Datasets:

griffin

/

ChemSum

like 11

Pharmacophore Modelling and Synthesis of Quinoline-3-Carbohydrazide as Antioxidants

From well-known antioxidants agents, we developed a first pharmacophore model containing four common chemical features: one aromatic ring and three hydrogen bond acceptors. This model served as a template in virtual screening of Maybridge and NCI databases that resulted in selection of sixteen compounds. The selected compounds showed a good antioxidant activity measured by three chemical tests: DPPH radical, OH° radical, and superoxide radical scavenging. New synthetic compounds with a good correlation with the model were prepared, and some of them presented a good antioxidant activity.

pharmacophore_modelling_and_synthesis_of_quinoline-3-carbohydrazide_as_antioxidants

1. Introduction<!>2.1. Training Set<!>2.2. Pharmacophore Model Generation<!>2.3. Antioxidant Activities of Identified Compounds<!>2.3.1. DPPH Radical Scavenging<!>2.3.2. OH° Radical Scavenging<!>2.3.3. Superoxide Radical Scavenging<!>3. Synthesis<!>Conclusion 3 .<!>Assay of Hydroxyl Radical (OH°) Scavenging Activity<!>DPPH Radical Scavenging Activity<!>O2 − Radical Scavenging Activity<!>4.2. Computational Methods<!>4.3.1. General Methods<!>4.3.2. General Method of Preparation of Compounds 6a-b<!>4.3.3. Ethyl 4-hydroxy-2-oxo-1,2-dihydroquinoline-3-carboxylate (6a)<!>4.3.4. Ethyl 4-hydroxy-1-methyl-2-oxo-1,2-dihydroquinoline-3-carboxylate (6b)<!>4.3.5. General Method for Compounds 7a-b<!>4.3.6. Hydroxy-2-oxo-1,2-dihydroquinoline-3-carbohydrazide (7a)<!>4.3.7. 4-Hydroxy-1-methyl-2-oxo-1,2-dihydroquinoline-3-carbohydrazide (7b)<!>4.3.8. General Method for Compounds (8a–h)<!>4.3.9. N′-[(E)-(2,4-Dihydroxyphenyl)methylidene]-4-hydroxy-2-oxo-1,2-dihydroquinoline-3-carbohydrazide (8a)<!>4.3.10. 4-Hydroxy-N′-[(1E)-(2-hydroxy-5-methylphenyl)methylene]-2-oxo-1,2-dihydroquinoline-3-carbohydrazide (8b)<!>4.3.11. 4-Hydroxy-N′-[(1E)-(2-hydroxy-5-methoxyphenyl)methylene]-2-oxo-1,2-dihydroquinoline-3-carbohydrazide (8c)<!>4.3.12. N′-[(1E)-(2,6-Dichloro-3-nitrophenyl)methylene]-4- hydroxy-2-oxo-1,2-dihydroquinoline-3-carbohydrazide (8d)<!>4.3.13. N′-[(1E)-(2,6-Dichloro-3-nitrophenyl)methylene]-4-hydroxy-1-methyl-2-oxo-1,2-dihydroquinoline-3-carbohydrazide (8e)<!>4.3.14. 4-Hydroxy-N′-[(1E)-(2-hydroxy-5-methylphenyl)methylene]-1-methyl-2-oxo-1,2-dihydroquinoline-3-carbohydrazide (8f)<!>4.3.15. 4-Hydroxy-N′-[(1E)-(2-hydroxy-5-methoxyphenyl)methylene]-1-methyl-2-oxo-1,2-dihydroquinoline-3-carbohydrazide (8g)<!>4.3.16. 4-Hydroxy-1-methyl-N′-[(1E)-(2-nitrophenyl)methylene]-2-oxo-1,2-dihydroquinoline-3-carbohydrazide (8h)

<p>Free radicals play an important role in the pathogenesis of many diseases, accounting for continuing interest in the identification and development of novel antioxidants that prevent radical-induced damage.</p><p>In humans, several pathologies involve the overproduction of reactive oxygen species (ROS): these oxygen species such as the superoxide radical anion (O2 −°) and hydrogen peroxide (H2O2) are formed by the partial reduction of molecular oxygen. Formation of the hydroxyl radical (HO°), another ROS, is thought to occur through the one-electron reduction of H2O2. This reaction is facilitated by transition metals that are in a reduced valence state (e.g., reduced copper or iron) [1]. Additionally, there are a large number of other reactive species that are formed from the reaction ROS with biological molecules (e.g., polyunsaturated lipids, thiols, and nitric oxide (NO)) [2]. For example, O2 −° reacts with NO to form peroxynitrite anion (ONOO−), which is unstable at physiological pH and rapidly decomposes. It forms potent nitrating and oxidizing species [3, 4] or hypochlorite (XOCl) that is a powerful oxidant produced by activated neutrophils via the reaction of H2O2 and Cl−, catalysed by the heme enzyme myeloperoxidase [5].</p><p>A lot of natural and synthetic products like quercetin 1, curcumine 2, resveratrol 3, Trolox 4, and N-acetylcystein 5 are known for their antioxidant activity [6–10]. Some heterocyclic compounds, either natural (phytoestrogens) or obtained by synthesis, having coumarin or quinoline rings, were studied for their biological activity. They are used especially as radicals scavenger like quercetol and coumestrol [11, 12] or the copper or iron chelating molecules such as clioquinol [13, 14].</p><p>After first studies realized in our laboratory [15–17] on new compounds with quinoline and coumarin structures and with the aim of discovering a very strong antioxidant, we decided to introduce in our research the three-dimentional generation and database searching. The increasing number of successful applications of 3D-pharmacophore-based searching in medicinal chemistry clearly demonstrates its utility in the modern drug discovery paradigm [18, 19]. In the absence of such three-dimensional structure-based, we attempted to identify the hypothetical 3D-ligand-based pharmacophore model by using the common features hypothesis generation approach (HipHop) implemented in the program Catalyst [20]. In particular, HipHop algorithm finds common feature pharmacophore models among a set of highly active compounds and carry out a qualitative model (without taking care of the activity data) which represents the essential 3D arrangement of functional groups common to the set of molecules that explains the specific activity, antioxidant in the current study.</p><p>The generation of a pharmacophore model for antioxidant from a training set of five molecules using catalyst/HipHop gave ten hypothesis, the best one was used for the databases search. The identified compounds were tested and discussed to validate a pharmacophore hypothesis.</p><p>Then, this pharmacophore was used to predict and select the synthesis of new quinoline derivatives, many compounds were prepared and their antioxidant properties evaluated by hydroxyl radical °OH scavenging activity and by their antiradical activity against 2,2-diphenyl-1-picrylhydrazyl radical (DPPH•) and anion superoxide.</p><!><p>Five molecules Quercetin 1, curcumine 2, resveratrol 3, Trolox 4, and N-acetyl cystein 5 as shown in the Scheme 1 were selected for the training set representing the best known natural antioxidants [5–10]. All structures were generated using editor sketcher in DS Catalyst software package and to build conformational models of up to 250 conformers for each molecule, the "best conformer generation" option and 10 kcal/mol energy cutoff were chosen.</p><!><p>Our Pharmacophoric analysis was carried out using the Catalyst/HipHop procedure to evaluate the common feature required and the hypothetical geometries of these ligands in their most active forms.</p><p>In the hypothesis generation based on the atom types in the molecules of the training set, the following chemical functions were selected in the feature dictionary of Catalyst: Hydrogen bond acceptor, hydrogen bond donor, aromatic ring, positive ionisable and hydrophobic groups.</p><p>Ten hypothesis (Hypo 1 to Hypo 10) were obtained using the default parameters of catalyst. These hypothesis had scores from 33.52 to 36.68 (Table 1) so we studied if they mapped to all the important features of the active compound, we searched the correlation between best values, conformational energies, and activity of the training set (data not shown) and we selected the highest ranked pharmacophore hypothesis (Hypo1) for the database search.</p><p>This selected pharmacophore model contains four chemical features: one aromatic ring (RA) (orange colour) and three hydrogen bond acceptors (HBA2, HBA3 and HBA4) (green colour). The RA maps the aromatic ring attached to position 2 of benzopyrane group of quercetin, the HBA2 maps the hydroxyl group at position 4 of aromatic ring, HBA3 and HBA4 maps respectively the hydroxyl groups at position 7 and 5 as shown in Figure 1.This alignment represents a good match of features of the pharmacophore model with the ligand (fit value = 3.99/4).</p><p>We employed this model as 3D-search query against the NCI, Maybridge, and minimaybridge structure databases (each contained thousands of compounds) using the "fast flexible search" approach implemented within Catalyst. The pharmacophore captured 300 hits for each database, we selected sixteen compounds Scheme 2 on the basis of fit value Log P and availability.</p><!><p>Free radical scavenging is one of the best known mechanisms by which antioxidants inhibit lipid oxidation. DPPH, Superoxide, and hydroxyl radical scavenging activity evaluation are standard assays in antioxidant activity studies and offer rapid techniques for screening the radical scavenging activity (RSA) of specific compounds. The stable free radical 2;2-diphenyl-1-picrylhydrazyl (DPPH) is a useful reagent to investigate the scavenger properties of polyphenols. It is now widely accepted that the reaction between phenols and DPPH proceeds through two different mechanisms: The direct hydrogen atom transfer (HAT) and the sequential proton less electron transfer. Superoxide anion produced by activated human neutrophils can be a source of additional harmful ROS and no radical species as the singulet oygen in vivo. The hydroxyl radical is considered the most damaging free radical for living cells because it leads to deleterious oxidations of cellular components including protein, DNA and lipids. The RSA of 16 identified compounds was estimated using these three methods.</p><!><p>A freshly prepared DPPH solution exhibits a deep purple colour with a maximum absorption at 517 nm. This purple colour generally disappears when an antioxidant is present in the medium as shown in Scheme 3. Thus, antioxidant molecules can quench DPPH free radicals (by providing hydrogen atoms or by electron donation, conceivably via a free-radical attack on the DPPH molecule) and convert them to colourless/bleached product [21, 22].</p><p>The RSA against DPPH radical of 16 identified molecules were examined and compared (Table 2). Results are expressed as a percentage of the ratio of the decrease in absorbance at 517 nm, to the absorbance of DPPH solutions in the absence of compounds at 517 nm.</p><!><p>We used the benzoic acid method [23]. The benzoic acid was hydroxylated by OH° formed by Fenton reaction at C3 or C4 positions of the aromatic ring and the fluorescence was measured at 407 nm emission with excitation at 305 nm. This fluorescence generally decreases when an antioxidant is present in the medium. Antioxidant molecules prevent the hydroxylation of benzoic acid by providing hydrogen atom.</p><p>The RSA OH° result of molecules identified were summarized in (Table 2), this results are expressedas (1)  RSA  OH°%=[Absorbance  in  the  presence  of  sample][Absorbance  in  the  absence  of  sample] ×100.</p><!><p>Superoxide radical scavenging activity was determined by absorbance measurement of the blackish blue formazan product by superoxide addition to nitro blue tetrazolium (NBT) substrate, according to the method of Nishikimi et al. [24]. Superoxide was generated chemically by the reduction of phenazine methosulfate (PMS), using β-NADH as the electron donor in the presence of dissolved molecular oxygen in the reaction solution.</p><p>The RSA O2− results in molecules identified were summarized in Table 2. The percentage scavenging effects were calculated from the decrease in absorbance against control. This absorbance was measured at 560 nm.</p><p>From analysis of Table 2, we can conclude that all the identified compounds present a scavenging effect. For The results RSA of DPPH radicals, seven compounds (AW 00493, BTB 14348, HTS 0630, NSC 2541, NSC 3028, NSC 412, and NSC 740) have the same or better activity than the standard N-acetylcysteine at 50 μmol·L−1.</p><p>Based on the result of superoxide radical scavenging we demonstrated that all compounds show a dose-dependent effect.</p><p>Concerning the RSA of radical hydroxyl, all compounds have same or better results than standards at 50, 100, and 150 μmol·L −1</p><p>These results show that the theoretical pharmacophore has got a discriminant power. It allows the selection of antioxidants molecules from a databasis containing thousand of compounds.</p><!><p>We envisaged doing the synthesis of compounds presenting a good correlation with the pharmacophore established to verify if it allows predicting the activity of molecules. We chose the quinoline derivatives to continue the work already realized on the synthesis of new quinoline derivatives by our laboratory [15–17].</p><p>Different molecules were proposed and first mapped on selected pharmacophore using ligand pharmacophore protocols. For these compounds, we obtained fit values from 2.6 to 3.3</p><p>In Figure 2 we present the compound 8c (fit value = 3.3/4) mapping with a previously selected pharmacophore, we can see that the RA maps the aromatic ring of phenolic group, the HBA2 maps the hydroxyl group at position 5 of phenolic group, HBA3 and HBA4 map respectively the carbonyl groups at position 2 of quinoline and carbonyl of carbohydrazide.</p><p>The synthetic route to prepare desired substituted 4-hydroxy-2-oxophenylmethylene-1,2-dihydroquinolin-3-carbohydrazide is described in Scheme 4.</p><p>The condensation N-H or N-methyl anhydride isatoique with ethylmalonate in dimethylformamide gave ethyl 4-hydroxy-2-oxo-1,2-dihydroquinoline-3-carboxylate 6a or his derivatives N-methyl 6b [25]. The 6a and 6d were converted in resulting 7a and 7b with hydrazine hydrate in methanol, finally they reacted with different aldehydes. This procedure gave compounds 8a–c, as a mixtures of E-and Z-isomers in a E  :  Z : 9 : 1, 7 : 3 and 4 : 1 ratio, respectively, whereas the target molecules 8d–h were isolated as pure E isomers. The E configuration compounds 8 was characterized in the 2D-NMR (1H-1H) spectra by NOESY experiments, and analyzing by the NOE effects on the hydrogens for the NH amide of carbohydrazide moiety, and the CH of imines.</p><p>All the compounds summarized in (Table 3) were obtained in moderate to good yields ranging from 56% to 94%. All these products were isolated from reaction mixture by recrystallisation from ethanol, and their structures were characterized by 1H NMR, IR spectra and elementary analysis.</p><p>The antioxidant activity for these compounds was measured by two methods DPPH and anion superoxide (Table 4), the hydroxyl radical scavenging is not applicable for these compounds because of their fluorescence at the studied wavelength (407 nm emission with excitation at 305 nm).</p><p>All synthesized compounds exhibit antiradical activity against DPPH radical and anion superoxide tests. The products 8a and 8c having a good fit value present better results. The product 8h has a lower result. It's probably due to absence of hydroxyl group</p><!><p>The present study is a successful example for a rational identification of antioxidants agents. This was accomplished by generating a three-dimensional pharmacophore model based on a training set of five well-know antioxidants. The model containing one aromatic group and three hydrogen bond acceptors was selected and used to identify new quinoline derivatives antioxidant agents.</p><!><p>In a screw-capped test tube, 0.2 mL of sodium benzoate (10 mmol), 0.2 mL of FeSO4·7H20 (10 mmol) and EDTA (10 mmol) were added. Then the sample solution and a phosphate buffer (pH 7.4, 0.1 mol) were mixed to give a total volume of 1.8. Finally, 0.2 mL of H2O2 solution (10 mmol) was added, and the whole incubated at 37°C for 2 h. After incubation, the fluorescence was measured on spectrofluorimeter Shimadzu RF 10AXL at wavelengths 407 nm for emission and 305 nm for excitation.</p><!><p>The capacity of compounds to scavenge the "stable" free radical DPPH was monitored according to the method of Hatano et al. [26]. Various concentrations of methanolic compounds solutions (0.3 mL) were mixed with methanolic solution containing DPPH radicals (1.5·10−4 M, 2.7 mL). The mixture was shaken vigorously and left to stand for 2 h in the dark (until stable absorption values were obtained). The reduction of the DPPH radical was determined by measuring the absorbance at 517 nm. The RSA was calculated as a percentage of DPPH colouration using (2)%  RSA  =[(ADPPH−AS)ADPPH]×100, where A S is the absorbance of the solution when the compound has been added at a particular level and A DPPH is the absorbance of the DPPH solution. Mean values from three independent samples were calculated for each compound and standard deviations were less than 5%.</p><!><p>The reaction mixture (1 mL) contained 700 μL of various concentrations of methanolic compounds solutions, 100 μL of β-NADH (1 mM in water), 100 μL of NBT (1 mM in 1 M-phosphate buffer, pH 7.8 and 100 μL of PMS (120 μM in water) added in that order and the mixture allowed to react at RT for 10 min. The control contained all the reaction reagents except the test material. The reaction was terminated by adding 40 μL of concentrated HCl (10 mM) and absorbance was measured at 560 nm against blanks that contained all compound except test material and PMS.</p><p>The percentage scavenging effects was calculated from the decrease in absorbance against control.</p><!><p>All molecular modelling studies were performed using discovery studio 2.1 with catalyst module. All structures were generated using 2D/3D editor sketcher and minimized to the closest minimum using the CHARMm-like force field implemented in the program [27]. A stochastic research coupled to a poling method [28] was applied to generate conformers for each compound by using "Best conformer generation" option with a 20 kcal/mol energy cutoff (20 kcal/mol maximum compared to the most stable conformer).</p><p>The pharmacophore-based investigation involved using the catalyst/Hip/Hop program to generate feature based 3D pharmacophore alignments [29]. This was performed in a three step procedures: (a) a conformation model of each molecule in the training set was generated, (b) each conformer was examined for the presence of certain chemical features, (c) a three dimensional configuration of chemical feature these steps were performed with a module common feature pharmacophore generation.</p><!><p>Reactions were monitored by TLC using precoated silica gel aluminum plates containing a fluorescent indicator (Macherey-Nagel). Detection was done with UV (254 nm). Melting points were determined on a Kofler block and were uncorrected. Infrared spectra were recorded on a Shimadzu FTIR-8201 PC spectrometer in KBr (ν in cm−1). 1H NMR spectra were recorded on a Bruker AC 300 spectrometer. Microanalyses were carried out by the Service Central d'Analyses, CNRS, Vernaison (France). All reagents were pure analytical grades and used without further purification.</p><!><p>The corresponding anhydride isatoic (1 eq) was suspended in DMF (10 mL) at 0°C. Sodium hydride (2 eq) and diethyl malonate (5 eq) were added slowly. The reaction mixture was heated at 85°C for 5 h. Then, 10 mL of water were added and the mixture was acidified with concentrated hydrochloric acid. The resulting solid was filtered, washed with water and dried, yielding the desired compound.</p><!><p>Following the General procedure in Section 4.3.2, the reaction of isatoic anhydride (1 g, 6.13 mmol) with NaH (0.29 g, 12.3 mmol), and diethyl malonate (4.91 g, 30.7 mmol) gave product 6a (1 g, 70%): mp 134°C; IR (KBr) υ 3406, 3193, 1658, 1604 cm−1; 1H NMR (DMSO-d 6, 300 MHz) δ 11.47 (s, 1H), 7.94 (d, J = 8.1 Hz, 1H), 7.62 (t, J = 7.2 Hz, 1H), 7.27 (d, J = 8.1 Hz, 1H), 7.20 (t, J = 7.5 Hz, 1H), 4.34 (q, J = 6.9 Hz, 2H), 1.31 (t, J = 7.2 Hz, 3H). Anal. Calcd. for C12H11NO4: C, 61.80%; H, 4.75%; N, 6.01%. Found: C, 61.72%; H, 4.78%; N, 6.10%.</p><!><p>Following the General procedure in Section 4.3.2, the reaction of N-methyl anhydride isatoic (1 g, 5.65 mmol) with NaH (0.27 g, 11.3 mmol), and diethyl malonate (4.52 g, 28.2 mmol) gave product 6b (0.56 g, 41%): mp 104°C; IR (KBr) υ 1631, 1593, 1562 cm−1; 1H NMR (DMSO-d 6, 300 MHz) δ 8.05 (d, J = 7.8 Hz, 1H), 7.45 (t, J = 7.5 Hz, 1H), 7.52 (d, J = 8.4 Hz, 1H), 7.31 (t, J = 7.2 Hz, 1H), 4.33 (q, J = 7.2 Hz, 2H), 3.54 (s, 3H), 1.30 (t, J = 7.2 Hz, 3H). Anal. Calcd. for C13H13NO4: C, 63.15%; H, 5.30%; N, 5.67%. Found: C, 63.24%; H, 5.27%; N, 5.61%.</p><!><p>The quinoline-3-carboxylate (1 eq) and its derivatives were suspended in methanol (20 mL). hydrazine (1.5 eq) was added and the mixture was heated at 100°C for 30 min. The precipitated compound was collected by filtration and used without further purification.</p><!><p>Following the General procedure in Section 4.3.5 the reaction of hydrazine (0.52 g, 16.3 mmol) with compound 6a (2 g, 8.58 mmol) gave product 7a (1.57 g, 84%): mp > 260°C; IR (KBr) υ 3167, 1674, 1616, 1531 cm−1; 1H NMR (DMSO-d 6, 300 MHz) δ 11.89 (s, 1H), 10.97 (s, 1H), 7.97 (d, J = 7.9 Hz, 1H), 7.68 (t, J = 7.2 Hz, 1H), 7.36 (d, J = 8.3 Hz, 1H), 7.29 (t, J = 7.3 Hz, 1H), 2.79 (s, 2H). Anal. Calcd. for C10H9N3O3: C, 54.79%; H, 4.14%; N, 19.17%. Found: C, 54.85%; H, 4.08%; N, 19.90%.</p><!><p>Following the General procedure in Section 4.3.5 the reaction of hydrazine (0.48 g, 15 mmol) with compound 6b (2 g, 8.10 mmol) gave product 7b (1.20 g, 64%): mp > 260°C; IR (KBr) υ 3328, 3240, 1647, 1589 cm−1; 1H NMR (DMSO-d 6, 300 MHz) δ 11.00 (s, 1H), 8.01 (d, J = 7.2 Hz 1H), 7.81 (t, J = 7.3 Hz, 1H), 7.62 (d, J = 8.6 Hz, 1H), 7.38 (t, J = 7.3 Hz, 1H), 4.95 (s, 2H), 3.63 (s, 3H). Anal. Calcd. for C11H11N3O3: C, 56.65%; H,4.75%; N, 18.02%. Found: C, 56.52%; H, 4.79%; N, 18.11%.</p><!><p>The corresponding quinoline-3-carboxyhydrazides 7a-b were stirred with 2,4-dihydroxybenzaldehyde or its derivatives in dimethyl sulfoxide and four drops of orthophosphoric acid for 15 min at room temperature. The mixture was then heated at 100°C for 1 h. The compound was collected by filtration and washed with water.</p><!><p>Following the General procedure in Section 4.3.8, the reaction of 2,4-dihydroxybenzaldehyde (0.31 g, 2.3 mmol) with compound 7a (0.5 g, 2.3 mmol) gave product 8a (0.75 g, 97%): mp > 260°C; IR (KBr) υ 3205, 1658, 1558 cm−1; 1H NMR (DMSO-d 6, 300 MHz) δ 13.19 (s, 1H), 12.01 (s, 1H), 11.08 (s, 1H), 10.10 (s, 1H), 8.51 (s, 1H), 7.96 (d, J = 7.9 Hz, 1H), 7.66 (t, J = 7.7 Hz, 1H), 7.34 (t, J = 7.3 Hz, 2H), 7.27 (t, J = 7.7 Hz, 1H), 6.34 (d, J = 8.4 Hz, 1H), 6.29 (s, 1H). Anal. Calcd. for C17H13N3O5: C, 60.18%; H, 3.86%; N, 12.38%. Found: C, 60.24%; H, 3.84%; N, 12.31%.</p><!><p>Following the General procedure in Section 4.3.8, the reaction of 2-hydroxy-5-methyl-benzaldehyde (0.15 g, 1.10 mmol) with compound 7a (0.25 g, 1.14 mmol) gave product 8b (0.31 g, 81%): mp > 260°C, IR (KBr) υ 3001, 1651, 1612, 1581, 1546 cm−1; 1H NMR (DMSO-d 6, 300 MHz) δ 13.65 (s, 1H), 8.87 (s, 1H), 8.27 (d, J = 7.9 Hz, 1H), 7.97 (t, J = 7.5 Hz, 1H) 7.67 (s, 1H), 7.57 (t, J = 7.5 Hz, 1H), 7.39 (d, J = 8.1 Hz, 1H), 7.11 (t, J = 8.6 Hz, 1H), 2.75 (s, 3H). Anal. Calcd. for C18H15N3O4: C, 64.09%; H, 4.48%; N, 12.46%. Found: C, 64.12%; H, 4.46%; N, 12.38%.</p><!><p>Following the General procedure in Section 4.3.8, the reaction of 2-hydroxy-5-methoxy-benzaldehyde (0.28 g, 1.84 mmol) with compound 7a (0.40 g, 1.83 mmol) gave product 8c (0.59 g, 92%): mp > 260°C; IR (KBr) υ 3355, 1670, 1651, 1577, 1542 cm−1; 1H NMR (DMSO-d 6, 300 MHz) δ 13.41 (s, 1H), 10.43 (s, 1H), 8.92 (s, 1H), 8.58 (s, 1H), 7.98 (d, J = 7.9 Hz, 1H), 7.68 (t, J = 7.9 Hz, 1H), 7.36 (d, J = 8.3 Hz, 1H), 7.23 (m, 1H), 7.12 (d, J = 2.8 Hz, 1H), 6.90 (m, 3H), 3.69 (s, 3H). Anal. Calcd. for C18H15N3O5: C, 61.19%; H, 4.28%; N, 11.89%. Found: C, 61.26%; H, 4.25%; N, 11.82%.</p><!><p>Following the General procedure in Section 4.3.8, the reaction of 2,6-dichloro-3-nitro-benzaldehyde (1 g, 4.55 mmol) with compound 7a (0.50 g, 2.30 mmol) gave product 8d (0.87 g, 92%): mp > 260°C; IR (KBr) υ 3298, 3093, 1735, 1620, 1596, 1539 cm−1; 1H NMR (DMSO-d 6, 300 MHz) δ 13.92 (s, 1H), 11.80 (s, 1H), 8.60 (s, 1H), 8.16 (d, J = 8.5 Hz, 1H), 8.01 (d, J = 8.1 Hz, 1H), 7.88 (d, J = 8.7 Hz, 1H), 7.66 (s, 1H), 7.35 (s, 1H), 7.26 (s, 1H). Anal. Calcd. for C17H10Cl2N4O5: C, 48.48%; H, 2.39%; Cl, 16.83%; N, 13.30%. Found: C, 48.44%; H, 2.37%; Cl, 16.86%; N, 13.25%.</p><!><p>Following the General procedure in Section 4.3.8, the reaction of 2,6-dichloro-3-nitro-benzaldehyde (0.78 g, 3.55 mmol) with compound 7b (0.30 g, 1.28 mmol) gave product 8e (0.51 g, 91%): mp 230°C; IR (KBr) υ 3058, 1635, 1577, 1558, 1519 cm−1; 1H NMR (DMSO-d 6, 300 MHz) δ 8.66 (s, 1H), 8.15 (d, J = 7.2 Hz, 2H), 7.87 (d, J = 8.7 Hz, 1H), 7.79 (s, 1H), 7.60 (s, 1H), 7.36 (s, 1H), 3.64 (s, 3H). Anal. Calcd. for C18H12Cl2N4O5: C, 49.67%; H, 2.78%; Cl, 16.29%; N, 12.87%. Found: C, 49.71%; H, 2.76%; Cl, 16.31%; N, 12.82%.</p><!><p>Following the General procedure in Section 4.3.8, the reaction of 2-hydroxy-5-methyl-benzaldehyde (0.15 g, 1.10 mmol) with compound 7b (0.25 g, 1.07 mmol) gave product 8f (0.34 g, 93%): mp > 260°C; IR (KBr) υ 2985, 1624, 1589, 1558, 1519 cm−1; 1H NMR (DMSO-d 6, 300 MHz) δ 12.55 (s, 1H), 9.95 (s, 1H), 7.81 (s, 1H), 7.30 (d, J = 7.7 Hz, 1H), 7.00 (t, J = 8.1 Hz, 1H), 6.82 (d, J = 8.6 Hz, 1H), 6.58 (d, J = 7.5 Hz, 1H), 6.53 (s, 1H), 6.30 (d, J = 8.3 Hz, 1H), 6.01 (d, J = 8.3 Hz, 1H), 2.83 (s, 3H), 1.41 (s, 3H). Anal. Calcd. for C19H17N3O4: C, 64.95%; H, 4.88%; N, 11.96%. Found: C, 64.92%; H, 4.86%; N, 11.95%.</p><!><p>Following the General procedure in Section 4.3.8, the reaction of 2-hydroxy-5-methoxy-benzaldehyde (0.16 g, 1.05 mmol) with compound 7b (0.25 g, 1.07 mmol) gave product 8g (0.33 g, 84%): mp > 260°C; IR (KBr) υ 2966, 1643, 1562, 1535 cm−1: 1H NMR (DMSO-d 6, 300 MHz) δ 13.49 (s, 1H), 10.51 (s, 1H), 8.65 (s, 1H), 8.15 (d, J = 7.5 Hz, 1H), 7.82 (d, J = 7.0 Hz, 1H), 7.65 (d, J = 8.1 Hz, 1H), 7.40 (t, J = 7.2 Hz, 1H), 7.15 (s, 1H), 6.92 (m, 2H) 3.74 (s, 3H), 3.67 (s, 3H). Anal. Calcd. for C19H17N3O5: C, 62.12%; H, 4.66%; N, 11.44%. Found: C, 62.19%; H, 4.63%; N, 11.41%.</p><!><p>Following the General procedure in Section 4.3.8, the reaction of 2-nitrobenzaldehyde (0.78 g, 5.16 mmol) with compound 7b (0.20 g, 0.86 mmol) gave product 8h (0.29 g, 92%): mp > 260°C; IR (KBr) υ 1670, 1566, 1523 cm−1; 1H NMR (DMSO-d 6, 300 MHz) δ 8.67 (s, 1H), 8.15 (s, 1H), 8.11 (d, J = 7.6 Hz, 1H), 8.07 (s, 1H), 7.83 (t, J = 6.9 Hz, 1H), 7.67 (m, 2H), 7.48 (s, 1H), 7.26 (s, 1H), 3.58 (s, 3H). Anal. Calcd. for C18H14N4O5: C, 59.02%; H, 3.85%; N, 15.29%. Found: C, 59.11%; H, 3.82%; N, 15.25%.</p>

PubMed Open Access

Superhydrophobic Thin Films Fabricated by Reactive Layer-by-Layer Assembly of Azlactone-Functionalized Polymers

We report an approach to the fabrication of superhydrophobic thin films that is based on the \xe2\x80\x98reactive\xe2\x80\x99 layer-by-layer assembly of azlactone-containing polymer multilayers. We demonstrate that films fabricated from alternating layers of the azlactone functionalized polymer poly(2-vinyl-4,4-dimethylazlactone) (PVDMA) and poly(ethyleneimine) (PEI) exhibit micro- and nanoscale surface features that result in water contact angles in excess of 150\xc2\xba. Our results reveal that the formation of these surface features is (i) dependent upon film thickness (i.e., the number of layers of PEI and PVDMA deposited) and (ii) that it is influenced strongly by the presence (or absence) of cyclic azlactone-functionalized oligomers that can form upon storage of the 2-vinyl-4,4-dimethylazlactone (VDMA) used to synthesize PVDMA. For example, films fabricated using polymers synthesized in the presence of these oligomers exhibited rough, textured surfaces and superhydrophobic behavior (i.e., advancing contact angles in excess of 150\xc2\xba). In contrast, films fabricated from PVDMA polymerized in the absence of this oligomer (e.g., using freshly distilled monomer) were smooth and only moderately hydrophobic (i.e., advancing contact angles of ~75\xc2\xba). The addition of authentic, independently synthesized oligomer to samples of distilled VDMA at specified and controlled concentrations permitted reproducible fabrication of superhydrophobic thin films on the surfaces of a variety of different substrates. The surfaces of these films were demonstrated to be superhydrophobic immediately after fabrication, but they became hydrophilic after exposure to water for six days. Additional experiments demonstrated that it was possible to stabilize and prolong the superhydrophobic properties of these films (e.g., advancing contact angles in excess of 150\xc2\xb0 even after complete submersion in water for at least six weeks) by exploiting the reactivity of residual azlactones to functionalize the surfaces of the films using hydrophobic amines (e.g., aliphatic or semi-fluorinated aliphatic amines). Our results demonstrate a straightforward and substrate-independent approach to the design of superhydrophobic and reactive polymer-based coatings of potential use in a broad range of fundamental and applied contexts.

superhydrophobic_thin_films_fabricated_by_reactive_layer-by-layer_assembly_of_azlactone-functionaliz

Introduction<!>Materials<!>General Considerations<!>Synthesis of Poly(2-vinyl-4,4-dimethylazlactone) (PVDMA)<!>Synthesis of Oligo(2-vinyl-4,4-dimethylazlactone)<!>Hydrolysis of Oligo(2-vinyl-4,4-dimethylazlactone)<!>Fabrication of Multilayered Films<!>Post-Fabrication Functionalization of PEI/PVDMA Films<!>Contact Angle Measurements<!>Fabrication and Characterization of Superhydrophobic PEI/PVDMA Films<!>Influence of Azlactone-Functionalized Oligomers on PEI/PVDMA Film Morphology<!>Post-Fabrication Modification of PEI/PVDMA Films with Hydrophobic Amines: Fabrication of Stable Superhydrophobic Surfaces<!>Summary and Conclusions<!>

