impurity identification in LC-MS, operated in electrospray positive mode (Q Exactive Focus). LC was performed on a C4 analytical column (150 mm × 1 mm, 5 μm, BioBasic-4, Thermo Scientific). The mobile phase consisted of 0.1% FA (v/v) in water (solvent A) and 0.1% (v/v) FA in acetonitrile (solvent B). The separation was achieved using a linear gradient from 25 to 60% of B over 37 min at a 40 μL/min flow rate. The experimental MS parameters are summarized in Table S1 of the Supporting Information.</p><!><p>The Met-PCT [3–116] solution at 0.1 g/L was also analyzed on a nanoelectrospray tribrid Eclipse instrument. LC separation was performed on a C4 analytical column (75 μm × 150 mm, 5 μm, Acclaim PepMap 300, Thermo Scientific). The mobile phase consisted of H2O/ACN 98:2 (v/v), 0.1% FA (solvent A) and H2O/ACN 10:90 (v/v), 0.1% FA (solvent B). The separation was achieved using a linear gradient from 25 to 60% B in 37 min at a 300 nL/min flow rate. The sample was analyzed in data-dependent acquisition mode, using four different fragmentation modes (HCD, EThDC, CID, and UVPD). The experimental MS parameters are summarized in Table S2 of the Supporting Information.</p><!><p>The SI-traceable quantification of Met-PCT [3–116] standard was performed by amino acid analysis (AAA) as described previously.24 Briefly, the Met-PCT [3–116] content was determined by quantifying phenylalanine, proline, valine, and leucine by ID-LC-MS using a five-point calibration curve after gas-phase hydrolysis in acidic conditions. Conditions of the gas-phase hydrolysis were optimized by carrying out the gas-phase hydrolysis (ELDEX Workstation) at different conditions: 110, 130, or 150 °C for 40 h and 130 °C for 24 and 72 h. In each condition, four processed replicates were performed. The amino acid mix was analyzed on the Q Exactive Focus instrument in the selected ion monitoring mode. Isocratic separation was performed using a C18 column (150 mm × 2.1 mm, 1.7 μm, BEH C18, Waters) in H2O/ACN/FA 98:2:0.1 (v/v/v). The final protein concentration was estimated as the average of the four amino acid titrations determined from optimal hydrolysis conditions with 29 processed replicates over six independent experiments.</p><!><p>Calibration and quality control (QC) materials were blank serum samples spiked with the recombinant Met-PCT [3–116]. Detailed preparation is available in document 2 of the Supporting Information. Briefly, a set of six calibration samples (concentration of Met-PCT [3–116] ranging from ∼0.25 to 13.74 μg/L) and three QC samples (concentration of Met-PCT [3–116] 1.0, 4.0, and ∼9.0 μg/L) was prepared by spiking Met-PCT [3–116] at different concentrations and SIL Met-PCT [3–116] at ∼1.7 μg/L in blank serum gravimetrically. The mass ratio between unlabeled and SIL protein ranged from ∼0.15 to 7.5 for calibration materials and 0.51–5.08 for QC materials. The calibration and QC materials were aliquoted (500 μL) and stored at −80 °C ± 10 °C .</p><p>QC materials used to determine the lower limit of quantification (LLOQ) were prepared by spiking Met-PCT [3–116] at two different concentrations (∼0.25 and ∼0.50 μg/L) and SIL Met-PCT [3–116] at ∼1.7 μg/L in blank serum. The mass ratios between unlabeled and SIL protein were 0.15 and 0.30.</p><p>QC materials used to determine the higher limit of quantification (HLOQ) were prepared by spiking Met-PCT [3–116] at ∼132 μg/L in blank serum. The sample was then diluted in blank serum to a concentration of ∼6.5 μg/L followed by the addition of SIL Met-PCT [3–116] at ∼1.7 μg/L. The mass ratio between unlabeled and SIL protein was 3.8.</p><!><p>Patient samples were prepared gravimetrically by mixing about 480 μL of the sample with 20 μL of SIL Met-PCT [3–116] at a concentration of ∼40 μg/L to reach a final concentration of ∼1.5 μg/L. The calibration, QC materials, and patient samples were processed as described previously.24 Briefly, 500 μL of serum was subjected to protein denaturation using SDC detergent and precipitated using acetonitrile. Next, the supernatant was diluted and purified on a C18 solid-phase extraction (SPE) cartridge (HLB C18, Waters). Extracted proteins were reduced (DTT), alkylated (IAA), and digested with 4.6 μg trypsin gold. Finally, the tryptic digest was purified on an HLB C18 SPE cartridge. The elution buffer was evaporated to dryness in a centrifugal vacuum concentrator, reconstituted with 100 μL of 0.1% formic acid, 2% MeOH in water (v/v/v) (noted final extract), and stored at −20 °C ± 5 °C until LC-MS/MS analysis.</p><!><p>The proteolytic digests were analyzed in parallel reaction monitoring (PRM) mode29 on the Q Exactive Focus instrument. Briefly, tryptic peptides were separated on a C18 analytical column (1 mm × 150 mm, 3 μm, Acclaim PepMap 100, Thermo Scientific) using 0.05% AA in water (v/v) as solvent A and 0.05% AA in methanol (v/v) as solvent B at a flow rate of 80 μL/min. Peptides were eluted with the following gradient of mobile phase B: 2% for 2 min, linear from 2 to 22% in 8 min, linear from 22 to 38% in 1 min, linear from 38 to 42% in 14 min, and from 42 to 98% in 1 min.</p><p>SALESSPADPATLSEDEAR (noted SAL) and FHTFPQTAIGVGAPGK (noted FHT) proteotypic peptides have been previously selected for PCT quantification.24 Two transitions per peptide were selected, one used as peptide quantifier and another as peptide qualifier (see Table S3, Supporting Information). Raw data were processed with Xcalibur software v4.1 (Thermo Scientific). Signal extraction in the LC profile was performed within a mass tolerance of 10 ppm for PRM data.</p><!><p>After defining the calibration curve, the analytical performance for PCT quantification in human serum using a protein-based calibration approach with SIL protein as internal standard was validated based on matrix-matched material according to FDA and EMA guidelines30,31 regarding linearity, trueness, precision, dilution, autosampler stability of extracted peptides, and carryover. The trueness and precision were performed using matrix-matched QC materials in three processed replicates over three independent experiments using freshly prepared calibrators for each experiment. Protocol and criteria for method validation are described in document 3 of the Supporting Information.</p><!><p>Uncertainty was evaluated according to the ISO Guide 98-3GUM using the bottom-up approach.32 The combined uncertainty of the experimental values for QC, LLOQ, HLOQ levels, and patient pools for individual concentration obtained per peptide (uSAL and uFHT) was calculated by propagating the uncertainty associated with all relevant sources of measurement uncertainty, including primary calibrator uncertainty and gravimetric preparation of calibrators (ucal), gravimetric preparation of samples (usam), regression model (ulin), and intermediate precision (uprec).</p><p>The uncertainty (umean) of mean concentration was calculated by combining the uncertainties of two individual concentrations per peptide.The final uncertainty (ufinal) of mean concentration was calculated by taking into account the uncertainty between peptides (uinterpeptide) obtained from analysis of variance (ANOVA).Finally, the expanded uncertainty (U) was expressed by multiplying the final uncertainty with a coverage factor k = 2, corresponding to a confidence level of ∼95%. The relative expanded uncertainty (%) was expressed by the ratio between the expanded uncertainty and the measurement result.</p><!><p>Developing a candidate reference measurement to quantify PCT in human serum requires each analytical process step to be metrologically traceable to SI units. Figure 1 illustrates the workflow for SI-traceable quantification of PCT in human serum of the developed method.</p><!><p>A total ion chromatogram obtained after LC-MS analysis of the primary calibrator is presented in Figure 2A. The MS spectrum corresponding to the major chromatographic peak at 15.8 min is presented in Figure 2B. A monoisotopic mass of 12 749.12 Da was identified, which agreed well with the theoretical value of Met-PCT: 12749.11 Da (Δmass −0.39 ppm). The identity of Met-PCT [3–116] was also confirmed by top-down MS/MS analysis. By combining four fragmentation modes on the charge state 14 of the major compound in buffer stock solution (m/z = 912.2293), 60% of Met-PCT [3–116] sequence coverage was obtained and the identity of the major compound in the buffer stock solution of the primary calibrator was confirmed (see Figure S1, Supporting Information).</p><!><p>Characterization of the Met-PCT [3–116] primary calibrator. (A) Extracted ion chromatogram obtained by injecting 5 μg of the protein standard. The base peak represents Met-PCT [3–116]. 1—Oxidized form; 2—acetylated form; and *—monocharged compounds. (B) Multicharged mass spectrum corresponding to the base peak Met-PCT [3–116] at 15.8 min.</p><!><p>The analytical challenge of developing a protein-based primary calibrator is identifying and quantifying all impurities impacting either AAA or LC-MS/MS quantification of PCT in a matrix, which can be burdensome. To limit this issue as much as possible, the primary protein calibrator should be highly purified. Upon request to the Promise manufacturer, the recombinant PCT protein was subjected to three orthogonal chromatographic strategies: ion exchange, reverse-phase, and size-exclusion chromatography. However, as some impurities are similar (e.g., proteoforms), a 100% pure recombinant protein is almost unattainable, even with a high cost in terms of yield. Figure 2A shows the presence of additional peaks around the peak of Met-PCT. The most intense impurities identified by accurate mass measurement were oxidized PCT, acetylated PCT, and four truncated forms of PCT (see Table S6, Supporting Information). The associated peak area obtained from extracted chromatogram after deconvolution of different species was compared. The relative areas of oxidized PCT and acetylated PCT (in the stock solution) peaks correspond to 5.07 and 2.14% of the Met-PCT peak area.</p><p>The relative peak areas of truncated forms of PCT were less than 0.6%. The top-down analysis confirmed that acetylation occurs on one of the three N-terminal residues of PCT (Met–Phe–Arg). Thus, this impurity affects neither the AAA results nor the LC-MS/MS quantification of the targeted peptides SAL or FHT. Regarding the oxidized form of Met-PCT, it was not yet possible to unambiguously identify the oxidation site based on top-down analysis. However, no oxidized form of peptide SAL or FHT (±5 min from retention time of peptide SAL or FHT) was detected based on LC-MS analysis of samples after trypsin digestion of Met-PCT in the buffer. Moreover, the most frequent residues subject to oxidation are methionine and cysteine: they are not among the residues targeted by AAA (phenylalanine, proline, valine, and leucine), and they are not found in the two targeted peptide sequences. The truncated forms observed with a delta mass of about −1300 Da had a retention time close to the recombinant protein one. The purification steps performed by the supplier of the recombinant protein, including size-exclusion chromatography, suggest that these low abundant truncated forms were artifacts generated during the LC-MS analysis and were not present in the original sample. The two modified forms with delta mass of +29 and −17 Da coeluted with the recombinant protein. The absence of chromatographic separation of these modified forms from the recombinant protein when using different elution gradients also suggests that these low abundant forms are artifacts generated during the LC-MS analysis. Therefore, the raw amino acid analysis results were not corrected, highlighting the benefits of working with highly purified materials.</p><!><p>AAA determined the concentration of the primary calibrator to establish the traceability of the results to the SI units. After optimizing the conditions of gas-phase hydrolysis, the highest concentration measured by AAA, with the lowest variation between the four amino acids (leucine, phenylalanine, proline, valine), was obtained at 130 °C for 40 h (see Table S4, Supporting Information). These experimental conditions allowed hydrolyzing the valine amide liaison, challenging to cleave without degrading the amino acids produced. These optimized conditions were then applied to the quantification of the four amino acids in the primary stock solution of Met-PCT [3–116] (N = 29). The mass fraction of Met-PCT [3–116] (average from four amino acid results) in the stock solution was 807 ± 72 μg/g (k = 2) (see Table S5, Supporting Information).</p><!><p>To ensure the accuracy of PCT concentration, the identification of each peptide was verified based on PRM LC-MS/MS data. The extracted ion chromatograms showed the coelution of two selected product ions, with the peptide of interest and its internal standard. The identification of the SAL peptide is presented in Figure 3.</p><!><p>Identification of the SAL peptide for PCT quantification in human serum. (A) MS/MS PRM spectrum of targeted precursor ion SAL2+ (selected product ions y13+ and y10+ for quantification and confirmation in red) in processed human serum spiked with a PCT at 5 μg/L; (B) extracted ion chromatograms obtained when measuring blank serum spiked with a PCT at LLOQ level showing coelution of two selected product ions; (C) extracted ion chromatograms obtained when measuring blank serum spiked with a PCT at the LLOQ level and labeled PCT at 1.5 μg/L showing coelution of the SAL peptide and its internal standard. Precursor ions were isolated within an isolation window of 1.5 m/z. Raw chromatograms were extracted without smoothing.</p><!><p>Most product ions of the SAL peptide were identified in PRM data obtained from processed human serum. While the FHT peptide contains two residues of proline, which readily generates internal fragmentation from its N-terminal side, detected ions could not be attributed only to the primary peptide backbone fragmentation (see Figure S2, Supporting Information). Therefore, the two most intense product ions were selected, one for quantification and another for confirmation. For PRM data generated from triplicated analyses, the peak areas of selected transition were then extracted to establish a calibration curve based on isotope dilution and quantitative analysis.</p><!><p>The regression model is linear over the range 0.25–13.74 μg/L for SAL and 0.47–13.74 μg/L for FHT (Figure 4A). The Pearson regression coefficient was above 0.998 for both peptides. Detailed data obtained for each peptide from three independent days are presented in Table S7 and Figure S3 of the Supporting Information.</p><!><p>Method validation and estimation of uncertainties for PCT quantification in human serum. (A) Linearity of the signal response obtained with nonzero protein-based matrix-matched calibrators for the SAL peptide and FHT peptide. Linearity results obtained from three independent experiments (linearity equation) were obtained by averaging three independent experiments. (B) Intraday (n = 3) and interday (n = 3, 3 days) trueness and precision at three QC levels. Blue lines represent the acceptation limit ± 15% for the trueness value. Precision was expressed as an error bar. (C) Estimation of uncertainties of PCT concentration of QC materials. Expanded uncertainty was expressed by an error bar. Relative expanded uncertainty was presented by the dashed line (SAL peptide), dotted line (FHT peptide), and solid line (mean of two peptides). (D) Method application to quantify PCT concentration in patient pool samples compared to those obtained by immunoassay (in solid blue line). Expanded uncertainty was expressed by an error bar. Relative expanded uncertainty was presented by the black dashed line (SAL peptide), black dotted line (FHT peptide), and solid black line (mean of two peptides).</p><!><p>Trueness and precision of the method's validation are presented in Figure 4B and detailed in Table S8. The intraday (n = 3) bias and interday (n = 3, 3 days) bias ranged from −2.8 to 1.6 and −1.2 to 0.2% for SAL and −7.3 to 8.0 and −6.2 to 6.3% for FHT. The intraday precision and intermediate precision (interday) were below 3.3 and 2.3% for SAL and 9.5 and 7.6% for FHT. For all QC materials, intraday bias was between −3.3 and +5.7%, and interday bias was between −3.0 and −0.7% for the mean concentration. Intraday precision was below 5.1%, and interday precision was below 4.0% for QC materials.</p><!><p>Extracted ion chromatograms from human serum at the LLOQ level are presented in Figure 3 and Figure S2. The LLOQ level was 0.25 μg/L for SAL and 0.47 μg/L for FHT. Therefore, PCT concentration was calculated by the average of two concentrations obtained from two peptides for concentration above 0.47 μg/L and by SAL only below this limit.</p><p>The mean bias and precision CV were 4.2 and 5.5%, respectively, for a concentration of 0.25 μg/L and −0.7 and 7.5% for a concentration of 0.51 μg/L.</p><!><p>The HLOQ quantification at a concentration above the highest calibrators was quantified after 20× dilution. It showed bias and precision of 1.6 and 2.3% for SAL and 5.5 and 0.2% for FHT. The method can quantify PCT for a concentration up to 132 μg/L.</p><!><p>Autosampler stability of 7 days at +7 °C was demonstrated for all QC levels (bias from the initial concentration <20%). The two peptide concentrations remained stable in the autosampler.</p><!><p>No carryover was observed for the two peptides.</p><p>The present method uses a SIL protein as an internal standard that differs from the other LC-MS/MS methods developed to quantify PCT.22−24 The SIL protein added at the beginning of the sample preparation process is ideal for protein quantification with the bottom-up approach.25,27 It compensates for the bias caused by incomplete digestion or material loss during sample preparation and LC-MS/MS analysis.24,26 These limitations have been underlined in a previous study in which PCT was quantified through peptide-based calibration using SIL peptides as internal standards.24 A correction factor has been applied to compensate for digestion incompleteness and material loss before the digestion step. Moreover, the FHT peptide could not be used as a quantifier peptide as it may be subject to miscleavage not corrected by the approach used. In the present study, both endogenous and SIL-PCT are simultaneously proteolyzed. PCT quantification with low bias and high precision was archived without using a correction factor when quantifying both SAL and FHT peptides for concentrations above 0.47 μg/L, allowing to increase the specificity of the method. These two selected peptides are located in two different regions of PCT and are not in the same region of epitopes usually targeted by commercially available immunoassays.24</p><p>Furthermore, as reported in the literature, PCT is present under three different isoforms characterized by the cleavage of one or two N-terminal amino acids.7 Our method quantifies the total serum PCT, including these three isoforms as measured by most commercially available immunoassays.</p><p>The calibration range, HLOQ, and LLOQ of the method encompass the clinical range of PCT concentrations found in serum from sepsis or septic shock patients. Therefore, the candidate RMP is intended to be used to measure PCT in sepsis patients and support activities of the IFCC working group on the standardization of PCT assays (WG-PCT) to monitor the accuracy and comparability of immunoassay results and evaluate if the correlation between available immunoassays at different clinical cut-off concentrations is sufficient to conduct standardization. While the analytical sensitivity of the candidate RMP covers almost all of the ranges of concentrations measured by immunoassays, if the standardization of the PCT assay is confirmed to be needed and feasible, further studies are required to improve LLOQ to cover LLOQ of all commercial immunoassays (0.02–0.2 μg/L). This improvement could be achieved through instrumental developments (e.g., reducing LC flow rates and dimensions, using a more sensitive mass spectrometer) and improving the sample preparation step (e.g., using immunoenrichment). Miniaturization of sample handling could suffer from low reproducibility when analyzing low abundant analytes in complex and concentrated samples such as serum.33</p><!><p>As a proof of concept to evaluate how results from the candidate RMP compare with those from immunoassays, the developed method quantification was further applied to five pools of patient samples on two independent experiments. The interassay precision ranged from 1.5 to 7.7% and from 6.5 to 10.5% for SAL and FHT, respectively (Figure 4D and Table S11, Supporting Information). The mean concentration was obtained with a precision below 5.1%. The concentration measured by immunoassay was higher than the one obtained by ID-LC-MS/MS, with a relative difference between ID-LC-MS/MS and immunoassay ranging from 18 to 55%.</p><p>This relative difference observed between LC-MS/MS and the immunoassay could be explained by differences in calibration and/or differences in specificity potentially caused by cross-reactivity issues. Although most PCT immunoassays employ two antibodies targeting different regions of PCT, it cannot be excluded that immunoassays measure other forms than the three full-length isoforms of PCT. However, it should be noted that only five samples were measured, and only one immunoassay was involved. Therefore, this did not allow making a definitive explanation and advocates for a larger study. Indeed, the result obtained from this assay could be different from the other assays because PCT assays were reported to employ different types of antibodies with different epitope specificities toward the multiple molecular forms of PCT.7 The correlation between commercial immunoassays and the candidate RMP should be established soon for all available immunoassays and not only Brahms PCT-sensitive Kryptor immunoassays. Also, a larger number of samples of proven commutability are required to establish a correlation, which was not demonstrated in the present study. These studies will be designed by IFCC WG-PCT and will help to confirm the magnitude and investigate the origin of differences observed in PCT concentration.</p><!><p>The uncertainty of the calibrator and the linear regression are presented in Tables S9 and S10 in the Supporting Information for each calibrator level. The relative expanded uncertainty of each concentration level of QC materials and pools of patient serum samples are presented in Figure 4C,D, respectively, and summarized in Table S12 of the Supporting Information. For all levels, the relative expanded uncertainty (k = 2) was below 18% and below 30% when using the SAL peptide and FHT peptide, respectively. The relative uncertainties were lower for the results obtained using the SAL peptide. The relative expanded uncertainty (k = 2) ranged from 7 to 18% for mean concentration, except for LLOQ FHT and Pool3 samples (about 24%). The relative contributions of the different components to the final uncertainty of individual concentration per peptide are presented in Figure S4 of the Supporting Information. The uncertainties associated with the value assignment of the primary calibrator (ucal), the linearity of the calibration curve (ulin), and the precision of measurements (uprec) appeared as the primary sources of measurement uncertainty. Their relative contributions varied depending on PCT concentration. The main contribution to the final uncertainty for low PCT concentrations was the uncertainty associated with the linear regression or the precision experiment, while for high PCT concentration, it was the uncertainty associated with the calibrator's purity. The uncertainty of the precision experiment was higher for the FHT than for the SAL peptide.</p><p>To ensure that laboratory measurements are clinically usable, it has been recommended that no more than one-third of the maximum allowable uncertainty of routine assays should be consumed by higher-order references.34 In addition to the correct implementation of calibration traceability, the achievement of appropriate analytical performance specifications for RMPs and CRMs is essential but can be challenging for low abundant proteins like PCT. Relative expanded uncertainties of results obtained with our method are generally 7–18%, but they reached up to 24% in some cases (low PCT concentration level). These uncertainties are probably too high for assigning a target value to a standalone CRM but are acceptable if this remains an isolated event when the RMP is used to measure a panel of patient samples (e.g., correlation study between available immunoassays and candidate RMP). As high uncertainties might lead to a modest correlation between the candidate RPM and the immunoassays and might compromise the ability to properly evaluate the accuracy of immunoassays, reducing measurement uncertainties would be beneficial. The major source of uncertainty at low PCT concentration was the uncertainty associated with the linear regression (up to 54%): this source of uncertainty could be reduced by employing a narrow working concentration instead of a large concentration range (0.15–7.5 in mass ratio).35 This may be difficult to handle when a large number of samples of unknown PCT concentrations over an expanded range of concentration should be measured (correlation study between available immunoassays and candidate RMP) but very much manageable in the case of a value assignment of pairs of CRMs at a given concentration. It should also be noted that the high uncertainty observed in one pooled sample with low PCT concentration was caused by variability between concentrations of the two measured proteotypic peptides. As this was observed only in one pool of patient samples, a more extensive study involving a larger number of pooled samples and single donation samples, as the one planned to assess standardization feasibility, will help further demonstrate the magnitude and source of uncertainties at this PCT-level concentration.</p><!><p>We developed and validated an ID-LC-MS/MS method for the SI-traceable quantification of PCT in human serum covering most clinical cut-off concentrations. We used a protein-based calibration strategy relying on a PCT recombinant protein as primary calibrator, and the corresponding isotope-labeled recombinant protein as an internal standard. Using recombinant protein as the primary calibrator and internal standard improved the method's accuracy compared to a previously developed method based on peptide calibrators. A correction factor is not required anymore with the present method, as the protein-based internal standard accounts for incomplete digestion and material loss during sample preparation. The present method thus appears suitable to determine PCT concentration in external quality assessment materials and secondary CRMs that could be used to monitor the accuracy and comparability of commercially available immunoassays for PCT at clinically relevant concentrations. The candidate RMP will support the activities of IFCC WG-PCT and especially evaluate the feasibility for the standardizing PCT assays.</p><!><p>Description of the chemicals and reagents, preparation of solutions, method validation; tables: MS parameters for intact analysis and top-down protein analysis, amino acid sequences and PRM parameters of selected peptides, AAA results obtained from optimization and final optimized conditions, main impurities of the stock solution, detailed data for the calibration curve, intraday and interday bias and precision, uncertainties of the calibrators, uncertainties of the linear regression, peptide and PCT concentration of the patient serum samples, relative expanded uncertainty of measurement results; figures: Met-PCT [3–116] characterization by top-down analysis, identification of the FHT peptide for PCT quantification, linearity of PCT quantification by LC-MS/MS and uncertainty estimation (PDF)</p><p>ac1c03061_si_001.pdf</p><!><p>⊥ Department of Laboratory Medicine and Pathology, University of Washington, Seattle, Washington, United States</p><!><p># A.B. and J.V. contributed equally to this work and both should be considered as senior authors. All authors confirmed they have contributed to the content of this paper and have approved the final article.</p><!><p>This work was supported by the European Metrology Programme for Innovation and Research (EMPIR) joint research projects [15HLT07] "AntiMicroResist" and [18HLT03] "SEPTIMET", which have received funding from the EMPIR program cofinanced by the Participating States and the European Union's Horizon 2020 research and innovation program. H.-H.H. and M.D.-M. received a CIFRE doctoral fellowship provided by ANRT (Association Nationale de la Recherche et de la Technologie). Mass spectrometry equipment was supported by SESAME 2018 no. EX039194 of Région Ile-de-France and IBiSA 2019 grants.</p><p>The authors declare no competing financial interest.</p>

PubMed Open Access

Asymmetric Reductive Dicarbofunctionalization of Alkenes via Nickel Catalysis

Alkenes are an appealing functional group that can be transformed into a variety of structures. Transition-metal catalyzed dicarbofunctionalization of alkenes can efficiently afford products with complex substitution patterns from simple substrates. Under reductive conditions, this transformation can be achieved while avoiding stoichiometric organometallic reagents. Asymmetric difunctionalization of alkenes has been underdeveloped, in spite of its potential synthetic utility. Herein, we present a summary of our efforts to control enantioselectivity for alkene diarylation with a nickel catalyst. This reaction is useful for preparing triarylethanes. The selectivity is enhanced by an N-oxyl radical additive.

asymmetric_reductive_dicarbofunctionalization_of_alkenes_via_nickel_catalysis

Introduction<!>Reaction Development<!>Mechanistic Investigation

<p>Dicarbofunctionalization of alkenes represents a compelling approach to efficiently access substituted, saturated molecular scaffolds from simple, abundant starting materials.1 Traditional methods to difunctionalize α,β-unsaturated carbonyl compounds have been achieved through a two-step process, in which a nucleophile undergoes conjugate addition with an activated alkene and the resulting enolate is then trapped with an electrophile (Scheme 1A).2 This strategy is limited to activated alkenes in conjugation with a carbonyl group, and functional groups must be compatible with the highly reactive organometallic nucleophiles.</p><p>Transition metal catalysis has allowed for one-step difunctionalization of alkenes using a nucleophile and an electrophile (Scheme 1B).1a This redox-neutral strategy has employed palladium catalysts and aryl or vinyl nucleophiles and electrophiles to achieve the regioselective difunctionalization of vinylarenes,3 dienes,4 and, using a directing group, unactivated alkenes.5 A few asymmetric variants have been reported, but the scope has been restricted to diene substrates6 or intramolecular reactions.7 Nickel catalysts have expanded the scope of nucleophiles and electrophiles to include alkyl groups,8 but asymmetric reactions have been restricted to intramolecular examples.9</p><p>Reductive coupling reactions have recently emerged to eschew stoichiometric nucleophiles and thus expand the substrate scope to include moieties sensitive to organometallic nucleophiles.10 Nevado developed a Ni-catalyzed reductive dicarbofunctionalization of alkenes using alkyl and aryl iodides in combination with tetrakis(dimethylamino)ethylene (TDAE) as a reductant (Scheme 1C).11 In this reaction, Ni mediates the formation of an alkyl radical, which is added to the alkene. Combination of the radical with Ni forms a Ni-alkyl intermediate, which undergoes reductive elimination to afford the product. While the C(sp3) electrophile is restricted to tertiary alkyl iodides, this method is notable for its mild conditions.</p><p>This example highlights the radical reactivity of nickel compared to its Group 10 congener palladium, which favors two-electron pathways.12 The access to radical intermediates by nickel catalysts provides new opportunities for controlling enantioselectivity. In particular, for palladium-catalyzed arylation of alkenes, the enantioselectivity is dictated by either the alkene coordination or migratory insertion step (Scheme 2, steps i and ii),13 while for a nickel-mediated pathway involving single-electron redox chemistry, radical capture by nickel (v) or reductive elimination (vi) could be enantio-determining.14 We applied the radical properties of nickel catalysts to develop an intermolecular asymmetric difunctionalization of alkenes (Scheme 1D).</p><!><p>We began our investigation by developing conditions for racemic diarylation of styrene using bromobenzene. A dimeric nickel(I) catalyst has been previously shown to enable the carbofunctionalization of dienes15, and we were able to utilize this catalyst to conduct diarylation on styrene to generate 1,1,2-triphenylethane in good yield, so these conditions were used as a basis from which an asymmetric reaction could be evolved (Equation 1). Polar, aprotic solvents such as N,N-dimethylacetamide (DMA) or dimethylpropyleneurea (DMPU) were necessary in order to achieve high yields of the desired difunctionalization products. Heterogeneous reductants such as manganese and magnesium offered some diarylation product, while zinc powder gave the highest yield. Among various classes of chiral ligands, bioxazoline (biOx) ligands, commonly used in nickel catalysis,16 proved to be the most effective for achieving high yield and enantioselectivity.</p><p>Racemic diarylation of styrene with bromobenzene</p><p>We observed a marked decrease in enantioselectivity from 38% to only 2% when the styrene substrate was distilled prior to use rather than used directly from its commercial bottle (Table 1, entries 1 and 2). Due to their propensity for auto-polymerization, styrenes are typically stabilized with radical inhibitors such as 2,6-di-tert-butyl-4-methylphenol (BHT). We speculated that the presence of such an additive, or the stabilized O-radical resultant from quenching of a radical by BHT, could be responsible for this discrepancy. While the use of BHT did not result in higher enantioselectivity (entry 3), the addition of (2,2,6,6-tetramethylpiperidin-1-yl)oxyl (TEMPO) had a dramatic effect, raising the enantiomeric excess to 49% (entry 4). Furthermore, equimolar TEMPO loading with respect to nickel was paramount in order to achieve optimal results, with lower loading engendering worse enantioselectivity (entry 5), while higher loading shut down the reaction (entry 6).</p><p>These observations led us to evaluate various accessible N-oxyl radicals. The enantioselectivity exhibits a strong correlation with the percent buried volume (%Vbur), a computationally derived steric descriptor,17 of the N-oxyl radical. Less bulky N-oxyl additives gave higher enantioselectivities (Figure 1). 9-Azabicyclo[3.3.1]nonane N-oxyl (AB-NO) delivered the difunctionalized product with the highest enantiomeric excess (Table 1, entry 7). Enantioselectivity was further improved by modifying the substituent on the biOx ligand, lowering the reaction temperature, and using a mixture of DMPU and tetrahydrofuran (THF) as the solvent (entry 8).</p><p>With optimized conditions in hand, we explored the substrate scope of this reaction (Scheme 3).18 Electron-donating and -withdrawing substituents had little effect on the enantioselectivity, although higher catalyst loadings were required to obtain good yields for electron-poor vinylarene substrates. Disubstituted alkenes were not well-tolerated, with 1,1-disubstituted vinylarenes showing no reactivity, although diarylation of indene gave a 17% yield of the trans-difunctionalized product.</p><!><p>We gained insight into the mechanism by considering several observations. First, formation of a dimer at the benzylic position suggests the presence of a benzylic radical (Scheme 4A). Second, the diarylation of indene resulted solely in the formation of the trans-diarylation product (Scheme 4B). One would expect a syn-addition from a typical migratory insertion, so the observed product diastereomer suggests the formation of a radical that combines with nickel on its less-hindered face (Scheme 4C). Third, the presence of excess N-oxyl radical with respect to nickel inhibited the reaction, mostly likely via interference of radical intermediates. Finally, no evident linear correlation was found between the observed enantiomeric ratio and the electronic effect of vinylarenes, suggesting that the enantio-determining step could involve a radical species.</p><p>We performed control experiments to investigate the possible in situ generation of an organozinc reagent, which could afford the difunctionalized alkene as in redox-neutral strategies. When phenylzinc chloride was used in place of bromobenzene in the absence of zinc powder, no diarylation product was observed. This experiment suggests that direct oxidative addition of zinc to the aryl bromide does not enable the reaction.</p><p>A mechanistic hypothesis consistent with our findings is presented in Scheme 5. Oxidative addition and reduction of a nickel(I) bromide species to the aryl bromide electrophile results in an arylnickel(I) species.19 This intermediate undergoes migratory insertion with the alkene substrate. Oxidative addition of another molecule of aryl bromide to the arylnickel(I) makes a nickel(III) species. This open-shell intermediate can eject a stabilized benzylic radical, which accounts for our observations of both the benzylic dimer byproduct and the trans-diastereoselectivity with indene. This reversible radical ejection is consistent with computational work performed by Kozlowski and co-workers.14 Recombination of this radical with the chiral nickel catalyst followed by reductive elimination furnishes the diarylated product and regenerates the nickel(I) bromide.</p>

PubMed Author Manuscript

Prion seeded conversion and amplification assays

The conversion of the normal prion protein (PrPC) into its misfolded, aggregation-prone and infectious (prion) isoform is central to the progression of transmissible spongiform encephalopathies (TSEs) or prion diseases. Since the initial development of a cell free PrP conversion reaction, striking progress has been made in the development of much more continuous prion-induced conversion and amplification reactions. These studies have provided major insights into the molecular underpinnings of prion propagation and enabled the development of ultra-sensitive tests for prions and prion disease diagnosis. This chapter will provide an overview of such reactions and the practical and fundamental consequences of their development.

prion_seeded_conversion_and_amplification_assays

Introduction<!>Cell-free conversion assays<!>Protein-Misfolding Cyclic Amplification (PMCA) and recombinant Protein-Misfolding Cyclic Amplification (rPMCA)<!>Prion strains and species barrier studies using PMCA<!>Amyloid Seeding Assay (ASA)<!>Quaking-Induced conversion (QuIC) reactions<!>Conclusions

<p>A major issue in coping with any infectious disease is the ability to detect the responsible pathogen. In the case of the transmissible spongiform encephalopathies (TSEs) or prion diseases of mammals, it is increasingly apparent that the pathogen is a misfolded multimeric form of the host's prion protein (PrP) [15]. This infectious protein, PrPSc, can instigate its own propagation by binding to its normal counterpart, PrPC or PrPsen, and inducing its conversion to a form that tends to be higher in beta sheet content, polymeric, and more protease-resistant. The lack of an agent-specific nucleic acid genome negates the possibility of ultrasensitive detection of prions by nucleic acid amplification methods such as PCR. The fact that the infectious agent is mainly comprised of a host protein also restricts the use of antibody-based detection methods to those based on conformational epitopes or epitope exposure. However, the apparent seeded/templated conformational conversion mechanism of prion propagation can be exploited to detect the presence of prions. Here we summarize recent developments in the characterization and detection of prions using assays based on seeded conversion of PrPC.</p><!><p>The ability of PrPres to induce the conversion of PrPC to PrPres was initially demonstrated in cell-free reactions in which brain-derived PrPres was incubated with radioactively labeled PrPC, which, under suitable conditions, bound to the PrPres and became similarly partially protease-resistant [34]. These first generation cell-free conversion (CFC) reactions were shown to be highly specific in ways that correlated with prion transmission barriers [9,30,35,47,48] and strains [6]. However, the newly generated PrPres was usually substoichiometric relative to the initial PrPres seed, and was not demonstrably associated with new infectivity [29]. As a result, these CFC reactions [reviewed in [14]] were not suitable for sensitive detection of PrPres or prions.</p><!><p>In 2001 Soto and colleagues described a new type of cell-free prion conversion reaction called protein-misfolding cyclic amplification (PMCA) which has greatly improved efficiency, continuity and sensitivity compared to the initial CFC reactions [52]. In the typical PMCA reaction, crude brain extracts are used as a source of the PrPC which is induced to convert by prions or PrPres in the test sample. Under these conditions, PrPres can be amplified to levels that are detectable by immunoblotting. Prion amplification by PMCA involves repeated cycles of incubation and sonication during which growing multimers of PrPres are fragmented by sonication to increase the effective seed concentration. Sensitivity can also be increased by performing serial rounds of PMCA by transferring a small proportion of one reaction into fresh PrPC substrate for the subsequent round. Originally developed using the hamster adapted 263K scrapie strain, the PMCA has now been adapted to many other species [53] such as sheep [56], deer [36], mice [41] and humans [31,32]. Also, Pastrana and colleagues showed the ability of the PMCA to detect relatively PK sensitive forms of PrPSc (i.e. sPrPSc) [45].</p><p>PMCA is capable of extremely sensitive detection of PrPres in tissues, including hamster and mouse blood [13,25,50,54] or environmental samples such as water [42]. To improve the assay's practicality an automated microplate horn system dubbed serial automated PMCA (saPMCA) was developed [50,51]. This system allows the detection of as little as 1.2 ag or ~ 26 molecules of PrPres after seven reaction rounds. The ability of the PMCA to detect miniscule amounts of PrPres and its applicability to human CJD, sheep scrapie and deer CWD support its use as a TSE pre-clinical diagnostic test. Recently, Chen et al. have reported a quantitative PMCA (qPMCA) approach that allows the determination of the concentration of small amounts of prions in biological samples [16]. This quantitation strategy is based on the direct correlation between the amount of PrPSc in a given sample and the number of PMCA rounds necessary to detect it. However, limitations of the assay for routine clinical applications include the need for 1) brain extracts as a source of PrPC substrate, 2) multiple reaction rounds over several days to 3 weeks to achieve the best sensitivity, and 3) PK digestion and immunoblotting of PMCA products which would impede high-throughput applications.</p><p>To improve the speed and practicality of the PMCA Atarashi et al used recombinant PrPC (rPrPC) as a substrate instead of brain derived PrPC [2]. rPrPC has the advantage of being easily manipulated genetically, purified in large quantities, and added to PMCA reactions at concentrations sufficient to accelerate conversion. The resulting reaction, designated rPrP-PMCA or, more briefly rPMCA, was shown initially to be able to detect as little as 50 ag of PrPres and to differentiate between scrapie-infected and uninfected hamsters using 2 µl of cerebral spinal fluid (CSF).</p><p>A ground-breaking consequence of being able to propagate PrPres in various cell-free reactions was the opportunity to directly evaluate the protein-only hypothesis for TSE prions. Initial indications that synthetic recombinant PrP amyloid fibrils alone can be infectious was reported by Legname and colleagues, who showed that such fibrils could induce or accelerate transmissible neurodegenerative disease in transgenic mice (Tg9944) that overexpressed the same truncated PrP mutant that was used to make the fibrils [37]. However the rPrP fibrils were non-infectious for wild-type mice, indicating that infectivity titers were extremely low. Moreover the initial report left open the possibility that prions were being generated spontaneously in the Tg9944 mice, but evidence to the contrary has recently been published [17].</p><p>Much more robust TSE infectivity for wild-type rodents has since been propagated in brain homogenate-based PMCA reactions by Castilla, Soto and colleagues [12]. Their study showed evidence that the biochemical, structural and biological characteristics of PMCA-propagated PrPres was almost indistinguishable from PrPres produced in vivo, except for the observation of substantially longer incubation periods obtained upon inoculation of the former into animals.</p><p>Weber and colleagues then focused on the cause of the prolonged incubation periods obtained with sPMCA-generated PrPres [59–61]. They described a sonication-dependent reduction in PMCA-generated PrPres aggregate size and suggested that enhanced clearance of such aggregates might explain the longer incubation periods. This interpretation is consistent with their ability to shorten the incubation periods by adsorbing the sPMCA PrPres products to nitrocellulose particles prior to inoculation.</p><p>Green and colleagues demonstrated PMCA amplification of naturally occurring CWD infectivity [27]. They reported an equal level of infectivity being present in both the CWD PMCA conversion product and the original 04–22412 CWD cervid brain homogenate inoculum.</p><p>Studies by Deleault, Supattapone and colleagues demonstrated that infectious PrPres propagation can be achieved in a greatly simplified system containing largely purified PrPC from brain and co-purified lipids [21]. Moreover, as previously reported by the same group using another in vitro PrPSc amplification assay [20], the PMCA's amplification efficiency was improved when carried out in the presence of accessory polyanions, i.e. single stranded synthetic Poly A RNA. In the presence of such RNA molecules, infectious PrPres was even generated spontaneously, i.e., without seeding with brain-derived PrPSc. The PMCA-propagated material had a titer of ≈ 5 × 103 LD50 per ml when intracerebrally (i.c.) inoculated into wild type hamsters. Mechanistic studies have described the selective integration of poly A RNA molecules into stable complexes with PrP molecules in the process of prion formation in vitro, suggesting that even in the absence of PrPSc, polyanionic molecules can induce a molecular reorganization of purified PrPC resulting in a conformation similar to that of PrPSc [26].</p><p>More recently, Wang and colleagues [58] described the generation of a recombinant prion with features typical of in vivo-generated prions using three components: rPrPC, POPG (1-palmitoyl-2-oleoylphosphatidylglycerol) and RNA. Following intracerebral inoculation in wild type mice clinical stage of disease was reached at ~150 days. Biochemical characterizations of the PrPres generated, as well as the clinical symptoms, histopathology and second passage behavior induced by its inoculation strongly supported the conclusion they had generated TSE infectivity using these three molecular components.</p><p>A concurrent study by Kim and colleagues reported that prions able to induce disease in wild type hamsters can be generated from purified bacterial rPrPC in the absence of any mammalian co-factors using prion-seeded rPMCA [33]. This rPMCA product showed variable attack rates upon inoculation into hamsters and therefore contained low levels of infectivity (incubation time from 119 – 401 days) on the first passage. However, upon second passage all animals became ill with an average incubation period of about 80 days. Lesion profiling indicated that rPMCA had altered the strain characteristics of prions (263K strain) that were initially used to seed the serial rPMCA reactions.</p><p>The modulation of conversion of PrPC into PrPres by cofactors was also studied by Abid et al. by using a heterologous PMCA reaction [1]. Their results suggest that the conversion factor involved in prion replication is present in several tissues (e.g. brain, liver, kidney, heart) from different mammalian species and absent in total extracts from other evolutionary lower species such as bacteria and drosophila. This cofactor was found within lipid rafts and most likely was neither a protein nor other molecule that can be denatured by heat. Furthermore, they present evidence that when nucleic acids were depleted from brain homogenate, some other factor promoted the PrP conversion suggesting that more than one type of molecule can act as a cofactor. In a similar study, using both hamster and mouse PMCA, Deleault and colleagues described species specific difference in the use of cofactors for PrPSc propagation [22]. They reported that in the case of mouse PMCA only brain and liver homogenates appeared to contain the conversion cofactor, which also appeared to be protease-resistant and heat stable.</p><p>To investigate the role of PrPC glycosylation in modulating conversion efficiency Nishina and colleagues tested the ability of un-, mono- or di-glycosylated PrPC to support prion amplification using both hamster and mouse the PMCA reactions [43]. Their data shows that whereas unglycosylated mouse PrPC is required to propagate homologous RML prions, diglycosylated PrPC is necessary to propagate hamster Sc237 prions, suggesting that the stoichiometry of the PrPC glycoforms influences the efficiency of PrPres formation in vitro. However, more recently, the Supattapone lab also used PMCA to show that PrP glycosylation is not necessary for strain-specific neurotropism [46].</p><!><p>Prion strains are characterized by distinct incubation periods, clinical symptoms and brain lesion profiles, as well as differences in biochemical features of PrPres (e.g. electrophoretic mobility, glycoform pattern, infrared spectrum, and conformational stability). Prion strains (or mixtures of strains) can usually be serially passaged stably in a hosts of a given species and genotype. However, under some circumstances, new strains or mixtures of strains can arise, especially after passage from one host genotype to another. A wealth of evidence suggests that the properties of prion strains are usually maintained by the faithful propagation of different conformers and/or aggregation states of PrPSc [6,7,19,40,55]. However, the occasional biological instability of prion strains implies that propagation of such conformational states can be subject to permutation, most notably when the prion seed has to act on heterologous PrPC molecules.</p><p>Several PMCA studies support this concept of prion strain propagation. Castilla et al showed that PMCA generated PrPres seeded with five different murine and four human prion strains retained their specific biochemical properties and, upon injection into wild type animals, the PMCA generated PrPres caused disease with features comparable to the parental strain [11]. Green and colleagues reported that features of the 04–22412 CWD prion strain were kept after PMCA reaction [27]. Collectively, these data are consistent with the idea that prion strain features are encoded, at least to a large extent, by the PrPres conformation. Furthermore, Green and colleagues describe the adaptation of the RML mouse prion strain to Tg(CerPrP) mice, overcoming the mouse-cervid species barrier and creating a new prion strain using PMCA [27]. In a similar study, Castilla and colleagues describe the generation of new prion strains by hamster-mouse interspecies PMCA amplification [10]. In particular, hamster PrPC substrate and mouse brain-derived PrPSc or vice versa, produced new prion strains which caused diseases with pathological and biochemical features that were unlike those of other known prion strains. Barria and colleagues developed mouse and hamster PMCA reaction conditions that allowed spontaneous generation of PrPres in the absence of initial seeding with PrPSc [5]. The spontaneous PrPres was infectious in wild type animals but caused a new disease phenotype, suggesting the creation of a novel prion strain. Finally, the 263K scrapie-seeded recombinant PrP prions propagated in rPMCA produced distinct lesion profiles through two passages in vivo, providing evidence that rPMCA with rPrPC substrate alone with no mammalian cofactors lead to stable changes in strain characteristics [33]. These studies indicate that new prion strains can be generated with interspecies PMCA, unseeded PMCA, or PMCA using solely rPrPC as substrate. Usually, when amplifying PrPSc the PMCA maintains the strain features of the initial seed, probably through precise templating of the PrPC misfolding process towards the formation of an exact replica of itself. When the PMCA is carried out in the absence of seed, with a heterologous or recombinant PrPC substrate, additional conformational options presumably become available which enhance the likelihood of forming a new prion conformer or strain.</p><p>Collectively, and remarkably, PMCA-based prion propagation mimics prion propagation in vivo to the extent that one can observe not only the stable propagation of prion strains within a given host, but also the permutation of strains and the spontaneous generation of new strains. However, as prion propagation seems largely to be a protein folding problem, we would not expect to PMCA reactions to recapitulate all aspects of prion strain propagation and transmission barriers that are seen in vivo. In intact cells or tissues, interactions between PrPres and PrPC are highly constrained in three dimensions by GPI anchoring to membranes and by localized interactions between the PrP isoforms and other molecules in their physiologically controlled microenvironments. In contrast, PMCA reactions occur in detergent lysates or extracts in which most such constraints on intermolecular interactions are removed.</p><!><p>The Amyloid Seeding Assay described by Colby and colleagues is a multi-well plate prion amplification assay that uses Thioflavin T (ThT) to detect amplification products [18]. ThT is an amyloid dye that undergoes an enhancement of fluorescence yield when bound to protein amyloid fibrils and is used in this and many other amyloidogenesis assays [38]. The ASA utilizes phosphotungstic acid (PTA) precipitated PrPSc as a seed and recombinant PrP (rPrPC) stored in 6 M guanidine hydrochloride as a substrate. The final guanidine hydrochloride concentration in the reaction (0.4 M) is such that the substrate is likely in a partially unfolded/destabilized state. Other reaction parameters include incubation at 37⁰C, the presence of a 3 mm glass bead in each well to enhance agitation, and continuous shaking of the plate. Notably, the ASA can detect protease-sensitive PrPSc from transgenic mice over expressing the PrP (101L) mutation [17]. ASA applicability to various rodent scrapie experimental models and capability to distinguish between brain samples from sporadic CJD (sCJD) patients and negative control normal brains were described. Furthermore, a 98 % correlation of prion detection by ASA and neuropathological lesions in transgenic mice was described. Nevertheless, as noted by the authors, one weakness of the assay is the need to analyze a high number of replicates per sample because of the variability of the kinetics of ThT positive fibril formation. This problem is exacerbated by the fact that under the ASA conditions, spontaneous (unseeded) rPrPC fibril formation also occurs, but usually with a longer lag phase than those seen with prion-seeded reactions. As detailed below for the real time QuIC assay, spontaneous fibrillization can be largely avoided under other reaction conditions.</p><!><p>To avoid technical complexities associated with sonication in PMCA reactions new assays were developed by Atarashi and colleagues in which sonication was substituted by intermittent shaking as a means to break up prion protein aggregates and produce new PrP seeds in reaction tubes [3,4,44,62]. Such shaken conversion reactions have been dubbed Quaking-Induced Conversion (QuIC) reactions. The first-generation QuIC reactions, herein abbreviated Standard QuIC or SQ, were developed as individual microtube-based reactions that contained detergents and used hamster-adapted 263K scrapie as a seed and hamster rPrPC as substrate [4]. As with the rPMCA [2], scrapie seeds induced the conversion of rPrPC to a specific set of proteinase K-resistant bands (rPrP-res(Sc)) that were visualized on immunoblots. The ability to detect as little as 100 ag PrPSc was demonstrated. Through careful selection of reaction parameters such as shaking regimen, detergent concentrations, incubation time and reaction temperature, virtual elimination of spontaneous (unseeded) conversion of the substrate to proteinase K-resistant product (rPrP-res(spon)) seed can be achieved.</p><p>SQ has been used successfully to discriminate between scrapie affected and control hamsters using CSF [4] or nasal lavages [8]. The assay was also applied to the detection of prion seeding activity in brain samples from scrapie affected sheep and a human vCJD patient [44]. Furthermore, good discrimination between cerebral spinal fluid samples from scrapie positive and normal sheep was observed.</p><p>To address limitations of SQ and ASA, Atarashi, Wilham and colleagues developed a new prion-seeded rPrP conversion assay that combines features of the ASA (i.e. multiwell plate format and ThT detection of conversion products) and the SQ (e.g. intermittent shaking, rPrPC preparation, and lack of chaotropic salts) [3,62]. This new assay was called Real-Time (RT) QuIC, or herein RTQ, because of its ability to almost continuously monitor the progress of the QuIC reaction in a shaking, temperature-controlled fluorescence plate reader. As with the ASA [18], the multiwell plate format gives the RTQ is more amenable to high-throughput testing of samples. However, in contrast to the ASA, the RTQ conditions can, depending on the rPrPC substrate, virtually eliminate the problem of unseeded, prion-independent amyloid formation. The prion-seeded RTQ conversion products were similar to the ones previously described with SQ [4,44] and showed distinct PK-resistant bands of ~20, 18, 14 and 13 kDa, while control reactions seeded with normal tissue had virtually no PK-resistant products. Circular dichroism (CD) and Fourier transform infrared (FTIR) studies of the RTQ substrate (hamster rPrPC 90–231) and conversion product indicate that the prion-induced structural changes in rPrPC shared some similarities with those occurring in vivo upon conversion of PrPC to PrPSc. Thus, RTQ has promise not only as a prion detection assay, but also as a tool to study the mechanism of prion-induced PrP conversion.</p><p>Wilham and colleagues also describe the use of RTQ to quantitate prion seeding activity in biological samples. Serial dilutions of a given sample are used as seeds and the seeding dose (SD) giving 50% ThT-positive replicate reactions (SD50), i.e., the 50% endpoint dilution, is estimated. The SD50 is analogous to the 50% lethal dose (LD50) determined in an endpoint dilution animal bioassay. As is commonly done in determining LD50 values, the estimation of SD50 values can be aided by using Spearman-Kärber [23] or Reed-Muench [49] analyses. This end-point dilution approach to prion quantification is potentially applicable to any prion-seeded amplification assay (e.i. PMCA, rPMCA, ASA). With the RTQ, SD50 concentrations obtained for four hamster scrapie brain homogenate stocks were comparable to LD50 concentrations obtained with hamster end-point dilution bioassays, indicating similar sensitivities for these two types of assays. However, RTQ has several major advantages over animal bioassays, including practicality, high-throughput potential, rapidity and reduced cost.</p><p>Quantitation of prions in CSF samples from scrapie positive hamsters by RTQ gave SD50 values of 105.6 and 104.7 per ml, respectively. Detection of prions in brain samples from TSE positive sheep and deer was also described. One important version of RTQ has been shown to have 81% sensitivity and 100% specificity in discriminating sporadic-CJD and non-CJD patients based on CSF samples [3,3].</p><p>Of particular interest are prion amplification assays that are capable of detecting prions in blood components such as plasma. However, blood typically has extremely low prion concentrations [i.e., ~13 LD50 per ml, [28]] and contains inhibitors of some of the most sensitive tests such as PMCA [13] and another assay [57]. Recently, a prion specific immunoprecipitation has been integrated with both SQ and RTQ to increase sensitivity and isolate prions from inhibitors such as those present in plasma. When coupled with the immunoprecipitation step the RTQ reaction allowed more sensitive detection of variant CJD brain homogenate diluted into human plasma and also rapid discrimination of plasma and serum samples from scrapie-infected and uninfected hamsters (unpublished observations). These developments should improve prospects for the practical detection of minimal levels of prions in tissues, fluids or environmental samples.</p><!><p>Since the development of the first PrP in vitro conversion reaction [34] much more efficient, continuous and sensitive prion-seeded conversion assays have been developed. These techniques have been used to investigate prion composition and propagation mechanisms as well as prion strain and transmission barrier phenomena. Moreover, these reactions serve as bases for ultra-sensitive prion detection that should facilitate TSE diagnostic tests and screening assays for medical, agricultural and environmental prion contamination.</p><p>A pre-clinical TSE diagnostic test should be sensitive enough to detect minimally infectious or even subinfectious quantities of prions and allow amplification/detection of multiple prion strains in a wide variety of biological tissues. As reported by Wilham and colleagues, inhibitory matrix interference can be overcome by diluting the sample until the reaction is no longer affect by the inhibitors present [62]. Another strategy is to capture and concentrate prions from complex mixtures in a manner that is compatible with amplification/detection as we have recently accomplished using immunoprecipitation and others have reported using steel [24] or magnetic particles [39]. Our studies indicate that when used in combination with an improved RTQ reaction, immunoprecipitation provides for sensitivities that are several orders of magnitude greater than those obtained using the metallic particles.</p><p>The fact that so far the infectivity of PrP in in vitro conversion products as been shown to be lower than that of bona fide PrPSc suggests that we are still missing important information about the conversion process and how to create an infectious prion The prion-seeded conversion reactions described in this chapter provide valuable tools to investigate these issues.</p>

PubMed Author Manuscript

Electroosmotic pumps for microflow analysis

With rapid development in microflow analysis, electroosmotic pumps are receiving increasing attention. Compared to other micropumps, electroosmotic pumps have several unique features. For example, they are bi-directional, can generate constant and pulse-free flows with flow rates well suited to microanalytical systems, and can be readily integrated with lab-on-chip devices. The magnitude and the direction of flow of an electroosmotic pump can be changed instantly. In addition, electroosmotic pumps have no moving parts. In this article, we discuss common features, introduce fabrication technologies and highlight applications of electroosmotic pumps.

electroosmotic_pumps_for_microflow_analysis

1. Introduction<!>2. Unique features<!><!>2. Unique features<!>3. Fabrication<!>3.1. Open channel<!>3.2. Packed column<!>3.3. Porous monolith<!>3.4. Porous membrane<!>4.1. Microflow-injection analysis (\xce\xbc-FIA)<!><!>4.1. Microflow-injection analysis (\xce\xbc-FIA)<!>4.2. Microfluidic liquid chromatography (\xce\xbc-LC)<!>4.3. Other applications<!>5. Conclusion<!>

<p>Integration and automation of multiple analytical processes and miniaturization of analytical instruments have been active research areas in the past half-century. Traditionally, chemical assays were implemented using a batch approach.</p><p>The first autoanalyzer, an air-segmented flow instrument, was developed in the late 1950s [1] and was improved in the early 1960s [2]. Air bubbles had been an issue in this instrument.</p><p>In the mid-1970s, a flow-injection (FI) technique was invented [3] and that eliminated air bubbles in the flow stream. Following FI, a sequential-injection (SI) method [4] was developed in the early 1990s. In the early 2000s, a lab-on-valve (LOV) device was developed [5] and was recently reviewed [6].</p><p>In the above systems, various "tubings" were used as conduits to transfer sample and reagent solutions, as chambers to perform chemical reactions, and as flow cells to facilitate detection. Since the early 1990s, a microfluidic or lab-on-chip (LOC) device has emerged as a new platform to integrate analytical processes and miniaturize analytical instruments [7]. In LOC devices, the various tubings were microfabricated into chips using standard photolithographic technologies developed in the semiconductor industry. To miniaturize FI, SI and LOV systems or to enable LOC devices to perform FI and SI analysis, micropumps are needed to manipulate solutions within the fluidic networks. Many micropumps have been developed and were reviewed recently [8].</p><p>In this article, we review a special class of micro pumps – electroosmotic pumps (EOPs).</p><!><p>EOPs use electroosmosis or electroosmotic flow (EOF) to drive liquids around within fluidic conduits. EOF is the movement of uncharged liquid relative to a stationary charged surface due to the application of an externally electric field [9]. It is a phenomenon inherent to a solid-liquid interface [10]. In 1809, Reuss [11] had discovered the phenomenon of electroosmosis, wherein the application of an electric field across a porous dielectric material caused liquid to flow. Using EOF to drive liquid for chemical analysis was pioneered by Pretorius et al. [12]. Although Theeuwes had developed EOPs for controlled drug delivery in 1975 [13], they were not popular in analytical chemistry until the 1990s, when they were applied in miniaturized analytical systems [14–17].</p><p>Compared with other pumps used in micro-analytical systems, EOPs have several unique features:</p><!><p>EOPs are bi-directional and their flow directions can be switched almost instantly. This feature comes from the nature of EOF. By changing the polarity of the electric field, the direction of EOF can be switched instantly. This property enables EOPs to perform rapid, accurate manipulations of solutions, critically important in studying fast chemical reactions on the microscale (e.g., ~10-μm channel).</p><p>EOPs are capable of generating constant and pulse-free flows. This is essential to microscale flow analysis, although minute flow-rate fluctuations (pulsed flows) are tolerable in regular FI/SI systems in which the conduits often have dimensions > 200 μm.</p><p>EOPs can be readily integrated with LOC platforms, due to the common materials and processes involved in the production of both devices. This is crucial to integrated microfluidic systems when multiple-step chemical analysis is performed. Piezoelectric and silicone-membrane-based micropumps [18,19] can be fabricated on LOC platforms, but their versatility, reliability and accuracy are inadequate for high-quality analytical applications (e.g., pipetting nL solutions) [20].</p><p>EOPs have no moving parts. The moving parts in all mechanical pumps are usually lifetime-limiting components and sources of inaccuracies and failures. EOPs eliminate such components.</p><!><p>These are a few primary advantages of EOPs. Table 1 presents a comparison between an EOP and other common types of pumps used in microflow analysis. One inherent issue associated with EOPs is contamination of pumping elements by pump solutions, especially when a system solution is used as the pump solution.</p><!><p>Based on the types of pumping elements used, EOPs can be categorized as open channel, packed column, monolith column and porous membrane. The following introduces the processes employed in making these pumps.</p><!><p>In open-channel EOPs, both silica capillaries and microchannels on chips have been used as the pumping elements.</p><p>Fig. 1 presents an EOP made from an open capillary [14]. The pumping element was a segment of a bare fused silica capillary (C1). A and B were two reservoirs holding a buffer (e.g., sodium tetraborate) solution. The reagent holding coil (HC) was a piece of capillary tubing filled with the reagent solution to be delivered. V1 was a selection valve.</p><p>When the valve was set at the position as shown in Fig. 1, the two syringes (S1 and S2) were isolated from the pump system. As the valve was switched to another position, S1 connected to C1, and S2 connected to HC. C1 could be rinsed with the solution in S1, and HC could be replenished with the reagent in S2. HV was a high-voltage power supply. A key component of this EOP was the grounding joint, which separated the pumping element (C1) from the chemical-reaction system (HC and the fluidic network after HC). It was constructed using a piece of Nafion ion-exchange-membrane tubing (M).</p><p>Fig. 2 shows a detailed construction of this joint. When C1 and connection capillary were joined, a small gap was left between them. Two PVC-tubing sleeves were used to secure and to seal M to C1 and the connection capillary. This membrane joint allowed electrical connectivity through ion conduction between the solutions inside and outside the membrane tubing, but solutions could not flow across the membrane. V1 was normally set to join the connection capillary and HC, as shown in Fig. 1. When a negative high voltage was applied, pump solution flowed backwards, which allowed the free end of the capillary to aspirate solutions. When a positive high voltage was applied, pump solution flowed forward to propel the solutions in HC to the rest of the fluidic network.</p><p>The pump rate can be controlled by adjusting the electric field strength applied to C1. Alternatively, the pump rate can be boosted by employing parallel pumping capillaries. To construct a multiplexed-capillary EOP, a union is required to combine all the outlets of the capillaries into one outlet. Making such a union with small dead volumes was an issue for open-capillary EOPs [15–17]. This issue can be easily addressed using microfabrication technologies.</p><p>Fig. 3 shows an example of an open-channel EOP [21]. The pumping element was a group of shallow microfabricated channels called pumping channels (1). All pumping channels had a common inlet (2) and outlet (3) reservoir. The pumping element was exposed only to buffer solutions and did not come into contact with the sample solution. A porous glass disk (5-mm diameter, 0.8–1-mm width, and 4–5-nm pore size) was utilized to facilitate application of the electric field between reservoirs 2 and 3, and also to prevent EOF leakage to reservoir 3. Reservoir 3 was fabricated from a PEEK external nut, and the porous glass disk was secured at the bottom of reservoir 3 with a corresponding internal nut. As a positive electric field was applied between reservoirs 2 and 3, EOF was generated to propel the solution(s) downstream.</p><p>The pumping elements in open-channel EOPs were usually produced using standard photolithographic technologies [22].</p><!><p>Packed columns are commonly used as pumping elements in EOPs, which are normally capable of producing higher flow rates than open-channel EOPs due to their increased porosities. When fine particles are used, the pore diameters become very small and hence high pumping pressures can be achieved (e.g., 8000 psi [23]).</p><p>The methods for producing packed columns for EOPs are identical to those for packing columns in capillary LC or capillary electrochromatography. The first step is to create a short (e.g., 1 mm) frit close to one end of the capillary (e.g., [24]) by introducing a cohesive paste made up of one part sodium silicate solution (14% NaOH aqueous solution) and three parts non-porous silica particles, then baking the capillary in an oven at 350°C for 20 min. In the second step, a non-porous silica bead slurry is pressurized into the capillary using a slurry packing apparatus. In the final step, after the beads are densely packed inside the column, a second frit is fabricated near the open end of the capillary by sintering. Such a column, with the two frits retaining the beads inside it, constitutes a packed-column EOP.</p><p>The overall configuration of a packed-column EOP is similar to that of an open-channel EOP except for the grounding joint. In packed-column EOPs, the grounding joint is often made from a metal electrode directly in contact with the pump solution (e.g., [25]) albeit porous columns have been used [26]. This configuration encounters a problem (i.e. gas bubbles generated from the electrolysis are trapped inside the fluidic system). A porous Teflon membrane device has been employed to release these bubbles [25]. Alternatively, platinum (catalyst) can be used to recombine H2 and O2 into H2O [27].</p><!><p>Microporous monolithic columns are outstanding options as pumping elements for in manufacturing EOPs. A porous polymer monolith is a single, continuous piece of highly cross-linked porous polymer that can be prepared by a simple polymerization process from liquid precursors, including monomers, cross-linker, free radical initiator and porogenic solvents. The polymer completely fills the volume of a specifically-designed chamber or a section of a capillary or a microchannel. An excellent feature of the monolithic column is the elimination of frits. The monolithic materials used for EOP development can be polymer-based or silica-based [28,29].</p><p>Silica monoliths are frequently used for development of EOPs. The following presents a typical procedure for preparing the monolith [30]. The inner surface of a capillary was first cleaned and activated by washing the capillary with NaOH, HCl, water, acetone and diethyl ether. A monolithic silica matrix, comprising 0.5 mL of 0.01 M acetic acid, 54 mg of poly(ethylene glycol) and 0.2 mL of tetramethoxysilane, was introduced into an appropriate length of the capillary. The ends of the capillary were then connected with a Teflon tube to form a loop and placed in an oven at 40 °C for 24 h. After the silica gel made within the capillary was washed with water and 0.2 M ammonium hydroxide, the columns were placed into the 40°C oven for another 24 h. The column was heated in a temperature programmable oven at 80°C, 120°C, 180°C, and 300°C for 4 h at each temperature, purged with helium at 180°C for 1 h, and cooled to room temperature at the rate of −1.0°C/min. Other protocols have also been used to produce silica monolith [31,32].</p><p>A potential problem for silica monolith is that silica can be dissolved slowly in alkaline solutions. This dissolution will change the structure of the monolith and hence the EO flow rate. In the worst cases, the monolith will collapse. An organic polymer monolith will overcome this problem.</p><p>Detailed methods for making polymer monolith have been described [33,34]. The following protocol presents an example for making a monolithic column with an average pore size of ~1 μm. The inner wall of a capillary was first derivatized with 3-(trimethoxysilyl)propyl methacrylate. A polymerizing solution, containing 1.2 g of 75% 2-(methacryloyloxy)ethyl]trimethylammonium chloride in water, 0.3 g ethylene dimethacrylate, 1.96 g 1-propanol, 0.84 g 1,4-butanediol, 0.15 g water and 13 mg azo(bisisobutyronitrile), was then introduced into the capillary. This capillary, with both ends sealed by silicone septa, was submerged in a water bath at 55–60°C for polymerization for ~24 h. After the monolith was flushed with methanol to remove all unreacted components, a monolithic column was produced. The pore dimension of the monolith can be tuned over a broad range (0.01–10 μm) by controlling the amount of the porogenic solvent, the percentage of cross-linking monomer in the polymerization mixture, and the polymerization temperature [35,36].</p><!><p>Porous membranes (porous frits are categorized as a special kinds of porous membrane) have also been used as the pumping element in EOPs. A unique feature of these materials is that they provide high-density arrays of short pumping channels. This feature leads to low-voltage EOPs, because the short pumping channels enable application of a low voltage to produce a high electric field across the pumping element.</p><p>Porous membranes are often commercially available (e.g., glass frit [27], ion exchange membrane [37], and silica and alumina membranes [38]). These materials can also be prepared in laboratories. For example [39], a porous membrane can be produced by putting a layer of polysilicon film onto a porous silicon template using an oxidized, low-temperature, chemical-vapor-deposition process. The pores go straight across the template, and are distributed hexagonally with diameters of the order of 6 μm and pitch distances of 8.5 μm [39]. Because the polysilicon film grows on the entire surface of the template, the final pore size can be controlled by controlling the thickness of the polysilicon. These membranes have an advantage over porous glass frits in that the porous silicon tortuosity approaches unity. Porous frits can also be produced via a sintering process [40].</p><p>Porous-membrane EOPs are configured the same way as the packed-column EOPs. However, because the membranes or the frits are short, a sandwich structure is frequently used to support the pumping element [27,37–39].</p><!><p>To demonstrate EOPs for FI applications, Dasgupta and Liu [14] constructed a two-line FI system for chloride determination. Referring to Fig. 4, there were two EOPs similar to that described in Fig. 1, but each pump comprised four single open-capillary (40 cm × 75 μm i.d. × 375 μm o.d.) pumps in order to increase the flow rate. The pump electrolyte (2 mM sodium tetraborate) was used directly as the carrier stream in which the sample was injected via V2 (100-nL injection volume), so no holding coil was necessary in this line. A holding coil, HC, was used in the reagent [0.8 mM Hg(SCN)2 and 0.2 M Fe(NO3)3] line.</p><p>The chemistry behind this method involved a reaction between chloride and mercuric thiocyanate, forming mercuric chloride and releasing thiocyanate. The released thiocyanate reacted with ferric ion in the same solution, forming a blood-red complex that could be detected by an absorbance detector.</p><p>In this system, the flow rate of each pump was ~1.7 μL/min. A good linear relationship was obtained for Cl− determination (R2 = 0.996 for 50–600 ppm [Cl−]), with a relative standard deviation of ≤0.8% (n = 13).</p><p>A similar EOP has also been used for SI analysis of nitrite-nitrogen and ammonia-nitrogen [16]. Fig. 5 presents the configuration of the SI system. The HC is a capillary holding coil with a length of 40 cm and an inner diameter of 250 μm. The EOP was connected to the common port of a selection valve (V2), which was pneumatically operated.</p><p>The Griess-Saltzman reactions [41] were employed for nitrite-nitrogen analysis and the nitroprusside-catalyzed Berthelot reaction [42] was utilized for ammonia-nitrogen determinations. This application took advantage of an EOP capable of instantly switching flow direction. Excellent linear relationships (R2 = 1.000) were obtained for both nitrite-nitrogen (20–400 μM) and ammonia-nitrogen (50–600 μM).</p><p>Pu and Liu [22] manufactured an EOP on a microchip and demonstrated its application for an enzyme-inhibition assay. Fig. 6A shows the photomask design of the pump chip and Fig. 6B the SI system. The microchip EOP contained a pumping element of 32 parallel channels, each with a depth of 20 μm and a total length of ~27 cm. An isolation channel was also fabricated on the chip to separate the sample and reagent solutions from the pumping solution. The isolation channel served as a sample and reagent holding coil as well as for the SI analysis. The experimental set-up was identical to a conventional SI system except for the microchip EOP. The system was applied to β-galactosidase-catalyzed hydrolysis of fluorescein di(β-D-galactoside) and inhibition of this hydrolysis reaction by diethylenetriaminepentaacetic acid.</p><p>There were three basic steps to accomplish the assay:</p><!><p>a negative high voltage was applied to the EOP to aspirate sample and reagents into the isolation channel;</p><p>after the sample and reagents were reacted adequately, a positive high voltage was applied to the pump to propel the product to the detector for measurement; and,</p><p>the system was reset (this included washing the sampling capillary attached to the selection valve and loading this capillary with the next sample solution).</p><!><p>For either of the above SI systems, if the EOP was replaced with a syringe pump, the reproducibility would deteriorate considerably because syringe pumps cannot control flows that precisely.</p><!><p>EOPs offer a simple, cost-effective means to generate adequate pressures and flow rates for capillary or chip-based liquid chromatographic (LC) separations. Paul et al. [43] constructed a high-performance LC (HPLC) system using a packed-column EOP. The pump column (0.15 mm i.d. and 3 cm long) was packed with 1.5-μm diameter non-porous silica beads, and the separation column (0.1 mm i.d. and 11 cm long) was packed with 3-μm diameter octadecylsilane-coated porous silica beads. Separations of polycyclic aromatic hydrocarbons were performed on this system.</p><p>Chen et al. [25] built an EOP capable of generating pressures in excess of 3 MPa and flow rates in the μL/min range. The pump comprised three parallel fused-silica capillary columns (25 cm × 530 μm i.d.) packed with 2-μm silica beads. Hollow metal tubes were used as grounding electrodes. Fig. 7 shows the one-stage EOP and the μ-HPLC system. The μ-HPLC system comprised a four-port injection valve with a internal loop of 200 nL, 15 cm × 320 μm i.d. 5 μm Spherigel C18 stainless-steel analytical column and an on-column ultraviolet-visible (UV-Vis) detector.</p><p>To evaluate the performance of the system, a mixture of thiourea, benzene, toluene, naphthalene, phenanthrene, biphenyl and anthracene were separated. Acetonitrile/water was used as the mobile phase. The number of theoretical plates of the column was 2.3–3.2×104/m using the EOP, and 1.4–2.3×104/m using a mechanical pump. The retention-time (tr) error expressed in RSD% was within 0.8% with the EOP for all compounds tested, while the tr error was 3.6% with the mechanical pump.</p><p>Lazar et al. [44] developed a microfluidic HPLC system (Fig. 8) for protein analysis. The system comprised two EOPs, a valving component, a separation channel with an on-column preconcentrator, and an electrospray ionization (ESI) interface. The separation channel (5) was 2 cm long and 50 μm deep. The packing material comprised 5-μm Zorbax SB-C18 particles, loaded manually in the channel from the LC waste reservoir (11) using a syringe. The packing material was retained in the separation channel with microfabricated multi-channel filter structures (~100 μm in length and ~1.5–1.8 μm in depth). The two EOPs (1A and 1B) were identical, and each comprised 200 nano-channels (2 cm long, ~1.5–1.8 μm deep). The voltage for EOF generation in the pumps was applied to reservoirs 2A/2B and 3. The voltage applied to reservoir 3 was also the voltage for generating the electrospray. EOF leakage from outlet reservoir 3 was prevented by a porous glass disk (5-mm diameter, 0.8–1-mm width, 40–50-Å pore size), which was secured to the bottom of the reservoir. Sample loading was accomplished through a double-T injector (4) via the EOF valving structures (8 and 9), each comprising 100 nano-channels (2 cm long, ~1.5–1.8 μm deep). As the hydraulic resistances of the EOPs and the valving structure are much larger than that of the separation channel, the parallel channels act as a valve that is open to material transport when an electric field is applied and closed to material transport when the electric field is removed. A fused-silica capillary (10 mm long, 20 μm i.d. and ~90 μm o.d.) was inserted into the LC channel (5) for generating the electrospray (10). The flow rates generated were sufficiently stable for nano-LC separations with detection by mass spectrometry (MS). The analysis enabled confident identification of 77 proteins.</p><p>For HPLC applications, piston pumps may be utilized, although they are difficult to integrate into microfluidic chip devices.</p><p>Table 2 summarizes the applications of EOPs in microflow analysis.</p><!><p>EOPs have been used for other applications.</p><p>Jin et al. [53] reported a proteolytic digestion chamber on a microchip. Various solutions were electroosmotically transported to the chamber, and the peptide products were subsequently analyzed using capillary electrophoresis (CE) and matrix-assisted laser desorption/ionization time-of-flight MS (MALDI-TOF-MS). The digestion results from this microchip device were comparable to those from a conventional water-bath digestion apparatus.</p><p>Dasgupta and Liu [54] connected an EOP to the detection end of a CE capillary to control the flow pattern in the CE capillary. The majority of CE applications involve samples having lower ionic strength than that of the running electrolyte to achieve sample stacking. In such cases, the auxiliary EOP can be used to optimize the stacking profile and thus improve the separation efficiencies.</p><p>EOPs have also been used in micro-energy systems. In particular, EOPs have been utilized for water management in proton-exchange-membrane (PEM) fuel cells [55] and for methanol/water-mixture delivery in direct methanol fuel cells (DMFCs) [56]. Kim et al. [56] described the use of an EOP made of porous glass to deliver methanol/water for a DMFC, and Buie et al. [57] developed a miniature, convection-free DMFC that utilized a planar EOP for methanol delivery to the anode. EOPs have also been used for cooling electronic devices [58].</p><!><p>EOPs have been demonstrated to be efficient for fluid propulsion and aspiration. They can be used in FI, SI, LOV and LOC systems. EOPs can generate stable, pulse-free flows from a few nL/min to several mL/min. Generally, if high pump rates are utilized, maximum pumping pressure will be sacrificed. This may not be of concern, because typically a high-flow-rate FI/SI system has relatively low flow resistance, unless a packed-bed reactor or a filtration system is used. The most important features of an EOP are that it can propel and aspirate solutions, it can generate pulseless, stable flow, its flow rate and direction can be changed instantly, it has no moving parts, and it can be readily integrated with microfluidic and microelectromechanical system (MEMS) devices. EOF pumps are ideally suited to microfluidic systems for which pulseless, stable flow is desired.</p><p>Flow stability in an EO pumped system has been a problem in some applications, because solutions of different compositions go through the capillaries. Sometimes, flow rates change because of the adsorption of compounds from the samples or sample matrix onto the surfaces of the pumping elements. This problem can be avoided if the pump fluid is separated from the sample and reagent solutions in the analytical system. When properly designed, EOPs can be used as a stand-alone pump or integrated into a microchip device to generate adequate flow rates and pressures for development of miniaturized HPLC.</p><!><p>Stand-alone electroosmotic pump (Reprinted with permission from [14]).</p><p>Membrane joint (Reprinted with permission from [14]).</p><p>Open-channel electroosmotic pump (EOP). 1 – Open-channel EOP, 2 – Micropump inlet reservoir, and 3 – Micropump outlet reservoir. The bottom figure shows an expanded view of reservoir 3 containing the porous glass disk. (Reprinted with permission from [21]).</p><p>Two-line flow-injection analysis (FIA) system with electroosmotic flow (EOF) pumping: B, Pump electrolyte-solution container; T1, T2, Capillary unions; Vla and Vlb, Four-way valve stacks a and b; S1 and S3, Syringes holding pump-buffer solution; S2 and S4, Syringes holding carrier and reagent solutions, respectively; HC, Reagent holding coil; T3, Low-volume tee union (Reprinted with permission from [14]).</p><p>Capillary-format sequential injection analysis (SIA) system. HV, High-voltage power supply; A, B, Pumping electrolyte-solution containers; M, Membrane joint; Cl, Pumping capillary; T, 4 x 1 union; HC, Holding coil; Vl, Four-way valve; Sl and S2, Syringes; V2, 6 x 1 selector valve; RI, R2, R3, Reagents; aux, Unused auxiliary solution port (Reprinted with permission from [16])</p><p>(A) Photomask design of the pump chip. (B) Micro-electroosmotic pump-sequential injection analysis (μ-EOP-SIA) system (Reprinted with permission from [22]).</p><p>One-stage electroosmotic pump (EOP) and the micro high-performance liquid chromatography (μ-HPLC) system. (A) The EOP system: 1, Solvent reservoir, covered with an insulating sheath; 2, High-voltage direct-current source module; 3, Pt wire; 4, Hollow electrode (grounded); 5, Capillary conduit; 6, Packed columns, three packed columns connected in parallel; 7, Gas-releasing device; 8, Representation of direction of gas flow); 9, Liquid-pressure sensor; 10, Open/close valve; 11, Measurement point of flow rate. (B) The μ-HPLC system: 12, Four-port injection valve; 13, Analytical capillary HPLC column; 14, On-column ultraviolet-visible (UV-Vis) detector; 15, Chromatographic data station; 16, Waste-liquid bottle (Reprinted with permission from [25]).</p><p>(A) Photomask design of the pump chip. (B) Micro-electroosmotic pump-sequential injection analysis (μ-EOP-SIA) system (Reprinted with permission from [44]).</p><p>Function comparison between electrosmotic pumps (EOPs) and other common pumps in microflow analysis</p><p>The rating is based on the following scale (from the best to the worst): E (excellent), V (very good), G (good), F (fair) and P (poor).</p><p>This pump represents reciprocal high-pressure HPLC pumps.</p><p>Summary of EOP applications in microflow analysis</p><p>10 mM NBu4ClO4 & 20 mM 1,10-phenanthroline(pH 4.9)</p><p>5 mM 2-[bis(2-hydroxyethyl)aminoethane sulfonic acid and 25 mM NBu4ClO4</p><p>n.a. stands for "not applicable" or "not available".</p>

PubMed Author Manuscript

Non-covalent interactions in controlling pH-responsive behaviors of self-assembled nanosystems

Self-assembly and associated dynamic and reversible non-covalent interactions are the basis of protein biochemistry (e.g., protein folding) and development of sophisticated nanomaterial systems that can respond to and amplify biological signals. In this study, we report a systematic investigation of non-covalent interactions that affect the pH responsive behaviors and resulting supramolecular self-assembly of a series of ultra-pH sensitive (UPS) block copolymers. Increase of hydrophobic and \xcf\x80-\xcf\x80 stacking interactions led to the decrease of pKa values. In contrast, enhancement of direct ionic binding between cationic ammonium groups and anionic counter ions gave rise to increased pKa. Moreover, hydration of hydrophobic surfaces and hydrogen bonding interactions may also play a role in the self-assembly process. The key parameters capable of controlling the subtle interplay of different non-covalent bonds in pH-triggered self-assembly of UPS copolymers are likely to offer molecular insights to understand other stimuli-responsive nanosystems. Selective and precise implementation of non-covalent interactions in stimuli-responsive self-assembly processes will provide powerful and versatile tools for the development of dynamic, complex nanostructures with predictable and tunable transitions.

non-covalent_interactions_in_controlling_ph-responsive_behaviors_of_self-assembled_nanosystems

Introduction<!>UPS block Copolymer synthesis by ATRP method<!>Effect of hydrophobic interactions on pH responsive behavior<!>Effect of \xcf\x80\xe2\x80\x93\xcf\x80 stacking on pH responsive behavior<!>Summary of effect of tertiary amines substituents in changing pH responsive behavior<!>Effect of hydrogen bonding and ionic bonds on pH responsive behavior<!>Effect of polymer concentration on pH responsive behavior<!>Conclusions

<p>Multifunctional nanomaterials are of growing interest and importance in a variety of optical, electrical, thermal and mechanical systems in a wide range of applications such as sensing, self-healing, adaptable surface adhesion and drug delivery1–3. Self-assembly has emerged as the most practical strategy in the synthesis of well-organized and stable nanostructures4. Such strategies have been exploited in a number of stimuli-responsive nanosystems such as thermo-sensitive hydrogels (e.g. Poly(N-isopropylacrylamide)5, elastin like proteins6), pH responsive nanoprobes7,8 and oligonucleotide-functionalized gold nanoparticles9,10. Despite widespread interest, molecular understanding of underlying supramolecular chemistry remains elusive. Part of the challenge is the lack of identification and evaluation of key structural and environmental parameters that affect their stimuli-response and accompanied supramolecular self-assembly, which hampers our capability in the rational design of responsive nanomaterials in a predictable fashion11.</p><p>Upon external physical or chemical stimuli, responsive nanomaterials can respond and adapt to surrounding environment through conformational or chemical changes. In many cases, the conformational changes come in the form of external signal-triggered supramolecular self-assembly. At the molecular level, supramolecular self-assembly is the spontaneous association of molecules under equilibrium conditions into thermodynamically stable and structurally well-defined nanostructures joined by non-covalent bonds12,13. Supramolecular chemistry offers a toolbox of multiple non-covalent interactions for the formation of well-defined nanostructures14. These interactions range from relatively weak forces such as hydrophobic interactions, hydrogen bonding, or π–π stacking interactions to ionic bonds. The keys to develop functional nanomaterials via external signal-induced self-assembly are to understand and control the non-covalent connections between building blocks or molecules, and to understand and overcome the intrinsically unfavorable thermodynamic barrier involved in bringing many molecules together in a single aggregate of a nanoparticle4. A major challenge in studying self-assembly in natural and synthetic systems is the interdependence of many non-covalent interactions and their compensating effects on the composite behaviors of the nanosystem15. To pinpoint specific contribution from individual non-covalent interaction, we need to establish a model system in which key parameters can be changed easily and independently.</p><p>Recently, we reported a library of methacrylate-based ultra-pH sensitive (UPS) nanoprobes with sharp pH transitions that are finely tunable in a broad range of physiological pH (4.0–7.4)16,17. Besides the introduction of acid labile moieties like acetal groups18,19, the incorporation of ionizable groups such as amines or carboxylic acids serves as the general strategy in the development of pH sensitive nanomaterials20–24. The nanoprobes consist of a block copolymer, PEO-b-PR, where PEO is poly(ethylene oxide) and PR is an ionizable tertiary amine block. The UPS nanoprobes achieved over 100 fold fluorescence intensity increase within 0.3 pH units, which is critical for broad tumor imaging and endosome maturation studies25,26. The hydrophobic and π–π stacking interactions can be intensified by increasing the hydrophobic chain length and incorporating aromatic moieties in the copolymer, respectively. The hydrogen bonding network and hydration of macromolecular solutes in aqueous solution are also dependent on the presence of salt and its concentration. The strength of ionic bonds between charged polymers and counter ions are known to be ion species dependent. By selectively tailoring structural (e.g., substituents of amines) and solution parameters (ion species and concentrations), we are able to strengthen or weaken ion pair interactions in pH-triggered supramolecular self-assembly of UPS block copolymers.</p><p>Herein we report a systematic investigation of key factors that impact the pH responsive behavior and resulting self-assembly of UPS block copolymers. This study aims to provide fundamental understanding of the effect of multiple non-covalent interactions (e.g., hydrophobic interactions, π–π stacking, hydrogen and ionic bonding) on the pH-triggered supramolecular self-assembly. It also offers useful insights for further development of polymeric pH sensors with predictable and tunable transition pH.</p><!><p>We used the atom transfer radical polymerization (ATRP) method with CuBr as a catalyst and N,N,N′,N′,N″-pentamethyl-diethylenetriamine (PMDETA) ligand for the copolymer syntheses (Scheme 1). The PEO-b-PR copolymers with homopolymeric PR block were synthesized using a single methacrylate monomer (Fig. S1 and Fig. S2) as previously described16. At low pH, micelles dissociate into cationic unimers with protonated ammonium groups. When pH increases, neutralized PR segments become hydrophobic and the block copolymers self-assemble into core-shell micelles (Fig. 1, Fig. S2 and S3). The delicate balance between the hydrophobic and hydrophilic segments as a result of external pH changes drives the formation of thermodynamically stable micelles. The phase transition pH, where reversible supramolecular self-assembly occurs, depends on the pKa values of UPS block copolymers around which tertiary amines are reversibly protonated.</p><p>pH titration offers critical information on how pH sensors respond to external pH changes along the titration coordinate. We performed the pH titration experiments by adding HCl solution into UPS block copolymers' micelle solution prepared following a solvent evaporation procedure16. Without specific indication, pH titration experiments were performed at weight concentration of 2.0 mg/ml of copolymers in the presence of 150 mM NaCl to mimic the physiological level of salt concentration. We treated the initial micelle solution as having a protonation degree of 0% when no HCl was added. We considered the tertiary amines as 100% protonated when the addition of HCl yielded the sharpest change of pH (Fig. S4). The apparent pKa value was determined as the pH at which 50% of all the tertiary amines were protonated27–29.</p><!><p>Naturally existing self-assembly systems like proteins and synthetic assemblies designed for biomedical applications such as UPS nanoprobes are primarily used in aqueous media. In aqueous solution, the hydrophobic interactions constitute the predominant driving force for the polymeric molecules to self-organize into nanostructures. We began the investigation of structure-property relationship by determining how changing hydrophobic interactions affected the pH responsive behavior and self-assembly of UPS block copolymers. Early studies in thermo-responsive polymers have shown that hydrophobicity of polymers had substantial impact on the transition temperature of nanomaterials6,30,31.</p><p>An obvious strategy in increasing hydrophobic interactions of amphiphilic PEO-b-PR block copolymers is to increase the hydrophobic chain length. For proof of concept studies, we synthesized a series of PEO114-b-nPDPAx block copolymers with fixed hydrophilic poly(ethylene oxide) chains but systematic changes in the hydrophobic PR chain length (x = 5, 10, 20, 60 and 100). pH titrations of these copolymers were performed at the same molar concentration of tertiary amines at 6.75 mM. The transition pH of PEO114-b-nPDPA100 block copolymers, with the longest hydrophobic segment, yielded the lowest pKa at 6.2 (Fig. 2a). In contrast, PEO114-b-nPDPA5, with the shortest hydrophobic chain length, displayed the highest transition pH around 6.7. The plot of pKa values as a function of hydrophobic chain length of PEO-b-nPDPA copolymers showed a dramatic hydrophobic chain length-dependent transition pH shift (Fig. 2b). It is also interesting to note that longer hydrophobic chain length also resulted in sharper pH transition, as measured by ΔpH10–90% (the pH range where protonation degree of all tertiary amines increases from 10% to 90%) (Fig. 2a and S5).</p><p>Strengthening the hydrophobic interactions can also be achieved by increasing the hydrophobicity of amine substituents of UPS block copolymers. To accomplish this goal, we synthesized a series of ultra-pH sensitive PEO114-b-PR80 block copolymers with identical poly(methacrylate) backbone and similar chain length but different linear terminal alkyl groups on the side chain. All polymers showed a strong buffer effect as proven by the plateau along the majority of the pH titration coordinates (Fig. 3a). The PEO-b-PD5A with the most hydrophobic pentyl group yielded the lowest pKa at 4.4. Meanwhile, the PEO-b-iPDPA with the least hydrophobic isopropyl group as an amine substituent showed the highest pKa, close to 6.6. PEO-b-nPDPA and PEO-b-PDBA had pH transitions at 6.2 and 5.3, respectively. These results demonstrated that PEO-b-PR block copolymers with more hydrophobic amine substituents have a lower pKa. We calculated the octanol-water partition coefficients (LogP) of the repeating unit of the PR segment (neutral form) and used them as a quantitative measure of molecular hydrophobicity and strength of hydrophobic interactions. The plot of pKa values as a function of LogP (Fig. 3b) showed a linear correlation.</p><p>To confirm the effect of enhancing hydrophobic interactions on the transition pH of UPS block copolymers, we synthesized another series of PEO114-b-PR80 block copolymers with the same backbone and similar chain length, but cyclic terminal alkyl groups. For PEO-b-PC6A, we observed the highest pH transition at 7.3. Incorporation of one or two extra methyl groups on the piperidine ring resulted in lower transition pH values of 6.8 and 6.1, respectively (Fig. 3c). A plot of pKa values as a function of LogP also showed a linear correlation (Fig. 3d). Besides six-membered rings as cyclic substituents, we also synthesized another block copolymer, PEO-b-PC7A, with seven-membered rings as cyclic substituents. C7A repeating units had similar hydrophobicity (LogP = 2.33) as that of C6S1A (2.25) and they indeed showed very close pKa value around 6.9 (Fig S6).</p><p>Based on the above experiments, we concluded that stronger hydrophobic interactions (both in amine substituents and hydrophobic chain length) generally lead to the decrease of transition pH of UPS block copolymers. An increase of hydrophobic interactions will stabilize the micelles and shift the equilibrium to the direction of neutralization of protonated tertiary amines, corresponding to the decrease of pKa. In this case, the pH-triggered supramolecular self-assembly occurred at lower pH. This is in accordance with early reports that the lower critical solution temperature (LCST) of poly(N-isopropylacrylamide)(PNIPAM) based copolymers can be controlled by changing the hydrophobic chain length32–34.</p><!><p>π–π stacking has been reported to direct the formation of structured ensembles via the self-assembly of individual magnetic particles35,36. We then investigated whether the incorporation of aromatic rings also affected the pKa values of UPS block copolymers. PEO-b-PMBA block copolymers were synthesized via polymerization of (methylbenzylamino) ethyl methacrylate (MBA, LogP = 2.91) monomers. The hydrophobicity of amine substituents of PEO-b-PMBA was similar to that of PEO-b-nPDPA, but the pKa value (4.86) was significantly lower (Fig. S7). This suggested that the introduction of aromatic rings and resulting π–π stacking may further decrease the pKa of UPS block copolymers, in addition to hydrophobic interactions.</p><p>To further investigate the π–π stacking effect on the pH response of UPS block copolymers, we synthesized a series of PEO-b-P(MBA-r-C7A) copolymers. The molar fraction of the two monomers can be precisely controlled prior to polymerization, leading to a random copolymerized P(MBA-r-C7A) block with predesigned MBA molar ratio. The pH titration experiments showed that incorporation of more hydrophobic MBA monomers into PEO-b-P(MBA-r-C7A) copolymers resulted in the decrease of pKa (Fig. 4a). Further quantification by plotting the pKa values of PEO-b-P(MBA-r-C7A) copolymers as a function of MBA molar ratio showed a linear correlation (Fig. 4b).</p><!><p>Upon external physical or chemical stimuli, responsive nanomaterials can respond and adapt to surrounding environment through conformation or chemical changes such as self-assembly. The pH-triggered reversible micellization of UPS block copolymers represents a supramolecular self-assembly process, which employs a multitude of non-covalent interactions (e.g., electrostatic and hydrophobic interactions, hydrogen bond, π–π stacking, etc.) to achieve thermodynamically stable nanostructures. Non-covalent interactions besides hydrophobicity may also impact the pH responsive behavior and shift the pKa of UPS block copolymers in aqueous solution. Figure 5 summarized the pKa values of all the UPS block copolymers used in this study as a function of their hydrophobicity of PR segment repeating units. The hydrophobicity of PR segment repeating units in PEO-b-P(C7A-r-MBA) copolymers was calculated as the statistical average of C7A and MBA repeating units. All PEO-b-PR block copolymers with aliphatic alkyl groups as amine substituents, either linear or cyclic, followed linear pKa-LogP correlation. The pKa of PEO-b-PR block copolymers did not change much by replacing the six-membered ring cyclic substituents (C6S1A) with seven-membered rings (C7A) as long as the hydrophobicity of repeating units stayed the same. The geometry of the alkyl substituents, namely PEO-b-iPDPA vs. PEO-b-nPDPA, appeared to affect the pKa of the copolymers, but fit in the same linear curve in the pKa-LogP plot.</p><p>Interestingly, the incorporation of π–π stacking via an introduction of aromatic rings also affect the pH responsive behavior of UPS block copolymers, as proved by the fact that the pKa values of PEO-b-P(C7A-r-MBA) copolymers did not fit in the linear correlation of pKa-LogP that existed in PEO-b-PR block copolymers with only aliphatic alkyl substituents. The observed drastic decrease in pKa of PEO-b-P(MBA-r-C7A) copolymers as a function of hydrophobicity of PR segment repeating units suggested the π–π stacking further lowered the pKa of UPS block copolymers. At the micelle state, the aromatic rings on PR segment were close to each other to form aromatic stacking. At the unimer state, the π–π stacking effect was significantly minimized because the aromatic rings were far from each other as a result of electrostatic repulsion between cationic ammonium groups. Incorporation of π–π stacking stabilized the micelles and shifted the equilibrium to the direction of neutralization of protonated tertiary amines, corresponding to the decrease of pKa. These data also suggested π–π stacking as an additional strategy in fine-tuning the transition pH of the UPS block copolymers.</p><!><p>Self-organization of molecules via multiple intra- and intermolecular hydrogen bonds have served as an important strategy in the development of self-assembled structures12,37. Ions have been known to greatly affect multiple chemical and biological processes in aqueous solution because of their ability to interfere with hydrogen bonding networks and solvent polarity of water38,39. Early reports indicated the increase of NaCl concentration resulted in the decrease of LCST of thermo-responsive PNIPAM38. We first investigated whether a change of ion concentration could impact the pH response of UPS block copolymers. Here we used PEO-b-nPDPA block copolymer and NaCl as a model system. A series of pH titration experiments of PEO-b-nPDPA block copolymers in aqueous solution were performed in the presence of various NaCl concentrations (Fig. 6a). As the NaCl concentration increased from 1 to 150 mM while keeping the polymer concentration the same, the apparent pKa values of PEO-b-nPDPA block copolymers increased from 5.1 to 6.2 (Fig. 6a). Quantitative correlation indicated an exponential increase of pKa as a function of NaCl concentration (Fig. 6d).</p><p>Hofmeister anions have been well known for their effects on the solubility of proteins in aqueous solution, though the underlying mechanism remains elusive40,41. The Hofmeister anion series have been divided into water structure makers (kosmotropes) and breakers (chaotropes) with distinct effect on protein solubility. SO42− and ClO4− anions are classical kosmotropes and chaotropes, respectively. Cl− is considered as a neutral anion. We investigated whether ion species can have a different effect on the transition pH of UPS block copolymers. A series of pH titration experiments of PEO-b-nPDPA micelle solutions were performed using H2SO4 and HClO4 in the presence of Na2SO4 and NaClO4 salts, respectively. As shown in Fig. 6b, increase of SO42− concentration also resulted in the increase of pKa values, although not as notable as Cl−. Most notably, increase of ClO4− concentration resulted in the most drastic pKa increase (Fig. 6c). Quantitative analysis showed that the pKa values of PEO-b-nPDPA were directly proportional to the logarithmic of ionic strength. The slopes of pKa values as a function of ion strength (dpKa/dLog[I]) for SO42−, Cl− and ClO4− were 0.16, 0.49 and 0.85, respectively (Fig. 6d).</p><p>These data demonstrate that both the anion species and concentration have significant impact on the transition pH of PEO-b-PR block copolymers. Increase of salt concentration generally leads to the increase of transition pH of UPS block copolymers. The salt effect on the transition temperature of thermo-responsive nanostructures has been studied in multiple systems6,42. However, such effect on the responsive behavior of pH-sensitive nanomaterials is less investigated. As addressed by previous reports38, interactions among anions, macromolecules and hydration water molecules all have potential impact on the stimuli-triggered supramolecular self-assembly behaviors of responsive nanomaterials.</p><p>We attempt to rationalize the effect of Hofmeister anions on the pH responsive behaviors of PEO-b-nPDPA by three plausible mechanisms (Fig. S8). First, the hydrated anions are capable of polarizing adjacent water molecules which may form hydrogen bonds to the nitrogen atoms on the tertiary amines. The polarization is likely to weaken the hydrogen bond and make the lone pair electron of amine nitrogen more accessible to protons. In this case, increase of salt concentration will favor the protonation of tertiary amines and lead to the increase of pKa values. Second, anions may also interfere with the hydration of hydrophobic surfaces by water molecules by increasing the surface tension at the water/hydrophobic interface. This dehydration effect will lead to the decrease of pKa because of decreased solubility of hydrocarbons. The decrease of pKa as a result of salt-induced increase of surface tension can partly offset the H bond-induced pKa increase. The order of anions' ability in decreasing H-bond interactions between water and tertiary amines and strengthening surface tension is: SO42− > Cl− > ClO4−. Third, direct bonding between anions and cationic ammonium groups can neutralize the positive charges of protonated tertiary amines through formation of ion-pairs. The neutralization will shift the equilibrium to the direction of protonation of amines, corresponding to the increase of pKa. The ability to form stable ion pair interactions is much stronger with chaotropic anions (e.g., ClO4−, I−) than kosmotropic anions (SO42−). As we reported previously39, the order of ion pair strength between ammonium groups and specific anions is: ClO4− > Cl− > SO42−, consistent with prior reports on the binding of anions to amides of PNIPAM38. We attribute that the ion pair interactions play a more dominant role than the other two factors for the observed influence of ClO4− anions on the pKa shift of PEO-b-nPDPA copolymer. PEO-b-PR block copolymers may serve as a good model system for the further delineation of solvation effect, hydrogen bonding and ionic interactions between Hofmeister anion series and synthetic macromolecules.</p><!><p>The effect of ion species and concentration in affecting the pKa values of UPS block copolymers suggest the solution parameters also play a critical role in the pH-triggered supramolecular self-assembly process. Functional nanomaterials for biomedical applications are usually designed to transport therapeutic or diagnostic modalities from the point of administration to the site of action. One potential challenge is to assure that dose dilution in the journey from injected sites to the action sites, such as in blood, will not compromise the performance of the nanomaterials. To address this question, we investigated whether the polymer concentration, a key solution parameter of polymers in water, may affect the transition pH of UPS block copolymers. We used PEO-b-nPDPA as a representative system in this study. The critical micelle concentration (CMC) of PEO-b-nPDPA, measured in the 0.2 M sodium phosphate buffer at pH = 7.4, was around 1 μg/ml (Fig. S9). We performed all the experiments in the concentration range above the CMC.</p><p>In the presence of low NaCl concentration (e.g.,1 mM), the pKa values of PEO-b-nPDPA increased significantly with the increase of polymer concentration (Fig. 7a and 7c). More specifically, when the polymer concentration was increased from 0.2 to 10 mg/ml, pKa values jumped from 4.7 to 5.3 (Fig. 7a and 7c). In the presence of 150 mM NaCl to mimic the physiological conditions, however, the pKa values of PEO-b-nPDPA copolymers stayed the same at 6.2 (Fig. 7b and 7c). High salt concentration normalized the variations in pH transition from different polymer concentrations.</p><p>To further validate the observation, we conjugated a fluorescent dye, tetramethylrhodamine (TMR), to the hydrophobic PDPA segment to evaluate the fluorescent transition pH as a function of polymer concentration in the presence of 150 mM NaCl. At high pH, neutralization of ammonium groups lead to the formation of micelles and resulting quenching of fluorescence. At low pH, protonation of tertiary amines resulted in the dissociation of micelles into unimers, accompanied by resurrection of fluorescence. The fluorescence intensity of PEO-b-nPDPA-TMR copolymers was measured in a series of sodium phosphate buffers with different buffering pH. Plot of normalized fluorescence intensity as a function of pH suggested the fluorescence on/off transition pH stay almost the same (Fig. 7d and Fig S10) over 100-fold change in polymer concentration (0.02 to 2 mg/ml), consistent with the above pH titration data. It is worth noting that the administrated concentration of UPS nanoprobes for in vivo tumor imaging studies varies from 0.5 to 2.0 mg/ml. The nanoprobe concentration in plasma 24 h after intravenous injection was approximately 0.02 to 0.1 mg/ml25. These data suggest that physiological level of salt concentration (i.e., 150 mM NaCl) may be critical to reduce variability from probe dilution and maintain the nanoprobe integrity in biological studies. Besides tumor imaging, the administered concentration of responsive polymers may vary significantly for different applications such as drug delivery or biosensing. Elucidation of polymer concentration-transition pH relationships will help predict and improve the performance of pH responsive nanomaterials.</p><!><p>In this study, we systematically investigated a variety of non-covalent interactions that impacted the pH responsive behavior and resulting supramolecular self-assembly of amphiphilic UPS block copolymers. Increase the strength of both hydrophobic interactions and π–π stacking effect stabilized the micelles, which favored the neutralization of protonated tertiary amines and resulted in the decrease of pKa values. Formation of ion pairs between protonated ammonium groups and chaotropic anions drove the protonation of tertiary amines and led to the increase of pKa values of UPS block copolymers. A series of key parameters that affected the responsive behaviors of UPS block copolymers such as chain length, hydrophobicity of substituents of tertiary amines and salt concentration were identified and evaluated, which help establish useful guidelines for rational development of nanomaterials based pH sensors with predictable and tunable transition pH. Without NaCl, transition pH of UPS block copolymers was polymer concentration-dependent. In the presence of 150 mM NaCl, the transition pH of UPS block copolymers remain unchanged over 100-fold range in polymer concentration. Ensuring the performance of pH sensors across a wide dose range is crucial for biomedical applications such as molecular imaging and drug delivery.</p><p>Introduction of stimuli-responsive moieties has become a general strategy in the design of responsive, functional nanomaterials. Temperature-sensitive materials usually contain amide bonds surrounded by hydrophobic groups (e.g., PNIPAM5, elastin-like peptides6). pH responsive materials are composed of amines43, carboxylic acids44, or pH-sensitive labile bonds24. Disulfide bonds are used to synthesize redox-responsive systems45. These stimuli-triggered supramolecular self-assembly sensors arise from a multitude of molecular interactions that exist ubiquitously in various natural and synthetic macromolecular systems. Self-assembly based on selective and precise control of non-covalent interactions provides powerful and versatile tools for the development of complicated nanostructures at the molecular level. The development of functional nanomaterials has a growing emphasis on identification and optimization of specific design parameters crucial to performance. The key parameters that affect the pH-triggered self-assembly from this study may also serve as useful guideline for tailoring the structure of other stimuli-responsive systems. Moreover, different responsive groups can be introduced in the same polymeric structures for the development of multi-responsive nanomaterials46.</p>

PubMed Author Manuscript

Mechanisms of Resistance to Photodynamic Therapy

Photodynamic therapy (PDT) involves the administration of a photosensitizer (PS) followed by illumination with visible light, leading to generation of reactive oxygen species. The mechanisms of resistance to PDT ascribed to the PS may be shared with the general mechanisms of drug resistance, and are related to altered drug uptake and efflux rates or altered intracellular trafficking. As a second step, an increased inactivation of oxygen reactive species is also associated to PDT resistance via antioxidant detoxifying enzymes and activation of heat shock proteins. Induction of stress response genes also occurs after PDT, resulting in modulation of proliferation, cell detachment and inducing survival pathways among other multiple extracellular signalling events. In addition, an increased repair of induced damage to proteins, membranes and occasionally to DNA may happen. PDT-induced tissue hypoxia as a result of vascular damage and photochemical oxygen consumption may also contribute to the appearance of resistant cells. The structure of the PS is believed to be a key point in the development of resistance, being probably related to its particular subcellular localization. Although most of the features have already been described for chemoresistance, in many cases, no cross-resistance between PDT and chemotherapy has been reported. These findings are in line with the enhancement of PDT efficacy by combination with chemotherapy. The study of cross resistance in cells with developed resistance against a particular PS challenged against other PS is also highly complex and comprises different mechanisms. In this review we will classify the different features observed in PDT resistance, leading to a comparison with the mechanisms most commonly found in chemo resistant cells.

mechanisms_of_resistance_to_photodynamic_therapy

INTRODUCTION<!>1.1 Photosensitizer Subcellular Distribution. The Role of Mitochondria<!>1.2 Multidrug Resistance and P-Glycoprotein<!>1.2.1 Reversal of MDR Phenotype by PDT Treatments<!>1.3 ABCG2-Mediated Transport of Photosensitizers<!>2. DNA ALTERATIONS AND GENE EXPRESSION<!>2.1 DNA Repair and Karyotype<!>2.2 Induction of Early Response Genes and Signal Transduction Pathways<!>3. APOPTOSIS, AUTOPHAGY AND SIGNAL TRANSDUCTION PATHWAYS ASSOCIATED WITH PROGRAMMED CELL DEATH. THE ROLE OF PS SUBCELLULAR LOCALIZATION<!>3.1 Activation of Caspases<!>3.2 Expression of Apoptotic and Antiapoptotic Proteins<!>3.3. Autophagy and the Ubiquitin-Proteasome System<!>3.4 Lipid-Derived Second Messengers: Ceramide<!>3.5 Phosphatidylinositol-3 Kinase<!>3.6 Calcium/Calmodulin-Dependent Kinases<!>4. CELLULAR ANTIOXIDANT DEFENCE MECHANISMS<!>5. HEAT SHOCK PROTEINS<!>6. MORPHOLOGY, CELL ADHESION, CYTOSKELETON AND METASTASES<!>7. INDUCTION OF CYCLOOXYGENASES<!>8. NITRIC OXIDE<!>9. SURVIVIN<!>10. HYPOXIA<!>11. CROSS-RESISTANCE BETWEEN PDT AND OTHER THERAPIES<!>12. DO PDT RESISTANT CELLS AND CHEMORESISTANT CELLS SHARE THE MECHANISMS OF RESISTANCE?<!>13. CONCLUSIONS AND FUTURE DIRECTIONS

<p>Photodynamic therapy (PDT) involves the administration of a photosensitizer (PS) either systemically or locally, followed by illumination with visible light [1, 2]. The PS absorbs light and, in the presence of oxygen, transfers the energy, producing cytotoxic oxygen species [3].</p><p>Unlike chemotherapy and radiotherapy, PDT involves the combination of two agents: light and the PS. It is possible to obtain differences in the level of resistance when it is expressed in terms of fixed PS concentration and when it is expressed in terms of fixed light dose [4]. Some authors have increased drug dose or drug dose exposure [4, 5] or light dose [6, 7] during the induction of resistance, and in both cases high levels of resistance have been found.</p><p>The mechanisms of resistance ascribed to the PS may be shared with general mechanisms of drug resistance, and may be related to: (i) different uptake rate or efflux, (ii) altered intracellular trafficking of the drug, (iii) decreased drug activation, and (iv) increased inactivation of drug.</p><p>When the photoactivation of the PS occurs, reactive oxygen species are formed [3], and during a first stage, an increased inactivation of toxic species can occur via antioxidant enzymes [8, 9]. In addition, heat shock proteins (HSPs) play a role as intra-cellular chaperones for other proteins, folding and assisting in the establishment of proper protein conformation, preventing unwanted protein aggregation and helping to stabilize partially unfolded proteins [10] and thus help to recover from PDT damage. In a second step after photodamage, an increased repair of drug induced damage to proteins, membranes and even DNA may happen. At this stage, induction of stress response genes occurs after PDT, resulting in modulation of proliferation and cell detachment and inducing survival pathways among other multiple extracellular signaling pathways [11].</p><p>PDT resistant cell lines have only been isolated in vitro, and usually the magnitude of resistance obtained with PDT protocols is less than that reported for most drug-resistant cell lines [6]. Cell lines from multiple PDT-treated tumors have not been yet isolated. However, in vivo, other mechanisms related to host-drug or host-tumor interactions may be relevant.</p><p>In 1991, Luna and Gomer [6] isolated cell lines resistant to PDT. RIF-1 fibrosarcoma cells were exposed to two protocols of Photofrin II (PII)-PDT: short exposure (initial injury associated primarily with the plasma membrane) and long exposure to PII-PDT (associated with damage to organelles and enzymes). In both protocols, the resistant variants displayed a stable level of photosensitization, and a 2.5- to 3.0- log and 1.2- to 1.5-log increase in survival respectively at the highest light doses. In the same year, Singh et al. [5] isolated and characterized two different clones originated from the same parental cell line, exposed up to 8 cycles of PII-PDT with the long exposure protocol, obtaining the so called RIF-8A cell line. The degree of resistance was also similar (2-log difference in cell kill). When the resistant cells were injected to mice, resistance to in vivo PDT was observed. In addition, resistant cells explanted immediately following in vivo PDT, were also resistant to the treatment. However, the resistance index was lower, suggesting that the direct cytotoxic effects of PDT on the tumor cells are not sufficient to cause the PDT response, and supports the role of host-related factors such as damage to microvasculature [12]. Again using the same RIF-1 cells as the parental cell line, Mayhew et al. [4] isolated two strains resistant to polyhematoporphyrin (PHP) and to zinc (II) pyridinium-substituted phthalocyanine (Zn-PCP), and demonstrated a 5.7- and 7.1-fold increase in resistance, respectively.</p><p>It has been demonstrated that the cause of resistance is highly dependant on the cell origin and the PS employed. However, it was not possible to identify any cellular characteristics that can be predictive of their ability to generate resistant PDT variants. The structure of the PS is believed to be a key point in the development of resistance, and this feature is probably related to its particular subcellular localization. The hydrophobic drug PII localize in plasma membrane or in intracellular membranes including mitochondria, depending on the incubation time [13] whereas the hydrophilic aluminium disulphonated phthalocyanine (AlPCS4) and Nile blue A are mainly located in lysosomes [14, 15]. Singh et al. [16] employed these 3 PS with different intracellular localization to induce PDT-resistant variants (Fig. 1). They found various degrees of resistance, and only from the colon adenocarcinoma HT29 line it was possible to generate resistant variants employing the three PS. From the bladder cell HT1376 line, only resistance to Nile blue A was achieved, whereas the SK-N-MC neuroblastoma did not develop any resistance at all. Thereafter, induced resistance appears to be towards the drug itself and not necessarily toward photosensitization.</p><p>We have demonstrated [7] that it was possible to isolate cells resistant to PDT employing a precursor of a PS. In recent years, 5-aminolevulinic acid (ALA)-mediated PDT has become one of the most promising fields in PDT. ALA is the pro-drug of the PS Protoporphyrin IX (PpIX). After ALA administration, cells generate PpIX through the haem biosynthetic pathway. ALA induces PpIX accumulation preferentially in certain tumor cells, primarily due to the reduced activity of ferrochelatase, the enzyme responsible for the conversion of PpIX into heme [16] and a relative enhancement of deaminase activity, the enzyme responsible for the passage of ALA to Uroporphyrin [17].</p><p>We developed two clones resistant to ALA-PDT from a murine mammary adenocarcinoma cell line. The clones exhibited 6.7 and 4.2 –fold increase in resistance, respectively. On the contrary, no evidence of PDT resistance was found in response of human glioma spheroids to repetitive ALA-PDT [18], showing again that resistance to PDT comprises a broad number of aspects and not all the cell types and cell models develop resistance to the same PS.</p><p>It is important to study the mechanisms of PDT resistance because this information can be used to improve combinations of treatments such as PDT plus chemotherapy or radiotherapy. It is also relevant to further elucidate the mechanisms of action of PDT, and to study the relationship between PDT cytotoxicity and cellular changes. In addition, comparison of the photosensitivity of tumors grown in vivo from cells with different PDT sensitivities induced in vitro, may also help to elucidate the role of the vasculature in PDT-induced damage [12]. Developing methods to measure PDT dosimetry, and establishing the number of PDT cycles required for optimal treatment and the cellular mechanisms modified by PDT are all necessary in lieu of an increase in the PDT applications [19–22]. In this review we will classify the different features observed in relation to PDT resistance, leading to a comparison with the mechanisms most commonly found in chemoresistant cells.</p><!><p>RIF-1 PII-PDT-resistant clones isolated by Luna & Gomer [6] accumulated a slightly increased amount of PS per cell, but on the other hand, a slightly lower amount of PII on a per mg protein basis. However, a decreased expression and function of alpha-2 macroglobulin receptor/low density lipoprotein receptor, involved in the transport of PII, suggested that the modulation of PII uptake and/or subcellular localization occurs in PDT resistant cells [23].</p><p>The PII-PDT resistant clones isolated by Singh et al. [5] displayed also similar amounts of porphyrin fluorescence per unit cell volume compared with the parental lines. They performed colocalization studies employing Rhodamine 123 and ION-nonyl acridine orange (NAO). Rhodamine 123 accumulates in the mitochondrial matrix and is a good indicator of mitochondrial membrane function [24]. NAO binds to cardiolipins in the inner mitochondrial membrane and is an indicator of mitochondrial number density [25]. These studies reflected that at long time exposure to the PS, Rhodamine 123 and PII have a weaker colocalization in the resistant variants. In addition, PII competence with NAO was more marked in the resistant lines, suggesting that the inner mitochondrial membrane is a significant PII binding site and may be related to the mechanism of resistance [26].</p><p>In our laboratory [7] we found that the amount of porphyrins synthesized per cell in the resistant clones to ALA-PDT was similar to the parental cell line, but when it was expressed per mg protein, there was a 2-fold decrease. This means that less porphyrins are available to target the same amount of proteins and, as it has already been demonstrated; proteins are a target for PDT [27]. If the amount of porphyrins and not the target molecule is the limiting factor in photodamage, this feature can lead to development of resistance.</p><p>In addition, we reported in the same clones, alterations in the enzymes of the haem pathway that produces PpIX, with a higher proportion of hydrophilic porphyrins (Table 1, Fig. 2). It has been shown that hydrophilic porphyrins such as coproporphyrin, and uroporphyrin are poor photosensitizers [28, 29], and this feature was found to be related to its membrane partitioning behavior and consequently, with lower cell uptake [30]. In addition, different subcellular localization of these hydrophilic porphyrins may also contribute to the resistance.</p><p>We have also found an increased number of mitochondria per cell, which is particularly important in the case of ALA-PDT, since the last steps of ALA conversion into PpIX take place in mitochondria [7]. There is little doubt that mitochondria are critical targets in the actions of PDT. PS, in particular porphyrins, significantly damage mitochondria [31, 32] causing inactivation of numerous mitochondrial enzymes, inhibition of adenosine triphosphatase and uncoupling of oxidative phosphorylation [33, 34]. Moreover, as it will be discussed, PS localized in mitocondria are able to induce apoptosis very rapidly [35, 36].</p><p>Several mitochondrial alterations have been reported in cells exposed to various selective pressures, including antineoplasic treatments [37–39]. The downregulation of mitochondrial RNA may represent a general mechanism by which cells protect themselves against oxidative stress. At least 7 of the 15 mRNA and rRNA encoding gene products in mitochondria are down-regulated by oxidative stress, probably representing an early stage "shut down" of mitochondria [40]. Shen et al. [41] used messenger RNA differential display to identify genes that were differentially expressed, and they found a reduction of mitochondrial 16S rRNA and NADH dehydrogenase subunit 4 in the PDT-resistant variants of HT29 human colon adenocarcinoma.</p><p>Singh et al. [42] assessed the response to mesoporphyrin-PDT of a cell line lacking mitochondrial DNA, and they found that this line was extremely resistant to PDT as well as Doxorrubicin (DXR) treatment, but not to alkylating agents or γ-irradiation. Although mitochondria have been extensively involved in the apoptotic cell death [43], the resistance was not due to changes in apoptosis. DXR activation in mitochondria requires reduction to semiquinone free radicals by Complex I in mitochondria, and then it is reoxidized to reactive oxygen species. These studies indicate that free radicals produced by mitochondria must play a critical role in cell death induced by both DXR and PDT.</p><p>Changes in mitochondria of the RIF clones resistant to PII-PDT were described by Sharkey et al. [44]. The study showed that the mitochondria of the resistant RIF-8A cells were smaller, more electron dense and higher in cristae density than the parental RIF-1 cell line. The total mitochondria area per cell in the resistant line was double that of the parental line. In addition, the ATP content and succinate dehydrogenase activity of the resistant cells were higher and oxygen consumption rates were similar to the parental cell line. That is, for an equivalent rate of oxygen consumption, resistant cells contain a greater intracellular ATP pool, suggesting an altered energy metabolism. On the other hand, the RIF-8A resistant cell line accumulated less Rhodamine 123, suggesting a decrease in mitochondrial potential.</p><p>We found [7] that ALA-PDT resistant clones displayed higher protein content and increased number of mitochondria, together with a higher oxygen consumption. However, when normalized per protein content, the number of mitochondria was similar for both cell lines. It is also noteworthy that although the number of mitochondria is higher in the resistant cells, PpIX synthesis which takes places in mitochondria, is not increased</p><!><p>Multidrug resistance (MDR) is the major confounding factor in solid tumour chemotherapy [45, 46]. MDR is a complex phenomenon that may be caused simultaneously by several mechanisms functioning in one and the same cell. Although various mechanisms involved in MDR can be identified, it remains a major problem in oncology. These mechanisms include: (i) the enhanced activity of drug pumps, i.e. ABC or alternative transporters, (ii) modulation of cellular death pathways, (iii) alteration and repair of target molecules, and (iv) other less common mechanisms. Together they build a complex network of cellular pathways and molecular mechanisms mediating an individual MDR phenotype [47].</p><p>P-glycoprotein (P-gp), also called ABCB1 is coded by the MDR1 gene, and it is one of the ATP-binding cassette (ABC) drug transporters held responsible for the phenomenon of multidrug resistance acting as drug efflux pump for antineoplastics with broad substrate specificity.</p><p>No overexpression of P-gp was found in cell lines obtained from multiple PDT treatments. The RIF-1 PDT-resistant cells isolated by Luna & Gomer [6] did not exhibit a MDR phenotype, by means of mRNA analysis of P-glycoprotein. In line with these findings, the PII-PDT resistant variant RIF-8A isolated by Singh et al. [5] from the same parental cell line, showed (i) similar uptake of DXR, (ii) no cross-resistance, and (iii) similar amounts of P-gp expression. However, the same RIF-8A cells were cross-resistant to cisplatin treatment [48].</p><p>In addition, the reverse situation is also variable, since many but not all MDR resistant cell lines have been shown to be resistant to PDT. Singh et al. [5] showed that the Chinese ovary hamster CHO-MDR line derived from multiple chemotherapy treatments, showed cross-resistance upon exposure to PII-PDT, and this was correlated with a lower PII accumulation. Similarly, Kessel et al. [49], employed copper benzochlorin iminium salt, a cationic PS, and demonstrated cross-resistance to PDT in P388/ADR murine leukemia cells resistant to DXR, because of impaired cellular accumulation of the PS due to P-gp efflux pump activity.</p><p>On the other hand, Kessel and Erickson [50] have shown that the same murine leukemia P388/ADR cells were not cross-resistant to mesoporphyrin-PDT, demonstrating the dependence of the PS structure on the affinity for the P-gp.</p><p>MCF-7 TX200 mammary carcinoma cells overexpressing multidrug resistant associated protein 1 (MRP1) or P-gp showed equal intracellular accumulation of chlorins, porphyrin-based PS and pheophorbides compared to the MCF-7 control cells [51]. Moreover, MCF-7/DXR cells (DXR resistant variant) were more sensitive to meta-tetra(hydroxyphenyl)chlorin (m-THPC)-PDT than its parental cells [52].</p><p>Robey et al. [51] suggested that the pro-PS ALA is not a P-gp substrate. They observed a negligible increase in intracellular ALA levels upon incubation of MCF-7 TX200 transfected cells with a P-gp inhibitor. Li et al. [53] also observed similar PpIX levels in MDR resistant leukemia cells exposed to ALA, and only small differences upon incubation with and without the P-gp inhibitor verapamil.</p><p>On the other hand, employing ALA derivatives, Chu et al. [54] showed that P-gp has certain affinity for either hexyl-ALA or PpIX. Human uterine sarcoma cells MES-SA-Dx5 overexpressing P-gp, showed reduced intracellular levels of PpIX derived from hexyl-ALA but to a limited degree, and this mechanism could be reversed by using P-gp inhibitor verapamil. P-gp expression was also related to a slight reduction in hexyl-ALA photosensitivity.</p><p>There are also some interesting but still unexplained findings about the relationship between MDR cell phenotype and PDT response. Tsai et al. [55] found that MCF-7/DXR cells accumulated a lower level of PpIX from ALA, as compared to the parental MCF-7. However, ALA-PDT was still less effective for MCF-7/DXR cells than MCF-7 cells even with similar amounts of PpIX, indicating that the resistant cells might possess intrinsic mechanisms that render them less sensitive to ALA-PDT and is not related to MDR efflux of PpIX.</p><p>Another unexpected but significant finding was that hexyl-ALA-PDT induced a drug and light dependant decrease in MDR1 mRNA levels in uterine fibrosarcoma MES-SA-Dx5 cells (resistant to DXR) together with a concomitant decreased expression of P-gp [54]. Similarly, Pheophorbide-PDT of multidrug resistant HepG2 cells induced c-Jun N-terminal Kinase activation, leading to a down-regulation of P-gp [56].</p><p>The influence of PS distribution on the MDR phenotype of P-gp overexpressing cells was reported in the last few years. Selbo et al. [57] showed that MES-SA-Dx5 cells were more resistant to PDT with disulfonated meso-tetraphenylporphine (TPPS2a). The process was not mediated by P-gp classical mechanism, as there were no differences in the uptake and efflux of TPPS2a as compared with the parental cell line. The authors suggested possible differences in endocytic vesicle localization of TPPS2a, speculating that the lysosomal targeting by PDT induces a stronger cytotoxic effect than PDT of endosomes. According to later investigations of Chu et al. [54] and Tang et al. [56], this finding may be related to an indirect downregulation of MDR, or alternatively, one of the MDR-associated mechanisms different from drug efflux.</p><p>Merlin et al. [58] found that neither Chlorin e6 accumulation nor efflux was different in MCF-7 and MCF-7/DXR overexpressing P-gp, but its subcellular distribution was different between both the cell lines. The presence of P-gp inhibitor restored the distribution of the PS and was found to potentiate the Chlorin e6-PDT to a similar extent in both cell lines.</p><p>Based on the information available, it appears that MDR confers a degree of PDT resistance in certain cases, and this resistance is strongly dependant on the structure of the PS and its affinity for the P-gp. However, there are no rules for cross resistance, and dependence on the cell type and PS type are important factors, as well as intracellular distribution of the PS.</p><!><p>Regarding vascular-targeted PDT, Preise et al. [59] have shown that P-gp-expressing human HT29/MDR colon carcinoma cells in culture were resistant to PDT with palladium-bacteriopheophorbide WST09 (TOOKAD). TOOKAD is a chemical entity of a new generation of hydrosoluble PS, with high binding to albumin and pure focal intravascular effect (Fig. 3). In contrast to the rest of PS, the molecule does not extravasate and remains constrained in the circulation. However, TOOKAD-PDT induces tumor necrosis with equal efficacy in HT29/MDR-derived xenografts and their wild type counterparts, demonstrating that the vascular-targeted PDT bypasses drug resistance.</p><p>In addition, photochemical internalisation (PCI) was reported to overcome chemoresistance in several MDR cell lines employing different PS and therapeutic drugs [57, 60, 61]. PCI involves localization of PS together with the drug of choice in endocytic vesicles within target cells, where the PS is specifically localized to the vesicular membrane [62]. Endocytosed macromolecules and PS are exposed to light, causing an efficient delivery of the drugs into the cytosol.</p><p>Lou et al. [60] showed that MCF-7 and MCF-7/ADR cells were equally sensitive to PCI with TPPS2a. In MCF-7/ADR cells preloaded with DXR, the drug was released into the cytosol after PCI treatment and entered cell nuclei, as was seen in MCF-7 cells without PCI, thus reversing the MDR phenotype by endo-lysosomal release of the drug. On the other hand, no PCI-induced increase in DOX sensitivity could be observed in MES-SA and MES-SA-Dx5 cells employing TPPS2a as a PS [57].</p><p>Adigbli et al. [61] employed PCI and co-administration of hypericin with mitoxantrone without alterations in P-gp expression, and were able to overcome the resistance of bladder MGHU-1 and breast cancer MCF-7 cells and their P-gp-overexpressing MDR subclones.</p><p>Selbo et al. [57] also evaluated the reversal of resistance induced by PCI of macromolecules that are not the target of ABC drug pumps such as the plant toxin gelonin and adenovirus. MES-SA and MES-SA-Dx5 cells were equally sensitive to PCI of gelonin (ribosome-inactivating protein) even though the endocytosis rates were lower in the MDR cells. The two cell lines are equally sensitive to PCI of gelonin at the lower light doses. At higher light doses the MES-SA/Dx5 cells are more sensitive to PCI of gelonin than the MES-SA cells. After adenoviral infection, PCI enhanced the fraction of transduced cells substantially, in both cell lines, suggesting the potential use of PCI of macromolecular therapeutic agents that are not targets of P-gp as a strategy to kill MDR cancer cells.</p><!><p>In addition to P-gp, another ABC transporter capable of causing cancer drug resistance has been described. Overexpression of a novel member of the G subfamily of ABC transporters was described in the cell line MCF-7/AdrVp. The new transporter was termed the breast cancer resistance protein (BCRP), and was formally designated as ABCG2 [63, 64]. Like all members of the ABCG subfamily, ABCG2 is a half transporter. The spectrum of anticancer drugs effluxed by ABCG2 includes mitoxantrone, camptothecin-derived and indolocarbazole topoisomerase I inhibitors, methotrexate, flavopiridol, and quinazoline ErbB1 inhibitors [65]. ABCG2 is believed to function as a component of the organism's defence against toxicity by restricting the entry of genotoxins from the intestinal tract into the organism and by facilitating the removal of toxic metabolites from the organism via bile or urine [66].</p><p>Studies with an ABCG2 knockout mouse have provided evidence for the ability of the transporter to efflux PpIX and protect cells from phototoxicity [67]. Tsunoda et al. [68] have also demonstrated correlation between the expression of ABCG2 and resistance to PII-PDT. Robey et al. [51] hypothesised that it may be involved in resistance to PDT. Based on their finding that the PS Pheophorbide a, is an ABCG2 substrate [69], they explored its ability to transport PS with a similar structure. ABCG2-overexpressing NCI-H1650 MX50 bronchoalveolar carcinoma cells were found to have reduced intracellular accumulation of Pyropheophorbide a methyl ester, Chlorin e6 and PpIX from ALA. On the contrary, intracellular accumulation of hematoporphyrin IX, meso-tetra(3-hydroxyphenyl)porphyrin (m-THPP), and m-THPC was not altered (Fig. 4). On the other hand, ABCG2-transfected human embryonic kidney HEK-293 cells were resistant to PDT with pheophorbide a, pyropheophorbide a methyl ester, Chlorin e6 and ALA but not to m-THPC. These studies suggest that the degree of ABCG2 mediated-resistance changes significantly with the PS employed and cell line.</p><p>Of great importance in the outcome of PDT, it has been suggested that decreased expression of ABCG2 may be a widespread phenomenon in human cancers. Gupta et al. [70] showed down-regulation of ABCG2 mRNA with malignant change in 12 different tissues in arrays of paired normal and cancer cDNAs. They reported also down-regulation at the mRNA level of ABCG2 in human specimens of colorectal and cervical cancer.</p><p>In addition, tyrosine kinase inhibitors can block the function of ABCG2. Liu et al. [71] tested the effects of these inhibitors on the response of PDT-treated cells. They employed human and mouse cell lines with a range of ABCG2 expression, as well as a control cell line transfected with ABCG2: i) BCC-1 cells from human basal cell carcinoma and RIF-1 fibrosarcoma cells (high expression), ii) Colo 26 colon carcinoma cell (moderate expression) and iii) human head and neck squamous cell carcinoma line FaDu (no expression). Efflux of 2-(1-hexyloxethyl)-2-devinyl pyropheophorbide-a (HPPH), PpIX from ALA and Benzoporphyrin Derivative monoacid ring A (BPD-MA) was shown in ABCG2+ cells but PII and a novel derivative of HPPH conjugated to galactose were minimally transported (Fig 5). HPPH and PpIX were more effectively transported than BPD-MA, showing a PS-dependent variation on the affinity for this transporter. The tyrosine kinase inhibitor Imatinib mesylate increased accumulation of HPPH, PpIX, and BPD-MA in ABCG2+ cells, but not in ABCG2- cells, and enhanced PDT efficacy both in vitro and in vivo in a RIF-1 tumour model, demonstrating that the inhibition of ABCG2 transport may enhance efficacy and selectivity of clinical PDT. The structure of the PS is a keypoint in the resitance mediated ABCG2. In this study, the multimeric molecule Photofrin is not an ABCG2 substrate, and in addition, monomeric agents but carbohydrate conjugation to a pyropheophorbide molecule blocks transport, as do the modifications in porphyrins and chlorins [51].</p><p>It still remains under debate as to whether or not Photofrin is a substrate of BCRP. Usuda et al.. [72] showed that human epidermoid carcinoma A431 cells overexpressing ABCG2, were resistant to PII-PDT but not to Mono-L-aspartyl chlorin e6 (NPe6-PDT), which has a similar structure to m-THPC (Fig 4), and the resistance was reversed by Fumitremorgin C, a non tyrosine kinase inhibitor of ABCG2. In accordance, a higher expression of ABCG2 in tumor samples obtained from patients with centrally located early lung cancers was inversely correlated with efficacy of PII-PDT but the correlation was restricted to small lesions.</p><p>Jendzelovský et al.. [73] reported for the first time the modulation of ABC transporters by a PS. They showed an increased activity of MRP1 and ABCG2 in HT-29 colon cancer cells treated with hypericin treatment without light. In addition to baseline ABCG2 expression, hypoxia, which is very common in tumors, has been found to up-regulate expression of ABCG2 and to increase cell survival by decreasing intracellular accumulation of heme and other porphyrins [74]. Therefore, hypoxia may inhibit PDT not only because the photodynamic process requires oxygen, but also through ABCG2-mediated decrease in intracellular photosensitizer levels.</p><p>To sum up, similarly to P-gp, resistance to PDT conferred by ABCG2 transporter varies significantly with the cell line and the PS employed, and ABCG2 inhibitors can reverse this PDT resistant phenotype.</p><!><p>PDT activates several signaling pathways, which in turn alter the expression of many different downstream genes. After PDT, inhibition of DNA and RNA polymerases and synthesis of DNA, RNA and protein have been demonstrated [41]. Some of the signal transduction pathways triggered by PDT are stress responses aimed at cell protection, while others are likely to contribute to the cell death process. Many PS bind to various cytoplasmic membranes but are not found in the nucleus and do not bind to DNA. Nevertheless, some DNA damage is produced that can lead to mutagenesis, the extent of which is dependent on the PS, the cellular repair properties, and the target gene [75]. Moreover, a number of investigators have shown that DNA damage is induced following PDT [76–79].</p><!><p>Cell variants resistant to PII-PDT derived from the radiation-induced murine RIF cells [5] are cross resistant to UV light. And the LY-R murine leukaemia cell line, which is deficient in the repair of UV-induced DNA damage, is also cross-sensitive to PDT [78,80]. This suggests some overlap in the type of cellular damage induced by UV and PDT and/or an overlap in the pathways for the repair from damage.</p><p>DiProspero et al. [81] employed an assay for adenovirus DNA synthesis as an indicator of recovery from PDT in the RIF-8A cells resistant to PII-PDT. An increased capacity for viral DNA synthesis was observed in the RIF-8A cells compared with the parental cells, suggesting that the increased resistance to PDT resulted from an elevated recovery and/or repair from DNA damage. The capacity of UV-irradiated cells for viral DNA synthesis was also greater for RIF-8A cells, indicating a cross-resistance to UV. While RIF parental cell line shows a mixture of diploid and tetraploid subpopulations, some of the RIF PDT-resistant variants have a complete tetraploid phenotype [6], suggesting that DNA damage is somewhat involved in photodamage.</p><p>On the other hand, employing the C3H 10T 1/2 mouse embryo cell system, Gomer et al.., [82] found that at the DNA level, PII-PDT does not induce any mutations. In addition, the effect of loss of DNA mismatch repair activity on the sensitivity to m-THPC-PDT was tested employing DNA mismatch repair-deficient cells, and it was found that this feature did not contribute to PDT resistance [83]. This controversy suggests that upon certain conditions, the DNA damage can influence resistance to PDT depending on PDT conditions, the PS employed, and the cell model used.</p><!><p>Activation of the early response genes does not require protein synthesis and is usually of a transient nature. Oxidants such as physical, biological, and chemical stresses including ultraviolet irradiation, growth factors, and tumor promoters induce a family of early response genes [84, 85]. Protein products of the early response genes act as transcription factors and thereby regulate the expression of a variety of genes via specific regulatory domains. The application of high-resolution microarray platforms to the gene expression after PDT has revealed the involvement of several genes related to survival signals that could be responsible of the development of resistance to PDT. Early-response genes mostly upregulated after PDT-mediated oxidative stress encodes transcription factors such as JUN, FOS, MYC, EGR-1, NF-κB, ERK, JNK and p38MAPK, among others [11, 86–90] (Table 2).</p><p>FOS and JUN play a role in cell proliferation, apoptosis, and stress response [91] are generally upregulated by stress and cell damage, and are the most commonly induced early response genes after PDT. The JUN and FOS proteins together form the activator protein-1 (AP-1), which has a function in apoptosis modulation, cell proliferation, and cell survival. Krammer group studied gene expression after ALA-PDT employing cDNA-array techniques. They found a strong induction of expression of the immediate early genes c-JUN and c-FOS, FOSB and p55-c-FOS after ALA-PDT of the squamous cell carcinoma line A-431 [11, 90, 91]. Similarly, ALA-PDT induced continuous upregulation of c-FOS in one normal urothelial (UROtsa) and two tumor cell lines (RT4, urothelial; HT29, colonic) [92]. Accordingly, Luna et al.. [86] have shown that PII-PDT mediates induction of FOS through protein kinase–mediated signal transduction pathways.</p><p>Pheophorbide-PDT of multidrug resistance HepG2 cells induces c-JUN N-terminal Kinase activation leading to activation of intrinsic apoptotic caspases and down-regulation of P-gp [56].</p><p>Activation of several cell survival signal transduction pathways including protein kinase C (PKC) [92], Etk/Bmx tyrosine kinase [94], protein kinase B (PKB/Akt) [95, 96], mitogen-activated protein kinases (MAPK) [97], Phosphatidylinositol 3-kinase (PI3K) [96, 97], and extracellular signal regulated kinases (ERKs) [98] have also been reported after PDT treatment.</p><p>CDNA arrays were also used as a tool to discover several signal transduction pathways induced by PDT treatment. Hypericin-PDT was found to induce in the human squamous cell carcinoma cell line A-431 several genes involved in various metabolic processes, stress-induced cell death, autophagy, proliferation, inflammation and carcinogenesis thus pinpointing the coordinated induction of a cluster of genes involved in the unfolded protein response pathway after endoplasmic reticulum stress and in antioxidant response [90].</p><p>Mitogen-activated protein kinases (MAPK) signal transduction pathways are involved in the regulation of numerous physiological processes during development and in response to stress. Analysis of PDT-treated cells after p38MAPK inhibition or silencing unraveled that the induction of an important subset of differentially expressed genes regulating growth and invasion, as well as adaptive mechanisms against oxidative stress, is governed by this stress-activated kinase. P38 MAPKs are members of the MAPK family, and p38MAPK inhibition blocked autonomous regrowth and migration of cancer cells escaping PDT-induced cell death [99].</p><p>Sanovic et al. [90] found that the most highly upregulated gene following hypericin- PDT of A-431 cells model is DUSP1, the dual specificity phosphatase 1. DUSP1 is an early immediate gene acting as inactivator of MAPK, and it is overexpressed after oxidative/heat stress and growth factors [100]. Upregulation after PDT reached a maximum of 243-fold which is very likely to be induced by oxidative stress. Being a negative regulator of ERK, JNK or p38MAPK, DUSP1 is presumably the main switch for inactivating all these pathways, especially proliferation signaling, and inducing apoptosis [90]. DUSP1, was both up-regulated in normal and tumor cells after ALA-PDT [92].</p><p>Following exposure to stress agents, various degrees of histone H3 modification at the DUSP1 chromatin may occur and it has been suggested that chromatin remodelling after stress contributes to the transcriptional induction of DUSP1 [101]. A concomitant upregulation of H3b was also found in Hypericin-PDT treated cells, thus suggesting that DUSP1 is possibly activated via H3 histone modifications [90]. Similar results were found by Buytaert et al. [99] with an upregulation of the genes encoding for histones H2A and H2B after hypericin-PDT.</p><p>ETR101 is another immediate early gene involved in cellular stress response, of which mRNA levels were up-regulated in colonic HT29 cells after ALA-PDT but down-regulated in tumor urotheilal RT4 cells [92], showing some cell line specificity.</p><p>The growth and differentiation factor 15 (GDF15) protein is a signal transducion factor in cellular response to injuries and seems to be expressed in an organ-independent manner and after a severe deadly stimulus as a cellular response and in attempt to survive. GDF15 gene was found to be regulated following ALA-PDT. The normal urothelial cell line UROtsa with apoptotic behavior following photodynamic therapy showed no regulation of this gene. On the other hand, RT4 tumor urothelial as well as HT29 tumor colon cells with a clear necrotic response to photodynamic therapy showed a strong activation of GDF15 RNA expression levels [92].</p><p>Ras proteins comprise a group of small GTP-binding proteins with essential roles in controlling the activity of crucial signaling pathways regulating normal cellular proliferation [102]. Mutations at the hot-spots in Ras proteins lead to defects in GTPase activity and constitutive activation of downstream signals. Ras proteins are constitutively activated in around 20 to 30% of human tumors, indicating the importance of this signaling pathway during carcinogenesis. Indeed, constitutive activation of Ras protein contributes significantly to several aspects of the malignant phenotype, including the alteration of tumor-cell growth and invasiveness [102, 103]. In this sense, it has been described that overexpression of Ras proteins are also involved in the resistance to cell death [104, 105].</p><p>We have found that Ras oncogene confers resistance to ALA-PDT [106] as well as PDT with other PS such as PII, merocyanine 540 (MC540), BPD-MA, acridine orange and m-THPC [107] (Fig. 6). In the mammary Ras transfected cells employed, PII, BPD-MA and m-THPC mainly localizes in mainly mitochondria and endoplasmatic reticulum. On the other hand, acridine orange exhibits a lysosomal pattern and MC540 is localized in plasma membrane, Golgi, mitochondria and reticulum. In this case, PS of very different structure and subcellular localization, are equally resistant to photodynamic treatment, showing that Ras oncogene induced resistance, appears to be toward PDT and not toward a particular PS.</p><p>It has also been shown that oncogenic activation of H-Ras as well as PI3K in murine keratinocytes can prevent cell death induced by immunological disruption of E-cadherin adhesion [96]. In addition, Zn-phthalocyanine (Zn-PC)-PDT photodamage is bypassed in cells showing constitutive activation of H-Ras and PI3K concomitant with the expression of phosphorylated Akt (Fig 6).</p><!><p>The induction of apoptosis by most physiological stimuli or toxic agents, proceeds through a series of signaling pathways, and PDT, as it has been previously addressed, has been found to upregulate numerous signaling pathways. Some of these signals act as mediators or promoters of apoptosis in PDT-treated cells, and some are stress responses whose function is to promote repair or tolerance of damage [108], which can be related to the appearance of resistance.</p><p>Several indirect evidences indicate that the complex machinery of apoptosis is directly related to induction of resistance particularly in models of gene transfections. However, a few studies have demonstrated direct evidence of altered apoptosis pathways in cells rendered resistant to PDT by multiple treatments.</p><p>The initial step in the photodynamic process involves localization of the photosensitizing agent at subcellular loci. These can be highly specific or quite broad, and have been reported to include the endoplasmic reticulum (ER), mitochondria, Golgi, lysosomes (Fig 7) and plasma membrane (Fig 8). Most PS are relatively hydrophobic and will be attracted to membranes. There are some exceptions to this rule, e.g., the sulfonated porphyrins/phthalocyanines and NPe6. Even these molecules, although having substituents that render them watersoluble, bind to membranes because of their hydrophobic ring systems [109]. Kessel group have extensively studied subcellular localization of PS and its relationship with the type of cell death [36].</p><p>Depending on localization of the photosensitizing agent, the process can induce photodamage to the endoplasmic reticulum, mitochondria, plasma membrane, and/or lysosomes. When ER or mitochondria are targeted, antiapoptotic proteins of the Bcl-2 family are especially sensitive to photodamage. However, targeting of the plasma membrane by a PS may lead to either a marked delay or inhibition of apoptosis and rescue responses are initiated, even if other sub-cellular sites such as antiapoptotic proteins are also targeted for photodamage [35, 110] (Fig 7).</p><p>The same group [111] evaluated the PDT responses to two structurally related photosensitizing agents, using P388 murine leukemia cells. Photodamage mediated by tin etiopurpurin (SnET2) involved lysosomes and mitochondria and yielded a rapid apoptotic response within 1 h after PDT. A drug analog, tin octaethylpurpurin amidine (SnOPA), targeted lysosomes, mitochondria and cell membranes; apoptotic nuclei were not observed until 24 h after PDT. These results suggest that membrane photodamage can delay or prevent an apoptotic response to PDT thus resulting in resistance to PDT. Similarly, Dellinger et al. [112] reported that cells exposed briefly to a high concentration of Photofrin, then irradiated, exhibited an aborted form of apoptosis and attributed this to leakage of cytoplasmic material through photodamaged membranes.</p><p>To provide an explanation for the ability of SnOPA and a monocationic porphyrin (MCP) (Fig 8) to delay or inhibit the apoptotic response to mitochondrial or lysosomal photodamage, Kessel group tested the hypothesis that this might derive from relocalization of PS during irradiation, resulting in photoinactivation of enzymes required for the apoptotic process [110] They provided evidence that relocalization to the cytosol occurs during irradiation.</p><p>Fluorescence localization studies on three sensitizers: SnET2, 9-capronyloxy-tetrakis (methyoxyethyl) porphycene (CPO) and m-THPC which had initially been classified as targeting mitochondria, revealed that these agents bind to a variety of intracellular membranes [113]. The apoptotic response to these PS is derived from selective photodamage to the antiapoptotic protein Bcl-2 while leaving the proapoptotic protein bax unaffected. Both CPO and m-THPC induced rapid apoptotic response whereas SnET2 was also associated with lysosomal photodamage eliciting a delayed apoptotic response.</p><p>A different localization pattern was observed using two dicationic porphyrins bearing positively charged –N(CH3)3 groups on adjacent or opposite phenyl groups attached to the bridging carbons of a porphyrin structure. The compound were 5,10-di[4-(N-trimethylaminophenyl)-15,20-diphenylporphyrin (DADP-a) and 5,15-di[4-(N-trimethylaminophenyl)-10,20-diphenylporphyrin (DADP-o) respectively (Fig 7) [114]. DADP-a is an amphyphillic structure that can penetrate the plasma membrane and selectively binds to mitochondria. The DADP-o structure has a different charge distribution, resulting in lysosomal affinity. DADP-a- PDT resulted in a rapid loss of the mitochondrial membrane potential, usually a prelude to apoptotic cell death. In contrast, DADP-o-PDT induced extensive lysosomal photodamage, being the first pathway more efficacious than the latter.</p><p>NPe6 (Fig 7) is an amphiphilic PS that bind to endosomal/lysosomal membranes. Upon exposure to light, such membranes will be damaged and become leaky before significant inactivation of lysosomal enzymes can occur. PDT employing NPe6 induces apoptotic response to lysosomal photodamage reflecting an indirect effect mediated by the apoptotic lysosomal pathway involving release of cathepsin B and cleavage of Bid to a truncated form. The latter product can interact with mitochondria resulting in release of cytochrome c, following activation of casapses −3 and −9. This could result in apoptotic response if release of lysosomal enzymes causes sufficient mitochondrial degradation to facilitate release of cytochrome c into the cytosol, triggering the apoptotic program [115].</p><p>Both apoptosis and autophagy can occur after PDT, autophagy being associated with enhanced survival at low levels of photodamage to some cells [116], serving as a pro-survival response via the recycling of damaged organelles. Autophagy offers protection from the phototoxic effects of low-dose PDT, but can serve as an alternate death mode when the PDT dose is increased [117, 118].</p><!><p>According to Almeida et al. [119], who have reviewed intracellular signaling mechanisms in PDT, two major apoptotic pathways have been characterized, the death receptor mediated and the mitochondria-mediated. In both pathways, the activation of initiator caspases (caspase-8 or caspase 9) leads to the activation of effector caspases (caspase 3, caspase 6, and caspase 7). In addition, the lysosomal pathway is a prelude for the mitochondria-mediated apoptosis after lysosomal photodamage [115].</p><p>As addressed above, monocationic PS such as MCP were initially localized in the plasma membrane, and during the first minutes of irradiation, porphyrins migrated to the cytosol [110]. If irradiation continues, photodamage to procaspases −3 and −9 occurs, thereby preventing an apoptotic response. These results may not necessarily be applicable to any PS that initially binds to the plasma membrane, but indicate that the absence of an apoptotic response can result from photodamage to critical elements of the apoptotic program.</p><p>Wild et al. [92] studied RNA expression and protein profiling of a normal cell line (UROtsa, urothelial) and two tumor cell lines (RT4, urothelial; HT29, colonic) following ALA-PDT. Whereas RNA expression of CASP8 was unchanged in the 3 cell lines, a delayed activation of caspase-8 protein was only found in UROtsa cells, whereas no changes were seen in both tumor cell lines, leading to the conclusion that activation of the casapase 8 pathway may serve as a secondary way for the cell to ensure demise in case of damage. Accordingly, Granville et al. [120], found an activation of caspase 8 in HeLa cells treated with BPD-MA-PDT, although they have also found activation of caspases 3, 6 and 7 after PDT.</p><p>Ruhdorfer et al. [11] studied the alteration of the gene expression pattern in the squamous cell carcinoma A-431 after ALA-PDT by cDNA-array technique, and found that the product of the 'Fas-associated via death domain' (FADD) gene was strongly induced. FADD is as an adaptor molecule which interacts with different cell surface molecules and transmits apoptotic signals to the cell. The receiver is procaspase-8, which in the death-inducing signaling complex is activated to caspase-8, leading to the execution of apoptotic cell death.</p><!><p>The Bcl-2 family of proteins acts at a central decision point in the apoptotic pathway. The family is divided into two functional groups: i) antiapoptotic members: Bcl-2, Bcl-XL and CED-9 and ii) proapoptotic members including Bax, Bak, BNIP3, as well as the BH3-only subfamily (Bik, Blk, Hrk, BimL, Bad, Bid) [121, 122] (Table 3).</p><p>Activation of antiapoptotic Bcl-2 proteins have been early observed after PDT-treatment [123, 124], thus being a mechanism supposed to be altered in PDT resistant cells. As explained above, when the mitochondria and/or the endoplasmic reticulum are targeted by photodynamic therapy, photodamage to the anti-apoptotic protein Bcl-2 is observed. On the other hand, lysosomal photodamage ultimately results in activation of the pro-apoptotic protein Bid, also leading to apoptosis [36].</p><p>Shen et al. [41] examined the expression of apoptosis-regulating genes in PDT resistant cells. They found an increased expression of Bcl-2, and heat shock protein 27 (HSP27) together with downregulation of Bax in the HT29 cell PDT-resistant variants. On the other hand, they found an increased expression of the proapoptotic BNIP3 by the use of mRNA differential display, and confirmed by Northern blotting and Western blotting. In addition, the mutant of the tumor suppressor protein, p53 was reduced substantially in the PDT-resistant variants. The same group reported that PDT-resistant HT29 cell lines showed a significant increase in cisplatin sensitivity concomitant with an increase in both spontaneous and cisplatin-induced apoptosis. Cisplatin sensitivity of the PDT-resistant HT29 variants was also correlated with increased BNIP3 and decreased mutant p53 protein levels, but not HSP27 protein levels [125].</p><p>The relevance of the involvement of Bcl-2 was supported by the fact that CHO cells transfected with the antiapoptotic protein were two times more resistant to PDT [126, 127]. Granville et al. [123] subsequently confirmed the ability of overexpressed Bcl-2 to suppress apoptosis in HL60 cells treated with BPD-MA-PDT. This group also found that overexpressed Bcl-2 or Bcl-xL did not prevent the release of cytochrome c from mitochondria but instead blocked the activation of several caspases [120]. In addition, human gastric adenocarcinoma MGC803 cells transfected with the antisense Bcl-2 sequence in a retrovirus vector followed by treatment with hypocrellin-PDT rendered more sensitive to PDT [128]. Antisense Bcl-2 also sensitized A-431 cells to Pc 4-PDT [124].</p><p>However, the usefulness of Bcl-2 expression as a predictor of PDT response is controversial. Kawaguchi et al. [129] showed no correlation between expression of Bcl-2 or p53 and local recurrence after PDT in a series of biopsies of squamous cell carcinomas of the bronchus previous to treatment with PII-PDT. The levels of Bcl-2 have also been measured in biopsies of esophageal tumors treated with PII-PDT, but again no apparent correlation was found [130]. On the contrary, a screening of biopsies from patients with esophageal cancer treated with PDT suggested that Bcl-2 expression is associated with favorable response to PDT [131]. This finding can be explained by experimental studies showing that PDT induces selective degradation of the Bcl-2 protein, leading to apoptosis by decreasing the Bcl-2/Bax ratio. On the other hand, no association of p53 with response to PDT was noticed.</p><p>Xue et al. [132] found that photodamage to Bcl-2 could be induced by Pc 4-PDT in several different cell lines, including human tumor lines. Usuda et al. [133] found that PDT with the same PS, sensitized breast cancer MCF-7c3 cells through Bcl-2 damage. Human breast cancer MCF-7c3 cells expressing stably transfected procaspase-3 were chosen based on its efficient induction of apoptosis in response to Pc 4-PDT. MCF-7c3 cells were transfected with wild-type Bcl-2 or certain deletion mutants lacking one of the membrane anchorage regions (each of which can be photodamaged) which resulted in relative resistance to Pc 4-PDT. This indicates that the deleted regions, which include a caspase-3 cleavage site, are not necessary for the inhibition of PDT-induced apoptosis. In contrast, Bcl-2 mutants, lacking the C-terminal transmembrane domain and do not bind to membrane which is not photodamaged, afforded no protection. These results indicate that the extent of Bcl-2 photodamage may determine the sensitivity of cancer cells to apoptosis and to overall cell killing caused by PDT. Furthermore, overexpression of Bcl-2 also inhibited the activation-associated conformational change of the proapoptotic protein Bax, and higher doses of Pc 4-PDT were required to activate Bax in cells expressing high levels of Bcl-2.</p><p>MCF-7c3 cells treated with Bax antisense oligonucleotides resulted in a 50% inhibition of PDT-induced apoptosis. Similarly, following Pc 4-PDT, apoptosis was completely blocked in Bax-negative human prostate cancer DU-145 cells, and restoration of Bax expression restored apoptosis. However, despite the inhibition of apoptosis, the Bax-negative DU-145 cells were as photosensitive as Bax-replete MCF-7c3 cells, suggesting that for Pc 4-PDT, the commitment to cell death occurs prior to Bax activation [134].</p><p>A similar pattern of Bcl-2 photodamage and cell death is found for other PS. Usuda et al. [135] showed that PII-PDT damaged Bcl-2 and induced apoptosis. However, NPe6-PDT did not damage Bcl-2 and showed a delayed apoptosis as compared with PII-PDT. Bcl-2 overexpressing cells were considerably more resistant to NPe6-PDT than parental MCF-7c3 cells, concluding that PII-PDT damages different molecular targets, and that the extent of Bcl-2 photodamage can determine the sensitivity of cancer cells to apoptosis and to overall cell killing caused by PDT using the mitochondrion-targeting photosensitizer PII, but not the PS lysosomal-targeting NPe6. On the other hand, NPe6-PDT can induce lysosome disruption and initiate the intrinsic apoptotic pathway, since the use of small interfering RNA for Bid afforded a significant protection against cell NPe6-PDT in human lung adenocarcinoma cells [136].</p><p>Ichinose et al. [137] demonstrated that overexpression of wild-type Bcl-2 conferred also relative resistance of MCF-7 cells to PDT with ATX-s10, a PS which localizes not only to mitochondria but also to lysosomes. Pharmacological inhibition of lysosomal cathepsins B and D, protected MCF-7c3 cells from apoptosis caused by ATX-s10-PDT, showing that photolysosomal damage can initiate apoptotic response and this apoptotic pathway can be regulated by photodamage to Bcl-2 via mitochondrial damage. Caruso et al. [138] reported resistance to NPe6-PDT of Tao variant of 1c1c.7 murine hepatoma cells having lysosomal fragility, revealed as reduced cathepsin B and D activities of endosomes/lysosomes. The onset of apoptosis was delayed, and the magnitude of the apoptotic response was muted in Tao cells exposed to NPe6-PDT.</p><p>P53 is a tumor suppressor protein, and also the most frequently mutated gene in human tumors. The increased p53 levelslead to transcription of target genes, cell cycle arrest, or apoptotic cell death depending on the cell type or context [139, 140]. Cells lacking functional p53 fail to undergo these responses, resulting in continued proliferation in the face of genetic damage, subsequent genetic instability, and tumor progression. P53-deficient cancer cells are often less responsive to chemotherapy, which lead to the suspicion that it could be somewhat involved in PDT resistance. However, there are no reports of downregulation of this protein in cells induced resistant to PDT.</p><p>Zhang et al. [128] showed that wild-type p53 transfected-HT29 human colorectal carcinoma cells were approximately two times more sensitive to PDT using a hypocrellin as PS. Similarly, human promyelocytic leukemia HL60 cells expressing wild type p53 were more sensitive to cell killing by PDT, with either PII or tin ethyl etiopurpurin I (SnET2), than cells in which the p53 genes were deleted or inactive. All of these cell lines underwent rapid apoptosis in response to PDT [141]. LS513 human colon carcinoma cell line expressing wild-type p53 was also more sensitive to PII-PDT compared to the mutated [142]. In addition, normal fibroblasts were more sensitive to PII-PDT than immortalized fibroblasts from a patient with Li-Fraumeni syndrome, in which the only p53 allele was mutated [144]. On the other hand, the introduction of the viral oncoprotein E6 to abrogate p53 function in LS513 and MCF-7 cells did not alter their PDT sensitivity [144].</p><p>In spite of the broad evidence of the role of p53 on photodynamic sensitivity, no association of p53 with response to PDT was found in two screenings of biopsies of patients treated with PDT [129, 131].</p><p>An interesting and unexpected finding was the role of p53 in porphyrin-PDT-mediated cell death by direct interaction with the drug, which leads to its accumulation and induction of p53-dependent cell death both in the dark and upon irradiation [145].</p><!><p>In mammalian cells, the autophagy-lysosomal system, in addition to the ubiquitin-proteasome system, represents one of the proteolytic systems for the clearance of PDT-damaged organelles and irreversibly oxidized cytosolic proteins, which are prone to cross-linking and formation of protein aggregates. Accumulating evidence indicates that PDT can stimulate autophagy with functional consequences varying from cytoprotection to the activation of autophagic cell death. The role of autophagy in PDT has been extensively reviewed by Reinners [109]. The induction of autophagy is a common outcome in PDT protocols. It occurs in a variety of cell types, and is not limited to PS that accumulate in specific organelles. PS that preferentially accumulate in late endosomes/lysosomes (i.e., NPe6), ER (i.e., hypericin, CPO), mitochondria (i.e., mTHPC, BPD-MA), or ER + mitochondria (i.e., Pc 4) all induced autophagy following irradiation.</p><p>In cells with defective apoptosis, authophagy it is believed to play a crucial role for cell sensitivity to PDT [146]. Loss of Bcl-2 function could lead to the initiation of autophagy [117]. There may also be an autophagic response to the photodamage of ER and/or mitochondria, in an attempt to recycle injured organelles, supporting the hypothesis that authopahy can serve as a protective mechanism [117, 147].</p><p>Dewaele et al. [148] attenuated macroautophagy using knockdown of the autophagy-associated protein Atg5 or chemical inhibition with 3-methyladenine. This resulted in reduced clearance of oxidatively damaged proteins and increased apoptosis in the Hypericin –PDT treated cells, thus revealing a cytoprotective role of macroautophagy in PDT. Paradoxically, genetic loss of macroautophagy improved clearance of oxidized proteins and reduced photokilling since up-regulation of chaperone-mediated autophagy (CMA) in Atg5(−/−) cells compensated for macroautophagy loss and increased cellular resistance to PDT.</p><p>Hypericin-PDT was also reported to induce a cytoprotective autophagic response in melanoma cells [149]. In addition, Atg7 knockdown of leukemia L1210 cells treated with CPO-PDT were more sensitive to the parental cells [109]. To sum up, stimulation of autophagy in apoptosis-competent cells increases cellular resistance to photokilling in PDT protocols. However, the same may not hold for cells incapable of mounting an apoptotic response. For example, knockdown of Atg7 increases resistance to Pc4-PDT in apoptosis-resistant MCF7 cells and Hypericin-PDT treatment of Bax−/−Bak−/− double knockout MEFs develop a much stronger autophagic response than their apoptosis-competent wild-type counterparts [109].</p><p>Carbonylation leads to exposure of hydrophobic patches within proteins, resulting in their partial unfolding, which favors their ubiquitination followed by recognition and degradation by proteasomes [150]. The ubiquitin-proteasome system (UPS) has been shown to play an important cytoprotective role through degradation of oxidatively modified proteins [150, 151]. Treatment with proteasome inhibitors is also associated with formation of intracellular protein aggregates, increased endoplasmic reticulum stress, and unfolded protein response induction in tumor models [152].</p><p>Szokalska et al. [153] observed that PII-PDT leads to carbonylation of cellular proteins and induction of unfolded protein response. Pretreatment of tumor cells with proteasome inhibitors sensitized EMT6, C-26 and HeLa cells to PDT-mediated cytotoxicity. Combination of these inhibitors and PDT led to potentiated antitumor effects, thus envisaging a possible role of the UPS in the resistance to PDT.</p><!><p>De novo ceramide can be associated with apoptotic sensitization after oxidative stress [154], and sphingomyelin synthases (SMS) have been shown to regulate cell growth and apoptosis. It has been demonstrated that de novo sphingolipids are associated with initiation of apoptosis after photodamage with Pc 4 –PDT [155]. The same group [156] have shown that overexpression of SMS1 is accompanied by attenuated ceramide response and apoptotic resistance after Pc 4-PDT and that RNA interference-dependent downregulation of SMS was associated with increased apoptosis after photodamage.</p><!><p>Among the many known signal transduction pathways, PI3K has been shown to promote cell survival and resistance to apoptosis [157]. Human prostate LNCaP cancer cells expressing a dominant-negative epithelial and endothelial derived tyrosine kinase (EtK) (substrate of PI3K) were resistant to Pc 4-PDT, suggesting that the PI3-kinase/Etk pathway is involved in the protection of prostate carcinoma cells from apoptosis in response to PDT.</p><!><p>Apoptosis is highly influenced by calcium as a mediator of signal transduction [158, 159]. In this regard, calcium/calmodulin-dependent kinases (CaM-Ks) rescue cancer cells from reactive oxygen intermediates by inducing the activation of antiapoptotic signaling pathways, such as Akt, ERK, and NF-kappaB in many different cell types. Rodriguez-Mora et al. [160] found that when MCF-7 cells were treated with PDT in the presence of a CaM-K inhibitor a greater level of cell killing was observed. In support of this finding, CaM-K inhibition increases hydrogen peroxide-induced apoptosis in MCF-7 cells through ERK phosphorylation.</p><!><p>PDT is well known to be antagonized by cellular antioxidant defence mechanisms, such as the glutathione system, superoxide dismutases (SOD), catalase or lipoamide dehydrogenase [161–164].</p><p>Human breast cancer MCF-7 cells transfected with the glutathione peroxidase gene were protected from PDT damage [165] due to removal of lipid hydroperoxides in living cells after 1O2 exposure. In addition, Dabrowski et al. [163] found that PDT toxicity induced by Hypericin was reduced in human kidney 293 cells over-expressing glutathione S-transferase P1–1. Overexpression of glutathione peroxidase-4 reversed nutrient-sensing protein kinase mTOR down-regulation and blocked macroautophagy progression and apoptosis.</p><p>Detoxification by glutathione conjugation has been correlated with drug resistance in cancer [167]. Some correlations on the involvement of glutathione system were found in cells resistant to PDT. Luna & Gomer [6] found a slight increase in the reduced glutathione levels in the RIF PII-PDT resistant cells, without any alterations in either glutathione peroxidase or superoxide dismutase levels. On the other hand, no differences in glutathione levels were found in the RIF-8A resistant variant characterized by Singh group [48]. In our ALA-PDT resistant clones [7] we found that the reduced glutathione content expressed on the basis of cell number increased two-fold. However, when expressed per μg protein no difference was observed among the cell lines. It is difficult to evaluate the impact of GSH due to the different protein content of the resistant lines. However the ratio of GSH: endogenous porphyrins, is higher in the resistant clones, and so is the ability to detoxify cytotoxic species per molecule of sensitizer.</p><p>In addition, the expression of the SOD2 isoform was shown to be regulated differently by ALA-PDT based on the cell origin. RNA SOD2 expression was up-regulated in tumor urothelial RT4 cells, not regulated in tumor colonic HT29, and slightly down-regulated in normal urothelial UROtsa cells [92] after ALA-PDT treatment.</p><!><p>Transcriptional and translational expression of heat shock proteins (HSPs) are associated with modulating cellular damage induced by various stresses including heat, oxidation, chemical exposure [168] and PDT [9]. HSP27, HSP34, HSP60, HSP70, HSP90, HSP110 [169, 170], glucose regulated proteins GRPs (GRP74, GRP78, and GRP100) [171] and HO-1 [172] have also been involved in defending PDT damage. Some of these proteins are presumed to be associated with rescue response of cells after PDT.</p><p>Upon PDT treatment, the RIF PII-PDT resistant cells [6] exhibited an increase in HSP70 and HO-1 mRNA, but these changes were not reflected in a higher protein synthesis. The same group has previously found that hyperthermia resistant cells overexpressing HSP70 were not cross resistant to PDT [173].</p><p>Verwanger and colleagues [91] photosensitized the human squamous cell carcinoma cells A-431 with ALA-PDT and employed the cDNA array technique to find increased expression of HSP70. They also found increased expression of HO-1 following dark incubation with ALA. The HO-1 expression did not increase further by irradiation. Hence the increased expression of HO-1 was probably caused by the need for heme degradation.</p><p>Hanlon et al. [174] found a higher expression of HSP60, which is a chaperone protein mainly found in mitochondria both in PDT resistant variants of colon cancer cells HT29 and in fibrosarcoma PDT resistant RIF-8A cells. In 2002, the same group [175] analyzed the expression of stress proteins in the same HT29 cells resistant to PII-PDT by means of microarray technology. They found an increase in HSP27 mRNA that is known to be part of the signaling pathway leading to apoptosis. Stable transfected cells with HSP27 complementary DNA showed an increased survival to PII-PDT suggesting that this protein plays a role in the resistance to PDT. Shen et al. [41] also found an increased expression of HSP27 mRNA in the HT29 human colon adenocarcinoma PDT-resistant variants.</p><p>Several studies have shown that various proteins involved in cellular stress response are induced after PDT treatment [35]; for instance, HSP1 was found to be phosphorylated and consequently activated after Pc 4-PDT of mouse lymphoma L5178Y cells [176].</p><p>Induction of HO-1 with hemin or stable transfection of colon adenocarcinoma C-26 cells with a plasmid vector encoding HO-1, increased resistance of tumor cells to PDT-mediated cytotoxicity. On the other hand, zinc (II) protoporphyrin IX, a HO-1 inhibitor, markedly augmented PDT-mediated cytotoxicity towards C-26 and human ovarian carcinoma MDAH2774 cells [177].</p><p>Cells pretreated with a calcium ionophore to increase overexpression of GRP, developed resistance to PDT within 16-h porphyrin exposure [171]. The study also indicated elevated levels of mRNA encoding, GRP-78, GRP-94 and an increase in GRP protein synthesis in RIF-1 cells exposed to 16-h porphyrin incubation prior to light exposure. However, a short (1h) porphyrin incubation prior to light treatment was associated with only minimal increases in GRP mRNA levels or GRP protein synthesis, indicating that specific targets of oxidative damage (modulated by porphyrin subcellular localization) are correlated with PDT-mediated GRP induction. In addition, a transient elevation of GRP mRNA levels in transplanted mouse mammary carcinomas following PDT was observed in vivo.</p><!><p>The extracellular matrix (ECM) is composed of collagens, elastin, proteoglycans and noncollagenous glycoproteins such as fibronectin and laminin. The ECM forms a complex, three-dimensional network among the cells of different tissues in an organ-specific manner. ECM is a dynamic structure that interacts with cells and generates signals through feedback loops to control the behavior of cells. Thus, ECM macromolecules are bioactive and modulate cellular events such as adhesion, migration, proliferation, differentiation, and survival [178].</p><p>Cell-to-ECM adhesion is regulated by specific cellular adhesion molecules known as integrins. Integrins are alpha-beta heterodimeric adhesion receptors that relay signals bidirectionally across the plasma membrane between the extracellular matrix, cell-surface ligands, cytoskeletal and signaling effectors [179]. The onset of drug-resistance to chemotherapy phenotypes is often associated with altered expression of adhesion and cytoskeletal components [180, 181].</p><p>Recently, it has become clear that cell-cell and cell-matrix interactions result in cytoskeletal reorganization and the activation of multiple signal transduction pathways that directly influence cell survival, growth and differentiation. Experimental evidence shows that anti-apoptotic pathways initiated by cell adhesion are operative in tumor cells and, furthermore, cause resistance to mechanistically distinct cytotoxics. The phenomenon has been called cell adhesion-mediated drug resistance (CAM-DR), and is based on the observation that cells that adhere to ECM components are protected from apoptosis induced by chemotherapeutic agents [182].</p><p>Cell adhesion to ECM proteins improves cell survival during radiation therapy. Integrin-mediated cell-matrix interactions impact favourably on normal and tumor cell survival after irradiation. Similarly to CAM-DR, this phenomenon is called cell adhesion-mediated radioresistance[183]. Cell size reduction and reduced adhesion to ECM proteins are found to be parameters associated with reversal of radioresistance induced by cells overexpressing integrin-linked kinases [184].</p><p>Inhibition of cell adhesion by PDT with BPD-MA was shown in 1997 by Margaron et al. [185]. In addition, it was demonstrated that downregulation of several adhesion molecules such as fibronectin could be the reason for the transient decrease in adhesion of human ovarian OVCAR 3 cancer cells to collagen IV, fibronectin, laminin and vitronectin after Verteporfin photosensitization [186].</p><p>The adhesive protein fibronectin and its integrin receptors play an important role in tumor development. Tumor cells are generally less adhesive than normal cells thereby contributing to tumor cell detachment and metastasis. Rudhorfer et al. [11] observed a dramatic downregulation of the fibronectin gene after ALA-PDT of the squamous cell carcinoma line A-431. This downregulation may simply characterize the beginning of the rounding up and the detachment process of cells after heavy damage. As a side effect, migration in vitro and metastasis in vivo, respectively, could be facilitated. After PDT of colon carcinoma cell lines with external porphyrin-based PS, a transient decrease in adhesiveness and in adhesion molecules expression was found [187]. According to the authors the decrease in adhesiveness could account for the decreased metastatic potential of PDT-treated cancer cells. However, either increased [188] or decreased [189] impact on metastasic ability of the PDT-surviving cells is likely to occur, and these differences may be ascribed to the different PS, light doses, cell model and even location of the tumor.</p><p>It has been described an effect of PDT on either decreasing or increasing adhesion to plastic, ECM and to endothelial cells [186, 190, 191]. PDT using BPD-MA inhibited cell adhesion, with no significant differences between matrices and without altering integrin expression [185]. In addition, PDT reduces invasiveness of smooth muscle cells and reduces fibroblast migration, generating a matrix barrier to invasive vascular cell migration, inhibiting experimental intimal hyperplasia [192]. Although the effects of PDT on the ECM are not well understood, it is clear that PDT induces changes in ECM.</p><p>One of the cellular PDT targets is cytoskeleton [193]. Three major eukaryotic cytoskeletal proteins are actin, tubulin and intermediate filaments. Any disturbances in these systems have been related to tumor progression and metastasis [191]. Changes in the cell shape (cell attachment, cytoskeleton) in the course of apoptosis execution promote the formation of apoptotic bodies. The clearance of the bodies is done mainly by cells of the immune system. A special case of apoptosis is 'anoikis, i.e., apoptosis induced by cell detachment of anchorage-dependent cells [195].</p><p>Extracellular signals, cell-detachment and cell shaping processes receive or transmit their information via intracellular signaling pathways such as p53MAPK, ERK1/2 or JNK. As cited above, Sanovic et al. [90] found promotion of p38MAPK, ERK, JNK and Ras signaling pathways supporting survival and/or apoptosis after Hypericin-PDT.</p><p>Sanovic et al. [90] also found upregulation of NEDD9 (neuronal precursor cell-expressed, developmentally down-regulated 9 (human enhancer of filamentation 1 (HEF1)) belonging to the CAS protein family after PDT. This protein localizes at focal adhesion sites, and its early upregulation could participate in apoptosis induction and execution. The NEDD9 overexpression is likely to activate JNK kinases, induce apoptosis and accelerate transition of 'flat' attached cells to rounded mitotic cells.</p><p>Integrins mediate cell adhesion and engage in crosstalk with different growth factor receptors. Phosphorylation of these receptors may occur following the binding of a growth ligand to the receptor and also occur by binding to integrins, without ligand binding. Genes encoding integrin β1, integrin 3 and integrin 6 were downregulated after Hypericin-PDT [90]. As a consequence, reduced signal transduction from ECM and impaired cell adhesion in the early phase of damage processing has appeared, all required for cell cycle stop and apoptosis. Similarly, downregulation of integrin 2 and β3 precursors after Hypericin-PDT were found in a bladder cancer cell model by Buytaert et al. [99]. Downregulation of thrombospondin-1, which is a ligand for integrin β1, was also observed in Hypericin-PDT treated A-431 cells [90]. The downregulation of the thrombospondin-1 precursor was also found after Hypericin-PDT by Buytaert et al. [99]. It has been demonstrated that β1-integrins play a role in cell detachment and apoptosis induction triggered by loss of E-cadherin following PDT with ZnPc [196], thus reinforcing the hypothesis of a role of integrins downregulation on the promotion of cell detachment and apoptosis.</p><p>The upregulation of the Rho family GTPase 3 (RhoE) by Hypericin-PDT is also likely to contribute to the process of cell detachment and the control of rearrangements of the actin cytoskeleton, since it inhibits integrin based focal adhesions and formation of actin stress fibres leading to cell rounding. [90]. Another detachment mechanism affected by PDT is the overexpression of Pleckstrin homology-like domain, family A, member 1 (PHLDA1), which also contributes to significant changes in cell morphology and decreased cell adhesion [90].</p><p>We have found that our ALA-PDT resistant cells were less invasive and migrant in vitro. The cells were also less metastatic in vivo compared to the parental LM3 adenocarcinoma cells. In addition, anchorage-dependent adhesion was also impaired in vivo in the resistant clones, evidenced by the lower tumor uptake, latency time and growth rate. However, both of the clones showed higher in vitro binding to the ECM protein collagen I, without overexpression of β1 integrin, which is the main molecule involved in collagen I binding [197]. In addition, the resistant clones exhibited also loss of actin stress fibers, as well as disorganized the actin cortical rim. E-cadherin, β-catenin (cell–cell adhesion proteins) and vinculin (cytoskeleton-associated protein) distribution was also disorganized, without differential expression in Western blot assays [197]. The reorganization of these cytoskeletal and adhesion proteins can probably be correlated with the lower metastatic phenotype.</p><p>Vimentin is a major cytoskeletal protein degraded in response to various inducers of apoptosis [198, 199]. In cells transfected with a caspase-resistant vimentin, apoptosis driven by PDT was partly suppressed and delayed, suggestingthat vimentin confers resistance to PDT by impairing caspase-3 translocation [200].</p><p>Plating efficiency of some of the PII-PDT resistant variants was reduced to 36–43% [6]. When the PII-PDT resistant variant cells were injected in syngenic mice, the number of cells required to produce tumors was 1,000 to 10,000 times higher than nonresistant cells. From our view point, this feature may also be related to changes in the extracellular matrix of the PDT resistant clones. Similarly, we have found an impairment of plating efficiency in our variants resistant to ALA-PDT [189], which was correlated with lower tumor take when injected to mice. In addition, when various cell lines with different histologies were exposed to PII-PDT, a general association was noted between PDT sensitivity and the plating efficiency, but no association was observed on PS uptake [201].</p><p>As addressed above, the fact that cellular shape is an important factor in the regulation of cell sensitivity to mitogens, it becomes evident that the proliferative rate is anchorage dependent [202]. The cellular shape is dictated by the extracellular material upon which the cells rest (in vivo condition) and by the substrate upon which the cells are maintained (in vitro condition). The substrate itself may, in turn, induce the cells to manufacture their extracellular material and specific cell surface proteins which control the cellular shape [203].</p><p>In our laboratory [7], we tested the hypothesis that cells would loose resistance upon ALA-PDT of cells in suspension, based on the observation that resistant clones spread more than the parental cell line. In this study we also observed that the resistance indices of the cells did not change. ALA-PDT was also performed in cells attached to fibronectin, but no differences in the resistant indices of the clones were observed.</p><p>Cell size has been suggested to be somewhat related to resistance to chemotherapy [204]. Similarly, there are some evidences that cell size can be related to PDT resistance as well. PII-PDT resistant variants from RIF fibrosarcoma cells [6] were larger and had an increased protein content; similar to the resistant variants isolated by Sharkley et al. [44]. An increased cell spreading together with an increased number of cells per colony was also observed. In addition, when several human leukemia cell lines and normal lymphocytes were tested, it was found that the resistance to BPD-MA-PDT was related to the cell sizes, with the smallest cells being the most vulnerable [205]. We also found [7] a 2-fold increase in the volume and protein content of ALA-PDT resistant variants as compared to the parental line. Similarly, plasma membrane is the main target for PDT damage [206] and since larger cells have a greater surface area, the treatment could be less effective in the resistant clones.</p><!><p>Cyclooxygenases (COX) catalyze the conversion of arachidonic acid to prostaglandin (PG) H2, the immediate substrate for a number of cell specific prostaglandin and tromboxane synthases. The production and release of Prostaglandin PGE2 and other prostanoids contribute to the development of important immunomodulatory responses. Two isoforms, COX-1 and COX-2, have been identified and their expression is regulated differently. COX-1 is constitutively expressed in most cell types and may be responsible for housekeeping functions. By contrast, the expression of COX-2, which is regulated both at the transcriptional and post-transcriptional level, is barely detectable in normal tissues but is rapidly induced in response to tumor promoters, oncogenes, cytokines, and mitogens [207–209]. A growing body of evidence suggests an association of COX-2 with tumor development, aggressive biological tumor behavior, resistance to standard cancer treatment, and adverse patient outcome [210].</p><p>Controversial results have been observed in cancer cells with modulated metabolism of arachidonic acid prior to PDT. It was found that PDT employing porphyrin- and chlorin-based photosensitizers induces the expression of COX-2 with subsequent release of PGE2 [211, 212]. Conversely, the combination of PDT with the selective NS-398 COX-2 inhibitor, resulted in enhanced photodamage in RIF-1 fibrosarcomas [211]. Similarly, Henderson and Donovan [213] and Penning et al.. [214] demonstrated that non specific COX-2 inhibitor indomethacin increased the sensitivity of tumor cells to PDT. At the same time NS-398 caused upregulation of COX-2 and induced apoptosis resistance in HeLa cells on hypericin- PDT [212, 215]. On the other hand, Kleban et al., [216] found that the arachidonic acid inhibitors inhibitors with known COX-independent action potentiated Hypericin-PDT, and the inhibitors of COX attenuated PDT. PII-PDT also induced the expression of COX-2 gene in C-26 cells [217]. The study also showed that the administration of a selective COX-2 inhibitor potentiated antitumor effects after PDT, but not during or before PDT.</p><p>Some recent reports have implicated p38 MAPK in the upregulation of the inducible COX-2. Overexpression of WT-p38 MAPK increased cellular resistance to PDT-induced apoptosis by blocking COX-2 up-regulation [212]. Phospholipase A2 inhibition caused an increase in the levels of free arachidonic acid, protected bladder cancer cells from Hypericin-PDT mediated apoptosis and attenuated the activation of p38 MAPK [215].</p><p>On the other hand, the inhibitor of endogenous PG synthesis indomethacin, increased resistance of glioma cells to PDT. The endothelial cells did not show an increase in resistance. In contrast to the studies performed using radiotherapy, exogenous PGs decreased the surviving fraction of human endothelial cells and glioma C6 cells treated by PII-PDT [218].</p><p>Kleban et al. [216] modulated arachidonic acid metabolism prior to Hypericin-PDT. They found the inhibition of lipooxigenase (LOX) activity upon PDT. The combination of low-dose Hypericin-PDT and 5, 12-LOX and 12-LOX inhibitors intensively strikes cell survival and proliferation.</p><!><p>Nitric oxide (NO) is a gaseous radical that can play either a cytotoxic or a cytoprotective role depending on the cell type and the experimental paradigm selected in the pathology.</p><p>Several pathways were found to be involved in chemoresistance mediated by NO. In malignant astrocytes, NO has been found to modulate radioresistance and chemoresistance against nitrosourea derivatives [219]. In neuroblastoma cells, NO inhibition of the transcription factor and proto-oncogene NMYC activity and expression of a large set of ATP binding cassette transporters influence the chemoresistance phenotype [220]. Inducible NO synthase also confers chemoresistance in head and neck cancer by modulating survivin [221]. In cisplatin-resistant ovarian cancer cells, blocking all NO synthases dramatically reverses the resistant phenotype through induction of apoptosis [222].</p><p>NO itself is not an effective oxidant, but can be converted to strong damaging oxidants under biological conditions. However, NO on its own may act as an antioxidant at low concentrations in lipid membranes by scavenging chain propagating oxyl and peroxyl radicals [223]. Radical interception by NO could contribute to overall cellular resistance to peroxidative stress. If this occurs during PDT, it might compromise treatment efficacy. Niziolek et al. [224] showed that photokilling, could be strongly suppressed by low, nontoxic levels of exogenous NO.</p><p>NO can elicit long-term cytoprotective antioxidant responses. The effects in this case are indirect; i.e., responsible NO is no longer on the scene when the oxidative challenge is presented. For example, endogenous NO produced via cytokine induction of nitric oxide synthase elicited similar long-term hyperresistance to H2O2 or high-level NO cytoxicity in hepatocytes [225, 226].</p><p>Similarly, hyperresistance to ALA-PDT was detected approximately 8 h post SPNO (exogenous NO donor), and maximized approximately after 20 h [227]. And in addition to its immediate radical-quenching effects, NO can evoke a delayed cytoprotective response in PpIX-sensitized COH-BR1 cells, since a concomitant increase in HO-1 levels and ferritin was observed. This observation suggested that a cytoprotective mechanism with mobilization of "signaling" iron was involved. The same group reported in 2010 [227] that NO has remarkable ability to support apoptosis. COH-BR1 tumor cells in glucose-containing medium died after ALA-PDT mainly due to necrosis with a low level of apoptosis. SPNO inhibited necrosis when introduced before PDT treatment at a nontoxic concentration but supported apoptosis such that the latter became predominant in the remaining cell death. Accompanying this was a large increase in caspase-3/7 activation. SPNO-supported apoptosis was more pronounced when glucose-deprived cells were compared with glucose-replenished, SPNO-treated counterparts. SPNO plus glucose also suppressed plasma membrane-damaging lipid peroxidation and loss of cellular ATP under photostress. The NO effect on PDT resistance is attributed to membrane protection with maintenance of sufficient glycolytic ATP to sustain apoptosis. They have also extrapolated the results of NO protection to PDT with other PS such as MC540 [228], a lipophilic dye that localizes primarily in the plasma membrane. Photodamage of MC540-sensitized mouse lymphocytic leukemia L1210 cells was inhibited when SPNO was introduced either immediately before or after light exposure. The mechanism of protection is related to interceptation of propagating radicals such as 5alpha-OOH, definitive singlet oxygen adduct of plasma membrane cholesterol.</p><p>In our laboratory [229] we found that the NO-resistant variant of murine breast adenocarcinoma LM3-SNP obtained after successive exposures to the NO donor sodium nitroprusside had no cross-resistance to ALA-PDT treatment. We have also induced resistance to ALA-PDT in LM3-SNP cells after multiple cycles of photodynamic treatment, showing that resistance to NO did not interfere in the development of PDT resistance. In addition, we found that various cell lines with different NO production levels were equally responsive to ALA-PDT [230]. Furthermore, the modulation of NO levels did not modify the intrinsic response of various cells lines to PDT treatment.</p><p>Bhowmick [231] et al. reported evidence for increased tumor cell resistance due to inducible NO synthase (iNOS) upregulation in a PDT model. After ALA-PDT treatment of breast tumor COH-BR1 cells, iNOS was upregulated, while other NOS isoforms were unaffected. Exposing cells to the NOS inhibitor L-NAME during photochallenge enhanced caspase-3/7 activation and apoptotic killing, suggesting that iNOS was cytoprotective. Consistently, a NO scavenger enhanced ALA-PDT-induced caspase-3/7 activation and apoptotic death.</p><!><p>Discovered 10 years ago, survivin has a dual role in the smooth progress of mitosis and in apoptosis resistance. Survivin plays an important physiological role in development, but it is absent in differentiated adult tissues. In contrast, aberrant survivin expression is found in most human cancers because of the activation of various signaling pathways. A complex survivin network appears to intersect multiple pathways in cell biology, related to several molecular partners and fine subcellular localizations. Based on its pro-oncogenic properties, basic and translational studies have shown a growing interest in survivin that has led to consider it as a prognostic marker and a promising target for anti-tumoral therapies. Initially, survivin was described as an inhibitor of caspase-9. However, over the last years, research studies have shown that the role of survivin in cancer pathogenesis is not limited to apoptosis inhibition but it also involves the regulation of the mitotic spindle checkpoint. Survivin also promotes angiogenesis and chemoresistance [232]. In various tumors, high survivin levels are correlated with poor prognosis, decreased apoptosis, increased angiogenesis, and chemoresistance in cancer cells [233, 234].</p><p>Survivin binds to HSP90 in cells and is therefore considered a HSP90 client protein [235]. HSP90 provides the necessary intracellular chaperone environment for proper folding and maturation of a variety of client proteins, many of which are involved in signal transduction and cell proliferation. A derivative of the antibiotic Geldanamycin, 17-allylamino-17-demethoxygeldanamycin (17-AAG), interferes with proper binding of client proteins, such as survivin and leads to misfolding of client proteins, ubiquination, and proteasome degradation.</p><p>Ferrario et al. [236] found in the human melanoma cell line YUSAC2/T34A-C4 that PII and chlorin-based PDT induced increased expression and phosphorylation of survivin together with increased PDT-mediated apoptosis and cytotoxicity. PDT treatment of melanoma cells expressing an inducible dominant-negative survivin gene, resulted in increased cleavage of the caspase substrate.</p><p>In addition, human BT-474 breast cancer cells treated with the combination of PDT and 17-AAG exhibited decreased expression of phosphorylated survivin, phosphorylated Akt, and Bcl-2. The decreased expression of these client proteins was accompanied by higher apoptotic indices and increased cytotoxicity, showing for the first time that targeting survivin and possibly other HSP90 client proteins improves PDT responsiveness in vitro [236].</p><p>ALA-PDT induced apoptosis and G0/G1 phase arrest in cervical cancer Me180 cells. ALA-PDT also suppressed the mRNA and protein expression of survivin in Me180 cells [237]. Immunohistochemistry of ALA-PDT treated Me180 tumors showed remarkable down-regulation of protein expression and mRNA of survivin [238].</p><p>Ferrario, [239] examined the effects of a combined modality protocol involving PDT and 17-AAG in mouse mammary carcinoma cells and tumors. PDT increased the expression of the anti-apoptotic and pro-angiogenic proteins survivin, Akt, HIF1-alpha, MMP-2 and VEGF in tumor tissue and 17-AAG significantly decreased the protein expression. Tumor bearing mice treated with PDT and 17-AAG had improved long-term tumoricidal responses when compared with individual treatment protocols. In conclusion, survivin has recently shown to be involved indirectly in resistance to PDT, and at the same time, as a target for PDT. However, to the best of our knowledge, upregulation of survivin was not found in PDT resistant cells.</p><!><p>Tumor hypoxia is a therapeutic concern since it can reduce the effectiveness of radiotherapy, some O2-dependent cytotoxic agents, and photodynamic therapy [240]. Tumor hypoxia can also negatively impact therapeutic outcome by inducing changes in the proteome and genome of neoplastic cells. Tumor hypoxia enhances survival and malignant progression by enabling the cells to overcome nutrient deprivation or to escape their hostile environment. The selection and clonal expansion of these favorably altered cells further aggravate tumor hypoxia and support a vicious circle of increasing hypoxia and malignant progression while concurrently promoting the development of a more treatment-resistant disease.</p><p>PDT-induced tissue hypoxia as a result of vascular damage and photochemical oxygen consumption may limit the efficacy of this treatment. A mechanism that protect tumor cells against PDT-mediated damage is stabilization of hypoxia-inducible factor1 (HIF1)-alpha [131,241]. It was reported that PDT induces hypoxia and expression of the vascular endothelial growth factor (VEGF) via the HIF1-alpha pathway, with subsequent promotion of tumor angiogenesis, thus enhancing tumor proliferation and survival. By DNA microarray analysis Okunaka et al. [242] demonstrated that VEGF mRNA expression was induced in the lung cancer cell line SBC-3 after ATX-s10-PDT.</p><p>Human esophageal normal Het-1A cell line induced with high-expression of HIF-1alpha by cobalt chloride-mediated chemical, clearly showed resistance to ALA-PDT. Moreover, transfection of the cells with anti-HIF1-alpha short interfering RNA (siRNA) knocked down the HIF-1alpha expression and restored the photosensitivity of the cells to ALA-PDT. However, HIF-1alpha expression was not induced by cobalt chloride in tumour esophageal KYSE-70 and KYSE-450 cell lines, and hence no difference in cell survival was found after ALA-PDT [179].</p><p>Some strategies have been developed in order to overcome PDT resistance due to hypoxia. It has been hypothesised [243] that by controlled temporary endo or peri-vascular occlusion of the collateral arterial branches upstream of the tumor, it is possible to redirect blood flow through the principal artery of the downstream tumor. The concept called "arterial flow focalization" increase oxygen supply, thus decreasing hypoxia-driven resistance to PDT. In addition, hypoxic cells can be also a preferential target of bioreductive drugs and hypoxia-directed gene therapy [244, 245].</p><!><p>Several investigators have looked into PDT susceptibility of cells resistant to various anticancer treatments (Table 4). Particularly, cells with either induced or transfected MDR have been tested for cross resistance with PDT, however these results have been conflicting. Some cells resistant to chemotherapy were found to be slightly more susceptible to PDT [52, 246, 247]. In general, about one third of the reports showed cross resistance to PDT in chemotherapy, radiotherapy and hyperthermia resistant cells, whereas the rest of the models showed no resistance.</p><p>In addition, some researchers have challenged if cell lines resistant to PDT are also resistant to chemotherapy, radiotherapy or hyperthermia (Table 5). Again, some cell lines were found to be cross resistant and others were not. One third of reports showed cross resistance and two thirds showed no cross resistance to chemotherapy etc in PDT resistant cells. In our laboratory, we found that ALA-PDT resistant clones were not resistant either to UV and hyperthermia treatment or to chemotherapy with DXR, cisplatin, methotrexate, 5-Fluorouracil and Mitomycin C treatment [7].</p><p>In the particular case of cross resistance to cisplatin, a 2-fold decrease in the number of platinum-DNA adducts were found, when the PII-PDT resistant cells RIF-8A were exposed to the drug [48]. Since several mitochondrial alterations have been described for this cell line with subsequent resistance to antineoplastic treatment, the authors hypothesize that an increased repair activity could result in an increased energy demand, and consequently a higher mitochondrial activity.</p><p>Cisplatin resistance–associated overexpressed protein (LUC7A) was cloned by Nishii et al. [248] from cisplatin-resistant cell lines. LUC7A mRNA was down-regulated in the colon cancer cell line HT29 after ALA-PDT and not regulated in two bladder cancer cell lines [92]. The fact that cisplatin resistance was not induced after PDT in those cell lines may provide a basis for combinatory therapy regimens. Lottner et al. [249] combined the cytostatic activity of cisplatin/oxaliplatin and the photodynamic effect of hematoporphyrin in the same molecule employing hematoporphyrin-platinum(II) conjugates. They found synergistic antiproliferative effects in vitro against J82 bladder cancer cells and UROtsa using four different hematoporphyrin-platinum(II) conjugates (Fig 9).</p><p>A PDT-mediated DXR transient resistance has been described for Chinese hamster fibroblasts after a single treatment of PII-PDT under short or long exposure times to the PS [250]. ATP depletion and cell cycle changes were positively correlated with decrease in drug sensitivity. However, induction of glucose-regulated stress proteins, antioxidant enzymes activities and intracellular drug levels were not responsible for the drug resistance.</p><p>Guo et al. [251] found that breast cancer MCF-7 cells resistant to paclitaxel and doxorubicin treatment were efficiently treated with the photoactivable drug calphostin C, through a mechanism that involves the induction of cytoplasmic vacuolization without activation of typical apoptotic pathways. Calphostin C, is not a classical PS, but it is a photoactivable inhibitor of phorbol-responsive protein kinase C isoforms [252].</p><p>Cross resistance of PDT resistant cells to UV light has been shown by DiProspero et al. [81] and Zacal et al. [125]. There are differences in sensitivity according to the UV illumination wavelength. For example, PDT-resistant HT29 human colon adenocarcinoma cells were cross-resistant to long-wavelength UVA (320–400 nm) but not to short-wavelength UVC (200–280 nm). The authors found that increased expression of Hsp27 and BNip3 and decreased expression of mutant p53 correlated with increased resistance to UVA. In contrast, increased expression of Hsp27 and BNip3 correlated with increased sensitivity to UVC, whereas increased expression of mutant p53 showed no significant correlation with sensitivity to UVC [253].</p><p>The study of cross resistance in cells with developed resistance against a particular PS challenged against other PS of similar or different characteristics (Table 6) is highly complex. Often cross-resistance is related to the structure of the PS.</p><p>The PII-PDT resistant RIF-8A cells were cross resistant to incubation with PpIX but not to ALA-induced PpIX, implying that the differences in mitochondrial localization and/or binding depending on the source of PpIX may be crucial in the outcome of PDT. Upon incubation of RIF-8A cells with ALA, PpIX fluorescence was higher than that obtained in the parental RIF-1 cells. Subcellular localization of PpIX was quite similar in both the strains. In addition a good correlation with Rhodamine123 fluorescence was observed in both lines. On the contrary, when exogenous PpIX was added, a strong correlation was seen in parental RIF-1 cells with Rh-123 fluorescence, but a weak correlation was found in RIF-8A cells between exogenous PpIX and mitochondrial sites [26].</p><p>We found that our ALA-PDT resistant clones from mammary carcinoma cells exhibited a slight resistance to exogenous protoporphyrin IX treatment but no cross resistance to BPD-MA and MC540 photosensitization. However intracellular accumulation of the three PS per protein was equal in both parental and resistant clones, showing that PS content is not crucial for photodynamic resistance in all the cases [7]. ALA ester derivatives hexyl-ALA and undecanoyl-ALA did show cross resistance with ALA-PDT (Fig 10), although both lypophyllic ALA derivatives do not enter the cell through the same transporter [254] and are probably not effluxed by the same mechanism [255]. This reinforces the hypothesis that cross resistance depends on PpIX formed from ALA or ALA derivatives and not ALA itself.</p><p>Mayhew et al. [4] found that the strains resistant to PDT with the anionic compound PHP exhibited cross resistance to other anionic PS such as i) exogenous PpIX, ii) Zn (II) tetrasulphonated phthalocyanine (TSPC) and iii) Zn (II) tetraglycine-substituted phthalocyanine (TGly) (Fig 11). However, the PHP-PDT resistant cells were not resistant to PDT with the cationic PS Zn-PCP and m-THPC, the neutral PS mTPyP (Fig 12) and PpIX from ALA.</p><p>However, the resistant variants to PDT with the cationic PS Zn-PCP did not exhibit cross resistance to any of the PS employed above mentioned PS, either the anionic (Fig 11) or the neutral and cationic (Fig 12). The conclusion of these studies was that there are at least two distinct mechanisms of PDT-resistance in these RIF-1 cells, and that the PHP-resistance is likely to depend, to some extent, upon the physical nature of the PS. Zn-PCP resistant RIF-1 cells are, on the contrary, not cross resistant to any other PS, and this is likely to be due to the alteration of a single cellular target, which is not shared with any other PS.</p><p>Luna & Gomer [6] found that the 16 h-PII PDT-resistant variants exhibited cross-resistance to a 1 h-PII PDT incubation protocol. The short incubation of PS is generally associated with plasma membrane damage, whereas extended incubation is usually believed to result in damage to specific cellular organelles [6,256]. However, the 1 h PII PDT-resistant variants did not exhibit resistance to the extended 16 h PII incubation PDT protocol. Mayhew et al. [4] found in their resistant variants, cross-resistance to short exposure to PHP (a compound equivalent to PII) and long exposure protocols. Such varying results employing similar PS suggest that the mechanisms of resistance are multiple and different for each induction protocol.</p><!><p>Nearly any type of chemoresistances is a multifactorial process involving induction of drug-detoxifying mechanism, quantitative and qualitative modification of drug targets, arrest of cell cycle, regulation of DNA replication or reparation mechanisms, and modulation of apoptosis. These modifications are acquired in response to a selection pressure by the drug treatment (acquired resistance) or expressed by cells already resistant and that will never respond to the drug treatment (intrinsic resistance). The specific mechanisms for chemoresistance have been extensively reviewed elsewhere [257, 258, 259].</p><p>Table 7 compares the mechanisms of resistance to PDT and chemotherapy. Although there are specific causes of PDT resistance, most of the features of PDT resistance have already been described for chemoresistance. The fact that in many cases no cross-resistance has been reported between both the treatments is in line with the enhancement of PDT efficacy by combination with chemotherapy [260–266]. On the other hand, in many cases, the same common features can be induced by different but overlapping pathways which can lead to cross resistance.</p><!><p>As in chemo- and radiation resistance, the mechanisms of resistance associated with PDT are complex and are reviewed in this article. The salient aspects of PDT-induced resistance mechanisms discussed here are: i) induction of resistance after multiple PDT treatments, ii) modulation of protein expression leading to resistance, iii) induction of specific genes involved in the mechanisms of resistance, and iv) studies of alterations in gene expression after PDT treatment.</p><p>An understanding of these resistant mechanisms could potentially help design new and robust treatments strategies such as combination of PDT with chemo or other therapies. Although chemoresistance is well established, in many cases, no cross-resistance between PDT and chemotherapy has been reported. For example, protective mechanisms such as damage to DNA repair help the chemotherapy escape process, but this resistance mechanism is very limited in PDT treated cells. In addition, the new cDNA array techniques provide the tools to further study the role of multiple survival pathways, demonstrating that PDT resistance, similar to chemoresistance, is a multifactorial phenomenon. The lack of cross resistance between PDT treatments with different PS in many studies confirms the complexity of the resistance processes and the specificity of the cell death pathways with each PS. As PDT evolves into a first line therapy, it is crucial to understand these resistance mechanisms and develop efficient treatment strategies to overcome these.</p>