<p>The term 'superhydrophobic' is generally used to describe surfaces and interfaces that exhibit advancing water contact angles in excess of 150° with low contact angle hysteresis.1–5 These properties generally cause water droplets to bead up and roll off of (rather than adhere to or spread on) the surface of a material and, thus, result in surfaces and interfaces that can completely resist wetting by aqueous media. Methods for the fabrication or functionalization of superhydrophobic surfaces are of significant interest because of the potential utility of these non-wetting materials in a broad range of consumer, industrial, and medically-oriented contexts (e.g., the design of 'self-cleaning' surfaces and textiles, new non-fouling surfaces, and membranes for oil/water separation, etc.). 1–5</p><p>The non-wetting behavior of superhydrophobic surfaces is generally understood to arise from a combination of two physicochemical factors: (i) the presence (and type) of micro/nanoscale surface roughness and (ii) the presence of surface functionality that is hydrophobic in nature (e.g., surfaces presenting fluorinated alkyl groups, etc).1 As a result, many approaches to the fabrication of superhydrophobic materials make use of methods for the introduction of hierarchical micro- and nanostructure to hydrophobic surfaces or methods for the deposition of hydrophobic molecules on surfaces that already display micro/nanostructured features. Methods such as chemical vapor deposition, various approaches to lithography, the deposition or dip-coating of nanoparticles, electrospinning, and layer-by-layer assembly have been used to fabricate superhydrophobic coatings on a variety of different types of surfaces.1–5 While each of these approaches has potential advantages with respect to the needs or constraints of a given application, several of these methods require the use of complex and expensive instrumentation or require different types of surface treatments that can involve the exposure of substrates to high temperatures (thereby limiting the types of substrates to which these methods can be applied). Straightforward and substrate-independent methods for the fabrication of superhydrophobic coatings under mild conditions (e.g., room temperature) could provide opportunities for the assembly of superhydrophobic surfaces on a broader range of substrates. Here, we report a solution-based, layer-by-layer approach to the fabrication of reactive, covalently crosslinked, and superhydrophobic thin films. These thin films possess both micro- and nanoscale surface features and chemically reactive groups that can be functionalized under mild conditions to produce films that retain their superhydrophobic properties for prolonged periods.</p><p>Past studies demonstrating methods for the layer-by-layer assembly6 of thin polymer films and composites have focused largely on the alternate and repetitive adsorption of oppositely charged polyelectrolytes on surfaces.7–11 These methods typically make use of aqueous solutions of polymer and result in films ('polyelectrolyte multilayers', or PEMs) that are assembled through multivalent electrostatic interactions. These methods are particularly useful for the fabrication of polymer-based surface coatings because they (i) permit incorporation of a broad range of materials, including components that can impart surface roughness (e.g., nanoparticles, carbon nanotubes, etc.), (ii) permit molecular-level and nanometer-scale control over film thickness and chemical composition, and (iii) can be deposited on a range of different types of materials, including substrates with complex or irregular geometries. 7–11 Layer-by-layer approaches have consequently been used to fabricate polymer thin films and composites of interest in a broad range of applications.7–11 Of particular relevance to the work reported here, several past studies have demonstrated layer-by-layer approaches to the fabrication of superhydrophobic surfaces,12–24 for example, by incorporating nanoparticles 12,13,19,21,23,24 or carbon nanotubes22 as structural elements of the films during assembly. While these past studies demonstrate the potential of layer-by-layer methods for the fabrication of superhydrophobic thin films, we note again that these polyelectrolyte-based approaches are generally confined to use with aqueous solutions and, thus, may not be useful for the fabrication of films on water-soluble substrates or substrates that react with (or are degraded by) water. In addition, many, but not all,15 PEM-based approaches to the design of superhydrophobic coatings have required several different surface treatments, including procedures such as chemical vapor deposition, to deposit non-water soluble, hydrophobic small molecules (e.g., aliphatic or fluorinated silanes) on the films to render the surfaces hydrophobic.12,13,18,19,21 Finally, we note that PEMs are assembled via weak electrostatic interactions that can be disrupted under certain environmental conditions. The work reported here makes use of an organic solvent-based approach to the layer-by-layer assembly of chemically crosslinked thin films that can be rapidly functionalized post-fabrication by treatment with organic solutions of hydrophobic primary amines.</p><p>We have described in several recent reports a 'reactive' layer-by-layer approach25 to the assembly of covalently crosslinked multilayered films that exploits the versatile reactivity of azlactone-functionalized polymers.26–30 Polymers bearing azlactone functionality, such as poly(2-vinyl-4,4-dimethylazlactone) (PVDMA, Eq. 1), are particularly useful for this approach because they react rapidly with primary amine-functionalized nucleophiles (Eq. 1).31 We have demonstrated that PVDMA can be deposited alternately on surfaces or interfaces with primary amine-functionalized polymers, such as branched poly(ethyleneimine) (PEI), to assemble covalently crosslinked polymer multilayers.26–30 The results of these past studies establish the layer-by-layer nature of the growth of these crosslinked multilayers on a broad range of surfaces and interfaces (e.g., on planar surfaces,26,27 on topologically complex substrates,26,29 and at interfaces formed between two immiscible liquids28), and that it is possible to assemble these multilayers using a broad range of organic solvents and optimized fabrication procedures.26–30 These past studies also reveal that the resulting polymer multilayers contain residual, unreacted azlactone functionality that is accessible for further reaction post-fabrication. These films thus provide a general and facile approach to the modification of surfaces with a broad range of chemical functionality.</p><p>The work reported here builds upon observations that arose during recent efforts in our laboratory to fabricate, characterize, and isolate freestanding PEI/PVDMA films. In a recent study, we reported that the layer-by-layer deposition of ~50 to 100 layer pairs (or 'bilayers') of PEI and PVDMA resulted in optically opaque films having a range of different micro- and nanoscale surface features (e.g., as characterized by scanning electron microscopy, SEM).30 Here, we describe an investigation into the origin of these observations and the development of methods that can be used to control surface roughness and fabricate films with specific surface properties (e.g., hydrophobic, hydrophilic, and superhydrophobic behavior, and films that transition in a time-dependent manner from superhydrophobic to hydrophilic upon prolonged exposure to aqueous media). We demonstrate that the roughness of these films (and thus their superhydrophobic properties) arises from the presence of cyclic azlactone-functionalized oligomers in the monomer used to synthesize PVDMA, and we report methods for the synthesis and intentional addition of these oligomers that can be used to fabricate surfaces with reproducible surface features and contact angles. Finally, we demonstrate that the superhydrophobic properties of these films are also influenced by the chemical functionality at the surface of the film (e.g., as dictated by the choice of the last polymer deposited during layer-by-layer fabrication or by post-fabrication treatment of residual azlactone groups with amine-based nucleophiles). Post-fabrication treatment with aliphatic or semi-fluorinated aliphatic amines yields films that are stable and remain superhydrophobic even after complete submersion in water for at least six weeks.</p><!><p>Branched poly(ethyleneimine) (PEI, MW = 25,000), reagent grade acetone, DMSO, 2,2′-azoisobutyronitrile (AIBN), trifluoroaceticacid (TFA), and 4,4,5,5,6,6,7,7,8,8,9,9,10,10,11, 11,11-heptadecafluoroundecylamine (HDFA) were purchased from Sigma Aldrich (Milwaukee, WI). Reagent grade THF and glass microscope slides were purchased from Fischer Scientific (Pittsburgh, PA). The monomer 2-vinyl-4,4-dimethylazlactone (VDMA) was a kind gift from Dr. Steven M. Heilmann (3M Corporation, Minneapolis, MN). Decylamine was purchased from TCI America (Portland, OR). All materials were used as received without further purification unless noted otherwise. Compressed air used to dry films and coated substrates was filtered through a 0.4 μm membrane syringe filter.</p><!><p>Gel permeation chromatography (GPC) was performed using a GPCmax-VE2001 Solvent/Sample module (Viscotek Corp., Houston, TX) and two PlusPore Organic GPC Columns (Polymer Laboratories, Amherst, MA) equilibrated to 40 °C. THF was used as the eluent at a flow rate of 1.0 mL/min. Data were collected using the refractive index detector of a Viscotek TDA 302 triple detector array and processed using the OmniSEC 4.5 software package. Molecular weights and polydispersities are reported relative to monodisperse polystyrene standards. Digital images were acquired using a Nikon Coolpix 4300 digital camera. Scanning electron micrographs were acquired on a LEO DSM 1530 scanning electron microscope at an accelerating voltage of 3 kV. Samples were coated with a thin layer of gold using a sputterer (30 s at 45 mA, 50 mTorr) prior to imaging. LC-MS data were obtained using a Shimadzu LCMS-2010 equipped with two LC-10ADvp pumps, an SCL-10Avp controller, an SIL-10ADvp autoinjector, an SPD-M10Avp UV/vis diode array detector, and a single quadrupole analyzer (for ESI). A Supelco 15 cm × 2.1 mm C18 wide-pore column was used for all LC-MS work. Standard RP-HPLC conditions for LC-MS were as follows: flow rate = 200 μL/min; mobile phase A = 0.4% formic acid in water; mobile phase B = 0.2% formic acid in acetonitrile. Static and dynamic water contact angles were measured using a Dataphysics OCA 15 Plus instrument and ImageJ (NIH). Glass substrates were cleaned with acetone, ethanol, methanol, and deionized water and dried under a stream of compressed air prior to the fabrication of multilayered films.</p><!><p>VDMA used for the polymerization of PVDMA was either distilled and stored with triethylamine (TEA) prior to polymerization (clear liquid) or was used without distillation (viscous, yellow liquid) (see text). The amounts of each reagent used for the polymerization of distilled or non-distilled monomer and the resulting polymer characteristics are noted below. VDMA was passed through a phenolic inhibitor removal resin followed by passage through a short plug of silica gel prior to polymerization. AIBN was weighed into a 25 mL round-bottomed flask equipped with a stir bar. Ethyl acetate was added and the solution was stirred to dissolve the AIBN. VDMA was added to the flask, the flask was capped with a septum, and the solution was purged with N2 for 10 minutes. The solution was stirred under N2 at 60 °C for 24 hours, after which time the viscous reaction mixture was cooled to room temperature and CH2Cl2 (5 mL) was added to the flask. The polymer was precipitated into hexanes to yield a white solid. The polymer was filtered and washed with hexanes, then redissolved in CH2Cl2 and precipitated once more in hexanes. PVDMA was isolated as a white solid. Polymerization using distilled VDMA: AIBN (11.8 mg, 0.0719 mmol), VDMA (1.0 g, 7.19 mmol), ethyl acetate (3 mL). 1H-NMR (300 MHz, CDCl3): δ = 1.37 (br s, (-CH3)2), 1.62–2.1 (br m, -CH2CH-), 2.69 (br s, -CH2CH-). FT-IR (ATR, cm−1): 2980–2900 (C-H), 1820 (lactone C=O), 1672 (C=N). Mn: 55,132; PDI = 3.6. Polymerization using undistilled VDMA: AIBN (27.2 mg, 0.166 mmol), VDMA (2.4 g, 17.4 mmol), ethyl acetate (6 mL). 1H-NMR (300 MHz, CDCl 3): δ = 1.37 (br s, (-CH3)2), 1.62–2.1 (br m, -CH2CH-), 2.69 (br s, -CH2CH-). FT-IR (ATR, cm−1): 2980–2900 (C-H), 1820 (lactone C=O), 1672 (C=N). Mn: 3,705; PDI = 2.4. For PVDMA polymerized in the presence of synthesized oligomer (synthesized as described below), distilled VDMA was used and oligo(VDMA) was added to the polymerization mixture at 7% (w/w) relative to VDMA. The polymerization was then carried out in analogy to the procedures described above. AIBN (11.8 mg, 0.0719 mmol), VDMA (1.0095 g, 7.19 mmol), oligo(VDMA) (71.5 mg, 0.504 mmol), ethyl acetate (3 mL). 1H-NMR (300 MHz, CDCl3): δ = 1.37 (br s, (-CH3)2), 1.62–2.1 (br m, -CH2CH-), 2.69 (br s, -CH2CH-). FT-IR (ATR, cm−1): 2980–2900 (C-H), 1820 (lactone C=O), 1672 (C=N). Mn: 20,724; PDI = 2.4.</p><!><p>Samples of oligo(VDMA) were synthesized in analogy to methods reported previously by Heilmann.32 Briefly, VDMA (1.0061 g, 7.2 mmol) was weighed into a 10 mL round-bottomed flask equipped with a stir bar. Ethyl acetate (1.7 mL) was added and stirred to dissolve the VDMA. TFA (37.5 μL, 0.505 mmol) was added, the flask was capped with a septum, and the solution was stirred at 65 °C for 24 hours. The reaction mixture turned yellow within 10 mins of stirring. After 24 hours, the dark orange mixture was precipitated into ~20 volumes of heptane and filtered to yield a yellow solid. The solid was redissolved in acetone (~1 mL) and precipitated a second time in heptane to give a yellow solid in 88% yield. 1H-NMR (300 MHz, CDCl3): δ = 1.2–1.8 (br m), 1.8–3.05 (br m), 3.57 (br s), 3.86 (br s). FT-IR (ATR, cm−1): 2974–2854 (C-H), 1816 (C=O, azlactone), 1735 (C=O, carboxylic acid of hydrolyzed azlactone), 1672 (C=O, amide of hydrolyzed azlactone). Mn: 806; PDI = 1.09.</p><!><p>Samples of oligo(VDMA) used for characterization by LC-MS were hydrolyzed prior to characterization using a protocol similar to that reported previously by Heilmann.32 Oligo(VDMA) (50.9 mg, 0.366 mmol) was weighed into a 4 dram vial equipped with a stir bar and dissolved in THF (0.5 mL). H2O (100 μL) and TFA (9 μL) were added to the vial, the vial was capped, and the solution was stirred at room temperature for 8 days. The mixture was precipitated into diethyl ether and filtered to produce a light yellow solid in 57% yield. FT-IR (ATR, cm−1): 2974–2854 (C-H), 1735 (C=O, hydrolyzed azlactone), 1672 (C=N, azlactone).</p><!><p>Solutions of PEI or PVDMA used for the fabrication of multilayered films were prepared in acetone (20 mM with respect to the molecular weight of the polymer repeat unit). Films were deposited manually layer-by-layer on glass substrates according to the following general protocol optimized and characterized extensively in our past studies of this acetone-based system:26–30 1) Substrates were submerged in a solution of PEI for 20 seconds, 2) substrates were removed and immersed in an initial acetone bath for 20 seconds followed by a second acetone bath for 20 seconds, 3) substrates were submerged in a solution of PVDMA for 20 seconds, and 4) substrates were rinsed in the manner described in step 2. This cycle was repeated until the desired number of PEI/PVDMA layers was reached (typically, 50–100 bilayers; the term 'bilayer' as used here refers to a single PEI/PVDMA layer pair). For these experiments, the concentrations of polymer solutions were maintained by addition of acetone as needed to compensate for solvent evaporation or by the replacement of solutions of polymer with fresh solutions (i) after every dipping cycle or (ii) after every 25 dipping cycles. Films were characterized or used in subsequent experiments immediately or were dried under a stream of filtered, compressed air and stored in a vacuum desiccator until use. All films were fabricated at ambient room temperature.</p><!><p>PEI/PVDMA films were functionalized post-fabrication by immersing film-coated substrates in solutions of an amine-functionalized nucleophile (e.g., decylamine or HDFA, 1 mM in THF) at room temperature for 20 hours using methods similar to those reported in past studies.26–30 Films were rinsed with THF and acetone after functionalization and dried with filtered air. Films were stored in a vacuum desiccator after fabrication and functionalization.</p><!><p>Contact angle measurements were made using a Dataphysics OCA 15 Plus instrument with an automatic liquid dispenser at ambient temperature. Static water contact angles were measured using a 4 μL droplet of deionized (18 MΩ) water in three different locations on a film measuring approximately 3 cm × 1 cm. Advancing and receding contact angles were measured during growth and shrinkage of 10 μL water droplets in three different locations on the film. Data are reported as the average (with standard deviation) of these individual measurements.</p><!><p>We recently reported an approach to the fabrication and isolation of freestanding and reactive polymer multilayers30 based on methods developed, characterized, and optimized by our group for the 'reactive' layer-by-layer assembly of PVDMA and PEI on a broad range of surfaces and interfaces. 26–29 During the course of these past studies, we sometimes observed the appearance of these films to change from uniform and optically clear to hazy or opaque after the deposition of ~25 bilayers of PEI/PVDMA. Additional characterization of these films by SEM revealed these changes in optical appearance to result from the presence of microscale and nanoscale topographic features on the surfaces of the films.30 The work reported here sought to investigate the origin of this surface roughness and determine whether this 'reactive' layer-by-layer process could be exploited to fabricate films with physicochemical properties that could be used to control the wettability of surfaces coated with these materials.</p><p>During our initial investigations, we observed large variations in the physical properties (e.g., roughness) of these films depending on the batch of polymer used to fabricate the films and, more specifically, whether or not the monomer used to synthesize the polymer was freshly distilled prior to use. For example, the fabrication of 100-bilayer PEI/PVDMA films using PVDMA that was synthesized using non-distilled VDMA (referred to hereafter as PVDMA ND) resulted in films that were opaque in appearance, as observed in our past studies and shown in the digital picture in Figure 1A. In sharp contrast, 100-bilayer PEI/PVDMA films fabricated using PVDMA synthesized using freshly distilled VDMA (referred to hereafter as PVDMADIST) resulted in clear, transparent films (as shown in Figure 1B). Closer inspection of the surface morphologies of these two films using SEM revealed the films to have substantially different surface characteristics. Figures 2A and 2B show top-down, high-resolution SEM images of 100-bilayer PEI/PVDMA films fabricated using PVDMAND and PVDMADIST, respectively. These images demonstrate that the surfaces of PEI/PVDMAND films are textured with micro- and nanoscale features, but that the surfaces of PEI/PVDMADIST films are smooth and relatively featureless at this magnification.</p><p>The influence of these differences in surface morphology on the properties of these films was further manifest as large differences in the wettability of substrates coated with these two different polymer films. As shown in Figure 3, advancing water contact angles of PEI/PVDMAND films were measured to be greater than 150° (as noted above, this is the contact angle threshold generally used to classify a surface as superhydrophobic). The advancing contact angles of PEI/PVDMADIST films were significantly lower (~75°) and more closely resembled the contact angles measured for 10-bilayer PEI/PVDMA films (~65°) reported in past studies.26,27</p><p>The results described above demonstrate clearly that the history of the VDMA monomer used to synthesize PVDMA plays a significant role in determining both the morphology and the resulting physical properties of PEI/PVDMA polymer multilayers (the results in Figure 2 also demonstrate that these differences do not arise simply from the influence of solvent or other physical or mechanical aspects of our layer-by-layer approach to fabrication). Because non-distilled monomer (PVDMAND) yielded films with properties that are of potential interest in the context of developing new superhydrophobic surfaces, we conducted a series of subsequent experiments to characterize the composition of the monomer and determine whether we could establish well-defined procedures and conditions that could be used to fabricate superhydrophobic films reproducibly (that is, to design rational approaches for assembly that would not rely on the age or history of the monomer, etc.). The results of these additional investigations are described in the section below. Additional characterization of water contact angles and other surface properties are described in subsequent sections.</p><!><p>Previous reports by Heilmann et al. have demonstrated that azlactone-functionalized vinyl monomers such as VDMA can undergo oligomerization reactions when stored for long periods of time in the absence of a proton scavenger, or when treated directly with catalytic amounts of acid.32 These reactions do not occur through a chain growth-type mechanism to yield short chains of PVDMA, but rather proceed by protonation of the azlactone nitrogen of VDMA followed by Michael addition to the olefin to form cyclic structures containing, most often, four, five, or six monomer units (although higher order oligomers can also be formed). A complete analysis of these oligomeric structures and a detailed discussion of the mechanism of formation of these oligomers have been reported previously (see also Figures S1 and S2 of the Supporting Information for the proposed mechanism and the structures of the oligomeric compounds that form).32 To determine whether our non-distilled supply of VDMA contained these oligomerization products, we added aliquots of a PVDMAND solution drop-wise to heptane (following a protocol reported by Heilmann et al. for the isolation of oligomers from VDMA)32 and observed the formation of a precipitate. Characterization of several different batches of monomer using this procedure revealed these samples to contain between 4% and 7% of this precipitate (w/w) depending upon the batch of monomer. Characterization of these precipitates using liquid chromatography-mass spectrometry (LC-MS) revealed a range of structures having the molecular weights shown in Table S1 of the Supporting Information. These molecular weights correspond to a mixture of VDMA-based oligomers composed of dimers through hexamers (with varying degrees of hydrolysis) that was similar to the results reported previously by Heilmann et al.32 Similar precipitation experiments using samples of freshly distilled VDMA monomer revealed these samples to be devoid of oligomers.</p><p>To investigate the potential influence of the oligomers identified above more completely, and to develop well-defined procedures for the fabrication of superhydrophobic films, we performed a series of experiments using samples of freshly distilled VDMA monomer and PVDMADIST (polymer synthesized from freshly distilled monomer) containing measured amounts of authentic, independently synthesized oligomer. We synthesized authentic samples of oligomer under controlled conditions by treating freshly distilled VDMA with TFA (see Materials and Methods for additional details).32 LC-MS characterization of samples of the resulting oligomers revealed them to be composed of a mixture of trimers, tetramers, pentamers, and hexamers (see Table S2 of the Supporting Information for additional details related to the characterization of these independently-synthesized oligomers). The composition of these mixtures was qualitatively similar to the mixtures of oligomers isolated from the non-distilled VDMA monomer (as described above) and to results reported previously by Heilmann et al. for the synthesis of these oligomers under similar controlled reaction conditions.32 These intentionally synthesized oligomers were used for all subsequent studies described below.</p><p>We next prepared two different samples of PVDMA containing these intentionally synthesized oligomers to characterize the influence of oligomer on the properties of our PEI/PVDMA films. The first sample was prepared by the conventional free radical polymerization of a solution of freshly distilled VDMA containing 7% (w/w) of the independently synthesized oligomer mixture (this sample is referred to hereafter as PVDMAOLIGO). The second sample was prepared by the addition of 7% (w/w) of oligomer to a sample of PVDMADIST (i.e., by the addition of oligomer to a sample of polymer synthesized previously using freshly distilled monomer). These two different PVDMA samples were then used to fabricate PEI/PVDMA films using procedures identical to those used to fabricate the films shown in Figures 1A and 1B.</p><p>The images in Figures 1C and 1D show digital pictures of PEI/PVDMA films 100 bilayers thick fabricated using either PEI and PVDMA OLIGO (Figure 1C) or PEI and PVDMADIST doped with 7% oligomer (Figure 1D). Inspection of these images reveals the visual appearance of films fabricated using PVDMAOLIGO (Figure 1C) to resemble more closely the appearance of films fabricated from PVDMAND (e.g., Figure 1A). Further inspection reveals the appearance of films assembled using PVDMADIST doped with 7% oligomer after polymerization to resemble more closely the appearance of films fabricated using PVDMADIST in the absence of added oligomer (e.g., Figure 1B). Further characterization of the morphologies of these films using SEM revealed that the surfaces of PEI/PVDMAOLIGO films (Figure 2C) were rough and had micro- and nanoscale features similar to those observed for PEI/PVDMA ND films (Figure 2A). In contrast, the surfaces of films fabricated using PEI and PVDMA DIST doped with 7% oligomer (Figure 2D) were smooth and relatively featureless, similar to films fabricated using PEI and PVDMADIST (Figure 2B). Characterization of the water contact angles of these films revealed that PEI/PVDMAOLIGO films (Figures 1C and 2C) were also superhydrophobic and exhibited advancing contact angles of ~155° (Figure 3). Films assembled using PVDMADIST doped with oligomer after polymerization (Figures 1D and 2D), however, exhibited contact angles similar to those of PEI/PDVMADIST films (advancing contact angle ~72°, as shown in Figure 3).</p><p>The results of the experiments above, when combined, demonstrate that the influence of added oligomer on film roughness arises substantially from its presence during the polymerization of PVDMA, and not simply from its presence in solution during film fabrication. These results thus suggest that the polymerization of VDMA in the presence of oligomer influences the structure and properties of PVDMA in ways that influence its behavior during assembly. We note, in this context, that despite in-depth characterization of the formation of oligomers in samples of VDMA, 32 the influence of these oligomers on the polymerization of VDMA has not, to our knowledge, been broadly characterized. One possibility is that these oligomers could be incorporated into the structure of a growing polymer during polymerization in ways that alter substantially the structure or solution behavior (e.g., conformation) of the resulting polymers and, thus, the manner in which they are deposited during fabrication. (For example, the development of surface roughness in polyelectrolyte-based multilayers has been attributed in past studies to changes in the conformations of polyelectrolytes that occur upon changes in the pH and ionic strength of polymer solutions used for film fabrication.)33–35 We note in this context, that these cyclic oligomers are not functionalized with vinyl groups (e.g., see structures shown in Figure S2 of the Supporting Information). We note further, however, that Heilmann et al. have proposed enol tautomers as intermediates during the formation of these oligomers32 and that, if present, these enol tautomers could potentially be incorporated into a growing polymer chain during free radical polymerization. Another possibility is that these oligomers could behave as chain transfer or chain termination agents during polymerization. Support for this possibility is provided by the results of additional GPC characterization and the observation that PVDMA polymerized in the presence of oligomers typically yields polymers with molecular weights (e.g., Mn <20,000 g/mol) that are lower than PVDMA polymerized in the absence of oligomers (e.g., Mn >50,000 g/mol) under otherwise identical conditions. Initial characterization by nuclear magnetic resonance spectroscopy did not reveal significant differences in PVDMA synthesized in the presence or absence of added oligomer. Additional experiments will be required to establish more completely the potential influence of these oligomers on polymerization kinetics and potential changes in polymer structure and/or solution conformation.</p><p>We also considered the possibility that low molecular weight PVDMA or free oligomer could potentially diffuse freely out of a film during fabrication and form crosslinked, particulate complexes upon contact with solutions of PEI during or after each dipping cycle used to assemble the films. Such complexes could then deposit on the surface of a film during subsequent deposition cycles and contribute to the formation of surface roughness. In this context, we note that the formation and intentional incorporation of particulate interpolyelectrolyte complexes into polyelectrolyte multilayers has been used to fabricate superhydrophobic PEMs.36 To investigate this possibility, we performed additional experiments in which the PEI solutions used during fabrication were exchanged with fresh PEI solutions after the deposition of every PEI layer to reduce substantially the potential for complex formation and/or re-deposition to occur. These experimental conditions also resulted in the formation of rough, textured surfaces. These results suggest that the formation of complexes in the PEI solutions either does not contribute significantly to the surface morphologies observed, or that the formation and re-deposition of the complexes occurs rapidly (e.g., on the time scale of the 20-second polymer deposition times used to fabricate these films). Finally, we note that the results of the experiments described above demonstrating that films fabricated using PVDMADIST solutions containing free, intentionally added oligomer are smooth and featureless are also not consistent with a mechanism for the formation of roughness that is based on the diffusion of oligomer or the formation of PEI/oligomer complexes. Although additional experiments will be required to fully elucidate the underlying mechanism for the formation of these rough surfaces, we do conclude that (i) the primary influence of the oligomer arises from its presence during the polymerization of VDMA, and (ii) that the addition of controlled amounts of intentionally synthesized oligomer to samples of freshly distilled VDMA can be used to fabricate rough and superhydrophobic films reproducibly. Additional characterization of the superhydrophobic properties of these films is discussed below.</p><!><p>Additional characterization of contact angles of the rough films described above revealed that they were substantially dependent upon the structure of the last polymer deposited during fabrication. For example, whereas films terminated with a layer of PVDMA exhibited contact angles in excess of 150° (as described above), droplets of water placed on the surfaces of films terminated with a final layer of PEI spread upon contact (i.e., static contact angles were approximately 40°). These results are also consistent with the layer-by-layer nature of the assembly process used to fabricate these films. Further characterization of PVDMA-terminated films, however, also revealed the superhydrophobic properties of these films to change over time (for example, upon prolonged exposure to water; discussed in greater detail below). These results, when combined, demonstrate that surface chemistry also plays an important role in defining the superhydrophobic behavior of these rough films.</p><p>The results of past studies demonstrate that the residual azlactone functionality in PEI/PVDMA films can be exploited to chemically modify the surfaces and properties of these materials by post-fabrication treatment with a range of small molecule amines.26–30 Our next experiments sought to determine whether the treatment of rough PEI/PVDMA films with hydrophobic amines could be used to either (i) increase the superhydrophobicity of these films or (ii) create 'stable' films that would retain their superhydrophobic properties for prolonged periods (relative to those described above) upon exposure to water. We selected the hydrophobic amines n-decylamine and heptadecafluoroundecylamine (HDFA) for these studies based on the past use of these and similar structural motifs for the preparation of hydrophobic and superhydrophobic surfaces.3,5</p><p>We fabricated 100-bilayer PEI/PVDMADIST and PEI/PVDMAOLIGO films on glass substrates and treated these films with either decylamine or HDFA (1 mM in THF; see Materials and Methods for additional details). Figure 4 shows a plot of the dynamic contact angles measured for unmodified, decylamine-treated, and HDFA-treated PEI/PVDMADIST and PEI/PVDMAOLIGO films. Treatment of PEI/PVDMADIST films with decylamine and HDFA resulted in a moderate increase in the hydrophobicity of the surfaces (i.e., the average advancing contact angle increased from ~75° to ~95°). We note that contact angles of ~95° are similar to contact angles measured for decylamine-functionalized 10-bilayer PEI/PVDMA films fabricated on silicon substrates that we reported previously.26,27 In contrast, the advancing contact angles of already-superhydrophobic PEI/PVDMAOLIGO films treated with either decylamine or HDFA did not increase measurably (we note, however, that treatment with these hydrophobic amines did result in slightly lower contact angle hysteresis; see Figure 4).</p><p>Although treatment of PEI/PVDMAOLIGO films with hydrophobic small molecules did not increase the contact angles, the long-term stability of the superhydrophobic properties of these amine-treated films was significantly improved relative to those of unmodified (i.e., azlactone-containing) films. For example, superhydrophobic PEI/PVDMAOLIGO films that were not modified by treatment with hydrophobic amines became water-absorbent after several days of submersion in water (e.g., after six days; see image in Figure 6D). This behavior is consistent with the ring-opening hydrolysis of unreacted azlactone functionality in these untreated materials and the subsequent time-dependent formation of hydrophilic carboxylate groups (discussed in more detail below). In contrast, the advancing contact angles of decylamine-treated and HDFA-treated PEI/PVDMAOLIGO films did not change significantly even after total submersion in water for at least six weeks. Figure 5 shows a plot of the dynamic contact angles of decylamine-functionalized and HDFA-functionalized films immediately after fabrication and after three or six weeks of immersion in a water bath at room temperature. The contact angle hysteresis increased slightly for decylamine-treated films over time in water and, consequently, water droplets did not roll off of these surfaces as readily as they did immediately after functionalization. The advancing and receding contact angles for PEI/PVDMAOLIGO films modified with HDFA, however, did not change substantially even after six weeks in water and water droplets rolled readily off of these films.</p><p>Further characterization of unmodified, decylamine-treated, and HDFA-treated films after submersion in water revealed that changes in contact angles were not a result of changes in the microstructure (i.e., roughness) of the surfaces of the films. The images in Figures 6A–C show top-down SEM images of an unmodified PEI/PVDMAOLIGO film that was submerged in water for one week (Figure 6A), and decylamine-modified (Figure 6B) and HDFA-functionalized (Figure 6C) PEI/PVDMAOLIGO films that were immersed in water for six weeks. Inspection of these images reveals that the micro- and nanoscale morphologies of the films were rough and textured, similar to that observed for PEI/PVDMAOLIGO films prior to treatment with the hydrophobic amines and exposure to water (see Figure 2C for comparison). Figures 6D–F show images of water droplets (4 μL) placed on unmodified, decylamine-treated, and HDFA-treated PEI/PVDMAOLIGO films that were soaked in water for either one week (for unmodified films) or six weeks (for treated films). As revealed by the absence of an observable water droplet in Figure 6D, unmodified PEI/PVDMA OLIGO films readily absorbed water after being exposed to water for one week.</p><p>The images shown in Figures 6E and 6F further demonstrate that decylamine-treated and HDFA-treated films continue to resist wetting by water, even after exposure to water for several weeks. We note that HDFA-treated PEI/PVDMAOLIGO films (Figure 6F) exhibited higher static contact angles than decylamine-treated films (Figure 6E) after six weeks in water and thus appear to be more durable over long periods of exposure to water. The results of these experiments demonstrate that exposure to water does not change substantially the structure or morphology of these covalently crosslinked multilayers, but that the water contact angles do change over time if the films are not treated with a hydrophobic amine. As noted above, residual azlactone functionality in PEI/PVDMA films will hydrolyze in the presence of water (or upon exposure to water vapor) to form carboxylic acid groups, and, it is therefore likely that the loss of superhydrophobicity observed for unmodified films exposed to water results from the formation of more hydrophilic functional groups on the surfaces of these films. Modification of these films with hydrophobic amines results in the formation of stable amide linkages (e.g., see Equation 1) that are less susceptible to hydrolysis and makes possible the fabrication of superhydrophobic surfaces that are stable when submerged in water for extended periods.</p><!><p>We have reported a facile and straightforward approach to the fabrication of superhydrophobic surface coatings that makes use of methods for the 'reactive' layer-by-layer assembly of azlactone-functionalized polymers. Our results demonstrate that polymer multilayers fabricated from PEI and PVDMA 100 bilayers thick exhibit micro- and nanoscale surface features that impart non-wetting, superhydrophobic behavior to film-coated substrates. These rough films exhibit advancing contact angles of ~155° without any additional surface treatment. The roughness of these films was attributed to the presence of azlactone-functionalized oligomers during the polymerization of PVDMA. Films fabricated using either (i) PVDMA synthesized in the absence of these oligomers or (ii) samples of PVDMA to which oligomers were added intentionally prior to fabrication resulted in smooth, featureless surfaces that were not superhydrophobic). Although rough PEI/PVDMA films exhibited superhydrophobic properties without any additional treatment, these properties were lost upon prolonged exposure to water. However, the reactivity of residual azlactone functionality in PEI/PVDMA films permitted facile post-fabrication modification with aliphatic and fluorinated aliphatic amines to produce surfaces that retained their superhydrophobic properties for extended periods of time (e.g., after at least six weeks of complete submersion in water).</p><p>The methods reported here do not require complicated, expensive, or harsh treatment processes, and should therefore facilitate the fabrication of superhydrophobic films on a wide range of different materials and substrates. In addition, because these methods are based on layer-by-layer assembly in organic solvents, this approach should be well suited to the assembly of superhydrophobic coatings on the surfaces of water-soluble or pH-sensitive materials that cannot be used as substrates using methods for the aqueous-based assembly of superhydrophobic polyelectrolyte multilayers.</p><!><p>Digital photographs of 100-bilayer PEI/PVDMA films fabricated using PVDMAND (A), PVDMADIST (B), PVDMAOLIGO (C), and PVDMADIST doped with 7% (w/w) synthesized oligomer (D).</p><p>Top-down SEM images of 100-bilayer PEI/PVDMA films fabricated using PVDMAND (A), PVDMADIST (B), PVDMAOLIGO (C), and PVDMADIST doped with 7% (w/w) synthesized oligomer (D). Scale bars = 2 μm.</p><p>Contact angles measured for PEI/PVDMA films 100 bilayers thick fabricated using PEI and either PDVMAND, PVDMADIST, PVDMAOLIGO, or PVDMADIST doped with 7% (w/w) synthesized oligomers. Advancing contact angles are represented by the dark gray bars and receding contact angles are represented by the light gray bars.</p><p>Contact angles measured for unmodified, decylamine-treated, and HDFA-treated PEI/PVDMADIST and PEI/PVDMAOLIGO films. Advancing contact angles are represented by the dark gray bars and receding contact angles are represented by the light gray bars.</p><p>Dynamic contact angles as a function of time for decylamine-treated and HDFA-treated PEI/PVDMAOLIGO films submerged in water. Advancing contact angles are represented by dark gray bars and receding contact angles are represented by light gray bars.</p><p>(A–C) SEM images of 100-bilayer films fabricated from PEI and PVDMA OLIGO that were (A) unmodified, (B) treated with decylamine, or (C) treated with HDFA. The images were acquired after soaking the unmodified film in water for one week and the decylamine-treated and HDFA-treated films in water for six weeks (see text). (D–F) Images of water droplets (4 μL) on (D) an unmodified film that was immersed in water for one week, (E) a decylamine-treated film that was submerged in water for six weeks, and (F) an HDFA-treated film that was immersed in water for six weeks (F). Scale bars = 2 μm.</p>

PubMed Author Manuscript

Disruption of mitochondrial quality control genes promotes caspase-resistant cell survival following apoptotic stimuli

In cells undergoing cell-intrinsic apoptosis, mitochondrial outer membrane permeabilization (MOMP) typically marks an irreversible step in the cell death process. However, in some cases, a subpopulation of treated cells can exhibit a sublethal response, termed “minority MOMP.” In this phenomenon, the affected cells survive, despite a low level of caspase activation and subsequent limited activation of the endonuclease caspase-activated DNase (DNA fragmentation factor subunit beta). Consequently, these cells can experience DNA damage, increasing the probability of oncogenesis. However, little is known about the minority MOMP response. To discover genes that affect the MOMP response in individual cells, we conducted an imaging-based phenotypic siRNA screen. We identified multiple candidate genes whose downregulation increased the heterogeneity of MOMP within single cells, among which were genes related to mitochondrial dynamics and mitophagy that participate in the mitochondrial quality control (MQC) system. Furthermore, to test the hypothesis that functional MQC is important for reducing the frequency of minority MOMP, we developed an assay to measure the clonogenic survival of caspase-engaged cells. We found that cells deficient in various MQC genes were indeed prone to aberrant post-MOMP survival. Our data highlight the important role of proteins involved in mitochondrial dynamics and mitophagy in preventing apoptotic dysregulation and oncogenesis.

disruption_of_mitochondrial_quality_control_genes_promotes_caspase-resistant_cell_survival_following

<!>Screening strategy and assay development<!><!>Screening strategy and assay development<!>The siRNA screen uncovers potential regulators of MOMP<!>The interplay of MQC and MOMP<!><!>Other MQC-related hits in category III<!><!>An assay to measure clonogenic cell survival despite caspase activation<!><!>MQC deficiency promotes minority MOMP<!>Conclusion<!>Cell lines<!>siRNA screen: cell transfections and treatments<!>High-throughput image acquisition and analysis<!>Confocal microscopy<!>CRISPR/Cas9-mediated gene depletion<!>Assay for clonogenic survival after caspase activation via MOMP (outlined in Fig. 2)<!>Data availability<!>Supporting information<!>Conflict of interest<!>Supporting information