PubMed Author Manuscript

Predicting protein folding cores by empirical potential functions

Theoretical and in vitro experiments suggest that protein folding cores form early in the process of folding, and that proteins may have evolved to optimize both folding speed and native-state stability. In our previous work, we developed a set of empirical potential functions and used them to analyze interaction energies among secondary-structure elements in two \xce\xb2-sandwich proteins. Our work on this group of proteins demonstrated that the predicted folding core also harbors residues that form native-like interactions early in the folding reaction. In the current work, we have tested our empirical potential functions on structurally-different proteins for which the folding cores have been revealed by protein hydrogen-deuterium exchange experiments. Using a set of 29 unrelated proteins, which have been extensively studied in the literature, we demonstrate that the average prediction result from our method is significantly better than predictions based on other computational methods. Our study is an important step towards the ultimate goal of understanding the correlation between folding cores and native structures.

predicting_protein_folding_cores_by_empirical_potential_functions

Introduction<!>Choice of experimental data and protein folding core prediction targets<!>Prediction of folding cores based on an empirical potential function<!>Evaluation of overlap between predictions and experiments<!>Results and discussion

<p>Understanding the mechanisms by which proteins fold is one of the grand challenges of molecular biology. Theoretical studies suggest a funnel-like free energy landscape for protein folding, which helps to explain how an extended polypeptide chain consistently folds into its stable native three-dimensional conformation in a speedy fashion [1-4].</p><p>Theoretical and in vitro experiments suggest that protein folding nuclei, or cores, form early in the folding process [5-13]. This finding, in turn, supports Hammond's postulate [14] that thermodynamics and kinetics are closely correlated in proteins and that proteins may have evolved to optimize both folding rate and native-state stability [15]. Our earlier combined experimental-theoretical study on Pseudomonas aeruginosa apo-azurin and another β-sandwich protein demonstrated this correlation, in which the stable folding cores predicted by our energetic method also harbored the key residues involved in the folding-transition [5].</p><p>Among the experimental methods to probe the protein folding process, protein hydrogen-deuterium exchange (HX)2 helps identify protein regions that are shielded from solvent and thus "protected" from deuterium exchange (i.e., resulting in a slower rate of exchange). Based on HX experiments, the hydrogen-bonded amide protons (NHs) that are most protected from deuterium exchange in the protein native-state are often found in the same protein regions as the NHs protected earliest during the protein folding reaction, as well as those NHs that are most protected in partially-folded intermediate states of the protein [13,16,17]. In contrast, NHs in turns and loops are rarely among the very slowest protons to exchange. Therefore, HX is useful in identifying the slow-exchanging NHs that make up the protein folding core.</p><p>Several computational models have been developed that try to connect folding theory with experimental data on protein unfolding/folding kinetics. Examples are graph-theoretical approaches based on effective contact order [18,19], several variants of a motion planning method [20-23], molecular dynamics simulations of unfolding fluctuations around the native-state [24,25], an unfolding approach using a secondary-structure contact network and minimum cuts [26], a simplified lattice-protein model of native-state HX [27], and a method that exploits a correlation between slowest exchanging cores and low conformational entropy [28]. The two most relevant examples of computational models, with respect to this study, are the Floppy Inclusions and Rigid Substructure Topography (FIRST) method [29] and the Gaussian Network Model (GNM) [15,30]. In the FIRST method, inter-atomic covalent and hydrogen bonds and hydrophobic interactions are replaced by rigid bars whose lengths and bond angles are constrained—only bond rotations are allowed. FIRST then identifies the rigid and flexible parts of the all-atomic protein model by selectively breaking hydrogen bonds in order of weakest to strongest. The GNM method coarse-grains a protein into an elastic network of residues, whereby pairs of residues within a cut-off distance are connected by virtual elastic springs, and it predicts the stable folding cores by studying the collective motions of the elastic network. In GNM, slow mode minima imply hinge sites, whereas high frequency mode peaks indicate stable "kinetically hot" residues.</p><p>Despite some success with these computational methods, there remains room for improvement. Empirical potential functions have been used previously to study changes in protein stability [31-33]. In our former work [5], we developed an empirically-weighted set of statistical potential functions and used them to analyze interaction energies among secondary-structure elements in two β-sandwich proteins. In the current study, we test the power of our empirical potential functions by applying them to the prediction of protein folding cores as revealed by HX experiments, using a large set of proteins with different structures.</p><p>Here, and in earlier studies [13,15], the experimental folding cores are defined as those that make up the folding core elements, which are the secondary-structure elements (SSEs) containing the slowest exchanging residues (those with the greatest protection factors) identified in HX experiments. Using a set of 29 unrelated proteins that were extensively studied in the literature, we show that, on average, our predictions correlate better with the experimentally-identified folding cores than those of two GNM methods and a third method using the FIRST software. We believe that our prediction method may be useful to facilitate a better understanding of the factors that dictate protein folding and native-state stability.</p><!><p>HX experiments are typically subdivided into three types based on their detection purposes and experimental settings [13]: slow exchange core experiments (for NHs most protected in the native-state), pulsed exchange experiments (for NHs first protected during folding), and folding competition experiments (for NHs most protected in partially-folded species). The folding core secondary-structure elements (SSEs) revealed by these three methods are often identical or very similar. Thus, we follow Rader and Bahar [15] in using experimental data from slow exchange core experiments, the most abundant experimental folding core data in the literature, as our prediction targets. In addition, the secondary-structure definitions are based on the Protein Data Bank SHEET and HELIX records.</p><p>To train our empirical potential function and then compare our computational predictions with experimental results, we used a set of 29 proteins (listed in Table 1) that were extensively studied in the literature [13,15,34-40].</p><!><p>The computational prediction method using our all-atom empirical potential function is described in detail in our previous work [5]. The stability cores are ranked by the interaction energy between multiple SSEs (groups of two, three or four) using a scoring function:(1)Score=3.45E‒Packing+5.0E‒AS+1.9E‒HB.Here, the three terms in the scoring function represent the effects of side-chain packing (ĒPacking), solvent accessible surface area (ĒAS), and hydrogen bonding interactions (ĒHB), respectively. The parameters for these three terms are statistically derived from a non-redundant structure database of 2701 non-homologous soluble proteins [41], and the weight for each term is chosen by fitting to the folding core results of two proteins with the most consistent HX data [13], listed in Table 1b. These two proteins, staphylococcal nuclease [42,43] and ribonuclease H [44,45], both have α-helix and β-sheet SSEs, and they are excluded from the set of 27 proteins used for cross-validation.</p><p>For comparison with the experimental HX results by Li and Woodward [13], we define the folding core as the group of SSEs with the lowest interaction energy. The interaction energies are calculated for groups of two, three and four SSEs, and each grouping type is considered a separate but related method for predicting the folding core.</p><!><p>To compare our approach to previous methods and experimental results, we adopted the method for evaluating overlap employed by Rader and Bahar [15]. There are two related measures for the overlap between methods A and B (A and B may be experimental or computational prediction methods):(2)s(A,B)=o(A,B)NA⋅NBN,(3)z(A,B)=o(A,B)−NA⋅NBN.Here, N is the total number of residues in the target protein, NA and NB are the numbers of folding core residues revealed by methods A and B, respectively, and o(A, B) is the overlap in the number of residues revealed by methods A and B. These two quantities s(A, B) and z(A, B) measure the extent of difference between the observed overlap, o(A, B), and the expected overlap for random matches, NA · NB/N. Thus, s = 1 and z = 0 correspond to random matches and larger values of s and z indicate greater correlation between methods A and B.</p><!><p>Fig. 1 illustrates the folding cores predicted by HX experiments and the empirical potential function for a few examples within the 27-protein test set. Folding core elements are mapped as dark ribbons on the light gray 3D cartoon backbone of the protein structure. Each column represents one of the four methods (HX experiments; two-, three- and four-SSE interaction groups). Fig. 2 summarizes the comparisons of the four methods for all 27 test proteins using the reduced representation from Rader and Bahar [15]. The x-axis corresponds to the residue index, and the stacked bars represent the experimentally-determined or predicted folding core elements. With the exceptions of ha-LA, CTX-3, and Eqlzm, the predictions yielded by the empirical potential function have substantial overlap with the experimental results. Fig. 3 overlaps experimental phi values with the folding core elements determined by the four methods for 10 of the 27 test proteins.</p><p>Table 2 lists the two measures of overlap (i.e., s and z in Eqs. (2) and (3)) for each of the 27 proteins in the test set in Table 1a. The columns of Table 2 compare the overlap between HX (X) and predictions based on the interaction energies (Eq. (1)) for groups of two, three and four SSEs, as well as the prediction results of other computational methods. These other methods are the fast mode peak residues (H) [30], FIRST (F) [29] and GNM global modes (G) [46] methods. The results show that our method consistently out-performs the three previous studies in terms of the mean values of s and z. The lowest mean value ⟨s⟩ = 2.254 by our method is better than that of H, F and G. For z, the smallest mean value by our method is for the two-SSE case (⟨z⟩ = 5.718), which is better than the mean values by H, F and G.</p><p>For proteins HCA-1, CI-2 and cSH3, all versions of our method are better at matching the HX-detected folding cores than the other methods. However, for ha-LA and Eqlzm, the H, F and G methods are generally better than our method in predicting the HX-detected folding cores. For nearly half of the test proteins (13 of 27), all versions of our method match the HX results with greater than 100% improvement over random agreement (s > 2.0), whereas G can claim only 6 of 27, H can claim 10 of 27 and F can claim 11 of 27 with s > 2.0. In addition, for Bnase and RnaseT1, all methods but G match the HX results with roughly 200% or better improvement over random agreement (s > 3.0). The success of our method in predicting the folding cores of Bnase and RnaseT1 may be due to the use of nucleases RnaseH and Snase in our training set. Interestingly, all methods perform poorly for pAB, which is a small three-helix protein. It is possible that for such a small and symmetrical protein, all elements have rather similar contributions to overall stability.</p><p>In addition, we tested our method on a few proteins (Cytc, ha-LA, scFv and IL-1b) whose secondary-structure definitions, namely the number of SSEs, were modified in the PDB header within the past three years. For ha-LA and scFv, the folding core predictions changed with the increase in the number of SSEs, whereas the predictions remained the same for IL-1b. Furthermore, although the overlap measures s and z declined for ha-LA and scFv with the increase in SSEs, we found no overall correlation between the number of SSEs and our performance in terms of s and z. In fact, we found little correlation between the number of SSEs and overlap performance for all the proteins in the test set (see Supplementary materials).</p><p>For ten of the proteins in our data set, the transient folding-transition states have been assessed by the phi-value approach [18,47-54]. This is an experimental approach to indirectly obtain residue-specific structural information about interactions in the transition state pioneered by Fersht [55]. It is often assumed that the folding core found in HX experiments corresponds to the region adopting native-like structure in the kinetic folding-transition state [13]. For some of the proteins having polarized, highly-organized transition-state structures (e.g., cSH3, Bnase, Ubiquitin and ha-LA), as identified by phi values, our method selects the same structural elements as those harboring residues with high phi values (see Fig. 3). In contrast, for proteins with diffuse folding-transition states (i.e., GB1, CI-2, RnaseA and T4 lysozyme), there is less correlation between phi values and our predicted folding cores (or between HX data as well). Taken together, we conclude that the stable folding cores, as identified by our empirical method or by HX data, often match the kinetic folding-transition states although these sometimes differ; for proteins folding via diffuse transition states involving many partially-formed interactions, the stable folding cores must be assessed by methods other than phi values.</p><p>In summary, we have developed an empirical potential function that can detect protein stability cores revealed by HX experiments. The average prediction results of our method are better than those of previous computational attempts. Although there is still room for improvement in the model, we believe the method reported here provides a more accurate way of estimating stability cores of proteins that can be useful in elucidating the mechanisms of protein folding.</p>

PubMed Author Manuscript

Characterization of Three Tetrabromobisphenol-S Derivatives in Mollusks from Chinese Bohai Sea: A Strategy for Novel Brominated Contaminants Identification

Identification of novel brominated contaminants in the environment, especially the derivatives and byproducts of brominated flame retardants (BFRs), has become a wide concern because of their adverse effects on human health. Herein, we qualitatively and quantitatively identified three byproducts of tetrabromobisphenol-S bis(2,3-dibromopropyl ether) (TBBPS-BDBPE), including TBBPS mono(allyl ether) (TBBPS-MAE), TBBPS mono(2-bromoallyl ether) (TBBPS-MBAE) and TBBPS mono(2,3-dibromopropyl ether) (TBBPS-MDBPE) as novel brominated contaminants. Meanwhile, the mass spectra and analytical method for determination of TBBPS-BDBPE byproducts were presented for the first time. The detectable concentrations (dry weight) of TBBPS-MAE, TBBPS-MBAE and TBBPS-MDBPE were in the ranges 28-394 μg/g in technical TBBPS-BDBPE and 0.1-4.1 ng/g in mollusks collected from the Chinese Bohai Sea. The detection frequencies in mollusk samples were 5%, 39%, 95% for TBBPS-MAE, TBBPS-MBAE and TBBPS-MDBPE, respectively, indicating their prevailing in the environment. The results showed that they could be co-produced and leaked into the environment with production process, and might be more bioaccumulative and toxic than TBBPS-BDBPE. Therefore, the production and use of TBBPS derivatives lead to unexpected contamination to the surrounding environment. This study also provided an effective approach for identification of novel contaminants in the environment with synthesized standards and Orbitrap high resolution mass spectrometry.Recently, increasing studies have been carried out to identify novel brominated contaminants in the environment, especially the derivatives, byproducts and degradation products of brominated flame retardants (BFRs) 1-3 . For example, the mono-modified byproducts or degradation products of tetrabromobisphenol-A (TBBPA) derivatives, including TBBPA mono(allyl ether) (TBBPA-MAE), TBBPA mono(2,3-dibromopropyl ether) (TBBPA-MDBPE), have been found in various environment samples, such as soil, sediment, earthworm and mollusks 2,3 . More importantly, these byproducts and degradation products showed higher bioaccumulation and toxicity than main technical TBBPA products 2,4,5 . Due to the similar structures and production process 6,7 , there should be more mono-modified

characterization_of_three_tetrabromobisphenol-s_derivatives_in_mollusks_from_chinese_bohai_sea:_a_st

<!>Structure Confirmation of Synthesized TBBPS-MAE, TBBPS-MBAE and TBBPS-MDBPE.<!>TBBPS-MAE, TBBPS-MBAE and TBBPS-MDBPE in Technical Products and Mollusk<!>Environmental Risk Prediction of TBBPS Derivatives.<!>Discussion<!>Methods<!>Sample Collection.<!>Analytical Method Validation. TBBPS-MAE, TBBPS-MBAE and TBBPS-MDBPE in samples were

<p>byproducts of TBBPA derivatives co-produced and leaked into environment, which could cause widespread contamination and deserve our more attention.</p><p>As important alternatives of TBBPA, the most widely used BFR, tetrabromobisphenol-S (TBBPS) and TBBPS bis(2,3-dibromopropyl ether) (TBBPS-BDBPE) are extensively produced and applied in electronic devices, plastics, rubber and textiles 8 . As a result, TBBPS and TBBPS-BDBPE have been detected in waste water at a concentration up to 10 μ g/L 9 . TBBPS-BDBPE was also found in herring gull eggs collected from colonies in the Laurentian Great Lakes 10 . Because TBBPS-BDBPE is synthesized by modification of the two phenol groups of TBBPS 7 , the mono-modified byproducts of TBBPS-BDBPE might also be co-produced with technical products and leaked into environment as potential contaminants, such as TBBPS mono(allyl ether) (TBBPS-MAE), TBBPS mono(2-bromoallyl ether) (TBBPS-MBAE) and TBBPS mono(2,3-dibromopropyl ether) (TBBPS-MDBPE). However, the byproducts of TBBPS derivatives were largely ignored in most studies, and there are even no pure standards available. Much still remains unknown about their environmental distribution and risks.</p><p>The lack of analytical methods is another main obstacle for identifying novel contaminants. Because of the thermolability of TBBPA, TBBPS and their derivatives, gas chromatography mass spectrometry (GC-MS) is not applicable for direct analysis of these compounds 11,12 . The high performance liquid chromatography coupled with tandem MS (HPLC-MS/MS) has been developed for analysis of TBBPA and TBBPS derivatives 3,8,10,13 . However, electrospray ionization (ESI) source was reported with poor sensitivity because of the weak polarity of TBBPA and TBBPS derivatives 8 . Although atmospheric pressure chemical ionization (APCI) and atmospheric pressure photoionization (APPI) mass spectrometry methods have been developed, they were not sufficient for the trace level determination of these derivatives in the environment matrices 10,13 . With the rapid development of high resolution mass spectrometry (HRMS), such as time of flight (TOF) HRMS and Orbitrap HRMS, the novel contaminants could be identified and quantified through the full scan acquisition spectrum [14][15][16][17] . An attractive advantage of full scan HRMS is that there is no number limitation of analytes in one single injection, which is enormously beneficial to the retrospective analysis of untargeted contaminants 18 . Furthermore, the exact mass information is helpful for identification of compounds without standards, which largely extends its application 19 . In this view, the combination of ultra HPLC (UHPLC) with Orbitrap Fusion HRMS technique would provide a high accuracy as well as a low detection limit for the mono-modified byproducts of TBBPS-BDBPE.</p><p>The aim of this study was to identify three potential byproducts of TBBPS-BDBPE as novel brominated contaminants. The standards of TBBPS-MAE, TBBPS-MBAE and TBBPS-MDBPE were synthesized with high purity. A sensitive and accurate method for simultaneous determination of these novel TBBPS derivatives was developed with UHPLC-Orbitrap HRMS. Their distribution in mollusk samples collected from the Chinese Bohai Sea and potential risks were discussed in detail. The strategy used in this work could also be an effective approach for identifying other novel brominated pollutants related to BFRs.</p><!><p>The standards of TBBPS-MAE, TBBPS-MBAE and TBBPS-MDBPE were self-synthesized with purities of 99%, 98% and 96%, respectively. The synthesis schemes and 1 HNMR spectra of these compounds are provided in Figure S1 (Supporting Information).</p><p>Orbitrap Fusion HRMS was employed to further identify the target compounds in the full scan mode (Table 1, The recoveries for TBBPS-MAE, TBBPS-MBAE and TBBPS-MDBPE were all higher than 76% with ultrasound method and accelerated solvent extraction (ASE) method at the spiking amounts of 10 ng in 0.5 g Neverita didyma (Nev) samples (n = 7, Table 2). The recoveries of ultrasound method were slightly higher than ASE method and the standard deviations (SDs) were lower than ASE method. Finally, the samples were extracted by ultrasound method and cleaned by ENVI-carb cartridges. The mean recoveries for TBBPS-MAE, TBBPS-MBAE and TBBPS-MDBPE were all higher than 70% and the SDs were all less than 10% at three different spiking amounts, 100 ng (n = 5), 10 ng (n = 7) and 1 ng (n = 5) (Table 2). The method DLs (MDLs) of TBBPS-MAE, TBBPS-MBAE and TBBPS-MDBPE were 0.04 ng/g dry weight (dw), 0.08 ng/g dw and 0.06 ng/g dw, respectively. The mean recovery of the internal standard, 13 C labeled 3,5-dibromophenol (ISDBP), was 101% ± 7% at the spiking amount of 10 ng (n = 7). The recoveries of ISDBP from the real samples ranged from 81% to 104% with a mean recovery of 91% and SD of 6% (n = 38). The matrix effects ranged from 0.86 to 1.05 (Table 2) at three different spiking concentrations, 1 ng/mL, 5 ng/mL and 10 ng/mL. The pretreatment method of TBBPS-MAE, TBBPS-MBAE and TBBPS-MDBPE from the real samples was reliable and repeatable. Details for the development and optimization of extraction and solid phase extraction (SPE) cleanup procedure were provided in Supporting Information.</p><!><p>Samples. The technical product of TBBPS-BDBPE purchased from a BFRs factory (purity > 90%) was dissolved in methanol at a concentration of 100 μ g/mL and determined by Orbitrap Fusion HRMS. The concentrations of TBBPS-MAE, TBBPS-MBAE and TBBPS-MDBPE in this technical TBBPS-BDBPE were 28 μ g/g, 87 μ g/g and 394 μ g/g, respectively.</p><p>The concentrations of TBBPS-MAE, TBBPS-MBAE and TBBPS-MDBPE in total 38 mollusk samples collected from Bohai Sea during 2009 to 2013 were also analyzed with an external standard method (Table S1). TBBPS-MAE was only found in two mollusk samples with concentrations of 0.1 ng/g dw and 0.2 ng/g dw. TBBPS-MBAE was detectable in 15 samples with the concentrations ranging from 0.1 to 1.6 ng/g dw. In these 15 samples, thirteen ones had the concentrations between 0.1 and 0.3 ng/g dw. TBBPS-MDBPE was detectable in 36 samples with the concentrations ranging from 0.3 to 4.1 ng/g dw, among which 20 ones contained the compound higher than 1.0 ng/g dw. The detection frequencies of these three compounds were in the order of TBBPS-MDBPE (95%) > TBBPS-MBAE (39%) > TBBPS-MAE (5%). As shown in Fig. 3, the mean concentration of TBBPS-MDBPE was higher than that of TBBPS-MBAE, and the concentration of TBBPS-MAE was the lowest. A typical mass chromatogram for the three compounds detected in mollusk sample is shown in Fig. 2(C,D).</p><!><p>The physical-chemical properties of TBBPS, TBBPS derivatives and other well concerned contaminants were calculated by US EPA EPI Suite V4.1, which has been widely employed for screening of potentially persistent and bioaccumulative contaminants [20][21][22][23][24][25][26] . As shown in Table 3, the log K ow values of TBBPS and the derivatives ranged from 5.21 to 9.52, log K oa values ranged from 16.83 to 21.83, log K oc values ranged from 4.16 to 6.33, and log K aw values were all lower than − 8.81. The bioconcentration factor (BCF) values of TBBPS-MAE, TBBPS-MBAE and TBBPS-MDBPE were 10730, 13200 and 8829, respectively, which were significantly higher than those of TBBPS (1266), TBBPS bis(allyl ether) (TBBPS-BAE) (4207) and TBBPS-BDBPE (775).</p><p>In addition, the potential toxicity of TBBPS-BDBPE, TBBPS-BAE, TBBPS, TBBPS-MAE, TBBPA-MBAE and TBBPS-MDBPE were also estimated with the primary cerebellum granule cells (CGCs) as the model, which were usually used for neurotoxicity studies 13,27,28 . The IC 50 of TBBPS, TBBPS-MAE, TBBPA-MBAE and TBBPS-MDBPE were 0.45, 0.19, 0.20 and 0.17 μ M, respectively. The IC 50 of TBBPS-BAE and TBBPS-BDBPE were 13.1 and 11.2 μ M. TBBPS and the three derivatives with phenol groups, TBBPS-MAE, TBBPS-MBAE and TBBPS-MDBPE, inhabited 50% of the cell viability at a much lower concentration than TBBPS-BAE and TBBPS-BDBPE.</p><!><p>Since the standards of TBBPS-BDBPE byproducts were not commercially available, TBBPS-MAE, TBBPS-MDBPE and TBBPS-MBAE were self-synthesized and further characterized by 1 HNMR, UHPLC-Orbitrap Fusion HRMS (Thermo Fisher scientific, USA) and HPLC-UV. The results indicated the successful synthesis with high purity (> 96%) of TBBPS-MAE, TBBPS-MBAE and TBBPS-MDBPE. These compounds could be used as the standards for the further analysis.</p><p>In order to identify the target compounds in samples, an accurate and sensitive method was developed. By using the highly sensitive Orbitrap Fusion HRMS, the IDLs for TBBPS-MAE, TBBPS-MBAE and TBBPS-MDBPE were in the range 0.06-0.1 pg which were lower than those acquired with HPLC-ESI-MS/ MS. The target compounds could be identified according to the accurate m/z values of precursor ions within a mass tolerance of 5 ppm. Meanwhile, in the full scan mode, the isotope information was positively observed with the quantification process. As for optimizing the extraction method, the ultrasound method showed slightly higher recoveries and lower SDs which meant it was a stable and reliable extraction method. Meanwhile, the matrix effects were all close to 1.0 which indicated the interference from the matrix could be ignored. ENVI-Carb SPE cartridges could effectively eliminate the interference and concentrate the target compounds. The pretreatment method was reliable and repeatable for the identification of TBBPS-MAE, TBBPS-MBAE and TBBPS-MDBPE in mollusk samples. Meanwhile, ISDBP was selected as an appropriate internal standard for the recovery monitor of TBBPS-MAE, TBBPS-MBAE and TBBPS-MDBPE from the mollusk samples. The recoveries of ISDBP were all higher than 80% in all the samples, indicated the recoveries of target compounds from real samples were reliable. With the proposed method, the existence of TBBPS-MAE, TBBPS-MBAE and TBBPS-MDBPE in technical TBBPS-BDBPE and mollusk samples from the Chinese Bohai Sea was studied in detail. In the sampling area of this work, several BFRs factories produce TBBPS and TBBPS-BDBPE on a large scale. TBBPS-MAE, TBBPS-MBAE and TBBPS-MDBPE were detectable in technical TBBPS-BDBPE of the BFRs factory with the concentrations ranged from 28 to 394 μ g/g. Consequently, in mollusk samples, TBBPS-MAE, TBBPS-MBAE and TBBPS-MDBPE were detectable at a level of ng/g dw. Therefore, the BFRs factories might be the point sources of TBBPS-MAE, TBBPS-MBAE and TBBPS-MDBPE. They were probably produced with the manufacture process, and leaked into the environment through the production and application process. The production process of TBBPS-BDBPE might influence the concentration and detection frequency of the byproducts. TBBPS-BDBPE is synthesized from TBBPS-BAE, and TBBPS-BAE is synthesized from TBBPS. Therefore, TBBPS-MDBPE, which has the most similar structure with TBBPS-BDBPE, is the main byproduct of TBBPS-BDBPE. As a result, TBBPS-MDBPE was detected at the highest concentration and detection frequency. While the technical TBBPS-BAE is produced as intermediate of TBBPS-BDBPE, its structure related byproduct, TBBPS-MAE, showed the lowest concentration and detection frequency. TBBPS-MAE and TBBPS-MBAE were detected in the mollusk samples at the similar concentration level with previously reported for TBBPA derivatives, TBBPA-MAE and TBBPA-MDBPE 3 . TBBPS-MDBPE showed higher concentration level (> 1.0 ng/g dw) than TBBPA-MDBPE (< 1.0 ng/g dw) in the mollusk samples. 3 The detected concentrations of TBBPS-MAE, TBBPS-MBAE and TBBPS-MDBPE were lower than Tris-(2,3-dibromopropyl) isocyanurate (TBC) 29 , hexabromocyclododecane (HBCD) 29 and polybrominated diphenyl ether (PBDE) 29,30 which were also detected in the mollusk samples with the concentration ranges of below detection limit (nd) to 12.1 ng/g dw, nd to 28.8 ng/g dw and 0.01 to 59 ng/g dw, respectively. The detection frequency of TBBPS-MDBPE (95%) was comparable with that reported for HBCD (99%) and PBDE (100%) 29,30 . The difference of the concentrations of TBBPS-MAE, TBBPS-MBAE and TBBPS-MDBPE between years was not significant. Interestingly, the detection frequency of TBBPS-MDBPE was 95% which indicated it was probably one widely dispersed brominated compound. Significant difference was not observed among the concentrations of TBBPS-MAE, TBBPS-MBAE and TBBP-MDBPE in different mollusk species. The concentrations of TBBPS-MAE, TBBPS-MBAE and TBBPS-MDBPE remained at a similar level which indicated these compounds could persistently accumulate in the mollusks. The property of persistent accumulation in the biota system may result in their potential health risks posing on the aquatic ecosystem.</p><p>Furthermore, the environmental risks of TBBPS-MAE, TBBPS-MBAE and TBBPS-MDBPE were evaluated. Their physical-chemical properties were calculated by US EPA EPI suite. The log K ow values of TBBPS-MAE, TBBPS-MBAE and TBBPS-MDBPE were close to TBBPA (7.20) 20 , TBC (7.37) 22 and HBCD (7.74) 29 and higher than 5. These results indicated the accumulation of TBBPS-MAE, TBBPS-MBAE and TBBPS-MDBPE in organic materials such as fat-rich organisms. The high K oa values of TBBPS-MAE, TBBPS-MBAE and TBBPS-MDBPE indicated low respiratory elimination rate and high bioaccumulation ability in respiratory organisms. Meanwhile, the low K aw values implied that large amount of TBBPS-MAE, TBBPS-MBAE and TBBPS-MDBPE would participate in water rather than the air at the boundary exchange process. The comparable K oc values with TBBPA (5.24) and TBC (4.92) indicated TBBPS-MAE, TBBPS-MBAE and TBBPS-MDBPE could be absorbed by the sediment and soil with a considerable amount. Usually, the chemicals with log K ow > 5 and BCF > 5000 are considered to be bioaccumulative 31 . In addition, the BCF values of TBBPS-MAE, TBBPS-MBAE and TBBPS-MDBPE were higher than TBBPS-BDBPE, TBBPA derivatives and other environmental contaminants in Table 3. They were more bioaccumulative than TBBPS-BDBPE and TBBPA derivatives. As novel brominated contaminants, the toxicity of TBBPS-MAE, TBBPS-MBAE and TBBPS-MDBPE were essential for their environment risks assessment. By using the CGCs as a model, the IC 50 of TBBPS-MAE, TBBPS-MBAE and TBBPS-MDBPE were all lower than 0.2 μ M, suggesting that they were more toxic than TBBPS-BDBPE and TBBPS-BAE with IC 50 higher than 11 μ M (Figure S3). This is probably accused by the phenol group in the structure which probably increased the toxicity of the BFRs 32 . The production of technical products would bring some novel compounds with more severe toxicity into the environment. These byproducts in the environment showed potential health risk to human. Further studies about the toxicity of these byproducts are urgently needed. By using Orbitrap HRMS, we also found several unknown brominated compounds. With the accurate results determined by Orbitrap HRMS, the compound contained Br is easier to be identified because of the special properties: 1) m/z value of the decimal part decreases with the number increase of 81 Br; 2) the isotope ratio is different for compounds contained different number of Br. In this study, some untargeted brominated organics also showed up in the samples. Their spectra are shown in Fig. 4 and the molecular formulas were calculated by Xcalibur software. The three compounds detected in mollusk samples showed the properties of containing bromine atoms. Within delta ppm < 5, the m/z detected at RTs 2.8 min (isotope ratio, 1:3:3:1), 2.6 min (isotope ratio, 4:6:4) and 1.7 min (isotope ratio, 1:2:1) were calculated to be [C 6 H 2 OBr 3 ] − , [C 12 H 5 O 4 Br 4 S] − and [C 6 H 2 O 3 NBr 2 ] − , respectively. They might be three kinds of bromophenols. These three untargeted compounds were further analyzed with Orbitrap Fusion HRMS with reference to the standards, including TBBPS, 2,6-dibromo-4-nitrophenol (DBNP) and four kinds of tribromophenol (Figure S4). The untargeted peak detected at RT 2.64 min (Fig. 4B2) showed similar RT and mass spectra with TBBPS (RT 2.62 min, isotope ratio 1:4:6:4:1). The untargeted peak detected at RT 1.76/1.73 min (Fig. 4A1,B1) had the similar RT and mass spectra with DBNP (RT 1.72 min, isotope ratio 1:3:1). DBNP was also identified as novel bromophenol compounds showing toxicity and potential risk to human [33][34][35] . DBNP could formed in the chlorination of drinking water and saline sewage effluent 36 . The untargeted peak detected at RT 2.82 min (Fig. 4A2) presented the similar mass spectra with all the four kinds of tribromophenol and the similar RTs with 2,3,4-tribromophenol, 2,4,6-tribromophenol and 2,4,5-tribromophenol. It might be 2,4,6-tribromophenol as it is one kind of mass-produced BFRs in the sampling area. We did not quantify these untargeted compounds because of the lack of reliable pretreatment method. However, their presence in the mollusk samples could be determined with the quantification of our target compounds by Orbitrap Fusion HRMS. In all the 38 mollusk samples, 7 samples were detected to contain tribromophenol, 8 samples for TBBPS and 17 samples for DBNP. The detection frequencies of these three compounds were all higher than TBBPS-MAE. The anthropogenic activities might result in the emergence of 2,4,6-tribromophenol, DBNP and TBBPS in the environment as they were not reported as natural compounds 37 . The untargeted compounds might also become novel brominated contaminants. In this view, further investigation is needed to be conducted on the identification and environmental fate of these compounds. It is worth mentioning that the Orbitrap Fusion HRMS is a powerful tool for the quantification of novel contaminants and qualitative analysis of unknown contaminants with one injection.</p><p>Most BFRs, such as poly brominated diphenyl ether (PBDE), TBBPA and TBBPS derivatives, usually share the ether bond linked structure 6 . For the production of ether bond derived organic aromatic chemicals with several bromine atoms, the left-over starting reagents, co-produced phenol and less brominated byproducts could be potential environmental contaminants together with the desired BFRs products. The byproducts generated from manufacture production or degradation draw great attention because they were found in various environment compartments as novel or emerging BFRs [38][39][40][41] . For example, the byproducts of TBBPA and its derivatives, TBBPA-MAE, TBBPA-MDBPE, TBBPA mono(2-hydroxylethyl ether), TBBPA mono(glycidyl ether), dibromobisphenol A and tribromobisphenol A have been determined in water, soil and biota system 2,3,42 . In this work, we found the manufacture process of TBBPS-BDBPE resulted in the occurrence of TBBPS-MAE, TBBPS-MBAE and TBBPS-MDBPE inevitably. The existence of the phenol byproducts in the aryl-O linked technical products might be a global problem. Although aryl-O bond of organic chemicals is considered very stable in chemical reactions, its cleavage is easy to fulfill under the bacterial biodegradation, the UV irradiation and super-reduced conditions [43][44][45] . TBBPA bis(2,3-dibromopropyl ether) (TBBPA-BDBPE) was also found to transform to TBBPA via ether breakage in aquatic mesocosm 46 . The compounds with ether bond are not as stable as suppositional under environmental conditions. TBBPA-MAE and TBBPA-MDBPE were also predicted to be the degradation products of TBBPA bis(allyl ether) (TBBPA-BAE) and TBBPA-BDBPE by the University of Minnesota Pathway Prediction System 2,3 . Through the same microbial transformation, TBBPS derivatives showed the potential ability of ether bond cleavage and form the mono-modified degradation products, TBBPS-MAE and TBBPS-MDBPE. The co-produced byproducts in manufacture process and microbial degradation in the environment contribute to the occurrence of mono-modified byproducts in the environment. The study about the byproducts and degradation products of these ether linked BFRs will supplement the information for novel brominated contaminants.</p><p>In conclusion, TBBPS-MAE, TBBPS-MBAE and TBBPS-MDBPE were identified as three novel brominated contaminants, which showed higher bioaccumulation properties and potential severe toxicity compared with TBBPS-BDBPE. They could be co-produced and leaked into the environment along with production process of TBBPS-BDBPE. The occurrence of the mono-modified byproducts or degradation products of the extensively used brominated products might be a widespread problem. This work could promote the further study of the environmental fate and risks of widely used TBBPS and TBBPS-BDBPE. The strategy used in this work, integrating the synthesis of standards and Orbitrap HRMS identification, could also be an effective approach for identifying other novel brominated pollutants related to BFRs.</p><!><p>Chemicals and Materials. TBBPS (98%) was purchased from Beijing Apisi biotechnology co. ltd., and was used without further purification. Ammonium hydroxide (50%) was purchased from Sigma-Aldrich. Methanol, acetone, hexane and methylene dichloride (DCM) were all HPLC grade. Ultra-pure water was generated by a Milli-Q advantage A10 system. TBBPS-MAE, TBBPS-MBAE and TBBPS-MDBPE were synthesized and purified in our lab. The synthesis procedures were described in Supporting Information. The purities of these three compounds were 99%, 98% and 96% as determined by HPLC-UV (214 nm).</p><!><p>From 2009 to 2013, in August of each year, 11 species of mollusks were collected from one coastal city -Shouguang, Shandong Province. These 11 selected species of mollusks were Rapana venosa (large and small, RapL and RapS), Crassostrea talienwhanensis (Ost), Scapharca subcrenata (Sca), Cyclina sinensis (Cyc), Mya arenaria (Mya), Mactra veneriformis (Mac), Chlamys farreri (Chl), Neverita didyma (Nev) and Meretix meretrix (large and small, MerL and MerS) (Fig. 3, Table S1). After sampling, the mollusks were frozen and transported on ice to the laboratory, and then cleaned by water. The collected samples were disposed according to the previous method 29 . The samples were freeze-dried, grinded, and preserved at − 20 °C until analysis. A total of 38 mollusk samples were obtained and analyzed.</p><p>Sample Pretreatment. Ultrasound Extraction. Mollusk (0.5 g) samples were mixed with 2 g anhydrous Na 2 SO 4 ; spiked with 10 ng 13 C labeled 3,5-dibromophenol (ISDBP); extracted with 10 mL DCM/ hexane (8/2, V/V) for three times by sonication (30 minutes per time). After centrifugation, the extraction solution was collected and the solvent was removed with rotary evaporator and re-dissolved in 3 mL DCM/hexane (1/1, V/V) before SPE process. SPE Procedures. The SPE cartridges (Supelclean TM ENVI-Carb TM , 0.5 g, 6 mL) were first conditioned by 5 mL acetone, 5 mL DCM and 10 mL hexane and then the samples were loaded. Then the cartridges were cleaned by 5 mL hexane and 5 mL DCM/hexane (1/1, V/V). Finally, the cartridges were eluted with 10 mL acetone (containing 0.5% NH 3 •H 2 O) and the elution were collected and blown to dryness by gentle nitrogen gas flow. The residue was solvent-exchanged to 1 mL methanol and analyzed by UHPLC-Orbitrap Fusion HRMS.</p><p>Instrument Parameters for UHPLC-Orbitrap Fusion HRMS, HPLC-UV and HPLC-ESI-MS/MS Analysis. The details were described in Supporting Information.</p><!><p>identified by retention time and accurate m/z of the precursor ions comparison with the corresponding standards. Quantification of the target compounds in the environmental samples was performed by peak area of the accurate precursor ions of compounds within 5 ppm mass tolerance. For TBBPS-MAE, TBBPS-MBAE and TBBPS-MDBPE, precursor ions at m/z 604.69189, 682.60248 and 764.52643 were used as quantification ions, and m/z 602.69415, 684.60022 and 762.52856 were used as the qualitative ions.</p><p>All calibration standards and spiking solutions were prepared by serial dilution in methanol. A linear calibration curve with seven points ranged from 0.05 to 100 ng/mL was used to quantify the target compounds with a determination coefficient (R 2 ) higher than 0.99 (Table 2). The concentrations of target compounds were determined by an external standard method. Every 9 samples were prepared with one blank sample (only anhydrous Na 2 SO 4 added), and analyzed with methanol as solvent blank to make sure no cross contamination. The DLs were determined by the lowest mass value of the target compounds that Orbitrap Fusion HRMS detected. The IDLs were determined for five times within 20% relative standard deviation for the signals. The MDLs were based on replicate analysis (n = 10) of Nev sample spiked at a mass concentration of 5 times of the IDLs and calculated with the method previously used for HRMS 47 . The recoveries were determined at the spiking amounts of 1 ng (n = 5), 10 ng (n = 7) and 100 ng (n = 5) in 0.5 g Nev samples (not containing target compounds). The internal standard 13 C labeled 3,5-dibromophenol (ISDBP) was used to monitor the pretreatment process and not used for the concentration calculation. The detailed procedures for the SPE optimization and results were described in Supporting Information. The matrix effects were determined according the method reported elsewhere previously 43,48 . Detailed information regarding the synthesis routines and 1 HNMR and MS 2 spectra of TBBPS-MAE, TBBPS-MBAE and TBBPS-MDBPE, the 1 HNMR data, instrumental analysis information of Orbitrap Fusion HRMS, HPLC-UV and HPLC-ESI-MS/MS, the optimization of pretreatment method, the concentration of every mollusk sample, the cell information and cytotoxicity test method, the cytotoxicity of TBBPS and its derivatives, the HRMS chromatograms and spectra of different brominated phenols are provided in the Supporting Information.</p>