<p>Edited by Ursula Jakob</p><p>Apoptosis is a ubiquitous cellular self-elimination process that is critical for the homeostasis of various cell populations. Dysregulated apoptosis is integral to cancer progression and contributes to multiple diseases, including immune and neurodegenerative disorders. Many cancer therapies rely on the enhanced apoptotic death of tumor cells. Apoptosis frequently involves a "cell-intrinsic" pathway involving mitochondria. The central event in mitochondria-dependent apoptosis is mitochondrial outer membrane permeabilization (MOMP), which is induced by B-cell lymphoma 2 (BCL2)–associated X, apoptosis regulator (BAX) and Bcl-2 homologous antagonist/killer (BAK), the key proapoptotic BCL-2 family proteins (1, 2). BAK is constitutively located on the mitochondrial outer membrane (MOM), whereas BAX is mostly soluble in the cytoplasm. When cells receive an apoptotic stress signal, molecules belonging to a subset of the BCL-2 family termed "Bcl-2 homology domain 3 (BH3)–only proteins" activate BAX and BAK. Consequently, BAX translocates to the MOM, and both BAX and BAK integrate into the outer membrane and trigger MOMP by inducing the formation of large membrane pores (3, 4, 5, 6, 7). In concert with these BCL-2 family members, additional MOM proteins facilitate BAX-induced pore formation (4). Antiapoptotic BCL-2 family members (including BCL-2, BCL-extra large, and myeloid cell leukemia-1) inhibit MOMP by sequestering the BH3-only proteins and by antagonizing the pore-forming activity of BAX and BAK.</p><p>The apoptotic pores in the MOM allow proteins normally residing in the mitochondrial intermembrane space to escape into the cytoplasm. Certain of these proteins are apoptogenic, including cytochrome c, second mitochondria-derived activator of caspase (SMAC)/direct inhibitor of apoptosis-binding protein with low pI, and OMI/high-temperature requirement factor A1. These proteins induce apoptotic protease-activating factor-1 (APAF-1)–dependent activation of caspase-9 and the effector caspases (most prominently caspase-3), which cleave certain protein substrates that actively promote cell destruction and engulfment (8, 9, 10). Even when caspase activity is blocked, MOMP per se usually leads to cell death, as it compromises energy metabolism in the permeabilized mitochondria (11, 12, 13). Although mitochondrial membrane potential and cellular ATP are maintained for some time after MOMP, cells treated with caspase inhibitors typically exhibit a gradual decline in mitochondrial respiration and a complete loss of clonogenic survival over the course of 2 or 3 days (12).</p><p>Recent reports describe a phenomenon, termed "minority MOMP," in which cells exposed to relatively weak cell stressors can evade MOMP-dependent death (14, 15, 16). In these atypical cells, only a fraction of the cell's mitochondria undergoes MOMP. As a result of minority MOMP, small amounts of cytochrome c and SMAC are released into the cytosol, and downstream caspases become activated at sublethal levels. This low-level caspase activation leads to a correspondingly limited activation of the apoptotic caspase-activated DNase (DNA fragmentation factor subunit beta) endonuclease, which enters the cell nucleus and produces a degree of DNA damage. The surviving cells exhibit genome instability and an increased propensity to become oncogenic (17).</p><p>The mechanisms controlling the uniformity of MOMP in cells, and hence the frequency of minority MOMP, are unclear. To discover genes that affect MOMP, we conducted a focused high-content siRNA screen that analyzed 1318 genes that had been annotated in databases as having some relationship to mitochondria. A novel aspect of our screen was the identification of genes whose silencing produced a heterogeneous MOMP phenotype in which some mitochondria within the cell were permeabilized, whereas others remained intact. Our screen yielded functionally diverse candidate genes, including a significant group of genes associated with mitophagy, autophagy, and mitochondrial dynamics, the processes constituting the system of mitochondrial quality control (MQC) (18, 19). We hypothesized that an increased heterogeneity of MOMP observed in MQC-compromised cells could increase the frequency of minority MOMP (17). If so, downregulation of MQC-related genes would tend to allow cells to survive stresses that produce a degree of caspase activation, as a result of limited MOMP. To test this hypothesis, we developed a single-cell assay to measure the long-term clonogenic cell survival of cells that had exhibited a measurable degree of caspase activation, in response to treatment with a BH3-mimetic drug. Our results confirmed that deficiencies in various MQC proteins enhanced the survival of caspase-engaged cells. We conclude that mitochondrial dynamics and mitophagy play a critical role in limiting the frequency of minority MOMP. Thus, proper functioning of the MQC system can be important to prevent oncogenesis that results from an incomplete execution of the mitochondrial apoptosis program.</p><!><p>To identify novel cell-intrinsic regulators of MOMP, we designed a high-content image-based siRNA screen. As we were primarily interested in mitochondrial function, we carried out a focused siRNA screen targeting genes with annotated relationships to mitochondria (Table S1). Our strategy took advantage of a HeLa cell line expressing two fluorescent reporter proteins (described previously by Llambi et al. (20)) that enabled us to simultaneously interrogate both BAX activation and MOMP. In living cells, Venus-BAX is cytoplasmic (diffuse), and OMI-mCherry is localized to mitochondria. When apoptosis is induced, Venus-BAX is translocated to mitochondria and OMI-mCherry is released into the cytoplasm and degraded. Normally, once apoptotic BAX translocation is initiated in a given cell, mitochondrial intermembrane space proteins (including cytochrome c or OMI) are released in a synchronous and rapid fashion (14, 20, 24). The release of proteins from the mitochondrial interior can only be restricted under special circumstances, for example, cristae junction remodeling by overexpression of a mutant form of the optic atrophy 1 (OPA1) protein (25) or MOMP inhibition by a recently described endolysosome-linked mechanism (26). However, in some cells, the release of intermembrane proteins is not an all-or-none event, and mitochondria within the same cell display heterogeneity in MOMP response (16, 17). We hypothesized that downregulation of certain genes could increase MOMP heterogeneity and promote post-MOMP cell survival, even in caspase-proficient cells.</p><!><p>Imaging-based screening assay for regulators of apoptotic MOMP response.A, representative epifluorescence microscopy of untreated and etoposide-treated cells expressing Venus-BAX (green; YFP channel) and OMI-mCherry (red; Texas Red channel). Bottom panel illustrates enlarged cells with indicated phenotypes. Nuclei were stained with Hoechst-33342 (blue; DAPI channel). B, identification of Venus-BAX and OMI-mCherry puncta using MetaXpress granularity application module. A representative enlarged image of etoposide-treated cells (left panel) and corresponding image segmentations. The granularity module identifies Venus-BAX (middle panel) and OMI-mCherry puncta. Note that cells with BAX "granules" do not contain OMI "granules" and vice versa. Nuclear segmentation settings correctly identify fragmented (apoptotic) and normal size nuclei. C, BAX siRNA inhibits Venus-BAX puncta formation and OMI-mCherry release in etoposide-treated cells. D, a confocal image of live Venus-BAX/OMI-mCherry cells transfected OMA1 siRNA. Arrows indicate cells with heterogeneous MOMP. The scale bars in (A–D) represent 20 μm. E and F, examples of phenotype quantification using granularity application module and high-throughput microscopy. E, quantification of indicated phenotypes in untreated and etoposide-treated cells transfected with a nontargeting (control) siRNA; "nondetermined" phenotype corresponds to a small fraction of nonfluorescent cells. F, effects of indicated siRNAs on the phenotype distribution. Data are mean ± SEM (n = 3 replicate wells). G, sorted assay scores for the primary screen siRNA set in triplicate plates. Numbers indicate the percentage of cells positive for both Venus-BAX and OMI-mCherry foci. Values shown are mean and SEM, n = 3. Hits were identified from the top 10% tail of the BAX/OMI score. BAX, BCL2-associated X, apoptosis regulator; DAPI, 4′,6-diamidino-2-phenylindole; MOMP, mitochondrial outer membrane permeabilization.</p><!><p>A BAX siRNA pool (which targets both endogenous BAX and ectopic Venus-BAX expression) potently inhibited etoposide-induced apoptosis and markedly reduced Venus-BAX fluorescence (Fig. 1, C and F). However, knockdown of BAK did not inhibit apoptosis, indicating that in these etoposide-treated cells, MOMP was predominantly BAX dependent. Transfection efficiency in high-throughput screening experiments was verified using a cell death–inducing transfection marker (siTOX) that consistently produced ∼90% loss of cells within 2 days post-transfection (not shown). Efficient knockdown of other proteins produced by siRNA pools was confirmed in supporting experiments (Fig. S1A).</p><!><p>The image-based assay was used to screen 1318 gene-specific siRNA pools. We ranked gene scores in three groups corresponding to the cell phenotype categories noted previously: genes whose knockdown (I) increased the percentage of Venus-BAX puncta, (II) decreased the percentage of cells with Venus-BAX puncta, or (III) increased the percentage of BAX/OMI double-positive cells. When we applied a 10% cutoff in the score distribution range, we identified ∼200 initial "hits" (with scores approximately threefold to fivefold above or below controls for the categories I and II). Hits in category III were from the top 10% tail of the score distribution (Fig. 1G). For a secondary screen, we chose to reassay 95 of the primary screen hits, based on their effect scores and our judgment concerning their biological interest (we decided not to pursue some "housekeeping" genes). In this secondary screen, we tested the four siRNAs from each siRNA pool individually. To reduce the likelihood of off-target effects, we required at least three of the four individual siRNAs to give concordant results (34). Applying this more stringent criterion yielded a final list of 63 candidate genes (Table S2). Several hits in categories I and II were consistent with previous reports. For example, the effect of siRNA targeting Bcl-2/adenovirus E1B 19-kDa interacting protein 3 (BNIP3) (category II) is consistent with the known cell death–promoting activity of this protein (35). Acting in the opposite manner, the siRNAs scoring in category I increased the percentage of cells undergoing MOMP. For example, several candidate genes in this group are required for metabolism of the mitochondrial lipid, cardiolipin (phospholipid scramblase 3 [PLSCR3], PRELI domain–containing protein 1 [PRELIDI], and TP53-regulated inhibitor of apoptosis 1 [TRIAP1]). Although cardiolipin is important for BAX pore formation (6, 36), in certain paradigms, cardiolipin deficiency potentiates the release of apoptogenic proteins (37, 38). In particular, the deficiency of p53-regulated protein TRIAP1 impaired cardiolipin level in mitochondria, compromised bioenergetics, and potentiated cytochrome c release (38). Potential novel regulators of MOMP include metabolic enzymes, long-chain acyl-coenzyme A dehydrogenase, which catalyzes one of the early steps in the circle of mitochondrial beta-oxidation of fatty acids, and fumarylacetoacetate hydrolase domain–containing protein 1 with putative oxaloacetate decarboxylase activity in mitochondria. The roles of these proteins in mitochondrial metabolism and cell senescence are emerging (39, 40), and their inhibitory effects on MOMP could be of interest for further investigations. In this study, we did not further pursue the hits in categories I and II but focused on the unique phenotype produced by the hits in category III. Of note, in nearly all cells containing both BAX and OMI puncta, BAX and OMI were not colocalized, implying that some mitochondria within a cell had undergone MOMP, whereas others had not (similar to the phenotype shown in Fig. 1D). Perhaps the most striking outcome of our screen is that a majority of hits in category III are related to the MQC system.</p><!><p>The concept of MQC is that mitochondrial dynamics (fission and fusion) work in tandem with mitophagy (the autophagic elimination of dysfunctional mitochondrial fragments) to maintain mitochondrial structural and functional integrity. If defective mitochondria cannot be repaired through fusion with functional organelles, they are prone to excessive fragmentation and elimination by mitophagy. It is postulated that asymmetrical fission segregates defective mitochondria, making the mitochondrial population inherently heterogeneous. These isolated small mitochondria are then redirected into a preautophagic pool (41, 42). In cells dysfunctional for mitochondrial dynamics or mitophagy, damaged mitochondria would be predicted to accumulate, and the organelles would become abnormally heterogeneous with respect to bioenergetic function and the distribution of proteins involved in MOMP (43, 44). Our screen tended to confirm this prediction.</p><!><p>A list of genes whose downregulation increased mitochondrial heterogeneity (BAX/OMI-positive phenotype) in siRNA screen</p><p>Abbreviations: ETC, electron transport chain; MAM, mitochondria-associated ER membrane; MAVS, mitochondrial antiviral-signaling protein; mTOR, mammalian target of rapamycin; UPR, unfolded protein response.</p><p>MQC functions (mitochondrial dynamics, mitophagy) of the hits are highlighted in bold. Other highlighted functions that can affect mitochondrial heterogeneity include regulation of mitochondrial respiration and apoptosis. Genes in boldface (RMDN3, ATG12, and BNIP3L) were selected for further experiments. Numbers indicate assay scores obtained in the secondary screen. Other secondary screen hits are listed in Table S2.</p><!><p>Besides well-defined components of mitophagy pathways, our screen yielded multiple candidates that could affect MQC as part of their function. For example, protein tyrosine phosphatase–interacting protein 51 (encoded by the regulator of microtubule dynamics protein 3 [RMDN3] gene) was shown to be important for mitochondria–endoplasmic reticulum (ER) tethering, and therefore, its putative role in MQC and apoptosis could be linked to multiple functions regulated by ER–mitochondrial contacts, such as ER stress–induced Ca2+ release (54), autophagosome formation, and mitochondrial fission (55, 56).</p><p>Another functional group represented by multiple hits in the screen (ATCAY, disrupted-in-schizophrenia 1 [DISC1], trafficking protein kinesin–binding 1, and spermatogenesis-associated 19 [SPATA19]) involves mitochondrial motility on microtubules, in which motor proteins such as kinesin 1 interact with certain mitochondrial membrane proteins, for example, mitochondrial Rho GTPase 1 (57, 58). We suspect that these genes appeared as hits in our screen because microtubule-based motility is important for mitochondrial fusion, including the transient fusion event known as "kiss and run" (59, 60). Since fusion is a key element of MQC, a loss of mitochondrial motility would be expected to increase the functional heterogeneity of the mitochondrial population within each cell.</p><p>Other interesting candidates include multifunctional E3 ubiquitin ligases ring finger protein 5 (RNF5) and MARCH5. Mitochondrial targets and pathways regulated by ring finger protein 5 remain to be identified. MARCH5 reportedly regulates activities of mitofusin-1 (MFN1) and dynamin-related protein 1 (DRP1) or its outer membrane receptor mitochondrial fission factor, the key mediators of mitochondrial fusion and fission, respectively (61, 62, 63). However, DRP1 and MFN themselves did not show significant effects in our screen, perhaps because of the redundancy. Among multifunctional proteins that may influence mitochondrial dynamics indirectly, leucine zipper-EF-hand containing transmembrane protein 1 (LETM1) is the mitochondrial Ca2+/H+ antiporter with pleiotropic effects on ATP production, mitophagy, cristae structure, and mitochondria morphology (64). In particular, mitochondria lacking LETM1 are prone to undergo DRP1-independent fission (65). In our screen, LETM1 knockdown also increased mitochondrial fragmentation (not shown) as well as MOMP heterogeneity. Other candidate genes in category III are highlighted in Table 1. Overall, our screen suggests that a deficiency in MQC disturbs the normal apoptotic response and, in particular, promotes heterogeneous MOMP.</p><!><p>Cell-based "minority MOMP" assay. Cells were treated with ABT199 or ABT737 for 5 h, and a caspase dye was added for the final 30 min. Cells were harvested, and 600 cells from the FITC-positive gate were sorted into 48-well plates containing the conditioned media. Grown colonies were stained with crystal violet, and colony areas were measured using the macro developed by Guzman et al. (23) (see the Experimental procedures section). MOMP, mitochondrial outer membrane permeabilization.</p><!><p>We chose the BH mimetics for this assay, instead of etoposide used in the screen, for their well-defined molecular targets in MOMP (66, 67) and no known toxic effects on other apoptotic pathways. Ichim et al. ((17); Fig. 1E) demonstrated that these mimetics induced minority MOMP in HeLa and U2OS cells. Therefore, we reasoned that these reagents would cause heterogenous MOMP as in the screen.</p><!><p>The caspase reporter dye faithfully detects activated caspases. Dot plots of WT U2OS and APAF-1 CRISPR KO cells stained with a caspase dye (CellEvent; Life Technologies) after treatment with ABT737 at 2.5, 5, or 10 μM for 5 h. APAF-1, apoptotic protease-activating factor-1.</p><p>Heterogeneous MOMP promotes cell survival after caspase activation. RMDN3, ATG12, and BNIP3L, candidate genes from the screen for the BAX/OMI-positive phenotype, were targeted by CRISPR in U2OS cells, and corresponding KO cells were subjected to the clonogenic survival assay. The percentages of surviving cells were plotted. p Values above the bar graphs show the difference in survival between the WT U2OS cells and each KO line. ATG12, autophagy-related 12; BAX, BCL2-associated X; MOMP, mitochondrial outer membrane permeabilization; RMDN3, regulator of microtubule dynamics protein 3.</p><p>Minority MOMP is enhanced by deletion of mitophagy-related or mitochondria dynamics–related genes.A, mitophagy defective penta-KO cells and their parental WT HeLa cells. B, OPA1 KO MEFs and their matched WT MEFs. C, MFN DKO MEFs and their matched WT MEFs. D, DRP1 KO MEFs and DRP1-positive WT MEFs used in the MFN set in (C). n denotes the number of independent experiments that were averaged. Error bars represent standard deviation. DKO, double KO; DRP1, dynamin-related protein 1; MEF, mouse embryonic fibroblast; MFN, mitofusin; MOMP, mitochondrial outer membrane permeabilization; OPA1, optic atrophy 1.</p><!><p>To test further whether MQC is important to mitigate minority MOMP, we used the clonogenic assay described previously to investigate the effects of deleting selected candidate genes from our screen as well as other known MQC factors. We generated CRISPR cells (U2OS) with perturbed expression of ATG12, RMDN3 (protein tyrosine phosphatase–interacting protein 51), and BNIP3L. Knockdown of each of these genes produced MOMP heterogeneity in our imaging assay (Table 1). To obtain stable KO cells, we generated CRISPR-edited cell pools and derived clonal populations from single cells (as described in the Experimental procedures section). Validation of the CRISPR-based gene-editing efficiency is shown in Fig. S2. As shown in Figure 4, we found that cells depleted of either RMDN3 or ATG12 showed significantly higher post-MOMP survival than WT cells. However, BNIP3L KO did not show an increase in minority MOMP. Reasons for this are unknown but could be related to other apoptosis-related activities of BNIP3L.</p><p>To examine further the hypothesis that compromised MQC promotes minority MOMP, we performed experiments using previously derived cells in which genes known to be directly involved in mitochondrial dynamics or mitophagy had been deleted, namely MFN1 and MFN2 double KO, OPA1 KO, DRP1 KO mouse embryonic fibroblasts (MEFs), and "penta-KO" HeLa cells that had been shown to be deficient in mitophagy because of the deletion of five mitophagy receptor genes (21). As shown in Figure 5, all these MQC-deficient cells exhibited an enhancement of cell survival in our clonogenic assay, compared with their matched controls. The effect in DRP1 KO cells was not statistically significant; in this case, a weaker effect could reflect the slowing of cell proliferation that was reported with DRP1 deletion (68). Taken together, the loss of these genes, important for MQC, promoted the proliferative survival of caspase-engaged cells, again implying an increased frequency of minority MOMP.</p><!><p>Taken together, our data allow us to conclude that the MQC system is critical for mitigating the phenomenon of minority MOMP. This result has implications for oncogenesis; even a small increase in cell viability resulting from sublethal caspase activation could potentially raise the frequency of oncogenic cell transformation (69, 70). Therefore, treatments developed to limit minority MOMP could improve the cytotoxic effects of anticancer treatments. For example, proteins that are known to inhibit mitophagy, such as seven in absentia homolog 3 (21), are potential therapeutic targets. Also, we predict, based on our observation that placing sorted caspase-engaged cells on ice eliminated their ability to survive clonogenicity (see the Experimental procedures section), that cryotherapy could prevent minority MOMP when used in combination with proapoptotic cancer therapeutics such as BH3-mimetic drugs. In conclusion, our study provides new insights into the mechanisms of aberrant apoptosis that are important for developing therapeutic strategies.</p><!><p>HeLa cells stably expressing Venus-BAX and OMI-mCherry cells (20) were constructed by the laboratory of Douglas Green (Department of Immunology, St Jude Children Hospital). U2OS cells with APAF-1 CRISPR KO were obtained from Dr Stephen Tait (Beatson Institute) (17) and MFN 1 and MFN2 double-KO cells were obtained from Dr David Chan (California Institute of Technology). Immortalized OPA1 KO and WT MEFs were obtained from American Type Culture Collection (deposited by Dr David Chan). DRP1 KO MEFs were provided by Dr Stefan Strack (University of Iowa); HeLa cells lacking five mitophagy receptors (TAX1BP1, NDP52, OPTN, NBR1, p62; Penta KO) were provided by Dr Richard Youle (the National Institutes of Health) (21). Unless indicated otherwise, cells were maintained in Dulbecco's modified Eagle's medium (DMEM; Life Technologies) containing 10% fetal bovine serum (FBS) (GeminiBio) and 100 units/ml penicillin/streptomycin at 37 °C with 5% CO2.</p><!><p>For the primary screen, gene-specific siRNA pools targeting 1318 mitochondria-annotated genes (Table S1) were cherry-picked from a Dharmacon genome-wide siRNA library (siGENOME). The list of genes was generated based on a "mitochondria/mitochondrial membranes" queries in the National Institutes of Health compiled databases (http://www.ncbi.nlm.nih.gov/gene). Pools of four individual gene-specific siRNAs and two nontargeting siRNAs were arrayed in 384-well master plates. Each plate also contained Dharmacon siGENOME TOX and siGlo Red oligonucleotides as transfection indicators. In addition, an siRNA pool targeting BAX was used as a positive control for the inhibition of apoptosis and BAX puncta formation. For reverse transfection, 4.4 μl siRNA picked from each 1 μM stock siRNA solution in the master plate was mixed with 30.6 μl prediluted Lipofectamine RNAiMAX transfection reagent (Life Technology). RNAiMAX reagent was diluted 47 times in Gibco Opti-MEM reduced serum medium. The transfection mixtures (35 μl/well) were incubated for 20 min at room temperature in a 384-well mixing plate. During the incubation period, Venus-BAX/OMI-mCherry HeLa cells were harvested by trypsinization and resuspended in antibiotic-free DMEM with 10% FBS at 30,000 cells per ml. After incubation, lipid–siRNA complexes were dispensed into triplicate tissue/culture Costar black-wall clear bottom 384-well plates (10 μl/well); siRNAs targeting the same gene were separated in the different plates. Plate handling, transfection reagent/siRNA mixing, and dispensing were performed with Hamilton Star Automated Liquid Handler contained within a Baker BioPROTECT class II biosafety cabinet. Cells were added at 40 μl (1200 cells) per well on top of the transfection mixture; final concentration of siRNA was 25 nM in the 50 μl per well volume. The plates were centrifuged at low speed (∼300g) for 1 min and transferred to a humidified CO2 incubator (37 °C). To minimize plate edge effects, the wells in two rows and columns at the plate edges were not used for transfections and contained only medium with cells. After ∼48 h, the medium was aspirated and replaced with fresh culture medium containing 400 μM etoposide (Sigma–Aldrich) and 20 μM of the caspase inhibitor Q-VD (Q-VD-OPH; SM Biochemicals LLC). Control wells were left untreated (no etoposide). By the time of etoposide treatment, most cells transfected with TOX siRNA had already shown morphological changes consistent with cell death, indicating efficient transfection. Following 24 h of etoposide treatment, the medium was removed, and cells were fixed with 0.5% glutaraldehyde as described previously (22). After two washes with PBS, cells were stained with Hoechst-33342 (Molecular Probes) diluted 1000 times in PBS. Plates were then washed twice with PBS, filled with 50 μl PBS per well, sealed, and stored at 4 °C. For the secondary screen, four individual siGENOME siRNAs per gene were obtained separately from Dharmacon to target 95 gene candidates chosen from the primary screen. Individual siRNAs were arrayed in replicate 384-well plates, and the experiments were conducted as described previously.</p><!><p>Image collection and processing were done using the Molecular Devices MetaXpress High Content Image Acquisition platform. Images of the cells were acquired in an Image Xpress Micro device at 20× magnification with the following filters: 4′,6-diamidino-2-phenylindole (5060B), for Hoechst nuclear staining; YFP (2427A), for Venus-BAX fluorescence; and Texas Red (4040B), for OMI-mCherry fluorescence (numbers indicate the Semrock part number for the filters). Images were collected from 16 sites clustered in the center of the well. In our assay development, the punctate patterns of Venus-BAX translocation and OMI-mCherry retention in mitochondria were analyzed using the granularity module in MetaXpress (version 5.1) software (Molecular Devices). Individual cells were identified based on nuclear segmentation (4′,6-diamidino-2-phenylindole channel images). Z′ factor for the BAX-positive or BAX-negative phenotype assay was calculated as 0.85 based on etoposide-treated and untreated conditions as positive and negative controls, respectively.</p><p>High-throughput analysis of primary and secondary screen data was done with a customized algorithm (created as a "journal macro" in MetaXpress) developed for automated phenotype quantification. We configured image analysis to count the percentage of cells that satisfied either of two different criteria: (1) a given cell contains Venus-BAX granules of diameters within the expected diameter range, with total intensity above a certain threshold or (2) a given cell contains both suprathreshold Venus-BAX and OMI-mCherry granules (not necessarily colocalized) within the appropriate diameter range. The thresholds were adjusted for stringency to limit false-positive scores from background fluorescence and debris. Plate-to-plate variability was minimal for BAX foci and acceptably low for cells double positive for BAX and OMI granules.</p><!><p>Confocal images of Venus-BAX/OMI-mCherry cells were acquired with a 60× oil immersion objective on an Olympus FluoView FV10i automated confocal laser scanning microscope (Olympus Scientific Solutions America Corp).</p><!><p>CRISPR experiments were performed using modified synthetic single-guide RNAs (sgRNAs) from Synthego. Target sequences for guide RNAs were selected with the Synthego CRISPR design tool. RNA oligonucleotides were reconstituted in 10 mM Tris–HCl, 1 mM EDTA, pH 8.0 buffer according to recommendations from the manufacturer. Ribonucleoprotein complexes were formed from sgRNA and recombinant Cas9 two nuclear localization signal proteins (New England Biolabs) mixed at an sgRNA to Cas9 ratio of 4.5:1. After 20 to 30 min of incubation at room temperature, assembled ribonucleoproteins were delivered into cells by electroporation using a ThermoNeon device and 10 μM tips (Thermo Fisher Scientific). For one sample (∼2 × 105 cells), 3 μl of sgRNA (from 30 mM stock solution) were mixed with 1.5 μl Cas9 protein (from 20 μM stock solution). U2OS cells were electroporated at 1230 v/10 ms/4 (pulse voltage, width, and pulse number) settings. Immediately after electroporation, cells were transferred to 12-well plate with prewarmed antibiotic-free DMEM with 10% FBS. After 2 to 3 days of incubation, a portion of the control (unedited) and CRISPR cells were harvested for genomic DNA isolation using a QIAGEN genomic DNA purification kit. The remaining cells were left in culture for further propagation and clonal selection. Genomic DNA concentrations were measured using a NanoDrop instrument (Thermo Fisher Scientific). Primer design and PCR amplification of the edited region were done according to Synthego recommendations. Sanger sequencing of PCR amplicons was performed at Genewiz or Eton Bioscience; the resulting DNA sequencing chromatograms were analyzed using the Inference of CRISPR Edits algorithm (Synthego). Inference of CRISPR Edits analysis of CRISPR-edited genomic regions typically demonstrated at least 70% editing efficiency (i.e., 70% KO cells in the pool); sgRNA sequences used are shown in Fig. S2. For clone selection, CRISPR cells were expanded for 2 to 3 additional days, harvested, and subjected to sorting into 96-well plates using BD FACSAria-3 or FACSAria-4 Fusion instruments; typically, one to four single cells were dispensed into one well. After clonal expansion, cells were analyzed for KO efficiency as described previously. For each gene of interest, two clones with verified KO were combined for use in further experiments.</p><!><p>Cells (CRISPR KO U2OS cells and MEFs [KO and matched WT]) were plated out at 3 × 105 per well containing 1 ml media in 6-well plates the day before the experiment, two wells per condition. The cells were then treated with ABT-199 or ABT-737 at 2.5, 5, or 10 μM for 5 h. In the last 30 min of incubation, a caspase reporter dye, CellEvent Caspase3/7 Green (Life Technology; R37111), was added at 30 μl per well. Cells were harvested with trypsin/EDTA and washed with medium and PBS. Finally, cells were resuspended in 300 μl of sorting buffer consisting of 1% bovine serum albumin in PBS containing 20 μM of Q-VD, where Q-VD was included to stop the caspase reaction. We noticed that, when these caspase-activated cells were left at room temperature or 4 °C for more than 30 to 40 min, their ability to survive was compromised. Therefore, care was taken to sort the cells immediately after harvesting. To minimize artifacts from sample handling delay, we operated two cell sorters (FACSAria; BD) simultaneously for WT and gene-defective cells. Also, the sorting chamber and the plate holder were kept at 37 °C during the run to prevent temperature shock to the cells. Six hundred cells within the population gated for green fluorescence were sorted in duplicate into a 48-well plate containing 300 μl of conditioned medium per well. The cells were grown for 8 to 12 days, during which an additional 500 μl of conditioned medium was added at days 4 to 5. Colonies were then stained with 6% glutaraldehyde containing 0.5% crystal violet. Nontreated cells distributed in the no-green gate were sorted as aforementioned and used as a control representing 100% growth. Because individual cell colonies tended to merge over time, we measured the total area occupied by colonies, using the ImageJ macros developed by Guzman et al. (23). The percentage of surviving caspase-engaged cells in the total population was normalized against the 100% growth control. The data are summarized from two to three independent experiments, as indicated. p Values were calculated using the mean and the standard deviation in each set of KO and WT cells with two-way ANOVA analysis using Prism 7 (GraphPad Software, Inc).</p><!><p>All representative data are contained within the article.</p><!><p>This article contains supporting information.</p><!><p>The authors declare that they have no conflicts of interest with the contents of this article.</p><!><p>Supplemental Figures S1 and S2, Tables S1 and S2</p>

PubMed Open Access

Fast Photochemical Oxidation of Proteins (FPOP) for Comparing Structures of Protein/Ligand Complexes: The Calmodulin-peptide Model System

Fast Photochemical Oxidation of Proteins (FPOP) is a mass-spectrometry-based protein footprinting method that modifies proteins on the microsecond time scale. Highly reactive \xe2\x80\xa2OH, produced by laser photolysis of hydrogen peroxide, oxidatively modifies the side chains of approximately one half the common amino acids on this time scale. Owing to the short labeling exposure, only solvent-accessible residues are sampled. Quantification of the modification extent for the apo and holo states of a protein-ligand complex should provide structurally sensitive information at the amino-acid level to compare the structures of unknown protein complexes with known ones. We report here the use of FPOP to monitor the structural changes of calmodulin in its established binding to M13 of the skeletal muscle myosin light chain kinase. We use the outcome to establish the unknown structures resulting from binding with melittin and mastoparan. The structural comparison follows from a comprehensive examination of the extent of FPOP modifications as measured by proteolysis and LC-MS/MS for each protein-ligand equilibrium. The results not only show that the three calmodulin-peptide complexes have similar structures but also reveal those regions of the protein that became more or less solvent-accessible upon binding. This approach has the potential for relatively high throughput, information-dense characterization of a series of protein-ligand complexes in biochemistry and drug discovery when the structure of one reference complex is known, as is the case for calmodulin and M13 of the skeletal muscle myosin light chain kinase, and the structures of related complexes are not,.

fast_photochemical_oxidation_of_proteins_(fpop)_for_comparing_structures_of_protein/ligand_complexes

INTRODUCTION<!>Materials<!>Protein and peptide stock solution<!>Formation of Complexes<!>FPOP Protocol<!>ESI MS Analysis with a Q-TOF Mass Spectrometer<!>Trypsin Proteolysis Protocol<!>LC-ESI-MS/MS Analysis of Protein Digest<!>LC-MS Feature Annotation<!>Calculation of Modification Extent<!>ESI MS of calmodulin submitted to FPOP<!>Information Content of LC-ESI-MS/MS proteolyzed FPOPsamples<!>Minimization of post-FPOP oxidative-modification bias<!>Changes of modification at peptide and residue level<!>Spectral-contrast-angle comparison of the modification patterns of peptides<!>Comparative structure assignment from FPOP data<!>CONCLUSION