Scientific Reports - Nature

The Phase Behavior of γ-Oryzanol and β-Sitosterol in Edible Oil

The phase behavior of binary mixtures of γ-oryzanol and β-sitosterol and ternary mixtures of γ-oryzanol and β-sitosterol in sunflower oil was studied. Binary mixtures of γ-oryzanol and β-sitosterol show double-eutectic behavior. Complex phase behavior with two intermediate mixed solid phases was derived from differential scanning calorimetry (DSC) and small-angle X-ray scattering (SAXS) data, in which a compound that consists of γ-oryzanol and β-sitosterol molecules at a specific ratio can be formed. SAXS shows that the organization of γ-oryzanol and β-sitosterol in the mixed phases is different from the structure of tubules in ternary systems. Ternary mixtures including sunflower oil do not show a sudden structural transition from the compound to a tubule, but a gradual transition occurs as γ-oryzanol and β-sitosterol are diluted in edible oil. The same behavior is observed when melting binary mixtures of γ-oryzanol and β-sitosterol at higher temperatures. This indicates the feasibility of having an organogelling agent in dynamic exchange between solid and liquid phase, which is an essential feature of triglyceride networks.

the_phase_behavior_of_γ-oryzanol_and_β-sitosterol_in_edible_oil

Introduction<!><!>Introduction<!>Materials<!>Preparation of Binary γ-Oryzanol + β-Sitosterol Mixtures<!>Preparation of Organogel<!>Differential Scanning Calorimetry (DSC)<!>X-ray Scattering<!><!>Binary Mixtures<!><!>Thermal Behavior of Pure Compounds and Binary Mixtures (DSC Data)<!><!>SAXS Data of Binary Mixtures<!><!>SAXS Data of Binary Mixtures<!>Ternary Mixtures (Organogels)<!><!>Thermal Behavior of Ternary Mixtures (Organogel)<!><!>SAXS Data of Ternary Mixtures (Organogel)<!>Melting Behavior of Binary Mixtures<!><!>Melting Behavior of Binary Mixtures<!>Conclusions

<p>Mixtures of plant sterols (e.g., γ-oryzanol and β-sitosterol) can be potentially used as an alternative to crystalline fats like triacylglycerols (TAGs) for structuring the oil phase of some food products. Certain crystalline TAGs contribute to raising blood cholesterol as a result of their high saturated fatty acids content and have been identified as a risk factor for cardiovascular diseases [1–4]. In contrast, reports showed that the intake of plant sterols has a blood cholesterol-lowering effect; for instance, the blood cholesterol content can be reduced by up to about 10 % upon a total intake of 2 g plant sterols/day [5]. Therefore, plant sterol-containing food products are expected to have a positive impact on human health. Recently, some studies reported on the use of plant sterol (ester + sterol) mixtures (γ-oryzanol and β-sitosterol) for the preparation of organogels and emulsions [6–12]. It was found that γ-oryzanol and β-sitosterol self-assemble into hollow double-walled tubules (ca. 10 nm in diameter) forming a transparent and firm organogel [10]. The formation of tubules is driven by enthalpy changes, indicating that the molecular interaction between γ-oryzanol and β-sitosterol determines the properties of the resulting gel to a large extent [12, 13]. The physical and mechanical properties of the gel, including firmness, transparency, melting and tubular microstructure, were found to be dependent on the γ-oryzanol to β-sitosterol ratio. For instance, the firmest gel was obtained with 1:1 molar ratio of γ-oryzanol to β-sitosterol and the gels rich in β-sitosterol were more hazy and had a higher melting point.</p><!><p>Dashed black lines: schematic representation of the ternary phase diagram of the β-sitosterol + γ-oryzanol + canola oil mixture according to AlHasawi and Rogers [14]. The diagram is based on weight percentages. The numbers in the figure indicate the regions belonging to different phases, and refer to the figure numbers in Ref. [14]. Dotted gray lines: ternary and binary mixtures as used in the present study, where sunflower oil was used as the edible oil. Solid gray dots: binary mixtures as used in melting experiments in the present study</p><!><p>The objective of the present paper was to study certain aspects of the ternary γ-oryzanol + β-sitosterol + edible oil phase diagram in more detail. Two specific questions were addressed: (1) How do the properties of binary γ-oryzanol and β-sitosterol mixtures change as a function of the ratio of both components? In particular, the question of the whether the microstructure of the mixture changes gradually as a function of composition or whether sudden transitions occur is of interest. (2) How does the structure of the tubules in the ternary mixture of ternary γ-oryzanol + β-sitosterol + oil change if the amount of liquid oil is gradually reduced? The boundaries are schematically indicated by the dotted lines in Fig. 1. These questions were addressed using differential scanning calorimetry (DSC) and small-angle X-ray scattering (SAXS).</p><!><p>The γ-oryzanol (Tsuno Rice Fine Chemicals, Wakayama, Japan) and tall oil sterol (78.5 % β-sitosterol, 10.3 % β-sitostanol, 8.7 % campesterol, and 2.5 % of other minor sterols, Unilever, the Netherlands) were used in this study to prepare the binary mixtures. The organogels were prepared with sunflower oil (Reddy, NV Vandemoortele, Breda, the Netherlands) as a lipid phase. All materials were used as received.</p><!><p>To prepare well-mixed binary mixtures, the γ-oryzanol powders and β-sitosterol granules were first melted at ca. 165 °C in the oven and the molten solution was manually mixed. The mixed solution was then left overnight in the oven (the oven was switched off) to solidify while cooling to room temperature. The solidified binary mixture was manually ground and was ready for characterization with DSC and SAXS. The pure compounds were also subjected to the same treatment as of the binary mixtures.</p><!><p>For the preparation of the organogel, γ-oryzanol and β-sitosterol were dissolved in sunflower oil using a magnetic stirrer equipped with a heater. The hot solution was subsequently cooled down until a gel was formed and stored in the fridge at 4 °C for at least 1 week before characterization. Organogels with different γ-oryzanol to β-sitosterol ratios and different total sterol concentrations were prepared.</p><!><p>Thermal transitions in binary γ-oryzanol/β-sitosterol mixtures and organogels were studied using a differential scanning calorimeter (Perkin Elmer Diamond DSC, Perkin-Elmer Co., Norwalk, CT). The ground binary mixtures or organogels samples (7–15 mg of each sample) were sealed in stainless steel cups and then scanned with the DSC system. Binary mixtures were scanned using the following procedure: heating from 5 to 160 °C, hold at 160 °C for 10 min, then cooling from 160 to 0 °C, hold at 0 °C for 60 min, and finally heating again from 5 to 160 °C. All heating/cooling rates were 10 °C/min. For some of the binary mixtures, an annealing step was applied in which the sample was cooled down from 160 to 90 °C, then kept at 90 °C for 300 min after which it was further cooled down to 0 °C. Previous work had shown that a temperature-holding step reduced the risk of entering a metastable state in oryzanol-rich samples.</p><p>The organogels were scanned with a slightly different program: heating from 0 to 120 °C, hold at 120 °C for 10 min, then cooling from 120 to 0 °C, hold at 0 °C for 10 min, and reheating from 10 to 120 °C with 10 °C/min as the heating/cooling rate.</p><!><p>Small-angle and wide-angle X-ray scattering (SAXS, WAXS) experiments were performed at the high-brilliance ID2 beamline of the European Synchrotron Radiation Facility (ESRF) in Grenoble, France [16]. Details of the experimental setup are given elsewhere [8]. SAXS/WAXS data were collected in two runs in either the range 0.079 nm−1 < q < 4.7 nm−1 and 3.1 nm−1 < q < 29.5 nm−1 or 0.094 nm−1 < q < 4.5 nm−1 and 3.9 nm−1 < q < 42.2 nm−1, respectively, where q is the scattering vector defined by q = 4π·sinθ/λ (with θ the scattering angle and λ the wavelength of the incoming X-ray beam). Scattering data were corrected for scattering from the oil phase by subtraction of the pure oil signal.</p><p>A number of complementary X-ray diffraction (XRD) measurements were performed (at Unilever R&D Vlaardingen) using a Bruker D8-Discover in a θ/θ configuration. Data was collected in the range 0.64 nm−1 < q < 7.7 nm−1 [11]. Data in these complementary experiments were not corrected for the contribution of oil.</p><!><p>Melting temperature of γ-oryzanol + β-sitosterol binary mixtures and ternary mixtures in sunflower (organogel) as a function of β-sitosterol weight fraction in the sterol mixture and oil concentration. The data shown in this figure were obtained from the second heating DSC scans. Oil in sample: circles 0 %, filled circles 10 %, squares 60 %, filled squares 84 %, triangles 92 %. The dashed lines emphasize the presence of a local maximum in the melting curves. Together with the lowering of the melting point of the pure components this leads to a W-shaped melting curve</p><!><p>The underlying mechanism can be studied by a combination of both DSC and SAXS [18].</p><!><p>Simplified binary phase diagram for the binary γ-oryzanol + β-sitosterol system, as inferred from the DSC data. The vertical line CF indicates the existence of a compound, the curve through points ABCDE is the liquidus, the area below the liquidus consists of two-phase areas of the compound OmSn and either sitosterol (S) or oryzanol (O). For details, see text</p><!><p>On the basis of the DSC data and assuming the formation of a 1:1 molar compound (40 % w/w β-sitosterol), the binary phase diagram as shown in Fig. 3 can be constructed. Depending on composition and temperature the phase diagram has the following properties: a completely mixed liquid phase containing components O (γ-oryzanol) and S (β-sitosterol) above the liquidus (curve ABCDE), four liquid–solid two-phase regions, and two solid–solid two-phase regions. Point A indicates the melting point of pure component O, while point E indicates the melting point of pure component S. The temperature maximum also denoted as a dystectic point is indicated by point C. As suggested by the experimental data and dictated by theory (e.g., Nerad et al., 18), a smooth curve is drawn around the dystectic point. As pointed out above the compound separates the binary system into two "pseudo-binary" subsystems. Similar to many phase diagrams revealed in lipid science there is no information on the state of mixing in the solid phases and the diagram here hence assumes complete immiscibility. This means that at temperatures lower than given by the eutectic point B the solids in the mixed system are dependent on the composition, i.e., either the solid compound coexisting with pure solid γ-oryzanol or solid compound coexisting with pure solid β-sitosterol.</p><!><p>SAXS data of different γ-oryzanol + β-sitosterol binary mixtures. Data taken at 10 °C after at least 1 week storage at 5 °C. γ-Oryzanol to β-sitosterol mass ratio from top to bottom: 100:0, 90:10, 80:20, 70:30, 60:40, 50:50; 40:60, 30:70, 20:80, 10:90, and 0:100</p><!><p>For the γ-oryzanol + β-sitosterol mixtures, Fig. 4 shows three regions. First of all, a region for compositions rich in β-sitosterol can be identified (ca. 70–100 % β-sitosterol), in which sharp peaks associated with the pure β-sitosterol crystal structure can be observed.</p><p>A similar region can be found for ca. 70–100 % γ-oryzanol, although the scattering shows fewer sharp features than for β-sitosterol—especially in the WAXS range around 10 nm−1. It is interesting to note that the main scattering peak at ca. 1.4 nm−1 shifts to slightly higher q values from 1.29 to 1.44 nm−1 when the γ-oryzanol is diluted in β-sitosterol. Using d = 2π/qi, where qi identifies the wave vector peak position associated with a particular peak, this indicates that the average bilayer distance decreases from 4.89 to 4.35 nm as would be expected as a result of the mixing in of a shorter molecule (a weak bilayer peak in γ-oryzanol powder can be found at 4.95 nm−1, whereas for β-sitosterol a bilayer spacing of ca. 3.8 nm−1 is expected [19]). The second, weaker peak at ca. 2.7 nm−1 originates from half the bilayer distance.</p><p>The third region in Fig. 4 shows qualitatively different behavior from the pure substances. Although we have not been able to quantitatively interpret the scattering data, it is clear that the pattern is much closer to that observed for the tubules in the organogel. One hypothesis is that the data reflects a flattened version of the tubules, as no liquid oil is present to fill the tubules. Tentative support for this interpretation can be obtained from SEM images of somewhat flattened tubules in de-oiled samples [15].</p><p>As pointed out above, the phase behavior as shown in Fig. 3 is solely based on DSC data. Combining the SAXS data with some general knowledge it is fairly straightforward to further detail the phase behavior. The assumption of complete immiscibility has to result in a very regular evolution of the scattering data. In this case the scattering data should be a simple linear combination of the pattern of the two solid structures coexisting which each other. In contrast it turns out on combination of diffractograms that none of the intermediate scattering patterns relating to compositions between 0 to 40 % (w/w) β-sitosterol can be made up from the patterns of the compound crystal and the pure γ-oryzanol. As mentioned before, the data rather suggest a gradual change as a function of the β-sitosterol inclusion level. However, the existence of a single mixed crystalline structure over the range from 0 to 30 % w/w) β-sitosterol inclusion is in conflict with the rules of equilibrium phase behavior. Starting from point B (eutectic) lower temperatures have to relate to a region of two coexisting solid phases. This conflict is possibly resolved by accepting that the DSC and X-ray data of pure γ-oryzanol relate to different crystalline structures. It appears that the possibly stable structure found in the mixing range from more than 5 % to more than 30 % (w/w) β-sitosterol extends metastably into the range of higher γ-oryzanol concentrations.</p><!><p>Full binary phase diagram for the binary γ-oryzanol + β-sitosterol system based on both DSC and X-ray scattering data. Number ϕ indicates the number of coexisting phases. The one-phase (1ϕ) compositions have broadened to regions of finite width compared to the simplified diagram, the curve through points ABGCDE is the liquidus, the two-phase (2ϕ) areas below the liquidus have shrunk. For details, see text</p><!><p>The identified positions of the transitions are in qualitative agreement with the data by AlHasawi and Rogers [14]. These authors indicate transitions at approximately 30 and 70 % β-sitosterol. In the range of high γ-oryzanol concentrations these authors did not identify a transition. However, as pointed out above the system is prone to remain in metastable states. Consequently is the transition at high γ-oryzanol concentrations (around 95 %, w/w) primarily a must dictated by the clearly identified eutectic point.</p><p>However, the complexity introduced on the basis of the SAXS data and displayed in Fig. 5 is in better agreement with the work of AlHasawi and Rogers [14] than the phase diagram shown in Fig. 3.</p><!><p>The studies on the binary mixtures established that oil-free mixtures below the melting temperature do not show the features of full tubule formation despite some features in the scattering data being reminiscent of the sterol tubules. The next stage is to proceed to mixtures containing edible oils to study the transition from binary mixture to tubules in the presence of liquid oil.</p><!><p>Schematic projection of the melting plane for the γ-oryzanol + β-sitosterol + sunflower oil ternary mixture assuming a binary phase behavior as described in Fig. 3. The dashed line xyk 2 describes the solidification trajectory of a sample of composition x in the melting plane, lines k 1 k 2 and k 3 k 4 connect the binary eutectic points in the diagram, the dotted vertical line represents the composition of the compound (the dotted line would be drawn at an angle as in Fig. 1 if the intention had been to draw a semiquantitative diagram)</p><!><p>The diagram in Fig. 6 predicts a central region dominated by the OmSn compound, in combination with either S or O crystals. Along the axes of the diagram we find O crystals along the O–Oil line, and S crystals along the S–Oil line. The line along the O–S axis passes for the most part through a range for the OmSn compound, except at the edges which are again dominated by pure crystals (either O or S). For the oil-free cases, region I corresponds to line AB, region II to line BC, region III to line CD, and region IV to line DE in Fig. 3, but extends also along the oil axis. Close to the oil vertex, there is a small region which reflects the crystallization of the edible oil at low T. Since the ternary mixtures are not cooled below 0 °C, the oil remains liquid and the top part of the diagram in Fig. 6 remains unexplored. We can compare Fig. 6 to the ternary phase diagram as presented by AlHasawi and Roger [14] (Fig. 1 as obtained by different experimental techniques). It should be realized that the microscopic images in their diagram are of the final structure at 30 °C, while the diagram in Fig. 6 allows for following the crystallization path upon cooling.</p><!><p>SAXS data of organogels prepared with (60:40 w/w mixture of γ-oryzanol/β-sitosterol) in sunflower as a function of oil concentration. Data taken at 10 °C after at least 1 week storage at 5 °C. From top to bottom: 0, 5, 10, 20, 40, 68, 84, and 92 % oil (w/w)</p><!><p>A pattern builds up from 5 % towards 40 % oil, starting from the small d peak but involving more and more larger d peaks at increasing oil content (in order of appearance: d = 2.30 nm, d = 3.15 nm, and d = 6.53 nm) under simultaneous weakening of the 4.19-nm feature. Taking the d = 2.30 nm peak as the shortest distance, the sequence of ratios of the d values for 5 and 10 % oil (1:1.37:1.83) comes close to the 1:√2:√3 sequence of a liquid crystalline cubic structure, as was noted before by AlHasawi and Rogers [14]. Given the fact that the smallest two d values coincide with those observed later at higher dilutions for tubules, the tentative cubic structure probably shares some elements with the tubules. The commonality should occur at a length scale well below a single twist of a helical ribbon (i.e., no more than a few stacked molecules); however, model calculations show that even very short tubules show all the reflections of infinitely large tubules. We therefore do not expect that these tentative cubic structures can be assigned to the full central region in the phase diagram and also because there is ample supporting evidence that tubule formation occurs over most part of the central region of the phase diagram [15]. For 20 % oil, the appearance of a peak close to d = 6.5 nm completes the set of features that characterize the tubule diffraction, albeit at different intensities than in a fully developed interference pattern due to tubules. The simultaneous intensity increase of the peaks associated with tubules with an intensity decrease of the d = 4.19 nm feature suggests the coexistence of two different structures in these systems with small amounts of oil.</p><p>It is possible to provide a rough estimate of the amount of oil needed to develop the tubules fully. Assuming the parameters obtained from earlier neutron scattering experiments gives rin = 2.43 nm and rout = 4.74 nm [10]. Assuming maximum hexagonal packing of perfectly aligned tubules and using (1 − π√3/6) + (π√3/6) · (rin/rout)2 = 0.33, it is found that a minimum oil content of about 33 % is needed to develop the tubules fully (where we ignore the fact that the outside tubule has a lower density because only half of the molecules in the β-sitosterol + γ-oryzanol tubule contain the ferulic acid moiety).</p><p>Indeed, Fig. 7 shows that the tubule scattering pattern is found at oil contents of 40 % and above, which is consistent with this lower boundary. Note that the apparent loss in intensity of the peaks associated with the d = 2.30 nm and 3.15 nm features is due to an increase in intensity of a broad liquid oil feature, which has not been subtracted.</p><p>The slope at low q indicates that formation of linear structures (like tubules) occurs from 10 % oil onwards. In addition, it is noteworthy—though mentioned previously [8]—that the WAXS area lacks any sharp features for the samples with 92 % oil, indicating the absence of long-range translational crystalline order in these systems.</p><!><p>The next interesting question that is raised by these results is whether tubule formation requires liquid triglyceride oil. In a way, this question has been answered already in a previous study, in which it was shown that tubules are formed in a range of apolar fluids [15]. These results suggest that tubule formation would also occur in a fluid γ-oryzanol + β-sitosterol mixture. This idea can be tested by heating a binary γ-oryzanol + β-sitosterol solid mixture and seeing whether tubule formation occurs spontaneously when sufficient γ-oryzanol and/or β-sitosterol is in the liquid state creating mobility for the molecules.</p><!><p>SAXS data of melting binary 20:80, 40:60, 60:40, and 80:20 γ-oryzanol + β-sitosterol mixtures in the temperature range from top to bottom between 20, 40, 60, 78, 95, 112, 127, and 144 °C (for the bottom right figure the three highest temperatures are left out because they are above the melting temperature). Curves have been shifted vertically for convenience</p><!><p>Analysis of the DSC curves for these samples during the first heating suggest that 2–15 % of the β-sitosterol + γ-oryzanol mixture is liquid over the temperature range between 80 and 100 °C. This figure is, however, quite sensitive to the choice of the baseline. The discussed occurrence of a liquid phase is actually in conflict with the phase diagram depicted in either Fig. 3 or 5 because at temperatures below the solid–liquid two-phase regions no liquid should be present. A meaningful explanation for the presence of a "low-temperature" liquid phase is based on the assumption that the phase behavior is actually more complicated than depicted in Fig. 3 or 5. The occurrence of this liquid could be the result of melting of a metastable polymorphic form combined with the slow crystallization of the more stable polymorph that has been the subject of the DSC analysis that Figs. 3 and 5 are based on.</p><p>Although the transition ranges for the dilution and melting experiments do no match exactly, the qualitative agreement between the ternary system and the system in melt transition is sufficient to argue that the liquid mixtures of β-sitosterol and γ-oryzanol function similarly to liquid triglyceride oils in promoting tubule formation.</p><!><p>The present paper investigated the phase behavior of binary and ternary mixtures of γ-oryzanol + β-sitosterol with and without sunflower oil. It complements the earlier study by AlHasawi and Rogers [14], which investigated a larger area of the phase diagram, in the sense that the present study takes a deeper look into a few specific changes in these systems and does not consider the full phase diagram of the ternary system. Overall, both studies are in agreement.</p><p>The melting behavior of the binary mixtures indicates the presence of two eutectic points and shows a maximum dissolution temperature. A phase diagram describing the melting points was derived on the basis of the DSC data. However, taking into account the DSC data together with the SAXS data affords a more complex phase diagram that is in agreement with the data presented by AlHasawi and Rogers [14] and basic thermodynamical phase rules. The suggested phase diagram contains four distinct mixed solid phases that are separated by three two-phase regions located around 5, 30, and 70 % (w/w) β-sitosterol. Next to the two eutectic points identified earlier the complex phase diagram suggests the presence of a peritectic point. It has to be acknowledged though that the SAXS data and melting points (DSC) are gathered at significantly different temperatures. Furthermore SAXS shows that the organization of γ-oryzanol and β-sitosterol in the intermediate composition range is different from the structure of the tubules found in oil-diluted samples.</p><p>The phase behavior of the ternary mixtures does not indicate a sudden transition from binary compound to tubule. Rather, a gradual transition seems to occur as more and more liquid oil becomes available on dilution of the γ-oryzanol + β-sitosterol mixtures with edible oil.</p><p>Finally, it seems that this behavior can also be observed in binary mixtures of γ-oryzanol and β-sitosterol. Once a mixed liquid phase is present, as for example on the disintegration of mixed metastable polymorph, the same behavior as in oil-diluted systems is found. This finding is important because it suggests that in the organogel systems the sterol (ester) molecules are dynamically transferring from tubules to solution and vice versa.</p>

PubMed Open Access

Cytoskeleton Dynamics in Drug-treated Platelets

Platelet activation is a key process in blood clot formation. During activation, platelets go through both chemical and physical changes, including secretion of chemical messengers and cellular shape change. Platelet shape change is mediated by the two major cytoskeletal elements in platelets, the actin matrix and microtubule ring. Most studies to date have evaluated these structures qualitatively, whereas this paper aims to provide a quantitative method of examining changes in these structures by fluorescently labeling the element of interest and performing single cell image analysis. The method described herein tracks the diameter of the microtubule ring and the circumference of the actin matrix as they change over time. Platelets were incubated with a series of drugs that interact with tubulin or actin, and the platelets were observed for variation in shape change dynamics throughout the activation process. Differences in shape change mechanics due to drug incubation were observable in each case.

cytoskeleton_dynamics_in_drug-treated_platelets

Introduction<!>Platelet Isolation<!>Immunofluorescence Imaging<!>Drug Treatment<!>Image Analysis<!>Results and Discussion<!>Image Processing<!>No Treatment<!>DMSO<!>Paclitaxel<!>Vincristine<!>Image Processing<!>No Treatment<!>DMSO<!>Latrunculin A<!>Cytochalasin D<!>Conclusions<!>

<p>Platelets play a diverse set of roles in the body. The most well-characterized aspects of platelets are the roles they play in hemostasis and thrombosis, but they have also been implicated in processes such as inflammation and the migration of cancer cells [1-3].</p><p>In vessel injury, endothelial cells expose a variety of adhesion molecules and secrete small molecules. When platelets encounter the site of injury, the exposed adhesion molecules bind to the platelets and impede their travel. The binding of the adhesion molecules and the small molecules secreted by the endothelial cells initiate an activation cascade in the platelets that then leads to clot formation. Two main processes characterize platelet activation: (1) the secretion of small molecules and proteins and (2) a major cytoskeleton-mediated shape change [4,5]. Small molecule secretion functions to propagate the activation signal and initiate the wound healing process by other cells, and there has recently been significant progress in characterizing this secretion process [6-11]. Preliminary work has studied the accompanying shape change, wherein platelets undergo a major cytoskeletal rearrangement where the cell body swells up and then flattens out to form extensions called lamellipodia and filopodia, but the timeline of shape change has not been quantitatively evaluated [12]. The two main components that make up the platelet cytoskeleton are the microtubule ring and the actin matrix. Dissimilar to most cells, the microtubules in platelets form a circumferential loop at the outer edge of the platelet while the actin matrix is spread throughout the platelet [13, 14]. Our goal herein is to develop a way to make a direct connection between platelet secretion and morphological change, allowing both fundamental insight into platelet biology and critical studies about drug or disease effects on blood platelets. With the methods developed, this work demonstrates that as platelets in suspension undergo activation, both the actin matrix and microtubule ring decrease in size. As activation progresses, the actin matrix reaches a stable size whereas the microtubule ring shrinks to a certain extent and then breaks up into small microtubule fragments. It is possible to then compare the results from the imaging data with dynamic secretion measurements and obtain information by correlating the two. The results presented herein, when compared with data previously obtained, verify that the actin matrix acts as a barrier to dense-body granule secretion and that the microtubule ring is not involved in dense-body granule secretion [7].</p><!><p>To isolate platelets, approximately 10-15 mL of rabbit blood was drawn from the midear artery of a rabbit after sedation according to IACUC protocol # 1311-31082A. The blood was centrifuged at 500 rcf with a brake speed of 0 for 15 minutes, at which point the supernatant, platelet rich plasma (PRP), was transferred to a clean centrifuge tube. The PRP was mixed with an equal volume of acid citrate dextrose solution (ACD; 85 mM trisodium citrate dihydrate, 66.6 mM citric acid monohydrate, 111 mM D-glucose) to prevent clotting during the platelet isolation. The PRP was then centrifuged at 750 rcf for 9 minutes to pellet the platelets, and the supernatant was removed. Next, the platelet pellet was resuspended in Tyrodes buffer (137 mM NaCl, 2.6 mM KCl, 1 mM MgCl2•6H2O, 5.55 mM D-glucose, 5 mM HEPES, 12.1 mM NaHCO3) and PGI2 (0.5 μM). To ensure that the platelets had time for recovery, they were not used until 1 hour after isolation. Visual inspection of the platelets upon resuspension was performed to detect any morphological changes in the platelets, indicating activation. The platelet cell count was determined using a hemocytometer, with a typical isolated platelet concentration between 1-2 × 108 platelets/ml. The actin and microtubule experiments were performed on different days, and platelet preparations resulted in a lower concentration of platelets for the microtubule conditions.</p><!><p>Immmunofluorescence imaging was performed on fixed platelets. Initially, platelets, at a concentration of 1 × 107 platelets/mL, were activated using human thrombin (5 units/mL, Sigma-Aldrich) at room temperature. Aliquots from the activated PRP were removed at 50 s intervals and activation was quenched by addition of the PRP to 8% formaldehyde in Tyrodes buffer [15]. Fixation was allowed to proceed for 20 minutes in the same 8% formaldehyde solution, after which the platelets were pelleted by centrifuge for 5 minutes at 2500 rcf. The fixative solution was removed, and the platelets were resuspended in 0.1% Triton X-100 in Tyrodes buffer (Sigma Aldrich) containing 0.1 mM EGTA (Sigma Aldrich). After 10 minutes of permeabilization, the platelets were again washed via pelleting, supernatant removal, and resuspension. Next, the platelets were incubated in a 1% BSA solution in Tyrodes buffer for 30 minutes to block nonspecific antibody binding. After another wash step, the platelets were incubated with either a Cy3-conjugated anti-β-tubulin-antibody (Abcam, ab11309) to label the microtubule ring, with the antibody diluted 1:100 in 1% BSA in Tyrodes, or a FITC-conjugated anti-β-actin-antibody (Abcam ab11005) to label the actin matrix, with the antibody diluted 1:250 in 1% BSA in Tyrodes buffer. The antibody incubation was performed overnight at 4°C. Finally, the platelets were washed again and allowed to settle onto poly-L-lysine-coated coverslips (1 μg/mL, 0.1% w/v poly-L-lysine in H2O, Sigma Aldrich) for imaging.</p><p>The fixed and labeled platelets were imaged using a Nikon Eclipse TE2000-U microscope and a Photometrics QuantEM:512SC camera. A 100x 1.40 NA oil immersion objective was used to obtain sufficient magnification of the platelets to enable clear visualization of the microtubule ring.</p><!><p>Drug-treated platelets were incubated with cytochalasin D, latrunculin A, vincristine, or paclitaxel (Sigma Aldrich) at a concentration of 10 μM for 45 min at room temperature prior to activation. The concentration used was chosen based on previous work examining the platelet cytoskeleton [7, 16]. As cytochalasin D, latrunculin A, and paclitaxel are insoluble in water, they were first dissolved in DMSO. As a control, platelets were also incubated with an equivalent DMSO concentration prior to activation to account for any effects DMSO itself might have on platelets [17].</p><!><p>For each condition, approximately 20 82 μm-by-82 μm fluorescence images were recorded, with several platelets visible in each image. Sample shape change progressions are shown in Figures 1 and 3 for the microtubule ring and actin matrix, respectively (For drug treatment progressions see Electronic Supplemental Material Fig. S1 –S6). For the microtubule ring-stained platelets, the diameter of each microtubule ring was measured at three locations to account for any non-circular character of the microtubule ring (Figure 2A). For the actin-treated platelets, the circumference of the actin matrix was measured. An ellipse was drawn inside the fluorescently labeled actin matrix, touching but not exceeding the edges (Figure 4A). For each condition, 25 platelets were measured. By determining the average measured value for each time point, a plot was created showing the change in diameter or circumference over time. Each plot was fit using a one-phase decay curve. Statistical analysis was performed using Graphpad Prism. One-way ANOVA was used to compare the shape change as a function of time. A comparison of fits was also used to determine significance between the various treatments. Any p < 0.05 was considered significant.</p><!><p>There are several pharmacological agents with modes of action based on cytoskeletal elements; these drugs present a perfect platform to prove the utility of these platelet imaging analyses. To disrupt microtubule dynamics, platelets were incubated with either vincristine or paclitaxel. At the concentrations used, paclitaxel is known to stabilize microtubules, and vincristine is known to cause destabilization of microtubules [18]. The same procedure was applied to examine the role of the actin matrix in platelet secretion wherein platelets were incubated with either cytochalasin D or latrunculin A, both of which are known to inhibit polymerization of actin filaments. The modes of inhibition vary between cytochalasin D and latrunculin A, where the former binds to the filament and prevents addition of monomers while the latter binds to the monomer to prevent addition to the filament [19, 20].</p><!><p>The labeled microtubules exhibited well-defined fluorescence, clearly showing that the labeled structures formed a ring. The rings were not completely circular, many being ellipsoidal in shape. To get an accurate measurement of the size change, the diameter of the microtubule ring was measured three times, where the longest, shortest, and an intermediate diameter were chosen (Figure 2A). In addition, platelets that appeared to have settled on the coverslip at an angle were not measured, as their dimensions were skewed, appearing thin and long.</p><!><p>Untreated (control) platelets showed a change in microtubule ring diameter from 3.3 ± 0.1 μm to 1.7 ± 0.1 μm during the 240 s time course following activation with thrombin. The primary diameter change occurred over the first 40 s of activation (p < 0.0001); while the diameter appears to decrease over the remaining 150 s, the changes are not statistically significant (Figure 2B).</p><!><p>A control was also performed using DMSO, a necessary reagent to dissolve the paclitaxel. The diameter of the DMSO-treated platelets started and ended at 3.4 ± 0.1 μm and 1.7 ± 0.1 μm, respectively (Figure 2B). These diameters were not statistically different from those measured from the non-treated platelets (p > 0.05). However, during activation, the DMSO-treated platelets exhibited a more start-stop approach to microtubule shrinkage, where the diameters had statistically significant decreases between 0 and 40 s (p < 0.0001) and 90 and 140 s (p < 0.05). This was further exemplified by the fact that one phase exponential fit tested via ANOVA did not fit both data sets (p < 0.05).</p><!><p>The paclitaxel-treated platelets were statistically compared to the DMSO-treated platelets as the paclitaxel solutions were made up in DMSO. The paclitaxel-treated platelets started off with a smaller ring diameter, 3.1 ± 0.1 μm, than the DMSO-treated platelets. After activation, the paclitaxel-treated platelets initially exhibited a slower decrease in ring diameter compared to the DMSO-treated platelets (p < 0.001). Around 140 s however, the shrinkage dynamics of the microtubule ring in the paclitaxel-treated platelets became greater in magnitude than those of the DMSO-treated platelets, resulting in a final diameter of 1.5 ± 0.0 μm (Figure 2C). Similar to the DMSO-treated platelets, the primary shape change occurred between 0 and 40 s (p < 0.0001) and 90 and 140 s (p < 0.05) of activation.</p><!><p>Treating platelets with vincristine resulted in the destruction of the microtubule ring dynamics. These platelets had microtubule rings that started off with an average diameter of 2.3 ± 0.1 μm and ended at a diameter of 2.5 ± 0.1 μm. The diameter fluctuated throughout the 240 s with a low of 2.2 ± 0.0 μm at 90 s and a high of 2.5 ± 0.1 μm at 240 s.</p><!><p>Labeling of the actin matrix is not as straightforward as labeling the microtubule ring because the cytoskeletal element is not as well defined. The actin-based fluorescence images showed diffuse fluorescence that appeared throughout the platelet. There were some structures visible in the form of dark spots within the fluorescent area, possibly resulting from granules, but they were uncommon and thus left uncharacterized (Figure 5). Like the microtubule rings, the fluorescent structure was not uniformly circular in shape and so it was necessary to determine the best measurement to characterize the structure. The methods tried include: fitting a circle or ellipse to the exterior or interior of the actin matrix or tracing the edges of each platelet. While tracing the edges initially appeared to be the best method, it became apparent that the time required to accurately trace the edge of each actin matrix was substantial. In addition, the edges were not always clear in the images, making this method somewhat subjective. More of the images exhibited platelets with an elliptical shape than a round shape, and so using the ellipse to approximate the platelet size was found to be most efficient and effective. When comparing measurements made on the exterior or interior of the fluorescent area, the interior measurements visually resulted in a closer fit to the true circumference than the exterior measurements. Thus a best-fit ellipse was used to approximate the actin matrix circumference, though this may result in a slight underestimation of total actin coverage.</p><!><p>The platelets that were not subjected to drug treatment showed a change in actin matrix circumference from 10.9 ± 0.3 μm to 9.1 ± 0.2 μm over the 240 s imaged (Figure 4B). The key characteristic found for actin matrix shape change was that the majority of the shrinkage occurs within the first 40 s after platelet activation. While the size did not appear to be completely static during the later time points, the differences in the circumference were not statistically significant (p > 0.05).</p><!><p>Similar to the microtubule disruption experiments, a control set of platelets were treated with DMSO as both latrunculin A and cytochalasin D are insoluble in water. The DMSO-treated platelets exhibited a more gradual decrease in circumference compared to the untreated platelets, going from 9.5 ± 0.4 μm to 8.2 ± 0.2 μm during the time monitored (Figure 4B). The decrease in size occurred over the first 90 s rather than the first 40 s in the untreated platelets (p ≤ 0.05). Similarly however to the untreated platelets, later changes were not statistically significant (p > 0.05). It is also important to note that, in addition to the change in time course, the DMSO-treated platelets started and ended up with a smaller circumference than the untreated platelets, despite the low DMSO concentrations used, indicating that the DMSO does influence the normal actin dynamics.</p><!><p>The latrunculin A-treated platelets started off the same size as the DMSO-treated platelets but did not change size in a statistically significant manner at any point during activation (p > 0.05). The circumference varied slightly, within a range in 0.5 μm where they were at their smallest at the 0 s time point and their largest during the 90 s time point, going from 9.6 ± 0.2 μm to 9.9 ± 0.2 μm (Figure 4C).</p><!><p>The decrease in size of the cytochalasin D-treated platelets occurred faster than that of the DMSO-treated platelets, where they reached a stable size within the first 40 s of activation. In addition, the cytochalasin D-treated platelet circumferences started off 1.5 μm larger than the DMSO-treated platelets and ended 0.5 μm larger, going from 11.0 ± 0.3 μm to 9.1 ± 0.2 μm, despite the DMSO-treated platelets being larger at the 40 s time point (Figure 4D).</p><p>Based on the fact that there are some significant changes in platelet cytoskeletal elements upon drug treatment, we considered correlation between these changes and platelet secretion of chemical messenger species. To this end, the microtubule ring and actin matrix dynamics were compared to previously published data showing the release of serotonin from platelets incubated with the same drugs at the same doses and incubation times. Previous results showed that when the platelets were incubated with the microtubule destabilizing drugs, the serotonin secretion from dense-body granules was not affected. However, when the actin matrix was disrupted, serotonin secretion was affected. With this previously published information in concert with the two data sets presented herein, we can see that the microtubule ring is not involved in the dense body granule release process despite the changes that were observed during activation [7]. Perhaps the microtubule ring behavior is more closely associated with alpha granule release (not assessed here) or shrinks to the inside of the platelet to minimize interactions with the granules during activation [21]. Future work will explore these hypotheses by performing simultaneous imaging of the alpha granules and the cytoskeletal elements.</p><!><p>Here, the dynamic changes of the cytoskeleton have been quantitatively tracked. Changes in the microtubule ring were tracked during the activation process by measuring the diameter of the ring at various time points. The images show that during activation, the microtubule ring first exhibits a decrease in diameter with retention of its circular or elliptical shape. However as activation proceeds, the microtubule ring begins to break apart into distinct pieces. The dynamics of the actin matrix are more difficult to quantify due to the more diffuse and abstract feature shape; however, changes in the actin matrix were effectively tracked by placing a best fit ellipse into the fluorescent area representing the labeled actin to measure the circumference. Like the microtubule ring, the actin matrix first decreases in size. After the initial decrease in circumference, at about 40 s post-activation the actin matrix reaches a steady circumference that remains throughout the rest of the activation process.</p><p>Future studies will involve tracking the cytoskeleton dynamics of live rather than fixed cells and a further exploration of the dynamics within the first 40 s of activation, where it became apparent through this work that the majority of cytoskeletal changes are occurring. Also, various microscopy techniques will be applied in an effort to increase the spatial resolution so that features within the actin matrix are more visible. Through live cell imaging techniques with increased resolution it will be possible to further elucidate the intertwining roles of the platelet cytoskeleton and granule release.</p><!><p>Microtubule ring – No treatment. Timeline of microtubule ring shrinkage dynamics over the course of 240 s following activation. The ring structure of the microtubules can be seen initially but becomes less well-defined as activation progresses. Scale bar 10 μm.</p><p>A) The microtubule ring of each platelet was measured by taking the diameter of the ring at the smallest, largest, and intermediate lengths. B-D) Comparison of the microtubule ring shrinkage curves for non-treated and DMSO-treated platelets, DMSO-and paclitaxel-, and DMSO- and vincristine-treated platelets, respectively. Error bars indicate standard error of the mean (SEM).</p><p>Actin matrix – No treatment. Timeline of actin matrix shrinkage dynamics over the course of 240 s following activation. Scale bar 10 μm.</p><p>A) The actin matrix of each platelet was measured by using the best fit ellipse to measure the circumference. B-D) Comparison of the actin matrix shrinkage curves for non-treated and DMSO-treated platelets, DMSO- and latrunculin A-, and DMSO- and cytochalasin D-treated platelets, respectively. Error bars indicate SEM.</p><p>Dark spots seen throughout the platelet during actin matrix labeling may be due to granules within the platelet. Platelets with these features exhibited one or more spots.</p>

PubMed Author Manuscript

Chemometric Methods for Spectroscopy-Based Pharmaceutical Analysis

Spectroscopy is widely used to characterize pharmaceutical products or processes, especially due to its desirable characteristics of being rapid, cheap, non-invasive/non-destructive and applicable both off-line and in-/at-/on-line. Spectroscopic techniques produce profiles containing a high amount of information, which can profitably be exploited through the use of multivariate mathematic and statistic (chemometric) techniques. The present paper aims at providing a brief overview of the different chemometric approaches applicable in the context of spectroscopy-based pharmaceutical analysis, discussing both the unsupervised exploration of the collected data and the possibility of building predictive models for both quantitative (calibration) and qualitative (classification) responses.

chemometric_methods_for_spectroscopy-based_pharmaceutical_analysis

Introduction<!>Chemometrics as tool for fraud/adulteration detection<!>Exploratory data analysis<!><!>Exploratory data analysis<!>Principal component analysis<!><!>Principal component analysis<!><!>Principal component analysis<!>Selected examples<!>Regression<!>Partial least squares (PLS) regression<!>Selected application of regression methods to pharmaceutical problems<!>Classification<!><!>Classification<!>Discriminant methods<!>Partial least squares discriminant analysis (PLS-DA)<!>Class-modeling methods<!>Soft independent modeling of class analogies (SIMCA)<!>Selected applications of classification approaches for pharmaceutical analysis<!>Validation<!>Other selected applications<!>Conclusions<!>Author contributions<!>Conflict of interest statement

<p>Quality control on pharmaceutical products is undoubtedly an important and widely debated topic. Hence, in literature, various methods have been proposed to check quality of medicines, either qualitative (e.g., for the identification of an active pharmaceutical ingredient, API; Blanco et al., 2000; Herkert et al., 2001; Alvarenga et al., 2008) or quantitative (quantification of the API; Blanco et al., 2000; Yao et al., 2007; Cruz Sarraguça and Almeida Lopes, 2009); involving either destructive or non-invasive online techniques. Recently, due to the benefits they bring, several non-destructive methodologies based on spectroscopic techniques (mainly Near-Infrared NIR) combined with chemometric tools have been proposed for pharmaceutical quality check (Chen et al., 2018; Rodionova et al., 2018).</p><p>Despite the development of analytical methodologies and the commitments of national and supranational entities to regulate pharmaceutical quality control, substandard and counterfeit medicines are still a major problem all over the world.</p><!><p>Poor-quality pharmaceuticals can be found on the market for two main reasons: low production standards (mainly leading to substandard medicines) and fraud attempts. Counterfeited drugs may present different frauds/adulterations; for instance, they could contain no active pharmaceutical ingredient (API), a different API from the one declared, or a different (lower) API strength. As mentioned above, several methodologies have been proposed in order to detect substandard/counterfeit pharmaceuticals; among these, a major role is played by those based on the application of spectroscopic techniques in combination with different chemometric methods. The relevance of these methodologies is due to the fact that spectroscopy (in particular, NIR) combined with exploratory data analysis, classification and regression method can lead to effective, high performing, fast, non-destructive, and sometimes, online methods for checking the quality of pharmaceuticals and their compliance to production and/or pharmacopeia standards. Nevertheless, the available chemometric tools applicable to handle spectroscopic (but, of course not only those) data are numerous, and there is plenty of room for their misapplication (Kjeldahl and Bro, 2010). As a consequence, the aim of the present paper is to report and critically discuss some of the chemometric methods typically applied for pharmaceutical analysis, together with an essential description of the figures of merit which allow evaluating the quality of the corresponding models.</p><!><p>In the large part of the studies for the characterization of pharmaceutical samples for quality control, verification of compliance and identification/detection of counterfeit, fraud or adulterations, experimental signals (usually in the form of some sorts of fingerprints) are collected on a series of specimens. These constitute the data the chemometric models operate on. These data are usually arranged in the form of a matrix X, having as many rows as the number of samples and as many columns as the number of measured variables. Accordingly, assuming that samples are spectroscopically characterized by collecting an absorption (or reflection/transmission) profile (e.g., in the infrared region), each row of the matrix corresponds to the whole spectrum of a particular sample, whereas each column represents the absorbance (or reflectance/transmittance) of all the individuals at a particular wavenumber. This equivalence between the experimental profiles and their matrix representation is graphically reported in Figure 1.</p><!><p>Graphical illustration of the equivalence between the collected experimental data (in this case, NIR spectra for 6 samples) and the data matrix X. Each row of the data matrix corresponds to the spectrum of a sample, whereas each column contains the value of a specific variable over all the individuals.</p><!><p>Once the data have been collected, exploratory data analysis represents the first step of any chemometric processing, as it allows "to summarize the main characteristics of data in an easy-to-understand form, often with visual graphs, without using a statistical model or having formulated a hypothesis" (Tukey, 1977). Exploratory data analysis provides an overall view of the system under study, allowing to catch possible similarities/dissimilarities among samples, to identify the presence of clusters or, in general, systematic trends, to discover which variables are relevant to describe the system and, on the other hand, which could be in principle discarded, and to detect possible outlying, anomalous or, at least, suspicious samples (if present). As evident also from the definition reported above, in the context of exploratory data analysis a key role is played by the possibility of capturing the main structure of the data in a series of representative plots, through appropriate display techniques. Indeed, considering a general data matrix X, of dimensions N × M, one could think of its entries as the coordinates of N points (the samples) into a M-dimensional space whose axes are the variables, which makes this representation unfeasible for the cases when more than three descriptors are collected on each individual. This is why exploratory data analysis often relies on the use of projection (bilinear) techniques to reduce the data dimensionality in a "clever" way. Projection methods look for a low-dimensional representation of the data, whose axes (normally deemed components or latent variables) are as relevant as possible for the specific task. In the case of exploratory data analysis, the most commonly used technique is Principal Components Analysis (PCA) (Pearson, 1901; Wold et al., 1987; Jolliffe, 2002).</p><!><p>Principal component analysis (PCA) is a projection method, which looks for directions in the multivariate space progressively providing the best fit of the data distribution, i.e., which best approximate the data in a least squares sense. This explains why PCA is the technique of choice in the majority of cases when exploratory data analysis is the task: indeed, by definition, for any desired number of dimensions (components) F in the final representation, the subspace identified by PCA constitutes the most faithful F-dimensional approximation of the original data. This allows compression of the data dimensionality at the same time reducing to a minimum the loss of information. In particular, starting from a data matrix X(N × M), Principal Component Analysis is based on its bilinear decomposition, which can be mathematically described by Equation (1):</p><p>The loadings matrix P(M × F) identifies the F directions, i.e., the principal components (PC), along which the data should be projected and the results of such projection, i.e., the coordinates of the samples onto this reduced subspace, are collected in the scores matrix T(N × F). In order to achieve data compression, usually F≪M so that the PCA representation provides an approximation of the original data whose residuals are collected in the matrix E(N × M).</p><p>Since the scores represent a new set of coordinates along highly informative (relevant) directions, they may be used in two- or three-dimensional scatterplots (scores plots). This offers a straightforward visualization of the data, which can highlight possible trends in data, presence of clusters or, in general, of an underlying structure. A schematic representation of how PCA works is displayed in Figure 2.</p><!><p>Graphical illustration of the basics of PCA. The samples, here represented in a three-dimensional space, are projected onto a low-dimensional subspace (highlighted in light red in the leftmost panel) spanned by the first two principal components. Inspection of the data set can be carried out by looking at the distribution of the samples onto the informative PC subspace (scores plot) and interpretation can be then carried out by examining the relative contribution of the experimental variable to the definition of the principal components (loadings plot).</p><!><p>Figure 2 shows one of the simplest possible examples of feature reduction, since it describes the case where samples described by three measured variables can be approximated by being projected on an appropriately chosen two-dimensional sub-space. However, the concept may be easily generalized to higher-dimensional problems, such as those involving spectroscopic measurements. Figure 3 shows an example of the application of PCA to mid infrared spectroscopic data. In particular, the possibility of extracting as much information as possible from the IR spectra recorded on 51 tablets containing either ketoprofen or ibuprofen in the region 2,000–680 cm−1 (661 variables) is represented.</p><!><p>Graphical illustration of the application of PCA on a spectral (mid-infrared) data. Fifty-one spectra recorded on samples containing either ibuprofen (blue) or ketoprofen (red) are recorded in the region 680–2,000 cm−1 (A). When PCA is applied to such a dataset, one obtains a scores plot (B) showing that two cluster of samples, corresponding to tablets containing ibuprofen (blue squares) or ketoprofen (red circles) are separated along the first component. Interpretation of the observed differences in terms of the spectroscopic signal is made possible by the inspection of the loadings on PC, which are shown in a "spectral-like" fashion in (C).</p><!><p>A large portion of the data variability can be summarized by projecting the samples onto the space spanned by the first two principal components, which account for about 90% of the original variance, and therefore can be considered as a good approximation of the experimental matrix. Inspection of the scores plot suggests that the main source of variability is the difference between ibuprofen tablets (blue squares) and ketoprofen ones (red circles), since the two clusters are completely separated along the first principal component. To interpret the observed cluster structure in terms of the measured variables, it is then necessary to inspect the corresponding loadings, which are also displayed in Figure 3 for PC1. Indeed, for spectral data, the possibility of plotting the loadings for the individual components in a profile-like fashion, rather than producing scatterplot for pairs of latent variables (as exemplified in Figure 2) is often preferred, due to its more straightforward interpretability: spectral regions having positive loadings will have higher intensity on samples which have positive scores on the corresponding component, whereas bands associated to negative loadings will present higher intensity on the individuals falling at negative values of the PC. In the example reported in Figure 3, one could infer, for instance, that the ketoprofen samples (which fall at positive values of PC1) have a higher absorbance at the wavenumbers where the loadings are positive, whereas ibuprofen samples should present a higher signal in correspondence to the bands showing negative loadings.</p><p>Based on what reported above, it is evident how the quality of the compressed representation in the PC space depends on the number of components F chosen to describe the data. However, at the same time, it must be noted that when the aim of calculating a PCA is "only" data display, as in most of the applications in the context of exploratory analysis, the choice of the optimal number of components is not critical: it is normally enough to inspect the data distribution across the first few dimensions and, in many cases, considering the scores plot resulting from the first two or three components could be sufficient. On the other hand, there may be cases when the aim of the exploratory analysis is not limited to just data visualization and, for instance, one is interested in the identification of anomalous or outlying observations, or there could be the need of the imputation of missing elements in the data matrix; additionally, one could also need to obtain a compressed representation of the data to be used for further predictive modeling. In all such cases, the choice of the optimal dimensionality of the PC representation is critical for the specific purposes and, therefore, the number of PCs should be carefully estimated. In this respect, different methods have been proposed in the literature and a survey of the most commonly used can be found in Jolliffe (2002).</p><p>Among the applications described above, the possibility of using PCA for the identification/detection of potential outliers deserves a few more words, as it could be of interest for pharmaceutical quality control. Actually, although outliers—or anomalous observations, in general—could be, in principle, investigated by visually inspecting the scores plot along the first components, this approach could be subjective and anyway would not consider some possible data discrepancies. Alternatively, when it is used as a model to build a suitable approximation of the data, PCA provides a powerful toolbox for outlier detection based on the definition of more objective test statistics, which can be easily automatized or, anyway, embedded in control strategies, also on-line. This is accomplished by defining two distance measurements: (i) a squared Mahalanobis distance in the scores space, which follows the T2 statistics (Hotelling, 1931) and accounts for how extreme the measurement is in the principal component subspace, and (ii) a squared orthogonal Euclidean distance (the sum of squares of the residuals after approximating the observation by its projection), which is normally indicated as Q statistics (Jackson and Muldholkar, 1979) and quantifies how well the model fits that particular individual. Outlier detection is then carried out by setting appropriate threshold values for the T2 and Q statistics and verifying whether the samples fall below or above those critical limits. Moreover, once an observation is identified as a potential outlier, inspection of the contribution plot can help in relating the detected anomaly to the behavior of specific measured variables.</p><!><p>PCA is customarily used for the quality control of drugs and pharmaceuticals; several examples of the application of this technique to solve diverse issues have been reported in the literature. One of the most obviously relevant ones is fraud detection. For example, in Rodionova et al. (2005) PCA was applied to both bulk NIR spectroscopy and hyperspectral imaging (HSI) in the NIR range to spot counterfeit drugs. In particular, bulk NIR was used to differentiate genuine antispasmodic drugs from forgeries, whereas HSI on the ground uncoated tablets was employed to identify counterfeited antimicrobial drugs. In both cases, the spectroscopic data were subjected to PCA, which allowed to clearly identify clusters in the scores plot, corresponding to the two kinds of tablets, i.e., genuine and counterfeited. In the case of the imaging platform, where the signal is stored as a data hypercube [i.e., a three-way numerical array of dimension number of horizontal pixels Nx, number of vertical pixels Ny and number of wavelengths Nλ, in which each entry corresponds to the spectral intensity measured at a certain wavelength and a specific spatial position (x-y coordinates)], a preliminary unfolding step is needed. Unfolding is the procedure allowing to reorganize a higher-order array into a two-way matrix, which can be then processed with standard chemometric techniques. In the case of hyperspectral data cubes, this is carried out by stacking the spectra corresponding to the different pixels one on top of each other, in a way to obtain a matrix of dimensions (Nx × Ny and Nλ).</p><p>Another relevant application of exploratory analysis is related to quality check. For instance, PCA can be applied to investigate formulations not meeting predefined parameters. In Roggo et al. (2005), PCA was used to inquire a suspicious blue spot present on tablets. Samples were analyzed by a multi-spectral (IR) imaging microscope and PCA analysis was performed on the unfolded data-cube, indicating that the localized coloration was not due to contamination, but actually given by wet indigo carmine dye and placebo (expected ingredients of the formulation).</p><p>PCA can also be used for routine quality checks at the end of a production process. For example, in Myakalwar et al. (2011) laser-induced breakdown spectroscopy (LIBS) and PCA were combined with the aim of obtaining qualitative information about the composition of different pharmaceuticals.</p><!><p>As discussed in the previous section, exploratory analysis is a first and fundamental step in chemometric data processing and, in some cases, it could be the only approach needed to characterize the samples under investigation. However, due to its unsupervised nature, it provides only a (hopefully) unbiased picture of the data distribution but it lacks any possibility of formulating predictions on new observations, which on the other hand may be a fundamental aspect to solve specific issues. In practice, very often quality control and/or authentication of pharmaceutical products rely on some forms of qualitative or quantitative predictions. For instance, the quantification of a specific compound (e.g., an active ingredient or an excipient) contained in a formulation is a routine operation in pharmaceutical laboratories. This goal can be achieved by combining instrumental (e.g. spectroscopic) measurements with chemometric regression approaches (Martens and Naes, 1991; Martens and Geladi, 2004). Indeed, given a response to be predicted y and a vector of measured signals (e.g., a spectrum) x, the aim of regression methods is to find the functional relationship that best approximates the response on the basis of the measurements (the predictors). Mathematically, this can be stated as:</p><p>where ŷ is the predicted response (i.e., the response value approximated by the model), f(x) indicates a general function of and x and e is the residual, i.e., the difference between the actual response and its predicted value. In many applications, the functional relationship between the response and the predictors f(x) can be assumed to be linear:</p><p>where x1, x2 … xM are the components of the vector of measurements x and the transpose indicates that it is normally expressed as a row vector, while the associated linear coefficients b1, b2 … bM, which weight the contributions of each of the M X-variables to y, are called regression coefficients and collected in the vector b. Building a regression model means to find the optimal value of the parameters b, i.e., the values which lead to the lowest error in the prediction of the responses. As a direct consequence of this consideration, it is obvious how it is mandatory to have a set of samples (the so-called training set) for which both the experimental data X and the responses y are available, in order to build a predictive model. Indeed, the information on the y is actively used to calculate the model parameters. When data from more than a single sample are available, the regression problem in Equations (2, 3) can be reformulated as:</p><p>where the vectors y^ and e collect the predictions and residuals for the different samples, respectively. Accordingly, the most straightforward way of calculating the model parameters in Equation (4) is by the ordinary least-squares approach, i.e., by looking at those values of b, which minimize the sum of squares of the residuals e:</p><p>ei being the residual for the ith sample and N being the number of training observations. The corresponding methods is called multiple linear regression (MLR) and, under the conditions of Equation (5), the regression coefficients are calculated as:</p><p>Here it is worth to highlight that, if one wishes to use the same experimental matrix X to predict more than one response, i.e., if, for each sample, instead of a single scalar yi, there is a dependent vector</p><p>L being the number of responses, then each dependent variable should be regressed on the independent block by means of a set of regression coefficients. Assuming that the L responses measured on the training samples are collected in a matrix Y, whose columns yl are the individual dependent variables,</p><p>the corresponding regression equations could be written as:</p><p>which can be grouped into a single expression:</p><p>where the residuals, i.e., the differences between the measured and predicted responses are collected in the matrix E, and the regression coefficients vectors are gathered in a matrix B, which can be estimated, analogously to Equation (6), as:</p><p>Equations (9–11) indicate that, as far as MLR is concerned, building a model to predict one response at a time or another model to predict multiple responses altogether would lead to the same results since, in the latter case, each dependent variable is anyway modeled as if it were alone. In either case, the solutions of the least-squares problem reported in Equations (6, 11) rely on the possibility of inverting the matrix (XTX), i.e., on the characteristics of the predictors. Indeed, in order for that matrix to be invertible, the number of samples should be higher than that of variables and the variables themselves should be as uncorrelated as possible. These conditions are rarely met by the techniques which are used to characterize pharmaceutical samples and, in particular, never met by spectroscopic methods. Due to these limitations, alternative approaches have been proposed in the literature to build regression models in cases where standard multiple linear regression is not applicable. In particular, since in order for the regression solution to exist, the predictor matrix should be made of few, uncorrelated variables, most of the alternative approaches proposed in the literature involve the projection of the X matrix onto a reduced space of orthogonal components and the use of the corresponding scores as regressors to predict the response(s). In this regard, one of the most widely used approaches is principal component regression (PCR) (Hotelling, 1957; Kendall, 1957; Massy, 1965; Jeffers, 1967; Jolliffe, 1982, 2002; Martens and Naes, 1991 Martens and Geladi, 2004) which, as the name suggests, involves a two-stage process where at first principal component analysis is used to compress the information in the X block onto a reduced set of relevant scores, as already described in Equation (1):</p><p>and then these scores constitute the predictor block to build a multiple linear regression:</p><p>C being the matrix of regression coefficients for this model. By combining Equations (12, 13), it can be easily seen how PCR still describes a linear relationship between the responses Y and the original variables X:</p><p>mediated by a matrix of regression coefficients BPCR(=PC), which is different from the one that would be estimated by Equation (11), since it is calculated by taking into account only the portion of the variability in the X block accounted for by the selected principal components. The use of principal component scores as predictors allows to solve the issues connected to the matrix (XTX) being usually ill-conditioned when dealing with spectroscopic techniques, but may be still suboptimal in terms of predictive accuracy.</p><p>Indeed, as described in Equations (12, 13), calculating a PCR model is a two-step procedure, which involves at first the calculation of PC scores and then the use of these scores to build a regression model to predict the response(s). However, these two steps have different objective functions, i.e., the criterion which is used to extract the scores from the X matrix is not the same which guides the calculation of the regression coefficients C in Equation (13). Stated in different words, the directions of maximum explained variance (especially when there are many uninformative sources of variability in the data) may not be relevant for the prediction of the Y. To overcome this drawback, an alternative approach to component-based regression is represented by the Partial Least-Squares algorithm (Wold et al., 1983; Geladi and Kowalski, 1986; Martens and Naes, 1991) which, due to its being probably the most widely used calibration method in chemometrics, will be described in greater detail in the following subparagraph.</p><!><p>Partial Least Squares (PLS) regression (Wold et al., 1983; Geladi and Kowalski, 1986; Martens and Naes, 1991) was proposed as an alternative method to calculate reliable regression models in the presence of ill-conditioned matrices. Analogously to PCR, it is based on the extraction of a set of scores T by projecting the X block on a subspace of latent variables, which are relevant for the calibration problem. However, unlike PCR, the need for the components not only to explain a significant portion of the X variance but also to be predictive for the response Y is explicitly taken into account for the definition of the scores. Indeed, in PLS, the latent variables (i.e., the directions onto which the data are projected) are defined in such a way to maximize the covariance between the corresponding scores and the response(s): maximizing the covariance allows to obtain scores which at the same time describe a relevant portion of the X variance and are correlated with the response(s). Due to these characteristics, and differently than what already described in the case of MLR (see Equation 11) and, by extension, PCR, in PLS two distinct algorithms have been proposed depending on whether only one or multiple responses should be predicted (the corresponding approaches are named PLS1 and PLS2, respectively). In the remainder of this section, both algorithms will be briefly described and commented.</p><p>When a single response has to be predicted, its values on the training samples are collected in a vector y; accordingly, the PLS1 algorithm extracts scores from the X block having maximum covariance with the response. In particular, the first score t1 is the projection of the data matrix X along the direction of maximum covariance r1:</p><p>While the successive scores t2 ⋯ tF, which are all orthogonal, account in turn for the maximum residual covariance. Therefore, PLS1 calculates a set of orthogonal scores having maximum covariance with y, according to:</p><p>R being the weights defining the subspace onto which the matrix should be projected, and then uses these scores as regressors for the response:</p><p>q being the coefficients for the regression. Similarly to what already shown in the case of PCR, Equations (16, 17) can be then combined in a single one to express the regression model as a function of the original variables, through the introduction of the regression vector bPLS1 (=Rq):</p><p>In contrast, in the multi-response case (PLS2), it is assumed that also the matrix Y, which collects the values of the dependent variables on the training samples, has a latent structure, i.e., it can be approximated by a component model:</p><p>U and Q being the Y scores and loadings, respectively. In particular, in order for the calibration model to be efficient, it is assumed that the X and the Y matrices share the same latent structure. This is accomplished by imposing that the component be relevant to describe the variance of the independent block and predictive for the responses. In mathematical terms, pairs of scores are simultaneously extracted from the X and the Y blocks so to have maximum covariance:</p><p>Where ti and ui are the X and the Y scores on the ith latent variable, respectively, qi is the ith column of the Y loading matrix Q while ri is the ith column of the X weight matrix R, which has the same meaning as specified in Equation (16). Additionally, these scores are made to be collinear, through what is normally defined as the inner relation:</p><p>ci being a proportionality constant (inner regression coefficient). When considering all the pairs of components, Equation (21) can be rewritten in a matrix form as:</p><p>where</p><p>Also in this case, by combining all the equations defining the model, it is possible to express the predicted responses as a linear function of the original variables:</p><p>where the matrix of regression coefficients BPLS2 is defined as RCQT.</p><p>Based on the above description, it is clear that, when more than one response has to be modeled, it is essential to decide whether it could be better to build an individual model for each dependent variable, or a single model to predict all the responses, as the results would not be the same. In particular, it is advisable to use the PLS2 approach only when one could reasonably assume that there are systematic relationships between the dependent variables.</p><p>On the other hand, independently on what model one decides to use, once the values of the regression coefficients (here generally indicated as B) have been estimated based on the training samples, they can be used to predict the responses for any new set of measurements (Xnew):</p><p>Here, it should be stressed that, in order for the calibrations built by PLS (but the same concept holds for PCR) to be accurate and reliable, a key parameter is the choice of an appropriate number of latent variables to describe the data. Indeed, while selecting a low number of components one can incur in the risk of not explaining all the relevant variance (underfitting), including too many of them (so that not only the systematic information is captured, but also the noise), can lead to overfitting, i.e., to a model which is very good in predicting the samples it has been calculated on, but performs poorly on new observations. To reduce this risk, a proper validation strategy is needed (see section Validation) and, in particular, the optimal number of latent variables is selected as the one leading to the minimum error during one of the validation stages (usually, cross-validation).</p><!><p>Regression methods in general, and especially PLS, are often combined with spectroscopy in order to develop rapid and (sometimes) non-destructive methodologies for the quantification of active ingredients in formulations. For instance, Bautista et al. (1996) quantified three analytes of interest (caffeine, acetylsalicylic acid and acetaminophen) in their synthetic ternary mixtures and different formulations by UV-Vis spectroscopy assisted by a PLS calibration model. Mazurek et al. proposed two approaches based on coupling FT-Raman spectroscopy with PLS and PCR calibration for estimation of captopril and prednisolone in tablets (Mazurek and Szostak, 2006a) and diclofenac sodium and aminophylline in injection solutions (Mazurek and Szostak, 2006b). The authors compared results obtained from calibration models built by using unnormalised spectra with the values found when an internal standard was added to each sample and the spectra were normalized by its selected band intensity at maximum or integrated. Another study on injection samples was proposed by Xie et al. (2010), using NIR spectroscopy combined with PLS and PCR to quantify pefloxacin mesylate (an antibacterial agent) in liquid formulations. PLS regression was also coupled to MIR (Marini et al., 2009) and NIR spectroscopy (Rigoni et al., 2014) to quantify the enantiometric excess of different APIs in the solid phase, also in the presence of excipients, based on the consideration that, in the solid phase, the spectrum of the racemic mixture could be different from that of either pure enantiomer. Specifically, it was possible to accurately quantify the enantiomeric excess of S-(+)-mandelic acid and S-(+)-ketoprofen by MIR spectroscopy coupled by PLS on the whole spectrum and after variable selection by sequential application of backward interval PLS and genetic algorithms (biPLS-GA) (Marini et al., 2009), while NIR was used to quantify the enantiomeric excess of R-(–)-epinephrine and S-(+)-ibuprofen (Rigoni et al., 2014). In the latter case, it was also shown that, when using the validated model to quantify the enantiomeric excess of API in the finished products, the influence of excipients and dosage forms (intact tablets or powders) has a relevant impact on the final predictive accuracy.</p><!><p>As already introduced in the previous section, in chemometric applications, in general, and in the context of pharmaceutical analysis, in particular, one is often interested in using the experimentally collected data (e.g., spectroscopic profiles) to predict qualitative or quantitative properties of the samples. While the regression methods for the prediction of quantitative responses have been already presented and discussed in section Regression, the main chemometric approaches for the prediction of qualitative properties of the individuals under investigation are outlined herein. These approaches are generally referred to as classification methods, since any discrete level that the qualitative variable can assume may also be defined as a class (or category) (Bevilacqua et al., 2013). For instance, if one were interested in the possibility of recognizing which of three specific sites a raw material was supplied from, it is clear that the response to be predicted could only take three possible values, namely "Site A," "Site B," and "Site C"; each of these three values would correspond to a particular class. A class can be then considered as an ensemble of individuals (samples) sharing similar characteristics. In this example, samples from the first class would all be characterized by having been manufactured from a raw material produced in Site A, and similar considerations could be made for the specimens in the second and third classes, corresponding to Site B and Site C, respectively. As it could already be clear from the example, there are many ambits of application for classification methods in pharmaceutical and biomedical analysis, some of which will be further illustrated in section Selected Applications of Classification Approaches for Pharmaceutical Analysis, after a brief theoretical introduction to the topic as well as the chemometric methods most frequently used in this context (especially, in combination with spectroscopic techniques).</p><p>As anticipated above, classification approaches aim at relating the experimental data collected on a sample to a discrete value of a property one wishes to predict. This same problem can be also expressed in geometrical terms by considering that each experimental profile (e.g., spectrum) can be seen as point in the multivariate space described by the measured variables. Accordingly, a classification problem can be formulated as the identification of regions in this multivariate space, which can be associated to a particular category, so that if a point falls in one of these regions, it is predicted as being part of the corresponding class. In this respect, classification approaches can be divided into two main sub-groups: discriminant and class-modeling methods. In this framework, a fundamental distinction can be made between discriminant and class-modeling tools, which constitute the two main approaches to perform classification in chemometrics (Albano et al., 1978). In detail, discriminant classification methods focus on identifying boundaries in the multivariate space, which separate the region(s) corresponding to a particular category from those corresponding to another one. This means they need representative samples from all the categories of interest in order to build the classification model, which will be then able to predict any new sample as belonging only to one of the classes spanned by the training set. In a problem involving three classes, a discriminant classification method will look for those boundaries in the multivariate space identifying the regions associated to the three categories in such a way as to minimize the classification error (i.e., the percentage of samples wrongly assigned). An example is reported in Figure 4A. On the other hand, class-modeling techniques look at the similarities among individuals belonging to the same category, and aim at defining a (usually bound) subspace where samples from the class under investigation can be found with a certain probability; in this sense, they resemble outlier tests, and indeed they borrow most of the machinery from the latter. Operationally, each category is modeled independently on the others and the outcome is the definition of a class boundary which should enclose the category sub-space:, i.e., individuals falling within that space are likely to belong to the class (are "accepted" by the class model), whereas samples falling outside are deemed as outliers and rejected. It is then evident that one of the main advantages of class modeling approaches is that they allow building a classification model also in the asymmetric case, where there is only a category of interest and the alternative one is represented by all the other individuals not falling under the definition of that particular class. In this case, since the alternative category is ill-defined, heterogeneous, and very likely to be underrepresented in the training set, any discriminant model would result suboptimal, as its predictions would strongly depend on the (usually not enough) samples available for that class. On the other hand, modeling techniques define the category space only on the basis of data collected for the class of interest, so those problems can be overcome.</p><!><p>Illustration of the difference between discriminant (A) and modeling (B) classification techniques. Discriminant classification techniques (A) divide the available hyperspace into as many regions as the number of the investigated categories (three, in the present example), so that whenever a sample falls in a particular region of space, it is always assigned to the associated class. Modeling techniques (B) build a separate model for each one of the categories of interest, so that there can be regions of spaces where more than a class is mapped and others where there is no class at all.</p><!><p>When the specific problem requires to investigate more than one class, each category is modeled independently on the others and, accordingly, the corresponding sub-spaces may overlap (see Figure 4B). As a consequence, classification outcomes are more versatile than with discriminant methods: a sample can be accepted by a single category model (and therefore be assigned to that class), by more than one (falling in the area where different class spaces overlap and, hence, resulting "confused") or it could fall outside any class-region and therefore be rejected by all the categories involved in the model.</p><!><p>As mentioned above, predictions made by the application of discriminant methods are univocal; namely, each sample is uniquely assigned to one and only one of the classes represented in the training set. This is accomplished by defining decision surfaces, which delimit the boundaries among the regions of space associated to the different categories. Depending on the model complexity, such boundaries can be linear (hyperplanes) or assume more complex (non-linear) shapes. When possible, linear discriminant models are preferred as they have less parameters to tune, require a lower number of training samples and are in general more robust against overfitting. Based on these considerations, the first-ever and still one of the most commonly used discriminant techniques is Linear Discriminant Analysis (LDA), originally proposed by Fisher (1936). It relies on the assumption that the samples of each class are normally distributed around their respective centroids with the same variance/covariance matrix (i.e., the same within-category scatter). Under these assumptions, it is possible to calculate the probability that each sample belongs to a particular class g p( g|x), as:</p><p>where x¯g is the centroid of class g, S the overall within-class variance/covariance matrix, πg the prior probability (i.e., the probability of observing a sample from that category before carrying out any measurement), C is a normalization constant and the argument of the exponential (x-x¯g)TS-1(x-x¯g) is defined as the squared Mahalanobis distance of the individual to the center of the category. Classification is then accomplished by assigning the sample to the category, to which it has the highest probability of belonging.</p><p>LDA is a well-established technique, which works well also on data for which the normality assumption is not fulfilled but, unfortunately, it can rarely be used on spectroscopic data for the same reasons MLR cannot be utilized for regression (see section Regression): calculation of matrix S−1 requires the experimental data matrix to be well-conditioned, which is not the case, when dealing with a high number of correlated variables measured on a limited number of samples. To overcome these limitations, LDA can be applied on the scores of bilinear models used to compress the data (e.g., on principal components), but the most commonly used approach involves a suitable modification of the PLS algorithm which makes it able to deal with classification issues; the resulting method is called partial least squares discriminant analysis (PLS-DA) (Sjöström et al., 1986; Ståle and Wold, 1987; Barker and Rayens, 2003), and it will be briefly described in the following paragraph.</p><!><p>In order for the PLS algorithm to deal with discriminant classification problems, the information about class belonging has to be encoded in a response variable Y, which can be then regressed onto the experimental matrix X to provide the predictive model (Sjöström et al., 1986). This is accomplished by defining Y as a "dummy" binary matrix, having as many rows as the number of samples (N) and as many columns as the number of classes (G). Each row in Y is a vector encoding the information about class belonging of the corresponding sample, whereas each column is associated to a particular class (the first column to class 1, the second to class 2 and so on up to the Gth). As such, the row vector corresponding to a particular sample will contain all zeros except for the column associated to the class it belongs to, where there will be a one. For instance, in the case of a problem involving three categories, a sample belonging to Class 2 will be represented by the vector yi = [0 1 0]. A PLS regression model is then calculated between the experimental data matrix X and the dummy Y [as described in section Partial Least Squares (PLS) Regression], and the matrix of regression coefficients obtained is used to predict the value of the responses on new samples, Y^new. Since the dependent variable is associated to the categorical information, classification of the samples is based on the predicted responses Y^new which, however, are not binary but real-valued. As a consequence, different approaches have been proposed in the literature to define how to classify samples in PLS-DA based on the values of Y^new. The naivest approach (see e.g., Alsberg et al., 1998) is to assign each sample to the category corresponding to the highest value of the predicted response vector. For instance, if the following predictions were obtained for a particular sample: y^new,k=[0.1 -0.4 0.8], it would be assigned to Class 3. On the other hand, other strategies have been also suggested, like the application of LDA on Y^new or on the PLS scores (Nocairi et al., 2004; Indahl et al., 2007), or the use of thresholds based on probability theory (Pérez et al., 2009).</p><!><p>As already stated, class-modeling methods aim at identifying a closed (bound) sub-space, where it is likely to find samples from a particular category, irrespective of whether other classes should also be considered or not. They try to capture the features, which make individuals from the same category similar to one another. Operationally, they define the class space by identifying the "normal" variability which can be expected among samples belonging to that category and, accordingly, introducing a "distance-to-the-model" criterion which accounts for the degree of outlyingness of any new sample. Among the different class-modeling techniques proposed in the literature, soft independent modeling of class analogies (SIMCA) is by far the most commonly used, especially for spectroscopic data, due to its ability of dealing with ill-conditioned experimental data matrices and, therefore, it will be briefly described below (for more details, the reader is referred to Wold, 1976; Wold and Sjöström, 1977, 1987).</p><!><p>The main idea behind SIMCA is that the systematic variability characterizing the samples for a particular category can be captured and accurately accounted for by a PCA model of appropriate dimensionality. This model is built by using only the samples from the investigated category:</p><p>where the symbols have the same meaning as in Equation (2), and the subscript indicates that the model is calculated by using only the training data from class g. The use of PCA to define the similarities among the samples belonging to the category of interest provides also the machinery to assess whether any new sample is likely to come from that class or not through the definition of two statistics normally used for outlier detection, namely T2 and Q. As already introduced in section Principal Component Analysis, the former is the squared Mahalanobis distance of a sample to the center of the scores space, indicating how far the individual is from the distribution of the "normal" samples in the space spanned by the significant PCs (Hotelling, 1931), while the latter is the (Euclidean) distance of the sample to its projection onto the PC space, describing how well that individual is fitted by the PCA model (Jackson and Muldholkar, 1979). In the context of SIMCA, once the PCA model of the gth category is calculated according to Equation (27), any specimen to be predicted is projected onto that model and its values of T2 and Q are used to calculate an overall distance to the model di, g (Yue and Qin, 2001), which constitutes the basis for class acceptance or rejection:</p><p>where the subscript indicates that the ith sample is tested against the model of the gth category. Accordingly, the boundary of the class space is identified by setting a proper threshold to the distance, so that if a sample has a distance to the model lower than the threshold it is accepted by the category and, otherwise, it is rejected.</p><!><p>As mentioned before, classification approaches are widely applied in quality controls of pharmaceuticals, in particular to detect counterfeit drugs, as, for instance, it is reported in da Silva Fernandes et al. (2012), where NIR and fluorescence spectroscopy were combined with different classification methods to distinguish among pure and adulterated tablets. In Storme-Paris et al. (2010), a non-destructive approach is proposed to distinguish genuine tablets from counterfeit or recalled (from the market) medicines. In order to achieve this, NIR spectra (directly collected on the tables) are analyzed by SIMCA. Results obtained suggest the validity of this approach; in fact, it allowed highlighting small differences among drugs (e.g., different coating), and it provided an excellent differentiation among genuine and counterfeits products. For the same purpose, namely counterfeit drug detection, NIR spectra were also widely combined with PLS-DA. Only to mention one, de Peinder et al. (2008) demonstrated the validity of this approach to spot counterfeits of a specific cholesterol-lowering medicine. Despite the fact that the authors highlighted the storage conditions sensibly affecting NIR spectra (because of humidity), the PLS-DA model still proved to be robust and provided excellent predictions.</p><!><p>Chemometrics relies mainly on the use of empirical models which, given the experimental measurements, should summarize the information of the data, reasonably approximate the system under study, and allow predictions of one or more properties of interest. Bearing this in mind, given the "soft" (i.e., empirical) nature of the models employed, there are many models one could in principle calculate on the same data and their performances could be influenced by different factors (number of samples and their representativeness, the method itself, the algorithm, and so on) (Brereton et al., 2018). Thus, selecting which model is the most appropriate for the data under investigation and verifying how reliable it is, is of fundamental importance and the chemometric strategies for doing so are collectively referred to as validation (Harshman, 1984; Westad and Marini, 2015). To evaluate the quality of the investigated models, the validation process requires the definition of suitable diagnostics, which could be based on model parameters but more often rely on the calculation of some sort of residuals (i.e., error criteria). In this context, in order to avoid overoptimism or, in general, to obtain estimates which are as unbiased as possible, it is fundamental that the residuals which are used for validation are not generated by the application of the model to the data it has been built on, since in almost all cases, they cannot be considered as representative of the outcomes one would obtain on completely new data. For such reason, a correct validation strategy should involve the estimation of the model error on a dataset different than the one used for calculating the model parameters. This is normally accomplished through the use of an external test set or cross-validation.</p><p>The use of a second, completely independent, set of data for evaluating the performances and, consequently, calculating the residuals (test set validation) is the strategy which best mimics how the model will be routinely used, and it is therefore the one to be preferred, whenever possible. On the other hand, cross-validation is based on the repeated resampling of the dataset, into a training and a test sub-sets, so that at each iteration only a part of the original samples is used for model building while the remaining individuals are left out for validation. This procedure is normally repeated up to the moment when each sample has been left out at least once or, anyway, for a prespecified number of iterations. Cross-validation is particularly suited when the number of available samples is small and there is no possibility of building an external test set, but the resulting estimates can be still biased as the calibration and validation sets are never completely independent on one another. In general, it is rather used for model selection (e.g., estimating the optimal number of components) than for the final validation stage.</p><!><p>In addition to some specific applications described above, in this paragraph additional examples will be presented to further emphasize the usefulness of chemometrics-based spectroscopy for pharmaceutical analysis.</p><p>Morris and Forbes (2001) coupled NIR spectroscopy with multivariate calibration for quantifying narasin chloroform-extracted from granulated samples. In another study, Forbes et al. (2001) proposed a transmission NIR spectroscopy method using multivariate regression for the quantification of potency and lipids in monensin fermentation broth.</p><p>Ghasemi and Niazi (2007) developed a spectrophotometric method for the direct quantitative determination of captopril in pharmaceutical preparation and biological fluids (human plasma and urine) samples. Since the spectra were recorded at various pHs (from 2.0 to 12.8), different models were tested, including the possibility of a preliminary spectral deconvolution using multi-way approaches. In particular, the use of PLS on the spectra at pH 2.0 allowed to build a calibration curve which resulted in a very good accuracy. Li et al. (2014) used Raman spectroscopy to identify anisodamine counterfeit tablets with 100% predictive accuracy and, at the same time, NIR spectroscopy to discriminate genuine anisodamine tablets from 5 different manufacturing plants. In the latter case, PLS-DA models were found to have 100% recognition and rejection rates. Willett and Rodriguez (2018) implemented a rapid Raman assay for on-site analysis of stockpiled drugs in aqueous solution, which was tested on Tamiflu (oseltamivir phosphate) by using three different portable and handheld Raman instruments. PLS regression models yielded an average error with respect to the reference HPLC values, which was lower than 0.3%. Other examples of application can be found in Forina et al. (1998), Komsta (2012), Hoang et al. (2013), and Lohumi et al. (2017).</p><!><p>Chemometrics provide a wealth of techniques for both the exploratory analysis of multivariate data as well as building reliable calibration and classification strategies to predict quantitative and qualitative responses based on the experimental profiles collected on the samples. Coupled to spectroscopic characterization, it represents an indispensable and highly versatile tool for pharmaceutical analysis at all levels.</p><!><p>AB and FM jointly conceived and designed the paper, and wrote the manuscript. All authors agreed on the content of the paper and approved its submission.</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