<p>Mass spectrometry (MS) has become an important tool for studying protein structure, dynamics, interactions, and function.1 In particular, detailed characterization of protein-ligand interactions is now possible,2 a critical step toward understanding biological function. One MS-based approach is protein footprinting,3 When a protein binds with a ligand, the solvent-accessible surface areas of its residues are affected by the resulting conformational changes and direct ligand interactions. Residues with reduced solvent accessibility, such as those at the binding site in the holo state, show attenuated labeling or cleavage relative to the comparison state, and this signal change is readily resolved by MS. Thus, protein footprinting is well suited to the study of protein-ligand interaction, although its outcome is of lower resolution than that provided by X-ray crystallography and NMR spectroscopy. An advantage of MS-based footprinting over these methods is that protein-ligand structure can be interrogated in a near physiological setting (i.e., at dilute concentration in solution with appropriate buffer, ionic strength, and bimolecular milieu).</p><p>Footprinting was first reported as a cleavage-based strategy to detect protein-DNA binding specificity.4 This demonstration inspired similar methods to be developed for the analysis of proteins5; Hanai6 and co-workers first demonstrated that modification reagents could footprint proteins. Two classes of modification strategies have emerged: reversible and irreversible modification.7 The most popular reversible modification is hydrogen deuterium (H/D) exchange,8 which detects solvent accessibility of exchangeable amide hydrogens on the backbone of proteins. PLIMSTEX9 and SUPREX,10 two HD exchange protocols, were developed to investigate protein-ligand interactions and affinities. The major disadvantage of reversible modification is that this labeling cannot survive if extensive purification and handling is needed to give residue-resolved information.</p><p>Besides reversible modification, several reagents can label irreversibly certain amino acids.11 One method is hydroxyl-radical (•OH) based protein footprinting, which was first introduced by Chance and coworkers12. Hydroxyl radicals can be generated by synchrotron radiolysis of water,13 laser photolysis of hydrogen peroxide14 and other methods.15, 16 Fast Photochemical Oxidation of Proteins (FPOP) employs laser photolysis of hydrogen peroxide to label proteins on the microsecond timescale.17 The advantages of •OH compared to other irreversible labeling reagents are its high reactivity, with approximately one half the common amino acids, and its size, which is similar to that of water. The growth in application of •OH based protein footprinting was recently reviewed.18 Both reversible and irreversible strategies elicit a real incentive for further development given the powerful analytical capabilities of modern MS-based proteomics.</p><p>In this paper, we use FPOP to show the effect of peptide ligands on calmodulin conformation. Our intention is to set forth a method for comparing structures of protein/ligand complexes when one or more reference structures are available. We chose as a model calmodulin (CaM), a small acidic calcium binding protein.19 Increased Ca2+ concentration promotes specific CaM-Ca2+ binding; with the uptake of four Ca2+ ions, CaM undergoes a conformational changes to a wholly different state capable of ligand binding.20 As a Ca2+ sensor and signal transducer, Ca2+-bound CaM binds to target proteins to alter their functions, acting as part of a calcium signal transduction pathway.21,22 CaM mediates processes such as inflammation, metabolism, muscle contraction, intracellular movement, short-term and long-term memory, nerve growth and the immune response.23 One family of CaM targets is Ca2+/Calmodulin–dependent protein kinases; an example is skeletal muscle myosin light chain kinase (SK-MLCK).24 Some peptides also bind CaM to inhibit the calcium signal transduction pathway. Two such peptides are melittin (Mel) and mastoparan (Mas). Mel, the principal component of honeybee venom, is a 26-residue peptide,25 whereas Mas is a 14-residue peptide toxin from wasp.26</p><p>To date, no high resolution 3D structures of CaM-Mel or CaM-Mas have been published, although studies by NMR,27 H/D exchange,28 fluorescence,29 enzyme cleavage,30 cross-linking31 and chemical modification were reported.32, 33 Previous studies demonstrated that both Mel and Mas share a target recognition and activation mechanism with the SK-MLCK binding domain (M13 peptide) (Figure 1), suggesting their structures are similar.34,35, 36</p><p>We tested the hypothesis of structural homology among the three CaM-ligand complexes by comparing their quantitative FPOP outcomes using spectra-contrast-angle evaluation.37 This method utilizes the CaM-M13 complex as a well-understood template structure for the interpretation of FPOP signals of the CaM-Mel and CaM-Mas complexes. By this example, we demonstrate a method for the detailed characterization of protein-ligand interactions. Such an approach should be applicable to protein-ligand systems where sub libraries of ligands that are known to bind can be screened for similarity of interaction relative to a standard protein-ligand complex; the approach has higher throughput than NMR and X-ray crystallography. Application of this method for the rapid analysis of protein-ligand complexes may prove useful for studies of disease-related protein complexes, especially in drug discovery.</p><!><p>Bovine CaM (calcium free) was obtained from Ocean Biologics Co. (Edmonds, WA). Hydrogen peroxide, water, acetonitrile, formic acid, calcium chloride, phosphate buffered saline powder, EGTA (ethylene glycol-bis(2-aminoethylether)-N,N,N',N'-tetra-acetic acid), L-methionine, L-glutamine, melittin from honeybee venom (MW 2846), mastoparan from Vespula lewisii (MW 1478), catalase from bovine liver, and trypsin from porcine pancreas at the highest purity available were purchased from Sigma-Aldrich (St. Louis, MO). Skeletal muscle myosin light chain kinase peptide (M13), 1KRRWKKNFIAVSAANRFKKISSSGAL26, was purchased from AnaSpec, Inc. (Fremont, CA).</p><!><p>All proteins and peptides were received as lyophilized powder. A stock solution of CaM was prepared in water, and its concentration determined by UV-absorbance at 280 nm (ε = 2980 M−1 cm−1). All FPOP samples contained 10 μM CaM in 10 mM phosphate buffered saline (PBS buffer, 138 mM NaCl, 2.7 mM KCl, pH = 7.4) and 20 μM L-glutamine. Peptides, Mel, Mas, M13, were used without further purification.</p><!><p>Eight types of samples were prepared from the CaM stock solution and PBS buffer: Ca2+-free-CaM (100 μM EGTA), Ca2+-free-CaM-Mel (100 μM EGTA, 20 μM Mel), Ca2+-free-CaM-Mas (100 μM EGTA, 20 μM Mas), Ca2+-free-CaM-M13 (100 μM EGTA, 20 μM M13), Ca2+-bound-CaM (100 μM calcium chloride), Ca2+-bound-CaM-Mel complex (100 μM calcium chloride, 20 μM Mel), Ca2+-bound-CaM-Mas complex (100 μM calcium chloride, 20 μM Mas), and Ca2+-bound-CaM-M13 complex (100 μM calcium chloride, 20 μM M13). The protein-peptide ratio in the sample solutions was 1:2. All samples were incubated overnight at 22 °C to allow for equilibration prior FPOP labeling.</p><!><p>The protocol used here was based on a previous report38 with minor changes. A 248 nm KrF excimer laser (GAM Laser Inc, Orlando, FL) tuned to 40 mJ/pulse, was used to irradiate the sample solution. The laser was focused through a 250 mm convex lens (Edmunds Optics, Barrington, NJ) onto 150 μm i.d. fused silica tubing (Polymicro Technologies, Pheonix, AZ) located 125 mm from the lens, giving a 2.5 mm irradiation window; the fused silica polyimide coating in this region was first removed by using a propane torch. The flow rate and pulse frequency were adjusted to guarantee that1.2 x the reaction volume vacated the window between laser shots. The laser pulse frequency was controlled by an external pulse generator (B&K Precision, Yorba Linda, CA).</p><p>Hydrogen peroxide was added to each sample to a final concentration of 15 mM just prior to FPOP. Samples of 50 μL were infused through the apparatus and collected in tubes containing 10 μL of 50 nM catalase and 20 μM methionine. Residual hydrogen peroxide was decomposed by catalase treatment at room temperature for 10 min before storage at 4 °C. FPOP labeling was done in triplicate for each sample type.</p><!><p>All ESI mass spectra were acquired in the positive-ion mode on a Waters (MicroMass) Q-TOF Ultima (Manchester, U.K.) equipped with a Z-spray ESI source. The instrument setup for protein analysis was similar to that reported previously.39</p><!><p>Digestion of CaM FPOP sample was conducted according to the previously reported protocol without modification.40</p><!><p>Digested samples (1 μL) were diluted in 100 μL water with 0.1% formic acid. An aliquot of 5 μL diluted solution was loaded onto a silica capillary column with a PicoFrit™ tip (New Objective, Inc., Woburn, MA) that was custom-packed with C18 reverse phase material (Magic, 0.075 mm × 150 mm, 5 μm, 120 Å, Michrom Bioresources, Inc., Auburn, CA). The gradient was pumped with an Eksigent NanoLC-1D ultra (Eksigent Technologies, Inc. Livermore, CA) at 260 nL/min, from 2% to 60% solvent B (acetonitrile, 0.1% formic acid) over 60 min, then to 80% solvent B for 10 min, followed by a 12 min re-equilibration step by 100% solvent A (water, 0.1% formic acid). The flow was directed by a PicoView Nanospray Source (PV550, New Objective, Inc., Woburn, MA) to an LTQ Orbitrap (Thermo-Scientific, San Jose, CA) with a spray voltage of 1.8-2.0 kV, and a capillary voltage of 27 V. The LTQ Orbitrap was operated in the standard, data-dependent acquisition mode controlled by Xcalibur 2.0.7 software. Peptide mass spectra (m/z range: 350-2000) were acquired at high mass resolving power (60,000 for ions of m/z 400) in FT mode. The top six most abundant multiply charged ions with minimal signal intensity at 1000 counts were subjected to CID (collision-induced dissociation) in the linear ion trap. Precursor activation was performed with an isolation width of 2 Da and an activation time of 30 ms. The normalized collision energy was set at 35%. The automatic gain control target value was regulated at 1 × 106 for the FT analyzer and 3 × 104 for ion-trap analyzer with a maximum injection time of 1000 ms for FT and 200 ms for ion trap.</p><!><p>The LC-MS/MS data were searched for modified and unmodified CaM tryptic peptides by using Mascot 2.2.06 (Matrix Science, London, UK)41 and the NCBI database. All known •OH-side chain reaction products were added into modification database for searching as variable modifications. Modification site assignments were validated by manual inspection of production spectra.</p><p>LC-MS/MS .raw files were imported into the Rosetta Elucidator system (v3.3.0.0.220, Rosetta Biosoftware, Seattle, WA) for retention time alignment of shared LC-MS features.42 The aligned retention time and peak volume of all high resolution extracted ion chromatogram features across a 5-70 min window were output to a spreadsheet for each sample. Each product-ion spectrum was paired with its LC-MS feature by using its retention time and precursor m/z; a table of this pairing was used by the Excel macro written in our lab to annotate all Elucidator-determined LC-MS features with the Mascot assignments linked to their product-ion spectra.</p><!><p>The fraction modified (Fox1 in Eq. 1) was calculated for any specific residue as the ratio of signal intensities of each peptide (Iox1) modified at this residue to the total intensities of modified and unmodified peptide signals spanning this residue (I + Iox1 + Iox2 + …+ Ioxn). The changes in modification (R) at a site between Ca2+-free and Ca2+-bound states (Eq. 2) were calculated at the peptide level by using the modification fraction (FCa free and FCa bound) of each peptide, wherein the signal from all modifications on the peptide contribute to the numerator in equation 1. Eq. 1Fox1=ΣIox1Σt=1nIoxt+ΣI Eq. 2Rox=FCabound−FCafreeFCafree</p><p>To identify changes induced by complexation, results from CaM (no complex) were used as controls in both comparisons.</p><!><p>Beginning with CaM in the presence of Mel or Mas, we detected by ESI MS multiply charged CaM molecules (Figure S1) that have a Gaussian charge-state distribution centered at the most abundant 15+ ion. We also saw ions for Mel and Mas at lower m/z, consistent with dissociation of the complex under ESI conditions. Trace signals for oxidized Mel/Mas were observed as well. The deconvoluted mass spectrum showed that the molecular weight (MW) of CaM is 16790 Da, consistent with CaM having the expected post-translational modifications of N-terminal acetylation (+42 Da) and trimethylation of K115 (+42 Da).40</p><p>After submitting CaM to FPOP, we found evidence in the mass spectra for significant oxidative modification (Figure S1b) as mass shifts +16, 32, 48… Da. The distribution of oxidation products is Poisson-like, serving as an indication that FPOP is fast and probed only a single state of the target protein. This conclusion is consistent with recent results43 showing that the distribution of oxidation products of CaM and two other proteins sensitive to oxidative-induced unfolding are well modeled by a Poisson distribution. The conformation sampled by FPOP for these experiments is singular and invariant during labeling (i.e., the protein did not unfold on the labeling timescale).</p><p>The abundances of the +16, +32, and +48 Da modification products, as estimated from their +15 charge state intensities (Figure S1b-d), show that slightly more protein is modified in the Ca2+-bound, ligand-free state than the Ca2+-free and Ca2+-bound CaM-ligand states. The Ca2+-free and Ca2+-bound structures are markedly different. Given that the global FPOP product distributions do not substantially distinguish these two, we turned to peptide and residue-level analysis for their and CaM-peptide complexes' structural footprints.</p><!><p>Mascot processing of the LC-MS/MS data of the trypsin-proteolyzed FPOP samples showed 93% sequence coverage for the three CaM complexes. The aforementioned PTMs were also identified by using Mascot. Of the 12 peptides identified, nine had product-ion spectra for both the oxidatively modified and unmodified forms for all samples, spanning over 70% of protein sequence. Calmodulin 38-74, a large tryptic peptide, was not detected probably owing to high retention upon sample handling and chromatography. This peptide was not included among the peptides used for quantitative analysis. Small tryptic peptides from ligand peptides (Mel/Mas/M13) were also observed. Owing to the excess ligand present at equilibrium prior to FPOP, labeling occurs for both bound and free ligands; their differential analysis is consequently not possible.</p><!><p>Several factors give rise to post-FPOP oxidation on peptides, including methionine oxidation during proteolytic digestion, sample handling,44 and electrospray ionization (ESI).45 Of these modification-biasing signals, ESI-induced oxidation is the easiest to distinguish and thereby to be excluded in our analysis. The reason that we are not misled by ESI-induced oxidation is the hydrophilicity of a peptide is nearly always increased with oxidative-modification so that its reverse-phase LC retention time is earlier than that of its unmodified peptide (Figure S2). We routinely observe this in all FPOP labeling experiments (data not shown). As coelution of an oxidized peptide with its unmodified counterpart is unlikely, we attribute any oxidatively modified peptides co-eluting with the unmodified counterpart to ESI-induced oxidation.</p><p>A recent study of post-FPOP sample handling showed that proteins stored in millimolar levels of hydrogen peroxide can oxidize while frozen and at lower temperatures.46 As a precaution, we stored FPOP-treated samples at 4 °C after removing hydrogen peroxide with catalase immediately following FPOP labeling. To this mixture was added free methionine to curtail further any post-FPOP oxidation.47</p><!><p>The extents of modification can be compared for CaM in the presence of various peptides with and without bound calcium (Figure 2). Comparisons between CaM-Mel/Mas/M13 complexes (Figure 2b-d) and CaM (Figure 2a) reveal that CaM peptides fall into three groups as discussed in following paragraphs.</p><p>Peptides 76-86, 78-86 and 31-37 belong to group I; they have similar trends of modification change for both CaM and CaM-Mel/Mas/M13 complexes. Peptides 76-86 and 78-86 peptides originate from the linker region of CaM and display increased labeling in their Ca2+-bound vs. the Ca2+-free states, whereas peptide 31-37 shows decreased labeling. More importantly, their modification extents are invariant with addition of the peptide ligands (Mel/Mas/M13), indicating that the regions represented by these peptides are not directly involved in any CaM-Mel/Mas/M13 interaction.</p><p>Peptides 14-21, 22-30, 107-126, and 127-148 belong to group II; they exhibit different labeling trends for the CaM-Mel/Mas/M13 complexes than for CaM itself. The extents of oxidative modification decrease for peptides 14-21, 107-126 and 127-148 when complexed with peptide ligands (Figure 2b,c,d) compared to Ca2+-bound CaM in the absence of peptide ligands, where increases occur (Figure 2a). Reverse trends pertain to peptide 22-30. The most significant decrease in oxidative modification occurs for peptide 14-21. The differential labeling between CaM and the CaM-Mel/Mas/M13 complexes indicates that, upon forming the complexes, these regions of CaM are either buried allosterically or directly protected by the ligand.</p><p>Group III peptides 1-13 and 95-106 do not share a single labeling trend when CaM binds to peptides. The N-terminus peptide, 1-13, from CaM and the CaM-Mel bound complex (Figure 2a, b) show a decrease in oxidative modification whereas that peptide from CaM-Mas and CaM-M13 complexes (Figure 2c,d) undergo increased labeling. For peptide 95-106, the trend for CaM, CaM-Mel, and CaM-Mas (Figure 2a,b,c) showed decreased labeling whereas the CaM-M13 (Figure 2d) showed increased oxidative labeling.</p><p>Taken together, we propose that the patterns for all nine peptides compose a unique "fingerprint" of the protein-ligand complex. Although these data are not residue-resolved, they are readily accessible for comparing structures of the complexes because the modification extent of a peptide is a function of the aggregate of its modifiable residues' solvent accessibilities (Figure 3a). Examining these fingerprints should establish whether the CaM-Mel/Mas structures are similar to the known structure of CaM/M13 complex.</p><p>Turning to product-ion spectra (Figure S3), we identified 14 modified residues. Seven are at the N-terminal domain, two from the linker region, and five from the C-terminal domain. Nine have side chains that contain sulfur (M36, M76, M109, M124, M144, and M145) or aromatic rings (F16, F19, Y99), all of which are highly reactive with •OH. The extent of modification is highest for Met (with modification fractions > 0.2). Five aliphatic amino acids (L4, I9, L18, I27, and I85) undergo detectable amounts of modification. The relative modification yields between these residues agree well with those reported in previous studies of the •OH reactivity of amino acids.48 Although we can detect other low-abundance modifications, we did not include them because their assignments based on product-ion spectra were problematic. Instead, we focused on the 14 oxidized residues as a source of comparative structural information of CaM complexes (Figure 4). Peptides containing more than one highly reactive site and undergoing double modifications are considered separately.</p><p>Higher resolution provided by residue level FPOP data taken by MS/MS should afford more detailed views of CaM-peptide complexes. Those "zoom-in" views help to elucidate residue-level interactions between the protein and the peptide ligand, a view that is missing in the peptide mass spectra. In the case of peptide 14-21, its modified residues F16, L18, and F19, however, do not share the same trends shown for the full peptide. L18 and F19 show higher modification levels for CaM than for the complexes, which is the same trend seen at the peptide level. F16, however, is modified to nearly the same extent for CaM as for the complexes. Thus, one model of complexation puts L18 and F19 in direct interaction with the peptide ligand and F16 away from this interface. These data for peptide 14-21 show that analysis at the amino-acid level, made efficient because the modifications are irreversible, provides a more resolved picture of protein-ligand interaction than methodologies that only use peptide-level data. Integrating high-resolution views with a reference structure of CaM-M13 can allow the structures of unknown CaM-Mel/Mas complexes to be compared (Figure 3b).</p><!><p>At this point, one may conclude from analysis at both the peptide level and the amino-acid residue level that the modification extents in the peptide-bound and free states for the three complexes are similar. To test more rigorously the similarity, we employed a spectral-contrast-angle (θ) analysis of the nine peptide modification trends. This spectra-contrast-angle approach can validate the comparison process by providing a confident value (θ) related to similarity. It was used previously to compare the product-ion spectra of oligodeoxynucleotide isomers.37 The changes of oxidative modification of the various peptides from CaM can be treated in similar way. Although the sequence coverage of the tryptic peptides is ~70%, this statistical approach still is suitable for making comparisons and does not require full coverage. For comparing the CaM-peptide complex with the peptide-free sample, the set of peptide signals can be represented as a basis vector in an N-dimensional space (ai). In this analysis, each signal is the change in modification level for the calcium-bound vs. calcium-free state. A comparison vector (bi) is comprised of the same signal data from a different sample replicate or CaM-peptide mixture. The spectral-contrast-angle (Eq. 3) of these vectors provides a single parameter reflecting their similarity; as the similarity increases, θ ➔0.</p><p>The first four entries in Table 1 are the average θ for the three pair-wise comparisons of replicates from the independent triplicates of a CaM-peptide or absent-peptide FPOP experiments. These entries provide an expectation level for near identity of θ < 35. The θ determined from the average of each CaM vs CaM-peptide comparison is significantly larger (Table 1, 5th entry). This implies a significant change in structure with peptide binding, as is expected, and the small standard deviation conveys the overall similarity of this change among the three peptide complexes. The pair-wise comparisons between each CaM-peptide complex give an average θ (Table 1 last entry), which is similar to the smallest angle from replicate experiments. This similarity further supports the conclusion that the two unknown CaM-peptide complexes have similar structures and that they resemble the known CaM-M13 structure. We propose that this statistical approach is a useful means of comparing protein/ligand structures for a series of ligands when one structure is known and can be used as a reference. Eq. 3cosθ=ΣtatbtΣtai2Σtbi2</p><!><p>The approach of utilizing the "fingerprint" of changes in extent of modification at the peptide and residue levels is more empirical than but complements that of Chance and coworkers49, 50 who showed that the combination of •OH surface mapping data with computational modeling provides important protein structural data. Our approach not only allows conclusions about the similarities of the protein-ligand structures but also permits establishment of some structural details. Proteins or peptides (including Mel, Mas) that bind to CaM are likely to bind to both N-term and C-term hydrophobic clefts by forming a long alpha helix in a manner similar to that of SK-MLCK M13 peptide.34 In addition to the NMR structure (PDB ID: 2BBM) of CaM-M13 complex, the structures of Ca2+-free CaM from NMR (PDB ID: 1CFD) and Ca2+-bound CaM from x-ray crystal structure (PDB ID: 1CLL) are also known (Figure 1). The FPOP results for the established structure of the CaM-M13 complex provide a basis for a more detailed structural analysis. For example, we see that L18, F19, M109, M124, M144 and M145 from both N and C terminal domains of the three complexes are protected against •OH reaction upon complex formation with the various ligand peptides. All six are hydrophobic and are located at the interface between CaM and M13, as seen in the NMR structure (Figure 3b). M109, M124, M144 and M145 are located inside the hydrophobic cleft of the C-terminus and make contact with the M13 hydrophobic side chains. For the CaM-M13 structure, L18 and F19, which are at the hydrophobic-cleft center of the N-terminus, become less solvent-accessible when the Ca2+-bound CaM binds with M13. On the other hand, F16 is not part of the hydrophobic cleft in the CaM-M13 complex, but it is hydrophobic and close in proximity to L18 and F19 (Figure 3b).</p><p>The L4 and I9 side chains have relatively similar solvent accessibilities in CaM and the CaM-M13 complex, suggesting that these residues share similar conformations in CaM and in the peptide complexes. Two other interesting sites are I27 and Y99, which are located in the EF hands. Y99 is in a short anti parallel beta sheet between adjacent calcium binding loops. The aromatic side chain of Y99 points outside the molecule. Owing to its high solvent accessibility and its inherent high •OH reactivity, this tyrosine residue is particularly sensitive to slight changes in structure. This suggests the peptide-binding-induced deprotection when M13 binds and induced protection when Mel and Mas bind are small changes although significantly different. I27, however, is on the small loop connecting two alpha helices of the first EF hand motif. Its side chain is inside the hydrophobic cleft in all three of the high resolution CaM structures (Ca2+-free CaM, Ca2+-bound CaM and CaM-M13 complex), consistent with the low reactivity and lack of difference between Ca2+-loaded CaM and the peptide complexes.</p><!><p>Comparison of FPOP data taken for a protein complex of known structure to the FPOP-induced modifications of unknowns by using spectral-contrast-angle provides a way to test whether a protein has similar or different structures when it binds to various ligands. Specifically, a statistical analysis of the modification extents peptide regions of CaM extricated by tryptic digestion shows that two peptide ligands (i.e., Mel and Mas) bind similarly in the presence of Ca2+ as does a third ligand (M13) that forms a known structure with CaM. We chose not to increase the certainty of the comparison by using more proteases in the digestion of FPOP samples to obtain more data for comparison, a suggestion made by one of the reviewers. We wished to show that this comparative approach is effective for distinguishing the structural differences even when the sequence coverage isn't 100%, as is the case here. An advantage of FPOP coupled with MS is that one can examine the modification extents at the amino-acid level to identify those residues that are involved in the conformational changes induced by ligand binding. Those residues that show differential modification levels in the structure of CaM-M13 complex are those that are expected on the basis of the high resolution 3D structure from NMR.</p><p>The MS-based FPOP method has the ability to probe structures on a comparative basis for small amounts of protein (pmoles) in relatively short times (fractions of a week) compared to NMR or X-ray crystallography. These simple comparisons may have application in screening many protein/ligand interactions including those in drug development. The FPOP data serve as a "fingerprint" for the ligand interactions. Combining the "fingerprints" with those from H/D exchange, specific chemical modification (e.g., acetylation of Lys), and cross linking can give detailed information about protein-ligand interactions with reasonable throughput.</p>

PubMed Author Manuscript

Chemoenzymatic Synthesis and Lectin Array Characterization of a Class of N-Glycan Clusters

N-glycans are major components of many glycoproteins. These sugar moieties are frequently involved in important physiological and disease processes via their interactions with a variety of glycan-binding proteins (GBP). Clustering effect is an important feature in many glycan-lectin interactions. We describe in this paper a chemoenzymatic synthesis of novel N-glycan clusters using a tandem endoglycosidase-catalyzed transglycosylation. It was found that the internal \xce\xb2-1,2-linked GlcNAc moieties in the N-glycan core, once exposed in the non-reducing terminus, was able to serve as acceptors for transglycosylation catalyzed by Endo-A and EndoM-N175A. This efficient chemoenzymatic method allows a quick extension of the sugar chains to form a class of glycan clusters in which sugar residues are all connected by native glycosidic linkages found in natural N-glycans. In addition, a discriminative enzymatic reaction at the two GlcNAc residues could be fulfilled to afford novel hybrid clusters. Lectin microarray studies revealed unusual properties in glyco-epitope expression by this panel of structurally well-defined synthetic N-glycans. These new compounds are likely valuable for functional glycomics studies to unveil new functions of both glycans and carbohydrate-binding proteins.

chemoenzymatic_synthesis_and_lectin_array_characterization_of_a_class_of_n-glycan_clusters

Introduction<!>Enzymatic transglycosylation onto the \xce\xb2-1,2-linked GlcNAc residues in the Asn-linked GlcNAc2Man3GlcNAc2 core<!>Further extension of the sugar chains in the glycan clusters by tandem enzymatic transglycosylation<!>Lectin array characterization of the synthetic N-glycan clusters<!>Conclusion<!>Materials and methods<!>Synthesis of 1-(6-biotinamido-hexanoylamino)-1-deoxy-2-acetamido-2-deoxy-\xce\xb2-D-glucopyranose (4)<!>Synthesis of biotinylated complex type N-glycan (6)<!>Synthesis of the biotinylated complex type N-glycan (7)<!>Synthesis of the N-glycan cluster (9)<!>Synthesis of glycan clusters 10a and 10b<!>Synthesis of N-glycan cluster 10c from the mono-transglycosylated compound 10a<!>Transglycosylation with the Man9GlcNAc oxazoline (9) by EndoM-N175A<!>Synthesis of N-glycan cluster 13<!>Synthesis of N-glycan cluster 14<!>Synthesis of N-glycan cluster 15<!>Lectin microarray analysis<!><!>A 45-lectin microarray platform for probing N-glycan clusters<!>Glyco-epitope profiles of novel synthetic N-glycans detected by lectin arrays