The right place at the right time: Aurora B kinase localization to centromeres and kinetochores

The fidelity of chromosome segregation during mitosis is intimately linked to the function of kinetochores, which are large protein complexes assembled at sites of centromeric heterochromatin on mitotic chromosomes. These key \xe2\x80\x9corchestrators\xe2\x80\x9d of mitosis physically connect chromosomes to spindle microtubules and transduce forces through these connections to congress chromosomes and silence the spindle assembly checkpoint. Kinetochore-microtubule attachments are highly regulated to ensure that incorrect attachments are not prematurely stabilized, but instead released and corrected. The kinase activity of the centromeric protein Aurora B is required for kinetochore-microtubule destabilization during mitosis, but how the kinase acts on outer kinetochore substrates to selectively destabilize immature and erroneous attachments remains debated. Here we review recent literature which sheds light on how Aurora B kinase is recruited to both centromeres and kinetochores and discuss possible mechanisms for how kinase interactions with substrates at distinct regions of mitotic chromosomes are regulated.

the_right_place_at_the_right_time:_aurora_b_kinase_localization_to_centromeres_and_kinetochores

Aurora B kinase regulates kinetochore-microtubule attachment stability in mitosis<!>Current models for kinetochore-microtubule attachment regulation by Aurora B kinase<!>Aurora B and the CPC are recruited to centromeres via phosphorylated histones<!>pH2A-T120 and pH3-T3 recruit the CPC to distinct locations within the centromere<!>A role for centromere-localized Aurora B kinase in chromosome segregation<!>Aurora B kinase localizes to kinetochores<!>How is Aurora B kinase recruited to the kinetochore to phosphorylate kinetochore substrates?<!>Closing comments

<p>Equal division of genetic material during mitosis requires that each sister chromatid of a mitotic chromosome stably attach to spindle microtubules emanating from each of the two opposite spindle poles. A complex network of proteins built on the centromere region of mitotic chromosomes, collectively called the kinetochore, mediates these attachments. Successful chromosome segregation also requires the precise regulation of kinetochore-microtubule attachment stability. In early mitosis, the mitotic spindle begins to form and establish its proper geometry at the same time that microtubules begin to dynamically probe for chromosomes, and as a result, erroneous kinetochore-microtubule attachments are likely to form (Figure 1) [1–6]. In order to prevent premature stabilization of kinetochore-microtubule attachments and to limit the accumulation of erroneous attachments, microtubule turnover at kinetochores is high during early mitosis; conversely, as mitosis progresses and chromosomes begin to bi-orient, kinetochore-microtubule turnover decreases and stably-bound microtubules accumulate at kinetochores [2, 6–10]. These stable attachments allow kinetochores to harness the forces generated from depolymerizing microtubule plus ends to power chromosome movements and to silence the spindle assembly checkpoint, which delays anaphase until all kinetochores are properly connected to microtubules [2, 6, 8, 11–13].</p><p>The Chromosomal Passenger Complex (CPC) is comprised of INCENP, Borealin, Survivin and Aurora B kinase, the enzymatic component of the complex which phosphorylates multiple substrates on mitotic chromosomes to ensure proper chromosome segregation [14–17]. One of the numerous functions of Aurora B kinase is to regulate kinetochore-microtubule attachment stability [14, 18–21]. For this purpose, Aurora B phosphorylates outer kinetochore-associated substrates, including Hec1 of the NDC80 complex, which directly links kinetochores to microtubules [8, 22, 23]. Phosphorylation of Hec1 decreases the affinity of NDC80 complexes for microtubules, and as a result, reduces kinetochore-microtubule attachment stability [24–27]. As mitosis progresses, Aurora B kinase-mediated phosphorylation of Hec1 and other outer kinetochore substrates decreases, resulting in increased stabilization of kinetochore-microtubule attachments, which in turn promotes chromosome congression and silencing of the spindle assembly checkpoint [12, 13, 24–27].</p><!><p>Although proper regulation of kinetochore-microtubule attachments requires that Aurora B phosphorylate substrates at outer kinetochores, the kinase itself, along with the rest of the CPC, resides prominently at inner centromeres during mitosis [28–32]. To explain how the centromere-localized kinase is capable of regulating the function of outer kinetochore proteins, researchers have proposed the "spatial positioning" model (Figure 2), which posits that activated Aurora B kinase emanates from the inner centromere as a diffusible gradient to phosphorylate its substrates [19, 33–38]. In this model, Aurora B is recruited to and activated at inner centromeres in early mitosis, prior to formation of kinetochore-microtubule attachments [31, 34–50]. Kinetochores in this case lack pulling forces from attached microtubules and are physically close to the inner centromere. As such, kinetochore substrates are situated within the reach of the kinase and are highly phosphorylated. As kinetochore-microtubule attachments are generated, microtubule-based pulling forces stretch kinetochores away from centromeres and outside the boundaries of the Aurora B gradient, which results in decreased phosphorylation of kinetochore substrates and subsequent stabilization of kinetochore-microtubule attachments [19, 33–38, 51]. In support of this model, Liu et al. [36] demonstrated that an ectopically targeted FRET sensor capable of detecting Aurora B kinase activity was phosphorylated when positioned at centromeres, but not at kinetochores, when kinetochores were properly bi-oriented. Additionally, ectopically targeting Aurora B to kinetochores using a Mis12-INCENP fusion protein destabilized kinetochore-microtubule attachments and delayed spindle assembly checkpoint silencing. This led the authors to conclude that stabilization of attachments in metaphase results from the spatial separation of outer kinetochore substrates from centromere-localized Aurora B kinase [36]. It is important to point out however, that other models describing Aurora B regulation of attachment stability (discussed below) predict that irreversibly targeting Aurora B to kinetochores in metaphase, when kinase activity at this region is known to be low, would lead to destabilization of attachments. Thus, the data presented by Liu et al., [36] do not necessarily rule out other mechanistic models for Aurora B regulation of kinetochore-microtubule attachment stability [36].</p><p>The spatial positioning model described above is rooted in the idea that Aurora B kinase emanates from the inner centromere as a steep, diffusible gradient capable of differentially phosphorylating substrates within a short distance (~50–100 nm) [25, 26, 36, 37, 51–53]. The presence of such a fine-tuned gradient is debated, and the mechanisms for how the proposed gradient is established and maintained remain unknown [21, 38, 51, 52, 54, 55]. Similar to the spatial positioning model, the "dog leash" model accounts for differential activity of Aurora B kinase towards its substrates based on their distance from the centromere but does not rely on a diffusible gradient of the kinase. In this proposed mechanism, Aurora B's "zone" of activity is restricted by its interaction with the CPC component INCENP, which contains a long single α-Helix (SAH) domain that may be capable of extending up to 80 nm to reach kinetochore substrates in early, but not late mitosis [21, 45, 47]. Consistent with this model, expression of a mutant version of chicken INCENP containing a shortened SAH domain in human cells resulted in decreased phosphorylation of outer kinetochore-, but not centromere-localized Aurora B substrates [45].</p><p>A third model to describe Aurora B regulation during mitosis suggests that spindle microtubules promote the activation of Aurora B kinase and facilitate phosphorylation of kinetochore substrates [31, 40–42, 44–50]. While the available data suggest a role for microtubules in activating Aurora B kinase, it seems unlikely that this is the major mechanism for regulating kinase activity at outer kinetochores in response to kinetochore-microtubule attachment, since the Hec1 tail domain and other kinetochore-associated Aurora B substrates remain highly phosphorylated when cells enter mitosis in the presence of the microtubule depolymerizing drug nocodazole [24, 37, 56].</p><p>A growing number of studies in both budding yeast and human cells have demonstrated that Aurora B kinase localizes not only to centromeres, but also to kinetochores, suggesting an alternative regulatory mechanism for controlling kinetochore-microtubule attachment stability, in which Aurora B kinase is recruited directly to kinetochores in early mitosis to phosphorylate its substrates and in turn, is evicted from kinetochores as stable attachments form (Figure 3) [21, 24, 43, 56–62]. In the sections below we discuss recent studies that shed light on the localization and functional properties of the CPC at both centromeres and kinetochores and how these new findings may lead to refinement of the current models for Aurora B kinase regulation of kinetochore-microtubule attachments during mitosis.</p><!><p>Aurora B kinase and its CPC cofactors are recruited to the centromere region of mitotic chromosomes just prior to nuclear envelope breakdown [14, 40, 63], and this recruitment depends on phosphorylation of histones H3 and H2A (Figure 4) [64]. A significant body of work has demonstrated that Haspin kinase phosphorylates histone H3 at Thr3 (pH3-T3), which recruits the CPC component Survivin [64–67]. The BIR domain of Survivin directly interacts with pH3-T3 [28, 56, 66, 68, 69], while a separate helical domain of Survivin forms a three-helix bundle with Borealin and INCENP, which is connected to Aurora B through the C-terminal IN-Box of INCENP [32, 69]. It has also been demonstrated that Bub1 phosphorylates histone H2A at Thr120 which recruits the Shugoshin proteins Sgo1 and Sgo2 to centromeres, which in turn recruit the CPC [64, 70–73]. In metazoans, a number of studies have suggested that this linkage is mediated through Borealin [64, 70–73]. Antibodies to both pH3-T3 and pH2A-T120 localize to centromeres, and loss of either phospho-mark reduces centromeric CPC localization, which has led to a model in which Aurora B kinase and the CPC are recruited to regions of centromeric chromatin where the two marks overlap [64, 68, 72, 74]. This concentrated pool of centromere-localized Aurora B kinase is proposed to phosphorylate both centromere and kinetochore substrates to ensure proper chromosome congression and segregation [34, 36, 37, 64, 66, 71, 75, 76].</p><!><p>In a series of recently published studies, three groups carried out experiments to directly test the above model for CPC localization; that is, to determine if CPC recruitment to inner centromeres requires the overlap of pH3-T3 and pH2A-T120 in human cells [56, 61, 62]. Using antibodies to both phospho-marks, the authors of all three studies reported that pH2A-T120 and pH3-T3 did not show significant overlap in cells; instead, immunostaining revealed that pH3-T3 localized distinctly as a single focus at the inner centromere, while pH2A-T120 localized as a pair of dots flanking the inner centromere [56, 61, 62]. This is consistent with previous data suggesting distinct localization patterns of the two histone marks [21, 58, 64, 66, 71, 77]. Line-scans and two-color localization experiments further revealed that pH2A is localized inside of the inner kinetochore protein CENP-C, on the order of ~100 nm, in both prometaphase and metaphase cells, which places this mark within the "kinetochore-proximal outer centromere" region, distinct from the pH3-T3-marked chromatin at the inner centromere [56, 61, 62]. Importantly, all three studies reported that each histone mark individually was sufficient to recruit Aurora B kinase and the CPC. Each group analyzed U2OS cells containing an ectopic Lac operator (LacO) array stably integrated in the short arm of chromosome one that were expressing fusions of either LacI-Haspin or LacI-Bub1 [78]. In cells expressing LacI-Bub1, the chromatin surrounding the LacO array was positive for pH2A-T120 but not pH3-T3, and the single phosphorylation mark (pH2A-T120) was sufficient for recruitment of Aurora B kinase in a manner dependent on Sgo1 [56, 61, 62]. Similarly, when Haspin was directed to the ectopic locus through expression of a LacI fusion protein, the local chromatin was positive for pH3-T3 but not pH2A-T120, and this single modification was also sufficient to recruit Aurora B kinase and its CPC partners [56, 61, 62]. Moreover, each histone mark was sufficient to recruit a population of the CPC to spatially distinct regions within the centromere region of mitotic chromosomes in human cells (Figure 4). While depletion of Haspin or inhibition of its kinase activity resulted in loss of both the pH3-T3 mark and accumulation of Aurora B at inner centromeres, a population of Aurora B remained localized to the kinetochore-proximal outer centromere, coincident with pH2A-T120 [56, 58, 61, 62]. Conversely, inhibition of Bub1 kinase activity resulted in loss of the pH2A-T120 mark, but Aurora B kinase and components of the CPC remained localized at the inner centromere coincident with pH3-T3 [56, 61, 62, 74]. Inhibition of both Bub1 and Haspin kinase activities; however, resulted in no detectable Aurora B and CPC components at centromeres [49, 56, 61, 62, 74].</p><p>In the studies described above, Aurora B kinase localized prominently to inner centromeres as a single focus, but was not clearly discernable at pH2A "marked" kinetochore-proximal outer centromere regions in control cells [56, 61, 62]. This population of the CPC became readily detectable however, when phosphorylation of H3-T3 was prevented through Haspin knockout or Haspin inhibition [56, 58, 61, 62], suggesting crosstalk between the two centromere-localized populations of the CPC. These results point to the possibility of a multifaceted loading process whereby a pool of Aurora B kinase is loaded directly to the inner centromere binding sites provided by pH3, and a second population of the complex is initially recruited to the kinetochore-proximal outer centromere by pH2A/Sgo1 and subsequently relocated to the inner centromere region. This mechanism is similar to what has been suggested for Sgo1, which first loads to kinetochores in early mitosis, which is required for its subsequent relocalization to the inner centromere, where it functions to protect cohesion and prevent premature sister chromatin separation [38, 73, 77, 79]. Furthermore, authors from the recent Liang et al. [56] study suggest that relocalization of Aurora B kinase from the pH2A-T120 binding sites to the inner centromere pH3-T3 binding sites in metaphase may be required to silence the spindle assembly checkpoint in response to kinetochore-microtubule attachment. The authors reported that experimentally-induced retention of Aurora B and the CPC at the kinetochore-proximal outer centromere in metaphase cells resulted in a small increase (by ~20%) in Aurora B kinase-mediated phosphorylation of the kinetochore scaffolding protein KNL1 which led to sustained checkpoint signaling [56]. In sum, three recent studies report the identification of discrete populations of Aurora B kinase within the centromere region that are recruited by distinct histone modifications. These studies suggest that the different populations of the CPC functionally interact and cooperatively contribute to the robust accumulation of Aurora B kinase at the inner centromere of mitotic chromosomes.</p><!><p>In light of the finding that each of the two histone marks recruits a distinct population of Aurora B and the CPC to centromeres, the authors of the studies described above [56, 61, 62] tested if either population is required for Aurora B kinase activity at kinetochores or for proper chromosome segregation. The three studies reported that inhibition of either pathway alone did not result in chromosome segregation errors or reduced phosphorylation of kinetochore Aurora B kinase substrates [56, 61, 62]. However, in cells inhibited for both Bub1 and Haspin kinase activities, chromosome segregation was compromised, although it was noted that the defects were less severe than those observed in cells inhibited for Aurora B kinase itself [56, 62]. Thus, the inner centromere and kinetochore-proximal outer centromere populations of Aurora B likely have redundant roles in ensuring proper chromosome segregation. Strikingly, in cells inhibited for both Bub1 and Haspin kinase activities, which resulted in a complete loss of centromere-localized CPC, Aurora B kinase localization remained high at kinetochores, and kinetochore substrate phosphorylation by Aurora B kinase was not largely reduced when compared to control cells (Figure 4) [56, 61, 62]. As such, the chromosome segregation defects resulting from loss of centromere-localized Aurora B could not be attributed to loss of phosphorylation of kinetochore substrates [56, 61, 62]. Importantly, these results provide evidence that centromere accumulation of the Aurora B kinase is not strictly coupled to Aurora B activity at kinetochores. This idea is consistent with a number of previous studies, the first of which demonstrated that centromere-localized Aurora B is not required for the regulation of kinetochore-microtubule attachments in chicken DT40 cells [80]. Here, cells depleted of endogenous Survivin and expressing a mutant version of Survivin defective for centromere localization completed mitosis normally with no detectable chromosome segregation defects [80]. In addition, kinetochore-microtubule attachment defects observed in HeLa cells depleted of the large kinetochore-associated scaffolding protein KNL1, which resulted in loss of kinetochore-associated Aurora B kinase activity, could not be rescued by ectopic targeting of the CPC to centromeres [57]. Finally, several studies have suggested that centromere accumulation of the CPC is uncoupled from kinetochore-associated Aurora kinase activity in budding yeast. Campbell and Desai [43] reported that budding yeast cells expressing INCENP/Sli15 mutants that fail to localize to centromeres exhibited normal chromosome bi-orientation and Aurora/Ipl1 kinase-mediated error correction. Recent studies have further demonstrated that the CPC is recruited directly to kinetochores in budding yeast, and this population is sufficient for Aurora kinase/Ipl1 activity at kinetochores and error-free chromosome segregation in the absence of centromere-localized CPC [59, 60].</p><p>An obvious question that emerges from the recent studies in human cells [56, 62] is what causes the observed chromosome segregation errors in the absence of centromere-localized Aurora B if the kinase is still able to phosphorylate kinetochore-associated substrates normally? A plausible explanation is that centromere-localized Aurora B phosphorylates centromere-localized substrates to promote proper chromosome segregation. Previous studies have demonstrated that Aurora B kinase regulates the activity and localization of MCAK, a centromere-localized kinesin-13 motor that promotes microtubule depolymerization, an activity that is implicated in correcting erroneous kinetochore-microtubule attachments [3, 9, 75, 76, 81–83]. Interestingly, the centromere localization of MCAK is perturbed in cells that are depleted of Haspin or inhibited for Bub1 kinase activity [62, 66, 67, 84, 85], consistent with the notion that alterations in MCAK activity and/or localization might contribute to the chromosome segregation errors observed in cells lacking centromere-localized Aurora B kinase. Further investigation into the role of MCAK and other centromere-localized Aurora B kinase substrates is needed to resolve this issue. It is interesting to note that while antibodies to CPC components localize prominently to the inner centromere in early mitosis, antibodies to phosphorylated, active Aurora B (pT232) and phosphorylated, active INCENP (pS893/pS894) show minimal inner centromere localization in early mitosis, but levels significantly increase as mitosis progresses [24, 61]. Consistent with this idea, Sgo1/2-PP2A antagonizes Aurora B activity at the centromere in early mitosis, which may explain this change in activity as mitosis progresses [86]. These results suggest that centromere substrates of Aurora B kinase that contribute to proper chromosome segregation may be phosphorylated and perhaps activated in late mitosis rather than in early mitosis.</p><p>Finally, it is important to point out that recent studies demonstrating that centromere-localized CPC is not explicitly required for phosphorylation of kinetochore substrates do not rule out the possibility that the centromere- and kinetochore-localized pools of the CPC may exhibit cross-talk and impact each others' localization or activity. In fact, studies have demonstrated that while delocalization of CPC from centromeres did not result in decreased activity of Aurora B at kinetochores in human cells and in M-phase Xenopus egg extracts, the regulation of kinase activity in response to kinetochore-microtubule attachment was compromised [27, 65, 87]. Why this is the case is not clear, but in the future it will be important to resolve how the centromere pool of the CPC contributes to proper regulation of Aurora B kinase substrate phosphorylation at kinetochores in response to microtubule attachment.</p><!><p>As previously mentioned, in addition to its centromere localization, Aurora B has also been detected at the kinetochore. Antibodies directed to active, phosphorylated Thr232, which resides in the T-loop of the kinase domain of Aurora B and is required for full activity of the kinase, recognize kinetochores in early mitosis, centromeres in late prometaphase and metaphase, and to the spindle midzone in anaphase [24, 54, 88–90]. A similar localization pattern was observed for phosphorylated Aurora B Ser331, a site whose phosphorylation is required for optimal Aurora B activity [91]. Furthermore, antibodies to phosphorylated, active INCENP (pS893/pS894) recognize both kinetochores and centromeres in early mitosis, and this localization shifts primarily to centromeres in late prometaphase and metaphase and the spindle midzone in anaphase [61]. Together, these studies raise the possibility that a population of Aurora B kinase, and likely the entire CPC, is recruited directly to kinetochores in early mitosis where it phosphorylates its kinetochore substrates to promote kinetochore-microtubule turnover. In such a model, as mitosis progresses and kinetochores accumulate bound microtubules, Aurora B is evicted from kinetochores, resulting in decreased kinetochore substrate phosphorylation, stabilization of attachments, and silencing of the spindle assembly checkpoint (Figure 3) [12, 13, 35, 61, 92–94].</p><!><p>A major task that remains to be tackled is identifying the kinetochore binding sites for the CPC. In budding yeast, two research groups have made considerable headway on this front by demonstrating that inner kinetochore COMA (Ctf19/Okp1/Mcm21/Ame1) complex recruits the CPC through a direct interaction between INCENP/Sli15 and Ctf19 [59, 60]. Importantly, this kinetochore-associated population is sufficient to support Ipl1/Aurora kinase activity and normal chromosome segregation in the absence of centromere-localized CPC [59, 60]. In metazoan cells, however; the kinetochore binding sites for Aurora B and the CPC remain unknown. In a recent study discussed above [61], authors approximated the location of Aurora B kinase in early mitotic cells to ~22 nm outside of the inner kinetochore protein CENP-C, which places it ~8 nm inside of the N-terminus of the outer kinetochore protein Hec1. Many kinetochore proteins localize in this region of the kinetochore, making it difficult to predict specific binding sites. A previous study reported that Aurora B kinase localization to kinetochores is dependent on the kinetochore protein KNL1 [57], and Broad et al. [61] demonstrated that Aurora B localization is at least partially dependent on Bub1. Moreover, a recent study found that KNL1 undergoes significant conformational changes upon kinetochore-microtubule attachment [92]. Together, these findings make it tempting to speculate that KNL1 may directly or indirectly provide binding sites for CPC components in early mitosis. As mitosis progresses and kinetochore-microtubule attachments accumulate, KNL1 may undergo conformational changes that occlude these sites, leading to eviction of Aurora B kinase and its CPC cofactors (Figure 2). This speculative model remains to be tested.</p><p>Importantly, the available data suggest that Aurora B kinase may be recruited to multiple locations within the kinetochore to facilitate different functions. Based on the recent studies described above, we predict that Aurora B is recruited directly to outer kinetochores to phosphorylate outer kinetochore substrates involved in kinetochore-microtubule attachment regulation [21, 24, 57, 61, 89]. However, as mentioned above, in budding yeast, the CPC is recruited to the inner kinetochore COMA complex, whose homolog in humans is the CENP-O/P/Q/U complex, a component of the Constitutive Centromere Associated Network (CCAN) [59, 60]. Indeed, Aurora B kinase has substrates at the inner kinetochore that are important for mitotic progression. For example, the Mis12 complex component Dsn1 is phosphorylated by Aurora B in early mitosis to promote kinetochore assembly by facilitating an interaction between the inner-kinetochore protein CENP-C and the Mis12 complex [95–104]. A recent study from Bonner et al. [102] demonstrated that delocalization of the CPC from centromeres in M-phase Xenopus egg extracts resulted in decreased Dsn1 phosphorylation and consequently, failure to assemble the outer kinetochore. The authors of this study found that the SAH domain of INCENP was required, in a microtubule-independent manner, for Aurora B kinase-mediated phosphorylation of Dsn1 and kinetochore assembly. Ectopic targeting of INCENP lacking its central SAH domain to the inner kinetochore protein Nsl1 and subsequent recruitment of the CPC to this region rescued both Dsn1 phosphorylation and kinetochore assembly [102]. The authors concluded that the INCENP SAH domain is critical for localizing the CPC to the inner kinetochore, in close proximity to the Mis12 complex so that Aurora B is able to efficiently phosphorylate Dsn1 to promote outer kinetochore assembly [102]. These results also bring to light the idea that multiple kinetochore functions rely on kinetochore-associated Aurora B kinase activity, and the CPC may be recruited to different locations within the kinetochore to support these activities.</p><!><p>A growing number of studies has demonstrated that centromere-localized Aurora B kinase is not explicitly required for Aurora B kinase activity at kinetochores. Based on these studies and data from numerous model systems, the current model for Aurora B kinase-mediated regulation of kinetochore-microtubule attachment stability (i.e. the "spatial positioning model") should be re-evaluated. A growing number of studies have mapped Aurora B kinase to kinetochores, and ectopically targeting the CPC to kinetochores in multiple cell types rescues loss of the centromere-localized population. In budding yeast, at least one kinetochore binding site for the CPC has been identified, which suggests that Aurora/Ipl1 kinase localizes to kinetochores to specifically phosphorylate kinetochore substrates. This straightforward mechanism for CPC function at budding yeast kinetochores, in which the kinase is recruited to the region of mitotic chromosomes where its substrates are located, is likely utilized in metazoan cells as well. The next major challenge is to identify the binding site, or more likely, binding sites, for the CPC at metazoan kinetochores.</p>

PubMed Author Manuscript

Direct Observation of Chain Lengths and Conformations in Oligofluorene Distributions from Controlled Polymerization by Double Electron-Electron Resonance

Synthetic polymers are mixtures of different length chains, and their chain length and chain conformation is often experimentally characterized by ensemble averages. We demonstrate that Double-Electron-Electron-Resonance (DEER) spectroscopy can reveal the chain length distribution, and chain conformation and flexibility of the individual n-mers in oligo-(9,9-dioctylfluorene) from controlled Suzuki-Miyaura Coupling Polymerization (cSMCP). The required spin-labeled chain ends were introduced efficiently via a TEMPO-substituted initiator and chain terminating agent, respectively, with an in situ catalyst system. Individual precise chain length oligomers as reference materials were obtained by a stepwise approach. Chain length distribution, chain conformation and flexibility can also be accessed within poly(fluorene) nanoparticles.

direct_observation_of_chain_lengths_and_conformations_in_oligofluorene_distributions_from_controlled

Introduction<!>Results and Discussion<!>Table 1.<!>Oligomer<!>Conclusion<!>ASSOCIATED CONTENT

<p>Knowledge of the chain length distribution and of chain conformations are essential to understand and design synthetic procedures, materials properties, and nanoscale structures. 1,2 Consequently numerous methods to access these parameters have been developed to a high level and for practical use. The most prominent and ubiquitious method to access molecular weight distributions is size exclusion chromatography. More recently, mass spectrometry has also been advanced considerably for the lower molecular weight regime, though it is not quantitative. These methods are based on separation of the sample, prior to detection of its individual chain length components. Likewise, an experimental determination of chain conformations on mixtures of different length chains yields an ensemble average only.</p><p>We now report double-electron-electron-resonance (DEER) 3,4 studies of spin-labeled oligofluorene mixtures that give direct access to the chain length distributions and monitor conformational ensembles and flexibilities 5 of the oligomers.</p><p>As a probe we chose oligofluorenes from controlled Suzuki-Miyaura coupling polymerization (cSMCP). 6,7 Oligofluorenes are attractive materials due to their photo-and electroluminescence and light-induced charge generation. 8,9 cSMCP proceeds in a chain growth fashion. 10 The controlled character of cSMCP allows for an introduction of functional endgroups at both chain ends. [10][11][12] cSMCP has been demonstrated for the synthesis of a scope of polyarylenes, like poly(fluorenes), poly(thiophenes) and poly(phenylenes). 7,13 DEER is a pulsed electron paramagnetic resonance (EPR) method to determine distance distributions between paramagnetic centers. Nitroxide groups, such as 2,2,6,6-tetramethylpiperidinyloxyl (TEMPO) are commonly used as spin labels. DEER is established in biological chemistry where it is commonly used to determine distances between a pair of spins placed in a single defined type of molecule, like a protein. 14,15 In the same way, DEER has also been applied to defined monodisperse synthetic molecules. 16,17</p><!><p>Spin labels were attached at the chain ends of poly(9,9dioctylfluorene) (PF8) directly during polymerization by employing spin-labeled initiators and terminating agents in an otherwise established controlled SMCP protocol. 10,11,18 The polymerization was initiated by an in situ system using chloro[(tri-tert-butylphosphine)-2-(amino-biphenyl)] palladium(II) as Pd(0) source, and TEMPO-labeled 4-bromobenzoic acid which adds oxidatively. Chains were quenched by addition of TEMPO-labeled 4-carboxyphenyl-boronic acid pinacol ester end capping agent, resulting in identical chain termini (Figure 1). The successful incorporation of the TEMPO labels was confirmed by MALDI-TOF MS, identi-fying the doubly labeled polymer chains as major component by the isotopic pattern (cf. Figure S 4.1). As expected, the maximum of the obtained chain length distribution determined by GPC could be adjusted by the ratio of initiator to monomer in the reaction mixture. Thus, the described in situ system offers a robust and versatile method for the synthesis of doubly TEMPO labeled conjugated polymers in a one-step approach. For establishing a chain conformation analysis, doubly labeled monodisperse oligomers with a precise number of repeat units (herein referred to as doubly labeled precise oligomers) were required as a reference. According to reported procedures, the non-labeled oligo(fluorenes) were built up stepwise by repetitive cycles of alternating Suzuki-Miyaura coupling and bromination, starting from the mono-and dibromo-substituted fluorene monomers (Figure S 3.2). [19][20][21] Spin labels were attached to the oligomers by a final Suzuki-Miyaura coupling step with a TEMPO-substituted monofunctional arylboronic acid ester. These doubly labeled precise oligomers from stepwise synthesis are abbreviated as Pn, with n representing the chain length in the following. Oligomers up to n = 5 were synthesized. As the synthetic effort of the step-by-step approach considerably increases for each additional monomer unit, reference oligomers with n = 5 and 6, abbreviated as P5´ and P6´, were provided by semi-preparative GPC fractionation (cf. S 3.2).</p><p>As reference, DEER measurements in Q-band for the precise oligomers were performed upon shock-freezing in toluene-d8 (Figure 2A). The experimental data for all individual oligomers Pn (with n = 1-4), P5´ and P6´ was fitted (Figure 2A) with the worm-like chain model (WLC), 17,22,23 using DEERAnalysis. [24][25][26] Assuming that the monomer P1 with just one repeat unit is completely stiff, the width of the respective distance distribution can be accounted to the flexibility of the spin-label end group. It can be described by a Gaussian broadening 16,17 with σ = 0.06 nm. For all further experiments, we used this Gaussian broadening to take the flexibility of the spin-label end groups into account. The persistence length Lp is a global parameter for describing the set of DEER data, the contour length Lc was fitted for each oligomer individually. The obtained distance distributions are shown in Figure 2B, the fit parameters are listed in Table 1.</p><!><p>Oligomers and parameters of model-based fit with WLC. Global parameters: Lp = 14 nm and σ = 0.06 nm. For P4 an additional gauss distance was introduced (r = 3.03 nm, σ = 0.267 nm, 21 % weighting factor).</p><!><p>Lc As an important detail, for P4 we observed occurrence of three-fold bromination (21 %) in the course of the stepwise oligomer synthesis (see SI Figure S4.2) and attachment of a third spin-label in the center of the chain. This results in a broad additional distance contribution below the expected end-to-end distance, which is described by an additional Gaussian in the model (Table 1).</p><p>The set of doubly labeled oligomers can be described with Lp = 14 nm and Lc is approx. (2.1 + n • 0.8) nm. This is in good agreement, with a monomer length of 0.75 nm, found for poly(9,9-bis(2-ethylhexyl)fluorene-2,7-diyl) (PF2/6). 27 With the WLC parameters for every oligomer in hand, DEER was applied to an oligomer mixture (Mix) obtained from controlled Suzuki-Miyaura polymerization (Figure 3A). A monomer to initiator species of five had been applied in the polymerization experiment.</p><p>The accessible range in the distance distribution detected by DEER is restricted by the dipolar evolution time, which was limited to 10 μs in our experiments due to spin-spin relaxation. Under these conditions, distance contributions and their width can be detected with an upper limit of 7 nm. The limit for accurate determination of the shape of the distribution is 5 nm. 4 To describe the experimental data, we used a model containing a superposition of different oligomers each described by WLC using the parameters obtained for the respected monodisperse reference oligomers (P1-4, P5´, P6´) according to a Poisson distribution with expectation value λ. As parameters for reference oligomers for n > 6 were not derived individually, we fitted the contour length of an oligomer with n = 7 as an additional parameter in the analysis.</p><p>Note that the polymerization mixture contains oligomers with n > 7, which are beyond the detection limit in our experiment. This does not disturb the analysis of the data for the oligomers amenable to DEER observation, however. To illustrate the amount of higher oligomers not observed by DEER in the overall distributions, these were calculated according to the found Poisson distribution (Figure 3B, dotted line).</p><p>In summary, we find that the experimental data is in excellent agreement with i) a Poisson distribution for the chain lengths, with an expectation value λ = 5.4 and ii) a chain conformation described by a WLC with a persistence length of 14 nm. Even when determining the fractions for each chain length individually in a model free approach, we find a reasonable agreement with a Poisson distribution (see SI S9). The agreement of the experimental DEER data with a Poisson distribution also allow for additional mechanistic conclusions on the polymerization reaction. Initiation of chains occurs very efficiently. That is, the initiating Pd-species is formed rapidly and completely in the early stages of the reaction, and it starts growth of a chain efficiently.</p><p>In many instances, solid polyfluorene materials are of interest rather than solutions. An access to chain conformations and chain flexibility in e.g. films or nanoparticles are desirable as they are instrumental in determining, for example, particle shapes. To demonstrate the principal suitability of the method reported also for nanoparticles, we prepared spherical poly(fluorene) nanoparticles (NP) by emulsification. [28][29][30][31][32] The bulk polymer was blended with P1, P2 and P3 (4:3:3 doubly labeled species, 0. Prior to DEER measurements, the particles were purified by dialysis for the removal of excessive surfactant and freeze-dried. DEER measurements within the nanoparticle were performed with a dipolar evolution time of 3 µs due to increased spin-spin relaxation in the protonated environment of the nanoparticle. Tm = 1.9 µs compared to 8.0 μs in the deuterated toluene-d8 matrix. The experimental data can be described by using contour length and broadening due to flexibility of the spin-label end group as derived in solution and fitting the weights for P1, P2 and P3 as well as the global persistence length for the oligomers. Figure 4c shows the corresponding distance distribution for spin-labeled oligomers incorporated in nanoparticles. The found fractions of 32 %, 28 % and 40 % are in agreement with the expectations from sample preparation (see above). We derived a persistence length of Lp = 14 nm as found in solution. This suggests a worm-like chain nature with undisturbed flexibility being retained in the nanoparticles.</p><!><p>In conclusion, we have demonstrated the application of DEER distance measurements to a real-life synthetic polymer containing a multitude of different chain length species. The required spin-labeled chain ends could be introduced efficiently by controlled Suzuki-Miyaura coupling Polymerization (sSMCP), cSMCP in general being a state-of-the-art versatile protocol for the synthesis of numerous poly(arylene)s. The DEER data agrees with a Poisson distribution, expected for the case of a living polymerization with fast and efficient initiation, which is given here. The method allows for the quantification of the individual n-mer populations and their respective conformations and flexibilities directly on mixtures. Further, we demonstrated that DEER allows a quantitative analysis of the oligomer fractions as well as characterization of conformation and flexibility inside nanoparticles in principle.</p><p>The necessity of spin-labeled oligomers is clearly a limitation of the analysis method reported. Thus, it is rather complementary than competitive to established standard methods for chain length distribution and conformation analysis. Its strength is access to properties of the individual n-mers directly on mixtures.</p><!><p>Supporting Information. Experimental procedures and characterization data; EPR measurements; Evaluation of DEER experiments.</p>

ChemRxiv

Analyses of Ferrous and Ferric State in DynabiTab Using Mössbauer Spectroscopy

Antianemic medicament ferrous gluconate, ferrous fumarate, and a Dynabi tablet with a basic iron bearing ingredient were studied with the use of Mössbauer spectroscopy. Room temperature spectra of ferrous gluconate gave clear evidence that the two phases of iron were present: ferrous (Fe2+) as a major one with a contribution at and above 91 a.u.% and ferric (Fe3+) whose contribution was found to be ~9 a.u.%. In the case of ferrous fumarate, a single phase was measured corresponding to ferrous (Fe2+) state. A Dynabi tablet consists of ferrous fumarate and ferrous fumarate. The ferric phase in ferrous gluconate is able to be reached about ~3.6 a.u.% in a tablet.

analyses_of_ferrous_and_ferric_state_in_dynabitab_using_mössbauer_spectroscopy

1. Introduction<!>2. Experimental<!>3. Results and Discussion<!>4. Conclusion

<p>Mössbauer spectroscopy is widely used for studying various molecular including biomolecular systems as well as various materials containing Mössbauer isotopes such as 57Fe, 119Sn, 121Sb, 127I, and 197Au. The nuclear hyperfine field, quadrupole splitting, and isomer shift provide very precise information about the electronic and magnetic state of the nuclei, chemical bonds, structure of the local environment, and so on [1, 2]. A number of pharmaceutical compounds contain Mössbauer atoms such as Fe, Sn, and Au. Pharmaceuticals containing Fe are mainly used for iron deficiency treatment. Iron is a very important mineral in the organic system in the human body. This element is an integral part of many proteins and enzymes. In addition, iron is essential for the regulation of cell growth and differentiation. Iron deficiency causes anemia and other pathological changes in the body. Absorption of iron from food requires recognition of the chemical form of iron by gut receptors. Both the shape and charge are important in the recognition process. Iron in a dietary supplement should be ferrous (Fe2+) to be absorbed in the body because it is soluble. The ferric Fe3+ ions are chelated, such as citrate, phytate, and heme [3]. Therefore, the iron valence state is very important information because it may be related to the effect and toxicity of pharmaceutical products.</p><p>Various vitamins and dietary supplements contain Fe in the form of ferrous fumarate, ferrous sulfate, and ferrous gluconate. Recently several commercial samples of Elevit (Rottendorf Pharma GmbH, Germany), Vitrum (Unifarm, Inc., USA), Children's Multi Vitamin (Target Corporation Minneapolis, USA), Multi For Her (Nature Made Nutritional Products, USA), and Multitabs (Ferrosan A/S, Denmark) containing ferrous fumarate and Sorbifer Durules (Egis, Hungary), Hemofer® (Glaxo-SmithKline Medicines S.A., Poland), Falvit® (Jelfa, Poland), and Vitaral (Jelfa, Poland) containing ferrous sulfate were studied using Mössbauer spectroscopy [4–6].</p><p>In this study, standard samples of ferrous fumarate and gluconate were measured using Mössbauer spectroscopy. The iron valence state and dietary supplements of a Dynabi tablet containing both ferrous fumarate and ferrous gluconate were selected and measured. A tablet of Dynabi is a famous iron supplement (dietary supplement) produced by Korean pharmaceutical company used to treat anemia or other iron deficiencies. It is used to be prescribed to pregnant and parturient women, occasionally in Korea (South).</p><!><p>Mössbauer spectra were measured in a transmission geometry with a moving absorber at a temperature of 295 K and recorded in 1024 channels. For their analysis, spectra with a low iron content and poor signal-to-noise ratio were converted into 512 channels [7]. The spectrometer velocity was calibrated with a high-purity α-57Fe foil. The studied dietary supplement contains iron compounds per tablet (275 mg in Dynabi), which is produced by Dong-A pharmaceutical company in Republic of Korea. The reference compounds of both ferrous gluconate and ferrous fumarate were commercial products. Both compounds are produced by Spectrum Chemical Manufacturing Corporation in USA. All samples consisting of 100 mg powder were distributed homogeneously on a surface of ~3.2 cm2.</p><p>The Mössbauer investigation was made on powdered Dynabi samples. The reference materials such as ferrous fumarate and gluconate were measured as powder.</p><!><p>The molecular formulas of the ferrous fumarate and ferrous gluconate are C4H2FeO4 and C12H22FeO14·2H2O, respectively. The chemical structures are represented at Figure 1.</p><p>The room temperature Mössbauer spectra for the samples are fitted a doublet. However, according to a certificate analysis for the commercial ferrous gluconate, the ferric iron is included at about 2 wt.%. Thus, the Mössbauer spectrum for ferrous gluconate is fitted as two phases. The room temperature Mössbauer spectra of ferrous fumarate and gluconate are presented in Figure 2. The fitting parameters of all spectra are listed in Table 1.</p><p>The room temperature Mössbauer spectrum for ferrous fumarate is fitted as one doublet with a ferrous state, only as shown in Figure 2(a). The Mössbauer spectrum for ferrous gluconate is fitted as two sets of doublets, as shown in Figure 2(b). The separation of the ferrous and the ferric phases is based on values of two parameters such as isomer shifts (IS) and quadrupole splitting (QS). Their values depend not only on the valence state but also on the spin states of the low spin (LS) and high spin (HS) [9]. In ferrous gluconate, the minor phase is contributed by either the ferrous-LS or ferric-HS. The recoil free fractions (f-factor) of the ferrous and the ferric are different. However, this value is fixed at same valent state, though the spin states are different. Both IS and QS for the ferric ion are overlapped with ferrous-LS. However, the identification of the spin state and valence state of the minor phase in ferrous gluconate is unclear, yet [10, 11]. The surface integral of the fitted spectrum is applied to the composition of the sample. The area of the spectrum corresponds to atomic percentage (a.u.%) of the phase in the sample. For the inorganic compounds, the overestimation for the ferric phase was reported to be as high as 15%. A ferric or Fe3+ (LS) phase in the ferrous gluconate reaches ~9 a.u.% in the sample. The peaks corresponding to ferrous iron show a slightly weaker Lamb-Mössbauer factor than those of the ferric phase [9]. These doublets show different values of quadrupole splitting (ΔEQ) and isomer shifts (δ). Additional components connected with Fe3+ in samples may be considered as an impurity or a result of ferrous gluconate oxidation and the formation of ferric gluconate [8].</p><p>The room temperature Mössbauer spectrum of a dietary supplement containing both ferrous fumarate and ferrous gluconate is presented in Figure 3. This spectrum was fitted by tree doublets. The fitting parameters in the spectrum are listed in Table 2.</p><p>The parameters of the two doublets are connected with ferrous gluconate (C12H22FeO14·2H2O). The parameters of the spectrum such as quadrupole splitting and isomer shifts are same as those of pure ferrous gluconate. The remaining doublet is related to ferrous fumarate (C4H2FeO4). The quadrupole splitting for the ferrous fumarate shows a lower value than those of the Fe2+ compound of ferrous gluconate because of a slightly different symmetry. The doublet with δ and ΔEQ values of 1.11 and 2.25 mm/s is attributed to ferrous fumarate. A tablet of Dynabi consists of ferrous fumarate 175 mg (0.97 mole), ferrous gluconate 100 mg (0.21 mole), folic acid 0.4 mg (0.001 mole), and ascorbic acid 309 mg (1.75 mole). Mössbauer spectrum reveals only an iron connected phase. According to a certificate analysis of the commercial product, the molar ratios for ferrous fumarate and ferrous gluconate are above 82.3 and 17.7 a.u.%, respectively. However, the areas of ferrous fumarate and gluconate are fitted to be 87 and 13%, as shown in Table 2. In addition, the area of the ferric phase reached to be ~3.6%. Though the Fe2+/Fe3+ ratio obtained from spectra area of the subspectra does not contribute real concentration of the phase, the overestimation is to be as high as 15% [9]. The absorption area is able to be converted to ~27.7 a.u.% in ferrous gluconate. The ferric phase is significantly decreased compared with pure ferrous gluconate of ~9 a.u.%.</p><p>We may suppose that the doublet with δ and ΔEQ values of 0.12 and 0.9 mm/s is attributed to a ferric phase in ferrous gluconate, according to a certificate analysis. This results in ferrous gluconate oxidation and the formation of ferric gluconate [10, 11]. The doublet with δ and ΔEQ values of 1.09 and 3.06 mm/s is attributed to ferrous gluconate.</p><!><p>The results of the applications of Mössbauer spectroscopy to study industrial samples such as ferrous fumarate, ferrous gluconate, and a dietary supplement demonstrate a wide possibility of this technique. 57Fe hyperfine parameters of the studied pharmaceuticals indicate the existence of major iron ferrous and ferric (or ferrous-LS) compounds. The studied dietary supplement consists of ferrous fumarate (above 87 a.u.%) and gluconate (~13 a.u.%). The Mössbauer spectrum estimated the presence of a ferric compound, owing to impurity and partially modified ferrous gluconate.</p>

PubMed Open Access

Fast 3D shape screening of large chemical databases through alignment-recycling

BackgroundLarge chemical databases require fast, efficient, and simple ways of looking for similar structures. Although such tasks are now fairly well resolved for graph-based similarity queries, they remain an issue for 3D approaches, particularly for those based on 3D shape overlays. Inspired by a recent technique developed to compare molecular shapes, we designed a hybrid methodology, alignment-recycling, that enables efficient retrieval and alignment of structures with similar 3D shapes.ResultsUsing a dataset of more than one million PubChem compounds of limited size (< 28 heavy atoms) and flexibility (< 6 rotatable bonds), we obtained a set of a few thousand diverse structures covering entirely the 3D shape space of the conformers of the dataset. Transformation matrices gathered from the overlays between these diverse structures and the 3D conformer dataset allowed us to drastically (100-fold) reduce the CPU time required for shape overlay. The alignment-recycling heuristic produces results consistent with de novo alignment calculation, with better than 80% hit list overlap on average.ConclusionOverlay-based 3D methods are computationally demanding when searching large databases. Alignment-recycling reduces the CPU time to perform shape similarity searches by breaking the alignment problem into three steps: selection of diverse shapes to describe the database shape-space; overlay of the database conformers to the diverse shapes; and non-optimized overlay of query and database conformers using common reference shapes. The precomputation, required by the first two steps, is a significant cost of the method; however, once performed, querying is two orders of magnitude faster. Extensions and variations of this methodology, for example, to handle more flexible and larger small-molecules are discussed.

fast_3d_shape_screening_of_large_chemical_databases_through_alignment-recycling

Background<!><!>Background<!>Subset extraction<!><!>Algorithms<!>Reference shape selection<!>Alignment-recycling (AR)<!><!>Alignment-recycling (AR)<!><!>Alignment-recycling (AR)<!>Finding the right Transform-Tanimoto<!><!>Finding the right Transform-Tanimoto<!><!>Finding the right Transform-Tanimoto<!>Comparing speed and hit lists<!><!>Comparing speed and hit lists<!><!>Comparing speed and hit lists<!><!>Comparing speed and hit lists<!>Discussion<!>Conclusion<!>Dataset<!>Gaussian shape overlay<!>Diverse reference shape selection<!>Speed comparison<!>Authors' contributions<!>Additional file 1<!><!>Additional file 2<!><!>Acknowledgements

<p>Databases of chemical structures are a key component of chemical information infrastructures. Searching these databases requires specialized methods, for example, to find similar chemical structures.</p><p>There are many ways [1-4] to define "similarity" between chemical structures. Generally, chemical similarity is determined by comparison of "fingerprints" using the Tanimoto equation (Eq. 1). The fingerprints are often binary bit strings with each set bit, or pattern of set bits, representing the presence of a particular topological fragment in a molecule.</p><p>Tanimoto=ABA+B−AB MathType@MTEF@5@5@+=feaafiart1ev1aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacH8akY=wiFfYdH8Gipec8Eeeu0xXdbba9frFj0=OqFfea0dXdd9vqai=hGuQ8kuc9pgc9s8qqaq=dirpe0xb9q8qiLsFr0=vr0=vr0dc8meaabaqaciaacaGaaeqabaqabeGadaaakeaacqWGubavcqWGHbqycqWGUbGBcqWGPbqAcqWGTbqBcqWGVbWBcqWG0baDcqWGVbWBcqGH9aqpdaWcaaqaaiabdgeabjabdkeacbqaaiabdgeabjabgUcaRiabdkeacjabgkHiTiabdgeabjabdkeacbaaaaa@40B7@</p><p>where AB is the count of common set bits and A and B are the count of set bits Similarity measures of this type make it possible to perform searches of chemical databases, containing millions of compounds, in a matter of seconds. While fast, these "2D similarity" methods tend to prefer compounds of similar structural class or topology as the query; however, "3D similarity" methods use geometric constraints and are valued for their ability to find compounds belonging to diverse chemical families [5] (Figure 1).</p><!><p>Examples of output from a shape search using the proton pump inhibitor omeprazole on a subset of PubChem organized by structural class.</p><!><p>The computational cost of 3D methods, however, is dramatically greater than 2D methods, due to the relative complexity of generating, selecting, and comparing various 3D representations of chemical structures. The cost is particularly severe when the comparisons are done by structural overlay, when considering the additional step of determining an optimal 3D overlay. As such, some groups have focused, for example, on extending 2D methods for the discovery of topologically non-obvious similar compounds using reduced-graph approaches [6-8]. With the increase of available computer power, fast 3D structural overlay software, such as ROCS [9], has become attractive for large database screening.</p><p>ROCS performs rapid overlays of 3D chemical structures using atom-centered Gaussians to compute geometric overlap [10]. Similarity is measured with the shape Tanimoto equation (Eq. 2); unlike 2D, an estimate of molecular volume overlap is used, instead of bit counts.</p><p>shapeTanimoto=OABOA+OB−OAB MathType@MTEF@5@5@+=feaafiart1ev1aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacH8akY=wiFfYdH8Gipec8Eeeu0xXdbba9frFj0=OqFfea0dXdd9vqai=hGuQ8kuc9pgc9s8qqaq=dirpe0xb9q8qiLsFr0=vr0=vr0dc8meaabaqaciaacaGaaeqabaqabeGadaaakeaacqWGZbWCcqWGObaAcqWGHbqycqWGWbaCcqWGLbqzcqWGubavcqWGHbqycqWGUbGBcqWGPbqAcqWGTbqBcqWGVbWBcqWG0baDcqWGVbWBcqGH9aqpdaWcaaqaaiabd+eapnaaBaaaleaacqWGbbqqcqWGcbGqaeqaaaGcbaGaem4ta80aaSbaaSqaaiabdgeabbqabaGccqGHRaWkcqWGpbWtdaWgaaWcbaGaemOqaieabeaakiabgkHiTiabd+eapnaaBaaaleaacqWGbbqqcqWGcbGqaeqaaaaaaaa@4CF0@</p><p>where OAB is the volume overlap between conformer A and conformer B, OA is conformer A volume, and OB is conformer B volume</p><p>Several published applications of ROCS demonstrate its usefulness in practical medicinal chemistry projects [11-13]. ROCS can screen the dataset used in this work at the rate of ~1 800 conformers per second per (64-bit 3-GHz Intel dual core Xeon) processor. Although this is a remarkable speed for this kind of software, it can still take hours to perform a single search of a moderately sized 3D database containing millions of conformers.</p><p>Innovative overlay-based approaches [14,15] have been created to avoid brute-force comparison between a query conformer and each and every conformer in a 3D database. One approach [15] involves finding a small "dictionary" of 3D structures that represent the overall diversity of possible 3D shapes. These diverse shapes are then used to create a binary "3D fingerprint" for each conformer in a database, with each set bit corresponding to a computed similarity above a predefined threshold between the diverse shape and the database conformer. This technique shifts the substantial 3D computational overhead into the initial selection of diverse shapes and the generation of the 3D fingerprint for all conformers in the database. For each 3D similarity query, the workflow now becomes identical to that of 2D binary fingerprint methods: compute the fingerprint for the query; loop over the database contents; and determine the bits in common for computation of Eq. 1. After a 3D fingerprint is designed and created, such an approach can significantly reduce the time to search moderately sized 3D databases, e.g., by shape similarity, from hours to minutes.</p><p>There are two major differences between the results from brute-force ROCS shape overlay comparison and the 3D shape fingerprint [15] similarity method. Firstly, the two methods use very different measures for the Tanimoto values and are not guaranteed to give similar results. Secondly, the 3D shape fingerprint similarity approach does not provide a 3D alignment with the query, thus making the results difficult to analyze or visualize. In this study, we attempt to modify an earlier 3D shape similarity approach [15] to mimic results provided by brute-force ROCS similarity searching, but at a fraction of the computational expense. A novel aspect of our method, which we call "alignment-recycling", comes from recycling the translational and rotational matrices resulting from the shape overlay during the initial selection of diverse shapes.</p><!><p>At the time of project initiation, the PubChem Compound [16] database contained approximately 5.3 million unique chemicals and mixtures. We focused our attention on a subset of PubChem by targeting only single-component molecules with size and flexibility below lead-like [17] or drug-like [18] filtering cut-offs. Our strategy was to work with a simple but relevant subset that could be incrementally updated with more challenging compounds in future studies.</p><p>The distributions of non-hydrogen (heavy) atoms and rotatable bonds for PubChem single component structures are presented in Figure 2. For this study, we limited our work to the first half of each distribution, i.e., just those small molecules (less than twenty-eight non-hydrogen atoms) with low flexibility (less than six rotatable bonds). Furthermore, we removed all compounds with incomplete stereochemistry (stereo atoms or bonds), to avoid enumerating multiple stereo-configurations. We also removed ionic forms of structures, since their neutralized forms will be contained in the PubChem compound dataset. The PubChem compound subset selected is summarized in Figure 3. Despite restrictions, the final dataset resulted in approximately one million (1 035 040) unique PubChem compounds representing about 19% of PubChem at the time of dataset extraction. Most of the structures are drug-like organic compounds and, therefore, are well suited for the MMFF94s force field [19] implemented in the 3D conformer generator OMEGA 1.8.1 and 2.0 Beta [20] used for this study.</p><!><p>Distribution of PubChem single component compounds. A) According to number of heavy atoms. B) According to number of rotatable bonds. Yellow bars represent the range of compounds included in the dataset. For a better scaling of the histograms, covalent units above 70 heavy atoms and above 30 rotatable bounds were excluded from the plots.</p><p>Selection of study compound subset from the entire PubChem Compound database.</p><!><p>The alignment-recycling (AR) methodology is intended to obviate performing the optimization required to maximize the volume overlap of the query conformer to each and every conformer in a 3D conformer dataset. This is achieved by selecting representative conformers to completely cover the "shape space" of the 3D conformer dataset. The granularity of coverage is defined by an empirical cutoff named "Design-Tanimoto" (see section Reference shape selection). Each conformer in the dataset is overlaid to each representative conformer and the overlay information is retained, if the similarity with a representative conformer is of sufficient magnitude.</p><p>The empirical criterion to decide if two overlaid conformers can be considered similar is named "Transform-Tanimoto" (see section Alignment recycling). Its value greatly influences the number of reference shapes associated with each conformer. By means of analogy to a binary fingerprint, the Transform-Tanimoto threshold defines when a bit is set.</p><p>To search the dataset by shape similarity, the query fingerprint, to extend the analogy, is compared to the dataset fingerprints to find common reference shapes. The Tanimoto value computed between query and database fingerprints with AR is not that from Eq. 1, as is typical with 2D fingerprint methods and used by the 3D fingerprint method of Haigh et al. [15]. Instead, finding a common reference shape triggers computing, via Eq. 2, the shape Tanimoto between the query conformer and database conformer, as may be performed by a typical brute-force ROCS approach. In our method, the 3D conformer overlay used in computing the shape Tanimoto is generated by simply reusing the transformation, i.e., the rotation matrix and translation vector, from the overlay to the common reference shape. This trivial transformation, while specific to alignment of a reference conformer, when applied, can yield a relatively accurate shape overlay between the query and database conformers without the need to perform the conformer overlay alignment optimization. Usually, when the query and a database conformer are fairly similar, multiple reference shapes are found to be in common. In such cases, all reference shape alignments are reused to find a maximum shape Tanimoto between conformers.</p><!><p>As described in the Method section, we implemented the clustering algorithm of Haigh et al. [15] to select a diverse set of reference shapes. For this study, we chose a Design-Tanimoto value of 0.75, which, according to their work, represented the best trade-off between sampling speed and granularity. This means, by definition, no pair of reference shapes has a similarity above 0.75, after diversity selection, and that every conformer in the entire dataset is associated with at least one reference shape with a shape Tanimoto similarity above 0.75.</p><p>Diverse reference shape selection for the one million compound dataset was performed in two stages. In the first stage, only a single conformer representative generated by OMEGA 1.8.1 [20] was used. The single conformer dataset was entirely covered after the inclusion of 2 458 reference shapes. In the second stage, we sampled the conformational space of each compound using OMEGA 2.0 Beta[20] at an RMSD of 1.0 Å. This generated approximately fifteen million (14 925 817) conformers. The distribution of conformers per compound is strongly skewed towards low values, with 50% of the compounds having six or fewer conformers and only 10% of the compounds accounting for 49% of the total conformer count. Interestingly, 99.8% of the fifteen million conformers in the second stage can be clustered at a Design-Tanimoto of 0.75 using one of the initial 2 458 reference shapes of the single conformer subset, revealing a large amount of shape redundancy in the multi-conformer models. However, the shape space of the remaining 0.2% conformers increases the number of diverse reference shapes from 2 458 to 5 534. This potentially surprising result may be a consequence of the sphere-exclusion algorithm variant used for the reference shape selection. In the attempt to cover the entire dataset shape space with a minimum number of reference shapes, the algorithm tends to leave 'holes' in the shape space, thus producing unequally sampled regions. Given the substantial redundancy of conformer shapes in the multi-conformer model dataset, it is very likely that a large fraction of the additional 3 076 reference shapes is necessary to fill these holes. There is no direct indication that the additional reference shapes resulted from anything more than sampling deficiency, i.e., were not directly attributable to the size or flexibility of molecules. Use of more efficient sampling algorithms designed to avoid empty spaces, e.g., DISE [21], may lead to more efficient shape space coverage than that used in this study.</p><p>Because we aim at selecting a diverse set of shapes, the reference conformers appear to represent particular structural features to a greater extent than are present in the entire dataset. For example, only 20% of the PubChem dataset contain chiral centers; however, 33% of the reference shapes contain a chiral center. Similarly, a (non-exhaustive) trend is found between the dataset and reference conformers for triple bonds (8% versus 37%), lack of aromatic atoms (6% versus 21%), and presence of a ring system with more than six atoms (3% versus 28%). As a consequence, reference shapes generated from structures with less common features tend to cluster fewer database conformers than those coming from compounds with more common features.</p><!><p>AR takes advantage of information created upon comparison of the reference shapes to a conformer during shape fingerprint generation. When a conformer is overlaid on a reference shape, and the computed shape Tanimoto is above the Transform-Tanimoto, the data required to reproduce that alignment are saved (Figure 4). Such information has the form of a three-by-three rotation matrix and a translation vector. In contrast to ROCS, AR can only occur when the query conformer structure is found to have a reference shape in common with a database conformer. The alignment between the query conformer and database conformer is determined using the retained rotational matrices and translational vectors relative to that reference shape.</p><!><p>Conformer alignment to reference shapes. Q is the query conformer. D is the dataset conformer. Reference shapes are numbered from 1 to n. Any alignment with a shape Tanimoto above the Transform-Tanimoto value of, in this case, 0.73 is stored for reuse during database screening.</p><!><p>The procedure to align the query conformer, Q, and database conformer, D, is the following (as depicted in Figure 5). The three-by-three rotation matrix and the translation vector to overlay the database conformer D on the reference shape R are merged into a single four-by-four affine transformation matrix (MRD). Similarly, one can construct the four-by-four affine transformation matrix MQR by using the transpose of the three-by-three rotational matrix and the minus of the translational vector from the overlay of the query conformer Q on the reference shape R. The matrix MQD is produced by the matrix multiply of MQR with MRD. Conformer D is aligned on conformer Q by multiplying the coordinate vector of each atom of D with MQD. In some aspects, the method is conceptually similar to structural alignments performed in 3D-QSAR methodologies for which all the conformers of the dataset are aligned on the same reference template. In our case, the reference template is a reference shape pre-selected during the initial diverse selection.</p><!><p>Alignment recycling. Q is the query conformer. D is the dataset conformer. R is the reference shape. MQD is the 4 × 4 matrix used to align D onto Q. MQD is calculated on the fly through the product of the pre-computed query/reference (MQR) and reference/dataset (MRD) alignment matrices.</p><!><p>Each time an alignment is attempted after transformation matrices combination, the quality of the alignment is evaluated by a single point shape Tanimoto estimation via a Gaussian Grid approximation similar to ROCS, as detailed in the Methods section. The final number of matrix multiplications and alignments depends on the Transform-Tanimoto value as well as the number of reference shapes in the vicinity of the query conformer.</p><p>In practice, a single combination of transformation matrices cannot guarantee a result close to an optimal structural alignment. Some conformers may have different optimal alignments with a reference shape due to structural symmetry; however, the presence of multiple reference shapes greatly increases the chance of finding an alignment very close to the analytical maximum overlap solution. A convenient property of the method is that similar structures tend to have more reference shapes in common than dissimilar ones, thus far more CPU time is dedicated to the alignment of similar structures than for dissimilar structures.</p><!><p>Similarity searches often require a threshold as a simple criterion to prune the hit list. The threshold value is somewhat subjective although a reasonable range of useful values can be deduced from the literature involving ROCS. Rush et al. [11] mention a general rule-of-thumb that a shape Tanimoto value greater than 0.75 provides visual shape similarity, although they used a 0.85 threshold to select their ZipA-FtsZ protein-protein inhibitors. According to a regression plot from Bostrom et al. [22] and our own in-house experience, a RMSD cut-off of 1.0 Å used during conformational sampling with OMEGA 2.0 Beta roughly corresponds to a shape Tanimoto between 0.75 and 0.85. In their virtual screening study, Muchmore et al. [13] found a melanin-concentrating hormone receptor 1 antagonist with nanomolar IC50 at a shape Tanimoto above 0.80. Taking these studies into account, our range of interest in finding similar shapes is limited to ROCS shape Tanimoto between 0.75 and 1.0, alignments with lower similarity values were not considered for this work.</p><p>The suitable Transform-Tanimoto value, which determines if two structures have a reference shape in common and enables alignment via matrix multiplication, was determined empirically. For that, we performed a set of random overlays using both ROCS and alignment-recycling. Our objective was to keep the Transform-Tanimoto value as high as possible to limit the possible number of matrix combinations and, in doing so, save substantially on CPU time. We started by setting the Transform-Tanimoto value to the Design-Tanimoto value, i.e., 0.75. When applying the AR technique, alignment cases where two conformers do not share a common reference shape are assigned a shape Tanimoto value of zero. Because the initial Transform-Tanimoto threshold was not providing the quantity of hits to be consistent with the brute-force approach, primarily due to not finding appropriate reference shapes in common, we progressively decreased the Transform-Tanimoto value by 0.01.</p><p>The relation between ROCS and alignment-recycling at several Transform-Tanimoto values is plotted on Figure 6. The plots are based on ~1.3 million (1 283 211) alignments with a ROCS shape Tanimoto in the range 0.75–1.0. This subset is part of a training set of thirty million random shape-overlays, generated by comparing 2 000 random conformers against 15 000 random conformers from the fifteen million PubChem conformer dataset. The particular nature of the distribution is unveiled by binning the data every 0.01 shape Tanimoto and plotting the isocontour lines at commonly used thresholds for proportion estimation. The scale on the side of each plot gives an indication of the proportion of the data points between each isocontour. All the data points are contained between the minimum and the maximum of each bin, the other isocontour lines (i.e. first and last percentile, decile, and quartile) highlight the intrinsic distribution of the data among each bin. The plots indicate that there is no hard shape Tanimoto limit between finding and not finding a common reference shape between ROCS and AR, but rather some probabilistic distribution. For example, at 0.75 Transform-Tanimoto, 25% of the alignments with a 0.75 ROCS shape Tanimoto do not share an associated AR reference shape. This proportion decreases to less than 10% at a Transform-Tanimoto of 0.74, and less than 1% at 0.73. By means of comparison, an AR reference shape is always found in common for the Transform-Tanimoto values 0.75, 0.74, and 0.73 at ROCS shape Tanimoto values of 0.89, 0.86, and 0.82, respectively. Although we could have further decreased the Transform-Tanimoto to even lower values, in order to further decrease or eliminate the chance of not finding an AR reference shape in common, we felt that a Transform-Tanimoto equal to 0.73 produced satisfying results.</p><!><p>ROCS versus alignment-recycling (AR) shape Tanimoto. A) Transform-Tanimoto equal to 0.75. B) Transform-Tanimoto equal to 0.74. C) Transform-Tanimoto equal to 0.73. The quality of the correlation improves as the Transform-Tanimoto threshold is decreased. Isocontours represent the distribution of AR alignments for each 0.01 ROCS shape Tanimoto interval. Distribution is successively partitioned at first percentile, first decile, first quartile, median, last quartile, last decile and last percentile. The scale on the side of each plot is proportional to the number of alignments in each partition.</p><!><p>A more detailed comparison on the relative performance of AR at 0.73 Transform-Tanimoto is plotted on Figure 7. A negative difference shows that the ROCS overlay is better than the recycled overlay. In contrast to ROCS, AR does not aim at finding the exact global alignment solution. Instead, AR attempts to provide an alignment that is very close with little difference in terms of the shape Tanimoto and graphical display. Consequently, AR overlays are 0.01 shape Tanimoto less than the ROCS overlays 25% of the time (Figure 7: point A); however, this difference is visually minor as shown in Figure 8a. A decrease of 0.03 shape Tanimoto (Figure 7: point B), as shown in Figure 8b, brings some visual separation to the AR and ROCS alignments. At a difference of 0.05 shape Tanimoto (Figure 7: point C), there is a clear visual difference between the overlays, although they are still qualitatively the same (Figure 8c). The degree of alignment quality that may be required by a user depends strongly on the intended use of the alignment, and AR alignments could certainly be used as a very good starting point for subsequent shape overlap optimization, e.g., using ROCS.</p><!><p>Alignment-recycling (AR-0.73) minus ROCS shape Tanimoto. Isocontours represent the distribution of shape Tanimoto differences between AR-0.73 and ROCS alignments for each 0.01 ROCS shape Tanimoto interval. AR-0.73 performs better than ROCS when the difference is above 0, and vice-versa. Distribution is successively partitioned at first percentile, first decile, first quartile, median, last quartile, last decile and last percentile. The scale on the side of each plot is proportional to the number of alignments in each partition. Points A, B and C correspond to the examples shown in Figure 8.</p><p>Impact of lower alignment quality. Examples of ROCS versus AR-0.73 alignments from Figure 7. Left side: Query (yellow), ROCS alignment (dark blue) and AR alignment (cyan). Right side: Same as left side but with query removed.A) AR shape Tanimoto 0.01 worse than ROCS. B) AR shape Tanimoto 0.03 worse than ROCS. C) AR shape Tanimoto 0.05 worse than ROCS.</p><!><p>The distribution in Figure 7 also indicates that about 1% of the time alignment-recycling performs relevantly better (shape Tanimoto difference > 0.01) than ROCS. One possible explanation for this observation is that ROCS gets locked into a local minimum during overlap optimization. A more likely explanation is differences in the numerical precision of the ROCS Grid method versus ours (see section Gaussian shape overlay). Overall, the chance of getting a poor AR alignment, as compared to one produced by ROCS, is relatively rare, when using a 0.73 Transform-Tanimoto value and considering the full 0.75–1.0 shape Tanimoto range.</p><!><p>The test set used here for speed and hit list comparison contained 65 compounds extracted from a dataset of leads and drugs from Oprea et al. [17]. Each test set compound was represented by a single random low-energy conformer. Together, the 65 conformers span a diverse range of shapes derived from simple structures, e.g., salicylic acid, to fairly complex ones, e.g., morphine. The CPU time required to query the fifteen million conformer dataset using the various methods is shown in Table 1. To compute ROCS shape overlays for the entire conformer dataset takes, on average, 2.3 hours, while the time required to perform AR screening at 0.73 Transform-Tanimoto is about 1.3 minutes. This represents more than a 100-fold speedup.</p><!><p>CPU time to query the fifteen million conformer database with a single conformer</p><p>a Total time includes screening plus the required 5 534 initial alignments between reference shapes and the query.</p><!><p>This increase in throughput is not surprising. For each conformer, we are only ever optimizing the overlay to the query for the 5 534 reference shapes. Also, the conformer database reference shape fingerprints are quite sparse, having only 1, 40, or 141 reference shapes set at minimum, average, or maximum, respectively. In contrast, ROCS requires optimizing the overlay of the query conformer to all fifteen million database conformers. As a means of comparison, CPU times required to search the dataset at Transform-Tanimoto values equal to 0.74 and 0.75 are also shown in Table 1. These timings indicate a two- and four-fold decrease, respectively, directly related to a substantial decline in the number of reference shapes considered during screening. This also suggests that each additional 0.01 decrease in the Transform-Tanimoto will increase the AR method CPU requirement by a factor of two.</p><p>The overlap of AR using a 0.73 Transform-Tanimoto value (AR-0.73) and ROCS hit lists were examined to see if the AR-0.73 method produces results similar to ROCS using the shape Tanimoto similarity thresholds 0.75, 0.80, and 0.85. Figure 9 compares the count of compound hits using both methods. As shown in Table 2, AR-0.73 consistently produced ~20% fewer hits than ROCS on average, when using identical shape Tanimoto search thresholds. According to Figure 7, the AR-0.73 shape Tanimoto is, on average, 0.01 less than that resulting from an optimized alignment using ROCS. This suggests that a fairly small decrease in the AR-0.73 screening shape Tanimoto threshold, relative to that of ROCS, should bring the hit count, with similar alignment quality, into sync. Figure 10a shows how the relative count of hits grows as the AR-0.73 screening threshold is decreased, relative to ROCS. Figure 10a also shows that a similar count of query hits may be obtained at AR-0.73 shape Tanimoto values equal to 0.740, 0.792 and 0.844 for ROCS shape Tanimoto equal to 0.75, 0.80 and 0.85, respectively. Comparable hit counts, however, do not imply commonality of hit lists.</p><!><p>Average compound hit list size resulting from querying the fifteen million conformer database with a single conformer</p><p>ROCS versus AR-0.73 number of hits. Box plots showing the distribution of the number of compounds found at 0.75, 0.80 and 0.85 shape Tanimoto cut-offs. Crosses represent the mean of each distribution. AR-0.73 retrieves fewer compounds than ROCS using the same cut-off.</p><p>ROCS versus AR at lower thresholds. AR-0.73 is compared with ROCS at the shape Tanimoto cut-offs 0.75, 0.80, and 0.85, represented by black boxes, and also at lower AR-0.73 cut-offs until a similar number of hits are found, represented by grey boxes. A) AR-0.73 retrieves a similar number of hits (grey boxes cut the 1.0 ratio line) by decreasing the shape Tanimoto threshold by 0.01, 0.008 and 0.006 at 0.75, 0.80 and 0.85, respectively. B) Percentage of ROCS compounds found by AR-0.73. Around 90% of the ROCS compounds are found for hit lists of the same size (grey boxes). C) The union between AR-0.73 and ROCS hit lists is close to the maximum when hit lists have a similar size (grey boxes).</p><!><p>Figure 10b shows the ability of AR-0.73 to reproduce a growing percentage of the ROCS compound hit list as the AR-0.73 screening threshold is decreased, while keeping the ROCS screening threshold constant. As Figure 10c shows, however, that simply decreasing the AR-0.73 screening threshold only improves the union of the two compound hit lists to a point, after which diminishing returns sets in and the hit list overlap becomes worse. This result is expected considering decreasing the AR-0.73 screening threshold results in both ROCS hits missed by AR-0.73 and AR-0.73 hits that would be found by ROCS, if the ROCS screening threshold was not kept constant.</p><p>In each case, the AR-0.73 shape Tanimoto values 0.740, 0.792 and 0.844, originally highlighted as providing a similar number of compound hits, appear to also provide the best trade-off in maximizing the reproducibility of the ROCS hit lists for shape Tanimoto cut-off values equal to 0.75, 0.80 and 0.85. Using these shape Tanimoto values when comparing the two methods compound hit lists, one can expect the average similarity between the AR-0.73 and ROCS hit lists to be in the range of 81–87% (Table 3). The remaining 13–19% hit compounds that were not part of both hit lists belonged to one of the following four categories: hits found by AR-0.73, due to the use of a lower shape Tanimoto threshold, that would have been found by ROCS if the thresholds had been equal (5.7–6.1%); hits found by AR-0.73 but just missed by ROCS, due to finding a suboptimal solution during the maximization of the volume overlap or variation in the grid numeric precision (0.6–2.3%); hits missed by AR-0.73, but found by ROCS, due to the inability to find a reference shape in common (0.0–0.2%); and hits missed by AR-0.73, but found by ROCS, due to suboptimal volume overlap using alignment-recycling only, i.e., without overlap maximization optimization (6.7–10.1%). Regarding this last category, the AR-0.73 missed hits were only narrowly missed, with the missed hits having average shape Tanimoto values of 0.731, 0.787, and 0.840, just 0.009, 0.005, and 0.004 below the AR-0.73 similarity thresholds of 0.740, 0.792, and 0.844, respectively. This shows that even though the hit list intersection appears to decrease slightly with increasing shape Tanimoto value, the missed hits are increasingly proximate to the shape Tanimoto threshold.</p><!><p>Comparison between ROCS and AR-0.73 hit lists when using reduced similarity thresholds for AR-0.73</p><p>a Missed hits artificially caused by using a lower AR-0.73 threshold than ROCS that would have been found, if the ROCS threshold had been the same as AR-0.73.</p><!><p>The observed correction for maximum overlap of AR-0.73 and ROCS hit lists as a function of shape Tanimoto appears to be linear. If this relationship holds across the entire range of ROCS shape Tanimoto values of 0.75 to 1.0, one could employ Eq. 3 to select the appropriate AR-0.73 shape Tanimoto cut-off to use for a desired ROCS shape Tanimoto value to achieve maximum overlap of results.</p><p>STAR-0.73 = 1.04 * STROCS - 0.04</p><p>where STAR-0.73 is the suggested optimum AR-0.73 shape Tanimoto value to use for a corresponding shape Tanimoto value, STROCS, in the range of 0.75 to 1.0.</p><!><p>The AR-0.73 method consistently reproduces ROCS results emphasizing that conformers with similar shapes tend to overlay to each other in a similar way. As such, overlay of two conformers, A and B, to a reference conformer, R, may generate an excellent approximation to the ideal alignment of conformers A and B by simply (re)using the alignments AR and BR. After finding a suitable set of reference shapes, the CPU cost to search for similar conformers across datasets of millions can be dramatically reduced. While efficient, the alignment-recycling method, AR-0.73, outlined in this work does have its limitations.</p><p>AR-0.73, in its current form, cannot be used for sub-shape comparison since global alignments are used. One can, however, readily imagine a subshape-based 3D fingerprint, much like dictionary-based 2D fingerprints. The implementation of such a method is beyond the scope of this work.</p><p>If a similar (enough) reference shape is not present when comparing two conformers, poor shape alignments may result, causing hits to be found by ROCS but missed by AR-0.73. If no reference shape is found to be in common, AR-0.73 cannot produce an alignment.</p><p>As the molecular size and flexibility increase, the number of required reference shapes is likely to increase dramatically to generate accurate shape overlays, which is probably the most important drawback of the AR-0.73 method. Reductions in the Design-Tanimoto can counter large increases in the number of reference shapes; however, in our experience, such a reduction in the Design-Tanimoto threshold, and concomitant reduction of the Transform-Tanimoto, can result in a reduction in the average quality of reproduction of the optimal overlay and an increased computational cost due to the consideration of additional conformers in alignment-recycling portion of the method. The overlay quality can be dramatically improved, in this situation, by slightly altering the methodology provided in this work to perform a post overlay optimization, using the near-optimal alignment-recycling overlay as a starting point for shape overlay optimization, providing substantial computational savings in the absence of such information. This proposed methodology extension may provide the means to apply aspects of the alignment-recycling method to larger and more flexible small molecules by eliminating the requirement that the recycled alignment reproduce the optimal alignment, thus allowing the Design-Tanimoto and Transform-Tanimoto thresholds to be (substantially) reduced.</p><p>Another drawback to AR-0.73 is that the primary computational expense is borne before any shape similarity searches are performed. For the fifteen million conformers used in this study, it took about four CPU years to compute the shape fingerprints using 64-bit 3-GHz Intel dual-core Xeon processors. Computational cost of the fingerprint generation is essentially recovered, however, after performing the same number of searches as there are reference shapes.</p><p>AR-0.73, while substantially reducing the CPU cost of shape similarity searching, adds concomitant demands on storing alignments to the reference shapes that must be available during the search. For the fifteen million conformer dataset, the (non-optimized) storage requirement for the fingerprints and rotational/translational information is 32 GB. If one is not careful, simultaneous access to this data can be a significant bottleneck.</p><p>If the AR-0.73 method is used with a dynamic database of conformers, additional computational costs can be envisioned. As new conformers are added, new reference shapes must be added dynamically whenever existing reference shapes cannot represent a new conformer. Addition of a new reference shape will require the precomputation step of comparing all existing database conformers to the new reference shape. After many new reference shapes are added (> 50% more of the initial total), a complete re-sampling of the reference shapes may be warranted to improve overall search performance through a reduction in the number of reference shapes. Also, for efficiency purposes, as conformers are deleted from the database, care must be taken to ignore reference shapes that no longer represent any database conformer to prevent unnecessary comparisons to a redundant reference shape.</p><p>With the above caveats in mind, the AR-0.73 method as described should be useful to speed the search of any 3D conformer dataset, regardless of size or flexibility. There should be no need to further modify the Transform-Tanimoto and Design-Tanimoto values of 0.73 and 0.75, respectively, to provide, e.g., complementary results to a ROCS search in the shape Tanimoto range of 0.75 – 1.0. The diverse reference shapes used in this work (see Additional files1 and 2) should be useful in helping create the initial reference shapes required to implement this method for arbitrary conformer databases. It is also reasonable to believe that the spirit of this methodology could be made to work using other shape searching packages besides ROCS.</p><p>Alterations to the AR-0.73 parameters, Transform-Tanimoto and Design-Tanimoto, may be made depending on the desired purpose. If one was only interested in use of this methodology as a shape search screen to dramatically reduce the number of conformers considered prior to shape overlay optimization and to provide reasonable starting points for overlay optimization, reduced values of the two parameters could be used, resulting in substantially fewer reference shapes and a significant reduction in the pre-computation cost. If one was only interested in reproduction of hit lists with shape Tanimoto values of 0.90 or greater, the Transform-Tanimoto could be increased closer to the Design-Tanimoto values, providing a further speed up in the shape search speed by reducing the number of conformers considered by alignment-recycling.</p><p>Overall, it appears clear that the AR-0.73 method, while an approximation to the optimal shape overlay, is very capable at routinely producing the vast majority of the ROCS results in a fraction of the CPU time.</p><!><p>One of the main advantages of 3D overlay is that it allows visualization of the superimposed compounds and a better understanding of their similarity. Unfortunately, at the scale of large databases containing millions or billions of conformers, 3D alignment-based similarity searches are reserved to only entities with substantial computing capabilities and modeling resources. Even for such entities, it would be a major breakthrough to get nearly all of the desired alignments in just a couple of minutes using only a single CPU node. The alignment-recycling method described in this work shows promise in dramatically improving the speed of shape similarity searches of large databases through pre-computation of a small subset of shape overlays. Although the pre-computation requires significant computing resources, it is within the reach of modern, yet modest, computer clusters. The pre-computed transformation matrices to obtain the alignments with the subset can be effectively recombined to generate new alignments. Hit lists comparable to the Gaussian shape overlay optimizer ROCS can be obtained 100-times faster with only a small loss in alignment quality for smaller and relatively inflexible molecules. Suggested extensions and modifications to this methodology may prove handy in making 3D similarity a more tractable tool for use on large conformer databases.</p><!><p>The subset of PubChem used for the analysis was extracted using the following protocol:</p><p>• Extract all the live records from the PubChem Compound [16] database</p><p>• Split mixtures into single covalent units</p><p>• Remove each structure not compliant with MMFF94s as implemented in OMEGA [20]</p><p>• Neutralize each ionic structure using a hydrogen atom, if chemically sensible</p><p>• Remove duplicate structures by comparing CACTVS stereo hash codes [23]</p><p>• Remove structures with incomplete stereochemistry (i.e., cis/trans double bonds or R/S stereo centers that are undefined)</p><p>• Remove structures with more than twenty-seven heavy atoms and more than five rotatable bonds</p><p>• Build the single conformer dataset using OMEGA 1.8.1 [20]</p><p>• Build the multiple conformer ensemble using OMEGA 2.0 Beta [20] and RMSD 1.0 Å spacing</p><!><p>The volume of a molecule is generally represented as the finite union of overlapping spheres, each one representing an atom. Although the most intuitive, the hard-sphere model involves complicated analytical expressions and gradient discontinuities. Grant and Pickup [24] overcame these problems by replacing the hard-sphere density function by a soft-sphere Gaussian equivalent, allowing rapid computation of molecular volumes. The smoothness of the Gaussian function and the simplicity of its derivatives greatly facilitate shape overlay optimizations [10]. Grant and Pickup algorithms are currently implemented in the OpenEye OEShape C++ toolkit [25]. The ROCS application is built using this toolkit. When we refer to ROCS, we are actually referring to the OEShape toolkit.</p><p>ROCS provides multiple conformer overlap determination methods. The Grid method is faster when many conformers are fit on a single reference conformer, but it treats all the heavy atoms as carbon, loosing overlap quality in some cases. For initial shape space coverage, we found the Analytic overlap method provided the best trade-off between the speed of the Grid overlap method and the precision of the Exact overlap method. For the alignment-recycling versus ROCS comparison we used the default ROCS Grid approach, as it is the fastest.</p><p>In this study, the atom radii used are Delphi radii, available from the OpenEye OEChem C++ library [26], and only non-hydrogen atoms are considered during shape comparisons. The shape similarity measure used is the Gaussian shape Tanimoto depicted in Eq. 2.</p><p>Alignment-recycling evaluates alignment-quality after each matrix multiplication through a single point shape Tanimoto computation. We used our own implementation of the ROCS Grid method. The results from our method are in essence identical to the results produced by ROCS (R2 = 0.9998, SD = 0.00073, with N = 9 401 620 and maximum difference = 0.012).</p><!><p>The methodology for reference shape selection has been explained in great detail by Haigh et al. [15]. The dataset of conformers are clustered using a simple sphere exclusion algorithm. In the first step, a starting conformer is randomly selected as a reference shape. In the second step, all the conformers with a shape Tanimoto to the current reference shape greater than a pre-defined cut-off value (i.e., the "Design-Tanimoto" value) are assigned to the current reference shape cluster. For all the unassigned conformers, the shape Tanimoto to the most similar reference shape is stored. In the third step, the one conformer with the lowest stored similarity is selected as a new reference shape. The second and third steps are repeated until all conformers are assigned to a reference shape cluster. The Design-Tanimoto defines the resolution of coverage of the "shape space" of the dataset. The structure of the reference shapes is available in supporting information.</p><!><p>The test set from Oprea et al. [17] was extracted from the SD File available in the supporting information. Only 65 compounds met the PubChem subset selection criteria, e.g., for size and flexibility. We generated a 3D conformer model for each compound using OMEGA 2.0 Beta [20] and a RMSD cut-off of 1.0 Å. A single conformer was selected at random for each compound to perform the benchmark speed comparison. The conformer structure coordinates are available in supporting information. CPU time comparisons were performed using 64-bit 3-GHz Intel dual-core Xeon processors on the SuSE Enterprise 9.3 platform.</p><!><p>FF wrote the first draft of the manuscript, developed the algorithms, and performed the testing experiments. EB participated in preparation of the PubChem dataset, design of experiments, and helped write the manuscript. YB participated in the conformer generation, gave advice on the design of experiments, and the writing of the manuscript. SHB supervised the work and provided critical review of the manuscript. All authors read and approved the final manuscript.</p><!><p>diverse set of reference shapes. MDL SD File containing 5 534 reference shapes with their 3D coordinates. The first 2 458 structures were selected using the single-conformer dataset.</p><!><p>Click here for file</p><!><p>65 test compounds used to query the multi-conformer database. MDL SD File containing 65 randomly selected 3D conformers, one for each compound.</p><!><p>Click here for file</p><!><p>The authors are thankful to Anthony Nicholls for constructive comments, Wolf-D. Ihlenfeldt who helped write CACTVS scripts, and OpenEye Scientific Software for intuitive insights and useful 3D tools. This research was supported by the Intramural Research Program of the National Institutes of Health, National Library of Medicine.</p>