<p>N-linked glycosylation is a predominant covalent modification of proteins in eukaryotes. It is well documented that N-glycans of glycoproteins are involved in many important biological events including protein folding, ER-associate protein degradation, cell differentiation, cell adhesion, host-pathogen interaction, cancer metastasis, and autoimmunity 1. Glycoproteins are often characterized by their structural micro-heterogeneity in terms of the components of the attached glycans. It becomes clear that distinct N-glycans can confer significantly different effects on the structure and function of a given glycoprotein. This was exemplified by recent discoveries that the attachment of subtly different N-glycans at the conserved N-glycosylation site of the Fc domain could result in dramatically different impacts on the ADCC function of monoclonal antibodies and on the anti-inflammatory activity of intravenous immunoglobulin (IVIG) 2. On the other hand, not only does the fine structure of the glycans provide the basis for molecular recognition, but the way of their presentation, e.g., the monovalent vs. multivalent format, is also of paramount importance in governing the specificity and strength in carbohydrate-protein interactions 3. Recent advances in glycan and glycan-binding protein microarray technology have offered exciting new opportunities to unveil the mysteries of glycans in various cellular processes and disease states 4. Nevertheless, functional glycomics studies are still limited by the availability of structurally well-defined N-glycans and related glycoconjugates, which are difficult to obtain in homogeneous forms from natural source 5.</p><p>We and others have previously demonstrated that a class of endo-β-N-acetylglucosaminidases (ENGases), including the Endo-A from Arthrobactor protophormiae and the Endo-M from Mucor hiemali, were able to transfer an oligosaccharide en bloc from either natural N-glycans or synthetic sugar oxazolines to a GlcNAc-containing moiety to form a new β-1,4-glycosidic linkage in a regio- and stereo-specific manner, leading to the synthesis of complex oligosaccharides, N-glycopeptides and N- glycoproteins 6. Endo-A is specific for high-mannose or hybrid type N-glycans and has been applied for the synthesis of high-mannose type oligosaccharides and N-glycopeptides 7, whereas Endo-M is able to work on three major types (high-mannose, hybrid, and complex type) of N-glycans and has been particularly useful for synthesizing complex type N-glycopeptides 8. In particular, the recent findings that synthetic oligosaccharide oxazolines (the mimics of the oxazolinium ion intermediate of the enzymatic reaction) could be used for Endo-A catalyzed transglycosylation have significantly expanded the scope of the chemoenzymatic method for glycopeptide and glycoprotein synthesis 9–11. It was found that the highly activated sugar oxazolines corresponding to the truncated or modified N-glycans could serve as substrates for the Endo-A catalyzed transglycosylation but the ground-state products formed were refractory to enzymatic hydrolysis due to the slight structural modification. Moreover, the discovery of several ENGase-based glycosynthases, including EndoM-N175A; EndoA-N171A, and EndoA-E173Q that could promote transglycosylation with sugar oxazolines of natural N-glycans but lack the ability to hydrolyze the product, has enabled the synthesis of homogeneous glycoproteins carrying full-size natural N-glycans 12–14. Subsequent studies indicate that Endo-A and Endo-M could accommodate diverse structures in the aglycon portions of GlcNAc- or Glc-tagged acceptors for transglycosylation, permitting the introduction of N-glycans into a wide range of natural products, unnatural peptides, and even polysaccharides 15, 16. The relaxed substrate specificity of Endo-A and Endo-M, together with the powerful transglycosylation potential of the glycosynthase mutants, prompted us to examine the possibility to glue multiple N-glycans to the complex-type GlcNAc2Man3GlcNAc2-Asn core through tandem enzymatic transglycosylation. We report in this paper the chemoenzymatic synthesis of a class of novel N-glycan clusters containing multiple N-glycan cores, in which all monosaccharide residues are connected via defined native glycosidic bonds found in natural N-glycans. Lectin microarray analysis of the synthetic N-glycan clusters has revealed unusual lectin-carbohydrate recognition patterns that were not observed before.</p><!><p>Construction of an array of N-glycan clusters started with the preparation of a biotinylated bi-antennary complex type N-glycan (Scheme 1). Coupling of glycosylamine 2 and an activated biotin tag (3) gave the GlcNAc-LC-biotin (4), which was then used as an acceptor for enzymatic transglycosylation. We have previously reported the synthesis of an asialoglycan oxazoline (5) and have shown that it was a good substrate of the glycosynthase mutant EndoM-N175A for glycoprotein synthesis 13. Thus the reaction of oxazoline (5) and acceptor (4) (donor/acceptor, 3:1) under the catalysis of EndoM-N175A gave the biotinylated N-glycan product (6) in 72% yield. The two terminal galactose residues were then selectively removed by treatment with β-1,4-galactosidase from Bacteroides fragilis to provide GlcNAc2Man3GlcNAc2-LC-biotin (7), in which the two internal GlcNAc residues were exposed at the non-reducing terminus. Enzymatic extension of the sugar chain from the two GlcNAc residues would enable the synthesis of novel glycan structures. The advantage of introducing a biotin tag in the aglycon portion of the glycan primer (7) was apparent: it would facilitate the detection of the synthetic N-glycans bound in a glycan-binding protein microarray platform, or the biotinylated glycans could be directly immobilized on a streptavidin plates to provide a glycan array platform.</p><p>To test whether the two terminal β-1,2-linked GlcNAc residues of 7 are able to serve as acceptors for the next round enzymatic transglycosylation, we first examined the Endo-A catalyzed transglycosylation of 7 using Man3GlcNAc oxazoline (8) 10 as the donor substrate (Scheme 2). It was found that when oxazoline 8 and GlcNAc2Man3GlcNAc2-LC-biotin (7) (donor/acceptor, 6:1) were incubated with Endo-A in a phosphate buffer at 30°C, the transglycosylation reaction proceeded very efficiently and, after 2h, almost all the starting material 7 was converted to the doubly glycosylated product 9 (as revealed by HPLC). The product was readily isolated by HPLC in 88% yield. MALDI-TOF MS analysis showed a single m/z species at 3055.12, which matches well with the theoretical data of 9 that carries three Man3GlcNAc2 cores (calculated, M = 3032.14; found M+Na = 3055.12). The newly formed glycosidic bonds in the Man3GlcNAc2 cores were assumed to be in the natural β-1,4-glycosidic linkage between the two GlcNAc residues, on the basis of all previously reported stereo-specificity of Endo-A catalyzed transglycosylation 10, 11, 15. To further substantiate this assumption, compound 9 was treated with wild type Endo-M that specifically hydrolyzes the β-1,4-glycosidic bond in natural N-glycan. It was found that treatment of 9 with Endo-M gave GlcNAc-LC-biotin (1 equiv.), GlcNAc2Man3GlcNAc (1 equiv.) and Man3GlcNAc (2 equiv.) (data not shown), suggesting that the three Man3GlcNAc2 cores all are having the natural GlcNAc-β-1,4-GlcNAc linkage in the core. The successful simultaneous enzymatic transfer of two Man3GlcNAc moieties to the GlcNAc2Man3GlcNAc2-LC-biotin indicates that the two terminal β-1,2-linked GlcNAc residues, albeit in a rigid and crowded environment on the N-glycan core, are accessible to Endo-A catalyzed glycosylation.</p><p>Next, we examined the glycosylation of 7 by the glycosynthase mutant EndoM-N175A, using a large complex type glycan oxazoline (5) as the donor substrate. Interestingly, when a limited amount of oxazoline donor substrate was used, a selective glycosylation on the two terminal GlcNAc residues was observed. Thus, incubation of oxazoline 5 and acceptor 7 (donor/acceptor, 3:1) with EndoM-N175A at 30°C for 4h gave three products: the two mono-glycosylated products (10a and 10b) and the doubly transglycosylated product 10c in 25, 5, and 8% yields, respectively (Scheme 2). The GlcNAc linked to the mannose on the α-1,6-arm was much more favorable for transglycosylation than the GlcNAc linked to the mannose at the α-1,3-arm, with a selectivity of 5:1 (25% vs. 5%). Although both GlcNAc are β-1,2-linked to the mannose moiety, the regio-selectivity might result from the difference in steric hindrance, as the GlcNAc at the 6-arm would be spatially less hindered than the GlcNAc located on the α-1,3-arm. Nevertheless, double glycosylation could be achieved by further chemoenzymatic transglycosylation with excess amount of donor substrate for a longer incubation time. This was exemplified by further glycosylation of the purified mono-glycosylated glycan 10a and oxazoline 5 (3.5 molar equiv.) under the catalysis of EndoM-N175A, giving the doubly-glycosylated product (10c) in 50% yield. When Man9GlcNAc oxazoline (11) was used as the donor substrate, an even higher regio-selectivity in transglycosylation was observed. Incubation of 11 and 7 (3:1) with EndoA-N175A for 4h gave 27% of 12a, 3% of 12b, and trace amount of the doubly glycosylated product (12a:12b = 9:1) (Scheme 2). The higher selectivity might be explained by the bulky structure of tri-antennary glycan (11) in comparison with the bi-antennary complex glycan 5. It should be mentioned that the use of wild type Endo-M for the reaction between acceptor 7 and the oxazoline 5 or 11 failed to provide the transglycosylation products due to the enzymatic hydrolysis of the acceptor 7 as well as the products that were formed (data not shown).</p><p>It should be mentioned that in the synthesis of large complex N-glycan clusters, most of the transglycosylation reactions were carried out on a relatively small scale. Nevertheless, the specific enzymatic transglycosylation usually gave a very clear HPLC profile allowing easy quantification and isolation of the reaction products by HPLC. To characterize the transglycosylation products, the new compounds were purified by HPLC and were subjected to MALDI-TOF MS analysis, which confirmed that the products 10a and 10b were mono-transglycosylated product and the 10c was the doubly glycosylated product (see Experimental Section). To discriminate the two mono-transglycosylated products 10a and 10b, specific enzymatic transformation coupled with MS analysis was applied (Figure S1, Supporting Information). Treatment of 10a and 10b with β-N-acetylglucosaminidase from Xanthomonas manihotis resulted in the removal of the remaining terminal GlcNAc, which exposed the α-1,3-Man residue in 10a or the α-1,6-Man residue in 10b. When the two intermediates, which had the same molecular mass, were then treated further with an α-1,2/α-1,3-mannosidase from Xanthomonas manihotis, only the exposed α-1,3-man residue in 10a would be removed, leading to the formation of a species with a loss of 162 Da in molecular mass, while the α-1,6-Man residue exposed in 10b would not be hydrolyzed. Thus, MALDI-TOF MS analysis of the resulting products led to an unambiguous assignment of 10a and 10b as the clusters with the additional glycan attached at the α-1,6-arm and the α-1,3-arm positions, respectively. Discrimination between 12a and 12b was achieved by similar enzymatic transformations of 12a with β-N-acetylglucosaminidase and α-1,2/α-1,3-mannosidase coupled with MS analysis (Figure S2, Supporting Information).</p><p>The regio-selectivity observed in the EndoM-N175A catalyzed transglycosylation also provided an opportunity to introduce distinct N-glycans at the α-1,6- and α-1,3-arms of the core, permitting the synthesis of novel hybrid N-glycan clusters. Thus, after the first selective transglycosylation of 7 with oxazoline (5) to produce compound 10a, another distinct glycan, Man3GlcNAc, was successfully introduced to the remaining terminal GlcNAc at the α-1,3-arm by Endo-A to give the hybrid N-glycan cluster 13 in essentially quantitative yield (Scheme 3). This result represents a remarkable example on the potential of the enzymatic transglycosylation for expanding the diversity of the glycan cluster structures.</p><!><p>For those glycan clusters that contain terminal LacNAc moieties such as compounds 10a, 10b, 10c, and 13, further sugar chain extensions could be readily achieved by unmasking the terminal GlcNAc residues followed by repeat enzymatic transglycosylation. For example, treatment of 10c with β-1,4-galactosidase from Bacteroides fragilis resulted in selective removal of the four terminal galactose residues to provide compound 14, in which the four internal GlcNAc residues were exposed. Simultaneous glycosylation of these GlcNAc residues in 14 was achieved in a single step by its Endo-A catalyzed transglycosylation with an excess amount of Man3GlcNAc-oxazoline (8) (4 equiv. per terminal GlcNAc) to give N-glycan cluster 15 in essentially quantitative yield (Scheme 4). The HPLC and MALDI-TOF MS profiles of the purified N-glycan cluster 15 were shown in Figure 1. The m/z species observed in the MS spectrum is in good agreement with the calculated molecular mass (observed, m/z = 6625.98; calculated, [M + Na] = 6625.41). Glycan cluster 15 consists of four terminal Man3GlcNAc2 cores and three internal Man3GlcNAc2 cores. It would be worthwhile to emphasize that all the sugar residues in these enzymatically synthetic N-glycan clusters are connected to each other by the defined natural glycosidic bonds found in natural N-glycans, which usually demonstrate confined rotations of the chains than those linked through flexible spacers. Thus the N-glycan clusters described in this work represent a new class of N-glycan clusters that are different from other synthetic oligosaccharide clusters in which mono- or oligosaccharides are linked with various spacers.</p><!><p>The availability of a class of structurally well-defined synthetic N-glycan clusters has now provided an opportunity to investigate how the unusual configuration of multiple N-glycan cores in a molecule would contribute to their recognition with various glycan-binding proteins such as lectins. For a preliminary study, a lectin microarray consisting of 45 lectins was used (Figure 2a), in which the lectins were immobilized on a glass slide as previously reported 17. All the biotinylated N-glycan cluster were assayed at a concentration of 100 nM, and the bound glycans were detected by a fluorescence(Cy3)-labeled streptavidin. Figure 2b demonstrated a typical fluorescent imaging of the microarray profiles for glycan clusters 9 and 15. The lectin-binding profiles of all the synthetic N-glycan clusters were summarized in Figure 3. For each compound, its responses to the respective lectins were quantitatively determined and expressed as mean fluorescence intensity (MFI) of triplicate detections.</p><p>For the simple biotinylated N-glycans, 6 and 7, no apparent lectin interactions were observed under the screening conditions (at 100 nM glycan concentration) except for a weak binding of the GlcNAc-terminated glycan 7 to ConA, a lectin that is specific for terminal GlcNAc, Man, and Glc residues. The known sugar-binding specificity of the 45 lectins used in the present microarray analysis was listed in Figure S3 (Supporting Information). The cluster 9, which contains two terminal Man3GlcNAc2 cores, showed significant signals on spots for lectins Calsepa and ConA. These results are consistent with the known specificity of these two lectins, which recognize terminal mannose/GlcNAc residues and particularly high-mannose glycan 18. Glycans 10a and 10b picked up three lectins, Calsepa, ConA, and RCA120. RCA120 is a lectin specific for terminal β-Gal residues 19. The strong interactions between lectin Calsepa and 10a or 10b is an interesting observation. Calsepa is a mannose-binding-type Jacalin-realated lectin (mJRL) with binding preference to small high-mannose N-glycan (Man2–6). It is also reported that Calsepa interacts with sialo/asialo complex-type N-glyans with terminal Gal or GlcNAc. Interestingly bisecting GlcNAc dramatically enhances binding affinity of complex-type N-glycans with Calsepa 18. The strong interaction between Calsepa and 10a or 10b suggests that both the terminal Man3GlcNAc core and the remaining terminal GlcNAc in 10a and 10b might bind to two distinct binding sites in the lectin simultaneously in a concerted manner, thus dramatically enhancing its affinity to the lectin. It should be noted that Calsepa did not show detectable affinity to the complex-type N-glycans 6 and 7 at 100 nM, despite the presence of terminal GalNAc and GlcNAc in the glycans.</p><p>The lectin microarray also revealed interesting properties for glycan cluster 10c, which bears four terminal Galβ1-4GlcNAc moieties. First, it showed significantly enhanced affinity to the LacNAc-specific galectins RCA120 and ECA 20 as compared with the glycans 6, 10a, and 10b. The dramatic enhancement of affinity could be explained by the clustering effect of the Galβ1-4GlcNAc ligands in 10c. Secondly, the Calsepa binding signals were also enhanced. More surprisingly, glycan cluster 10c also demonstrated significant affinity to ConA, although it does not contain terminal GlcNAc/Man residues. These results suggest that the appropriate clustering arrangement of the internal Man3GlcNAc2 cores in 10c may generate novel new binding sites or interfaces for lectin ConA that hitherto favors terminal mannose/GlcNAc residues. As to the Man9GlcNAc2 containing cluster 12a, it was recognized by 5 lectins in the microarray. In addition to its expected affinity to the high-mannose recognizing lectins, including Calsepa, ConA, and HHL, it also picked up two unexpected lectins, ABA and UDA. Lectin ABA has a known specificity for terminal Galβ1-3GalNAc structure and a very weak affinity to terminal GlcNAc residue on complex-type N-glycans (Kd at 100 μM level) 21. The reason why 12a exhibits such a high-affinity to lectin ABA is yet to be elucidated, but perhaps both the Man9 structure and the terminal GlcNAc at the other arm are involved in a simultaneous interaction with two binding sites in the lectin. The glycan clusters 10a and 10b, which contain a terminal GlcNAc but carry a complex glycan on the other arm, were not recognized by ABA. Lectin UDA (Urtica dioica agglutinin) is known to be specific for chitin oligomers (N,N′,N″-triacetylchitotriose and higher chitin oligosaccharides) 22, 23. The significant interaction between UDA and compound 12a may suggest that the three internal GlcNAc residues may form a discontinuous glyco-epitope that would fit to the chitotriose-binding domain in the lectin 23. Alternatively, the Man9GlcNAc2 moiety itself may also contributes directly to the binding, as compound 10a, which contains a similar three internal GlcNAc motif but carries a complex N-glycan at the 6-arm, clearly did not bind to UDA. The glycan cluster 13, which possesses both terminal Galβ1-4GalNAc and high-mannose moieties, demonstrated the expected binding specificity for lectins Calpesa, ConA, and RCA120. But again, it also showed significant binding affinity to UDA. The glycan cluster 14, which possesses multiple terminal GlcNAcβ1,2-Man structures, demonstrated strong affinity to lectins Calsepa, ConA, and ABA, and moderate binding capacity to HHL and UDA. While its strong interactions with Calsepa, ConA and ABA are expected because of the clustering effect of multiple terminal GlcNAc residues, the interaction between glycan 14 and lectin HHL that has a known specificity for mannose is a new observation. Compound 15, which possesses 4 terminal Man3GlcNAc2 cores and 3 internal Man3GlcNAc2 cores, is another novel N-glycan cluster that demonstrated unusual lectin recognition properties. A comparison of this glycan cluster with the simpler glycan cluster 9 (2 terminal Man3GlcNAc2 cores and 1 internal Man3GlcNAc core), revealed a clear cluster effects in lectin recognition (Figure 3). Under the same assay conditions, glycan 9 recognized only two lectins, Calsepa and ConA that are specific for terminal mannose residues and high-mannose type N-glycans. However, glycan cluster 15 was recognized by 7 different lectins in the microarray. In addition to Calsepa and ConA, compound 15 showed high affinity to the other three high-mannose specific lectins, GNA 24, HHL, and NPA 25. The fact that glycan cluster 9 did not show sufficient affinity to these three lectins (negative at 100 nM) strongly suggest that these three lectins recognize particularly a high-density cluster of high-mannose type glycans, as demonstrated by the glycan cluster 15. Moreover, glycan cluster 15 also picked up two chitin oligosaccharide-specific lectins, UDA 23 and LEL 26 that recognizes chitin oligosaccharides with 3 or more GlcNAc moieties. A plausible explanation is that the compact cluster of the multiple GlcNAcβ1-4GlcNAc moieties in 15 forms a novel discontinuous epitope that mimics the chitin oligosaccharide to provide a high-affinity structure to interact with the binding domain in the lectin. Notably, among the synthetic glycan clusters, only compound 15 showed significant binding to lectin LEL that is known to be specific for N-acetyl-chitooligosaccharide moieties (Figure 3). These experimental data suggest that the unusual configuration of the N-glycan cores in the defined N-glycan clusters create novel lectin recognition motifs or glyco-epitopes that may implicate special roles of unusual N-glycans in a biological system. Recently, Dennis and co-workers has reported that the number and degree of branching (e.g., tetra-antennary vs. bi-antennary) of complex type N-glycans in cell-surface receptors are critical factors to regulate cell proliferation and differentiation, due to the distinct affinities of the differentially branched N-glycans to galectins 27. On the other hand, the highly branched and compactly packed N-glycans clusters described in this work, which show unusually high-affinity to some specific lectins, may be used as specific inhibitors to decipher the functional roles of N-glycans and lectins in a given biological process.</p><!><p>A facile synthesis of a class of novel N-glycan clusters was achieved via tandem chemoenzymatic tranglycosylation. The unusual configurations and highly branched packing of the subunit N-glycans in the clusters create new structural motifs, demonstrating novel carbohydrate-lectin recognition patterns. These synthetic N-glycan clusters should be valuable for functional glycomics studies to unveil new functions for both glycans and carbohydrate-binding proteins.</p><!><p>1-β-Azido-GlcNAc (1) was prepared by reported method 28. Succinimidyl-6-(biotinamido)hexanoate (NHS-LC-biotin, 3) was purchased from Pierce Biotechnology, Inc. CT-GlcNAc oxazoline (5) 13, Man3GlcNAc oxazoline 10, and Man9GlcNAc oxazoline 12 were synthesized following our previously reported procedures. The β-1,4-galactosidase, β-N-acetylglucosaminidase and α-1,2/1,3-mannosidase were purchased from New England Biolabs, Inc. Endo-A was overproduced following the literature 29. EndoM-N175A was overproduced according to the previously reported method 12. The activity unit of Endo-A was defined as following: 1 unit of Endo-A is the amount of enzyme required to hydrolyze 1 μmol Man9GlcNAc2Asn (substrate concentration, 10 mM) in one minute at 30 °C in a phosphate buffer (50 mM, pH 6.5). The unit of glycosynthase activity of EndoM-N175A was defined as following: 1 unit of EndoM-N175A is the amount of enzyme required to transfer 1μmol Man9GlcNAc oxazoline to GlcNAc-pNP at 1:2 donor/acceptor ratio in one minute. All the other reagents were purchased from Sigma/Aldrich and used as received.</p><p>Analytical RP-HPLC was performed on a Waters 626 HPLC instrument with a Symmetry300™ C18 column (5.0 μm, 4.6 × 250 mm) at 40°C. The column was eluted at a flow rate of 1.0 mL/min using a linear gradient of 0–90% MeCN containing 0.1% trifluoroacetic acid (TFA) for 30 min. The yields were calculated based on the HPLC quantification of both starting materials and products [absorbance (abs) at 214 nm], using the following formula: yield (%) = [product abs/(starting material Abs + product abs)] × 100. Preparative HPLC was performed on a Waters 600 HPLC instrument with a preparative C18 column (Symmetry300™, 7.0 μm, 19 × 300 mm). The column was eluted with a suitable gradient of water/acetonitrile containing 0.1% TFA. NMR spectra were measured with JEOL ECX 400 MHz and/or Inova 500 MHz NMR spectrometers. All chemical shifts were assigned in ppm. The ESI-MS Spectra were measured on a Waters Micromass ZQ-4000 single quadruple mass spectrometer. MALDI-TOF MS measurement was performed on an Autoflex II MALDI-TOF mass spectrometer (Bruker Daltonics). The instrument was calibrated by using ProteoMass Peptide MALDI-MS calibration kit (MSCAL2, Sigma/Aldrich). The matrix used for glycans is 2,5-dihydroxybenzoic acid (DHB) (10.0 mg/mL in 50% acetonitrile containing 0.1% trifluoroacetic acid). The measuring conditions in detail: 337 nm nitrogen laser with 100 μJ output; laser frequency 50.0 Hz; laser power 30–45%; linear mode; positive polarity; detection range 1000–10000; pulsed ion extraction: 70ns; high voltage: on; realtime smooth: high; shots: 500–2000.</p><!><p>1-β-Azido-GlcNAc (1) (10 mg, 38 μmol) was dissolved in MeOH (1.0 mL) containing 5% palladium on carbon (5.0 mg). The mixture was hydrogenated at r.t. under atmospheric pressure overnight. The residue was filtered via a Celite pad and the filtrate was concentrated. The obtained crude syrup of 1-β-amino-GlcNAc (2) was directly used without additional purification. A solution of the crude compound 2 and NHS-LC-biotin (3) (20 mg, 44 μmol) in a mixed solvent of phosphate buffer (50 mM, pH 7.5, 1.0 mL), MeCN (1.0 mL) and DMSO (1.0 mL) was shaken at r.t. overnight. The residue was subject to preparative HPLC for purification. The fractions containing product were combined and lyophilized to give GlcNAc-LC-biotin (4) as white powder (17 mg, 79% for two steps). 1H NMR (D2O, 400 MHz): δ 4.93 (d, 1H, J = 9.6 Hz, H-1 of GlcNAc), 4.47 (dd, 1H, J = 4.8, 8.0 Hz, H-7 of biotin), 4.29 (dd, 1H, J = 4.4, 8.0 Hz, H-8 of biotin), 3.74 (dd, 1H, J = 2.0, 12.4 Hz, H-6a of GlcNAc), 3.67 (t, 1H, J = 10.0 Hz, H-2 of GlcNAc), 3.61 (dd, 1H, J = 4.6, 12.4 Hz, H-6b of GlcNAc), 3.47 (t, 1H, J = 10.0 Hz, H-3 of GlcNAc), 3.38 (m, 2H, H-4 and H-5 of GlcNAc), 3.21 (m, 1H, H-4 of biotin), 3.03 (t, 2H, J = 6.8 Hz, CH2NHCO), 2.87 (dd, 1H, J = 4.8, 13.0 Hz, H-6a of biotin), 2.63 (d, 1H, J = 13.0 Hz, H-6b of biotin), 2.14 (m, 4H, CH2CONH), 1.86 (s, 3H, Ac of GlcNAc), 1.61-1.34 (m, 8H, CH2), 1.26 (m, 2H, CH2), 1.16 (m, 2H, CH2); 13C NMR (D2O, 100 MHz) δ 177.7, 176.6, 174.5, 165.3, 78.3, 77.6, 74.1, 71.2, 69.4, 62.1, 60.5, 60.2, 55.4, 54.3, 39.7, 39.0, 35.7, 35.5, 28.0, 27.8, 27.7, 25.5, 25.2., 25.0, 22.0; analytical HPLC: tR = 12.2 min; ESI-MS: calculated for C24H41N5O8S, M = 559.27 Da, found, 560.67 [M + H]+.</p><!><p>A solution of GlcNAc-LC-biotin (4) (1.0 mg, 1.8 μmol) and CT-GlcNAc oxazoline (5) (7.5 mg, 5.3 μmol) in a phosphate buffer (50 mM, pH 7.5, 100 μL) was incubated with EndoM-N175A (125 mU) at 30°C for 8h. The residue was subject to preparative HPLC for purification. The fractions containing product were combined and lyophilized to give CT-GlcNAc2-LC-biotin (6) as white powder (2.5 mg, 72%). 1H NMR (D2O, 400 MHz): δ 4.97 (s, 1H, H-1 of Man4), 4.92 (d, 1H, J = 9.6 Hz, H-1 of GlcNAc1), 4.78 (s, 1H, H-1 of Man4′), 4.62 (s, 1H, H-1 of Man3), 4.46 (m, 4H, H-1 of GlcNAc2, H-1 of GlcNAc5, H-1 of GlcNAc5′, H-7 of biotin), 4.33 (dd, 2H, J = 2.4, 7.8 Hz, H-1 of Gal6 and Gal6′), 4.28 (dd, 1H, J = 4.6, 7.8 Hz, H-8 of biotin), 4.11 (m, 1H), 4.06 (m, 1H), 3.97 (m, 1H), 3.19 (quintet, 1H, J = 4.8 Hz, H-4 of biotin), 3.03 (t, 2H, J = 6.8 Hz, CH2NHCO), 2.86 (dd, 1H, J = 4.8, 12.8 Hz, H-6a of biotin), 2.65 (d, 1H, J = 12.8 Hz, H-6b of biotin), 2.12 (m, 4H, CH2CONH), 1.94 (s, 3H, Ac), 1.91 (s, 3H, Ac), 1.90 (s, 3H, Ac), 1.85 (s, 3H, Ac), 1.61-1.34 (m, 8H, CH2), 1.28 (m, 2H, CH2), 1.16 (m, 2H, CH2); 13C NMR (D2O, 100 MHz) δ 177.7, 176.6, 174.7, 174.6, 174.5, 165.3 (7 × CO), 102.9 (C-1 of Gal6 and Gal6′), 101.5 (C-1 of GlcNAc2), 100.4 (C-1 of Man3), 99.7 (C-1 of Man4), 99,6 (C-1 of GlcNAc5 and GlcNAc5′), 97.0 (C-1 of Man4′), 78.6 (C-1 of GlcNAc1), 62.8 (C-8 of biotin), 60.5 (C-7 of biotin), 55.8 (C-4 of biotin), 39.7 (C-6 of biotin), 39.0 (2 × CH2NHCO), 35.7, 35.5 (2 × CH2CONH), 27.9, 27.8, 27.6, 25.4, 25.2., 24.8 (6 × CH2), 22.3, 22.2, 22.0 (4 × Ac); analytical HPLC: tR = 11.2 min; MALDI-TOF-MS: calculated for C78H130N8O48S, M = 1978.77 Da, found, 2002.68 [M + Na]+.</p><!><p>A solution of CT-GlcNAc2-LC-Biotin (6) (2.5 mg, 1.3 μmol) in a phosphate buffer (50 mM, pH 5.5, 300 μL) was incubated with β-1,4-galactosidase (100 U) from Bacteroides fragilis (New England Biolabs) at 37 °C for 24h. The reaction was monitored by analytical HPLC until complete removal of terminal galactose. The residue was subject to preparative HPLC. The fractions containing product were combined and lyophilized to give GlcNAc2Man3GlcNAc2-LC-biotin (7) as white powder (2.1 mg, quantitative yield). 1H NMR (D2O, 400 MHz): δ 4.98 (s, 1H, H-1 of Man4), 4.93 (d, 1H, J = 9.6 Hz, H-1 of GlcNAc1), 4.78 (s, 1H, H-1 of Man4′), 4.64 (s, 1H, H-1 of Man3), 4.48 (m, 2H, H-1 of GlcNAc2, H-7 of biotin), 4.43 (d, 2H, J = 8.6 Hz, H-1 of GlcNAc5 and GlcNAc5′), 4.29 (dd, 1H, J = 4.4, 8.0 Hz, H-8 of biotin), 4.12 (m, 1H), 4.05 (m, 1H), 3.97 (m, 1H), 3.20 (quintet, 1H, J = 4.8 Hz, H-4 of biotin), 3.03 (t, 2H, J = 6.8 Hz, CH2NHCO), 2.87 (dd, 1H, J = 4.8, 13.2 Hz, H-6a of biotin), 2.63 (d, 1H, J = 12.8 Hz, H-6b of biotin), 2.13 (m, 4H, CH2CONH), 1.94 (s, 3H, Ac), 1.91 (s, 6H, 2 × Ac), 1.86 (s, 3H, Ac), 1.62-1.34 (m, 8H, CH2), 1.27 (m, 2H, CH2), 1.15 (m, 2H, CH2); 13C NMR (D2O, 100 MHz) δ 177.7, 176.6, 174.7, 174.6, 174.5, 165.3 (7 × CO), 101.3 (C-1 of GlcNAc2), 100.4 (C-1 of Man3), 99.6 (C-1 of Man4, C-1 of GlcNAc5 and GlcNAc5′), 97.0 (C-1 of Man4′), 80.4, 79.5, 78.6 (C-1 of GlcNAc1), 78.2, 76.4, 76.2, 75.8, 74.4, 74.3, 73.5, 73.3, 73.2, 72.8, 72.7, 72.0, 70.2, 69.9, 69.5, 69.4, 67.3, 65.8, 65.7, 62.1 (C-8 of biotin), 61.7, 61.6, 60.6 (C-7 of biotin), 60.2, 59.8, 55.4, 55.3 (C-4 of biotin), 39.7 (C-6 of biotin), 39.0 (2 × CH2NHCO), 35.7, 35.5 (2 × CH2CONH), 27.9, 27.8, 27.6, 25.4, 25.2., 24.8 (6 × CH2), 22.3, 22.2, 22.0 (4 × Ac); analytical HPLC: tR = 11.4 min; MALDI-TOF-MS: calculated for C66H110N8O38S, M = 1654.66 Da, found, 1678.07 [M + Na]+.</p><!><p>A solution of GlcNAc2Man3GlcNAc2-LC-Biotin (7) (0.20 mg, 0.12 μmol) and Man3GlcNAc oxazoline (8) (0.50 mg, 0.72 μmol) in phosphate buffer (50 mM, pH 7.5, 15 μL) was incubated with Endo-A (2.0 mU) at 30 °C for 1h. The reaction was monitored by analytical HPLC until the completion of the reaction. The residue was subject to preparative HPLC purification. The fractions containing product were combined and lyophilized to give Man3GlcNAc2Man(1,6)- [Man3GlcNAc2Man(1,3)]-ManGlcNAc2-LC-biotin (9) as a white powder (88% yield). Analytical HPLC: tR = 11.1 min; MALDI-TOF-MS: calculated for C118H196N10O78S, M = 3033.14 Da, found, 3055.12 [M + Na]+.</p><!><p>A solution of GlcNAc2Man3GlcNAc2-LC-biotin (7) (0.50 mg, 0.30 μmol) and CT-GlcNAc oxazoline (5) (1.3 mg, 0.91 μmol) in a phosphate buffer (50 mM, pH 7.5, 20 μL) was incubated with EndoM-N175A (30 mU) at 30°C. The reaction was monitored by analytical HPLC. After 4h, HPLC showed the formation of three new products, which were readily purified by HPLC to give the 6-arm glycosylated product CT-GlcNAc2Man(1,6)-[GlcNAcMan(1,3)]-LC-biotin (10a) in 25% yield; the 3-arm glycosylated product CT-GlcNAc2Man(1,3)-[GlcNAcMan(1,6)]-LC-biotin (10b) in 5% yield, and the doubly glycosylated product 10c in 8% yield.</p><p>CT-GlcNAc2Man(1,6)-[GlcNAcMan(1,3)]-ManGlcNAc2-LC-biotin (10a): white powder, analytical HPLC: tR = 10.8 min; MALDI-TOF-MS: calculated for C120H199N11O78S, M = 3074.17 Da, found, 3097.78 [M + Na]+.</p><p>CT-GlcNAc2Man(1,3)-[GlcNAcMan(1,6)]-ManGlcNAc2-LC-biotin (10b): white powder, analytical HPLC: tR = 11.1 min; MALDI-TOF-MS: calculated for C120H199N11O78S, M = 3074.17 Da, found, 3098.36 [M + Na]+.</p><p>CT-GlcNAc2Man(1,6)-[CT-GlcNAc2Man(1,3)]-ManGlcNAc2-LC-biotin (10c): white powder, analytical HPLC: tR = 10.2 min; MALDI-TOF-MS: calculated for C174H288N14O118S, M = 4496.24 Da, found, 4519.69 [M + Na]+.</p><!><p>A solution of 10a (0.30 mg, 0.10 μmol) and CT-GlcNAc oxazoline (8) (0.50 mg, 0.35 μmol) in a phosphate buffer (50 mM, pH 7.5, 5.0 μL) was incubated with EndoM-N175A (30 mU) at 30 °C for 8h. The reaction mixture was subject to HPLC purification to give the doubly glycosylated product 10c (50% yield).</p><!><p>A solution of GlcNAc2Man3GlcNAc2-LC-biotin (7) (0.20 mg, 0.12 μmol) and Man9GlcNAc oxazoline (11) (0.60 mg, 0.36 μmol) in a phosphate buffer (50 mM, pH 7.5, 15 μL) was incubated with EndoM-N175A (10 mU) at 30°C. The reaction was monitored by analytical HPLC. After 4h, HPLC indicated the formation of the 6-arm glycosylated product 12a in 27% yield and the 3-arm glycosylated product 12b in 3% yield. The products were purified by HPLC. The doubly glycosylated product was formed in only trace amount (as indicated by MS analysis), which was not isolated for further analysis.</p><p>Man9GlcNAc2Man(1,6)-[GlcNAcMan(1,3)]-ManGlcNAc2-LC-biotin (12a): white powder, analytical HPLC: tR = 10.7 min; MALDI-TOF-MS: calculated for C128H213N9O88S, M = 3318.13 Da, found, 3341.78 [M + Na]+.</p><p>Man9GlcNAc2Man(1,3)-[GlcNAcMan(1,3)]-ManGlcNAc2-LC-biotin (12b): white powder, analytical HPLC: tR = 11.0 min; MALDI-TOF-MS: calculated for C128H213N9O88S, M = 3318.13 Da, found, 3341.57 [M + Na]+.</p><!><p>A solution of 10a (0.10 mg, 33 nmol) and Man3GlcNAc oxazoline (5) (0.10 mg, 0.14 μmol) in a phosphate buffer (50 mM, pH 7.5, 5.0 μL) was incubated with Endo-A (1.0 mU) at 30°C for 2h. Analytical HPLC monitoring indicated the formation of the hybrid glycosylated product 13 in a quantitative yield. The product was purified by HPLC to give CT-GlcNAc2Man(1,6)-[Man3GlcNAc2Man(1,3)]-ManGlcNAc2-LC-biotin (13). Analytical HPLC: tR = 10.4 min; MALDI-TOF-MS: calculated for C146H242N12O98S, M = 3763.40 Da, found, 3786.87 [M + Na]+.</p><!><p>A solution of 10c (0.15 mg, 33 nmol) in a phosphate buffer (50 mM, pH 5.5, 30 μL) was incubated with β-1,4-galactosidase (10 U) at 37°C for overnight. The reaction was monitored by analytical HPLC until the complete removal of all the terminal galactose moieties. The residue was subject to preparative HPLC purification. The fractions containing product were combined and lyophilized to give GlcNAc2Man3GlcNAc2Man(1,6)-[GlcNAc2Man3GlcNAc2Man(1,3)]-ManGlcNAc2-LC-biotin (14) (quantitative yield) as a white powder. Analytical HPLC: tR = 10.7 min; MALDI-TOF-MS: calculated for C150H248N14O98S, M = 3847.67 Da, found, 3870.97 [M + Na]+.</p><!><p>A solution of 14 (0.10 mg, 26 nmol) and Man3GlcNAc oxazoline (5) (0.30 mg, 0.44 μmol) in a phosphate buffer (50 mM, pH 7.5, 15 μL) was incubated with Endo-A (2.0 mU) at 30°C. HPLC monitoring indicated the complete transglycosylation after 2h incubation. The reaction mixture was subject to preparative HPLC. The fractions containing product were combined and lyophilized to give the N-glycan cluster (15) (quantitative yield). Analytical HPLC: tR = 9.6 min; MALDI-TOF-MS: calculated for C254H420N18O178S, M = 6602.41 Da, found, 6625.98 [M + Na]+.</p><!><p>Lectins were immobilized on glass slides according to the previously reported method 17. The lectin microarray profilings of the glycan clusters were carried out following the literature method 17, with some modifications. Briefly, the biotinylated N-glycan clusters were applied on lectin arrays at a concentration of 100 nM and incubated at 4°C for 5h. After washing to remove the unbound N-glycans, the lectin arrays were incubated with a fluorescence (Cy3)-labeled Streptavidin at a concentration of 1 μg/ml at r.t. for 30 min. The array-captured N-glycans were then visualized and quantified by scanning the arrays with a specialized GlycoStation (GP Biosciences Ltd, Yokohama, Japan). For each compound, the binding responses with the lectin arrays were quantitatively determined and expressed as mean fluorescence intensity (MFI) of triplicate detections. In all the assays, GlcNAc-LC-biotin (4) was used as the control. Overlay plots were produced using JMP-Genomics 4.0 software package (SAS Institute in Cary, North Carolina).</p><!><p>HPLC and MALDI-TOF MS profiles of N-glycan cluster 15.</p><!><p>a) lectin printing pattern; b) Fluorescent imaging of the lectin-recognition profiles of glycan clusters 9 and 15.</p><!><p>overlay plots were produced using JMP-Genomics 4.0 software package (SAS Institute in Cary, North Carolina). Each plot was an overlay plot of GlcNAc-LC-biotin (4) that served as a control in the assay (marked with a symbol "X") and the respective N-glycan cluster (marked with a symbol "□").</p>

PubMed Author Manuscript

A Biphilic Phosphetane Catalyzes N-N Bond Forming Cadogan Heterocyclization via PIII/PV=O Redox Cycling

A small ring phosphacycle (1,2,2,3,4,4-hexamethylphosphetane) is found to catalyze deoxygenative N-N bond-forming Cadogan heterocyclization of o-nitrobenzaldimines, o-nitroazobenzenes and related substrates in the presence of hydrosilane terminal reductant. The reaction provides a chemoselective catalytic synthesis of 2H-indazoles, 2H-benzotriazoles, and related fused heterocyclic systems with good functional group compatibility. On the basis of both stoichiometric and catalytic mechanistic experiments, the reaction is proposed to proceed via catalytic PIII/PV=O cycling, where DFT modelling suggests a turnover limiting (3+1) cheletropic addition between the phosphetane catalyst and nitroarene substrate. Strain/distortion analysis of the (3+1) transition structure highlights the controlling role of frontier orbital effects underpinning the catalytic performance of the phosphetane.

a_biphilic_phosphetane_catalyzes_n-n_bond_forming_cadogan_heterocyclization_via_piii/pv=o_redox_cycl

<p>Tricoordinate phosphorus reagents are versatile O-atom acceptors,1 and the conversion PIII→PV=O drives numerous valuable synthetic transformations.2–4 Despite great utility, the inefficient generation of stoichiometric phosphine oxide waste inherent to these methods is regarded as a key limitation. In recent years, though, several new catalytic methods involving cycling through phosphine oxide intermediates have emerged,5 both in redox-neutral6–7, and redox-driven modes.8–10 As part of a program aimed at developing designer main group compounds as broadly useful biphilic11 catalysts in organic synthesis, 12 we recently reported the use of phosphetanes as O-atom transfer catalysts operating via PIII/PV=O redox cycling.12c In this Communication, we advance this biphilic catalysis concept by describing a catalytic N-N bond forming heterocyclization enabled by phosphetane-catalyzed reductive O-atom transfer. Beyond broadening the repertoire of phosphine oxide redox-catalyzed organic transformations, this study documents a dominant electronic basis for the superior performance of phosphetane catalysts that provides a framework for future targeted design of geometrically deformed main group compounds as biphilic catalysts.</p><p>Cadogan13 and Sundberg14 have shown that phosphine-mediated exhaustive deoxygenation of o-functionalized nitrobenzene derivatives drives phenylnitrene-like azacyclizations. Most commonly, these transformations employ superstoichiometric amounts of phosphorus-based reagents at elevated temperatures; the standard Cadogan protocol calls for reaction in neat refluxing triethylphosphite (bp 156 °C), although milder conditions have been reported in some cases.15 We questioned whether inherently nontrigonal biphilic phosphines, which colocalize donor and acceptor behavior at P, might facilitate O-atom transfer from the nitro substrate under milder conditions and in a manner conducive to PIII/PV=O redox cycling (Figure 1).</p><p>Starting from our reported conditions for deoxygenative carbonyl functionalization as a point of departure, treatment of 1 with substiochiometric quantities of aminophosphetane P-oxide 3·[O] (15 mol%) and phenylsilane (2 equiv.) in PhMe at 100 °C was indeed found to afford indazole 2, albeit in modest yield (Table 1, entry 1). Further optimization studies converged on the simple 1,2,2,3,4,4-hexamethylphosphetane P-oxide 5·[O] as the optimal catalyst for this transformation (see SI for details); in the event, phosphetane P-oxide 5·[O] is found to catalyze deoxygenative heterocyclization delivering product 2 in 83% GC yield within 3 h at 100 °C (entry 3). Neither phospholane-based (6·[O] and 7·[O], entries 4, 5) nor acyclic (8·[O], entry 6) phosphorus precatalysts exhibit similar catalytic reactivity under otherwise identical conditions. Control experiments (entries 7 and 8) confirm that no reaction is observed in the absence of either 5·[O] or terminal reductant PhSiH3; the transformation is demonstrably under catalyst control. Consistent with the notion of PIII/PV=O redox cycling, the use of tricoordinate (σ3-P) precatalyst 1,2,2,3,4,4-hexamethylphosphetane 5 affords indazole 2 with qualitatively similar results as 5·[O] (entry 9). That said, the ease with which phosphine oxide 5·[O] –an air stable solid– is both handled on laboratory scale and reduced in situ to 516 recommended its use as precatalyst in subsequent studies.</p><p>Results of our studies into the scope of this catalytic transformation are collected in Table 2. With respect to N-substitution, aliphatic substituents (Table 2A) are well tolerated including 1° (9), 2° (10), and 3° (11) moieties, and basic nitrogen functionality may be incorporated without incident (compare 12 and 13). Likewise, aromatic substrates of diverse substitution (Table 2B) are suitable substrates; halogenated (17–19), electron-rich (20, 21), electron-poor (23, 25), unsaturated (24) and sterically demanding (26) N-aryl substituted substrates all undergo smooth catalytic indazole formation. Polyheterocyclic products (30, 31) may be similarly prepared. Free hydroxyl moieties do not inhibit deoxygenative heterocyclization (22, 32), although such substrates may undergo in situ silylation by the PhSiH3 reductant; desilylative workup ensures recovery of the free OH group. Challenging substrates for the current method include, unsurprisingly, those with multiple nitro moieties (Table 2C, 33); evidently the catalyst does not selectively recognize the o-imino nitro moiety. However, a range of other reducible functionality including nitriles (23), amides (29), and esters (36) are all carried through the deoxygenative cyclization reaction without incident.</p><p>The reaction is similarly amenable to non-benzaldimine substrates, permitting the synthesis of diverse heterocyclic structures through catalytic N-N bond formation (Table 2D,E). For instance, deoxygenative heterocyclization of 2-(2-pyridyl)nitrobenzene delivers the fused polyheterocyclic pyrido[1,2-b]indazole (39) in near quantitative yield under standard catalytic conditions within 4 h. Relatedly, indazolodihydroimidazoles (40), -tetrahydropyrimidines (41), and -dihydrooxazoles (42) are accessible under standard conditions. Stereochemistry adjacent to the reaction centers is retained upon cyclization (42). Beyond indazole synthesis, the preparation of N-aryl 2H-benzotriazoles (43–45) is achieved by deoxygenative heterocyclization of the corresponding substituted o-nitroazobenzene.</p><p>In situ spectral monitoring of catalytic reactions provides information regarding both the mechanistic course of the transformation and speciation of active phosphorus compounds during catalysis. 1H NMR spectra (400 MHz, toluene-d8, 100 °C) of a standard catalytic transformation (1 equiv. of 1, 15 mol% of 5·[O], 2 equiv. of PhSiH3, 1M in C7D8) show the appearance of product indazole 2 over the course of ca. 1.5 h at the expense of starting imine substrate 1; no long-lived intermediates are observed by 1H NMR spectroscopy. Under identical conditions, 31P NMR spectra show the rapid conversion of 5·[O] (δ 53.2 ppm) to an epimeric mixture of the corresponding phosphetane 5 (δ 32.4 ppm (major, anti); δ 18.9 ppm (minor, syn)), which persists in solution until the reaction is complete (see SI). From these data, we infer that 5 represents the catalytic resting state, with the initial deoxygenation of substrate 1 as the turnover limiting step.17 Regrettably, direct kinetic observation of reaction steps following initial deoxygenation is therefore precluded; presumably, though, the nitroso derivative formed by deoxygenation of 1 is an obligate intermediate that proceeds on to observed product in a series of fast following steps.18</p><p>Under stoichiometric pseudo-first order conditions with excess phosphine reagent, we find that consumption of substrate 1 is markedly faster with phosphetane 5 than nBu3P 8 (krel ≈ 8, see SI Figure S2); phosphetane 5 is evidently superior to acyclic 8 for Cadogan cyclization. In an effort to understand the origin of this rate difference, we undertook DFT modelling studies (Figure 2). Direct O-atom transfer from nitro-methane to phosphetane 5′ and trimethylphosphine 8′ (via TS15′ and TS18′, respectively) is found to be high in energy. Instead, a stepwise mechanism proceeding through pentacoordinate azadioxaphosphetanes I5′ and I8′ is found at lower energies. The rate-controlling transition structure along this pathway (i.e. TS25′ and TS28′) involves a (3+1) cheletropic addition19,20 of MeNO2 to 5′ and 8′, respectively. Subsequent decomposition of I5′ and I8′ via retro-(2+2) fragmentation (TS25′ and TS28′) then follows with low barrier. This (3+1)/retro-(2+2) pathway is mechanistically analogous and orbitally equivalent to known reactivity of phosphines(-ites) with O3.20</p><p>To understand further the superiority of the phosphetane with respect to deoxygenation, the (3+1) transition structures TS25′ and TS28′ were analyzed within the distortion/interaction model21 (Figure 3). Despite the presence of the small ring in 5′, the destabilizing distortion energy (ΔEd‡) within transition structures TS25′ and TS28′ is nearly identical (39.2 vs. 38.8 kcal/mol, respectively), being driven to an over-whelming extent by pyramidalization of the nitro substrate, not by geometric reorganization about phosphorus. By consequence, the lower overall energy of TS25′ arises from a significantly larger stabilizing interaction energy (ΔEi‡) for phosphetane 5′ (−17.5 kcal/mol) than Me3P 8′ (−9.7 kcal/mol). The inference from this analysis is that the differential performance of the small-ring phosphetane and trimethylphosphine – both of which compositionally are simple trialkylphosphines – is driven primarily by electronic (orbital) as opposed to enthalpic (ring strain) effects.</p><p>The difference in interaction energies for the (3+1) addition transition structures can be rationalized by frontier orbital inspection. Figure 4 depicts the Kohn-Sham HOMO and LUMO of each reactant distorted to the transition state structure. Whereas both phosphetane and trimethylphosphine exhibit HOMOs (nonbonding lone pair) of nearly identical energy (−6.95 eV), LUMO(5′) resides ca. 0.8 eV lower in energy than LUMO(8′). In effect, the geometric constraint enforced by the 4-membered ring of the phosphetane 5′ results in a marked lowering of the LUMO that preferentially permits synergistic interactions of HOMO(5′) → LUMO(MeNO2) and HOMO(MeNO2) → LUMO(5′) in a [π4s+ω2s] fashion.22 We note that the low frontier orbital energy gap of phosphetanes has previously been invoked by Chesnut and Quin to rationalize nonmonotonic chemical shift anisotropy trends in phosphacycloalkanes.23 Relatedly, Hudson24 and Westheimer25 have noted the extent to which ring constraint increases electrophilic character at phosphorus. The importance of orbital effects in reactions of strained ring systems has been noted by Hoz.26</p><p>In summary, we have found that a readily accessible27 phosphetane is a suitable catalyst for the Cadogan indazole synthesis. The method provides a simple phosphacatalytic approach to a valuable N–N bond forming mode that has previously been accomplished via (super)stoichiometric reagent chemistry,13,28 transition metal catalysis,29 or alternative high energy azide substrates.30 Whereas previous studies involving PIII/PV=O redox cycling have focused primarily on ring strain arguments underpinning catalytic turnover of phosphine oxides by silane reductants, the results above suggest a dominant electronic component to the overall biphilic function of the phosphetane catalyst. Work continues in an effort to establish further the biphilic reactivity of phosphetanes as generalized platforms for catalytic reductive O-atom transfer.</p>

PubMed Author Manuscript

Synthesis and cytotoxicity studies of steroid-functionalized titanocenes as potential anticancer drugs: sex steroids as potential vectors for titanocenes

Six titanocenyls functionalized with steroidal esters have been synthesized and characterized by infrared, 1H, and 13C NMR spectroscopy and elemental analysis. Among those steroids, dehydroepiandrosterone, trans-androsterone, and androsterone are androgens and pregnenolone is a progesterone precursor. Clionasterol is a natural steroid compound. These steroid-functionalized titanocenyls were tested by 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide assay for in vitro cytotoxicity for MCF-7 breast cancer and HT-29 colon cancer cells. All complexes exhibited more cytotoxicity than titanocene dichloride. The titanocenyls containing androgen and progesterone derivatives as pendant groups had higher antiproliferative activities than those with cholesterol steroid compounds. Of particular significance is titanocenyl\xe2\x80\x93dehydroepiandrosterone complex, which is 2 orders of magnitude more cytotoxic than titanocene dichloride and also shows much more sensitivity and selectivity for the MCF-7 cell line.

synthesis_and_cytotoxicity_studies_of_steroid-functionalized_titanocenes_as_potential_anticancer_dru

Introduction<!>Synthesis and characterization<!>Structural density functional theory discussion<!>Cytotoxicity studies<!>Conclusion<!>General procedure<!>Synthesis and characterization<!>Complex 1<!>Complex 2<!>Complex 3<!>Complex 4<!>Complex 5<!>Complex 6<!>Cytotoxicity assay<!>

<p>The discovery of the anticancer activity of cis-diaminodichloroplatinum(II) (cisplatin) marked the beginning of a rich field of transition-metal-based medicinal chemistry. Despite its remarkable success, however, cisplatin has several disadvantages, including notable toxic side effects such as nephrotoxicity, neurotoxicity, and emesis [1]. In addition, some tumors exhibit inherent resistance to cisplatin, whereas others develop resistance after initial treatment, thereby limiting its clinical usefulness [1]. These particular disadvantages have driven the search for new compounds exhibiting high cytotoxic activity along with reduced side effects and no cross-resistance.</p><p>In 1979, Köpf and Köpf-Maier opened a new chapter in medicinal chemistry with the discovery of the first metal-locene-based organometallic anticancer agent, titanocene dichloride, Cp2TiCl2. The fact that it possesses significant antitumor properties in cancer cell lines that are insensitive to cisplatin as well as lower toxic effects motivated the scientific community to continue investigating this species [2–9]. Unfortunately, the efficacy of Cp2TiCl2 in phase II clinical trials in patients with some cancer types [10, 11] was too low for the use of Cp2TiCl2 to be pursued. To overcome the aforementioned efficacy problem and to increase the cytotoxic activity of titanocene dichloride derivatives, many titanocenes complexes have been synthesized [12–37].</p><p>Our research group has reported the structural modification of titanocene by either replacing chloride with hydrophilic or biologically important ligands or functionalizing the cyclopentadienyl (Cp) ring to study structure– activity relationships [38–42]. The structural modification of titanocene dichloride to enhance its anticancer properties requires a careful selection of the functional group to be appended to the Cp ring or replacement of the ancillary ligands by more active ones. Recently we reported the synthesis, structure, and biological activity of amide-functionalized titanocenyl complexes on the colon cancer cell line HT-29 [43]. We were able to achieve cytotoxic activities (IC50 values) on HT-29 in the micromolar range, which are 2 orders of magnitude greater than for titanocene dichloride. Motivated by these optimistic results, we pursued the synthesis of six steroid-functionalized titanocene complexes.</p><p>Sex steroids as pendant groups will provide more selectivity to specific cancer cell lines, potentially resulting in target-specific anticancer drugs for hormone-dependent cancers. The present study was undertaken to investigate the synthesis and biological activity of six novel steroidfunctionalized titanocenes. To determine the biomedical potential of the steroid complexes, cytotoxicity testing was performed using 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay, which led to improved cytotoxicities for MCF-7 and HT-29 cell lines, when compared with titanocene dichloride.</p><p>The selection of the steroids as pendant groups was based on the following criteria. Androsterone (6) and its epimer trans-androsterone (5; epiandrosterone) are natural androgens produced by the enzyme 5α-reductase from the adrenal hormone dehydroepiandrosterone (4; DHEA) [44]. Androsterone is a hormone with weak androgenic activity and can also be produced by the metabolism of testosterone, whereas trans-androsterone is a more active species and has been used as a steroid hormone drug owing to its inhibitory effects on breast cancer [44]. DHEA is a natural steroid hormone produced by the adrenal glands, gonads, adipose tissue, and the brain. It is a precursor of androstenedione, testosterone, estradiol, and estrone and it is the most abundant hormone in the human body. It can act on the androgen receptor directly. DHEA has inhibitory effects on breast cancer [44, 45]. Chloresterol and dehydrochloresterol are precursors of steroid hormones and clionasterol (1) is a natural product analog of chloresterol. Pregnenolone (2) is a steroid hormone involved in the steroidogenesis of progesterone, mineralocorticoids, glucocorticoids, androgens, and estrogens [46]. It can be enzymatically converted to progesterone. Thus, all the sex steroid derivatives selected as pendant groups are biologically important compounds.</p><!><p>The synthesis of the cholesterol-functionalized titanocene dichlorides was performed using a published procedure [47]. This involves the activation of the coordinated Cp ring by an acyl chloride, starting from fulvene. The reaction of the titanocene acyl chloride with the corresponding substituted steroids affords the steroid-functionalized titanocenyl dichloride (Fig. 1). The syntheses of the nine titanocenyl amide complexes presented has been reported previously by our group [43]. In this work, six novel steroid-functionalized titanocene complexes were synthesized to complete the series of titanocenyls with a wide variety of different steroid rings (see "Materials and methods") (Fig. 2). These complexes are air- and moisture-stable and are soluble in dimethyl sulfoxide (DMSO) and water. They were characterized by IR and NMR spectroscopies and elemental analysis.</p><p>In the NMR spectra, these titanocenyl steroids showed –CH–O– signals between 4.52 and 4.96 ppm, corroborating the presence of the 3-αH or 3-βH protons, and the ester carbonyl carbons appeared at about 170 ppm. The IR spectral data of these species revealed characteristic absorption peak carbonyl bands at about 1,728–1,743 cm−1 corresponding to the ester groups. The ring signals of the substituted Cp ligand, in both 1H and 13C NMR spectra, are shifted downfield compared with the ring signals of the unsubstituted Cp ligands.</p><!><p>Despite our efforts to crystallize these six titanocenes, no crystal structures were obtained. To overcome this problem, density functional theory calculations were carried out for titanocenyls 2, 5, and 6 at the B3LYP level using the 6–31G** basis set. Selected bond lengths of the optimized structures of 2, 5, and 6 are listed in Table 1. The calculated structures are presented in Fig. 3.</p><p>The calculated structures for 2, 5, and 6 showed two η5-Cp rings and two chlorides in a distorted tetrahedral geometry around the Ti(IV) center. In addition, the steroid pendant groups are positioned away from the chlorides. The Cl–Ti–Cl bond angle for the three complexes is approximately 97.5°. The Cp–Ti–Cp angle for 2 is 123.68°, for 5 is 123.63°, and for 6 is 123.53°. Whereas the optimum calculated structures of 2 and 5 showed the pendant groups (steroids) to be pointing upward with repect to the Cp ligand plane, avoiding possible steric interactions with the chlorides, the Cp steroid of 6 was in the opposite direction, most likely as a result of its original steroid configuration on carbon 3. However, the steroid is rotated about 78° away from the chlorides, avoiding steric interactions.</p><p>The Ti–C(Cp) bond distances for the substituted Cp ring vary from 2.34898 to 2.6078 Å (2), from 2.3487 to 2.6136 Å (5), and from 2.3753 to 2.6141 Å (6), the longest Ti–C bond distance corresponding to the substituted carbon atom of the Cp ligand. The average Ti–C(unsubstituted Cp) bond distances for 2, 5, and 6 are very similar and fall within a very narrow range, 2.4403, 2.4398, and 2.4411 Å, respectively.</p><p>Although no X-ray structure has been reported for the titanocenyl–steroid complex, recent density functional theory calculations on alkenyl-, boryl-, and dimethylamino-substituted titanocenes have shown the pendant group on the Cp ring to be pointing upward, away from the chlorides, analogous to 2 and 5 [29, 36, 48]. Also, the Ti–C(Cp) bond distances and Ti–Cl bond distances and angles are almost identical to those for our calculated structures [12, 29, 36]. However, 6 showed the opposite direction of the pendant group and it was twisted about 78° to avoid steric interactions. Although the structure of 6 is completely different from that predicted for 2, 5, and other substituted titanocenes [12, 29, 36], the fact it is more stable (energetically) than when the pendant group (steroid) is pointing upward with respect to the Cp ligand plane strongly suggests that this is the preferred conformation for this steroid.</p><!><p>The cytotoxicities of these complexes for breast cancer MCF-7 and colon cancer HT-29 cell lines were determined using a slightly modified MTT assay at 72 h [48, 49]. As a reference, the cytotoxic activity of Cp2TiCl2 was tested at 72 h and IC50 values of 413 ± 2.0 and 570 ± 5.0 µM were obtained for HT-29 and MCF-7, respectively. In addition, two control experiments were run in 100% medium and 5% DMSO/95% medium. Both control experiments behaved identically, demonstrating that 5% DMSO in the medium does not have any cytotoxic effect on these cells. Also the amount of CH2Cl2 (included as solvate) under the experimental conditions tested had no effect on the cytotoxicity of the complexes.</p><p>The IC50 data obtained on MCF-7 breast cancer and HT-29 colon cancer cell lines as determined by MTT assay are shown in Table 2 and the dose curves are depicted in Figs. 4 and 5. According to the analysis of the data in Table 2, first, we can observe that all the functionalized titanocenes are more cytotoxic than titanocene dichloride to breast cancer cell line MCF-7 (IC50 = 570 µM) and colon cancer cell line HT-29 (IC50 = 413 µM). Thus, the steroid functionalization enhances the cytotoxic activity of the titanocenes when compared with Cp2TiCl2. Additionally, two groups of titanocenyl complexes can be recognized: those containing cholesterol rings with IC50 values over 200 µM, with the exception of complex 1, and highly active titanocenyls containing sex steroids with IC50 values below 50 µM for the MCF-7 cell line.</p><p>We should underline that the more highly cytotoxic complexes are those including sex steroid derivatives, whereas those including a cholesterol unit [titanocenyl–clionasterol, 1; titanocenyl–dihydrocholesterol, 3; titanocenyl– cholesterol, 7] showed less cytotoxicity. Thus, there is great potential for the effect of the former species on hormone-dependent cancers such as prostate, testicular, and ovarian cancer to be studied. Also, we can foresee that a titanocenyl–estradiol complex (not yet synthesized) could exhibit high cytotoxicity for the MCF-7 breast cancer cell line.</p><p>From a structural point of view, 7 and its analog complexes (1 and 3) deserve special attention. These titanocenyls have very similar structures and all of them have very low cytotoxicity (IC50 > 200 µM) for the HT-29 cell line. On the other hand, 1 has significant cytotoxicity for the MCF-7 cell line. Since these three titanocenyls have very similar structures, the cytotoxic activity of 1 on MCF-7 may be explained easily in terms of lipophilicity, since 1 has an extra –CH2CH3 group. An increase in lipophilicity could facilitate the entry of the drug through the cell membrane [50].</p><p>Structural differences among the sex steroid derivatives must be also discussed. Titanocenyls 2 and 4 are similar in structure except for the substitution on carbon 17. Complex 4 is 1.5 times more cytotoxic to MCF-7 (IC50 12.6 ± 2.3) than 2, whereas the opposite behavior is observed for the HT-29 cell line. In addition, the cytotoxic activity of 2 is very similar in MCF-7 and HT-29 cell lines. In other words, since complex 4 is more cytotoxic to the breast cancer cell line it might be a better target-specific drug for the treatment of hormone-dependent cancers. There is a possible explanation for the observed cytotoxicity. DHEA is a naturally occurring steroid synthesized in the adrenal cortex, gonads, brain, and gastrointestinal tract, and it is known to have chemopreventive and antiproliferative actions on tumors, in particular on breast cancer [44, 45]. Also, it can act on the androgen receptor directly. Although the results of the investigation may suggest that a steroid containing a ring ketonic carbonyl appears to be more active than a steroid with a ring-substituted acetyl functional group, we believe that 4 is more cytotoxic owing to the inhibitory effects that DHEA itself has on breast cancer rather than because of the difference in the structures of 2 and 4. Furthermore, when we analyze the IC50 values of 4, 5, and 6, the possible structure–activity correlation must be used with caution.</p><p>Upon analysis of complex 4 and complex 5, it is clear that the cytotoxic activity of complex 4 is higher than that of complex 5. The only difference between 4 and 5 is the presence of a functional double bond in the internal steroid ring, perhaps increasing its hydrophobic character and making it easier to cross the cell membrane, particularly in the MCF-7 cancer cell line. For the pair of epimers, complexes 5 and 6, the IC50 values suggest that the stereochemistry of the androgens has an effect on their cytotoxicity.</p><p>It can be envisaged that we will be able to construct some structure–activity relationship to determine the factors needed to enhance the anticancer activity. These observations indicate that further studies of these complexes for their activity in other cancer cell lines, in particular hormone-dependent cancers, is warranted.</p><!><p>The six new titanocenyl dichlorides 1–6 shown in Fig. 2 were prepared using the synthetic method developed previously by Gansäuer et al. [47]. This allowed us to study how the steroid pendant group influences the cytotoxic activity of titanocenyl complexes.</p><p>We have discussed some of the structure–activity parameters that may influence the cytoxicity of the new steroid titanocenyl complexes. Our results indicate that the sex steroid complexes have great potential as vectors for anticancer agents. In fact, titanocenyl–steroid complexes 2, 4, and 6 exhibit IC50 values for the MCF-7 cell line in the low micromolar range and deserve to be investigated in other cell lines. Evidently, besides estrogen receptors, MCF-7 expresses androgen and progesterone receptors [51–53]. Thus, species like 2, 4, and 6 could bind progesterone and androgen receptors before expressing their activity, serving the steroids as shuttles. Nevertheless, this possible mechanism needs to be investigated in more detail. We can foresee that these pendant groups could serve as vectors (shuttles) of the titanocenes to make chemotherapeutic agents for hormone-dependent and brain cancers.</p><!><p>All reactions were carried out under an atmosphere of dry nitrogen using Schlenk glassware or a glove box, unless otherwise stated. Reaction vessels were flame-dried under a stream of nitrogen, and anhydrous solvents were transferred by oven-dried syringes or cannula. Tetrahydrofuran was dried and deoxygenated by distillation over potassium benzophenone under nitrogen. Infrared (IR) spectra were obtained with dried KBr pellets. The NMR spectra were obtained with a Bruker DRX-500 spectrometer. For the samples prepared in CDCl3, chemical shifts were referenced relative to CHCl3 at 7.27 ppm (1H NMR) and CHCl3 at 77.00 ppm (13C NMR) as internal standards. Analytical data were obtained from Atlantic Microlab.</p><p>The breast adenocarcinoma cell line MCF-7 and the colon cancer cell line HT-29 were purchased from American Type Culture Collection (ATCC) and were maintained at 37 °C and 95% air/5%CO2, as previously reported [43]. The growth medium for MCF7 was Eagle's minimum essential medium supplemented with 10% (v/v) fetal bovine serum, 1% (v/v) antibiotic/antimycotic, nonessential amino acids, and 0.01 mg/ml bovine insulin. MTT and Triton X-100 used for the cytotoxicity assay were obtained from Sigma. All MTT manipulations were performed in a dark room.</p><!><p>Titanocene acyl chloride and its precursor were prepared as described by Gansäuer et al. [47]. As reported by Gansäuer et al. and our group, these complexes crystallize as CH2Cl2 solvates [43, 47]. Apparently the inclusion of solvent is important for the stabilization of the complex in the solid state. A CH2Cl2 signal appears in all the spectra.</p><!><p>Titanium carboxylate (0.25 mmol, 77.4 mg) was dissolved in SOCl2 (1.0 ml) and stirred for 2 h at room temperature. Excess SOCl2 was removed under a high vacuum, followed by drying for 24 h. The precipitate was dissolved in CH2Cl2 (2.0 ml), the resulting solution was added dropwise to a mixture of NaH (0.75 mmol, 18 mg) and the natural product 1 (0.25 mmol, 100 mg) in CH2Cl2 (2.0 ml), and the mixture was stirred for another 20 h. After filtration through Celite, the solvent was washed with a mixture of 1 N HCl and NaCl (1.0 g each 10 ml) (2 × 5.0 ml). The organic layer was dried with MgSO4 and the solvent was removed under reduced pressure. The crude product was then chromatographed on Bio-Bead SX3 (200–400 mesh) (before use, Bio-Bead S-X3 was swollen in CH2Cl2 for 24 h) and was eluted with CH2Cl2 to give 0.14 g (75%) of viscous red oil. The product was resolved in CH2Cl2/hexane at −20 °C and a red solid was obtained. 1H NMR (500 MHz, CDCl3), δ (ppm): 6.65 (dd, 3J = 2.5 Hz, 4J = 2.5 Hz, 2H; Cp), 6.59 (s, 5H; Cp), 6.49 (dd, 3J = 2.5 Hz, 4J = 2.5 Hz, 2H; Cp), 5.37 (dd, 3J = 5.5 Hz, 3J = 2.5 Hz, 1H; 6-H), 4.54 (dddd, 3J = 10.8 Hz, 3J = 10.7 Hz, 3J = 6.4 Hz, 3J = 4.4 Hz, 1H; 3-αH), 2.58 (s, 2H; –CH2–COO–), 2.26–2.24 (m, 2H), 2.05 (ddd, 2J = 12.5 Hz, 3J = 3.6 Hz, 3J = 3.6 Hz, 1H), 2.01 (dddd, 2J = 17.4 Hz, 3J = 5.2 Hz, 3J = 5.3 Hz, 5J = 2.1 Hz, 1H), 1.96 (ddd, 2J = 9.6 Hz, 3J = 9.6 Hz, 3J = 6.0 Hz, 1H), 1.88 (ddd, 3J = 9.5 Hz, 3J = 9.5 Hz, 3J = 6.0 Hz, 1H), 1.78 (dddd, 2J = 12.5 Hz, 3J = 9.0 Hz, 3J = 3.7 Hz, 3J = 3.7 Hz, 1H), 0.98–1.65 (m, 22H), 1.52 (s, 6H), 1.02 (s, 3H), 0.95 (d, 3J = 6.5 Hz, 3H), 0.89 (d, 3J = 6.5 Hz, 3H), 0.88 (d, 3J = 6.5 Hz, 3H), 0.84 (t, 3J = 6.5 Hz, 3H), 0.69 (s, 3H). 13C NMR (125 MHz, CDCl3), δ (ppm): 170.6, 146.5, 139.6, 122.7, 120.6, 120.2, 117.1, 117.0, 74.0, 56.8, 56.0, 50.0, 49.5, 46.1, 42.3, 40.4, 38.2, 36.9, 36.8, 36.6, 36.3, 33.9, 32.0, 31.9, 28.9, 28.2, 27.8, 27.8, 27.7, 26.4, 24.3, 23.0, 21.2, 19.6, 19.3, 18.9, 18.8, 12.0, 11.9. IR (KBr, cm−1): 3,109, 3,092, 2,935, 2,867, 1,734, 1,466, 1,440, 1,368, 1,338, 1,175, 1,131, 1,008, 838, 819. Anal. calcd for C44H66Cl2O2Ti*3/4CH2Cl2: C, 66.51; H, 8.42. Found: C, 66.15; H, 8.40.</p><!><p>1H NMR (500 MHz, CDCl3), δ (ppm): 6.65 (dd, 3J = 2.5 Hz, 4J = 2.5 Hz, 2H; Cp), 6.59 (s, 5H; Cp), 6.50 (dd, 3J = 2.5 Hz, 4J = 2.5 Hz, 2H; Cp), 5.38 (dd, 3J = 5.5 Hz, 3J = 2.5 Hz, 1H; 6-H), 4.52 (dddd, 3J = 10.8 Hz, 3J = 10.7 Hz, 3J = 6.4 Hz, 3J = 4.4 Hz, 1H; 3-αH), 2.58 (s, 2H; –CH2–COO–), 2.56 (dd, 3J = 9.0 Hz, 3J = 6.5 Hz, 1H; 17-H), 2.26–2.25 (m, 2H), 2.15 (s, 3H; –COCH3), 2.20 (ddd, 2J = 12.5 Hz, 3J = 3.6 Hz, 3J = 3.6 Hz, 1H), 2.05 (m, 2H), 1.90–1.78 (m, 2H), 1.75–1.0 (m, 12H), 1.51 (s, 6H), 1.03 (s, 3H), 0.65 (s, 3H). 13C NMR (125 MHz, CDCl3), δ (ppm): 209.6, 170.7, 146.5, 139.5, 122.4, 120.6, 120.2, 117.0, 117.0, 73.8, 63.7, 56.8, 49.9, 49.4, 44.0, 38.8, 38.1, 36.9, 36.8, 36.6, 31.8, 31.7, 31.6, 29.7, 27.8, 27.8, 27.7, 24.5, 22.8, 21.0, 19.3, 13.2. IR (KBr, cm−1): 3,111, 2,933, 2,850, 1,728, 1,702, 1,609, 1,439, 1,356, 1,330, 1,223, 1,195, 1,117, 1,014, 826. Anal. calcd for C36H48Cl2O3Ti*1/8 CH2Cl2: C, 65.92; H, 7.39. Found: C, 65.50; H, 7.65.</p><!><p>1H NMR (500 MHz, CDCl3), δ (ppm): 6.65 (dd, 3J = 2.5 Hz, 4J = 2.5 Hz, 2H; Cp), 6.59 (s, 5H; Cp), 6.49 (dd, 3J = 2.5 Hz, 4J = 2.5 Hz, 2H; Cp), 4.63 (dddd, 3J = 10.8 Hz, 3J = 10.7 Hz, 3J = 6.4 Hz, 3J = 4.4 Hz, 1H; 3-αH), 2.56 (s, 2H; –CH2–COO–), 2.00–1.96 (dddd, 3J = 9.6 Hz, 3J = 9.6 Hz, 3J = 9.0 Hz, 3J = 3.7 Hz, 1H), 1.87–1.79 (m, 2H), 1.75–1.70 (dddd, 3J = 9.0 Hz, 3J = 9.0 Hz, 3J = 3.7 Hz, 3J = 3.7 Hz, 1H), 1.70–1.65 (ddd, 3J = 9.6 Hz, 3J = 9.6 Hz, 3J = 3.7 Hz, 1H), 1.60–1.47 (m, 4H), 1.45–0.6 (m, 23H), 1.53 (s, 6H), 0.92 (d, 3J = 6.5 Hz, 3H), 0.89 (d, 3J = 6.5 Hz, 3H), 0.88 (d, 3J = 6.5 Hz, 3H), 0.82 (s, 3H), 0.65 (s, 3H). 13C NMR (125 MHz, CDCl3), δ (ppm): 170.7, 146.6, 120.6, 120.5, 120.2, 119.5, 117.2, 73.7, 56.4, 56.3, 54.2, 49.6, 44.7, 42.6, 39.9, 39.5, 36.7, 36.3, 35.8, 35.5, 34.1, 32.0, 29.7, 28.6, 28.2, 28.0, 27.8, 27.5, 24.2, 23.8, 22.8, 22.6, 21.2, 18.7, 12.2, 12.1. IR (KBr, cm−1): 3,111, 2,926, 2,867, 1,732, 1,467, 1,443, 1,349, 1,178, 1,128, 1,010, 820. Anal. calcd for C42H64Cl2O2Ti: C, 70.08; H, 8.96. Found: C, 69.89; H, 9.17.</p><!><p>1H NMR (500 MHz, CDCl3), δ (ppm): 6.65 (dd, 3J = 2.5 Hz, 4J = 2.5 Hz, 2H; Cp), 6.59 (s, 5H; Cp), 6.49 (dd, 3J = 2.5 Hz, 4J = 2.5 Hz, 2H; Cp), 5.40 (dd, 3J = 5.5 Hz, 3J = 2.5 Hz, 1H; 6-H), 4.54 (dddd, 3J = 10.8 Hz, 3J = 10.7 Hz, 3J = 6.4 Hz, 3J = 4.4 Hz, 1H; 3-αH), 2.59 (s, 2H; –CH2–COO–), 2.47 (ddd, 2J = 19.0 Hz, 3J = 9.0 Hz, 3J = 0.9 Hz, 1H; 16-βH), 2.36–2.21 (m, 2H), 2.15–2.05 (m, 1H), 2.10 (ddd, 2J = 19.0 Hz, 3J = 9.0 Hz, 3J = 0.9 Hz, 1H; 16-αH), 1.94 (dddd, 2J = 12.0 Hz, 3J = 9.0 Hz, 3J = 5.7 Hz, 3J = 0.9 Hz, 1H; 15-αH), 1.90–1.83 (m, 3H), 1.72–1.62 (m, 2H), 1.59–1.45 (m, 3H), 1.52 (dd, 2J = 12.0 Hz, 3J = 9.0 Hz, 1H; 15-βH), 1.51 (s, 6H), 1.32–1.25 (m, 3H), 1.14–0.90 (m, 2H), 1.04 (s, 3H), 0.89(s, 3H). 13C NMR (125 MHz, CDCl3), δ (ppm): 221.0, 170.6, 146.5, 139.8, 121.9, 120.6, 120.6, 120.3, 117.1, 117.0, 73.7, 51.8, 50.1, 49.3, 47.5, 38.1, 36.9, 36.8, 36.7, 35.9, 31.5, 31.4, 30.8, 29.7, 27.8, 27.9, 27.7, 21.9, 20.3, 19.4, 13.6. IR (KBr, cm−1): 3,110, 2,936, 2,865, 1,736 (br s), 1,438, 1,371, 1,330, 1,197, 1,117, 1,023, 824. Anal. calcd for C34H44Cl2O3Ti*1/8CH2Cl2: C, 65.04; H, 7.10. Found: C, 65.14; H, 7.39.</p><!><p>1H NMR (500 MHz, CDCl3), δ (ppm): 6.65 (dd, 3J = 2.5 Hz, 4J = 2.5 Hz, 2H; Cp), 6.59 (s, 5H; Cp), 6.50 (dd, 3J = 2.5 Hz, 4J = 2.5 Hz, 2H; Cp), 4.63 (dddd, 3J = 10.8 Hz, 3J = 10.7 Hz, 3J = 6.4 Hz, 3J = 4.4 Hz, 1H; 3-αH), 2.57 (s, 2H; –CH2–COO–), 2.45 (ddd, 2J = 19.0 Hz, 3J = 9.0 Hz, 3J = 0.9 Hz, 1H; 16-βH), 2.09 (ddd, 2J = 19.0 Hz, 3J = 9.0 Hz, 3J = 0.9 Hz, 1H; 16-αH), 1.94 (dddd, 2J = 12.0 Hz, 3J = 9.0 Hz, 3J = 5.7 Hz, 3J = 0.9 Hz, 1H; 15-αH), 1.81–1.80 (m, 2H), 1.74–1.19 (m, 13H), 1.52 (dd, 2J = 12.0 Hz, 3J = 9.0 Hz, 1H; 15-βH), 1.50 (s, 6H), 1.15–0.97 (m, 2H), 0.89 (s, 3H), 0.85 (s, 3H), 0.75–0.69(m, 1H). 13C NMR (125 MHz, CDCl3), δ (ppm): 221.3, 170.8, 146.5, 120.6, 120.5, 120.2, 117.1, 117.1, 73.5, 54.3, 51.4, 49.5, 47.8, 44.7, 36.7, 36.6, 35.9, 35.6, 35.0, 33.9, 31.5, 30.8, 29.7, 28.2, 27.7, 27.5, 21.8, 20.5, 13.8, 12.2. IR (KBr, cm−1): 3,109, 2,922, 2,852, 1,743, 1,727, 1,465, 1,447, 1,364, 1,339, 1,201, 1,121, 1,014, 821. Anal. calcd for C34H46Cl2O3Ti: C, 65.69; H, 7.46. Found: C, 65.44; H, 7.79.</p><!><p>1H NMR (500 MHz, CDCl3), δ (ppm): 6.65 (dd, 3J = 2.5 Hz, 4J = 2.5 Hz, 2H; Cp), 6.60 (s, 5H; Cp), 6.48 (dd, 3J = 2.5 Hz, 4J = 2.5 Hz, 2H; Cp), 4.96 (dd, 3J = 3.0 Hz, 3J = 2.5 Hz, 1H; 3-βH), 2.60 (s, 2H; –CH2– COO–), 2.45 (ddd, 2J = 19.0 Hz, 3J = 9.0 Hz, 3J = 0.9 Hz, 1H; 16-βH), 2.09 (ddd, 2J = 19.0 Hz, 3J = 9.0 Hz, 3J = 0.9 Hz, 1H; 16-αH), 1.94 (dddd, 2J = 12.0 Hz, 3J = 9.0 Hz, 3J = 5.7 Hz, 3J = 0.9 Hz, 1H; 15-αH), 1.81–1.80 (m, 2H), 1.74–1.19 (m, 14H), 1.52 (dd, 2J = 12.0 Hz, 3J = 9.0 Hz, 1H; 15-βH), 1.51 (s, 6H), 1.08–0.99 (m, 1H), 0.89 (s, 3H), 0.82 (s, 3H), 0.80–0.84(m, 1H). 13C NMR (125 MHz, CDCl3), δ (ppm): 221.5, 170.8, 146.5, 120.5, 120.4, 120.3, 117.2, 117.1, 69.9, 54.3, 53.5, 51.4, 49.5, 47.8, 40.1, 36.7, 35.9, 35.8, 35.0, 32.9, 32.8, 31.5, 30.8, 28.1, 27.8, 27.7, 26.1, 21.8, 20.1, 13.8, 11.4. IR (KBr, cm−1): 3,110, 2,931, 2,855, 1,734 (br s), 1,597, 1,447, 1,386, 1,368, 1,354, 1,250, 1,199, 1,117, 1,057, 1,015, 822. Anal. calcd for C34H46Cl2O3Ti*1/4CH2Cl2: C, 63.99; H, 7.30. Found: C, 63.60; H, 7.48.</p><!><p>Biological activity was determined using the MTT assay originally described by Mossman [49] but using 10% Triton X-100 in 2-propanol as a solvent for the MTT formazan crystals [50]. HT29 and MCF7 cells were maintained at 37 °C and 95% air/5% CO2 in McCoy's 5A (ATCC) complete medium, which had been supplemented with 10% (v/v) fetal bovine serum (ATCC) and 1% (v/v) antibiotic/antimycotic (Sigma). Asynchronously growing cells were seeded at 1.5 × 104 cells per well in 96-well plates containing 100 µl of complete growth medium, and were allowed to recover overnight. Various concentrations of the complexes (1–1,300 µM) dissolved in 5% DMSO/95% medium were added to the wells (eight wells per concentration; experiments performed in quadruplicate plates). The solutions of the complexes were prepared by first dissolving the corresponding titanocenyl in DMSO and then medium was added to a final composition of 5% DMSO/95% medium. In addition to experiments on the cells treated with the titanocenyls, two control experiments were performed: one without any addition of solvent mixture (5% DMSO/95% medium) and one with addition of 5% DMSO/95% medium to the cells. Both control experiments behaved identically, showing that 5% DMSO in the medium was not toxic to these types of cells. Although CH2Cl2 may have an effect on the cytotoxicity, under our experimental conditions at concentrations of 10−4–10−7 M, CH2Cl2 has no effect at all in terms of cytotoxicity. Actually, we estimated that the IC50 of CH2Cl2 for MCF-7 is 30 mM (3.0 × 10−2 M) and for HT-29 is higher than 50 mM (5.0 × 10−2 M).</p><p>The cells were incubated for an additional 70 h. After this time, MTT dissolved in complete growth medium was added to each well to a final concentration of 1.0 mg/ml and the mixture was incubated for an additional 2h. After this period, all MTT-containing medium was removed, the cells were washed with cold phosphate-buffered saline, and were dissolved in 200 µl of a 10% (v/v) Triton X-100 solution in 2-propanol. After complete dissolution of the formazan crystals, well absorbances were recorded in triplicate with a 340 ATTC microplate reader (SLT Lab Instruments) at 570 nm with background subtraction at 630 nm. The concentrations of the compounds required to inhibit cell proliferation by 50% (IC50) were calculated by fitting the data to a four-parameter logistic plot by means of SigmaPlot from SPSS.</p><!><p>Electronic supplementary material The online version of this article (doi:10.1007/s00775-010-0649-7) contains supplementary material, which is available to authorized users.</p>