PubMed Open Access

Which is the accurate data normalization strategy for microRNA quantification?

Different technologies, such as quantitative real-time PCR or microarrays, have been developed to measure miRNA expression levels. Quantification of miRNA transcripts implicates data normalization using endogenous and exogenous reference genes for data correction. However, there is no consensus about an optimal normalization strategy. The choice of a reference gene remains to be problematic, and can have serious impact on the actual available transcript levels and consequently, on the biological interpretation of data. In this review article we intend to discuss the reliability of use of small RNAs commonly reported in the literature as miRNAs expression normalizers and to compare different normalization strategies used, setting the base for establishing a global standard of miRNAs expression data normalization.

which_is_the_accurate_data_normalization_strategy_for_microrna_quantification?

Introduction<!>Technical challenges<!>Relative data normalization to endogenous reference genes<!>RNU6 and other RNUs as normalizers<!>miR-16 as normalizer<!>Other miRNAs as normalizers<!>Upshot of relative data normalization<!>Relative data normalization to exogenous reference genes<!>Absolute data normalization<!>Ideal data normalization models<!>Conclusion<!><!>Workflow for miRNA qPCR data normalization

<p>MicroRNAs (miRNAs) are small non-coding RNA transcripts of ~22 nucleotides length that exert an important regulatory role of cell activity at the transcriptional level. miRNAs are initially transcribed as immature pri-miRNAs, being processed in the cell nucleus and cytoplasm by Drosha and Dicer, respectively, and loaded into Ago proteins to form the RISC (RNA-induced silencing) complex (1). Mature functional miRNAs loaded in RISC interact with complementary sequences usually located at the 3′-UTR of mRNAs present in the cell cytoplasm, promoting mRNA cleavage or repressing their translation (2). Ultimately, this regulates gene expression by decreasing the production of effector proteins. MiRNAs can repress multiple targets within the same pathway resulting in amplification of their biological effects (3).</p><p>In physiological conditions, miRNAs regulate cell differentiation, cell proliferation and survival, metabolism, among many other functions (2). Additionally, disruption of their expression patterns implicates miRNAs in disease onset and progression (8), such as cancer (9), and their potential role as prognostic and predictive biomarkers in patient management has been described (10). Beyond the functions they exert in the cells that produce them, miRNAs may also be secreted and transferred to other cells, circulating in virtually all body fluids, either in protein complexes or enclosed inside extracellular vesicles, such as microvesicles and exosomes (11).</p><p>Since the collection of miRNAs produced by cells reflect their physiological state, these non-coding RNAs have been greatly explored as disease biomarkers (12). In this way, different methodologies have been applied to characterize qualitatively and quantitatively the miRNAs expression patterns associated with pathological versus normal conditions, such as quantitative real-time PCR (qPCR), microarray screening, northern blotting, ultra-high throughput miRNA sequencing (e.g., small RNA-seq, next generation sequencing), in situ hybridization with locked nucleic acids probes and hybridization in solution with tagged probes (e.g., nCounter® nanoString technology), among many others (13). To accurately determine the levels of analyzed miRNAs, their expression data is usually normalized relatively to endogenous and/or exogenous reference genes. However, different studies use different normalization strategies to report miRNA expression data. This leads to ambiguous data interpretation, misleading conclusions, and erroneous biological predicted effects, impairing comparisons between studies, and consequently none optimal normalization strategy seems to be consensual for the scientific community so far.</p><!><p>The outcomes of miRNA analysis depend on several aspects of the whole process, beginning with the nature of the sample, the way it is collected, preserved and processed, the technical method applied for miRNA detection and the strategy followed for data normalization and analysis (14).</p><p>The accurate comparison of miRNA expression between samples requires that equal amounts of miRNAs are used as input for the detection method applied. Usually, the determination of the quantity of sample input for miRNA detection techniques is based on total RNA quantification, but the real proportion of miRNAs may vary from sample to sample, especially if they have a different origin. Besides, miRNAs integrity is hardly determined, with microfluidic capillary electrophoresis being currently the best method to assess miRNA quality, even with results being compromised by mRNA degradation in the samples (15).</p><p>miRNAs may be isolated by different methods from cultured cells, fresh tissues, frozen and fixed tissues, or as cell-free circulating RNAs from conditioned cell culture media and body fluids, for instance whole blood, plasma, serum, urine, cerebrospinal fluid, etc. (16–20). The collection of fresh human samples requires their preservation by different methods, as coagulation prevention, freezing, fixing and paraffin-embedding. The preservation process induces molecular changes that may lead to global miRNA instability or enrichment/depletion of specific miRNAs in the samples (21). As an example, Kim et al. demonstrated that the capacity to detect endogenous and exogenous miRNAs in plasma samples strongly depends on the method used to prevent blood coagulation (22). In the same way, Farina et al. showed that freeze and thawing cycles differently affected the levels of specific miRNAs in serum samples (23). Furthermore, the time selected for sample collection is also important when analyzing circulating miRNAs, since their physiological levels may vary according to the circadian rhythm, meal ingestion and overall lifestyle (for instance, smoking and drug consumption) (24, 25). Another aspect to take into account when analyzing the levels of miRNAs in pathological conditions is the variation of their levels according to disease stage and progression (26, 27), as well as medical interventions and treatments in course (28, 29).</p><p>Currently, the study of circulating miRNAs strongly focuses on RNA isolated from microvesicles and exosomes actively secreted by cells. For the particular case of exosomes, different methods can be employed for their isolation, being the most common ultracentrifugation (using or not a density gradient), filtration, size-exclusion chromatography and precipitation (using polymeric solutions or beads with immunoaffinity to a exosomal protein marker). Considering their working principle, the different methods lead to an enrichment of specific vesicle subpopulations that likely carry different cargo (25, 30). Of note, large and dense complexes of proteins associated with other biomolecules, such as RNA, have been detected as co-precipitants in exosome pellets isolated by ultracentrifugation. Consequently, miRNA analysis on these samples does not reflect the real intra-exosomal content. Size-exclusion chromatography has been arising as a good isolation alternative to circumvent this issue (31). Immunoaffinity-based methods have been widely used to isolate exosomes from body fluids, targeting proteins described as disease specific biomarkers and carried on vesicles surface. This strategy may bias the definition of specific circulating miRNAs as a disease biomarker, when relatively compared to miRNAs expression in exosomes isolated in healthy conditions by a method of diminished specificity (32). Conversely, disease-specific miRNAs biomarkers may be missed due to their absence in vesicles selected by a particular protein biomarker chosen (33).</p><p>Ultimately, the global analysis of miRNA expression, especially for the confident discovery and validation of disease biomarkers, strongly depends on the size of cohorts/sets of samples analyzed. Very frequently, only small-sized test populations are studied, leading to an erroneous or biased misidentification of biomarkers. In a recently published meta-analysis reviewing miRNA biomarker discovery in non-neoplasic diseases, it was revealed that the majority of the studies published rely on populations under 100 individuals (median size = 69 subjects) (34).</p><p>The methodology chosen for miRNA detection also influences the outcomes of miRNA quantification. Nowadays, the most used methodologies are qPCR and hybridization on microarray platforms, with the former being the gold standard for detection of specific sets of miRNAs of interest and the later mostly applied for large scale profiling.</p><p>Measurement of miRNAs by qPCR is very specific and sensitive, allowing the detection of very small quantities of miRNAs, is relatively inexpensive and commercial ready-to-use kits are widely available. The most common qPCR detection techniques are stem-loop shaped RT-primer Taqman assays (Applied Biosystems), assays using locked nucleic acid primers (Exiqon), and assays with poly-A tailing primers (QIAGEN, Stratagene). With these approaches, only a limited set of miRNAs can be tested in a single reaction, and detection is greatly influenced by the specificity of the primers designed (35).</p><p>Conversely, microarrays allow the probing of a large set of miRNAs simultaneously, nowadays at much more competitive costs, since several off-the-shelf platforms are commercially available from companies as Affimetrix, Agilent and Exiqon, among others. This technique requires an input of higher amounts of RNA, and assay performance may be sometimes compromised by hybridization conditions that are not optimal to the whole probes in test. Likewise, the design of the probes to include in the platform may reveal troublesome, since they have to be specific enough for the target miRNAs and, at the same time, share similarities in the conditions required for hybridization (36).</p><p>Both approaches have advantages and drawbacks associated, but one of the most striking pitfalls is the low correlation between different techniques. In a comparative study by Sato et al. (37), miRNAs in liver and prostate human tissues were profiled using microarray platforms from different companies and expression log-ratios were ranked. Surprisingly, the median rank correlation across platforms was only 0.55, being the highest correlation found of 0.87. Interestingly, the correlation between microarray- and Taqman-based expression data was higher, with a median correlation coefficient of 0.7, and only one of the platforms presenting a correlation coefficient lower than 0.5, supporting the common practice of using qPCR-based techniques as a confirmatory method for the validation of microarray expression data.</p><p>In an attempt to minimize the effect imposed by the factors previously described on miRNA expression levels, an accurate data analysis should be performed, using appropriate reference genes for external and internal variations correction (38).</p><!><p>Quantification of miRNAs has come to the fore, and particularly produced a wealth of literature during the last five years (39). Here, we will focus on the issues that may arise with the data normalization in the gold-standard method RT-qPCR. There are two most commonly used methods to analyze data from real-time qPCR: absolute and relative quantification. Absolute quantification determines the input copy number, usually by relating the PCR signal to a standard curve. Relative quantification relates the PCR signal of the target transcript in a treatment group to that of another sample such as an untreated control (40). For the relative miRNA quantification, the PCR-derived cycle threshold (Cq) of target miRNAs is compared with that of a stably expressed endogenous miRNA from the same sample. The difference between these values is the ΔCq value (41). This normalization approach aims at removing differences due to sampling and quality of the samples. The guidelines for quality control and standardization of RT-qPCR imply the use of an optimal normalization method (42), however there are no universally accepted reference genes or so-called "housekeeping" transcripts for miRNA data normalization. This lack of consensus has resulted in various normalization strategies (43). Some studies had even the erroneous notion of completely abandon an endogenous normalization (44–48), and should of course be approached with some caution. Actually, it was reported that RT-qPCR can be used without endogenous controls to analyze miR-371-3 in the serum of patients with testicular cancer, if the technical procedure is performed under controlled conditions. This study compared the Cq and ΔCq values of miR-371a-3p, miR-372 and miR-373-3p by real-time PCR with and without 18S rRNA (ribosomal RNA) as a reference gene for data normalization. The miRNA transcript levels were measured in 25 testicular germ cell tumor patients, 4 non-testicular germ cell tumor patients and 17 age-matched male controls. A highly positive correlation between Cq and ΔCq values was found in all samples. The highest correlation was found for miR-371a-3p (R2: 0.956) (49). In spite of this positive correlation between Cq and ΔCq values detected in all serum samples by those authors (49), relative data quantification is absolutely indispensable to substantiate robust miRNA data (50). In this study (49), the detected significant correlation is related to the similar (good) quality and handling of the serum samples, but nobody can be sure that this is always the case. There is no evidence for a good sample quality, since total RNA can be partly degraded and bias the results. Therefore, miRNA quantifications without an internal control should be considered critically.</p><p>Fortunately, the majority of studies have carried out a relative quantification that is adapted to compare the expression levels of target genes with the expression of an endogenous reference gene. The choice of an endogenous reference gene is a critical step before starting an experiment, in order to avoid misinterpreted data and to identify true changes in miRNA expression levels. Usually, researchers select their endogenous reference gene for miRNA quantification according to reports in the literature or based on a distinguishable low standard deviation in miRNA microarrays. Considering the most recent and significant reports (Table 1), numerous reference genes, such as small nuclear/nucleolar RNAs, have been commonly used for miRNA quantification (51–54). These small RNA molecules share similar properties, such as RNA stability and size and are abundantly expressed. Although these transcripts display constant expression in single analyses, their expression levels can change under different experimental conditions and may be affected by specific types of disease (53). Therefore, the different normalizers should be first established for different sample types, and the combination of several normalizers might be more appropriate than a single universal normalizer (55).</p><!><p>The small non-coding RNAs RNU6 (RNU6A) and RNU6B are the most frequently used reference genes as normalizers (Table 1). However, RNU6 is not a miRNA, and consequently, does not reflect the biochemical character of miRNA molecules in terms of their transcription, processing and tissue-specific expression patterns (56, 57). In addition, the efficiency of their extraction, reverse transcription, and PCR amplification may differ from that of miRNAs. It has been argued that it is best to normalize miRNAs with reference genes belonging to the same RNA class (43). In this regard, Gee et al. showed that the use of snRNAs (small nuclear RNAs) as reference genes can introduce bias when quantifying miRNA expression and that they are rather important in cancer prognosis. This study (57) measured RNU6B, RNU44, RNU48 and RNU43 for the data normalization in RT-qPCR analyzing breast cancer and head and neck squamous cell carcinoma (HNSCC) patients. The expression of these snRNAs was as variable as the expression of the target miRNAs miR-21, miR-210 and miR-10b, and data normalization to these recommended reference snRNAs introduced bias in the associations between miRNA and pathology or outcome (57). Using microarray-based serum miRNA profiling followed by RT-qPCR, Xiang et al. screened and compared the expression levels of reference RNAs in patients with different tumors and healthy controls. They found large fluctuations in RNU6 expression and a relatively stable expression of miR-16. The difference of ΔCq values of RNU6 between the highest and lowest expression level was 3.29, while that of miR-16 was 1.23. They also subjected the serum samples to different freeze-thaw cycles, and showed that RNU6 expression gradually decreased after 1, 2 and 4 cycles of freezing and thawing, while the expression of miR-16 and miR-24 remained relatively stable (58). Lamba et al. compared the stability of RNU6 and RNU6B in hepatic tissue, and found that both snRNAs were not suitable for the use as endogenous controls for normalizing miRNA data in this tumor type. They used Taqman-based RT-qPCR to quantify the expression levels of 22 miRNAs along with RNU6 and RNU6B in 50 human liver samples. Both software programs NormFinder (59) and GeNormplus identified RNU6 to be among the least stable of all candidate snRNAs analyzed, and RNU6B was also not among the top genes in stability. In their analyses, miR-152 and miR-23b were identified to be the two most stable candidates, and eligible as endogenous controls for data normalization (60). Benz et al. analyzed RNU6B levels in the serum samples of healthy volunteers, intensive care unit patients and patients with liver fibrosis. They demonstrated that serum RNU6B levels displayed a high variability between the cohorts, and consequently, were dysregulated in a disease-specific manner. Most importantly, the expression levels were significantly upregulated in the serum of patients with critical illness and sepsis compared with controls and correlated with established markers of inflammation. In contrast, in patients with liver fibrosis, RNU6B levels were significantly downregulated (56). Furthermore, Ratert et al. also showed that notably, RNU6B is unsuitable for miRNA normalization. Based on miRNA microarray data, a total of 16 miRNAs were identified as putative reference genes. After validation by RT-qPCR, RNU6B, RNU48, miR-101, miR-125a-5p, miR-148b, miR-151-5p, miR-181a, miR-181b, miR-29c, miR-324-3p, miR-424, miR-874 and Z30 were evaluated by the programs geNorm (61), NormFinder, and BestKeeper. These algorithms recommended the combinations of four (miR-101, miR-125a-5p, miR-148b, and miR-151-5p) and three (miR-148b, miR-181b, and miR-874) reference miRNAs for data normalization (62). In miRNA expression studies on renal cell carcinoma, RNU6B was also unsuitable as normalizer. Validation experiments were performed on four miRNAs (miR-28, miR-103, miR-106a, miR-151) together with RNU6B, RNU44, and RNU48. MiR-28, miR-103, miR-106a and RNU48 were proved as the most stably expressed genes, but RNU6B was differentially expressed. If only a single reference gene can be used, miR-28 was recommended as normalizer, while the combinations of miR-28 and miR-103 or of miR-28, miR-103, and miR-106a were preferred (63). Torres et al. investigated the expression of twelve candidate snRNAs (RNU6, RNU44, RNU48, RNU75, RNU54, RNU49, RNU6B, RNU38B, RNU18A, miR-16, miR-26b, miR-92a) in tissue samples of 30 endometrioid endometrial carcinoma patients and 15 normal endometrium samples using RT-qPCR. The stability of candidate endogenous controls was also evaluated using the algorithm programs and an equivalency test. The results were then validated using a larger number of samples. RNU48, RNU75 and RNU44 were identified as most stably and equivalently expressed snRNAs between malignant and normal tissues. Both programs NormFinder and geNorm indicated that these three snRNAs were optimal for RT-qPCR data normalization in endometrioid endometrial tissues. The authors suggested that the average values of these snRNAs could be used as a reliable endogenous control in studies on endometrioid endometrial cancer (64). In a study on Parkinson's disease, RNU24 ranked top of the list of reference genes, followed by Z30. In contrast, miR-103a-3p was ranked as the worst reference gene so that in combination with other reference genes this miRNA was able to bias results. It is important to underline that miR-103a-3p alone or in combination with the other reference genes reversed the direction of the expression levels of the target miRNAs miR-29a-3p and miR-30b-5p. Also, RNU6B was not considered to be a reliable reference gene for Parkinson's disease blood samples, because the efficiency, the r2 and the stability values were too low (51).</p><p>These findings suggest that RNU6 may be unsuitable as an endogenous reference gene in the research of miRNA quantification. In contrast, Han et al. recommended RNU6 as reference gene for the relative quantification of miRNA expression levels in pleural effusion. Following miRNA microarray, the expression levels of candidate reference miRNAs, together with RNU6B, RNU44 and RNU48 were validated in 46 benign pleural effusion samples and 45 lung adenocarcinoma-associated malignant pleural effusion samples by RT-qPCR, and verified using the NormFinder and BestKeeper algorithms. RNU6B, as well as miR-192, were identified as single reference genes and the combination of these genes was preferred for the relative quantification of miRNA expression levels in pleural effusion (65).</p><!><p>Moreover, miR-16 is also frequently used as a normalizer, as it is highly expressed and relatively invariant across various samples (66). To normalize RT-qPCR data, McDermott et al. demonstrated that the combined use of miR-16 together with miR-425 generated more reliable results than using either one of these miRNA alone, or the use of RNU6. Following miRNA profiling of approximately 380 miRNAs, RT-qPCR was performed in 40 breast cancer patients and 20 healthy women. The analysis by geNorm and NormFinder algorithms showed that miR-16 and miR-425 were the most stably combination, achieving the lowest V-value of 0.185 (67). Song et al. analyzed the expression levels of miR-16 together with let-7a, miR-93, miR-103, miR-192, miR-451 and RNU6B in the serum samples of 40 gastric cancer patients and 20 healthy volunteers by RT-qPCR. The geNorm, NormFinder and Bestkeeper algorithms were used to select the most stably expressed reference gene from the seven candidates. The algorithms revealed miR-16 and miR-93 to be the most stably expressed reference genes, with stability values of 1.778 and 2.213, respectively, for serum miRNA quantification across all the patients and healthy controls. The effect of different normalization strategies was also compared. When the data were normalized to the serum volume, there were no significant differences of miRNA levels between patients and controls. However, when the data were normalized to miR-16 and miR-93, or the combination of miR-93 and miR-16, significant differences were detected. These results demonstrated that the use of reference gene for RT-qPCR data normalization has a great effect on the study outcome, and that miR-16 and miR-93 can be recommended as suitable reference genes for serum miRNA quantification in gastric cancer patients (54). In contrast, Schaefer et al. reported that data normalization to miR-16 may lead to biased results using tissue and normal adjacent tissue sample pairs from men with untreated prostate carcinoma collected after radical prostatectomy. In this study (68), the expression of four putative reference genes (miR-16, miR-130b, RNU6-2, SNORD7) was examined with regard to their use as normalizer. Candidate miR-130b and RNU6-2 showed no significantly different expression levels between the matched malignant and non-malignant tissue samples, whereas miR-16 was significantly downregulated in malignant tissue. GeNorm and NormFinder algorithms predicted miR-130b and the geometric mean of miR-130b and RNU6-2 as the most stable reference genes (68). To date, the expression of miR-16 has also been described to be deregulated in different diseases by several other studies (69–74). For example, in osteoclast differentiation, the expression of miR-16 is elevated, and miR-16 was characterized as a regulator of osteolytic bone metastasis.(70).</p><!><p>Using geNorm and NormFinder, Peltier and Latham found that miR-191 was the most consistently expressed miRNA across different human tissues, followed by miR-93, miR-106a, miR-17–5p, and miR-25. In contrast, RNU6 and snRNA5S were the least stable. Indeed, the difference in stability between miR-191 and snRNA5S was a standard deviation of nearly 61 Cq or a difference of 62-fold. Normalization to total RNA mass was also evaluated, but this reference approach ranked behind miR-191 and miR-93 in stability (75). Hu et al. designated miR-1228 as a promising stable endogenous control for quantifying circulating miRNAs in cancer patients. In this report circulating miRNAs was quantified in control individuals (healthy subjects and those with chronic hepatitis B and cirrhosis) and cancer patients (hepatocellular, colorectal, lung, esophageal, gastric, renal, prostate, and breast cancer). GeNorm and NormFinder algorithms as well as coefficient of variability (CV) were used to select the most stable endogenous control, whereas ingenuity pathway analysis (IPA) was adopted to explore the signaling pathways involved. MiR-1228 with CV = 5.4% and minimum M and S values presented as the most stable endogenous control across eight cancer types and three controls. IPA showed miR-1228 to be involved extensively in metabolism-related signal pathways and organ morphology, implying that miR-1228 functions as a housekeeping gene. Additionally, functional network analysis found that miR-1228 was associated with hematological system development, explaining its steady expression in the blood (76). Using TaqMan low density array and NormFinder algorithm, Zhu et al. recommended the combination of miR-26a, miR-221, and miR-22* as the most stable set of reference genes for the evaluation of circulating miRNA in hepatitis B virus (HBV)-infected patients and healthy individuals (77). To determine the levels of candidate reference genes (RNU1-4, RNU6-2, SNORD43, SNORD44, SNORD48, SNORA74A, miR-let-7a-1, miR-106a) for urological malignancies, Sanders et al. analyzed cel-miR-39-spiked serum of prostate cancer patients, bladder cancer patients, renal cell carcinoma patients and control subjects in a RT-qPCR. Recovery of cel-miR-39 (mean 11.6%, range 1–56%) was similar in control subjects and cancer patients. SNORD44 and SNORD74A levels were around the detection limit of the assay. Using the NormFinder and geNorm algorithms, SNORD43 was the most stable reference gene. A combination of two genes (SNORD43, RNU1-4) increased somewhat the stability, indicating that SNORD43 may be a suitable reference gene for the quantification of circulating miRNA in uro-oncological patients (78). For uterine cervical tissues, Shen et al. suggested that miR-23a and miR-191 are the optimal reference miRNAs. Following a microarray assay, the stability of candidate reference genes (miR-26a, miR-23a, miR-200c, let-7a, and miR-1979) was assessed in a cohort of 108 clinical uterine cervical samples by RT-qPCR. MiR-23a was identified as the most reliable reference gene followed by miR-191 (79). To screen suitable reference genes for hepatocellular carcinoma, gastric carcinoma, hepatic cirrhosis and hepatitis B, Tang et al. used GeNorm, Normfinder, BestKeeper, and comparative ΔCq algorithms integrated in RefFinder and measured plasma concentrations of RNU6, let-7a, miR-21, miR-106a, miR-155, miR-219, miR-221 and miR-16 in these patients and healthy volunteers. RefFinder revealed miR-106a and miRNA-21 as the most stably expressed reference genes, with comprehensive stability values of 1.189 and 1.861, respectively, whereas RNU6 was the most unstable miRNA (80). Following miSeq sequencing and qPCR, Wang et al. selected 5 genes (miR-193a-5p, miR-16-5p, RNU6, miR-191-5p and let-7d-3p) for stability analysis using geNorm and NormFinder software. These algorithms identified miR-193a-5p and miR-16-5p as the most stably expressed reference genes. One-way analysis of variance indicated that no significant differences were present in the serum levels among patients with non-muscle-invasive bladder cancer, patients with muscle-invasive bladder cancer and healthy controls. The combined use of miR-193a-5p and miR-16-5p demonstrated that normalization of miRNA data may produce reliable and accurate results for the detection of the significant upregulation of serum miR-148b-3p in bladder cancer (81). To find out the control gene for exosomal miRNA normalization, Le et al. evaluated the expression stability of 11 reference genes in circulating exosomes, and found that the combination of miR-221, miR-191, let-7a, miR-181a, and miR-26a can be an optimal gene reference set for normalizing the expression of liver-specific miRNAs. This combination enhanced the robustness of the relative quantification analyses (82).</p><!><p>Taken together, these findings show the endeavors of developing an optimal endogenous miRNA control to normalize miRNA expression levels. The suggested normalizers for target miRNAs are tissue and species specific. So far, the studies also demonstrate that no consensus exists regarding the normalization to a standard reference gene in various diseases making the miRNA results incomparable. On the one hand, the studies have evaluated and suggested convenient miRNAs, snRNA or rRNAs as ideal candidate reference genes for the data normalization in different diseases using specific algorithms, whereas on the other hand, other studies have shown their deregulation, even for the same disease. In this regard, normalization to a standard reference is most in its infancy. Furthermore, the selection of a normalizer should always follow validation screening tests on a subset of the samples under analyses.</p><p>However, these studies also demonstrate that the use of more than one reference gene increases the accuracy of quantification compared to the use of a single reference gene. More than ten years ago, Vandesompele et al. evaluated 10 housekeeping genes from different abundance and functional classes in various human tissues, and demonstrated that the conventional use of a single gene for normalization leads to relatively large errors in a significant proportion of samples tested. The geometric mean of multiple carefully selected housekeeping genes was validated as an accurate normalization factor (61). Chugh and Dittmer described the potential pitfalls in microRNA profiling and showed that the best way to approach the analysis of miRNA expression data is through global mean normalization of a set of reference genes that may be tissue-specific. This method takes a minimum of three stable housekeeping genes and takes the geometric mean to provide a reliable normalization factor that can control for outliers and differences in abundance between genes (43).</p><!><p>To ensure that miRNA quantification is not affected by the technical variability that may be introduced at different analysis steps, synthetic, non-human spike-in miRNAs are frequently used to monitor RNA purification and reverse transcription efficiencies. The C. elegans miRNA cel-miR-39 is almost exclusively used for normalization to an exogenous reference gene (83–85), but also cel-miR-54 (48), synthetic miRNAs Quanto EC1 and Quanto EC2 (86), as well as the simian virus gene SV40 (87) have been used. Because of the lack of established reference genes, some studies have only carried out miRNA quantification with data normalization to these artificial reference miRNAs. These exogenous miRNAs are usually added to the samples before reverse transcription of RNA, to avoid differences in template quality and warrant efficiency of the reverse transcription reaction. This spike-in method can eliminate some deviations of the experimental process and make the results more reliable, but does not correct for deviations in sampling and quality of samples. A major drawback of using spike-in controls is that only the handling of experiments is considered, but not the quality of tissues, body fluids or extracellular vesicles samples. However, age, body fluids collection, preparation or storing of tissue or fluid samples may result in changes of miRNA levels which may be caused by cell lysis or miRNA degradation. For example, samples with low total RNA quality showed the highest concentrations of miRNA (9, 88). Therefore, the data of this approach should be interpreted with caution. However, when normalization is based on a combination of an endogenous and an exogenous control miRNA, differences in miRNA recovery and differences in cDNA synthesis between samples may be compensated (89).</p><!><p>Some studies have carried out absolute data normalizations, and calculated the miRNA expression using standard curves developed by synthetic miRNAs and melting curves (44–47). Following miRNA expression array, Yau et al. quantified two target miRNAs (miR-221 and miR-18a) in 40 pairs of colorectal carcinoma tissues and 595 stool samples. Their miRNA quantification was based on standard curves and normalized per nanogram of the total input RNA. Derived from standard curves plotted from known amounts of synthetic miR-221 and miR-18a, a technical detection limit of two copies for miR-221 resulted in a Cq value of 42, and a technical detection limit of five copies for miR-18a resulted in a Cq value of 47. All Cq values of >42 for miR-221, and >47 for miR-18a were assigned to '0'. Samples with no amplification of miR-221 or miR-18a were also assigned a value of '0' (45). However, this study should be considered critically, since based on the MIQE (Minimum Information for publication of Quantitative real-time PCR Experiments) guidelines for qPCR, Cq values in the order of 40 and more are not reliable (90). Liu et al. examined the expression patterns of serum let-7 in 214 gastric cancer patients, 222 atrophic gastritis patients and 202 controls. The concentration (copy number) of let-7 was calculated using a standard curve. Melting curve analysis was performed to validate the specificity of the expected PCR products (46). Wang et al. normalized five serum miRNAs (miR-487a, miR-502, miR-208, miR-215 and miR-29b) to the serum volumes, because RNU6 and 5S rRNA were degraded in the serum samples and because of the lack of a consensus housekeeping miRNA for RT-qPCR. Additionally, they assessed the detection limits of the RT-qPCR assay and the dynamic range, and calculated the absolute concentration of target miRNAs based on a calibration curve developed by synthetic miRNAs with known concentration (47).</p><p>Although these studies show interesting results, the applied absolute normalization does not consider the influence of RNA quality on the performance of RT-qPCR. This normalization method is not optimal for an exact quantification of real miRNA amounts and only reliable for samples with a good RNA quality.</p><!><p>The lack of consensus on reference gene selection for miRNA expression data normalization has led to the spread of publications screening for suitable normalizers for samples of defined origins and/or implicated in different pathological conditions. New studies performing miRNA detection by qPCR may greatly benefit from this knowledge and thus a strategic experiment workflow is proposed (Figure 1). Prior to qPCR assay performance, researchers are advised to review the literature using samples of the same origin and physiological state, and processed as similarly as possible, in order to find candidate reference genes. Nevertheless, the suitability of these genes for the set of samples under analysis should always be then validated in a sample subset. If the candidate reference genes are stable, qPCR can be performed and data normalized using their expression levels. If not, the specific samples have to be screened for more suitable reference genes. Ideally, good reference genes should have low standard deviations of expression levels across samples, similar mean and median expression values, and be few affected by storage conditions and sample processing, with a high efficiency of extraction. In these cases, the addition of exogenous xenogeneic or synthetic miRNAs to the samples may be beneficial, and their expression levels may be considered simultaneously with endogenous appropriate normalizers for data correction. Considering the advantages and disadvantages of the different methods for miRNA qPCR data normalization, it is evident that the standard approach of relative data normalization with endogenous and exogenous reference genes benefits from a complementary data absolute normalization. Although it should not be used alone, absolute data normalization allows data correction for limitations intrinsic to the qPCR methodology and gives more insights about the overall status of miRNA expression.</p><p>Alternatively, qPCR data for specific miRNAs expression may be normalized following strategies that take into account the total miRNAs expression in the samples. Very frequently, qPCR analysis of specific miRNAs follows global analysis of total miRNAs expressed in a sample by other technique, as is the case of microarray expression data validation. In this case, information about the whole miRNA content of the sample is available that can be used for global normalization. Mestdagh et al. proposed the use of the mean expression value of whole miRNAs in a sample to normalize miRNA qPCR data (35). In this study, a high-throughput qPCR assay that allows the detection of 430 different human miRNAS and 18 small RNA controls was performed to determine global miRNA expression levels in samples of normal and tumor tissue. The mean miRNA expression value was then calculated considering all the transcripts with a maximal Cq threshold of 35 cycles. A comparative analysis of the stability of mean expression value and common reference genes performed using geNorm clearly showed the adequate application of this strategy for normalization, performing better than genes such as RNU48 and miR-191. Ideally this strategy should be more widespread for qPCR data validation, nevertheless it implies that a large number of genes are always profiled, which may not be cost-effective. To circumvent this issue, the authors propose the selection of reference genes with expression levels similar to values of the global mean expression level previously reported, using their geometric mean for qPCR data normalization. In addition, researchers should bear in mind that, overall, qPCR data normalization may greatly benefit from sample preservation and stability, and thus adequate protocols for sample processing should be standardized at least for samples of the same origin in the same lab.</p><!><p>Normalizing to a reference gene can eliminate differences due to sampling and quality of RNA, and can identify real changes in miRNA expression levels. Therefore, careful validation of reference genes for miRNAs is of crucial importance to obtain accurate miRNA data. Reliability of results reported in the literature, using the wrong reference gene or even performing no data normalization is questionable. It should be emphasized that the applicability of reference genes in some studies does not automatically apply to other studies and that the use of a single reference gene is not sufficient to obtain reliable miRNA data.</p><!><p> Disclosures/Conflict of interest </p><p>Dr. Calin is The Alan M. Gewirtz Leukemia & Lymphoma Society Scholar. Work in Dr. Calin's laboratory is supported in part by the NIH/NCI grants 1UH2TR00943-01 and 1 R01 CA182905-01, the UT MD Anderson Cancer Center SPORE in Melanoma grant from NCI (P50 CA093459), Aim at Melanoma Foundation and the Miriam and Jim Mulva research funds, the Brain SPORE (2P50CA127001), the Center for Radiation Oncology Research Project, the Center for Cancer Epigenetics Pilot project, a 2014 Knowledge GAP MDACC grant, a CLL Moonshot pilot project, the UT MD Anderson Cancer Center Duncan Family Institute for Cancer Prevention and Risk Assessment, a SINF grant in colon cancer, the Laura and John Arnold Foundation, the RGK Foundation and the Estate of C. G. Johnson, Jr. Andreia M. Silva is a recipient of a Ph.D. fellowship from FCT–Fundação para a Ciência e a Tecnologia (SFRH/BD/85968/2012).</p><!><p>Candidate reference genes may be selected from literature and validated in a subset of the samples under analysis. If not appropriate, samples should be profiled to find adequate endogenous reference genes. Alternatively, exogenous miRNAs may be spiked-in into the samples. Also, mathematical indicators calculated from previous miRNA global expression profiles of the same or similar samples, may be used for normalization.</p>

PubMed Author Manuscript

Tuning the Nanoaggregates of Sialylated Biohybrid Photosensitizers for Intracellular Activation of the Photodynamic Response

AbstractIn the endeavor of extending the clinical use of photodynamic therapy (PDT) for the treatment of superficial cancers and other neoplastic diseases, deeper knowledge and control of the subcellular processes that determine the response of photosensitizers (PS) are needed. Recent strategies in this direction involve the use of activatable and nanostructured PS. Here, both capacities have been tuned in two dendritic zinc(II) phthalocyanine (ZnPc) derivatives, either asymmetrically or symmetrically substituted with 3 and 12 copies of the carbohydrate sialic acid (SA), respectively. Interestingly, the amphiphilic ZnPc‐SA biohybrid (1) self‐assembles into well‐defined nanoaggregates in aqueous solution, facilitating cellular internalization and transport whereas the PS remains inactive. Within the cells, these nanostructured hybrids localize in the lysosomes, as usually happens for anionic and hydrophilic aggregated PS. Yet, in contrast to most of them (e. g., compound 2), hybrid 1 recovers the capacity for photoinduced ROS generation within the target organelles due to its amphiphilic character; this allows disruption of aggregation when the compound is inserted into the lysosomal membrane, with the concomitant highly efficient PDT response.

tuning_the_nanoaggregates_of_sialylated_biohybrid_photosensitizers_for_intracellular_activation_of_t

<!>Introduction<!><!>Introduction<!>Synthesis of dendritic ZnPc‐SA biohybrids<!><!>Synthesis of dendritic ZnPc‐SA biohybrids<!>Photochemical characterization<!><!>Photochemical characterization<!>Subcellular localization and phototoxicity experiments<!><!>Subcellular localization and phototoxicity experiments<!>Aggregation studies and intracellular ROS generation<!><!>Aggregation studies and intracellular ROS generation<!><!>Aggregation studies and intracellular ROS generation<!>Conclusions<!>Conflict of interest<!>

<p>V. Almeida-Marrero, M. Mascaraque, M. Jesús Vicente-Arana, Á. Juarranz, T. Torres, A. de la Escosura, Chem. Eur. J. 2021, 27, 9634.</p><!><p>Porphyrinoids are second generation photosensitizers (PS) with a wide range of uses in nanotechnology and materials science,[1, 2] including molecular photovoltaics[3] and biomedicine.[4] Among them, phthalocyanines (Pc) are green/blue dyes comprised of four isoindole subunits, condensed into an aromatic macrocycle with an 18 π‐electron conjugated system. The photophysical and photochemical properties of Pc have been deeply investigated,[5, 6] showing high yields of production of singlet oxygen (1O2) and other reactive oxygen species (ROS), and a strong absorption band in the phototherapeutic window, which render them as excellent agents for PDT.[7, 8, 9] However, their low solubility and strong aggregation in aqueous media limit their medical applications.[10] The therapeutic use of Pc also requires improving biocompatibility and cellular uptake,[11] which can be achieved by conjugation with hydrophilic targeting units of biological origin, leading to so‐called photosensitizing biohybrid materials.[4, 12, 13, 14]</p><p>N‐Acetylneuraminic acid, most commonly known as sialic acid (SA), is a carbohydrate that plays an extensive role in the organism,[15] and has been used for stabilization of nanocarriers towards cancer detection and targeting.[16, 17, 18, 19] Different biological/ synthetic polymers and colloidal nanoparticles with multiple SA units have demonstrated their surface‐protecting/masking properties, resulting in prolonged circulation times and enhanced pharmacokinetic parameters.[20, 21, 22, 23] Based on these advantageous properties, in this paper we describe a combined biohybrid strategy to tackle the issues that normally prevent an efficient photodynamic response of Pc against tumoral diseases. On one hand, decorating the ZnPc macrocycle with SA‐containing dendritic arms to ensure water‐solubility, biocompatibility and cell internalization. On the other hand, modulating the amphiphilic character and thus the aggregation behavior of the resulting ZnPc hybrids, to ensure an efficient ROS photogeneration once internalized. Such modulation results in a superior photodynamic performance of the asymmetrically substituted compound 1 (Figure 1), in experiments with three human superficial tumor cell lines: SCC‐13 (squamous cell carcinoma from face), A431 (squamous cell carcinoma from vulva), and HeLa (cervical adenocarcinoma), as they represent good models of the kind of tissues where this PS could be topically administered. This behavior is due to the PS self‐assembly into inactive, highly hydrophilic nanoaggregates in aqueous media, which facilitates cell penetration likely through endocytosis, and to disruption of aggregation once internalized, activating the PS when it gets inserted in the membrane of lysosomes within the cell (Figure 1b).</p><!><p>a) Structure of the dendritic ZnPc‐SA biohybrid PS 1 and 2. b) Cartoon showing the strategy to enhance the photodynamic performance of 1. The PS is inactive and highly aggregated into hydrophilic nanoparticles outside the cell (step A). Aggregation is disrupted after internalization, by incorporation into the membrane of lysosomes (step B, represented by red arrow), which results in an on‐site activation of the PS photodynamic properties (step C).</p><!><p>Carbohydrates are among the most promising biomolecular kinds that have been combined with Pc for biomedical applications, as they can improve the PS solubility and allow specific delivery to tumor cells.[24] There are various strategies for the linkage of sugars to Pc dyes, which include axial and peripheral substitution of the macrocycle.[25, 26, 27] Different spacers can be used for this purpose,[28, 29, 30, 31] with the possibility to tune the number and distribution of carbohydrate moieties over the Pc scaffold. This can affect the PS amphiphilicity, while the incorporation of multiple units of the same sugar in dendritic nanostructures can help to achieve high water‐solubility and multivalency in the interaction with specific biological receptors.[32, 33, 34] On these bases, we have synthesized two dendritic ZnPc‐SA compounds 1 and 2 (Figure 1a), bearing either 3 or 12 SA moieties, respectively, which present different self‐assembling features and photodynamic properties in buffered aqueous media and in subcellular organelles. The specific behavior of 1 can be ascribed to its amphiphilic character, owing to the asymmetric pattern of sialylated substituents displayed on the Pc macrocyclic platform. The present work therefore reveals how the intimate relationship between self‐assembly and photochemical activity induced by the PS in the subcellular medium where it is localized determines the observed photodynamic response.</p><!><p>The strategy to decorate ZnPc with multiple copies of SA is shown in Scheme 1 and, given the high molecular weight and functionality of the resulting hybrids, it was very challenging from the synthetic point of view. First, the glycodendron 3, bearing three SA moieties (with the carboxylic and alcohol functionalities protected as methyl ester and acetylated groups, respectively), was prepared by copper‐catalyzed Huisgen cycloaddition (click) reaction between 3,4,5‐tris(propargyloxy)benzyl chloride S1 and the azide‐containing derivative S2, followed by nucleophilic displacement of the chlorine atom with sodium azide (Scheme S1 in the Supporting Information). Compound 3 was then confronted in a second click reaction with either the asymmetrically substituted propargyloxy‐ZnPc 4 (prepared as in Scheme S2) or the symmetrically substituted tetraethynyl‐ZnPc 5 (prepared as in Scheme S3), leading to the corresponding dendritic ZnPc with protected SA units (1 p or 2 p). Carbohydrate deprotection led to the final products 1 and 2.</p><!><p>Synthetic route to the ZnPc‐SA dendritic compounds 1 and 2, for which the detailed molecular structure is depicted in Figure 1.</p><!><p>The click reactions of glycodendron 3 with ZnPc 4 or 5 were carried out with careful control of the stoichiometry (1.2 : 1 for the former and 4.8 : 1 for the latter), using CuSO4 ⋅ 5H2O as catalyst and sodium ascorbate as antioxidant agent, in mixtures of THF/water (see the Supporting Information for details). After 48 h, the reaction crudes were treated with QuadraSil MP metal scavenger resin to remove the excess of copper ions. The resulting products were purified by size exclusion chromatography (SEC) using Biobeads as stationary phase, obtaining 1 p and 2 p in 31 and 26 %, respectively. A deprotection step was then carried out by treatment with sodium methoxide in methanol (to remove the acetate groups), followed by addition of water to hydrolyze the SA methyl ester functionalities.[35] The ZnPc‐SA biohybrids 1 and 2 were finally treated with Dowex W50X8 (H+) ion exchange resin, and triturated with acetone to eliminate the excess of acetic acid generated in the deprotection process, obtaining in quantitative yield two blue/green solids that were highly soluble in water.</p><p>The characterization of compounds 1 p and 2 p was performed by 1H NMR, electrospray ionization mass spectrometry (ESI‐MS) and UV/vis spectroscopy (see the Supporting Information). Their 1H NMR spectra showed clear correlation of integrals between the aromatic signals associated to the ZnPc core and those arising from the glycodendron protons (Figure S1). ESI‐MS allowed identifying peaks corresponding to the molecular ion of both compounds in various charged states (Figure S2). For the two final products, 1 and 2, it was not possible to obtain a well resolved 1H NMR spectrum in D2O (and they are not soluble in sufficient concentrations in [D6]DMSO or [D7]DMF), due to their strong aggregation (especially in the case of 2), which flattens and broadens all the signals, and to the presence of regioisomers in the case of 1 (Figures S3 and S5). ]The full deprotection of 1 and 2 was however demonstrated by the absence of peaks corresponding to the acetyl and methyl ester protecting groups in such 1H NMR spectra (Figure S3), while confirmation of the identity of 1 and 2 (with molecular masses of 2112.47 and 5866.62, respectively) came from ESI‐MS (Figure S4). Interestingly, high fragmentation was observed in the MS spectra, as expected from analysis of other glycodendron molecules reported in the literature.[36, 37] Finally, the purity of both compounds was proven by HPLC, confirming the great difference in polarity between 1 and 2 (Figure S5 and comments therein).</p><!><p>The UV/vis spectra of compounds 1 and 2 were recorded in different solvents: DMF, water and phosphate buffered‐saline (PBS) solution (Figure 2a). None of the compounds was aggregated in DMF, their spectra presenting the typical sharp Q‐band of monomeric ZnPc, with absorption maxima at 675 nm and 689 nm, respectively. In water or PBS, their absorbance decreased drastically and an intense hypsochromically displaced absorption showed up in the range of 630–640 nm. This behavior is indicative of H‐aggregation, and so an in‐depth comparative study of the self‐assembly of 1 and 2 is shown two sections below.</p><!><p>a) UV/vis spectra of 1 (left) and 2 (right) in DMF (red), water (green) and PBS (blue) at 7 μM. b) Fluorescence spectra of 1 (blue), 2 (green) and non‐substituted ZnPc (red) in DMF at 1 μM; λ ex=665, 679, and 660 nm, respectively. c) Plot of the decrease in DPBF absorption with time, photoinduced by 1 (blue), 2 (green) and non‐substituted ZnPc (reference compound, red), which correlates with the amount of singlet oxygen produced by each of these PS.</p><!><p>The photosensitizing capacity of the two ZnPc‐SA dendritic derivatives was evaluated in the non‐aggregating solvent DMF. Figure 2b shows the emission spectra of 1 and 2, in comparison to that of non‐substituted ZnPc as reference. Irradiation was performed in all cases at a wavelength 10 nm lower than their absorption maxima, that is at 665 (for 1), 679 (for 2) and 660 nm (for non‐substituted ZnPc). The fluorescence quantum yield (Φ F) values of 1 and 2 were also determined to be 0.1 and 0.12, respectively (see section 1.3 in the Supporting Information). Singlet oxygen quantum yields (Φ Δ), in turn, were determined through the relative method,[38] based on measuring the rate of photodegradation of a chemical scavenger (1,3‐diphenylisobenzofuran – DPBF) that is directly proportional to the formation of 1O2. Figure S6 shows the decay of the scavenger absorption induced by each of these two PS during different irradiation intervals. Decrease in Q‐band intensity or appearance of new bands were not observed, ensuring the PS integrity over the experiment. Under these conditions, plotting the dependence of ln(A 0/A t) with irradiation time (t; with A 0 and A t being the DPBF absorbance values at 414 nm before and after the irradiation time t) afforded a straight line whose slope reflects the PS efficacy to generate 1O2 (Figure 2c), and from which Φ Δ values of 0.46 and 0.48 could be calculated for 1 and 2, respectively (Section 1.3 in the Supporting Information). These Φ Δ values are slightly lower than that of the non‐susbtituted ZnPc reference (Φ Δ (DMF)=0.56),[39] probably due to partial quenching of the excited state by the glycodendron units, yet they are good enough to permit an efficient photosensitizing action.</p><!><p>In order to get an initial assessment of the therapeutic potential of PS 1 and 2, we tested their capacity to enter into a representative cell line of human skin cancer, that is, SCC‐13. Despite of the strong aggregation of these compounds in water and PBS, we hypothesized that the presence of multiple copies of SA over the external surface of their assemblies could be beneficial for penetration across cellular membranes.[40] Subcellular localization of 1 and 2 within SCC‐13 cells was evaluated by fluorescence microscopy, performing colocalization studies with fluorescent markers for specific organelles. To this end, cells were incubated with each dendritic ZnPc‐SA at a concentration of 10 μM, and then briefly washed with PBS. Assuming that no intracellular relocalization occurs, and in order to have a better fluorescence signal, the incubation time in this case was of 18 h, even if phototoxicity was measured after 5 h of incubation. The PS subcellular localization was anyway the same after 5 or 18 h incubation, the differences between both conditions being just in fluorescence intensity. The intracellular emission of both PS was coincident with that of the lysosomes, since a yellowish fluorescence was observed by overlapping of red and green emissions from the ZnPc and LysoTracker®, respectively (Figure 3a and S7a,b). Moreover, the fluorescence intensity was at the same level after incubation of the cells with both PS derivatives 1 and 2, as monitored for more than 100 SCC‐13 cells using the J‐image program (Figure S7c). These results clearly point to an efficient and similar internalization of the two biohybrid PS.</p><!><p>a) Subcellular localization of ZnPc‐SA 1 in SCC‐13 cells after 18 h of incubation at 10 μM of PS (similar data are presented for hybrid 2 in Figure S7); PhC: phase contrast. Red fluorescence is from the ZnPc, green fluorescence from lysosomes, and blue fluorescence from mitochondria. The merged image shows the ZnPC‐SA and each organelle together. A blue (450–490 nm) exciting lamp was used for LysoTracker (top row) and UVA (360–370 nm) exciting lamp was used for MitoTracker (bottom row), while green (545 nm) exciting light was used for the ZnPc‐SA derivative. Scale bar: 10 μm. b) Survival of SCC‐13, A431 and HeLa cells after 5 h of incubation with compound 1 (0.5 μM) followed by different light doses. c) Survival of SCC‐13 cells after 5 h of incubation with compound 2 (0.5 and 10 μM) followed by different light doses. Cell survival was evaluated 24 h after irradiation by the MTT assay. Each point corresponds to the mean value±SD obtained from three independent experiments. *P<0.05, **P<0.01, ***P<0.001.</p><!><p>Next, the phototoxicity of PS 1 and 2 upon red light irradiation was evaluated by the 3‐(4,5‐dimethylthiazol‐2‐yl)‐2,5‐diphenyltetrazolium bromide (MTT) assay. Dark toxicity was first discarded by incubation of the cells for 5 h with two different concentrations of both compounds (0.5 and 10 μM, as the minimum and maximum concentrations employed in the subsequent photosensitization experiments) in absence of light, showing no significant cytotoxicity, with survival rates above 95 % (Table S1). For PDT treatment, cells were incubated for 5 h with the lower non‐cytotoxic PS concentration (0.5 μM) and then exposed to different red light doses (1, 3, 6 and 9 J/cm2, λ=635 nm). MTT evaluation 24 h after the treatment revealed a drastic decrease of cell survival induced by PS 1 with a light dose of 9 J/cm2 in the three cell lines (Figure 3b). The extent of this decrease in cell survival after PS 1 photosensitization is completely comparable to that induced by other effective PS (e. g., LDH‐ZnPcS8 or silicaCe6‐FA), and also to PS precursors such as methyl aminolevulinate, which is used in clinic for several treatments of in situ SCC and pre‐malignant situations.[41, 42, 43]</p><p>In contrast, PS 2 did not induce any phototoxicity at those conditions (Figure 3c, dark gray bars), and both its concentration and the light dose had to be increased (up to 10 μM and 49 J/cm2, respectively) to observe a significant decrease in cell survival (Figure 3c, light gray bars). This remarkable difference in photodynamic performance, with the asymmetrically substituted compound 1 clearly outperforming the symmetric ZnPc‐SA compound 2, could be related to the amphiphilic nature of the former, which let it get inserted in a disaggregated state in the membrane of lysosomes where it localizes within cells. We provide further support to this hypothesis in the next section.</p><p>In view of the promising photodynamic activity of hybrid 1, an in‐depth in vitro study of this compound was performed. Besides SCC‐13, the possible applicability of PS 1 to human superficial cancers was demonstrated with a complete set of internalization, phototoxicity and cell morphology experiments with two other cell lines: HeLa and A431, associated to cervix and vulva tumors, respectively. In fact, one of the main applications of PDT is in non‐melanoma skin cancer, where PS or precursors are topically applied followed by light irradiation.[43] In this direction, the subcellular localization of 1 in HeLa and A431 cells was also found to be in the lysosomes (Figure S8), while these two cell lines resulted even less resistant to the PDT treatment with 1 than SCC‐13 (Figure 3b, dark and light gray bars). The cell morphology of the treated cells was analyzed 24 h after the treatment using phase contrast microscopy, revealing for the three cell lines morphological changes that correlated with the irradiation time (Figure S9). The ratio of cells showing cytoplasmic retraction, with a rounded aspect similar to cells in apoptosis, gradually increased with the light dose applied. Although their visualization was difficult because dead cells get easily detached from the plate, they represented the majority for those irradiated with 9 J/cm2. Such images are in concordance with the results obtained from cell viability assays.</p><!><p>The huge difference in the photodynamic performance of compounds 1 and 2 made us reflect about the reason for such dissimilar behavior. It is common that anionic and hydrophilic aggregated drugs (like 1 and 2) enter the cell through endocytosis, and so are localized in lysosomes.[44] That explains that both PS 1 and 2 go to the lysosome and do not stay at the plasma membrane. However, that aggregated state normally results in a lower photodynamic efficacy shown by lysosomal localization, versus compounds that locate in other organelles, due to the excitonic coupling that occurs when a PS aggregates through π‐π stacking promoted by the aqueous medium. This is what happens with the ZnPc‐SA biohybrid 2, as Figure 2a (right) reveals a strong aggregation by π‐π stacking in water and PBS, and the measured octanol/water partition coefficient (log P OW=−2.23; Section 1.4 in the Supporting Information and Figure S10) confirms the very hydrophilic nature of those aggregates.</p><p>The ZnPc‐SA hybrid 1, in contrast, performs much more efficiently in the PDT experiments, which could be related to its amphiphilic structure, with three isoindole units substituted with tert‐butyl groups and the fourth one substituted with a very hydrophilic SA‐containing dendron. Indeed, Figure 2a (left) shows a lower tendency to aggregate (the band at 630–640 nm is significantly less prominent), and its log P OW value of −1.52 (Figure S10) indicates a lower preference for the aqueous phase. Reverse phase HPLC measurements in a 200‐C18‐42 (ACE 3C18‐AR, 150×3 mm, 3 μm) column, using a water/acetonitrile gradient as mobile phases and a flow rate of 0.5 mL/min, confirmed a great difference in polarity between 2 (t R=0.92 min in water/acetonitrile 60 : 40) and 1 (t R>6.4 min in 100 % acetonitrile; Figure S5), with the advantage that HPLC in such conditions reflects the compounds polarity in their monomeric state. Our hypothesis was then that PS 1 could get inserted into the membrane of lysosomes in its monomeric form,[45] with the macrocyclic core embedded in the hydrophobic region of the lipid bilayer and the hydrophilic dendron exposed to the aqueous environment on the membrane surface. To prove this hypothesis, we carried out a comparative study to correlate the aggregation behavior of hybrids 1 and 2 with their capacity for intracellular ROS generation.</p><p>The supramolecular organization of hybrids 1 and 2 was first examined through temperature‐dependent UV/vis experiments. Aggregation in PBS was too strong to be disrupted by heat, and so solvent mixtures with different proportions of PBS and DMF were tested to study the aggregation mechanism. Figure 4a, top panel, shows for example the absorption spectra of compound 2 in DMF/PBS (35 : 65), recorded after successive temperature decrease steps of 5 °C in the range from 80 to 5 °C, revealing two isosbestic points at 670 and 712 nm. The degree of aggregation (α agg) could be calculated from that set of spectra by a standard method (see Section 1.5 in Supporting Information). Plotting α agg as a function of temperature gave rise to a sigmoidal curve (Figure 4a top, inset), which is indicative of an isodesmic supramolecular polymerization mechanism.[46, 47, 48, 49] Analysis of the data following an isodesmic model resulted in a Kα value of 2.39×105 M−1, while the van't Hoff plot obtained from such analysis allowed determining the rest of thermodynamic parameters of the assembly process (Figure S11 and Table S2).</p><!><p>a) Cooling curves of 50 μM 1 in DMF/PBS (30 : 70; bottom) and 15 μM 2 in DMF/PBS (35 : 65; top) over temperature ranges of 90 to 0 °C and 80 to 5 °C, respectively, with temperature decrease steps of 5 °C. Inset: Fitting of the cooling curve of 2 to an isodesmic supramolecular polymerization model. b) DLS size distribution diagrams of 5 μM 1 (bottom) and 2 (top) in PBS. c) TEM micrographs of 50 μM 1 (bottom) and 2 (top).</p><!><p>For the case of compound 1, cooling curves were recorded at different proportions of DMF/PBS (ranging from 10 to 40 % DMF), but it was not possible to extract the supramolecular thermodynamic parameters, as the obtained α agg data did not fit into a sigmoidal curve for any of those solvent ratios. Along the series it could be observed, however, that the aggregation strength drastically decreased with the amount of DMF. Disruption of aggregates by heat was almost negligible in DMF/PBS 10 : 90 (Figure S12a), while in DMF/PBS 40 : 60, aggregation was almost inexistent at room temperature (Figure S12c). Finally, at a solvent mixture (DMF/PBS 30 : 70) comparable to that in which the optimum cooling curve was obtained for compound 2, it was possible to discern that the aggregation of 1 (Figure 4a bottom panel) is less prominent than that of hybrid 2 (top panel).</p><p>As the self‐assembly features of PS can determine their photodynamic response, the size and morphology of the aggregates formed by 1 and 2 were also characterized. Their hydrodynamic diameter was estimated by dynamic light scattering (DLS), carried out in PBS at a low concentration of PS (5 μM) in order to minimize the ZnPc absorption (Figure 4b). Compound 2 yields large and relatively polydisperse assemblies with an average diameter of 150±76 nm (Figure 4b, top). Compound 1, in turn, yields rather monodisperse, smaller assemblies with an average diameter of 48±11 nm (Figure 4b, bottom). Transmission electron microscopy (TEM) images from 50 μM PS samples deposited on hydrophilic glow‐discharged carbon‐coated copper grids confirmed those observations (Figures 4c and S13). Such a high concentration was used to ensure the presence of sufficient, visible amounts of objects on the grids coming from the sample, assuming that the same type of nanoaggregates would be present at the lower concentrations employed for PDT. While rods with a well‐defined diameter of 8±3 nm and lengths in the range of 40–50 nm were widespread over the grid for compound 2 (Figure 4c, top panel), the aggregation of 1 resulted in a majority of irregular, spherical assemblies with diameters in the range of 10–20 nm (Figure 4c, bottom panel). Importantly, the divergence in sizes observed by DLS and TEM analysis (larger for the former) is a common phenomenon in studies of nanostructures, and relates to the fact that DLS determines diffusion coefficients, from which average particle sizes can be derived, yet various factors can weight differently the resulting particle size distributions.[50, 51]</p><p>The above results provide useful insights about which could be the behavior of 1 and 2 in the intracellular medium. Compound 2 self‐assembles into larger structures that cannot be easily disrupted. The amphiphilic PS 1, on the contrary, tends to form smaller assemblies, which are less robust and could be disassembled in the presence of a membrane, resulting in monomeric PS molecules with no exciton coupling quenching the excited state. To check if that is the case when hybrid 1 reaches the lysosomes in the studied target cells (see above), we analyzed the intracellular ROS formation in SCC‐13 cells by fluorescence microscopy, when subjected to PDT with PS 1 and 2, using the 2,7‐dichloro‐dihydrofluorescein diacetate (DHF‐DA) fluorescent probe (Figure 5).</p><!><p>ROS production detected by the DHF‐DA fluorescent probe after PDT with the ZnPc‐SA biohybrids 1 (top) and 2 (bottom) under irradiation with different red light doses. 1st column: Control experiment in the absence of PS. 2nd column: Cells maintained in the dark, as a second control. 3rd and 4th columns: Cells irradiated with two different light doses. A fluorescence signal indicative of ROS production was observed by fluorescence microscopy (λ ex=436 nm). The intracellular fluorescence intensity was measured by ImageJ (B). *P<0.05, ***P<0.001. PhC: phase contrast images, for analysis of the effect of the PDT treatment on the cells morphology.</p><!><p>SCC‐13 cells were incubated with 0.5 μM 1 or 2 for 4 h, and 6 μM DHF‐DA was then added and incubated for 1 additional h. Immediately after exposure to red light (λ=635 nm, 3 or 9 J/cm2), the cell cultures were analyzed by fluorescence microscopy, under blue light excitation (λ ex=436 nm). As shown in Figure 5, the control samples, consisting of cells in absence of PS (1st column images) or non‐irradiated (2nd column images), exhibited a negligible green fluorescent signal. The experiment with PS 1 under light illumination showed, on the other hand, a clear light dose‐dependent increase in fluorescence (compared to baseline levels, see the inset graph), revealing prominent ROS production after PDT. Comparison with the response of PS 2 actually reveals a substantially higher ROS photoproduction by the ZnPc‐SA biohybrid 1, which confirms its in‐situ intracellular activation when this amphiphilic compound gets inserted into the lysosomal membrane.</p><!><p>Here, we have designed and synthesized a novel dendritic biohybrid PS with structural features and self‐assembly properties that are suitable to address key aspects of photodynamic therapy in the intracellular medium. A challenging synthetic strategy allowed us decorating ZnPc derivatives with multiple copies of SA, a carbohydrate that confers high water‐solubility, biocompatibility, internalization capacity, and enhanced PS transport. Self‐assembly of the resulting biohybrids, either in aqueous solution or in the membrane of target cellular organelles (lysosomes), seems to depend on the number and distribution of SA moieties over the ZnPc macrocycle. This has led to an insightful inference about the interplay between self‐assembly and photochemical activity in the engagement of PS transport, cellular internalization and on‐site activation, which overall are responsible for triggering an efficient photodynamic response against different superficial cancer cell lines.</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

Defining Signal Transduction by Inositol Phosphates

Ins(1,4,5)P3 is a classical intracellular messenger: stimulus-dependent changes in its levels elicits biological effects through its release of intracellular Ca2+ stores. The Ins(1,4,5)P3 response is \xe2\x80\x9cswitched off\xe2\x80\x9d by its metabolism to a range of additional inositol phosphates. These metabolites have themselves come to be collectively described as a signaling \xe2\x80\x9cfamily\xe2\x80\x9d. The validity of that latter definition is critically examined in this review. That is, we assess the strength of the hypothesis that Ins(1,4,5)P3 metabolites are themselves \xe2\x80\x9cclassical\xe2\x80\x9d signals. Put another way, what is the evidence that the biological function of a particular inositol phosphate depends upon stimulus dependent changes in its levels? In this assessment, examples of an inositol phosphate acting as a cofactor (i.e. its function is not stimulus-dependent) do not satisfy our signaling criteria. We conclude that Ins(3,4,5,6)P4 is, to date, the only Ins(1,4,5)P3 metabolite that has been validated to act as a second messenger.

defining_signal_transduction_by_inositol_phosphates

13.1 Introduction<!>13.2 Ins(1,3,4,5)P4<!>13.3 Ins(1,4,5,6)P4<!>13.4 Ins(1,3,4,5,6)P5 and InsP6<!>13.5 Ins(3,4,5,6)P4<!>13.6 Conclusions<!>