PubMed Author Manuscript

High expression of class III \xce\xb2-tubulin has no impact on functional cancer cell growth inhibition of a series of key vinblastine analogs

Clinical association studies have implicated high expression of class III \xce\xb2-tubulin as a predictive factor for lower response rates and reduced overall survival in patients receiving tubulin binding drugs, most notably the taxanes. Because of the implications, we examined a series of key vinblastine analogs that emerged from our studies in functional cell growth inhibition assays for their sensitivity to high expression of class III \xce\xb2-tubulin (human non-small cell lung cancer cell line A549 vs taxol-resistant A549-T24). Unlike taxol, vinblastine and a set of key analogs 3\xe2\x80\x9310 did not exhibit any loss in sensitivity toward A549-T24. The results suggest that vinblastine and related analogs are not likely prone to resistance derived from high expression of class III \xce\xb2-tubulin unlike the taxanes. Most significant are the results with 4\xe2\x80\x936, a subset of 20\xe2\x80\xb2 amide vinblastine analogs. They match or exceed the potency of vinblastine and they display more potent activity against taxol-resistant A549-T24 than even wild type A549 cells (1.2\xe2\x80\x932 fold), complementing our prior observations that they also display no sensitivity to overexpression of Pgp (HCT116/VM46 vs HCT116) and are not subject to resistance derived from Pgp efflux.

high_expression_of_class_iii_\xce\xb2-tubulin_has_no_impact_on_functional_cancer_cell_growth_inhibit

<p>The Vinca alkaloids are a family of natural products that continue to have a remarkable impact on anticancer drug discovery and treatment.1,2 Originally isolated in trace quantities from the periwinkle plant (Catharanthus roseus (L.) G.Don),3,4 vinblastine (1) and vincristine (2) are the most prominent members of this class and among the first plant-derived natural products used in the clinic for the treatment of cancer (Figure 1). These two compounds along with three clinically-approved semi-synthetic analogs, vindesine,5 vinorelbine6 and vinflunine,7 are integral oncology drugs employed today in highly successful combination drug therapies. Their mode of action, which involves disruption of microtubulin formation and dynamics through tubulin binding, remains one of the most successful approaches for the discovery and development of new oncology drugs.8</p><p>Although vinblastine and vincristine are superb drugs even by today's standards, a potential limitation to their continued use is the emergence of clinical resistance. Most recognized of the mechanisms of resistance to the Vinca alkaloids is that mediated by overexpression of the drug efflux pump phosphoglycoprotein (Pgp).9 Pgp overexpression, which also results in multidrug resistance (MDR), is responsible for the majority of all relapses in oncology. The discovery of vinblastine analogs not susceptible to Pgp efflux could serve as potential replacements for vinblastine in its current clinical uses or in instances of Pgp-derived vinblastine resistance. Even more significantly, it could expand the use of a vinblastine to new therapeutic applications for Pgp-derived MDR tumor treatments. Despite past efforts focused on vinblastine that have searched for analogs that effectively overcome Pgp-derived vinblastine resistance, little progress has been made.2 Recent advances in the total synthesis of vinblastine, vincristine and related natural products provided access to analogs of the natural products not previously accessible by semisynthetic modification of the natural products themselves.10–12 The latest of these efforts provided a powerful approach to access vinblastine analogs that contain systematic deep-seated modifications within either the lower vindoline-derived13–21 or upper catharanthine-derived subunits.22–29 As a result of these developments, we have disclosed several series of key analogs, systematically exploring and defining the impact individual structural features and substituents have on tubulin binding affinity and cancer cell growth inhibition. In these studies, we have shown that replacement of the C20′-OH with 20′ ureas was possible,22 that substantial23,24 and even remarkable25 potency enhancements were obtainable with such 20′ ureas, and that some exhibited further improvements in activity against vinblastine-resistant cancer cell lines.24 In an extension of these studies, an extensive series of vinblastine 20′ amides were disclosed that provided analogs that matched or exceeded the potency of vinblastine, but that are not subject to Pgp efflux and its derived vinblastine resistance.26 Although not discussed herein, well-defined structure–activity results and accompanying structural models were delineated that account for not only the on target tubulin binding affinity and resulting functional cell growth inhibition of the analogs (HCT116), but also defined the structural features and their characteristics needed for avoidance of Pgp efflux and resistance derived from Pgp overexpression (HCT116/VM46). The studies provided vinblastine analogs no longer susceptible to resistance derived from overexpression of Pgp, and they represented the discovery of a site and functionalization strategy for the preparation of readily accessible vinblastine analogs (3 steps) that improve binding affinity to tubulin (on target affinity) and functional potency in cell-based assays while simultaneously disrupting their efflux by Pgp (off target affinity and source of resistance), offering a powerful opportunity to discover new, improved, and durable oncology drugs.</p><p>Alterations to the target tubulin could also impact activity and contribute to or be responsible for Vinca alkaloid resistance. A series of association studies of clinical data have implicated high level expression of class III β-tubulin as both a prognostic and predictive factor for lower response rates or reduced overall survival in patients receiving tubulin binding drugs.30,31 However, most of the association studies and the supporting cellular studies have examined the impact of class III β-tubulin on taxanes and a much smaller sampling of its impact on Vinca alkaloids are represented in the association studies.30,31 Despite the obvious differences in the tubulin binding sites of the taxanes and Vinca alkaloids as well as their distinct functional behaviors (stabilization vs destabilization of tubulin dynamics), both taxanes and the Vinca alkaloids typically have been lumped together as potentially being negatively impacted by the high expression of class III β-tubulin.30,31 Because of these implications, we have examined a series of the key vinblastine analogs that emerged from our studies in additional cell-based functional cell growth inhibition assays for their sensitivity to the high expression of class III β-tubulin and report the results herein.</p><p>The key analogs that were examined alongside vinblastine (1) are 10′-fluorovinblastine (3),28 three vinblastine 20′ amides 4–6 that displayed no susceptibility to Pgp efflux and are insensitive to Pgp overexpression resistance,26 and a series of vinblastine 20′ ureas 7–10 that includes the ultrapotent analogs 8–10 (Figure 2).24,25 In our original work, these were screened for growth inhibition activity against HCT116 (human colon cancer cell line) and a matched resistant cell line (HCT116/VM46) that is approximately 100-fold resistant by virtue of the overexpression of Pgp. This well-designed set of cell-based assays simultaneously provided a direct measure of both functional activity (HCT116) and the analog susceptibility to Pgp efflux (resistance, HCT116/VM46). Key members were assessed for tubulin binding affinity for correlation with functional activity and those that showed no sensitivity to Pgp overexpression, including 4–6, were examined in efflux assays to confirm their behavior toward Pgp and related efflux transporters.26 The cell growth inhibition activity against the L1210 (mouse leukemia) cell line was also measured and the results were qualitatively and quantitatively (IC50) nearly identical to those observed with the HCT116 cell line. In these studies, the HCT116 human tumor cell line was found to accurately reflect activity observed against a larger panel of clinically relevant human tumor-derived cell lines.12,24–28 Alongside this prior data and herein, we now report the cell growth inhibition activity of the analogs 3–10 against the human A549 cell line (human non-small cell lung cancer) and a matched cell line A549-T24 insensitive to taxol (Figure 2).32 After its original generation and characterization, this taxol-resistant cell line (A549-T24) was found to embody an increased expression of class III β-tubulin that was correlated with the loss in taxol sensitivity with no alteration in Pgp expression.32 As a result, the comparison of cell growth inhibition of candidate drugs against A549 vs A549-T24 has been used to characterize potential resistance due to increased expression of class III β-tubulin.32</p><p>Consistent with prior reports, A549-T24 proved insensitive to taxol treatment, exhibiting a 10-fold loss in potency relative to wild type A549 cells (IC50 = 70 vs 7.1 nM).32 In contrast, vinblastine as well as all the vinblastine analogs 3–10 did not exhibit any loss in sensitivity toward A549-T24. In fact, most showed slight increases in potency (1–3.6 fold, avg = 1.8 fold), suggesting they may be even more effective in the presence of expressed class III β-tubulin. In retrospect, this may not be surprising. Class III β-tubulin increases the instability and dynamics of microtubule assembly, potentially countering the stabilizing effects of bound taxol, but enhancing the destabilizing effects of the Vinca alkaloids. Most significant within this series of analogs are the results with 4–6.26 They match or exceed the potency of vinblastine, they display equipotent or more potent activity against A549-T24 than even wild type A549 cells (1.2–2 fold) and, as detailed earlier,26 they uniquely display no sensitivity to the overexpression of Pgp (HCT116/VM46 vs HCT116) and are not subject to efflux by Pgp or related drug efflux transporters.26 The results not only suggest that (1) vinblastine and related analogs are not likely to be prone to resistance derived from high expression of class III β-tubulin unlike the taxanes, but also that (2) association studies of clinical data with tubulin binding drugs30,31 should treat taxanes and the Vinca alkaloids as distinct drug classes likely to exhibit different sources of on target resistance.</p>

PubMed Author Manuscript

The Clarithromycin Susceptibility Genotype Affects the Treatment Outcome of Patients with Mycobacterium abscessus Lung Disease

ABSTRACTMycobacterium abscessus accounts for a large proportion of lung disease cases caused by rapidly growing mycobacteria. The association between clarithromycin sensitivity and treatment outcome is clear. However, M. abscessus culture and antibiotic susceptibility testing are time-consuming. Clarithromycin susceptibility genotyping offers an alternate, rapid approach to predicting the efficacy of clarithromycin-based antibiotic therapy. M. abscessus lung disease patients were divided into two groups based upon the clarithromycin susceptibility genotype of the organism isolated. A retrospective analysis was conducted to compare the clinical features, microbiological characteristics, and treatment outcomes of the two groups. Several other potential predictors of the response to treatment were also assessed. Sixty-nine patients were enrolled in the clarithromycin-resistant genotype group, which included 5 infected with rrl 2058-2059 mutants and 64 infected with erm(41)T28-type M. abscessus; 31 were in the clarithromycin-sensitive group, i.e., 6 and 25 patients infected with genotypes erm(41)C28 and erm(41) M type, respectively. The results showed that lung disease patients infected with clarithromycin-sensitive and -resistant M. abscessus genotypes differed significantly in clarithromycin-based combination treatment outcomes. Patients infected with the clarithromycin-sensitive genotype exhibited higher initial and final sputum-negative conversion and radiological improvement rates and better therapeutic outcomes. Multivariate analysis demonstrated that genotyping was a reliable and, more importantly, rapid means of predicting the efficacy of clarithromycin-based antibiotic treatment for M. abscessus lung disease.

the_clarithromycin_susceptibility_genotype_affects_the_treatment_outcome_of_patients_with_mycobacter

INTRODUCTION<!>Patient characteristics.<!><!>Colony morphology.<!><!>Comparison of antibiotic sensitivity.<!><!>Combination antibiotic treatment and treatment response.<!><!>Combination antibiotic treatment and treatment response.<!><!>DISCUSSION<!>Study population.<!><!>Collection, identification, and preservation of bacteria.<!>Identification of colony morphology.<!>Genotype analysis.<!>(i) Whole-genome sequencing.<!>(ii) SNP analysis.<!>Drug sensitivity assay.<!>Treatment regimen and efficacy evaluation.<!>Statistical analysis.<!>Accession number(s).<!>

<p>The incidence of infections by nontuberculosis mycobacteria (NTM) has increased significantly in recent years (1–3). Of all NTM infections, treatment of Mycobacterium abscessus infections is the most challenging (4, 5). M. abscessus accounts for 65 to 80% of the cases of lung disease caused by rapidly growing mycobacteria and has emerged as an important pathogen for patients with bronchiectasis, chronic obstructive pulmonary disease, and cystic fibrosis (6–11).</p><p>M. abscessus is among the most antibiotic-resistant pathogens known (12). Although some antibiotics, such as amikacin, cefoxitin, and imipenem, are effective, only clarithromycin (CLA) exhibits convincing evidence of clinical efficacy for treatment of M. abscessus lung disease (8). Currently, CLA is the only effective antibiotic administered orally and, therefore, recommended as the core agent for treatment of M. abscessus infections (8).</p><p>Genotypic variations influence the sensitivity of M. abscessus to CLA. Two genotypes confer CLA resistance: a point mutation (A to C or A to G) in the 2058-2059 locus of the 23S rRNA (rrl) gene confers acquired resistance (13). An intact erm(41) gene, which exhibits a T/C polymorphism at the 28th nucleotide, confers inducible resistance when the 28th nucleotide is thymidine [erm(41)T28] (14, 15). Alternatively, CLA sensitivity is conferred when cytidine is the 28th nucleotide in intact erm(41), i.e., genotype erm(41)C28 (15). Deletion of erm(41) nucleotides 64 and 65, or deletion of nucleotides 159 to 432, also results in the loss of erm(41) gene function (M type) and a gain in CLA sensitivity (14, 16).</p><p>M. abscessus can be divided into M. abscessus subsp. abscessus and M. abscessus subsp. massiliense based upon the integrity or absence of the erm(41) gene. Korean and Japanese researchers first reported that M. abscessus subsp. abscessus and M. abscessus subsp. massiliense exhibited disparate clinical and microbiological characteristics (17, 18). Retrospective analysis and a prospective study conducted in 2017 confirmed these results and suggested that patients infected with a CLA-sensitive [erm(41)C28] genotype had a prognostic advantage (19, 20). Therefore, differences in the A and M subtypes may be due largely to genotypic differences that affect CLA sensitivity (21–23). Here, we report the results of a retrospective analysis undertaken to determine the relationship between genotype, CLA sensitivity, and the outcome of CLA-based treatment of M. abscessus lung disease.</p><!><p>One hundred M. abscessus lung disease patients who conformed to our recruitment criteria were enrolled and divided into CLA-resistant and -sensitive genotype groups according to the rrl and erm(41) sequevar. Sixty-nine (69%) patients were enrolled in the CLA-resistant genotype group, which included 5 (7.2%) rrl 2058-2059 mutant- and 64 (92.8%) erm(41)T28-type-infected patients; 31 (31%) belonged to the CLA-sensitive genotype group, which included 6 (19.4%) erm(41)C28- and 25 (80.6%) erm(41) M-type-infected patients. No significant differences were found in the ages and genders of the two groups (Table 1). The proportion of patients with hemoptysis was higher in the CLA-resistant genotype group than the CLA-sensitive genotype group (22/69 versus 4/31; P = 0.045). Furthermore, a significantly greater incidence of cavity-like manifestations occurred in computed tomography (CT) scans of patients infected with isolates with the CLA-resistant genotype than in patients infected with isolates with the CLA-sensitive genotype (50/69 versus 8/31; P < 0.001). CT scans of the CLA-sensitive-group patients, on the other hand, displayed a higher incidence of a tree-in-bud pattern (14/31 versus 16/69; P = 0.027).</p><!><p>Baseline characteristics of patients infected with M. abscessus belonging to CLA-resistant and -sensitive genotypes</p><p>COPD, chronic obstructive pulmonary disease; CHD, coronary heart disease; AFB, acid-fast bacilli; IQR, interquartile range.</p><p>Data are the numbers (%) of patients found in the CLA-resistant and -sensitive genotype groups unless otherwise indicated.</p><p>Patients treated for tuberculosis prior to the diagnosis of M. abscessus lung disease.</p><!><p>M. abscessus isolates manifest two distinct colony morphotypes: smooth and rough. The colony morphology of the isolates associated with both CLA susceptibility groups did not differ significantly (Table 1). Patients infected with M. abscessus characterized by a rough-type colony exhibited a higher incidence of cavities in CT images (P = 0.043) (Table 2). The colony morphotype did not exert a significant effect on any of the other parameters assessed.</p><!><p>Relationship between morphotype, results of initial CT scan, and treatment outcome</p><p>Number (percentage) of patients infected with isolates that give rise to rough and smooth colony types versus the disease parameter listed.</p><!><p>The sensitivity of all the M. abscessus isolates to 10 antibiotics tested is shown in Table 3 and Table S1 in the supplemental material. The five rrl 2058-2059 mutant isolates exhibited acquired resistance to CLA, i.e., they were resistant on day 3 of exposure and prior to induction. Twenty-seven of the 64 erm(41)T28 isolates also exhibited acquired resistance; 36 isolates were induced by 14 days exposure to CLA; and one isolate showed abnormal CLA sensitivity despite expressing an erm(41)T28 gene, albeit with a wild-type rrl gene. In sharp contrast, no CLA resistance was observed within the CLA-sensitive genotype group. Notably, although most isolates in the CLA-resistant genotype group were insensitive to CLA, only one isolate was insensitive to amikacin treatment. A considerable number of isolates in both the CLA-sensitive and CLA-resistant genotype groups were sensitive to linezolid. A large number of isolates in both groups were resistant to moxifloxacin, doxycycline, imipenem, and tobramycin; no significant difference in resistance to these antibiotics was found between groups.</p><!><p>Antibiotic resistance of all M. abscessus isolatesa</p><p>The erm(41) sequevar-dependent resistance of 100 M. abscessus isolates to the antibiotics indicated was determined by the microdilution method. The incubation time was 3 days (before) and 14 days (after) induction for CLA and 3 days for the other antibiotics listed.</p><p>Resistant isolates were distinguished according to the breakpoint provided by NCCLS document M24-A2. ND, no data. Tigecycline has no recommended breakpoint.</p><!><p>All patients enrolled in the study were treated with a standard combination of antibiotics based upon CLA. Patients infected with the CLA-sensitive genotype group isolates were significantly more likely to demonstrate initial sputum conversion (Fig. 1 and Table 4) (P = 0.011). Times to initial sputum conversion also differed significantly between the CLA-sensitive and CLA-resistant genotype groups (P = 004). Sputum relapse after initial conversion to negative occurred in both groups and did not differ significantly. The proportion of patients whose sputa converted and remained culture-negative during the follow-up period was significantly greater in the CLA-sensitive than in the CLA-resistant genotype group (61.3% versus 30.4%, respectively; P = 0.013). Radiographic improvement rates were significantly higher in patients infected with the CLA-sensitive genotype group isolates than in patients infected with the CLA-resistant genotype group (P = 0.006). The effective treatment response evaluated by radiology and microbiology was also significantly greater for the CLA-sensitive genotype group than for the CLA-resistant genotype group (P < 0.001).</p><!><p>Comparison of initial sputum smear/culture conversion between patients infected with the CLA-resistant [2058-2059 rrl mutant or rrl wild type/erm(41)T28] and -sensitive [rrl wild type/erm(41)C28 or rrl wild type/erm(41) M type] genotype groups. Patients infected with the resistant group isolates showed a significantly longer initial sputum medium conversion time: 12 months versus 7 months for the sensitive group (P = 0.004).</p><p>Treatment outcomes for CLA-resistant and -sensitive genotype groups</p><p>The data are the number and (percentage) of patients in each group unless otherwise indicated.</p><!><p>In multivariate analysis, the genotype was a reliable predictor of the response of M. abscessus lung disease to treatment (odds ratio [OR] = 0.185; 95% confidence interval [CI], 0.059 to 0.579; P = 0.004) (Table 5). All other characteristics, i.e., age, sex, body mass index (BMI), colony morphology, and CT imaging, were nonpredictors.</p><!><p>Univariate and multivariate analyses of factors affecting combination antibiotic treatment</p><!><p>The study reported here was the first undertaken to explore and correlate the differences in treatment outcomes of M. abscessus lung disease patients with the CLA susceptibility genotype of clinical isolates. We found that the treatment results for patients infected with isolates with the CLA-sensitive M. abscessus genotype were far superior to the results for patients infected with isolates with the CLA-resistant genotype evaluated in terms of sputum conversion rate, duration of initial sputum conversion, radiological improvement, and efficacy. Treatment outcome, however, was independent of all other factors examined, which included BMI, colony morphology, and radiological images.</p><p>In 2006, the M. abscessus complex was first divided into M. abscessus subsp. abscessus and M. abscessus subsp. massiliense based upon differences in the rpoB gene (24). In 2011, Bastian and coworkers reported that variations in the erm(41) genotype influenced the sensitivity of these subtypes to CLA in vitro (15). The clinical characteristics and treatment outcomes of patients infected with M. abscessus subsp. abscessus and M. abscessus subsp. massiliense differed in subsequent studies (17–20). Patients infected with M. abscessus subsp. massiliense usually responded better to treatment due, in part, to the CLA sensitivity of the organism. Several genotypes are associated with CLA sensitivity and -resistance: rrl mutant/wild type, erm(41)T28, erm(41)C28, and erm(41) M type. In the study described here, clinical isolates were grouped into these genotypes rather than M. abscessus subsp. abscessus and M. abscessus subsp. massiliense, and the responses of patients to standard, CLA-based treatment were assessed and compared. The response of the CLA-sensitive genotype group was significantly superior to that of the CLA-resistant genotype group judged in terms of the sputum conversion rate, radiological improvement, duration of initial sputum conversion results, and treatment efficacy. While lung disease patients infected with M. abscessus subsp. abscessus [erm(41)C28 genotype] isolates may exhibit a better response to combination CLA treatment, the response of patients infected with M. abscessus subsp. massiliense isolates expressing the 2058-2059 rrl mutation was often much worse. As such, CLA susceptibility genotyping is more accurate than subtyping as an approach to predicting the treatment outcomes of patients with M. abscessus lung disease (46.4% versus 42.9% true-positive rates, respectively).</p><p>The effect of BMI on the treatment outcomes of patients with NTM lung disease was demonstrated in several studies (25, 26). A recent retrospective study suggested that, in addition to CLA sensitivity, BMI was a factor that affected the success of M. abscessus lung disease treatment (20). This suggestion, however, was not confirmed by the present study. The overall BMIs of M. abscessus lung disease patients enrolled in our study were low; moreover, multivariate analysis failed to support its value in predicting an effective treatment outcome. This finding is consistent with results reported by other investigators (19). Similarly, predictions concerning the outcome of antibiotic therapy based upon symptoms or CT imaging are unrealistic. While hemoptysis and cavity-like manifestations were more common among the patients infected with CLA-resistant genotype M. abscessus, these factors failed to predict the prognosis upon multivariate analysis.</p><p>Patients infected with M. abscessus characterized by a rough-type colony exhibited a high incidence of cavities in CT images. This finding is consistent with the conclusion that rough-type strains usually exhibit higher virulence and pathogenicity. Unlike previous studies (19), however, we found that the initial colony morphology failed to correlate with the final radiologic improvement rate or treatment efficacy (Table 2). Jonsson and coworkers reported a significant increase in the number of rough colonies during the course of infection and the occurrence of smooth-to-rough colony conversion (27). Thus, we speculate that colony morphology is associated only with pathogenicity and the pathogenesis of infection and is not a reliable predictor of treatment efficacy.</p><p>The study described here has several limitations. First, only a relative small number of isolates exhibited the rrl mutation and erm(41)C28 genotypes; consequently, their characteristics may not be representative. Solidifying their characteristics will require the enrollment of more patients infected with isolates exhibiting the rrl mutant and erm(41)C28 genotypes in future studies. Second, a minority of patients relapsed following initially successful treatment (see Table S2 in the supplemental material). Conceivably, these relapses were due to subsequent infection by a different M. abscessus strain or genotype. In the absence of dynamic follow-up, our study failed to determine whether recurrence occurred due to reinfection by a different M. abscessus strain.</p><p>In conclusion, there was a significant difference in treatment outcomes for patients infected with CLA-resistant and -sensitive M. abscessus genotype isolates. The CLA-sensitive genotype group was significantly superior in sputum conversion rate, initial sputum conversion time, radiological improvement, and treatment efficacy. Accurate genotyping is an important factor in predicting the efficacy of combination therapy with CLA-based antibiotics. Rapid genotyping should help clinicians optimize therapeutic strategies, especially in cases of critically ill patients who cannot wait weeks for culture and susceptibility testing. Genotyping would also be effective as a diagnostic approach in areas where facilities for mycobacterial culture and susceptibility testing are unavailable.</p><!><p>A retrospective review of the medical records of all patients with M. abscessus lung disease was conducted between January 2012 and December 2015 at the Shanghai Pulmonary Hospital. Patient inclusion criteria were as follows: (i) age, >16 years; (ii) underwent initial diagnosis and treatment at the Shanghai Pulmonary Hospital in accordance with the 2007 American Thoracic Society/Infectious Disease Society of America (ATS/IDSA) guidelines; (iii) received oral CLA-based combination treatment; (iv) follow-up period lasted more than 6 months. The exclusion criteria were as follows: (i) age, <16 years; (ii) history of NTM lung disease; (iii) lack of critical visit data (e.g., regular sputum culture or CT examination), failure to follow up, or death from non-M. abscessus lung disease-related causes; (iv) treatment did not include oral CLA; (v) history of long-term macrolide drug treatment; (vi) diagnosed with active tuberculosis or received antituberculosis treatment within 3 months prior to study enrollment; (vii) coinfected with another nontuberculosis mycobacterium; (viii) refused to sign informed consent form; (ix) AIDS. In addition, patients with cystic fibrosis were not included in the study; notably, cystic fibrosis is extremely rare among Asian patients. A detailed, patient enrollment flow chart is shown in Fig. 2. This study was approved by the Ethics Committees of Shanghai Pulmonary Hospital and Tongji University School of Medicine, ethics number K17-150. All participants signed informed consent forms before enrollment.</p><!><p>Flow diagram of the study. One hundred M. abscessus lung disease patients who conformed to the inclusion criteria were enrolled. Sixty-nine patients were in the CLA-resistant genotype group, including 5 patients infected with rrl 2058-2059 mutants and 64 erm(41)T28-type-infected patients; 31 belonged to the CLA-sensitive group, including 6 erm(41)C28- and 25 erm(41) M-type-infected patients.</p><!><p>All the clinical M. abscessus isolates used in this study were preserved in the Clinical Microbiology Laboratory of Shanghai Pulmonary Hospital. Shanghai Pulmonary Hospital is one of the designated treatment centers for tuberculosis and NTM disease in China, attracting NTM disease cases nationwide. M. abscessus isolates were obtained from sputum and bronchoalveolar lavage fluid. Samples were transferred to Lowenstein-Jensen (L-J) agar plates after treatment with 4% NaOH. Smears prepared from the bacterial colonies that grew were stained and examined microscopically to identify the acid-fast organisms. To select further NTM, positive colonies were inoculated and cultured in L-J medium containing 0.5 mg/ml P-nitrobenzoic acid and 5 mg/ml 2-thiophenecarboxylic acid hydrazide for 1 to 2 weeks at 37°C. Bacterial isolates that grew rapidly were selected for molecular typing by PCR. The bacteria were digested with 1 mg/ml lysozyme and 1 mg/ml proteinase K, and the DNA was extracted with phenol-chloroform. First, the rpoB gene was amplified by PCR, and the DNA sequences were determined. To confirm the M. abscessus complexes, 754 bp of the DNA segment was subjected to BLAST analysis. Second, the erm(41) gene was amplified, and the DNA sequence was analyzed to identify and differentiate M. abscessus subsp. massiliense, M. abscessus subsp. abscessus, and M. abscessus subsp. bolletii. Finally, the PRA-hsp65 gene was compared to an online reference (http://app.chuv.ch/prasite/index.html) to confirm the M. abscessus subsp. abscessus and M. abscessus subsp. bolletii identifications. M. abscessus subsp. bolletii was excluded from the study because it is essentially absent in China. Identified M. abscessus subsp. abscessus and M. abscessus subsp. massiliense isolates, stored at −80°C, were subsequently recovered for microbiology and molecular biology studies.</p><!><p>Single colonies were obtained from frozen M. abscessus isolates by growth on Middlebrook 7H10 agar plates supplemented with 10% oleic acid-albumin-dextrose-catalase. The colonies were classified macroscopically as smooth or rough. If isolates gave rise to colonies of both morphotypes, a colony of each type was analyzed separately, and the identity was established by whole-genome sequencing.</p><!><p>Genomic information for all isolates was obtained by whole-genome sequencing. Single nucleotide polymorphism (SNP) analysis was performed using the NCBI GenBank database and BLAST algorithm. The following genotypes were of specific interest: erm(41) [including erm(41)C28, erm(41)T28, and erm(41) M type], rll wild type, and rrl 2058-2059 mutant.</p><!><p>Detailed methods were published previously by us (28). DNA was extracted according to the method of Somerville and coworkers (29), and paired-end libraries with insert sizes of ∼400 bp were prepared following Illumina's standard genomic DNA library preparation protocol (Illumina, San Diego, CA, USA). After shearing, ligating, and PCR, the qualified Illumina paired-end library was used for Illumina HiSeq sequencing (paired-end 150 bp × 2). The default parameters of the SPAdes software (version v.3.6.0) (http://bioinf.spbau.ru/en/spades) were used to assemble the genome draft (30). The assembled product was evaluated using QUAST (version v.2.3) (31; http://quast.bioinf.spbau.ru/).</p><!><p>The NCBI Nucleotide BLAST program was used for SNP analysis. The standard ATCC 19977 (NC_010397.1) M. abscessus strain served as the reference for rrl and erm(41)T28, CR5701 (HQ127366.1) was used as the reference strain for erm(41)C28, and CCUG48898 (AP014547.1) was the reference for M type.</p><!><p>Antibiotic sensitivity was determined by the microdilution method. Sulfonamides, moxifloxacin, cefoxitin, amikacin, doxycycline, tigecycline, CLA, linezolid, imipenem, and tobramycin are among the most common antibiotics used to treat M. abscessus infections; each was tested (TREK Diagnostic Systems, Brooklyn Heights, OH, USA). CLA resistance was assessed at 3 days and 14 days after M. abscessus exposure. Antibiotics' susceptible and resistant breakpoints were interpreted according to Clinical and Laboratory Standards Institute (CLSI) document M24-A2. Staphylococcus aureus (ATCC 29213; American Type Culture Collection, Manassas, VA, USA) served as the control reference strain.</p><!><p>All patients were treated with antibiotics as follows: an initial 4-week course of amikacin (15 mg/kg of body weight/day in two equal doses) combined with cefoxitin (200 mg/kg/day with a maximum of 12 g/day in three equal doses) by intravenous administration. CLA was also administered orally from the beginning of therapy. After 4 weeks, an oral regimen of CLA combined with levofloxacin or moxifloxacin was given. If an adverse reaction to either amikacin or cefoxitin occurred, the regimen was replaced with imipenem (500 mg three times a day), linezolid (600 mg once every 12 h), or tigecycline (100 mg initially, followed by 50 mg every 12 h). CLA was administered continually throughout the course of treatment as recommended in the guidelines.</p><p>All the patients underwent chest CT examination, as well as sputum smears and culture, regularly. Therapeutic efficacy was determined according to the results of microbiological examination and radiological changes. The clinical characteristics, sputum culture conversion rate and time, radiological improvement rate, and microbiological characteristics of each genotype group were compared. Culture conversion was defined as three consecutive negative cultures from sputum specimens. Effective treatment was defined as sputum culture negative or significant pulmonary lesion resolution without recurrence during the observation period. Ineffective treatment included failure to achieve culture and smear conversion, recurrence after initial culture conversion, and appearance of increased or stable lesions in CT scans.</p><!><p>All statistical analyses were conducted using SPSS20.0 (IBM, Armonk, NY, USA). The data were compared using Student's t test or the Mann-Whitney U test for continuous variables and the Pearson χ2 test or Fisher exact test for categorical variables. P values of <0.05 were considered statistically significant in a 2-tailed analysis. Times to initial culture conversion were compared using the Kaplan-Meier method. Potential predictors of the treatment response were assessed by multivariable logistic regression. In the logistic regression models, variables with P values of <0.1 in the univariable analysis were included in the multivariable analysis.</p><!><p>The accession numbers for all the M. abscessus isolates sequenced in this study are available at DDBJ/ENA/GenBank under BioProject PRJNA398137.</p><!><p>Supplemental material for this article may be found at https://doi.org/10.1128/AAC.02360-17.</p>

PubMed Open Access

Following DNA chain extension and protein conformational changes in crystals of a Y-family DNA polymerase by Raman crystallography

Y-family DNA polymerases are known to bypass DNA lesions in vitro and in vivo. Sulfolobus solfataricus DNA polymerase (Dpo4) was chosen as a model Y-family enzyme for investigating the mechanism of DNA synthesis in single crystals. Crystals of Dpo4 in complexes with DNA (the binary complex) in the presence or absence of an incoming nucleotide were analyzed by Raman microscopy. 13C, 15N labeled d*CTP, or unlabeled dCTP, were soaked into the binary crystals with G as the templating base. In the presence of the catalytic metal ions, Mg2+ or Mn2+, nucleotide incorporation was detected by the disappearance of the triphosphate band of dCTP and the retention of C* modes in the crystal following soaking out of noncovalently bound C(or *C)TP. The addition of the second coded base, thymine, was observed by adding cognate dTTP to the crystal following single d*CTP addition. Adding these two bases caused visible damage to the crystal possibly caused by protein and/or DNA conformational change within the crystal. When d*CTP is soaked into the Dpo4 crystal in the absence of Mn2+ or Mg2+, the primer extension reaction did not occur; instead a ternary protein/template/d*CTP complex was formed. In the Raman difference spectra of both binary and ternary complexes, in addition to the modes of d(*C)CTP, features appear due to ring modes from the template/primer bases being perturbed and from the DNA backbone, as well as from perturbed peptide and amino acid side chain modes. These effects are more pronounced in the ternary than in the binary complex. Using standardized Raman intensities followed as a function of time C(*C)TP population in the crystal maximized at about 20 min. These remained unchanged in the ternary complex but declined in the binary complexes as chain incorporation occurred.

following_dna_chain_extension_and_protein_conformational_changes_in_crystals_of_a_y-family_dna_polym

<!>EXPERIMENTAL PROCEDURES<!>1. The Raman spectrum of a Dpo4 single crystal<!>2. Raman intensity changes<!>3. Adding d*CTP to the hanging drop to form a ternary complex in the Dpo4 crystal<!>4. Extending the DNA chain by one base by soaking in dCTP in the presence of a divalent metal ion<!>5. Extending the DNA chain by two bases using d*CTP and dTTP<!>DISCUSSION