<p>The receptor-dependent, Ins(1,4,5)P3-mediated mobilization of intracellular Ca2+ stores (Streb et al. 1983) is now a textbook signal transduction pathway (see Taylor's 2011 chapter). When this signaling activity was discovered, it was certainly a paradigm-altering concept. That novel idea was also highly contagious. After it emerged that Ins(1,4,5)P3 is further phosphorylated to isomers of InsP4, InsP5 and InsP6 (Fig. 13.1), we (Shears 1989) and others (Irvine and Schell 2001; York et al. 2001) proposed that some of these "higher" inositol phosphates might also have important signaling roles. This notion of "orphan signals" (Menniti et al. 1990) certainly captured the imagination (see Ismailov et al. 1996; York et al. 2001). Furthermore, the dramatic digital analogy that paints the inositol ring as a "six-bit signaling scaffold" (York et al. 2001; York 2006) implies that there are signaling roles for many members of the inositol phosphate family. However, the thesis of this review is that, except for one notable exception, we are still a long way from confirming that these "orphan" inositol phosphates truly function as classic second messengers.</p><p>In anticipation that our colleagues might raise their eyebrows in reaction to that last sentence, we will quickly provide some clarification. The second messenger concept (Robison et al. 1968) owes its existence to the discovery of cyclic AMP (Rall and Sutherland 1958; Sutherland and Rall 1958), a diffusible molecule that, in response to an extracellular stimulus, is generated at (or released from) a particular subcellular site. In this example, a change in the concentration of cAMP leads to the modification of a protein kinase activity. If a soluble inositol phosphate is to act as an intracellular signal it must also exhibit a stimulus-dependent change in its concentration. Additionally, the "information" encoded by the inositol phosphate should be converted—transduced—from one chemical form to another. These are significant criteria in the current context.</p><p>Take InsP6 for example. This polyphosphate is an essential co-factor for adenosine deaminase, an mRNA editing enzyme; without InsP6 at its core, the protein does not fold correctly and is devoid of catalytic activity (Macbeth et al. 2005). This is unarguably an important function, but there is no evidence that InsP6 serves as a stimulus-dependent regulator of this enzyme. In fact, in add-back experiments, a maximal effect of InsP6 upon deaminase activity was attained at a polyphosphate concentration of 1 μM (Macbeth et al. 2005), which is 10–50-fold less than the concentration that is always present in cells (Pittet et al. 1989; Oliver et al. 1992; Barker et al. 2004; Bunce et al. 1993). That is, the deaminase should always be fully InsP6-activated by default. A similar argument can be made of the requirement that S. cerevisiae has for InsP6, in order that mRNA can be exported out of the nucleus (Alcázar-Román et al. 2006; York et al. 1999). Here, InsP6 acts by stimulating the ATPase activity of Dbp5, which is envisaged to be a molecular ratchet that pulls mRNA out of the nucleus (Yu et al. 2004). This function is fulfilled by only 0.1 μM InsP6, a small fraction of the total cellular concentration. Again, the role of InsP6 in this process is almost certainly as a cofactor rather than as the "regulator" or "signal" that it was originally proposed to be (York et al. 1999). Once more, let's acknowledge that these are important discoveries. They are just not examples of cell signaling activities.</p><p>Thus, in the current review it is our intention to assess the evidence that a biological response can be attributed to a stimulus-dependent alteration in the levels of a particular inositol phosphate. While Ins(1,4,5)P3 clearly fulfils that criterion, it will not be discussed here as it is the dedicated subject of another chapter (Taylor 2011). The inositol "pyrophosphates" are also reserved for the attention of a separate chapter (Saiardi 2011). Finally, the emphasis in this review is on the inositol phosphates themselves.</p><!><p>Receptor dependent Ca2+ mobilization arises not just by Ca2+ release from intracellular stores, but also through Ca2+ entry across the plasma membrane; the fact that the two processes are tightly linked lies at the heart of the "capacitative calcium entry" hypothesis (Putney 1986), which was subsequently refined and repackaged as "store-operated calcium entry" (Hoth and Penner 1992; Parekh and Penner 1997). In the immediate aftermath of Ins(1,4,5)P3 being discovered to release Ca2+ from intracellular stores (Streb et al. 1983), a 3-kinase was discovered that phosphorylated Ins(1,4,5)P3 (Irvine et al. 1986). At that time, it was not known how Ca2+ release was coupled to Ca2+ entry. Thus, Ins(1,3,4,5)P4 became a prime suspect for signaling Ca2+ entry (Irvine 1986). The initial evidence seemed highly incriminating: stimulus-dependent increases in cellular levels of Ins(1,4,5)P3 are followed shortly afterwards by several-fold increases in levels of Ins(1,3,4,5)P4 (Batty et al. 1985). That is a valuable credential if Ins(1,3,4,5)P4 is to be a cellular signal. Furthermore, Irvine's group (Irvine and Moor 1986) reported that micro-injection of Ins(1,3,4,5)P4 into sea urchin eggs raised the fertilization envelope in a manner that was dependent upon extracellular [Ca2+]. To explain how Ins(1,3,4,5)P4 might mediate a physiological activity that was dependent upon the Ca2+ outside the cell, it was proposed that this inositol phosphate somehow stimulates Ca2+ entry (Irvine and Moor 1986).</p><p>However, as (Schell 2010) noted in a recent review, we are still searching to identify an Ins(1,3,4,5)P4-based signaling cascade, now more than 20 years after Irvine's initial experiments. Not in the least because Ins(1,3,4,5)P4 has been a discordant beast, producing conflicting data in the hands of different groups. For example, the early results obtained from the sea urchin eggs proved impossible to reproduce in subsequent studies (Irvine and Moor 1987; Crossley et al. 1988). Irvine and colleagues were unable to account for this difficulty (Irvine and Moor 1987), and so they turned to other model systems in which they could test their hypothesis (Changya et al. 1989; Morris et al. 1987; Loomis-Husselbee et al. 1996). For example, they used Ca2+-activated K+ channels in mouse lacrimal acinar cells as a readout of the degree of Ca2+ mobilization, and they reported that Ins(1,3,4,5)P4 augmented the Ins(1,4,5)P3-induced Ca2+ response (Morris et al. 1987). Initially, a stimulation of Ca2+ entry by Ins(1,3,4,5)P4 was put forward as the explanation (Morris et al. 1987), although subsequently (Changya et al. 1989) it was suggested to be more likely that Ins(1,3,4,5)P4 somehow augmented intracellular Ca2+ mobilization. That idea was also consistent with a separate series of experiments performed with permeabilized L-1210 cells (Loomis-Husselbee et al. 1996). In contrast, Bird et al. (Bird et al. 1991; Bird and Putney 1996), who also worked with mouse lacrimal cells, reported that Ins(1,4,5)P3 by itself maximally activates Ca2+-activated K+ channels; Ins(1,3,4,5)P4 was not required. After some debate of this topic in the literature (see (Putney 1992; Irvine 1992)), Irvine and colleagues (Smith et al. 2000) later described how the two group's use of slightly different cell preparations might largely explain the contrary data. (Bird et al. 1991) had studied cells that had been in primary culture for up to 24 h. In contrast, Irvine's group had used freshly isolated cells. So, which laboratory used the most appropriate cell preparation? Irvine and colleagues (Smith et al. 2000) accepted that the continuous morphology of the endoplasmic reticulum that is observed in cultured cells reflects a more physiologically-relevant model; this organelle is somewhat fragmented in freshly isolated lacrimal cells. Thus, Irvine et al. (Smith et al. 2000) accept that their model might be less biologically relevant. Nevertheless, they (Smith et al. 2000) argued, the effects of Ins(1,3,4,5)P4 that they observed had to be an exaggeration of a physiological event, rather than an artifact. Perhaps, they suggested, Ins(1,3,4,5)P4 acted by influencing the continuity of endoplasmic reticulum. However, later studies (Brough et al. 2005) did not support that hypothesis.</p><p>In their later studies Irvine and colleagues (Changya et al. 1989; Smith et al. 2000) somewhat de-emphasized a possible role for Ins(1,3,4,5)P4 in regulating Ca2+ entry into cells. Others (Hermosura et al. 2000) have also shown that store-operated Ca2+ entry is not activated by Ins(1,3,4,5)P4. Yet, there continue to be occasional electrophysiological demonstrations of plasma membrane Ca2+ currents being stimulated by Ins(1,3,4,5)P4, although such data have not been placed in a physiologically-adequate context. In one example of this genre, (Luckhoff and Clapham 1992), 30 μM Ins(1,3,4,5)P4 was reported to enhance Ca2+-activated Ca2+ current in excised inside-out patches. However, such high levels of Ins(1,3,4,5)P4 are not biologically relevant. In most cases, cellular Ins(1,3,4,5)P4 levels after receptor activation would not be expected to exceed 3–4 μM (Guse et al. 1993; Barker et al. 1992; Huang et al. 2007), and initial work with optical sensors do not reveal any compartmentalization of Ins(1,3,4,5)P4 synthesis (Sakaguchi et al. 2010). (Luckhoff and Clapham 1992) were able to get lower concentrations of Ins(1,3,4,5)P4 to activate a plasma membrane current when Mn2+ was used as a Ca2+ surrogate, but neither the ion selectivity of this current, nor its biophysical characteristics, match those of store-operated calcium entry (Parekh and Penner 1997).</p><p>During their work with lacrimal cells, Irvine and colleagues (Changya et al. 1989; Smith et al. 2000) went to great lengths to prove that their effects of Ins(1,3,4,5)P4 upon Ca2+ mobilization could not be explained by inhibition of Ins(1,4,5)P3 metabolism (Changya et al. 1989; Smith et al. 2000). It is therefore somewhat ironic that Penner and colleagues (Hermosura et al. 2000) concluded that in certain circumstances inhibition of Ins(1,4,5)P3 metabolism by Ins(1,3,4,5)P4 was indeed a genuine mechanism by which Ca2+ mobilization can be enhanced. More precisely, it was argued that Ins(1,3,4,5)P4 set "a discriminatory time window for coincidence detection that enables selective facilitation of Ca2+ influx by appropriately timed low-level receptor stimulation". However, to achieve these effects in vitro, 5–20 μM concentrations of Ins(1,3,4,5)P4 were required (Hermosura et al. 2000), which, as mentioned above, are well above those that prevail inside cells.</p><p>Ins(1,3,4,5)P4 has also been reported to inhibit Ca2+ signaling. For example, in vitro Ins(1,3,4,5)P4 can inhibit Ins(1,4,5)P3 from binding to its receptor (Hermosura et al. 2000); Putney's group (Bird and Putney 1996) had discovered this phenomenon some years earlier, but they had expressed concern that the levels of Ins(1,3,4,5)P4 that were required were too high to be physiologically relevant. In yet another twist in this tale, Ins(1,3,4,5)P4 was reported to promote Ca2+ re-uptake into the endoplasmic reticulum of permeabilized T51B liver cells (Hill et al. 1988; Boynton et al. 1990). That particular observation has never been reproduced in another cell type, and neither has a mechanistic rationale been developed.</p><p>Genetic perturbation of Ins(1,3,4,5)P4 production has also failed to yield a consistent picture of the polyphosphate's putative role in Ca2+ mobilization. For example, Ca2+ signals in thymocytes were unaffected when Ins(1,3,4,5)P4 synthesis was compromised upon knock-out of the type B Ins(1,4,5)P3 3-kinase gene (ITPKB) (Pouillon et al. 2003). Yet, in B-lymphocytes from Itpkb−/− mice it was reported (Marechal et al. 2007) that there is a reduction in receptor-mediated Ca2+ signaling. Adding further to the confusion, Miller and colleagues (2007, 2009) observed the opposite effect; B-lymphocytes from their knock-out mice showed enhanced Ca2+ mobilization. The latter phenotype does have a quite facile explanation. Ca2+ mobilization might well be expected to be enhanced when the half-life of Ins(1,4,5)P3 is prolonged following loss of a significant route of Ins(1,4,5)P3 metabolism, i.e., an Ins(1,4,5)P3 kinase. Moreover, cells that have more robust Ins(1,4,5)P3 signals might be expected to have more depleted Ca2+ stores, and hence higher rates of store-dependent Ca2+ entry (Jia et al. 2008). However, Miller et al. (2007, 2009) interpreted their data quite differently. They argued that Itpkb−/− lymphocytes have lost an inhibitor of store-operated Ca2+ entry—Ins(1,3,4,5)P4—and that, they concluded, is why Ca2+ mobilization is enhanced in those cells.</p><p>To more directly pursue their hypothesis, Miller et al. (2007, 2009) used a cell-permeant analogue of Ins(1,3,4,5)P4. The addition of this analogue attenuated both receptor-dependent (anti-IgM) and receptor-independent (thapsigargin-mediated) increases in cytosolic [Ca2+] (Miller et al. 2007, 2009). In many of their experiments, Miller et al. reported that cellular Ca2+ levels were reduced immediately upon the addition of cell-permeant Ins(1,3,4,5)P4. The speed of those responses is, perhaps, unexpected, in view of a report (Li et al. 1997) that a delay of at least a minute should be expected, which, apparently, cannot be eliminated by increasing the concentration of the cell permeant analogue. One of the reasons for this lag is the time it takes time for the analogue to diffuse across the membrane. Additionally, each phosphate group is "protected" by two butyryloxymethyl groups (Li et al. 1997), all eight of which must be removed by intracellular esterases before the Ins(1,3,4,5)P4 can be liberated. These technical considerations raise a concern that the immediate effect of the cell permeant Ins(1,3,4,5)P4 might be non-physiological. Yet, it does seems that the response is specific: Miller et al. (2007) added cell-permeant Ins(1,4,5,6)P4 in control experiments. That had no effect upon the cell's Ca2+ responses. Nevertheless, a more direct demonstration of Ins(1,3,4,5)P4 inhibiting a physiologically-relevant pathway for Ca2+ entry, for example in an electrophysiological assay, would be helpful.</p><p>Is it possible to rationalize all of these different reported effects of Ins(1,3,4,5)P4 upon Ca2+ mobilization? And what are we to make of experimental observations where even the order of addition of Ins(1,4,5)P3 and Ins(1,3,4,5)P4 determines whether or not an effect upon Ca2+ fluxes is observed (Loomis-Husselbee et al. 1996)? One group (Hermosura et al. 2000) has argued that the somewhat confused and inconsistent literature reflects the actions of Ins(1,3,4,5)P4 being complex and multifaceted. Perhaps, they say, the precise molecular actions of Ins(1,3,4,5)P4 are cell-specific, varying with the strength of receptor activation, or subcellular localization of Ins(1,4,5)P3-metabolic enzymes, or heterogeneity of intracellular Ca2+ stores, or differences in Ins(1,4,5)P3 receptor subtypes, or other regulatory factors. However, we put it to the jury that the case for Ins(1,3,4,5)P4 being a cellular signal for Ca2+ mobilization is unproven. Judgement should be reserved until a specific signaling activity can be reproducibly demonstrated using physiologically-relevant concentrations of Ins(1,3,4,5)P4 and, moreover, an Ins(1,3,4,5)P4-sensitive signaling entity with a defined role in Ca2+ signaling would need to be identified.</p><p>Of course, Ins(1,3,4,5)P4 could have other signaling roles that do not involve Ca2+. Indeed, Itpkb−/− mice are immunologically compromised (Pouillon et al. 2003). For example, thymocytes in Itpkb−/− mice synthesize almost no Ins(1,3,4,5)P4 and their developmental program fails (Pouillon et al. 2003). Itpkb−/− mice also exhibit defective B-lymphocyte development (Miller et al. 2007; Marechal et al. 2007) and neutrophil migration is compromised (Jia et al. 2007). We will briefly discuss three recent developments that suggest there may be Ins(1,3,4,5)P4 "receptors" that can help explain the nature of these Itpkb−/− phenotypes.</p><p>One intriguing protein to which Ins(1,3,4,5)P4 binds tightly and specifically is the Ras-Gap that was originally named GAP1IP4BP (Cullen et al. 1995). An exhaustive study in which the levels of GAP1IP4BP were genetically manipulated (Walker et al. 2002) led to the conclusion that this protein did not exert any influence upon Ca2+ mobilization. Nevertheless, there has recently been renewed interest in GAP1IP4BP, which has been re-christened as RASA3. The Ins(1,3,4,5)P4-binding region of RASA3 is now known to be a PH domain, which also binds to PtdIns(4,5)P2 (Cozier et al. 2000). Could competition between Ins(1,3,4,5)P4 and PtdIns(4,5)P2 have some signaling significance? In CHO cells, receptor-dependent PLC activation (and hence Ins(1,3,4,5)P4 accumulation) did not affect the membrane-association of GAP1IP4BP/RASA3 (Cozier et al. 2000). However, in a subsequent study with COS cells (Marechal et al. 2007), in which ITPKB was over-expressed, RASA3 was dislodged from the plasma membrane upon PLC activation. Furthermore, 30 min pre-incubation of cells with cell-permeant Ins(1,3,4,5)P4 also caused RASA3 to translocate from the plasma membrane to the cytosol (Marechal et al. 2007). This may represent a new signaling function for Ins(1,3,4,5)P4 in B-lymphocytes (Marechal et al. 2007). It was proposed that Ins(1,3,4,5)P4 regulates the intracellular location and hence the activity of a RASA3-ERK signaling pathway that controls pro-apoptotic BIM gene expression (Marechal et al. 2007). As noted elsewhere (Sauer and Cooke 2010), the further development of this hypothesis would be helped by a demonstration that Ins(1,3,4,5)P4 acts in this manner in lymphocytes.</p><p>Ins(1,3,4,5)P4 has also been reported to be an inhibitory signal for neutrophil function, by competing with PtdIns(3,4,5)P3 for binding to the PH domain of AKT (Jia et al. 2007). A cell-permeant Ins(1,3,4,5)P4 analogue was shown to diminish receptor-dependent recruitment of AKT-PH domain to the plasma membrane (Jia et al. 2007). Moreover, neutrophils from Itpkb−/− mice exhibit up-regulated AKT activity because, it was argued, more of that kinase can translocate to the plasma membrane in the absence of Ins(1,3,4,5)P4 (Jia et al. 2007).</p><p>Another proposed "receptor" for Ins(1,3,4,5)P4 is the interleukin-2 tyrosine kinase ITK, which phosphorylates and activates PLC-γ. For ITK to regulate PLC-γ, the kinase must first translocate to the plasma membrane, courtesy of the affinity for PtdIns(3,4,5)P3 of the protein's pleckstrin homology domain. Interestingly, this receptor-dependent translocation process is inhibited in thymocytes prepared from Itpkb−/− mice (Huang et al. 2007). In that latter study it was further reported that, in vitro, Ins(1,3,4,5)P4 promotes the association of the PH-domain of ITK with an immobilized, short acyl-chain analogue of PtdIns(3,4,5)P3. An attractive feature of this phenomenon is that it was observed when using concentrations of Ins(1,3,4,5)P4 (1 μM) that are physiologically relevant. However, it is not clear exactly how Ins(1,3,4,5)P4 has this effect. In fact, as discussed above, it would normally be expected that Ins(1,3,4,5)P4 would compete with PtdIns(3,4,5)P3 for the same ligand-binding domain. To resolve this apparent paradox, Huang et al. (2007) have proposed that, if ITK were to oligomerize, then the binding of Ins(1,3,4,5)P4 to one subunit might allosterically enhance the affinity of another subunit for PtdIns(3,4,5)P3. However, that idea is not consistent with FRET analysis that has revealed cytoplasmic ITK to be monomeric in vivo (Qi et al. 2006; Qi and August 2009). Oligomerization of ITK only occurs after the kinase has already translocated to the plasma membrane (Qi et al. 2006). The folded state of monomeric ITK (Qi and August 2009) suggests that it alters its conformation upon its transfer to the plasma membrane. Thus, the alternative "induced-fit" hypothesis that Huang et al. (2007) have proposed is arguably a more likely explanation for their in vitro data. Here, initial binding of Ins(1,3,4,5)P4 is suggested to alter the conformation of ITK so that it gains an increased ligand affinity, but particularly for PtdIns(3,4,5)P3. The viability of the latter proposal is critically dependent upon accurate determinations of the relative affinities of PtdIns(3,4,5)P3 and Ins(1,3,4,5)P4. It would therefore be useful to determine these binding parameters using an in vitro technique that is closer to a physiological context than is the use of a soluble PtdIns(3,4,5)P3 analogue immobilized to beads. The preferred method (Narayan and Lemmon 2006) is to incorporate "natural" PtdIns(3,4,5)P3 into phospholipid vesicles, and then determine the affinities of the lipid (and the competing headgroup) by using surface plasmon resonance. Additionally, rather than characterizing just the PH domain fragment of ITK (Huang et al. 2007), it would be more physiological to use full-length protein.</p><p>In all three of the examples described above—RASA3, AKT and ITK—it will be important to address the concern (Irvine et al. 2006) that an Itpkb−/− phenotype could be an unpredictable consequence of a loss of non-catalytic (scaffolding?) activities of the type B Ins(1,4,5)P3 3-kinase, rather than a reduction in Ins(1,3,4,5)P4 levels. This question could be pursued by studying if the phenotype of the Itpkb−/− cells can be rescued by expression of ITPKB, and not by a kinase-dead Itpkb mutant. That being said, it also needs to be established that the biological effects that have been attributed to Ins(1,3,4,5)P4 are not actually performed by one or more of its metabolites (Fig. 13.1). For example, the elimination from cells of Ins(1,4,5)P3 3-kinase activity can lead to a reduction in levels of Ins(1,3,4,5,6)P5 and InsP6 (Leyman et al. 2007). Ins(1,3,4)P3 is another important metabolite of Ins(1,3,4,5)P4. Loss of the latter in Itpkb−/− cells (Pouillon et al. 2003) will uncouple the link between PLC activity and the Ins(3,4,5,6)P4-signaling cascade (see below). That is, cells that have reduced ITPK activity will also be encumbered by an inability to synthesize the Ins(3,4,5,6)P4 signal. That could modify cell function in a number of ways (see below). To take one pertinent example, the ClC3 Cl− channel that Ins(3,4,5,6)P4 regulates (Mitchell et al. 2008) plays an important role in neutrophil migration (Volk et al. 2008) which, as mentioned above, is defective in Itpkb−/− cells.</p><!><p>In yeast (Odom et al. 2000; Saiardi et al. 2000) and in flies (Seeds et al. 2004), Ins(1,4,5,6)P4 is formed from Ins(1,4,5)P3 by the kinase activity of Ipk2. In a 2000 study (Odom et al. 2000) evidence was presented that this synthesis of Ins(1,4,5,6)P4 was necessary for the function of an ArgR-Mcm1 transcriptional complex that regulated the expression of ornithine transaminase. The authors of that study assayed transcriptional control in wild-type and ipk2Δ cells using "growth on ornithine as a sole nitrogen source". However, this interpretation of the data has been disputed by others (Dubois et al. 2000), who reported that the slower growth of ipk2Δ cells was a general phenotype rather than being specific to the nutrient source. This is an important point; if slowed cell growth were to be a general phenotype, then it would no longer provide a specific readout of the expression of ornithine transaminase. Unfortunately, it has never been resolved why these two groups came to different conclusions. Yet, despite the controversy, the idea that the catalytic activity of Ipk2 might regulate transcription was taken up in subsequent studies: O'Shea and colleagues found that this kinase activity regulated Pho5 transcription (Steger et al. 2003). Additional genetic experiments indicated it was chromatin remodeling that was regulated by either Ins(1,4,5,6)P4 and/or Ins(1,3,4,5,6)P5 (another product of Ipk2 activity). These authors reported an increased accessibility of a Cla I restriction site in the Pho5 promoter in nuclei that were isolated from cells shifted to Pho5 inducing conditions (Steger et al. 2003). This increased Cla I accessibility was impaired in the ipk2Δ cells (Steger et al. 2003).</p><p>But are these phenotypes the direct consequence of altering the expression of the kinase activity of Ipk2? There are far-reaching effects upon many mRNA species in yeast strains in which the catalytic activity of Ipk2 is compromised (El Alami et al. 2003), and so it is important to separate primary regulatory events from secondary consequences. Additionally, metabolic homeostasis utilizes regulatory processes that control the expression of genes encoding metabolic enzymes. Thus, there are many links between gene regulation and metabolic status (McKnight 2003). It might be considered inevitable that control over phosphate supply to yeast, and the regulation of Pho5 expression, must be intertwined with regulation of inositol phosphate synthesis, which of course is a phosphate-consuming process. The big question, therefore, is whether this apparent effect of Ipk2 upon chromatin remodeling reflects a global metabolic control process, or instead is this really a more specific utilization of inositol phosphate turnover to control gene expression? This query could be resolved if we had a molecular justification for the intriguing genetic effects that were described by O'Shea and colleagues (Steger et al. 2003).</p><p>A molecular mechanism by which Ipk2 might control Pho5 expression has been put forward; it was proposed that Ins(1,4,5,6)P4 and Ins(1,3,4,5,6)P5 directly stimulate "nucleosome sliding" (Shen et al. 2003; Steger et al. 2003). Nucleosomes are the basic repetitive unit of chromatin: histone octomers around which are wrapped about 150 bp of DNA (Becker and Hörz 2002). Regulatory elements can be exposed when nucleosomes are nudged along the DNA helix by ATP-consuming, nucleosome remodeling factors. Wu and colleagues (Shen et al. 2003) studied nucleosome movement along a Drosophila hsp70 DNA fragment driven by the yeast SWI/SNF chromatin remodeling complex. It was reported that 500 μM of either Ins(1,4,5,6)P4 or Ins(1,3,4,5,6)P5 stimulated this nucleosome sliding (Shen et al. 2003). Unfortunately, as discussed elsewhere (Shears 2004), Wu and colleagues (Shen et al. 2003) studied the effects of inositol phosphates at concentrations that are 500–1700 times higher than estimated cellular levels. In such circumstances, it is difficult for us to accept that those effects are physiologically relevant.</p><p>It is even more difficult to imagine Ins(1,4,5,6)P4 being a signal that controls transcription in higher organisms. The mammalian homologue of Ipk2—often called IPMK—phosphorylates Ins(1,4,5)P3 to Ins(1,3,4,5)P4 (Fig. 13.1) rather than to Ins(1,4,5,6)P4 (Saiardi et al. 2001; Nalaskowski et al. 2002). So mammalian cells do not use IPK2/IPMK to regulate levels of Ins(1,4,5,6)P4. It is possible for Ins(1,4,5,6)P4 to be synthesized by dephosphorylation of Ins(1,3,4,5,6)P5, probably by the cytoplasmic and nuclear pools of PTEN that, in these particular locations, cannot access the alternative and better-known substrate, PtdIns(3,4,5)P3 (Caffrey et al. 2001; Otto et al. 2007). In any case, there is no evidence that cellular levels of Ins(1,4,5,6)P4 are receptor regulated (Menniti et al. 1990). (It has been published that Ins(1,4,5,6)P4 levels are elevated in src-transformed fibroblasts (Mattingly et al. 1991), but that result has not been studied further).</p><p>There is an environmental pathogen that can perturb Ins(1,4,5,6)P4 metabolism: Some years ago it was noted that the invasion of epithelial cells by Salmonella strongly activated the dephosphorylation of Ins(1,3,4,5,6)P5 to Ins(1,4,5,6)P4 (Eckmann et al. 1997). The evidence indicated that Ins(1,4,5,6)P4 might augment the secretion of salt and fluid that accompanies Salmonella infection (Eckmann et al. 1997). Whether or not this phenomenon might have physiological rather than just pathological relevance has not been established. Subsequently, it was demonstrated that one of the proteins that is required for the pathogen's virulence, SopB, was responsible for dephosphorylating Ins(1,3,4,5,6)P5 (Norris et al. 1998). A later study (Zhou et al. 2001) established that the main product was Ins(1,4,5,6)P4. However, there was no evidence that virulence itself depends upon Ins(1,3,4,5,6)P5 dephosphorylation (Zhou et al. 2001). Instead, cell invasion by Salmonella appeared to require inositol lipid dephosphorylation by SopB (Hernandez et al. 2004). Indeed, it now seems possible that Ins(1,3,4,5,6)P5 is little more than an off-target substrate for SopB. In any case, the groups that work with SopB now focus on its role in metabolizing inositol lipids rather than the inositol phosphates (Hernandez et al. 2004). Taking all these data into account, we conclude that Ins(1,4,5,6)P4 is not qualified to be described as a cellular signal.</p><!><p>Most nucleated cells synthesize 15–50 μM of both Ins(1,3,4,5,6)P5 and InsP6 (Pittet et al. 1989; Oliver et al. 1992; Barker et al. 2004; Bunce et al. 1993). Undoubtably, the initial proposals that Ins(1,3,4,5,6)P5 and InsP6 might be cellular signals (Heslop et al. 1985; Vallejo et al. 1987; Michell et al. 1988) were strongly influenced by the manner in which these polyphosphates were first discovered in animal cells, that is, as a consequence of studying metabolism of Ins(1,4,5)P3. However, intracellular signals typically are expected to exhibit significant stimulus-dependent changes in their concentrations. In contrast, cellular levels of Ins(1,3,4,5,6)P5 and InsP6 do not respond acutely to most extracellular stimuli, and even when they do, 25–35% changes in their concentrations seem to be an upper limit (Larsson et al. 1997; Pittet et al. 1989).</p><p>One dramatic exception emerged in a study (Gao and Wang 2007) of certain signaling events that lie downstream of the so-called Frizzled receptors. Activation of Frizzleds by the Wnt ligands regulates many aspects of embryonic development and adult tissue homeostasis. Wnt ligands can activate PLC and stimulate inositol phosphate accumulation (Slusarski et al. 1997), but Ins(1,3,4,5,6)P5 would normally be expected to be well-insulated from that response (Menniti et al. 1990). It was therefore unexpected when Gao and Wang (Gao and Wang 2007) demonstrated that Ins(1,3,4,5,6)P5 levels increased up to 2.5-fold in the few minutes following activation of the (over-expressed) rat Fz1 receptor by Wnt3a. In vitro data indicated that the biological consequence of this increase in Ins(1,3,4,5,6)P5 was activation of casein kinase II (CK2) and inhibition of GSK3β. The maximally effective concentration of Ins(1,3,4,5,6)P5 in each case was approximately 50 μM, which is approximately what is normally estimated to be present inside mammalian cells (Oliver et al. 1992; Pittet et al. 1989). Both of those effects of Ins(1,3,4,5,6)P5, if they occurred in vivo, would be expected to stabilize β-catenin, enhancing its transcriptional response to Wnt signaling (Gao and Wang 2007).</p><p>Let's first discuss the proposed regulation of CK2. It has been known for some years that Ins(1,3,4,5,6)P5 can activate CK2 in vitro, although it was originally reported by Solyakov et al. (2004) that InsP6 is more efficacious. However, neither Ins(1,3,4,5,6)P5 nor InsP6 affects the activity of purified, native CK2 (Solyakov et al. 2004; Gao and Wang 2007). Instead, it was reported that the inositol phosphates act by reversing the effect of an uncharacterized, heat-stable inhibitor of CK2 that is present in cell lysates (Solyakov et al. 2004). The lack of insight into either the nature of the inhibitor, or the mechanism of action of the polyphosphates, has prevented this hypothesis from developing further. Moreover, it is a popular viewpoint in the CK2 field that this kinase is normally constitutively active and therefore has no requirement to be stimulated (Ruzzene and Pinna 2010). Even if that prevailing opinion were to be incorrect, it is hard to see how a stimulus-dependent increase in Ins(1,3,4,5,6)P5 levels would have any effect upon CK2 that should already be constitutively activated by endogenous InsP6.</p><p>The inhibitory effect of Ins(1,3,4,5,6)P5 upon GSK3β is more encouraging because of its specificity: neither Ins(1,4,5)P3, Ins(1,3,4,5)P4 nor InsP6 had any effect (Gao and Wang 2007). The treatment of intact cells with inhibitors of Ins(1,3,4,5,6)P5 synthesis also prevented Wnt3a from inhibiting GSK3β (Gao and Wang 2007). Since Ins(1,3,4,5,6)P5 levels do not typically change in response to short-term receptor activation, it is possible that the response that Gao and Wang (Gao and Wang 2007) observed is specific to signaling through Frizzled receptors. As for possible mechanisms, Ins(1,3,4,5,6)P5 did not inhibit purified GSK3β, so an intermediary seems to be required (Gao and Wang 2007). Further work on this topic would seem to be appropriate.</p><p>Aside from that isolated response of Ins(1,3,4,5,6)P5 to activation of the Fz1 receptor, its steady-state levels—and also those of InsP6—do not respond acutely to receptor activation (see above). Thus, when there are reports that Ins(1,3,4,5,6)P5 and InsP6 have biological activity, it has been difficult to place these data in a signaling context. One illustrative example is a report that both Ins(1,3,4,5,6)P5 and InsP6 inhibit protein phosphatases (Larsson et al. 1997). How can this be of regulatory significance if the levels of these polyphosphates do not change acutely? (In this particular case, one might also ask how any signaling specificity could result from two inositol polyphosphates both being broad spectrum inhibitors of PP1, PP2A and PP5). Similarly, it is also difficult to place in a signaling context a report that Ins(1,3,4,5,6)P5 and InsP6 inhibit L-type Ca2+ channels (Quignard et al. 2003); a similar criticism can be made of proposals that InsP6 is a "regulatory factor" in mRNA export and gene translation (Monserrate and York 2010; York et al. 1999). To be fair, we do note that others interpret these data rather differently. For example, it has been proposed that one role for InsP6 is to "set" (Berggren and Barker 2008) the basal state of a number of signaling entities. In particular, there are a number of studies that argue InsP6 establishes the default activities of various beta-cell signaling complexes (Berggren and Barker 2008). Barker and colleagues (2004) have also proposed that this putative global effector role for InsP6 may be of regulatory significance as cells transit through the cell cycle, during which time they estimate that the level of InsP6 may fluctuate by as much as threefold. Nevertheless, those cell-cycle dependent metabolic changes have not been tied to a specific signaling event.</p><p>Of course, the situation would be different if, as has been suggested (Larsson et al. 1997; Barker et al. 2002; Otto et al. 2007), the high total cellular levels of Ins(1,3,4,5,6)P5 and InsP6 mask stimulus-dependent alterations in smaller, discrete "signaling" pools of these compounds. That is, significant changes in "local" concentrations of Ins(1,3,4,5,6)P5 and InsP6 could be missed during the analysis of inositol phosphates in entire cell populations. Not so long ago, such a concept would have been labeled as heretical: how could a small and apparently freely-diffusible molecule not be uniformly distributed throughout the cell? However, it is now recognized that a concentration gradient of cAMP across a cell can be maintained by the spatial separation of the adenylyl cyclases from cAMP phosphodiesterases (Zaccolo et al. 2006). Is there any evidence for spatial heterogeneity of enzymes of inositol phosphate metabolism? Indeed there are. Arguably the most dramatic example is the receptor-dependent relocalization of the Ins(1,4,5)P3 3-kinase in hippocampal neurones (Schell and Irvine 2006). In stimulated cells, the kinase moves away from the post-synaptic region of the neuronal spines and into the dendritic shaft. This translocation undoubtably influences Ins(1,4,5)P3-dependent Ca2+ mobilization (Schell and Irvine 2006). But what about the enzymes that metabolize higher inositol phosphates? IPK2/IPMK (Nalaskowski et al. 2002; Odom et al. 2000) and IP5K (York et al. 1999; Brehm et al. 2007) are both concentrated in the nucleus. Moreover, the only known mammalian InsP6 phosphatase—MIPP—is restricted to the lumen of the endoplasmic reticulum (Craxton et al. 1997; Ali et al. 1993). So it is also of interest that plants at least can utilize an ABC-transporter like protein to move InsP6 across membranes (Nagy et al. 2009; Shi et al. 2007). It would be a significant breakthrough in this field if a mammalian homologue could be identified that transported InsP6 across the endoplasmic reticulum so that it could be metabolized by MIPP. However, embryonic fibroblasts made from Mipp−/− mice showed only 30% higher levels of InsP6 than wild-type animals; more discouragingly, the animals exhibited no obvious phenotype (Chi et al. 2000). Is it possible that mammals express another InsP6 phosphatase that we've all missed? Certainly we are missing something: we (Yang et al. 2008) have reported that a 20–25% decrease in cell volume following hyperosmotic stress is accompanied by a proportionate decrease in the amount of cellular InsP6, so that its concentration is not altered. This observation indicates that, when the cell deems it necessary, the metabolism of InsP6 can be quite rapid. We really ought to find out how, and why.</p><p>Some time ago, Michell's group (Stuart et al. 1994) also considered this question of whether or not some inositol phosphates might be present inside cellular organelles. They ascertained that 80–90% of the cells's inositol phosphates, including Ins(1,3,4,5,6)P5 and InsP6, were immediately released into the surrounding medium when the plasma membrane was permeabilized. That observation argues strongly against inositol phosphates being inside membrane-delimited cellular organelles. It can also be argued that no more than about half of the InsP6 pool in intact cells can be "hidden" from the cytoplasm, since the other half is readily accessible to soluble kinases that synthesize the inositol pyrophosphates (Menniti et al. 1993).</p><p>On the other hand, InsP6 can bind to membranes, at least in vitro (Cooke et al. 1991). InsP6 is also a structural component of certain cellular proteins (Macbeth et al. 2005); that particular pool of InsP6 would not be expected to be freely exchangeable with the bulk phase. There are other cellular proteins that can bind InsP6, which could also reduce its free concentration in the cytosol (Barker et al. 2002). The punctate intracellular distribution of endogenous IP5K, particularly in nucleoli and so-called stress granules (Brehm et al. 2007), also suggests that, to a degree at least, the synthesis of InsP6 might be compartmentalized. Uncertainty over compartmentalization is unlikely to be resolved until appropriate sensors of the intracellular location of these polyphosphates can be developed. It is our opinion that the status of Ins(1,3,4,5,6)P5 and InsP6 as cellular signals depends upon this question being answered.</p><p>There is no doubt that Ins(1,3,4,5,6)P5 and InsP6 still fulfill important biological functions that do not depend upon dynamic changes in their concentrations. For example, both Ins(1,3,4,5,6)P5 and InsP6 are precursors for the inositol pyrophosphates, which are currently the recipients of considerable interest from the signaling community (Saiardi 2011). Another possible function for Ins(1,3,4,5,6)P5 and InsP6 (in vitro at least) is to compete with inositol lipids for binding to certain proteins (Komander et al. 2004; Kavran et al. 1998). It has been speculated that this phenomenon increases the signal-to-noise ratio for PtdIns(3,4,5)P3-dependent functions (Irvine and Schell 2001; Komander et al. 2004). The idea is that binding of soluble inositol phosphates to a protein target helps keep it away from membranes until PtdIns 3-kinase activity is elevated by an appropriate stimulus (this concept is arguably analogous to the proposal (Berggren and Barker 2008) that Ins(1,3,4,5,6)P5 and InsP6 "set" the basal activities of certain signaling entities). Likewise, the binding of Ins(1,3,4,5,6)P5 to PTEN may inhibit that enzyme's low protein phosphatase activity and possibly contribute to PTEN's cytoplasmic and nuclear localization in the absence of PtdIns 3-kinase signaling (Caffrey et al. 2001).</p><p>Also, as noted in the introduction, InsP6 is an essential cofactor for adenosine deaminase (Macbeth et al. 2005), and, in yeast at least, InsP6 stimulates mRNA export from the nucleus (Alcázar-Román et al. 2006; York et al. 1999). Additionally, by enhancing the interaction of Ku with other proteins, InsP6 stimulates DNA repair through non-homologous end-joining (Cheung et al. 2008). However, since all of these functions for InsP6 can be satisfied by just a small percentage of total cellular InsP6 levels, it is our contention that in these cases this inositol polyphosphate more likely functions as a cofactor rather than as a dynamic "regulator".</p><p>In view of all of these activities of Ins(1,3,4,5,6)P5 and InsP6, it is not surprising that, in mammals, embryonic lethality results from the knock-out of IPK2/IPMK (Frederick et al. 2005) or IP5K (Verbsky et al. 2005a). The knock-down of Ip5K in zebrafish embryos is also phenotypically dramatic: there is disturbance of asymmetric Ca2+ signaling that is important for embryonic patterning (Sarmah et al. 2005). However, those genetic experiments in themselves do not speak to any specific signaling role of Ins(1,3,4,5,6)P5 or InsP6. It is possible that these phenotypes are, in part, consequences of the loss of non-catalytic activities of inositolphosphate kinases (Odom et al. 2000) and/or the absence of more highly phosphorylated metabolites, such as the inositol pyrophosphates, which also function in development (Sarmah and Wente 2010).</p><!><p>Cellular levels of Ins(3,4,5,6)P4 are around 1 μM in resting cells, and they increase to the 5–10 μM range whenever PLC is activated (Ho and Shears 2002). Ins(3,4,5,6)P4 is a concentration-dependent inhibitor of a CaMKII-activated Cl− conductance that is located in the plasma membrane (Xie et al. 1996, 1998; Ho et al. 2001; Mitchell et al. 2008). At least in mammalian cells, the inhibition of Cl− channel conductance by Ins(3,4,5,6)P4 is an exquisitely specific regulatory process; it is not imitated by any of the many other inositol phosphates that exist inside cells (Ho and Shears 2002; Ho et al. 2000; Xie et al. 1996). In other words, it has been demonstrated that Ins(3,4,5,6)P4 is a receptor-regulated signal, its biological target is known, and Ins(3,4,5,6)P4 acts specifically. This inositol phosphate is undoubtedly an intracellular signal.</p><p>Ins(3,4,5,6)P4 can only be formed in animal cells by receptor-dependent dephosphorylation of Ins(1,3,4,5,6)P5 by an enzyme that is—unfortunately—known as ITPK1 (for Inositol Trisphosphate Kinase). This baptism by the Human Genome Nomenclature Committee seems to have been prompted by the fact that the protein was initially characterized as a 6-kinase activity that phosphorylates Ins(1,3,4)P3 to Ins(1,3,4,6)P4 (Shears et al. 1987; Balla et al. 1987). It is only recently that it has been determined that the trisphosphate kinase activity reflects a more complex, ADP-dependent phosphotransferase activity (Chamberlain et al. 2007; Ho et al. 2002). This is a unique phenomenon in the inositol phosphate field, that explains the molecular mechanism by which Ins(3,4,5,6)P4 levels are coupled to receptor-regulated PLC activity (Fig. 13.2). In its ADP-bound form, ITPK1 dephosphorylates Ins(1,3,4,5,6)P5 to Ins(3,4,5,6)P4. The Ins(3,4,5,6)P4 is released to the bulk phase in exchange for Ins(1,3,4)P3, but the nucleotide—now ATP—remains bound. The tenacity of this binding of adenine nucleotide has been verified by crystallographic data showing that less than 10% of the nucleotide is solvent exposed (Miller et al. 2005; Chamberlain et al. 2007). Thus, the inorganic phosphate that is removed from Ins(1,3,4,5,6)P5 is not released. Instead, it is passed on to the newly-bound Ins(1,3,4)P3, thereby phosphorylating it to Ins(1,3,4,6)P4, which the active-site then exchanges for a new molecule of Ins(1,3,4,5,6)P5, and the entire phosphotransferase cycle is repeated (Fig. 13.2). Importantly, the rate at which Ins(1,3,4,5,6)P5 is dephosphorylated to Ins(3,4,5,6)P4 is stimulated as the rate-limiting concentration of phosphate acceptor—Ins(1,3,4)P3—is increased (Ho et al. 2002). In turn, the cellular levels of Ins(1,3,4)P3—a metabolite of Ins(1,4,5)P3 (Fig. 13.1)—mirrors both the intensity and the duration of receptor-activated PLC activity (Batty et al. 1998; Batty and Downes 1994). In other words, the degree of PLC activity sets Ins(1,3,4)P3 levels, which controls Ins(3,4,5,6)P4 synthesis. This is the molecular basis for the integration of inositol phosphate signaling pathways via human ITPK1.</p><p>The Ins(1,3,4)P3 6-kinase activity of ITPK1 also plays a metabolic role (Fig. 13.1) in maintaining the size of the cell's Ins(1,3,4,5,6)P5 pool (Shears et al. 1987; Balla et al. 1987; Verbsky et al. 2005b). It is unclear if it is this metabolic function, or the signaling activities of ITPK1, which explain why mice which are hypomorphic for the Itpk1 allele are susceptible to neural tube defects (Wilson et al. 2009). A complete knock-down of ITPK1 expression is apparently lethal (Verbsky et al. 2005b).</p><p>The best characterized biological end-point for Ins(3,4,5,6)P4 action upon Cl− transport is to regulate epithelial salt and fluid secretion (Vajanaphanich et al. 1994; Carew et al. 2000). However, the recent identification of ClC3 as the target of Ins(3,4,5,6)P4, at least in mammalian cells (Mitchell et al. 2008), has greatly expanded the biological repertoire of this inositol phosphate. For example, ClC3 is responsible for the Ins(3,4,5,6)P4-regulated Cl− conductance in hippocampal neurones (Mitchell et al. 2008), which is thought to contribute to the overall regulation of the synaptic efficacy in generating action potentials (Wang et al. 2006). Long-term changes in synaptic efficacy comprise a cellular basis for information storage and memory formation (Bliss and Collingridge 1993). Thus, Ins(3,4,5,6)P4 is a molecule that has the potential to affect neuronal development. It therefore seems pertinent that Ins(3,4,5,6)P4 has also previously been suggested to have the characteristics of a "memory molecule", because its relatively slow rate of metabolism permit its physiological effects to long outlast the duration of the stimulus that initially prompts intracellular Ins(3,4,5,6)P4 to accumulate (Ho and Shears 2002).</p><p>ClC3 that is in the plasma membrane may have other roles, such as tumor cell migration (Mao et al. 2008; Cuddapah and Sontheimer 2010) and the regulation of apoptosis (Claud et al. 2008). We can therefore anticipate that Ins(3,4,5,6)P4 might also regulate these processes. It should be noted, however, that some of the workers in this field (Jentsch 2008; Jentsch et al. 2002) propose that ClC3 is not a plasmalemmal Cl− channel per se, but instead a regulator of other Cl− channels. That argument, if correct, does not devalue the role of Ins(3,4,5,6)P4 in regulating ClC3 function. For example, in one cell type in which our own data are consistent with the ClC3 regulating other Cl− channels, we have shown that ClC3 still mediates the effect of Ins(3,4,5,6)P4 upon plasma membrane Cl− fluxes (unpublished data).</p><p>While there is disagreement as to whether or not ClC3 is an independent plasma membrane Cl− channel, it is well recognized that ClC3 also resides in intracellular vesicles such as insulin granules (Barg et al. 2001) and the early endosomal compartment (Zhao et al. 2007; Gentzsch et al. 2003; Stobrawa et al. 2001; Hara-Chikuma et al. 2005; Mitchell et al. 2008). Here, ClC3 contributes to endosomal acidification (Jentsch 2008), although there is uncertainty concerning the exact mechanism. Nevertheless, when a cell-permeant analogue of Ins(3,4,5,6)P4 was added to cells so as to inhibit ClC3, the pH of certain vesicular sub-compartments became more alkaline (Renström et al. 2002; Mitchell et al. 2008). What is the biological significance of this regulation of intravesicular pH? With regard to insulin granules, it has been proposed that their intraluminal acidification is a priming process, without which they become less competent to fuse with the plasma membrane and release their cargo (Barg et al. 2001). In support of this idea, we have shown that alkalinization of insulin granules by Ins(3,4,5,6)P4 has the effect of reducing insulin secretion from pancreatic β-cells (Renström et al. 2002). In many other cell types, the acidification of endosomes and secretory vesicles serves other important functions, including modulation of certain ligand-protein interactions during endocytosis, enzyme targeting, and H+-coupled uptake of small molecules (such as neurotransmitters) (Nishi and Forgac 2002; von and Sorkin 2007). It appears that we have only scratched the surface of our understanding of the biological importance of Ins(3,4,5,6)P4.</p><p>It is unfortunate that we do not yet understand the mechanism by which Ins(3,4,5,6)P4 prevents CaMKII from activating ClC3. Data obtained to date indicate that Ins(3,4,5,6)P4 does not inhibit CaMKII activity per se (Xie et al. 1996; Ho et al. 2001; Ho and Shears 2002). Furthermore, in single channel analysis of CaMKII-activated Cl− channels, Ins(3,4,5,6)P4 was not inhibitory, so it is unlikely to act as a direct channel blocker (Ho et al. 2001). Presumably an intermediary protein mediates the action of Ins(3,4,5,6)P4. However, our efforts to identify an Ins(3,4,5,6)P4 "receptor" have so far been disappointingly fruitless (unpublished data). Our work on this important problem is ongoing.</p><!><p>The metabolic intermediates that accumulate during the dephosphorylation of Ins(1,4,5)P3 and Ins(1,3,4)P3 to inositol are not generally considered to be signaling molecules. The same could be true of at least some of the intermediates in the pathways of phosphorylation of Ins(1,4,5)P3 to InsP6. The concept that inositol is a combinatorial signaling scaffold (York 2006) is intellectually appealing, but it is still not obligatory that all of these inositol phosphates be cellular signals; only at least one of the end products might act in a signaling pathway.</p><p>We have argued here that there is no strong evidence that Ins(1,4,5,6)P4 is a cellular signal. As for Ins(1,3,4,5)P4, we have highlighted the confusing and often conflicting observations in the literature concerning proposed actions of this inositol phosphate. Under such circumstances, it is difficult to formulate a general signaling role. Ins(1,3,4,5,6)P5 and InsP6—particularly the latter—clearly have important roles as cofactors, but our conclusion is that we need more concrete evidence before we can claim that these molecules are truly cellular signals. As we have discussed, it may be that further information of cellular compartmentalization may be the savior of these molecule's signaling credentials. So, other than Ins(1,4,5)P3, that leaves, in our opinion, Ins(3,4,5,6)P4 as the only validated "classical" cellular signal from within this group of molecules.</p><!><p>Inositol phosphate metabolism. The figure shows the pathway of Ins(1,4,5)P3 metabolism. The numbers in circles indicate the different enzymes that are involved: 1, IPK2/IPMK; 2, Ins(1,4,5)P3 3-kinases; 3, Ins(1,4,5)P3/ Ins(1,3,4,5)P4 5-phopshatase; 4, ITPK1; 5, PTEN; 6, IP5K. There is a candidate for the question mark—MIPP—but it is uncertain how that enzyme can access its substrate (see text for details). The inositol pyrophosphates are not shown in this figure (they are the subject of a separate chapter (Saiardi 2011)). The enzymes that dephosphorylate Ins(1,4,5)P3 and Ins(1,3,4)P3 to inositol are not shown, as this review is only concerned with metabolites that have received attention as being cellular signals. Note that the positional specificity of IPK2/IPMK shows phylogenetic variation. In yeasts, Ins(1,4,5)P3 is phosphorylated primarily to Ins(1,4,5,6)P4, in mammals the product is predominantly Ins(1,3,4,5)P4, and the enzyme in flies produces roughly equal quantities of both InsP4 isomers</p><p>The phosphotransferase activity of ITPK1. The figure shows the phosphotransferase activity that is catalyzed by ITPK1. The phosphate group that is transferred between from Ins(1,3,4,5,6)P5 to Ins(1,3,4)P3 is highlighted in by the grey circle</p>

PubMed Author Manuscript

Divalent Cu, Cd, and Pb Biosorption in Mixed Solvents

Dead dried Chlorella vulgaris was studied in terms of its performance in binding divalent copper, cadmium, and lead ions from their aqueous or 50% v/v methanol, ethanol, and acetone solutions. The percentage uptake of cadmium ions exhibited a general decrease with decrease in dielectric constant values, while that of copper and lead ions showed a general decrease with increase in donor numbers. Uptake percentage becomes less sensitive to solvent properties the larger the atomic radius of the biosorbed ion, and uptake of copper was the most affected. FT-IR analyses revealed stability of the biomass in mixed solvents and a shift in vibrations of amide(I) and (II), carboxylate, glucose ring, and metal oxygen upon metal binding in all media. ΔνCOO values (59–69 cm−1) confirmed bidentate metal coordination to carboxylate ligands. The value of ν asCOO increased slightly upon Cu, Cd, and Pb biosorption from aqueous solutions indicating lowering of symmetry, while a general decrease was noticed in mixed solvents pointing to the opposite. M–O stretching frequencies increased unexpectedly with increase in atomic mass as a result of solvent effect on the nature of binding sites. Lowering polarity of the solvent permits variations in metal-alga bonds strengths; the smaller the metal ion, the more affected.

divalent_cu,_cd,_and_pb_biosorption_in_mixed_solvents

1. Introduction<!>2. Materials and Methods<!>3. Results and Discussion<!>4. Conclusions

<p>Algal biomass represents an inexpensive material, that is, very efficient in heavy metal removal from effluents and industrial waters. Different mechanisms have been proposed to describe metal biosorption by algae where several functional groups are involved; carboxylates, hydroxyls, amines, phosphates, and sulfates (in marine species usually) are the most involved in metal binding [1–4]. Chlorella vulgaris (division Chlorophyta) usually does not contain sulfates [5].</p><p>According to Pearson [6] and complexation behavior of ligands and cations in terms of electron pair donating Lewis bases and electron pair accepting Lewis acids, metal ions are termed hard, soft, and borderline. Hardness of metal ions (Lewis acids) will determine their preference to binding. Softer ions (e.g., Cd2+, low positive charge relative to large size, very polarizable) are expected to bind nitrogen and sulfur donor atoms of the ligand on the algal cell wall, whereas hard metals (e.g., Ca2+, high charge to radius ratio, not very polarizable) coordinate to carboxylate groups, and borderline ions (e.g., Cu2+ and Pb2+) would bind to any of the ligands according to conditions that may change the hardness of the ligand. The same concept in hardness is used to describe solvents in terms of their abilities to bind ions [7].</p><p>Solvent polarity is one of many factors affecting metal binding besides concentration, temperature, pH, and presence of competing species. Assessment of polarity scales employed bulk physical properties such as dielectric constant (relative permittivity) and other measures of chemical interactions such as donor numbers (important in coordination chemistry). The donor number (DN) [b] is a chemical measure (devised by Viktor Gutmann 1976) of Lewis basicity and is defined as the negative enthalpy value for the 1 : 1 adduct formation between a Lewis base and the standard Lewis acid antimony pentachloride, in dilute solution of the noncoordinating solvent 1,2-dichloroethane with a zero DN. The units are kilocalories per mole.</p><p>Most of the studies on metal binding by algae were performed in aqueous media [1–8]. Biosorption investigations on other types of biomass, such as fungi [9] and natural products waste [10] followed the same methods of biosorption from aqueous solutions. Using solvents other than water in such experiments are almost exclusive of extraction purposes of materials adsorbed by algae [11] and in limited cases for metal ion adsorption capacity studies [8]. In an earlier study [12] ethanol-water Cu(II), Cd(II), Fe(III), and Sn(IV) solutions proved to enhance metal uptake by Chlorella vulgaris after extensive reuse of the biomass.</p><p>Using vibrational spectroscopy to study the cell wall of algae [3, 9, 10] and to investigate the metal coordination sites proved to be very useful in understanding the nature of metal-algae binding. To our knowledge, such studies on metal-loaded algae from mixed solvent solutions received no attention to date. Hence this study aims to investigate the biosorption of divalent copper, cadmium, and lead from different mixed solvents and the effect of the solvent properties on metal ion uptake and on metal-sensitive vibrations in the infrared region.</p><!><p>The alga Chlorella vulgaris was generously provided by Dr. F. Al-Baz from the Botany Department at the National Research center, Cairo. The biomass was washed, air dried, ground, and sieved to a particle size of <355 μm. Analytical grade nitrate salts of copper, cadmium, and lead were used as well as methanol, ethanol, and acetone of the same purity. Deionized water was used for all experiments. FT-IR spectra were recorded as KBr (10%) pellets using a Perkin-Elmer Spectrum 1000 FT-IR spectrometer.</p><p>50% v/v mixed methanol/water, ethanol/water, and acetone/water were used in all experiments as mixed solvents. To determine the amount of the metal biosorbed by Chlorella vulgaris, 0.1 g biomass was added to 50 mL of 200 ppm metal solution, stirred for 10 minutes and left overnight, then filtered. Metal concentration was measured using ICP and the amount adsorbed calculated by difference. To prepare samples for IR measurements, 0.1 g of the biomass was added to 50 mL of 0.1 M metal solution, stirred for 10 minutes, and left overnight. Then filtered, washed, and dried at 80°C for one hour.</p><p>The dielectric constant (ε mix) of each mixed solvent was calculated using the equation</p><p>(1)εmix=εH2OVH2O+εsolventVsolventVtotal</p><p>and presented in Table 1.</p><!><p>Values of dielectric constants of mixed solvents are all above 40, which reduce chances of formation of ion pairs in solution, and all ions are expected to be present as totally solvated in all solvents. In addition, polarizing ability of each of these ions is not enough to encourage ion pairs to form. Accordingly, metal ion species suggested are of the type [M(H2O)6-n(Sol)n]2+, a distorted octahedral moiety. Uptake from dilute solutions such as in this study renders precipitation on the algal cell wall unlikely to occur.</p><p>The percentage uptake for each metal ion was calculated and presented in Table 1.</p><p>Inspecting the values in the table above, it appears that the uptake percentage of the metal ions does not follow a clear trend in relation to the same property. But we can divide our observations into (1) cadmium ions showing a general decrease in percentage biosorbed with decrease in dielectric constant values (2) the harder copper and lead ions exhibiting a general increase in biosorption with decrease in donor numbers (Figure 1). The decrease from water to both alcohols is understood, as stronger forces between solvent and metal minimizes opportunities of competing ligands on algal cell wall. The unexpected higher uptake percentage from acetone solutions compared to alcohol solutions could be attributed to the fact that acetone is a nonprotogenic solvent which reduces chances for exchange with protons on the biomass.</p><p>A plot of the percentage biosorbed versus ionic radii of the divalent metal ions (Figure 2) showed that, as the ionic radius increases from copper to lead, the uptake process becomes less sensitive toward changing the solvent.</p><p>Biological molecules such as algae show complex vibrational spectra that include overtones and combinational bands. But metal-ligand stretching frequencies and properties of functional groups coordinated to metal centers offer useful information. C–O stretching, NH2 rocking, and M–N and M–O stretching bands are metal sensitive and are shifted as the metal is changed, but NH2 vibrations are very sensitive to the effect of intermolecular interactions (e.g., hydrogen bonding) which makes it difficult to discuss the strength of the metal-nitrogen bond from the frequency shift. Doshi et al. [13] reported a blue shift of about 75–100 cm−1 of the band at 3304 cm−1 assigned to ν NH2 coupled with hydrogen-bonded hydroxyl stretching in Spirulina sp. upon treatment with metal ions. Alcoholic groups in the glucose ring may play a role in metal binding, although Guibal et. al. [14] considered it constant and used it as an internal standard for calculating band intensities.</p><p>Assigning bands to the corresponding vibrations for biomaterials like algae is not a direct and easy task. According to Nakamoto [15], results of some researchers [1, 3, 14] and earlier work [16, 17], metal sensitive bands for the free biomass (water washed) are assigned as in Table 2.</p><p>Mixed aqueous methanol, ethanol, or acetone did not seem to affect the functional groups responsible for metal binding in the biomass as can be seen from the IR spectra relative to water (Figure 3).</p><p>Most FTIR studies on algae and seaweeds [1, 3, 14] and algal extracts [18] revealed the metal interaction sites of carboxyl, amino, and hydroxyl groups on the algal surface.</p><p>Introducing copper, cadmium, or lead into the biomass shifted all metal sensitive frequencies. For all metal loaded biomass samples, the separation (ΔνCOO) between ν asCOO (asymmetrical stretching) and ν sCOO (symmetrical stretching) was in the range of 59–69 cm−1, which conforms with bidentate coordination [15]. Hence, solvent polarity affected the amount adsorbed of the metal and accordingly shifted vibrations of metal sensitive bonds, but the mode of bonding was not affected. The shift in ν asCOO (Δν asCOO) varied upon metal binding and changing solvent, and a wider range of this shift was observed with decrease in dielectric constant values (Figure 4), where Δν asCOO = (ν asCOO free solvent − ν asCOO metal-loaded biomass).</p><p>The positive shift in vibrations of samples prepared in aqueous solutions indicates lowering in symmetry for carboxylate coordination; on the contrary most samples from mixed solvents exhibited a negative shift presenting a more symmetrical environment for carboxylate. The same trend was observed in shifts of amide(II) bands in Figure 5. It can be concluded then that lowering polarity of the solvent permits variations in bond strength.</p><p>Figure 6 reveals an unexpected general trend; the stretching frequencies of M–O bonds increased with increasing atomic mass of the metal in all solvents. This may be a result of changes in the nature of the binding sites with changing solvents, which overcame the reduced mass effect.</p><p>The last figure showed that copper biosorption varied in different solvents more than cadmium or lead did, which is what was noticed in Figure 2 (chemical analyses results). We can conclude then that the smaller the metal atom will be, the more affected biosorption will be by changing the solvent polarity.</p><!><p>Cadmium ions showed a general decrease in percentage uptake with decrease in dielectric constant values, while the harder copper and lead ions exhibited a general increase in biosorption with decrease in donor numbers. The metal ion uptake process becomes less sensitive toward changing the solvent the larger the atomic radius.</p><p>Soaking dead dried Chlorella vulgaris in 50% v/v methanol-water, ethanol-water, or acetone-water did not alter the functional groups. ΔνCOO for all metal loaded biomass samples was in the range of 59–69 cm−1 confirming bidentate coordination to carboxylate ligands. The value of ν asCOO increased slightly upon Cu, Cd, and Pb biosorption from aqueous solvents indicating lowering of symmetry, while a general decrease was noticed from mixed solvents indicating the opposite. Lowering polarity of the solvent permits variations in bond strength. M–O stretching frequencies increased unexpectedly with increase in atomic mass of the metal in all solvents, a result—thought to be—because of the nature of binding sites being affected by solvents, making the reduced mass effect less pronounced.</p>

PubMed Open Access

Prospects on Strategies for Therapeutically Targeting Oncogenic Regulatory Factors by Small-Molecule Agents

Although the Human Genome Project has raised much hope for the identification of druggable genetic targets for cancer and other diseases, this genetic target-based approach has not improved productivity in drug discovery over the traditional approach. Analyses of known human target proteins of currently marketed drugs reveal that these drugs target only a limited number of proteins as compared to the whole proteome. In contrast to genome-based targets, mechanistic targets are derived from empirical research, at cellular or molecular levels, in disease models and/or in patients, thereby enabling the exploration of a greater number of druggable targets beyond the genome and epigenome. The paradigm shift has made a tremendous headway in developing new therapeutic agents targeting different clinically relevant mechanisms/pathways in cancer cells. In this Prospects article, we provide an overview of potential drug targets related to the following four emerging areas: (1) tumor metabolism (the Warburg effect), (2) dysregulated protein turnover (E3 ubiquitin ligases), (3) protein\xe2\x80\x93protein interactions, and (4) unique DNA high-order structures and protein\xe2\x80\x93DNA interactions. Nonetheless, considering the genetic and phenotypic heterogeneities that characterize cancer cells, the development of drug resistance in cancer cells by adapting signaling circuitry to take advantage of redundant pathways or feedback/crosstalk systems is possible. This \xe2\x80\x9cphenotypic adaptation\xe2\x80\x9d underlies the rationale of using therapeutic combinations of these targeted agents with cytotoxic drugs.

prospects_on_strategies_for_therapeutically_targeting_oncogenic_regulatory_factors_by_small-molecule

<!>HOW MANY DRUGGABLE ANTICANCER TARGETS ARE THERE? GENOME- VERSUS MECHANISM-BASED TARGETS<!>THE WARBURG EFFECT (TUMOR METABOLISM): GLYCOLYTIC ENZYMES AS TARGETS<!>DYSREGULATED PROTEIN TURNOVER: ONCOGENIC E3 LIGASES AS TARGETS TO RESTORE DYSREGULATED PROTEIN FUNCTION<!>Mdm2<!>Skp2<!>\xce\xb2-TRANSDUCIN REPEAT-CONTAINING PROTEIN (\xce\xb2-TrCP)<!>CASITAS B-LINEAGE LYMPHOMA (c-Cbl)<!>TARGETING PROTEIN\xe2\x80\x93PROTEIN INTERACTIONS<!>TARGETING UNIQUE DNA HIGH-ORDER STRUCTURES AND PROTEIN\xe2\x80\x93DNA INTERACTIONS<!>TARGETING THE DNA QUADRUPLEX<!>TARGETING PROTEIN\xe2\x80\x93DNA INTERACTIONS<!>OUTLOOK

<p>In light of remarkable technological breakthroughs in cancer "omics," the past decade has witnessed tremendous progress in our understanding of cancer biology [Vucic et al., 2012]. These advances have also been translated into new cancer biomarkers and therapeutic targets, leading to a shift in the paradigm of drug discovery toward a target-based rational design approach in lieu of empirical structure–activity relationship-based lead modifications. Such a paradigm is epitomized by the FDA's approvals of more than 10 kinase inhibitors that target mutational activation of kinase signaling in various types of malignancies [Zhang et al., 2009; Dar and Shokat, 2011], including BCR–ABL fusion in chronic myelogenous leukemia [Druker et al., 2001], BRAF mutations in melanomas [Flaherty et al., 2010], EGFR mutations in a subset of lung adenocarcinoma [Lynch et al., 2004; Paez et al., 2004; Pao et al., 2004], and ALK fusion in lung cancer [Koivunen et al., 2008].</p><p>Although these new therapeutic agents have led to improved clinical outcomes for many cancer patients, kinase inhibitors face two major challenges in clinical development, that is, specificity for target versus off-target kinases and emergence of drug resistance. Most kinase inhibitors developed so far act by competing with ATP for the ATP-binding sites located at the hinge region of target kinases [Zhang et al., 2009]. As there are a total of 518 kinases encoded in the human genome [Venter et al., 2001], it is inevitable that many of these drugs show complex clinical pharmacology in vivo by targeting multiple kinases [Zhang et al., 2009; Dar and Shokat, 2011], which raises potential concerns of untoward side effects arising from this polypharmacology. However, from a clinical perspective, such multikinase inhibitors might be therapeutically advantageous through enhanced efficacy by targeting a spectrum of kinases involved in cancer pathogenesis and progression. Examples include sorafenib [Ahmad and Eisen, 2004] and sunitinib [Fabian et al., 2005], both of which suppress tumor proliferation and angiogenesis by blocking multiple kinase pathways, including those mediated by RAF-kinase, vascular endothelial growth factor receptor (VEGF)2, VEGF3, platelet-derived growth factor receptor-β, KIT, and FLT3. With regard to drug resistance, cancer cells acquire a resistant phenotype to kinase inhibitors under selective pressure, in part, through target amplification or mutations at the gate-keeper residues that abrogate drug binding [Zhang et al., 2009]. Alternatively, cancer cells might adapt their signaling circuitry to develop compensatory mechanisms by taking advantage of redundant signaling pathways or feedback/crosstalk systems to counteract drug actions [Logue and Morrison, 2012].</p><p>Another frontier that has progressed rapidly in cancer therapeutic development is epigenetic-modulating drugs [Rodríguez-Paredes and Esteller, 2011]. The cancer epigenome is characterized by global changes in the patterns of DNA methylation and histone modifications arising from dysregulated expression of DNA methyltransferases (DNMTs) and histone-modifying enzymes, including histone acetyltransferases (HATs)/deacetylases (HDACs), lysine- and arginine-specific methyltransferases (HMTs)/demethylases (HDMs), kinases/phosphatases, and so on [Kouzarides, 2007]. Dysregulation of any of these epigenetic enzymes through mutations or altered expression results in aberrant gene expression associated with typical cancer traits. More important, in contrast to genetic mutations, the reversible nature of epigenetic changes in the patterns of DNA methylation and histone acetylation/methylation underlies the impetus of targeting this epigenetic machinery, particularly DNMTs [Heyn and Esteller, 2012; Singh et al., 2013] and HDACs [Marks, 2010], in cancer cells to restore the epigenome to its normal state. In the past few years, the epigenetic field has generated 4 FDA-approved drugs for the treatment of subtypes of leukemia and lymphoma, including the DNMT inhibitors 5-azacytidine (azacitidine, Vidaza) and 5-aza-2′-deoxycytidine (decitabine, Dacogen) for myelodysplastic syndrome and the HDAC inhibitors SAHA (vorinostat, Zolinza) and depsipeptide (romidepsin, Istodax) for the rare cutaneous T cell lymphoma and other hematological malignancies.</p><p>Although the biology of other epigenetic enzymes remains less well defined, inhibitors of many of these enzymes, especially those of sirtuins, HATs, HMTs, and HDMs, have shown promising preclinical tumor-suppressive efficacy in vitro and/or in vivo [Rodríguez-Paredes and Esteller, 2011].</p><!><p>Although the Human Genome Project has raised much hope/hype for the identification of druggable genetic targets for cancer and other diseases, this genetic target-based approach, however, has not improved productivity over the traditional approach. This discrepancy might, in part, be attributable to the complex process of in vitro and in vivo validation of these targets in relevant cell and transgenic animal models [Sams-Dodd, 2005]. Analyses of known human target proteins of currently marketed drugs reveal that these drugs target only a limited number of proteins as compared to the whole proteome. For example, a comprehensive analysis of all FDA-approved small-molecule drugs, a total number of 1,204, by Overington et al. [2006] indicated that these drugs, including 5 kinase inhibitors acting on 18 protein kinases, targeted 207 distinct human genome-derived proteins [Hopkins and Groom, 2002], a small number in comparison to the estimated 30,000 human protein-coding genes [Overington et al., 2006]. The majority of these 207 drug-targeted proteins in human cells fall into the following categories: G protein-coupled receptors, ligand-gated ion channels, nuclear receptors, phosphodiesterases, proteases, protein kinases, voltage-gated ion channels, and enzymes involved in DNA synthesis/mitosis [Overington et al., 2006]. With the exception of kinases, the identification of these targets, prior to the Human Genome Project, were based on laboratory or clinical findings associated with various pathological conditions. From a translational perspective, this mechanism-based approach avoids the one-gene-one-disease hypothesis, and can be applied much more broadly in the context of target identification.</p><p>In contrast to genome-based drug targets, mechanistic targets are derived from empirical research, at cellular or molecular levels, in disease models and/or in patients, thereby enabling the exploration of a greater number of druggable targets. This approach is illustrated by the therapeutic targeting of the Warburg effect by developing inhibitors of enzymes involved in glucose metabolism in tumor cells. In addition, a number of molecular defects resulting from dysregulated protein turnover or interactions with other macromolecules (proteins and DNA) have also been interrogated as targets (Fig. 1), which are delineated as follows.</p><!><p>Cells undergoing malignant transformation often exhibit a shift in cellular metabolism from oxidative phosphorylation to glycolysis, known as the Warburg effect, to gain growth advantage [Kroemer and Pouyssegur, 2008; Vander Heiden, 2011]. This glycolytic shift enables cancer cells to adapt to low-oxygen environments, to produce biosynthetic building blocks needed for cell proliferation, to acidify the local environment to facilitate tumor invasion, and to generate NADPH and glutathione through the pentose phosphate shunt to increase resistance to oxidative stress. As the Warburg effect is considered a fundamental property of neoplasia, targeting glycolysis represents a therapeutically relevant strategy for cancer treatment [Eisenstein, 2012]. Thus, development of small-molecule agents that target various aspects of glucose metabolism has been the focus of many recent investigations, which are summarized in Table I.</p><p>The emerging view of cancers as a metabolic disease opens up opportunities for the development of new strategies for cancer therapy. Many of the tumor metabolism-targeted agents listed in Table I exhibit in vivo efficacy alone or in combination with chemotherapeutic drugs in advanced cancers. Although it is generally believed that interference with energy metabolism gives rise to ATP depletion and metabolic stress, leading to cell death, data from this and other laboratories indicate that reduction of glycolytic rate by energy restriction elicits the activation of multiple signaling pathways, including those mediated by the NAD+-dependent HDAC Sirt1 (silent information regulator 1), AMPK, and endoplasmic reticulum (ER) stress [Wei et al., 2010]. This complicated signaling network affects many aspects of cellular functions in cell cycle regulation, survival, and aggressive phenotype, culminating in cancer cell death through autophagy and apoptosis. Thus, it is plausible to achieve synergy in killing cancer cells by using metabolism-targeted agents with other molecularly targeted agents, such as kinase inhibitors or HDAC inhibitors. Further understanding of the signaling mechanisms underlying the antitumor effects of these tumor metabolism-targeted agents will help foster novel strategies for cancer therapy.</p><!><p>The ubiquitin-proteasome system (UPS) plays a pivotal role in the regulation of key cellular functions, including cell cycle control, DNA repair, and growth factor receptor signaling, through targeted degradation of regulatory proteins [Devoy et al., 2005]. UPS-mediated protein degradation consists of two sequential steps initiated by ubiquitination of the target protein, followed by proteolysis via the 26S proteasome complex. The targeted ubiquitination is mediated through the concerted action of three enzymes: E1 ubiquitin-activating enzyme, E2 ubiquitin-conjugating enzyme, and E3 ubiquitin ligase (Fig. 3) [Nakayama and Nakayama, 2006]. In the past few years, inhibitors targeting different components of this ubiquitination system have been developed. For example, a mechanism-based neddylation inhibitor, MLN4924, was developed to target NEDD8 activating enzyme, an essential component of the NEDD8 conjugation pathway that controls the activity of the cullin-RING subtype of ubiquitin ligases [Soucy et al., 2009], while small-molecule inhibitors of murine double minute 2 protein (Mdm2) have progressed into preclinical/clinical development (see discussion below). From a therapeutic perspective, relative to E1 and E2, E3 ligases are of particular interest as drug targets for their role in conferring the selectivity for protein ubiquitination [Nalepa et al., 2006].</p><p>E3 ligases can be divided into three groups: the RING-finger E3s, the HECT (homologous to E6-AP COOH-terminus)-domain E3s, and the U-box E3s, each of which is characterized by a distinct protein interaction domain (RING-finger, HECT, or U-box domain) that serves to bind E2 ligases. As these E2-interacting domains are highly conserved, the specificity of E3 ligases is conferred by a variable substrate recognition motif that determines which substrate is to be ubiquitinated. Consequently, while proteasome inhibitors block the degradation of all ubiquitinated proteins indiscriminately, targeting a single E3 ligase allows for selective stabilization of a subset of ubiquitinated proteins. In light of this increased specificity, it is more therapeutically advantageous to target E3 ligases, as compared to the proteasome, to increase the stability/activity of selected tumor-suppressive proteins.</p><p>Among the three groups of E3 ligases, RING-finger E3 ligases have received much attention in the development of small-molecule inhibitors, which are represented by inhibitors of Mdm2 and S-phase kinase associated protein 2 (Skp2) in light of their well-characterized roles in regulating the degradation and/or activity of the tumor suppressors p53 [Kubbutat et al., 1997] and p27 [Carrano et al., 1999], respectively. The therapeutic targeting of Mdm2, Skp2, and other E3 ligases involved in regulating the stability of oncogenic or tumor-suppressive proteins are addressed as follows.</p><!><p>At the cellular level, Mdm2 and p53 are mutually regulated through an autoregulatory feedback loop: in response to stress signals, p53 transcriptionally activates Mdm2 gene expression, and in turn Mdm2 inhibits the transcriptional activity and promotes the ubiquitin-dependent proteasomal degradation of p53 [Wang et al., 2012b]. Thus, dysregulation of this regulatory loop results in malignant transformation of normal cells. Among all E3 ligases identified, Mdm2 is the most intensely pursued target, leading to the development of several structurally distinct classes of inhibitors, at least six of which have progressed into clinical trials in advanced solid tumors or acute myelogenous leukemia [Zhao et al., 2013]. Mechanistically, Mdm2 inhibitors are classified into two categories: inhibitors of the Mdm2-p53 protein–protein interactions, such as Nutlin-3 [Vassilev et al., 2004] and most other Mdm2 inhibitors [Zhao et al., 2013], and inhibitors of the Mdm2 E3 ligase activity [Yang et al., 2005].</p><p>Recently, MdmX, an Mdm2 homolog, has also received considerable attention as a target for therapeutic development [Zhao et al., 2013] because of its non-redundant and essential role as a negative regulator of p53 [Finch et al., 2002)]. MdmX has no E3 ligase activity, but forms heterodimers with Mdm2 through their RING domains to increase Mdm2 E3 ligase activity [Tanimura et al., 1999]. However, due to the high degree of sequence homology between Mdm2 and MdmX, many small-molecule inhibitors that were designed to target MdmX-p53 interactions also showed high affinity with Mdm2, thus becoming MdmX/Mdm2 dual inhibitors.</p><!><p>Substantial evidence indicates that Skp2, a Skp1-Cul1-F-box (SCF) E3 ubiquitin ligase, acts as an oncoprotein by targeting a wide range of signaling effectors, such as the tumor suppressor p27 [Carrano et al., 1999], for degradation. Moreover, it was demonstrated that Skp2 facilitates the activation of Akt through ubiquitination downstream of ErbB receptor signaling in Her2-positive breast cancer [Chan et al., 2012], and that Skp2 represents a key component for the Mre11/Rad50/NBS1 (MRN) complex-mediated ATM activation in response to DNA double-strand breaks through NBS1 ubiquitination [Wu et al., 2012]. Together, this oncogenic E3 ligase represents an important target for cancer drug discovery [Frescas and Pagano, 2008; Wang et al., 2012c].</p><p>Data from the authors' laboratory indicate that downregulation of Skp2 represents a cellular response in cancer cells to energy restriction induced by CG-5 (a novel glucose transporter inhibitor) and 2-deoxyglucose [Wei et al., 2012]. This Skp2 downregulation was attributable to Sirt1-dependent suppression of COP9 signalosome (Csn)5 expression in response to CG-5, leading to increased cullin 1 neddylation in the SCF protein complex and consequent Skp2 destabilization. This finding provides a proof-of-concept that the oncogenic Csn5/Skp2 signaling axis represents a "druggable" target by using this novel glucose transporter inhibitor. More recently, a Skp2 inhibitor, 3-(1,3-benzothiazol-2-yl)-6-ethyl-7-hydroxy-8-)1-piperidinylmethyl)-4H-chromen-4-one (compound #25), that selectively inhibited Skp2 E3 ligase activity, but not other SCF complexes, was identified using high-throughput in silico screening [Chan et al., 2013]. This Skp2 inhibitor phenocopied the effects observed upon genetic Skp2 deficiency, such as suppressing survival and Akt-mediated glycolysis and triggering p53-independent cellular senescence, and showed antitumor efficacy in multiple animal models.</p><!><p>Besides Skp2, another SCF E3 ligase that has received much attention is β-TrCP. In contrast to Skp2, the role of β-TrCP as a therapeutic target remains controversial because it plays a dichotomous role, either oncogenic or tumor-suppressive, in a cellular context-dependent manner considering its diverse substrate spectrum [Frescas and Pagano, 2008]. Although evidence suggests its oncogenic character is mediated through the activation of NF-κB signaling [Fuchs et al., 2004], β-TrCP also facilitates the degradation of a wide array of tumor-promoting proteins, including β-catenin [Hart et al., 1999], Snail [Yook et al., 2006], ATF4 [Lassot et al., 2001], cdc25A [Jin et al., 2003], Mcl-1 [Ding et al., 2007], cyclin D1 [Wei et al., 2008], and Sp1 [Wei et al., 2009], thereby suppressing cancer cell proliferation and invasion. The authors previously demonstrated that treatment of cancer cells with the glucose transporter inhibitor CG-5 led to decreased Skp2 accompanied by upregulated β-TrCP expression, as Skp2 targets β-TrCP for degradation via a cyclin-dependent kinase 2-dependent mechanism [Wei et al., 2012]. Mechanistic evidence indicates that this β-TrCP upregulation underlies the suppressive effect of CG-5 on cancer cell proliferation. Together, these findings raise a question of whether the inhibition of β-TrCP-mediated ubiquitination represents a therapeutically relevant strategy for cancer treatment.</p><!><p>Is a RING-type E3 ligase involved in ubiquitination and degradation of BCR–ABL, EGFR, and a series of other receptor and non-receptor protein kinases [Lu and Hunter, 2009]. Thus, wild type c-Cbl has been proposed to function as a tumor suppressor. A recent report indicates that arsenic sulfide (As4S4) upregulated the expression of c-Cbl by blocking its self-ubiquitination/degradation through the RING finger binding, thereby inducing degradation of BCR-ABL in chronic myelogenous leukemia (CML) [Mao et al., 2010]. This finding provides a molecular basis to design small-molecule agents that activate c-Cbl through a similar mode of mechanism.</p><!><p>The concept of targeting protein–protein interactions has been demonstrated by therapeutic antibodies, which block ligand-mediated activation of growth factor or cytokine receptors. In contrast, even just a decade ago, it was generally perceived unfeasible to develop selective small-molecule compounds that could interfere with protein–protein interactions effectively, of which the reason is multifold [Arkin and Wells, 2004]. For example, the interface involved in the protein–protein complex formation is typically large and associated with diverse protein topologies, and small molecules would have to compete with macromolecular partners for binding. However, recent advances in structural biology and bioinformatic analysis indicate that a few amino acids at the interface ("binding hotspots") contribute to the majority of the binding energy in protein–protein interactions [Valkov et al., 2012]. This paradigm shift in combination with advances in computation- and chemical library-based high-throughput screening technologies has proven the feasibility of developing small-molecule inhibitors of protein–protein interactions [Blundell et al., 2006; Berg, 2008b; Fry, 2008; Valkov et al., 2012]. To date, therapeutic development targeting a number of protein–protein interactions has been the focus of many recent reports (Table II). Especially noteworthy is that at least two BH3 mimetics, (−)-gossypol and ABT-263 (Navitoclax), have advanced to clinical trials [http://clinicaltrials.gov].</p><!><p>A large portion of currently used chemotherapeutics, such as platinum drugs, alkylating agents, topo II poisons, and DNA intercalating agents, act by inhibiting DNA replication and cell division through their reactions with DNA [Sheng et al., 2013]. However, relative to proteins, DNA historically has not been as well recognized as a mechanistic target for structure-based drug design for a number of reasons. First, the highly charged nature of DNA renders ligand recognition of target DNA less discriminative/specific. Second, transcription factors have generally been considered undruggable due to lack of suitable assay methods. In recent years, advances in nucleic acid chemistry and molecular and structural biology have created opportunities for potential drug discovery, which is addressed as follows.</p><!><p>G-quadruplexes (also known as G-tetrads or G4-DNA) are higher-order DNA structures formed from guanine (G)-rich sequences that are capable of forming a four-stranded structure (Fig. 4) [Burge et al., 2006]. Four guanine bases can associate through Hoogsteen hydrogen bonding to form a square planar structure called a G-tetrad, and two or more G-tetrads can stack on top of each other to form a G-quadruplex.</p><p>Potential G-quadruplex sequences have been identified in eukaryotic telomeres, and more recently in non-telomeric genomic DNA, for example, in the nuclease-hypersensitive promoter regions of many genes, such as c-myc, chicken β-globin gene, human ubiquitin-ligase RFP2, and the protooncogenes c-kit, bcl-2, vegf, H-ras, N-ras, and K-ras [Burge et al., 2006; Neidle, 2009]. As G-quadruplexes exhibit diverse topologies and structures, targeting these high-order DNA structures for selective therapeutic intervention represents a feasible strategy since it is reasonable to assume that each target G-quadruplex has a unique architecture. Consequently, small-molecule agents capable of stabilizing a G-quadruplex structure in upstream regions essential to the promoter activity of a protooncogene will result in down-regulation of its gene expression. This premise provides a mechanistic rationale to identify ligands for selective G-quadruplex binding [Neidle, 2009], of which the proof-of-concept is provided by several small-molecule G-quadruplex-stabilizing agents, including RHPS4 [Leonetti et al., 2004], BRACO-19 [Burger et al., 2005], telomestatin [Miyazaki et al., 2012], and TMPyP4 [Le et al., 2013], with interesting antitumor activities associated with telomere capping alteration and/or inhibition of various protooncogenes, such as c-myb and bcl-2.</p><!><p>In addition to blocking protein–protein interaction as discussed above, the function of transcription factors can also be inhibited by disrupting their interactions with DNA [Berg, 2008a]. For example, c-Myc could be targeted by blocking its dimerization with Max (Table II), or by inhibiting the recruitment of c-Myc-Max dimers to their DNA recognition motif, that is, the E-box element (5′-CACGTG-3′), to block c-Myc-induced transcriptional activation of target genes (Fig. 5).</p><p>The latter strategy was demonstrated by the identification of two DNA binding inhibitors of c-Myc/Max dimers, MYRA-A [Mo and Henriksson, 2006], and NSC308848 [Mo et al., 2006]. Evidence suggests that MYRA-A might target DNA-binding domains of c-Myc-Max dimers, in lieu of the DNA recognition motif.</p><p>Similar approaches were also taken to design inhibitors of other transcription factors, including hypoxia-inducible factor (HIF)-1 and signal transducer and activator of transcription (STAT)3, by targeting their dimerization or DNA binding, which have led to the identification of HIF-1 and Stat3 inhibitors targeting either mechanism [Berg, 2008a].</p><!><p>The paradigm shift in drug discovery toward a target-based approach in the past decade has made a tremendous headway in developing new therapeutic agents targeting different clinically relevant signaling mechanisms/pathways in cancer cells. However, despite apparent advantages of targeted therapies, challenges remain in improving clinical outcomes, which is, in part, attributable to the genetic and, equally important, phenotypic heterogeneities of cancer cells. Assumptions are made that gain/loss of function of a particular target protein or pathway is the major cause for the pathogenesis or progression of cancer. However, under targeted therapy-imposed selective pressure, cancer cells might adapt their signaling circuitry to develop compensatory mechanisms by taking advantage of redundant signaling pathways or feedback/crosstalk systems to develop drug resistance. Such a "phenotypic adaptation" represents a major challenge for targeted therapy, which underlies the rationale of using a therapeutic combination with cytotoxic drugs.</p>

PubMed Author Manuscript

Operando Spectroscopic and Kinetic Characterization of Aerobic Allylic C\xe2\x80\x93H Acetoxylation Catalyzed by Pd(OAc)2/4,5-Diazafluoren-9-one

Allylic C\xe2\x80\x93H acetoxylations are among the most widely studied palladium(II)-catalyzed C\xe2\x80\x93H oxidation reactions. While the principal reaction steps are well established, key features of the catalytic mechanisms are poorly characterized, including the identity of the turnover-limiting step and the catalyst resting state. Here, we report a mechanistic study of aerobic allylic acetoxylation of allylbenzene with a catalyst system composed of Pd(OAc)2 and 4,5-diazafluoren-9-one (DAF). The DAF ligand is unique in its ability to support aerobic catalytic turnover, even in the absence of benzoquinone or other co-catalysts. Herein, we describe operando spectroscopic analysis of the catalytic reaction using X-ray absorption and NMR spectroscopic methods that allow direct observation of the formation and decay of a palladium(I) species during the reaction. Kinetic studies reveal the presence of two distinct kinetic phases: (1) a burst phase, involving rapid formation of the allylic acetoxylation product and formation of the dimeric PdI complex [PdI(DAF)(OAc)]2, followed by (2) a post-burst phase that coincides with evolution of the catalyst resting state from the PdI dimer into a \xcf\x80-allyl-PdII species. The data provide unprecedented insights into the role of ancillary ligands in supporting catalytic turnover with O2 as the stoichiometric oxidant and establish an important foundation for the development of improved catalysts for allylic oxidation reactions.

operando_spectroscopic_and_kinetic_characterization_of_aerobic_allylic_c\xe2\x80\x93h_acetoxylation_

Introduction<!>Initial mechanistic observations<!>Interrogation of the catalyst speciation by X-ray absorption spectroscopy<!>Interrogation of catalyst speciation by NMR spectroscopy<!>Kinetic studies and isotope effects<!>Evolution of a \xcf\x80-cinnamyl-PdII resting state following the burst<!>Investigation of allyl acetate reductive elimination from [PdII(\xcf\x80-cinnamyl)(OAc)]2<!>Catalytic mechanism: Overview of observations<!>Analysis of the burst phase of the reaction<!>Analysis of the post-burst phase of the reaction<!>Conclusion

<p>The Wacker process for industrial production of acetaldehyde from ethylene was discovered in 1959,1 and three years later Vargaftik et al. reported the first example of PdII-mediated allylic acetoxylation of olefins.2 The latter reaction employed stoichiometric PdCl2 in acetic acid and generated linear allyl acetates from terminal olefins, such as 1-hexene. Shortly thereafter, Anderson and Winstein reported allylic acetoxylation of cyclohexene with stoichiometric Pd(OAc)2.3 In the subsequent half century, extensive efforts were directed toward the development of catalytic methods for this reaction through the combination of PdII sources with a stoichiometric oxidant capable of promoting reoxidation of Pd0 to PdII.4 The most widely used oxidant is benzoquinone,2–13 but other oxidants include CuCl2,14,15 inorganic and organic peroxides,16–19 and hypervalent iodine reagents.20,21 Molecular oxygen is also a viable oxidant, typically when used in combination with one or more redox active cocatalysts to facilitate regeneration of the PdII catalyst.22–30</p><p>The development of ligand-supported catalyst systems has stimulated renewed interest in Pd-catalyzed allylic oxidation reactions in recent years.7,8,12,31 These catalyst systems, which contrast the simple PdII salts [e.g., PdCl2 or Pd(OAc)2] primarily employed as catalysts during the first four decades of the field, facilitate catalyst optimization efforts and introduce new opportunities to achieve catalyst controlled selectivity in the reactions. For example, chelating ligands bearing one or two sulfoxides have been used in reactions that generate branched allyl acetates from terminal olefins,7,12 rather than the more commonly observed linear allyl acetates. This ligand-controlled regioselectivity has been implemented in a number of synthetic applications,32–34 and other ligand-supported catalysts have begun showing promise in enantioselective and diastereoselective reactions.9,35–38 Ligands can also support more efficient catalysis,39–41 as revealed by the ability of 4,5-diazafluoren-9-one (DAF) to support aerobic catalytic turnover, even in the absence of redoxactive co-catalysts.31</p><p>In spite of the long history and recent progress, Pd-catalyzed allylic oxidation reactions remain at a relatively early stage of development in comparison to other Pd-catalyzed reactions, with none of the existing catalyst systems exhibiting a full complement of desirable characteristics. For example, catalyst systems that support aerobic catalytic turnover generate only linear allyl acetate products with terminal olefins,30,31 while catalyst systems that enable branched product formation exhibit relatively poor catalytic performance (e.g., 3–10 turnovers, 48–72 h reaction times) and are not effective with O2 as the oxidant.8,12 The mechanistic origin of these limitations is poorly understood. While the basic reaction steps associated with Pd-catalyzed allylic C–H oxidation are well established, few insights are available into the catalytic mechanisms, such as the identity of the turnover limiting step, the composition of the catalyst resting state, and the influence of ancillary ligands on these features. Insights of this type have played a major role in advancing other classes of homogeneous catalytic reactions. For example, Pd-catalyzed cross-coupling methods are some of the most versatile methods in synthetic chemistry. The basic reaction steps involved in these reactions were well established by the early 1970s, but the development of more efficient catalysts, expansion of the synthetic scope and utility, and the emergence of new reaction classes (e.g., Buchwald-Hartwig reactions) in subsequent decades benefitted greatly from mechanistic studies of catalytic reactions.42–45 For example, insights into the kinetic bottlenecks of the catalytic cycle and characterization of the catalyst species, including on- and off-cycle species,46–48 provided a foundation for the discovery of new ligands,49–51 catalyst systems,52 and reaction conditions that overcame existing limitations.53,54 The iterative synergy between empirical and mechanistic studies is much less evident in the field of Pd-catalyzed allylic oxidation reactions; however, the emergence of ligand-supported catalyst systems amenable to this approach is still relatively recent.</p><p>Here, we report the first kinetic and mechanistic investigation of a Pd-catalyzed method for aerobic allylic C–H oxidation, focusing on DAF/Pd(OAc)2-catalyzed acetoxylation of allylbenzene, for which the basic steps are illustrated in Scheme 1. The initial report of this catalyst revealed that DAF is unique in its ability to support aerobic catalytic turnover among a series of (bi)pyridyl-based ligands (Scheme 2),31,55,56 and investigation of stoichiometric reactions with π-allyl-PdII species showed that DAF promotes reductive elimination57 of the allyl acetate product.31,58 The implications of these observations were not extended to catalytic conditions, however, and the present study provides a comprehensive analysis of the catalytic mechanism, including kinetic studies and operando X-ray absorption spectroscopy (XAS) and NMR spectroscopic analysis of the catalytic reaction, which reveal changes in the identity of the turnover limiting step and evolution of the catalyst species during the reaction. The results provide valuable insights into Pd-catalyzed allylic oxidation reactions and the ability of DAF to support aerobic catalytic turnover,59–69 and this study establishes an important foundation for future empirical and mechanistic efforts directed toward the identification of more efficient and synthetically useful allylic C–H oxidation catalyst systems.</p><!><p>In a recent study, we showed that DAF/Pd(OAc)2-catalyzed allylic acetoxylation of allylbenzene exhibits a kinetic burst at the beginning of the reaction and generates a dimeric PdI species under the catalytic conditions.70 These preliminary results, reproduced in Figure 1, provided a starting point for the present investigation, with the goal of understanding the origin and implications of these observations. The burst is accompanied by a change in the color of the reaction solution from yellow to red, but, as the reaction progresses past the burst, the reaction solution changes color again from red to yellow. The latter color persists for the remainder of the reaction.71 A red PdI dimer, [PdI(DAF)(OAc)]2 2, was independently synthesized and characterized by X-ray crystallography (Figure 1B), and it exhibits a UV-visible absorption spectrum that closely matches the UV-vis spectra observed from the catalytic reaction mixture (Figure 1C).70</p><!><p>The data noted above imply that the oxidation state of the Pd catalyst changes during the course of the reaction. Palladium K-edge X-ray absorption spectroscopy probes the excitation of Pd 1s electrons, and the "near edge" absorption energy, corresponding to the 1s → 5p orbital transition, provides insight into the oxidation state of Pd complexes.72,73 This technique has been used to probe a number of homogeneous Pd-catalyzed reactions,74–79 including recent analysis of a Pd-catalyzed aerobic oxidation of an aromatic C–H bond.80 In order to conduct analogous experiments for the allylic acetoxylation reaction, a reactor was designed and constructed to enable operando studies on the high photon-flux beamline at Argonne National Laboratories (see Supporting Information for details; Figures S3–S6).</p><p>Pd K-edge XANES spectra were obtained for a series of Pd0, PdI, and PdII reference compounds: Pd0 (Pd2(dba)3), PdI ([PdI(DAF)(OAc)]2 (2)), and PdII (1:1 DAF/Pd(OAc)2 (1); [PdII(π-cinnamyl)(OAc)]2 (3); and [PdII(DAF)(π-cinnamyl)]BF4 (4)). Similar spectra were then acquired during the course of the catalytic reaction,81 with a time course of the edge energy shown together with selected reference compounds in Figure 2A. Upon initiation of the reaction by injecting allylbenzene, the edge energy shifts rapidly from its initial position associated with 1 to a lower energy position, consistent with reduction of the catalyst during the burst (spectrum B). The new edge energy is very similar to that of the PdI dimer reference (cf. 2, Figure 2B). Analysis of the extended X-ray absorption fine-structure (EXAFS) region of the spectra revealed three prominent scattering peaks during the first 13 min of the reaction, and these peaks are very similar to those observed with an authentic sample of the PdI dimer 2 (Figure 3A). Collectively, the XANES and EXAFS spectra support the formation of PdI species 2 as a catalytic intermediate during the burst phase of the reaction.</p><p>As the reaction proceeds, the edge increases gradually to higher energy, ultimately reaching an energy intermediate between that of DAF/PdII(OAc)2 1 and the PdI dimer 2 (Figure 2A). An isosbestic point is evident in the XANES spectra during this phase of the reaction (cf. Figure 2A), suggesting that the PdI dimer converts directly into a new Pd species, which has a higher formal oxidation state. The XANES spectrum for the catalyst resting state in the post-burst phase (PB) is clearly different from the DAF/PdII(OAc)2 pre-catalyst 1, and it closely resembles the spectra of the independently prepared π-allyl-PdII species, the dimeric species [PdII(π-cinnamyl)(OAc)]2 (3) and the monomeric DAF-ligated π-allyl-PdII species, [PdII(DAF)(π-cinnamyl)]BF4 (4) (Figure 2C). The changes evident in the XANES spectra were also evident in a comparison of the inflection point in the spectra for the Pd species present during the reaction relative to the approximate edge energies of the PdI and PdII reference compounds (Figure 4). Analysis of the EXAFS spectral region during the post-burst portion of the reaction also reveal similarities between the post-burst PdII intermediate (PB) and the π-allyl-PdII reference compounds (Figure 3B).</p><!><p>The catalyst speciation during the reaction time course was also interrogated by NMR spectroscopic methods. The reaction employs a mixture of AcOH/NaOAc as the solvent and source of acetate nucleophile, and the predominant palladium species observed by 1H NMR spectroscopy at the start of the reaction was identified as an "ate" complex, Na[(κ1-DAF)PdII(OAc)3] (5). This assignment was supported by NMR integrations and independent titration studies involving addition of NaOAc to a solution of DAF and PdII(OAc)2 (Figures S20–S24). A time course of the reaction was monitored by 1H NMR spectroscopy by using a sealed NMR tube pressurized with 3.2 atm of static O2. The data reveal rapid product formation with concurrent conversion of the PdII ate complex 5 into the PdI dimer 2 (Figure 5). The concentration of PdI maximizes at the end of the kinetic burst of product formation (t ~ 20 min), after which the rate of cinnamyl acetate product formation decreases to a steady-state rate. Unligated DAF becomes evident in the reaction mixture and its concentration increases during the steady state reaction period (Figure 5 - expansion).</p><p>To investigate the catalyst speciation at early reaction times (i.e., during the burst period), thick-walled NMR tubes were prepared with different pressures of O2.82–84 Plots showing the formation of cinnamyl acetate and the PdI dimer (2) at 0.2–6.2 atm O2 are provided in Figure 6. Nearly identical rates of product formation (i.e., independent of the pO2) are evident during the burst phase; however, the rate of product formation after the burst phase increases with increasing pO2 (Figure 6A). Complementary analysis of the Pd species shows that the amount of PdI dimer 2 generated during the burst phase decreases with increasing pO2 (Figure 6B; see Supporting Information for further details).</p><p>To complement these data, the recently developed Wisconsin High Pressure NMR Reactor (WiHP-NMRR)85–87 was used to monitor the dependence of the rate on O2 pressure (pO2). This apparatus provides continuous gas circulation in the liquid during the NMR relaxation delay (d1). This dynamic gas supply allows for better control of dissolved O2 concentration during the reaction and avoiding complications that can arise from the lack of gas-liquid mixing in an NMR tube when a static pressure of O2 is used.88 Consistent with the data in Figure 6, the rate of product formation during the burst phase was not affected by changes in pO2, while a saturation dependence of the rate on pO2 was evident from kinetic data after the burst (Figure 7).</p><p>Another set of studies was conducted to probe the influence of substrate concentration on the reaction rate and Pd speciation during the early periods of the reaction (0.135–1.08 M allylbenzene). The overall rate of product formation increased with higher [allylbenzene] (Figure 8A). An increase in the rates of formation and decay of the PdI dimer 2 with increasing [allylbenzene] results in a similar maximum concentration of 2 under all conditions, but a shorter lifetime for this species at higher substrate concentration (Figure 8B).</p><!><p>More thorough kinetic studies were performed to probe the contributions of each of the reaction components to the reaction rate during the burst and post-burst phases of the reaction (Figures 9 and 10). These studies were conducted at 40 °C and 80 °C, respectively, by removing aliquots of the reaction mixture and analyzing product formation by gas chromatography (GC). Analysis of the burst phase was carried out at lower temperature to enable acquisition of more precise data, but the results are consistent with general trends observed at higher temperatures. A linear dependence of the rate on the 1:1 DAF/PdII(OAc)2 catalyst concentration [1] was observed during the burst phase, while a slight saturation dependence was evident in the post-burst regime (Figures 9A and 10A, respectively). In both the burst and post-burst phases, the rate exhibits a nearly linear dependence on [DAF] up to a 1:1 DAF/PdII(OAc)2 ratio; however, increasing [DAF] beyond a 1:1 ratio decreases the rate of the burst, while it does not affect the rate during post-burst turnover (Figures 9B and 10B). A nearly linear dependence on [allylbenzene] was observed in the burst and post-burst phases (Figures 9C and 10C), but the data intersect at the origin for the burst kinetics, while the post-burst kinetic data exhibit a substantial non-zero intercept. The positive order in [allylbenzene] in both cases correlates with the kinetic influence of allylbenzene on the rate of formation and decay of the PdI dimer 2 observed in the NMR spectroscopic studies shown in Figure 8B.</p><p>Two deuterium-labeled substrates were synthesized to probe intramolecular and intermolecular kinetic isotope effects (KIEs). 4-Allylbiphenyl (6) was selected as the substrate for these studies because it is not volatile and is easier to purify than the parent allylbenzene.89 Independent rates were measured with 6 and 6-d2 substrates, and intramolecular competition KIEs with 6-d1 were determined via 1H NMR spectroscopy by quantifying the product formation derived from C–H or C–D activation. The primary KIEs evident in the burst phase from both sets of experiments (Table 1, entries 1 and 2) are rather small, but very similar to that observed in a recent allylic amination reaction.90 In the period of the reaction after the burst, the intramolecular competition experiment revealed a KIE similar to that observed in the burst phase (cf. entry 3 vs entries 1 and 2), but essentially no KIE was observed from the independent rate measurements (entry 4).</p><!><p>The XAS data described above implicate the possibility of a π-allyl-PdII species as the catalyst resting state during the post-burst phase of the reaction. Initial attempts to observe such a species directly by 1H NMR spectroscopy were unsuccessful, however, possibly reflecting a high degree of fluxionality associated with such a species.91–93 The NMR time course in Figure 5 suggests that the DAF ligand dissociates from the Pd center during the reaction, a feature that could further contribute to the fluxionality of a π-allyl-PdII species and complicate its detection by 1H NMR spectroscopy.</p><p>To overcome this complication, we analyzed the catalytic acetoxylation of 4-fluoroallylbenzene to probe the reaction by 19F NMR spectroscopy. Operando analysis of the catalytic reaction was conducted throughout the time course using the WiHP-NMRR in an effort to identify possible intermediates. During the burst phase, 19F NMR resonances are evident only for the alkene starting material (−113.4 ppm) and cinnamyl acetate product (−110 ppm) (e.g., Figure 11a). As the reaction proceeds, however, a new resonance becomes evident at −109.2 ppm. (Figure 11b; see Figure S12 for additional spectra). The chemical shift of this peak, denoted I, is very similar that observed from an independently prepared sample of the 4-fluorophenyl derivative of the π-cinnamyl-PdII dimer 3 (i.e., 3F) and the same compound in the presence of DAF (DAF:Pd = 1:1) (Figures 11c and 11d). Integration of this resonance during the course of the catalytic reaction show that this species accounts for nearly all of the Pd catalyst present in solution (Figure 12), thereby providing direct support for evolution of the catalyst into a π-allyl-PdII resting state following the burst.</p><p>In light of these observations, the initial time course of the catalytic acetoxylation of allylbenzene was analyzed by comparing reactions conducted with Pd(OAc)2 and [PdII(π-cinnamyl)(OAc)]2 (3) as the source of Pd catalyst, under otherwise identical conditions. The data in Figure 13 show that no burst is observed when 3 is used as the Pd precatalyst, and the initial rate in this reaction resembles that of the post-burst rate observed from the reaction initiated with Pd(OAc)2 as the precatalyst.</p><!><p>Identification of a π-allyl-PdII resting state during the catalytic reaction prompted us to prepare the well-defined [PdII(π-cinnamyl)(OAc)]2 species 3 to explore its reactivity under catalytically relevant conditions (Figure 14). As a control experiment, negligible reductive elimination of cinnamyl acetate57 was observed under N2 in the absence of the DAF ligand (Figure 14, condition a). Some reactivity was observed upon addition of DAF to the solution of 3 under N2 (condition b), consistent with previously reported observations,31 and further enhancement of the reaction was observed upon adding an equivalent of DAF/PdII(OAc)2 (condition c). The latter conditions led to significant formation of the PdI dimer 2. These reactions were also enhanced by conducting the reaction under an atmosphere of O2 (conditions d and e). The last of these conditions (e) resemble the reaction conditions present during the burst phase of the catalytic reaction, when allylic C–H activation and formation of an allyl-PdII intermediate will take place in the presence of additional DAF/PdII(OAc)2 pre-catalyst, which can react with the Pd0 species formed upon reductive elimination of cinnamyl acetate and generate the PdI dimer 2 (see further discussion below). The presence of allylbenzene has negligible effect on the rate of reductive elimination (see Figure S17 in the Supporting Information), suggesting that the beneficial kinetic effect of allylbenzene after the burst (cf. Figure 10C) does not arise from the reductive elimination step, but rather from its contribution to catalyst speciation (i.e., by promoting conversion of PdI dimer 2 into the steady-state PdII catalyst; cf. Figure 8B).</p><!><p>The data presented above provide key insights into the DAF/PdII(OAc)2-catalyzed acetoxylation of allylic C–H bonds, including factors that affect the rate of the reaction and the identity of the catalyst resting state (Table 2). Briefly, the kinetic burst phase evident at the beginning of the reaction is characterized by a primary deuterium KIE (independent rate measurement) and a lack of rate-dependence on pO2, and the Pd catalyst evolves from a DAF-ligated PdII-acetate species into a DAF-bridged PdI dimer. The ensuing post-burst phase lacks a primary KIE, exhibits a rate-dependence on pO2, and the Pd catalyst resting state evolves from the DAF-bridged PdI dimer into π-allyl-PdII species with bound and unbound DAF ligand.</p><!><p>The deuterium kinetic isotope effect observed during the burst supports allylic C–H activation as the rate-limiting step during this phase.90,94,95 The resulting allyl-PdII intermediate undergoes reductive elimination of cinnamyl acetate to afford a Pd0 species that reacts with a DAF/PdII(OAc)2 species to afford the PdI dimer 2 (Scheme 3; cf. Figure 14). The mechanistic details of reduction elimination and PdII/Pd0 comproportionation steps are not well understood and warrant attention in future studies, but formation of analogous dimeric PdI species has been studied extensively in Pd-catalyzed cross-coupling reactions.52,96 PdI species have also been identified in oxidative carbonylation reactions,97–100 but the DAF-ligated dimer 2 represents the first example of a PdI intermediate identified in ligand-supported Pd-catalyzed aerobic oxidation reactions.70 A second example was recently reported by Albéniz and Maseras, who detected a PdI dimer in the aerobic oxidative homocoupling of alkynes with Ph3P as an ancillary ligand.101 The use of PdI dimers as air-stable catalyst precursors in many non-oxidative coupling reactions reactions52 raises prospects for similar opportunities in aerobic oxidation reactions.</p><!><p>The burst phase concludes as DAF-ligated PdII-acetate species are depleted and the PdI dimer concentration reaches a maximum. The collection of kinetic and spectroscopic insights obtained from the reaction following the burst are incorporated into the overall catalytic mechanism proposed in Scheme 4. Previous studies have shown that the PdI dimer can undergo oxidation to a DAF/PdII-acetate species in the presence of O2,70,102 and the NMR data in Figure 10 show that allylbenzene promotes the conversion of PdI dimer into a new PdII form of the catalyst, identified as the allyl-PdII species 3 (cf. Figure 12), which serves as the catalyst resting state following the burst. This structure is consistent with the appearance of free DAF ligand under these conditions. The stoichiometric reactivity studies with 3, depicted in Figure 14, show that reductive elimination of cinnamyl acetate57 is promoted by DAF, O2, and DAF/Pd(OAc)2; however, following the burst, excess DAF/Pd(OAc)2 species are no longer available in the reaction mixture to promote reductive elimination. This difference between the burst and post-burst conditions rationalizes the change in catalyst speciation and turnover-limiting steps (cf. Table 2). These changes, in turn, account for the lack of a deuterium KIE for C–H activation during the post-burst phase of the reaction and the similarity between the of post-burst turnover rate and the rate of the reaction when π-allyl-PdII dimer 3 is used as the pre-catalyst (cf. Figure 13).</p><p>DAF plays an important role in supporting catalytic turnover under steady-state (post-burst) conditions by facilitating reductive elimination of cinnamyl acetate from the allyl-PdII species. DAF coordination to the allyl-PdII dimer 3 is proposed to generate the monomeric DAF-ligated allyl-PdII species 4, which can undergo reversible formation of cinnamyl acetate (Figure 15).31,58 This reaction is driven by trapping of Pd0 (7) by O2, which is proposed to be the turnover limiting step during this phase of the reaction (Table 2).103,104 No build-up of PdI dimer 2 is observed after the burst phase due to the low steady-state concentration of the DAF/PdII-acetate species available to trap Pd0 species generated in the reductive elimination step. The reaction of O2 with the Pd0 species 7 is proposed to proceed in a manner similar to previously reported reactions between O2 and related well-defined nitrogen-chelated Pd0 model complexes.41,105–107</p><p>As noted in the Introduction, very few mechanistic studies of Pd-catalyzed allylic oxidation reactions have been reported. The two most thorough precedents were reported recently by the groups of Labinger/Bercaw10 and Fristrup,94 and focused on bipyrimidine (bpym)/PdII(OAc)2 and bis(sulfoxide)/PdII(OAc)2 catalyst systems, respectively. The turnover-limiting step in the former reaction was identified as alkene binding to (bpym)PdII(OAc)2, while the latter reaction was attributed to C–H activation. The presence of a π-allyl-PdII catalyst resting state and turnover-limiting reductive elimination are unprecedented, and these two observations likely reflect the use of O2, rather than benzoquinone, as the stoichiometric oxidant. Benzoquinone was used as the stoichiometric oxidant in both of the previously studied catalytic reactions, and it is known to promote reductive elimination from π-allyl-PdII species.7,58,108,109 These considerations highlight opportunities for the development of improved catalyst systems and/or ligand designs that support more effective reductive elimination while also enabling efficient reaction of Pd0 with O2. Efforts toward this end have been initiated.</p><!><p>This study has implemented a number of different experimental approaches to gain insights into a Pd-catalyzed allylic oxidation reaction that is capable of using O2 as the stoichiometric oxidant due to the use of 4,5-diazafluoren-9-one (DAF) as an ancillary ligand. Operando XAS and NMR spectroscopic studies provide clear evidence for two kinetic phases, characterized by different catalyst resting states and turnover-limiting steps. The initial burst phase features the formation of an unusual DAF-bridged dimeric PdI species, which then evolves into a π-allyl-PdII catalyst resting state following the burst phase of the reaction. Allylic C–H activation is the rate determining step during the burst phase, while catalyst reoxidation by O2 is turnover-limiting following the burst. The results described herein provide valuable new insights into this widely studied catalytic reaction, perhaps most importantly shedding light on the contribution of ancillary ligands to key steps in the catalytic mechanism. These insights provide a foundation for the exploration of new ancillary ligands and catalyst designs capable of efficient Pd-catalyzed allylic C–H oxidation.</p>

PubMed Author Manuscript

Expected values for gastrointestinal and pancreatic hormone concentrations in healthy volunteers in the fasting and postprandial state

BackgroundGastrointestinal hormones regulate intestinal transit, control digestion, influence appetite and promote satiety. Altered production or action of gut hormones, including glucagon-like peptide-1 (GLP-1), glucose-dependent insulinotropic polypeptide (GIP) and peptide YY (PYY), may contribute to the biological basis of obesity and altered glucose homeostasis. However, challenges in analytical methodology and lack of clarity on expected values for healthy individuals have limited progress in this field. The aim of this study was to describe expected concentrations of gastrointestinal and pancreatic hormones in healthy volunteers following a standardized meal test (SMT) or 75 g oral glucose tolerance test (OGTT).MethodsA total of 28 healthy volunteers (12 men, 16 women; mean age 31.3 years; mean body mass index 24.9 kg/m2) were recruited to attend a hospital clinic on two occasions. Volunteers had blood sampling in the fasting state and were given, in randomized order, an oral glucose tolerance test (OGTT) and standardized mixed liquid meal test with venepuncture at timed intervals for 4 h after ingestion. Analytical methods for gut and pancreatic hormones were assessed and optimized. Concentrations of gut and pancreatic hormones were measured and used to compile ranges of expected values.ResultsRanges of expected values were created for glucose, insulin, glucagon, GLP-1, GIP, PYY and free fatty acids in response to a standardized mixed liquid meal or OGTT. Intact proinsulin and C-peptide levels were also measured following the OGTT.ConclusionsThese ranges of expected values can now be used to compare gut hormone concentrations between healthy individuals and patient groups.

expected_values_for_gastrointestinal_and_pancreatic_hormone_concentrations_in_healthy_volunteers_in_

Introduction<!>Study design<!>Study visits and stimulation tests<!><!>Study visits and stimulation tests<!>Analytical methodology<!><!>Statistical analysis<!>Results<!><!>Results<!><!>Discussion<!>

<p>The rising prevalence of nutritional disorders has led to increased scientific interest in the mechanisms which regulate nutrient disposal, appetite and satiety. Hormones such as glucagon-like peptide-1 (GLP-1) and peptide YY (PYY), which are released from the gastrointestinal tract in response to nutrient ingestion, may mediate satiety and reduce appetite.1,2 GLP-1- and glucose-dependent insulinotropic polypeptide (GIP) function as incretins, promoting the release of insulin from the pancreas in response to a meal.3 The complex relationships between gastrointestinal and pancreatic hormones are now thought to regulate energy balance and may be affected by diseases such as type 2 diabetes mellitus (T2DM).4 Bariatric surgery, which provides an effective method of long-term weight control, is associated with increased postprandial concentrations of GLP-1, PYY and improved glucose tolerance.5,6 However, the expected circulating concentrations of many gastrointestinal and pancreatic hormones in healthy individuals have yet to be established, yet are essential for the interpretation of altered levels seen in metabolic and gastrointestinal conditions.</p><p>Despite the high levels of interest in gut hormones, progress in this field has been limited by the development of robust analytical methods for their quantitation in plasma. Many of these peptides are present in picomolar concentrations and require highly sensitive assays, which can measure concentrations as low as 1–10 pmol/L. Highly specific assays are also required, as many peptides have important structural similarities. Glucagon and oxyntomodulin, for example, differ only at their C-termini and are indistinguishable using antibodies that target the shared peptide sequence.7 A number of commercially available assays have been shown to be unfit for purpose.8,9 However, analytical methodology and the concordance of results obtained using different assays are improving. With this advancement in methodology, it is now possible to permit comparison of plasma peptide concentrations in health and disease.</p><p>Concentrations of several gastrointestinal hormones are altered after bariatric surgery,10 in individuals with T2DM4 and after gestational diabetes.11 However, expected concentrations of a comprehensive range of gut hormones in healthy individuals have not been described in depth. The aim of the current study was to provide ranges of expected values using reliable and readily-available assays for gastrointestinal and pancreatic hormones in healthy human subjects in the fasting state and after both an oral glucose tolerance test (OGTT) and a standardized mixed liquid meal test (SMT).</p><!><p>Healthy male and female volunteers aged 18–65 years old were recruited using advertisements placed in Addenbrooke's Hospital and the University of Cambridge. To fulfil the entry criteria, healthy volunteers needed to be free from chronic diseases and recent acute conditions, such as infections, diarrhoea or constipation, and have a body mass index (BMI) of 18–35 kg/m2. Healthy participants were either taking no medication or were stable on medication that was considered unlikely to interfere with the results of the study. Participants with known pre-existing anaemia, diabetes or endocrine disorders and pregnant or lactating women were excluded from this study. The study was given ethical approval by the local research ethics committee (13/EE/0195), and all participants gave full written consent.</p><!><p>Participants attended a Clinical Research Facility at 09:00 h on two occasions following an overnight fast. The evening before each visit, participants prepared for themselves a standardized pasta meal containing 15% protein, 30% fat and 55% carbohydrate which was designed to provide 33% of their daily calorie requirement based on an estimation of their metabolic rate and activity levels.12 After the meal, participants were allowed free access to water but were asked to avoid food, caffeinated and calorie-containing drinks overnight for 12 h prior to the study visit.</p><p>Participants attended for two visits and received a 75 g OGTT and a SMT in randomized order (Table 1 for composition). Blood was taken at baseline and at timed intervals (0, 15, 30, 45, 60, 90, 120, 150, 180, 210 and 240 min) after the OGTT or SMT for analysis of glucose, non-esterified-free fatty acids (FFA) and pancreatic and gut hormones. Participants were asked to remain sedentary throughout the testing process.</p><!><p>Nutritional contents of OGTT and Ensure plus standardized meal.</p><p>OGTT: oral glucose tolerance test.</p><!><p>The OGTT was produced using 82.5 g glucose monohydrate (equivalent to 75 g pure glucose) which was dissolved in 250 mL chilled water, and a further glass of 250 mL water was given afterwards to wash out the oral cavity. The SMT consisted of a 237 mL bottle of Ensure plus, a balanced nutritional supplement containing 11 g fat (28%), 13 g protein (15%) and 50 g carbohydrate (57%). Participants were also given 250 mL of water after the meal was consumed. Each drink (OGTT or SMT) was consumed within a 5 min time frame. Altogether, the OGTT and SMT provided 300 and 350 kcals, respectively.</p><!><p>All analyses were performed in the Core Biochemical Assay Laboratory in Addenbrooke's Hospital. The analytical methods used with performance measures are given in Table 2 (see supplemental information for more details). All analyses demonstrated acceptable linearity and recovery and passed standard quality control measures. Samples with known common interferences such as haemolysis, lipaemia and hyperbilirubinaemia were excluded from this analysis.</p><!><p>Summary of methods used for analysis of biochemical markers, gut and pancreatic hormones.</p><p>CV: coefficient of variation; EDTA: ethylenediaminetetraacetic acid; FFA: free fatty acids; GLP-1: glucagon-like peptide-1; GIP: glucose-dependent insulinotropic polypeptide; PYY: peptide YY.</p><!><p>Characteristics of participants are described as mean (±SD). A range of expected values was produced using the non-parametric method as recommended by the IFCC through the Clinical and Laboratory Standards Institute (CLSI).13 The 2.5th and 97.5th percentiles were calculated to give an interval which included 95% of values from this sample. Other approaches, such as bootstrapping and robust methods were not used in this study but may have some advantages in a small sample.14</p><p>The appropriate assessment of outliers is important in establishing ranges of expected values. Some outliers were identified and were assessed for each individual participant and each analyte. If any clear pathological cause was evident, the outlier was removed. However, in practice, the outliers could not be explained on the basis of known pathology and were not excluded. Some individuals were found to have elevated fasting glucose and insulin concentrations. Although these findings are suggestive of prediabetes, they are also extremely common in the reference population, and these data were therefore kept in the analysis.</p><!><p>A total of 28 healthy volunteers were recruited (12 male, 16 female; Table 3). Most healthy volunteers were aged 22–40 years (31.3 ± 10.9 years) with a lean body mass index 24.9 ± 3.7 kg/m2. Participants had normal haemoglobin, white cell count, alanine aminotransferase (ALT), thyroid-stimulating hormone (TSH) and creatinine at baseline with an HbA1c in the non-diabetic range (33.6 ± 3.7 mmol/L).</p><!><p>Baseline characteristics of study participants. Characteristics shown as mean (SD) for continuous variables and n (%) for categorical variables.</p><!><p>Data from these 28 participants were used to create ranges of expected values for glucose, insulin, glucagon, total GLP-1, PYY, GIP and FFA before and after an OGTT and SMT (see Figures 1 and 2; Tables S1 and S2). C-peptide and intact proinsulin were measured following an OGTT only (n = 20 participants; Figure 2).</p><!><p>Ranges of expected values for glucose, insulin, glucagon, intact proinsulin, C-peptide, GLP-1, PYY, GIP and FFA in the fasting state and after a 75 g oral glucose tolerance test (OGTT). Note that intact proinsulin and C-peptide were measured at time-points 0, 30, 60, 90, 120, 180 and 240 minutes only on samples from 20 participants. The remaining analytes were measured on samples from 28 participants.</p><p>FFA: free fatty acids; GLP-1: glucagon-like peptide-1; GIP: glucose-dependent insulinotropic polypeptide; LRL: lower range limit; 2.5th percentile; PYY: peptide YY; URL: upper range limit; 97.5th percentile.</p><p>Ranges of expected values for glucose, insulin, glucagon, GLP-1, PYY, GIP and FFA in the fasting state and after a standardized meal test (SMT), based on measurements taken from 28 participants.</p><p>FFA: free fatty acids; GLP-1: glucagon-like peptide-1; GIP: glucose-dependent insulinotropic polypeptide; LRL: lower range limit; 2.5th percentile; PYY: peptide YY; URL: upper range limit; 97.5th percentile.</p><!><p>In this study, ranges of expected values were provided for glucose, insulin, glucagon, total GLP-1, PYY, GIP and FFA in healthy volunteers in fasting and postprandial circumstances, in response to both an OGTT and a standardized mixed liquid meal. Ranges for intact proinsulin and C-peptide were produced in healthy volunteers following an OGTT. These ranges of expected values provide reference values for comparison with concentrations in various disease states in future studies. There is currently no standardized stimulation test for the assessment of postprandial gut hormone concentrations. Previous studies have used a variety of stimuli including standardized breakfasts, mixed liquid meals, glucose and ice cream,15–19 which presents considerable challenges in the comparison of findings. As gut hormone release is triggered by protein and fat, as well as glucose, a mixed meal is generally a well-tolerated and reliable stimulus of gut hormone secretion. The liquid standardized meal used in the current study circumvents the requirement for mechanical food breakdown in the stomach, and is therefore suitable for comparison with patient groups, such as postbariatric or other gastrointestinal surgery. Oral glucose tolerance tests using a 75 g glucose bolus are frequently used for the assessment of glucose metabolism, but are not well tolerated after gastric bypass surgery.</p><p>To assess variability in gut hormone secretion, our data suggest that sampling at 0 and 30 min time-points would encompass the majority of variability in hormone measurements in healthy subjects, although a longer period of sampling may be required in populations where gastric emptying is more variable and in response to solid meal stimuli. At later time points, the inhibitory effects of elevated GLP-1 and PYY concentrations on gastric emptying tend to reduce the rate of nutrient delivery into the small intestine. This means the gut hormone concentrations then become a complex balance between the strength of the feedback control of gastric emptying, the quantity of nutrients still awaiting digestion and absorption and the secretory potential of the small intestine.</p><p>Gut hormone analysis was based on methods that are routine within our laboratory and we have found to be reliable in most circumstances. Detailed methodology and performance data have been given for each analyte (Table 2 and supplemental material). Many other studies advocate the use of protease inhibitors and DPP-IV inhibitors which can be added to blood tubes prior to sampling or provided in the form of specialized pretreated tubes.20 DPP4 or protease inhibitor-treated tubes provide results up to 20–30% higher for some analytes,21 which reduces the need for high analytical sensitivity and may increase analytical performance. However, in this study, we chose specimen preparation and processing techniques that minimize degradation while using equipment and blood tubes readily available in most health-care environments and require no added reagents. Although rapid processing is essential for gut hormone and insulin analyses, these sample preparation methods were found in this study to be feasible both in the hospital research unit and have been tested in a mobile laboratory in the field. DPP4 inhibitors would be required for the measurement of active GLP-1 concentrations, but in the current study, we elected to assay total GLP-1 as a better potential measure of the hormone secretory rate.</p><p>Choosing a suitable group of individuals in perfect health for the development of ranges of expected values is challenging. The study group used here were relatively young and lean and were recruited using advertisements in University and Hospital buildings. Although most participants took no regular medication, several of the female participants were on the oral contraceptive pill, and this may have influenced the results. Gastrointestinal and pancreatic hormone concentrations were not studied at different specified times during the female menstrual cycle or in post-menopausal women. Our study group were predominantly European and ethnic differences in GLP-1 responses have been previously identified.22 We did not study diurnal variation in gut and pancreatic hormones although all visits took place in the morning, and for insulin, glucagon, GLP-1, PYY and GIP, diurnal changes are likely to be influenced primarily by nutrient ingestion at mealtimes.7 Although hormone concentrations can be influenced by age, gender and obesity, the current study has inadequate sample size to provide formal reference intervals or partitioned reference estimations for subgroups of the population.13</p><p>Despite its limitations, this study provides a comprehensive set of data for expected concentrations of gut hormones in healthy human subjects. Our data show reasonable consistency with other published work, although direct comparison is difficult due to the variety of stimulation tests and analytical methods used. For example, Vilsbøll et al. studied fasting and postprandial gut hormone responses in eight healthy individuals who had previously had a normal OGTT.23 A 260 kcal meal of bread, margarine and jam was given with milk. For GLP-1, fasting concentrations of 15–20 pmol/L (50–66 pg/mL) were obtained and concentrations peaked at 30 pmol/L (99 pg/mL) after the meal (1 pmol/L GLP-1 = 3.297 pg/mL GLP-1). For GIP, fasting concentrations were around 15 pmol/L (66 pg/mL) with a peak concentration of 80–90 pm (352–396 pg/mL) after the meal (1 pmol/L GIP = 4.4 pg/mL GIP). Visbǿll's population had lower concentrations of fasting and postprandial insulin (fasting 15–20 pmol/L; peak ∼180 pmol/L) compared with the population under investigation in the current study, but participants had been prescreened with an OGTT prior to enrolment and individuals with a family history of diabetes were excluded. Concentrations of fasting C-peptide in Visbǿll's study were 400 pmol/L with peak concentrations of 1500 pmol/L after the meal.23 In contrast, Alsalim et al. studied 24 healthy lean volunteers who had a meal of 511 kcal, consisting of a sirloin steak with potatoes, vegetables and sorbet.24 Concentrations of intact GIP were 10 pmol/L (44 pg/mL) fasting and peaked at around 75 pmol/L (330 pg/mL) postprandially. Insulin concentrations were around 30 pmol/L fasting and around 250 pmol/L postprandially. C-peptide concentrations were 300 pmol/L fasting and around 1000 pmol/L postprandially. Unfortunately, concentrations of total GLP-1 and GIP were not reported.24 In the current study, the use of a liquid meal, rather than solid food, is likely to explain the more rapid increase in gut hormones seen in comparison to Visbǿll and Alsalim's work.23,24</p><p>In the current study, glucagon concentrations after the SMT generally exhibited a small transient elevation, and thereafter remained similar to baseline. This likely reflects a complex balance between inhibitory signals acting on pancreatic alpha cells, including glucose and GLP-1, and stimulatory signals such as amino acids. After the OGTT, we expected to see a less complex picture, as glucagon should be predominantly suppressed by the elevated glucose and GLP-1 concentrations. Indeed, in the majority of healthy volunteers, post-OGTT glucagon concentrations reached a nadir by 45 to 60 min. A few participants (6/28), however, exhibited an elevated glucagon of >50% at some point in the first hour. It is currently not clear whether this represents a true increase in biologically active glucagon from the pancreas, release of glucagon from the intestine, or detection by the glucagon assay of longer peptides containing the glucagon sequence that might be released from the gut or pancreas. It is also possible that under physiological conditions, the glucagon-stimulating effects of GIP outweigh the glucagon-suppressing effects of GLP-1.25,26 It has been proposed that the gut can release glucagon under certain conditions, including following total pancreatectomy where pancreatic release should be impossible,27 or after Roux-en-Y gastric bypass surgery.28,29 However, when GLP-1 secretion is high, some glucagon assays cross-react with longer proglucagon-derived species containing the glucagon sequence, particularly glicentin, leading to debate over whether the gut ever releases intact glucagon.30,31 In addition to these reports from populations with altered pathology, and although we were unable to find active glucagon in human intestinal tissue by LC-MS,32 our data suggest that some healthy people may also exhibit post-OGTT glucagon elevation.</p><p>In conclusion, this comprehensive set of expected values for fasting and postprandial gastrointestinal and pancreatic hormone concentrations in healthy human subjects facilitates future comparison between healthy individuals and those with metabolic or endocrine disorders in response to an OGTT or standardized liquid meal.</p><!><p>Click here for additional data file.</p><p>Supplemental material, sj-pdf-1-acb-10.1177_0004563220975658 for Expected values for gastrointestinal and pancreatic hormone concentrations in healthy volunteers in the fasting and postprandial state by Claire L Meek, Hannah B Lewis, Keith Burling, Frank Reimann and Fiona Gribble in Annals of Clinical Biochemistry</p>

PubMed Open Access

GAS41 recognizes di-acetylated histone H3 through a bivalent binding mode

GAS41 is a chromatin-associated protein that belongs to the YEATS family and is involved in the recognition of acetyl-lysine in histone proteins. A unique feature of GAS41 is the presence of a C-terminal coiled-coil domain, which is responsible for protein dimerization. Here, we characterized specificity of the GAS41 YEATS domain, and found that it preferentially binds to acetylated H3K18 and H3K27 peptides. Interestingly, we found that full-length, dimeric GAS41 binds to di-acetylated H3 peptides with an enhanced affinity when compared to mono-acetylated peptides, through a bivalent binding mode. We determined the crystal structure of the GAS41 YEATS domain with H3K23acK27ac to visualize the molecular basis of di-acetylated histone binding. Our results suggest a unique binding mode in which full-length GAS41 is a reader of di-acetylated histones.

gas41_recognizes_di-acetylated_histone_h3_through_a_bivalent_binding_mode

INTRODUCTION<!>The GAS41 YEATS domain preferentially binds H3K18ac and H3K27ac peptides<!>Full-length GAS41 YEATS is dimeric in HEK293T cells<!>Full-length GAS41 YEATS domain binds with increased affinity to di-acetylated histone H3<!>Dimerization of YEATS domain increases the affinity toward the di-acetylated histone H3<!>Di-acetylated H3 peptide pulls-down GAS41 from eukaryotic cells<!>Structural basis for the recognition of di-acetylated histone H3 by the GAS41 YEATS domain<!>CONCLUSIONS<!>Cloning, Expression and Protein Purification<!>NMR Titration<!>Isothermal Titration Calorimetry<!>Determination of binding affinity using bio-layer interferometry<!>Circular Dichroism Spectroscopy<!>X-Ray Crystallography<!>GAS41 pull-down assay<!>NanoBiT protein-protein interaction assay

<p>GAS41, also known as YEATS4, is a nuclear protein encoded by the GAS41 (glioma-amplified sequence 41) gene that was identified in a glioblastoma cell line1. It is a member of the YEATS family of proteins, and is implicated in chromatin remodeling and transcriptional regulation 2, 3. GAS41 interacts with chromatin-modifying complexes, including TIP60 and SRCAP 4, 5, and promotes deposition of histone H2A.Z 6. Functional studies implicate GAS41 in cell growth and survival 7, 8, and in maintenance of embryonic stem cell identity 9.</p><p>Acetylation of histone lysines is abundant in cells, and a high level of acetylated histones typically correlates with actively transcribed genes 10. Proteins involved in transcriptional regulation recognize acetylated lysines utilizing two major families of reader domains: bromodomains 11 and YEATS domains 12. There are four YEATS domain-containing proteins in humans: ENL, YEATS2, AF9, and GAS41. Both AF9 and ENL bind histone H3 peptides with acetylated K9, K18, and K27, with low- to mid-micromolar affinities 13, 14. Recently, GAS41 has been shown to interact with H3K14ac and H3K27ac peptides with affinities ranging from 9.3 to 32.7 μM 6, 9, and with H3K122 succinylated peptide at low pH (pH = 6.0) 15. Notably, YEATS domains have been observed to bind histone peptides with crotonylated lysine with higher affinities 16–19. This is particularly evident for YEATS2, which favors crotonylated lysine with a seven-fold stronger binding affinity when compared to acetylated lysine 16.</p><p>In this study, we report that the YEATS domain of GAS41 binds acetylated histone peptides with moderate, mid-micromolar affinities, and preferentially recognizes acetylated H3K18 and H3K27. GAS41 contains a C-terminal coiled-coil domain, and we demonstrate that full-length GAS41 is dimeric in HEK293T cells. We found that full-length GAS41 binds with an enhanced affinity towards di-acetylated over mono-acetylated histone H3, in vitro and in pull-down experiments. We determined the crystal structure of the YEATS domain in complex with H3K23acK27ac, illustrating a bivalent mechanism of di-acetylated histone recognition by GAS41. Our findings suggest a unique recognition mode of acetylated histone by GAS41 through higher-order interactions.</p><!><p>Previous studies revealed that the human YEATS domain-containing proteins AF9 and ENL bind histone H3 peptides with acetylated K9, K18, and K27, with low- to mid-micromolar affinities 13, 14. To determine the binding affinity and specificity of the GAS41 YEATS domain, we tested a series of peptides derived from histones H3 and H4 with mono-acetylated lysine residues encompassing major sites for lysine acetylation. The recombinant GAS41 YEATS domain yields a well-dispersed 1H-15N HSQC spectrum, which enabled us to perform NMR titration experiments and determine the binding affinity of selected histone peptides (Supplementary figure S1). We found that all acetylated peptides bind to the YEATS domain with relatively modest affinities, and the most potent binding was observed for H3K27ac (KD = 58 μM) and H3K18ac (KD = 106 μM) (Figures 1A, C). All remaining peptides bind to the GAS41 YEATS domain with at least ten-fold lower affinity (Figure 1C). Contradictory to the recent report, we have observed low affinity of GAS41 YEATS toward H3K14ac (KD = 720 μM) (Supplementary figure S2). No binding was detected for the non-acetylated H3K27, indicating that acetylated lysine is essential for the interaction with the GAS41 YEATS domain (Figure 1B). To assess the contribution of acetylated lysine, we tested the tripeptide AKacA, and found that it binds to the GAS41 YEATS domain with a KD of ~1mM, making it comparable to weakly binding acetylated H3 and H4 peptides (Figure 1C). Next, we performed a sequence alignment of all tested peptides, and found that two common features of the two peptides binding with the highest affinities, namely H3K18ac and H3K27ac, are the presence of an arginine residue directly preceding the acetylated lysine and an additional lysine located four residues upstream of the acetyl-lysine (Figure 1D). This result suggests that these positively charged residues are likely important for the recognition of acetylated histone H3 by GAS41.</p><p>Our data demonstrate that the binding affinities of GAS41 YEATS toward acetylated H3 peptides are relatively modest, comparable to ENL and YEATS2 14, 16, and in agreement with recently reported data for GAS41 6. Several recent reports revealed that YEATS domains recognize crotonylated lysine residues with higher affinity than acetylated lysines 16–19. Among these proteins, YEATS2 was shown to have the largest selectivity (approximately seven-fold) toward crotonylated over acetylated lysine 16. To assess whether GAS41 YEATS binds crotonylated lysine, we tested the binding of the H3K27cr peptide and found an approximately two-fold improvement in the binding affinity (KD = 34 μM) compared with H3K27ac (Figures 1A, C). According to the recent studies, the GAS41 YEATS domain binds H3 peptide succinylated on K112 at acidic pH (pH = 6.0) 15. We have tested binding of an H3 peptide with succinylated K27 (H3K27su) at pH = 7.5, and found only modest binding with KD > 1mM (Figures 1C, S3) suggesting that GAS41 is not recognizing K27su at physiological pH.</p><!><p>The sequence of GAS41 reveals the presence of a coiled-coil domain immediately following the YEATS domain (Figure 2A), and previous studies have shown that the GAS41 coiled-coil region is dimeric in solution 20. We expressed the C-terminal fragment of GAS41 encompassing residues 149–227, which includes the coiled-coil domain. Using circular dichroism (CD) spectra, we demonstrate helical secondary structure of the C-terminal coiled-coil in solution (Supplementary figure S4). Thermal denaturation indicates melting transition with Tm = 29°C (Figure 2B), which is consistent with the previously reported stability of GAS41 coiled-coil derived peptides 20.</p><p>To test whether full-length GAS41 forms a dimer in cells, we employed the NanoBit protein-protein interaction assay 21. We expressed GAS41 as fusions with LgBit and SmBit proteins in HEK293T cells, and observed a strong luminescence signal, indicating that full-length GAS41 is indeed dimeric in eukaryotic cells (Figure 2C). To the contrary, we have not observed dimerization of the YEATS domain alone (Figure 2C). Indeed, the dimerization of GAS41 has been suggested in previous studies, and L211G and L218G mutations disrupting the coiled-coil motif impaired GAS41-mediated activation of p53 tumor suppressor pathway 22. We tested GAS41 L211G, L218G double-mutant in NanoBit assay, and found that these mutations disrupted GAS41 dimerization (Figure 2C).</p><!><p>We hypothesized that full-length, dimeric GAS41 may simultaneously recognize di-acetylated histone H3 through a bivalent interaction mode, which could result in enhanced binding affinity. To characterize the interaction of full-length GAS41 with acetylated peptides, we employed bio-layer interferometry using selectively biotinylated GAS41. First, we tested mono acetylated H3K18ac and H3K27ac peptides, and found mid-micromolar KD values (36 μM and 81 μM, respectively, Figures 3A,B) that are consistent with the NMR results for isolated YEATS domain (Figure 1C). Subsequently, we tested the binding of di-acetylated peptides and found low-micromolar affinities for H3K18acK27ac (KD = 3.2 μM) and H3K23acK27ac (KD = 5.0 μM) (Figures 3C, D). This represents an over ten-fold enhancement in the binding affinity of the full-length GAS41 toward di-acetylated H3 peptides.</p><!><p>To test whether dimerization of the YEATS domain increases the affinity towards di-acetylated H3 peptides in a model system, we purified a dimeric GST-YEATS construct and performed isothermal titration calorimetry (ITC) experiments (Supplementary Table 1). We found that GST-YEATS binds mono-acetylated H3K27ac peptide with 1:1 stoichiometry, and with a similar binding affinity as an isolated YEATS domain (KD = 39 μM, Figures 3E, F). We also confirmed that crotonylated lysine slightly increases the binding affinity of H3K27cr toward GST-YEATS (KD = 23.4 μM, Figure 3E). Subsequently, we tested the di-acetylated H3K23acK27ac peptide, and found that it binds to GST-YEATS with KD = 13.6 μM and 1:2 stoichiometry (single molecule of H3K23acK27ac binds two molecules of YEATS domain, Figure 3G). A very similar binding affinity and stoichiometry was observed for H3K18acK27ac (KD = 12.5 μM and 1:2 stoichiometry, Figure 3G). Therefore, dimerization of the YEATS domain in a model system using GST-fusion leads to two- to three-fold stronger binding of di-acetylated peptides than mono-acetylated H3K27ac. The stoichiometry clearly indicates that the dimerized YEATS domain recognizes di-acetylated H3, and enhanced affinity over mono-acetylated H3 is most likely resulting from a bivalent binding mode (Figure 3H). We also found a very similar effect for a double crotonylated H3K23crK27cr peptide (Figure 3G), which binds to GST-YEATS with KD = 11.1 μM and 1:2 stoichiometry, consistent with an approximately two- to three-fold enhanced affinity compared with that of H3K27cr.</p><!><p>We also assessed whether di-acetylated H3K23acK27ac peptide can interact with full-length GAS41 expressed in eukaryotic cells. Transfection of HEK293T cells with the construct encoding wild-type GAS41 resulted in robust overexpression of the full-length protein (Figure 4). We then used these cells to perform the pull-down, using biotinylated, di-acetylated H3K23acK27ac and mono-acetylated H3K27ac peptides. While both peptides pulled-down GAS41, the interaction with H3K23acK27ac is considerably stronger when compared to the H3K27ac peptide (Figure 4). To confirm that this effect is specific, we introduced W93A mutation that is known to impair YEATS domain function 6, 9 and found that it leads to the disruption of the binding (Figure 4). Altogether, pull-down experiments validate enhanced affinity of full-length GAS41 towards di-acetylated H3, and are consistent with KD values determined in biochemical experiments.</p><!><p>To determine how the GAS41 YEATS domain recognizes acetylated histone peptides, we used X-ray crystallography. First, we determined the structure of the YEATS domain at 2.1 Å resolution (Supplementary figure S5 and Table 1). The GAS41 YEATS domain adopts an immunoglobulin fold with a two-layer β-sandwich consisting of eight antiparallel β strands (Supplementary figure S5), and is similar to the previously reported structures of the YEATS domains from AF9, ENL, and YEATS2 (Supplementary figure S6A). We found a short α-helical region encompassing residues 123–129 as a new structural element not observed in other YEATS domains (Supplementary figure S5).</p><p>Analysis of the packing in the crystal structure revealed four YEATS domain molecules in the asymmetric unit, with two molecules showing accessible sites for the binding of acetylated peptides. To determine the crystal structure of a complex with histone peptide, we performed soaking of the YEATS domain crystals with di-acetylated peptides and found the best diffraction (2.4 Å resolution) for H3K23acK27ac complex. Interestingly, we found that di-acetylated H3K23acK27ac binds simultaneously to the two YEATS domains in the crystal structure, and that acetylated K23 and K27 residues occupy the corresponding pockets in two different molecules of the YEATS domain (Figure 5A). Importantly, this binding mode is consistent with a 1:2 binding stoichiometry from the ITC experiment (Figure 3G). Analysis of the crystal structure reveals that only very short segments of the H3 peptide are in contact with the YEATS domains, and the major contacts involve recognition of the acetylated K23 and K27 side chains (Figure 5B). Acetylated lysine fits into the channel comprised of the aromatic side chains of H71, Y74, W93, and F96. The acetyl group is recognized by a network of hydrogen bonds, including interactions with the backbone amides of W93 and G94, as well as the side chain of S73 (Figure 5B). The binding mode of the H3 fragment surrounding K27ac is very similar to the recently reported structure of the GAS41 YEATS domain bound to H3K27ac peptide (Supplementary Figure S7). Both structures show the presence of a hydrophobic interaction between H3P30 and GAS41 F121, which may explain enhanced specificity towards H3K27ac peptide (Supplementary figure S7). Notably, the structure of the YEATS domain with bound H3K23acK27ac reveals reversed direction of polypeptide around H3K23ac, suggesting some plasticity in recognition of acetylated lysines. Overall, the crystal structure of YEATS with bound H3K23acK27ac represents a snapshot of the bivalent recognition mode of di-acetylated histone H3 by full-length dimeric GAS41.</p><!><p>GAS41 belongs to a four-member family of human proteins, characterized by the presence of the conserved YEATS domain. GAS41 is the shortest protein in this family, and is composed of two domains, namely, the N-terminal YEATS and C-terminal coiled-coil (Figure 2A). In this study, we investigated the binding of the GAS41 YEATS domain to a series of histone H3- and H4-derived peptides with acetylated lysine residues, and found that it preferentially recognizes H3K18ac and H3K27ac with modest affinities (KD values ranging from 35 to 106 μM). The presence of the C-terminal coiled-coil suggests GAS41 dimerization 20, 22. Indeed, we have validated that full-length GAS41 is dimeric in HEK293T cells. Disorder prediction for GAS41 sequence indicates that the linker between the YEATS and the coiled-coil domain is relatively short and consists of approximately 20 amino acids (Figure 2A). This implicates the close proximity of the two YEATS domains in dimeric GAS41. In this study, we found that full-length GAS41 preferentially binds to di-acetylated H3 peptides with ~ten-fold enhanced affinity when compared with mono-acetylated H3 (Figures 3A, C). Importantly, we observed a more efficient pull-down of GAS41 using di-acetylated H3K23acK27ac when compared to mono-acetylated H3K27ac (Figure 4), further validating GAS41 as a reader of di-acetylated histone H3. Our results suggest a model in which full-length GAS41 recognizes di-acetylated histone in a unique, bivalent mode (Figure 5C). We further validated this model by testing the binding of dimeric GST-YEATS domain and found enhanced affinity towards di-acylated (including both di-acetylated and di-crotonylated) versus mono-acylated H3 peptides (Figures 3E, G).</p><p>Histone acetylation is an epigenetic modification important for regulation of gene expression. Recent ChIP-seq experiments revealed co-localization of GAS41 to gene promoters enriched in H3K14ac and H3K27ac 6. Acetylated lysines of H3K14, H3K18 and H3K27 are enriched at promoters of actively transcribed genes 6, 23, and enhanced affinity of GAS41 towards di-acetylated H3 may explain over 90% of GAS41 bound to highly-acetylated gene promoter regions 6. We also observed approximately two-fold enhanced binding of GAS41 YEATS to peptides with crotonylated versus acetylated lysine. Further studies are required to establish whether GAS41 is a di-acetyl-lysine or di-crotonyl-lysine reader in physiological conditions.</p><p>YEATS domain proteins are recently characterized readers of acylated histones, recognizing both acetylated and crotonylated lysines 13, 24. Among human homologs, only AF9 has been reported to bind acetylated histone proteins with relatively high affinity (KD = 2.1 to 3.7 μM) 13, 24, while the KD values for ENL and YEATS2 interactions are weaker (KDs above 30 μM) 14, 16, and are comparable to the affinity of the GAS41 YEATS domain for mono-acetylated H3 6. We found that the affinity of GAS41 toward di-acetylated H3 is significantly improved through a higher-order interaction involving dimerization of the YEATS domain. Whereas a bivalent binding mode has not been proposed for other members of the YEATS domain family, sequence analysis predicts the presence of the coiled-coil domain in YEATS2 (Supplementary figure S6C). This suggests that higher-order interactions might be present in other YEATS protein readers.</p><!><p>The synthetic gene encoding Human GAS41 YEATS (residues 13–158) was ordered from Life Technologies and subcloned using the BamHI and HindIII restriction sites into pQE-80L expression vector (Qiagen) with an N-terminal hexahistidine (His6) tag. The recombinant plasmid pQE80L-GAS41 (residues 13–158) was transformed into E. coli strain BL21(DE3). Transformants were grown in 15N- labeled M9 minimal medium containing ampicillin at 37 °C until reaching an OD600 between 0.6 and 0.8. After induction with 0.25 mM isopropyl 1-thio-β-D-galactopyranoside (IPTG), cultures were grown for an additional 16 h at 18 °C. Harvested cells were resuspended in lysis buffer (50 mM Tris, pH 7.5, 300 mM NaCl and 1 mM TCEP) and lysed using a cell disrupter. Clarified lysate was applied to Ni–NTA (Qiagen) affinity column. The column was extensively washed with lysis buffer containing 35 mM imidazole and eluted with lysis buffer containing 200 mM imidazole. The eluted pure fractions were pooled and dialyzed against 50 mM Tris, pH 7.5, and 200 mM NaCl buffer and then concentrated to ~60 μM. The final purified 15N-labeled GAS41(13–158) was used in NMR studies.</p><p>Codon-optimized cDNAs of full-length human GAS41 was synthesized by Life Technologies and amplified GAS41 YEATS (residues 1–148) by the polymerase chain reaction (PCR). The PCR product was subcloned using the BamHI and EcoRI restriction sites into pGST-parallel vector 25 with an N-terminal GST tag followed by a TEV cleavage site. The resulting plasmid pGST-GAS41(1–148) was transformed into E. coli strain BL21(DE3). Transformed cells were grown in Luria broth medium with ampicillin selection. After 18 h induction with 0.2 mM IPTG at 18°C, cells were harvested by centrifugation and resuspended in lysis buffer containing 50 mM Tris, pH 7.5, 300 mM NaCl and 1 mM TCEP and lysed using a cell disruptor. The soluble fraction of the cell lysate was then loaded onto a glutathione-Sepharose 4B (GE Healthcare) affinity column. The column was thoroughly washed with buffer containing 50 mM Tris, pH 8.0 and 500 mM NaCl and eluted with 10 mM reduced glutathione. GST-tagged GAS41(1–148) protein was used in ITC experiment. The eluted proteins were proteolytically cleaved with TEV protease, followed by S-75 size exclusion chromatography purification into buffer containing 20 mM Tris, pH 7.5, 300 mM NaCl. The final purified GAS41(1–148) was used in crystallization trial.</p><p>GAS41 coiled coil region (residues 149–227) was cloned from the synthetic human GAS41 gene by PCR. The PCR product was subcloned using the BamHI and HindIII restriction sites into pMocr vector with an N-terminal hexahistidine (His6) tag followed by a TEV cleavage site. GAS41 149–227 protein was expressed as inclusion bodies and solubilized in buffer with 6 M Guanidine HCl. Re-folding was performed using dialysis to 50 mM Tris, pH 7.5, 150 mM NaCl, 2 mM DTT buffer following the cleavage with TEV protease. N-terminal Mocr-His6 was extracted by re-application to Ni-NTA resin. Protein was dialyzed extensively against Storage Buffer (50 mM tris pH 7.5, 150 mM NaCl, 1 mM TCEP) and concentrated for storage at –80 °C.</p><!><p>All NMR spectra were collected at 30°C on a 600 MHz Bruker Advance III spectrometer equipped with cryogenic probe, running Topspin version 2.1. Binding of histone peptides to GAS41 YEATS (residues 13–158) were characterized by measuring chemical shift perturbations of selected amide resonances on the 1H-15N HSQC spectra of 60 μM 15N-labeled GAS4113–158 titrated with peptides at molar ratios of 1.0, 2.7, 4.3 and 6.0. Dissociation constants were determined from least-squares fitting of chemical shift perturbations as a function of ligand concentration δi={b−√(b2–4*a*c)}2a with a = (KA/δb) × [Pt], b = 1 + KA([Lti] + [Pt]), and c = δb * KA * [Lti], where δi is the absolute change in chemical shift for each titration point, [Lti] is the total ligand concentration at each titration point, [Pt] is the total protein concentration, KA = 1/KD is the binding constant, and δb is the chemical shift of the resonance in question in the complex. KD and δb were used as fitting parameters in analysis 26.</p><!><p>The measurements were performed using a VP-ITC titration calorimetric system (MicroCal) at 25°C. GST-tagged GAS41 YEATS (residues 1–148) was dialyzed extensively against ITC buffer consisting of 50 mM phosphate, pH 7.5, 150 mM NaCl and 2 mM β-mercaptoethanol. Histone derived peptides were directly dissolved in the ITC buffer at 500 μM concentration. The titration curve was obtained by injecting 10 μL aliquots of histone derived peptides into the cell containing 50 μM GST-GAS41(1–148), at a time interval of 200 s. All samples were degassed by vacuum aspiration for 20 min prior to measurements. ITC titration data were analyzed with a single-site fitting model using Origin 7.0 software.</p><!><p>To determine the affinity of full-length GAS41 towards H3 peptides we employed Bio-Layer Interferometry experiments using an Octet Red 96 instrument (ForteBio, Inc.). Mono-biotinylated GAS41 was obtained by co-expression of Avi-tagged GAS41 with BirA enzyme in E. coli BL21(DE3) cells as previously described 27. The purification procedure for biotinylated GAS41 was identical to that for the His tagged GAS41 YEATS (residues 13–158) protein. Prior to protein immobilization, streptavidin biosensor tips were incubated in 50 mM Tris, pH 7.5, and 200 mM NaCl buffer for 600 s. Subsequently, protein was loaded onto tips for 600 s, followed by 1200 s equilibration step. Peptide binding experiments were performed in following order: 600 s equilibration, 300 s association, and 600 s dissociation. Experimental set-up was performed using Octet Data Acquisition Software, and data were analyzed by Octet Data Analysis Software (Pall ForteBio, LLC.). Signal was plotted as a function of ligand concentration to determine KD, using Prism software (GraphPad Software, Inc.). To correct for drift during association, the slope during the drift period was fit by linear regression, and the product of slope and time (in seconds) was subtracted from the signal.</p><!><p>Spectra and thermal denaturation experiments were measured at 50 μM concentration of GAS41 (149–227) in 20 mM sodium phosphate, pH 6.5, and 250 mM sodium fluoride using a Jasco CD-810 spectropolarimeter with constant N2 flushing. Rectangular cells with 1 mm path length were used, and a circular water bath was used to control temperature of the optic cell chamber. Protein spectra were averaged from three wavelength-scans collected at 0.1 nm-intervals from 178 – 260 nm. Baseline spectrum of buffer was recorded prior to recording protein spectra. Mean residue ellipticity was determined by the equation Θ=(Θobs×mrw)/(10×l×c) where Θobs is observed ellipticity (in millidegrees), mrw is mean residue molecular weight, l is optical path length of the cell (in cm), and c is peptide concentration. For thermal denaturation experiments, we recorded CD spectra over temperature gradient from 5 – 95°C at 1 degree increments.</p><!><p>Protein crystals of GAS41 YEATS (residues 1–148) were obtained by sitting drop method using 6–8.5 mg/ml protein mixed in a 1:1 volume ratio with precipitation solution containing 100 mM CHES, pH 9.4 and 1.24 –1.36 M ammonium sulfate. The crystals were cryoprotected in reservoir solution supplemented with 25% ethylene glycol. X-ray diffraction data were collected on LS-CAT beam line 21-ID-F at the Advanced Photon Source. The data were then indexed, integrated, and scaled using the HKL2000 suite 28. The structure was determined by molecular replacement method with the CCP4 version of MOLREP 29 using the structure of yeast Taf14 YEATS domain (PDB code: 3QRL) as a search model. To obtain a histone peptide bound complex structure, native GAS41 crystals were soaked in a 1:2 mixture of protein buffer and reservoir solution containing saturating amounts of peptide. Model building and structure refinement were carried out using WinCOOT 30 and Phenix.refine 31. The data statistics are summarized in Table 1.</p><!><p>Full-length GAS41 and GAS41 W93A mutant was cloned into pCMV vector and used to transfect HEK293T cells using FuGENE6 (Promega). Cell lysates were prepared in lysis buffer (PBS with 1 % Triton X-100 and protease inhibitor cocktail) and sonicated. Biotinylated H3K23acK27ac and H3K27ac peptides were incubated with streptavidin-magnetic beads for 6 h at 4 °C. Cell lysates were precleared with streptavidin-magnetic beads for 2 h at 4 °C and incubated with the biotinylated peptides immobilized on streptavidin-magnetic beads overnight at 4 °C. The beads were washed 10 times with wash buffer (PBS with 0.2 % Triton X-100). Pulled-down proteins were analyzed using western blot with GAS41 antibody (sc-393708 from Santa Cruz).</p><!><p>The NanoBiT protein-protein interaction assays (Promega) were conducted according to the manufacturer's instructions. Briefly, full-length GAS41, YEATS (residues 1–158), and GAS41 L211G/L218G mutant were cloned into pBiT1.1-C[TK/LgBiT] and pBiT2.1-C[TK/SmBiT] vectors for NanoBiT system. All plasmids were verified by sequencing. HEK293T cells were plated into 96-well white plates at 1×104 cells per well. Each Lg-BiT and Sm-BiT plasmids were co-transfected using FuGENE HD (Promega) the next day and incubated for 48 h. After we added the Nano-Glo Live Cell Reagent to each well, the luminescence was measured on 0 min, 10 min, 30 min, and 60 min using the PHERAstar FS (BMG Labtech). The positive control (Lg-PRKAR2A and Sm-PRKACA) and the negative control vectors were provided by the manufacture.</p>

PubMed Author Manuscript