<p>DNA polymerases perform a diverse repertoire of biological functions including genomic replication, DNA damage repair, lesion bypass, and immunoglobulin diversification. So far, six families of DNA polymerases (A, B, C, D, X, and Y) have been classified and the Y-family established in 2001 is the newest.1, 2 Cellular DNA is frequently damaged by both endogenous and exogenous agents and processes. Although there are various DNA repair pathways, a large number of DNA lesions escape repair and stall replicative DNA polymerases and thus the replication machinery.3 However, the Y-family DNA polymerases can bypass DNA lesions, thereby rescuing cellular DNA replication. Notably, each living organism contains at least one Y-family DNA polymerase.2 For example, Sulfolobus solfataricus, an aerobic crenarchaeon that metabolizes sulfur and grows optimally at 80°C and pH 2-4,4 encodes one Y-family enzyme, DNA polymerase IV (Dpo4).4, 5 In addition to a typical polymerase core with a "right hand" geometry, consisting of Finger, Thumb, and Palm domains, Dpo4 also contains a fourth domain, designated as the "little finger" (LF) domain6 (Figure 1)7. In the ternary structure shown in Figure 1C, the active site of Dpo4 is relatively "loose" and solvent accessible when compared to the active site of a replicative DNA polymerase.6 Moreover, Dpo4, like all other Y-family DNA polymerases, is devoid of the proofreading exonuclease domain. Thus, Dpo4 catalyzes polymerization in damaged or undamaged DNA with low fidelity.8-16</p><p>So far, all kinetically characterized DNA polymerases catalyze nucleotide incorporation by following a minimal kinetic mechanism17 with a rate-limiting pre-catalytic conformational change.16-22 Since our recent stopped-flow fluorescence resonance energy transfer assays conclusively show that a pre-catalytic global conformational change associated with all four domains of Dpo4 is too fast to be rate-limiting, it was hypothesized that the rate-limiting conformational change corresponds to the subtle repositioning of active site residues which are critical for properly aligning two magnesium ions, the 3′-hydroxyl of the primer terminus, the α-phosphate of the incoming dNTP, and the conserved carboxylate residues within the active site.19 For a new phosphodiester bond formation during nucleotide incorporation, the primer 3′-OH makes an in-line nucleophilic attack on the α-phosphate of an incoming dNTP. Interestingly, the nucleotidyl-transfer reaction catalyzed by a truncated human DNA polymerase η (eta), another Y-family member, has recently been visualized at atomic level through time resolved X-ray crystallography.23 Previously, the large fragment of Bacillus stearothermophilus DNA polymerase I was found to catalyze several rounds of nucleotide incorporation in crystals.24 Here, single crystal Raman spectroscopy is employed to probe Dpo4-catalyzed DNA chain extension and relevant protein/DNA conformational changes within crystals.</p><p>The Raman method uses a Raman microscope that consists of an optical microscope that allows the operator to view a single crystal within a drop of holding solution mounted in a crystallization tray. A laser excitation beam travels on the optical axis of the microscope and is focused within the single crystal 25. Back scattered light from the focal volume travels back through the microscope and is carried by an optical fiber to a Raman spectrometer. The spectrometer provides a Raman spectrum from the focal volume. The basic experiment involves recording the spectrum of the Dpo4•DNA crystal, then adding the ligand/substrate to the drop that contains the crystal. The ligand soaks in the crystal and the Raman difference spectrum [Dpo4•DNA + ligand] minus [Dpo4•DNA] reveals chemical details of the reaction between the ligand and the Dpo4•DNA complex.</p><!><p>The DNA polymerase Dpo4 contains 352 amino acids and has an approximate molecular weight of 40 kDa. The enzyme was purified as described previously 15 and co-crystallized with DNA substrate prepared by annealing a 13-mer DNA primer and an 18-mer template strand shown in Scheme 1 following a published protocol.6 The DNA strands were purchased from Integrated DNA Technologies, Inc. Single crystals of the Dpo4•DNA complex were grown as described and suspended in a 5 μl hanging drop within a crystallization tray mounted on the stage of a Raman microscope.25, 26 Typically, crystals of the Dpo4•DNA complex were 500 × 150 × 150 μm and the excitation laser beam was focused through a flat 500 × 150 μm face.</p><p>2′-Deoxycytidine 5′-triphosphate (dCTP), thymidine 5′-triphosphate, and 15N and 13C labeled dCTP (d*CTP) were purchased from Sigma-Aldrich.</p><p>The Raman measurements were performed with the 647.1 nm line of a krypton laser; the laser power at the sample was 100 mW and the spectral data acquisition time was 100 seconds. Usually, a difference spectrum, the mathematical difference of two spectra, after and before ligand soaking into the crystal, was used to obtain the data.</p><!><p>A Raman spectrum of the crystalline DNA-enzyme complex is shown in Figure 2, where a signal to noise ratio of about 140:1 was achieved. This was the highest spectral quality obtained because adding ligand (e.g. dCTP) to the crystals resulted in deterioration in crystal morphology as is discussed in the next section. In Figure 2, the most intense features are due to well documented protein modes, e. g. amide I and III, and the aromatic amino acid side chains of Phe and Tyr.27 However, the four bases of DNA also make a significant contribution, as does the PO2− stretch of DNA backbone groups at 1094 cm−1 and the phosphodiester backbone has a stretching mode that contributes to the intensity at 784 cm−1. These assignments are listed in Table 1 and based on references.27-32 In Figure 2, the amide I feature at 1663 cm−1, the high intensity in the 1340 cm−1 region and the peak at 939 cm−1 attest to the presence of significant α-helix structure.27, 31, 32 This is consonant with the X-ray crystal structure that has 48% α-helix secondary structure. One band at 1058 cm−1 could not be assigned with certainty. Tentatively, this may be a mode from part of the DNA phosphodiester backbone that it is distorted away from the classic A or B forms, we have published data on a RNA polymerase that show features in 1000-1100 cm−1 that are assigned to distorted regions of DNA or DNA/RNA duplexes.32</p><!><p>In the Raman difference spectra discussed below, intensity changes in spectral features are monitored that have been assigned to dCTP, d*CTP, or protein and DNA modes from the Dpo4•DNA complex. There are three sources of intensity changes:</p><p>a) For dCTP (d*CTP) the intensity changes are due to population changes of the cytosine ring and the attendant triphosphate group inside the crystal.</p><p>b) Since the crystal forms a fixed set of axes, Raman dichroism can occur. This happens when groups change orientation in the crystal and thus change orientation with respect to the fixed orientation of the laser beam. For example, if purine and or pyrimidine rings change orientation during the experiment, the intensity of their Raman modes will change. It is maximal when the rings are at right angles to the incoming laser beam and minimal when the laser beam is parallel to the plane of the ring. This phenomenon is seen predominantly for base ring modes and amide I (mostly C=O) protein modes.</p><p>c) When adjacent purine and pyrimidine base modes stack or unstack this also affects their inherent Raman intensity. By analogy to absorbance spectroscopy, the changes in Raman intensity are termed Raman hypo- or hyper- chromism.27</p><!><p>A ternary complex consists of a protein, a DNA substrate and an incoming dNTP at the pre-insertion stage. If the dNTP becomes covalently linked to the primer via phosphodiester bond formation the result is a post-insertion binary stage. Scheme 2, adapted from Cox et al.,33 illustrates the reaction mechanism. Ternary complexes were formed by soaking 15N and 13C labeled dCTP (d*CTP) into the crystal in the absence of Mg2+ or Mn2+ ions using the active template when it will bind in the active site without phosphodiester bond formation.</p><p>Spectra at different times for "soaking in" d*CTP are shown in Figure 3A. The main d*CTP ring modes occur at 1215 and 763 cm−1 (the d*CTP Raman spectrum in aqueous solution is shown in Figure 5A) with a less intense mode at 1482 cm−1. In Figure 3A these peaks "grow in" with time, see below. An intense peak remains at 1126 cm−1 due to the d*CTP's triphosphate, confirming that the d*CTP has not been incorporated into the primer chain. It must be kept in mind that the d*CTP peaks seen in Figure 3A could contain a contribution from non-specifically bound ligand (i.e. not H-bonded to template G in the active site, see Scheme 2) as well as d*CTP correctly bound in the active site. It is also likely that non-specifically bound d*CTP contributes to the intensity changes seen in Figure 3 by perturbing groups within the Dpo4-DNA complex.</p><p>In Figure 3A several peaks due to dA and dG are identified. These are likely due to the adenine (dA4) in the template next to the dG5 that H-bonds to the incoming *dCTP ring and the two guanines in the template closest to the bound d*CTP (dG5 and dG6) (Scheme 1). The formation of the Watson-Crick base pair changes the environment of the neighboring dA4, dG5 and dG6, and hence the intensity of the ring modes. Since this is a fixed population the intensity changes are probably the effect of Raman dichroism although Raman hyper- or hypo- chromism may play a role.27 Other notable bands in Figure 3A come from amide I modes at 1671 and 1653 cm−1 and amide III modes at 1295 and 1254 cm−1. Likely Raman dichroism is responsible although changes in protein secondary structure may occur on addition of d*CTP. The intensity of the amide I modes in Figure 3A at 37 min is about 8% of the amide I mode in Figure 2. This implies that the change seen in Figure 3A is equivalent to the intensity of 0.08 times the total number of amino acids, 352, namely 29 amino acids. The amide intensity change in Figure 3A is equivalent to 29 amino acids. To generate this value it means that more than 29 amino acids have moved and changed in amide I intensity in the ternary compared to the binary complex. Possibly the d*CTP in the active site is causing dynamic protein fluctuations to be damped leading to a slight narrowing of the amide I profiles and this could be another cause of the apparent increase in amide I and III mode intensities. In Figure 3A, the appearance of the band at 785 cm−1 is due to a backbone phosphodiester stretch mode of DNA, and this represents 19% of the intensity of the 785 cm−1 band in Figure 2. This indicates that the DNA backbone has undergone significant conformational change in the ternary complex. The Raman data by themselves do not usually identify the specific locations of the observed protein and DNA changes.</p><p>In Figure 4 the intensities of the Raman peaks are plotted during soak-in using the inactive template which is dideoxy at the 3′C in the presence of 50 mM Mn2+. The peak heights are standardized with the Phe peak at 1004 cm−1 in the mother spectrum prior to the subtraction. For marker bands the following features are used; for triphosphate near 1120 cm−1, C at 1241 cm−1, and DNA phosphodiester backbone at 784 cm−1. All these marker bands show similar time dependence and the changes in the intensity are ascribed predominantly to population changes. This is an initial fast phase from 0 to about 20 minutes corresponding to maximum C(*C)TP population. This is followed by a plateau with possibly a small increase in intensity from 20-100 minutes – which was the limit of the crystal stability. In Figure 3, for d*CTP, at 19 minutes soak-in the difference spectrum is dominated by d*CTP features but by 37 minutes has more contribution assigned to template nucleic acid and protein features. At 37 min. the *C peak intensities are essentially unchanged from 19 minutes. However, the increase in amide III intensity at 1254 cm−1 and in ring modes associated with the nucleic acid scaffold suggests that slow conformational changes occur after the point where maximal *C population is reached.</p><p>When "soak-in" experiments were carried out on the Dpo4 binary complex, crystal cracking was invariably observed. One factor that may contribute to the cracking are protein and/or DNA conformational changes that, due to crystal packing forces, are not allowed the freedom to complete the changes that would be seen in solution. This would lead to unfavorable protein protein-constraints in the crystal that can be relieved by the crystal fragmenting.</p><p>The spectrum in Figure 3B is obtained under "soak-out" conditions. After "soaking in" d*CTP for 120 minutes the crystal was transferred to a holding solution that contained neither d*CTP nor Mg2+ or Mn2+. The difference spectrum was obtained after 40 minutes of "soak-out" conditions. The trace is almost a straight line showing that the d*CTP has left the crystal completely since no d*C ring modes remain. Moreover, the dA modes from the template disappear showing that the dA ring has resumed its position identical to that of the dA prior to soaking in d*CTP. Only very weak features persist. These are attributed to Tyr and Phe side chain modes, and to template dG ring closest to the d*CTP binding site that have not relaxed back completely to their pre-binding state in 40 minutes. The same conclusion can be drawn for the DNA backbone giving rise to the negative phosphodiester mode at 785 cm−1. That is the change in DNA backbone is greatly diminished after "soak-out".</p><!><p>In order to promote chain extension within the crystal, 50 mM Mn2+ ions (50 mM MnCl2)34 were added to the holding solution containing the crystal before soaking in 10 mM unlabeled dCTP. The resulting difference spectra are shown in Figure 6. At 62 minutes soak-in the spectrum seen in Figure 6A is obtained. The most intense peaks due to dC ring modes (see Figure 5B for the spectrum of aqueous unlabeled dCTP) are seen at 1529, 1293, 1255, and 785 cm−1. The medium intensity peak at 1122 cm−1 is due to dCTP's triphosphate and is evidence that some unreacted dCTP remains in the crystal. Many bands are assigned to dA and dG ring modes and these result from dCTP perturbing the environment of the bases on the primer and template, as in the ternary complex of the previous section. Similarly, binding bring about changes in Tyr side chain environment, evidenced by the "negative" 856 cm−1 feature, and the Tyr 10 and Tyr 48 in the active site are good candidates as a source of this change. The broad feature near 1658 cm−1 is due to Dpo4's amide I modes and shows that some perturbation to the protein α-helical secondary structure has also occurred although distinct amide III modes are not observed. The intense band at 785 cm−1 owes its intensity to a dC ring mode. Although a DNA backbone mode occurs at the same position it is unlikely that it contributes significantly since the insert band for *CTP shows weak intensity at 783 cm−1. Some of the intensity between 1030 and 1100 cm−1 is likely due to phosphodiester modes.32 For example the peak at 1094 cm−1 is due to the PO2− groups of the DNA backbone and this is evidence that the reaction is occurring since chain extension adds one additional PO2− to the chain.</p><p>Features due to pyrophosphate could not be unambiguously detected. The Raman spectra of pyrophosphate tetrabasic (Na4P2O7) in water over a range of pHs as shown in Supplementary Figure S1. The peaks with highest intensities are in the region 1000 – 1150 cm−1 where PO2− modes from the DNA backbone occur and specific peaks due to the formation of pyrophosphate are difficult to identify. The pyrophosphate peak near 716 cm−1 (Supplementary Figure S1) was not observed, possibly due to its relatively low intensity and broad or due to the pyrophosphate product leaving the crystal.</p><p>In Figure 6A there are undoubtedly contributions from dCTPs that are not specifically bound in the active site and have not been covalently linked to the primer chain. Thus, "soaking out" was employed. After 62 min "soaking in", we placed the crystal in holding solution that did not contain dCTP (50 mM Mn2+ metal remains). After 30 minutes of "soaking out" spectrum 6B was obtained. The intensity of the dC modes has diminished by about 65% showing that about one third of the original dCTP's in the crystal has been incorporated in the primer chain. The triphosphate peak at 1125 cm−1 is small showing that now only a small amount of unreacted dCTP remains. Many perturbations persist but, compared to the ternary complex, changes in the protein conformation are less, the broad amide I band at 1658 cm−1 (Figure 5A) is replaced by three smaller peaks at 1681, 1667 and 1648 cm−1. These are tentatively assigned, at least in part, to the "C=O modes" from the new G-C Watson-Crick base pair that has been formed.</p><p>Figure 7 shows the intensity variation with time of the *C ring mode under conditions when the reaction is occurring: this is with the reaction template and 50 mM MnCl2. The population of *C, gauged by ring mode intensity at 1214 cm−1 rises rapidly to about 23 minutes and then decreases. The triphosphate population using the 1119 cm−1 mode as a marker also rises rapidly and peaks at 23 minutes, and declines as *CMP is incorporated in the DNA primer strand. The similar maxima in both traces in Figure 7 suggest that catalysis becomes dominant at about the point of "full soak-in". The decrease in *C population in Figure 7 indicates that nonspecifically bound *CTP is expelled from the crystal after incorporation begins at around 23 minutes and suggest that a conformational change is occurring that leads to less favorable binding of non-specific *CTP.</p><p>It is noteworthy that soaking in dCTP in the presence of 1 mM dCTP and 50 mM Mn2+ resulted in only specific binding. The intensities of the cytosine peaks in the difference spectrum after soaking in the crystal in a solution with 1 mM labeled dCTP after 40 min are the same as the intensities of the cytosine peaks after the soaking out experiment (Figure 6B) obtained by washing the crystal during 40 minutes in a solution without dCTP (data not shown).</p><p>There is no way of separating the effects of specific (i.e. *C Watson-Crick binding to template G) and non-active site d*CTP binding in Figure 3. However, the spectrum in Figure 6A suggest that there are fewer intensity changes for binary complex with non-covalently bound dCTPs than in Figure 3A for the ternary complex with non-specifically bound d*CTPs. This is supported by data in Figure 8A which compare covalently labeled d*CTP in a pure binary complex (Figure 6B) with that prior to soak-out where non-covalent d*CTPs are also present. Although the quality of the data is not optimal, in Figure 6B, after soak-out, compared to the ternary complex in Figure 3 only minor changes are seen in the amide I region near 1650 cm−1 and the DNA backbone marker band near 780 cm−1. In total the data suggest that in the binary complex the protein and DNA backbone changes are probably minor compared to those seen in the ternary complex.</p><!><p>In order to obtain a binary complex where one d*C was added to the primer, the protocol in the previous section was repeated using 13C, 15N labeled dCTP (d*CTP). The results are shown in Figure 8A. d*CTP at 7 mM with 50 mM Mn2+ was soaked in for 40 min and then soaked out for 30 min in holding solution that did not contain d*CTP. Although the data quality in Figure 8A is inferior to that in Figure 6, the spectrum of the washed out crystal resembles closely the spectrum in Figure 6B except that the main dC ring modes are downshifted e.g. the mode around 785 cm−1 is "downshifted" 22 cm−1 due to the 13C and 15N substitutions (d*C). Figure 8A confirms that for the binary complex the changes in protein and DNA conformation are small compared to those for the ternary complex since the amide I and III features and DNA backbone feature near 785 cm−1 are weak in the difference spectrum. This binary crystal was then soaked in 2 mM dTTP in the presence of 50 mM Mn2+ for 100 min. Then, following soaking in dTTP the crystal was placed in holding solution that has no dTTP for 50 min; the resulting difference spectrum is seen in Figure 8B. The multiple soak process gave crystals that had cracked and lost most of their crystalline appearance. However, a difference Raman spectrum could be obtained that showed both d*C and dT ring modes, strongly suggesting that both d*C and dT have been incorporated into the primer chain. No evidence for unreacted triphosphate is seen in Figure 8B.</p><p>It is of high interest that intense dA and dG ring modes occur in Figure 8B as does a peak from the DNA phosphodiester backbone at 784 cm−1. The high intensities suggest that upon the second soak with dTTP much larger changes are seen in the primer and/or template than those seen on the first d*CTP incorporation. This is interpreted as resulting from primer and template translocation, fully or partially, through the active site of Dpo4 in order to bring the template dA base into register with the catalytic machinery.</p><!><p>The Raman difference spectra allow us to observe structural and conformational changes in a crystalline Dpo4•DNA complex occurring when the DNA polymerization reaction is catalyzed by the Y-family polymerase Dpo4. The Raman peaks in the difference spectra with the highest intensity are usually due to the added nucleotide, dCTP, d*CTP or dTTP, but peaks corresponding to changes in the protein/DNA conformations or template/primer base perturbations can also be observed. For the latter larger changes are observed in ternary compared to binary complexes. The evidence that chain incorporation of dNTP is based on the fact that the C, *C or T ring modes remain in the crystal after extensively "soaking out" the non-covalently bound ligands. This is supported by the observations that there is little evidence for the reactive dNTP triphosphate group after "soak-out". It is noteworthy that Xu et al.19 proposed that a local active site rearrangement is a rate-limiting conformational change step driving a single correct dNTP incorporation process. However, their studies were in aqueous solution and we will need to undertake future rapid mix – rapid quench experiments in solution as detailed below to examine conformational changes on the millisecond time scale.</p><p>Running the reaction in the crystal comes at a price. Significant degradation of the crystal morphology is observed. This could prevent future parallel X-ray crystallography experiments from being carried out which could use the Raman experiments as reference points for flash freezing in order to carry out time-dependent X-ray analysis32, 35-37. For future experiments the approach detailed by Nakamura et al.23 will be employed, in this study of the reaction investigating η DNA polymerase from the Y-family they co-crystallized the substrate and the enzyme in the absence of divalent metal ions, which prevents the chain incorporation reaction from occurring. They then triggered the reaction by soaking in Mg2+ and could observe high resolution maps by flash freezing at different time points during metal "soak-in". It is possible that a major cause of the crystal cracking seen in our experiments is the conformational changes the enzyme wishes to make to achieve translation being opposed by crystal packing forces.</p><p>Clearly it would be of great interest to compare the reaction in solution with that in the crystal. Until very recently solution studies using normal (non-resonance) Raman spectroscopy were technically impossible. However, a flash-freezing protocol that can examine the reaction in aqueous solution in the millisecond time scale has been developed in our laboratory 38 and this approach will be used to follow dNTP incorporation catalyzed by Dpo4 DNA polymerase in aqueous solution.</p><p>Our Raman database on the nucleic acid polymerases is being extended. The present DNAP is the smallest, a single subunit about 40 KDa containing a DNA primer and an 18 nucleotide template. Recently, results on a 115 KDa RNA polymerase from the N4 phage virion were presented where the initiation of the RNA chain was studied.39 This in crystallo initiation reaction was about 3 times faster than the present DNA chain extension. The simultaneous soak-in of GTP and ATP was observed to plateau at 7 minutes. A small protein conformational change was seen on about the same time and this was assigned to movement of the α-helix near the active site. Functional active site formation about two Mg2+ ions was also seen in less than 10 minutes as evidenced by Asp side chains liganding to Mg2+. Triphosphate intensity reduction showed that catalysis was complete shortly after 10 minutes. Basu and Murakami40 used the Raman kinetic data to flash freeze and solve the structures of the N4 crystals at time points between the 0 to 10 minutes of soak in. The predictions from Raman crystallography were borne out, they were able to obtain high quality X-ray structures for intermediates during the formation of the covalent nucleic acid backbone bond. The trio of polymerases is completed by recent Raman studies on a RNAP from the bacteria Thermos thermophilus (Tth RNAP) (unpublished work, this laboratory). This is a 5 subunit 380 KDa 'machine'. Time resolve in crystallo Raman data showed that the base to be incorporated, GTP, soaks in and "plateaus" after 30 minutes. This triggers a large reversible change in protein conformation probably from functionally important alpha helices that flank the active site and the crab-like "pincers" that form a channel to the active site. The protein conformational change is accompanied by a modest and reversible change in DNA backbone from the RNA/DNA hybrid between 0 and 60 minutes. The conformational changes lead to GTP incorporation between 65 and 100 minutes. Covalent bond formation occurs, apparently, shortly after the nucleic acid skeleton has translocated through the active site channel. Remarkably, the Tth RNA polymerase in the crystal appears to be primed for a second round of nucleotide triphosphate incorporation.</p><p>Thus, in all these NAPs the cognate NTP can be followed soaking into its crystal and starting or incorporating into a nucleic acid chain. The time scales range from minutes to 10s of minutes, probably 10 thousand to 100 thousand times slower than in solution. In each instance the time points of "interesting intermediates" that are candidates for X-ray analysis can be identified. In the case of RNAP from N4 this prediction has been met and their intermediates characterized by X-ray crystallography. DNAP and Tth RNAP reactions have yet to be studied by X-ray.</p>

PubMed Author Manuscript

Chemical Synthesis of Human Selenoprotein F and Elucidation of Its Thiol-Disulfide Oxidoreductase Activity

Selenoprotein F (SelF) is an endoplasmic reticulum-residing eukaryotic protein that contains a selenocysteine (Sec) residue.It has been suggested to be involved in a number of physiological processes by acting as a thiol-disulfide oxidoreductase, but the exact role has remained unclear due to the lack of a reliable production method. We document herein a robust synthesis of the human SelF through a three-segment two-ligation semisynthesis strategy. Highlighted in this synthetic route are the use of a mild desulfurization process to protect the side-chain of the Sec residue from being affected and the simultaneous removal of acetamidomethyl and p-methoxybenzyl protection groups by PdCl 2 , thus facilitating the synthesis of multi-milligram of homogenous SelF. The reduction potential of SelF was determined and the thiol-disulfide oxidoreductase activity was further supported by its ability to catalyze the reduction and isomerization of disulfide bonds.

chemical_synthesis_of_human_selenoprotein_f_and_elucidation_of_its_thiol-disulfide_oxidoreductase_ac

Introduction<!>Synthetic strategy<!>Chemical synthesis of SelF and its Trx-like domain<!>Probing of the thiol-disulfide reductase activity of SelF<!>Conclusion

<p>Selenoprotein F (SelF), or the 15-kDa protein (Sep15), is an endoplasmic reticulum (ER)-residing eukaryotic protein containing the 21 st essential (proteinogenic) amino acid selenocysteine (Sec, U). 1 An increasing number of studies have linked SelF gene polymorphisms and SelF dysregulation to various diseases, including several types of cancer, AIDS, and neurodegeneration, which reveals the importance of SelF's physiological functions. 2 As shown in Figure 1A, the mature human SelF consists of 134 amino acid residues (aa), including an N-terminal Cys-rich domain and a C-terminal thioredoxin (Trx)-like domain. Although with no typical ER retention peptide sequence, SelF is able to bind the UDP-glucose: glycoprotein glucosyltransferase (UGGT)-a large chaperone protein in the ER, via its N-terminal Cys-rich domain, thus also called the UGGT binding domain; 3 The C-terminal Trx-like domain contains a unique CGU redox motif in a dynamic loop, a key structure rendering SelF a competent thiol-disulfide oxidoreductase in the ER. 4 As such, SelF has been suggested to play a role in the quality control of the ER, by either rearranging (isomerase function) or reducing incorrectly formed disulfide bonds (reductase function) in misfolded glycoproteins bound to UGGT, 5, 6 but its exact biological</p><p>function is yet to be elucidated. Along this line, the in vitro characterization of SelF is generally missing due to the lack of reliable recombinant expression technique, and most studies are carried out with its Sec-to-Cys homologue. 4 In this context, the disulfide pairing mode of SelF(U65C) has been elucidated in our previous work through site-directed mutagenesis and enzymatic digestion (see Figure 1B), S1. [18][19][20][21][22][23][24][25][26][27][28][29] Despite these significant achievements, the currently used synthetic routes usually involve either no or limited post-ligation treatment, as a result most of the synthetic Sec-proteins either are small in size (usually <100 aa and with few Cys/Sec residues) or contain a Sec residue at the C-terminal region, which can be obtained by EPL through ligation with a synthetic peptide bearing an N-terminus Sec (Table S1, Entries 4, 8 and 11). As such, there is a clear need to develop a more straightforward strategy for the synthesis of complex selenoproteins having multiple Cys and Sec residues, like in the case of SelF (7 Cys and 1 Sec). We disclose herein the synthesis of SelF, where the Sec65 residue resides in the middle of its sequence, making it inconvenient to apply EPL. To maximize the overall synthetic yield, we adopted a three-segment two-ligation strategy, where Ala75 was chosen as one of the ligation sites as there is no Cys available in the C-terminal region, and a desulfurization step will be required (Figure 1B). While selective deselenization in the presence of Cys has been routinely carried out using tris(2-carboxyethyl)phosphine (TCEP), 30, 31 the selective desulfurization in the presence of Sec has not been reported to the best of our knowledge, which is one of the key challenges in our synthetic endeavor. With the current strategy, multi-milligram of homogeneous synthetic SelF was obtained, which allowed the elucidation of its thiol-disulfide oxidoreductase activity, thus providing evidences for its involvement in the quality control of the ER.</p><!><p>In analogy to the synthesis of the SelF(U65C) analogue, 7 the full-length protein was disconnected into three segments at Gly41-Cys42 and Gln74-Ala75. Noted that the Gln74 was mutated to an Ala residue in order to obtain a stable peptide hydrazide segment, and according to previous experience 7 this mutation should not affect the protein folding and function (vide infra). Unlike most of the reported synthetic Seccontaining proteins where the Sec residue is placed in the Nterminus of the peptide, i.e. the ligation site, we opted to place it as an internal residue in segment 2 and the side-chain was protected with a p-methoxybenzyl (Mob) group. Moreover, the Cys75 residue of segment 3 will have to be desulfurized to give the native Ala residue, and for this purpose, all the Cys residues in segment 2 were protected with an acetamidomethyl (Acm) group, which also prevents the lactamization of the resulting thioester. As such, segments 1 and 2 were obtained with standard Fmoc-SPPS, and segment 3 through N-terminal His-SUMO fusion protein expression and Ulp1 cleavage.</p><!><p>With all peptide segments in hands, the first NCL reaction was carried out between segments 2 and 3, which completed within 2 h (Figure 2). At this stage, the initial plan for a one-pot desulfurization at Cys75 failed due to the increased amount of 2,2'-Azobis[2-(2-imidazolin-2-yl)propane]dihydrochloride (VA-044) required in this case (vide infra). Instead, peptide 4 was purified and subjected to standard desulfurization conditions (i.e. 30-40 eq VA-044, 200 mM TCEP and 5% t-BuSH) (t-BuSH = tert-butylthiol). 34 Surprisingly, the unwanted peptide 5' with desulfurization at Cys75 and deselenization at Sec65-despite being Mob-protected, was the major product (Figures S13-14).</p><p>The peptide S2 (the hydrazide precursor of segment 2) was then used as a model peptide and its stability was tested against each desulfurization component, and no significant change was observed when incubating with TCEP or t-BuSH (Figures S15-16), which led us to conclude that VA-044 could most probably be the reason. To our delight, when reducing the amounts of VA-044 (to 2.5 eq), the desired desulfurized peptide 5 was obtained as the major product at 30 min, with an isolated yield of 39.5% (Figure S22). 35 It seems that the presence of the Mobprotected Sec residue in the sequence speeded up the desulfurization, as the same process would usually take longer (4-24 h) and need more VA-044 (e.g. 50 eq.) in the Cyscontaining homologue peptide.</p><p>7</p><p>Next, the global Acm removal of peptide 5 was attempted, and the Mob group of Sec65 was supposed to be retained at this stage to prevent possible deselenizaiton in the following NCL. While the Acm-removal with AgOAc went well during the synthesis of SelF(U65C), 7 it gave a complex mixture in this case, with all Acm and Mob groups being removed as well as severe peptide truncation (Figures S24-S25). We then switched to the PdCl 2 -dithiothreitol (DTT) method, 36, 37 and by accident we noticed that an increased amount of Pd 2+ could lead to a simultaneous Acm and Mob removal of a model peptide S5 (Figures S26-S27). It is worth noting that there is literature presence using Pd 0 for the deprotection of an allyl group from Sec in aqueous solution. 38 Gratifyingly, when using 150 eq of PdCl 2 quantitative global Acm/Mob removal of peptide 5 was accomplished (Figure S29), subsequently, the excess Pd reagent was removed by DTT and extensive washing, which after incubation with ascorbate and TCEP to reduce any possible peptide diselenide dimer afforded peptide 6. The discovery that the Mob group can be facilely removed by aqueous PdCl 2 solution is remarkable considering the generally harsh conditions required in the literature procedures, 39, 40 like the use of dimethyl sulfoxide (DMSO) 41 or 2,2'-dithiobis(5nitropyridine) (DTNP) 42, 43 in TFA. We envision that the chemistries relating to the selective desulfurization in the presence of Sec and simultaneous Acm/Mob removal disclosed herein may find further applications in the CPS field, where the use of Sec either for selenoprotein synthesis or as a Cys analogue, has become an increasingly popular strategy. 28, 31, 44, 45 The second ligation between peptide 1 and 6 was carried out in the presence of 0.1 M ascorbate and a reduced concentration of TCEP (5 mM) to prevent the now-free selenol side-chain from being removed, [46][47][48] furnishing the full-length protein 7. After 12 h, the ligation mixture was directly subjected to the redox refolding buffer, and the refolded protein 8 was obtained after HPLC purification with an isolated yield of 13.7% over two steps (~1.5 mg). Noted that the presence of thiolactone derivative of peptide 1 (indicated with * in Figure 3) was probably one of the reasons for the slow ligation rate and the slightly lowered recovery yield. In principle installation of Acm protection group at the Cys sidechain during the synthesis of peptide 1 could avoid the formation of thiolactone, we have, however, observed significant aspartimide formation at the Asp-Cys(Acm) sequence (data not shown), as also reported in the literature. 49 Nevertheless, with the current route enough amounts of protein was obtained for further studies, and the residual Pdcontent in 8 determined using ICP-MS was negligible (<0.061%). The proper folding of 8 was confirmed by ESI-MS (~-8 Da vs. 7) and the CD spectrum (vs. the expressed Cys analogue 7</p><p>). And importantly, the presence of a mixed selenenylsulfide bond between Cys63-Sec65 was established via a consecutive trypsin/chymotrypsin digestion (Figure S47), which agrees well the reported NMR structure of the fruit fly Sep15 protein as well as the SelF(U65C) homologue (both with a disulfide bond instead). Meanwhile, following a similar synthetic strategy, e.g. NCL between a short Sec(Mob)-containing peptide thioester segment S9 and segment 3, selective desulfurization, simultaneous Acm/Mob removal and refolding, the Trx-like domain SelF(63-134)(Q74A) 9 was obtained conveniently (Figures S34-44), which set the stage for a comparative functional study with the full-length protein 8 (vide infra).</p><!><p>With the synthetic protein in hand, it is now possible to determine its redox potential, a key parameter gauging its ability to act as a native thiol-disulfide oxidoreductase. 50, 51 Using a glutathione (GSH)-glutathione disulfide (GSSG) reference buffer, the fractions of the reduced SelF at selected GSH/GSSG ratios were measured by HPLC and plotted against the redox potential poised by the corresponding redox buffers (Figure 4A, Figure S50). The resulting data was fitted by Nernst equation and the potential of the SleF was established as -222 mV, whereas the Cys homologue has a redox potential of -205 mV. It agrees well with the reported value of the fruit fly Sep15 (-225 mV), which has a disulfide rather than a selenenylsulfide in the active-site redox motif. Encouraged by this result, we probed the thiol-disulfide oxidoreductase activities of the synthetic SelF protein. Firstly, the disulfide reductase activity was determined using insulin as a model substrate, 52 where the cleavage of the disulfide bond connecting chain A and chain B will lead to protein aggregation and the resulting turbidity can be followed by absorption spectroscopy for assessment. As shown in Figure 4B and ). It is worth noting that the double mutant SelF(U65C/Q74A) shows a similar, albeit low, reactivity compared to SelF(U65C), suggesting that the effect of mutation at this site (Gln74) is minimum. Interestingly, the Trx-like domain of SelF 9, with the key CGU motif, is also active in reducing the disulfides of insulin. Altogether these data indicate that SleF is a viable disulfide reductase and the presence of Sec in the redox motif is clearly the key for this activity. Moreover, we also tested the protein disulfide isomerase activity of the synthetic SelF by assessing its ability to catalyze the refolding of the scrambled RNase A. As shown in Figure S57, only in the case of SelF a small but appreciable amount of folded RNase A can be observed after 2 h incubation. While RNase A may not be the native substrate as presented in the ER, 4 the current data suggests that SelF could indeed contribute in the isomerization of disulfide bonds of the (misfolded) glycoprotein substrates of UGGT.</p><!><p>In summary, we have developed a robust synthetic strategy affording the full-length human SelF protein that contains an internal Sec residue and seven other Cys residues. Notable challenges addressed in the synthetic route are 1) the use of a reduced amount of VA-044 during desulfurization to protect the side-chain of the Sec residue from being affected and 2) the simultaneous removal of Acm and Mob protection groups by PdCl 2 , thus facilitating the synthesis of multimilligram of homogenous SelF for biological studies. The critical selenenylsulfide bond Cys63-Sec65 in the CGU motif of SelF was unambiguously established, representing the first experimental evidence for such connectivity in SelF. The redox potential of the synthetic protein was determined to be -222 mV, typical for those of the thiol-disulfide oxidoreductase. 4, 53, 54 We demonstrate that SelF is capable of catalyzing the disulfide reduction and to a less extent, isomerisation, in vitro, and the Sec residue is the key for these functions. These data suggest that SelF, together with UGGT, can indeed play a crucial role in the quality control of ER. Moreover, the synthetic strategy developed herein may find broad applications in the synthesis of other complex selenoprotiens with multiple Cys/Sec residues, and thus helping in the elucidation of their physiological functions.</p>

ChemRxiv

Ta3N5-Pt nonwoven cloth with hierarchical nanopores as efficient and easily recyclable macroscale photocatalysts

Traditional nanosized photocatalysts usually have high photocatalytic activity but can not be efficiently recycled. Film-shaped photocatalysts on the substrates can be easily recycled, but they have low surface area and/or high production cost. To solve these problems, we report on the design and preparation of efficient and easily recyclable macroscale photocatalysts with nanostructure by using Ta 3 N 5 as a model semiconductor. Ta 3 N 5 -Pt nonwoven cloth has been prepared by an electrospinning-calcination-nitridation-wet impregnation method, and it is composed of Ta 3 N 5 fibers with diameter of 150-200 nm and hierarchical pores. Furthermore, these fibers are constructed from Ta 3 N 5 nanoparticles with diameter of ,25 nm which are decorated with Pt nanoparticles with diameter of ,2.5 nm. Importantly, Ta 3 N 5 -Pt cloth can be used as an efficient and easily recyclable macroscale photocatalyst with wide visible-light response, for the degradation of methylene blue and parachlorophenol, probably resulting in a very promising application as ''photocatalyst dam'' for the polluted river. Environmental problems associated with harmful pollutants in water pose severe threats to human health. Among the water treating methods, photocatalysis offers a ''green'' and energy saving technology for completely eliminating organic pollutants in water [1][2][3] . A prerequisite for the development of photocatalysis application is to gain access to excellent photocatalysts. Generally, two kinds of photocatalysts have been well developed. One kind is the nanosized semiconductor photocatalysts, including nanoparticles 4 , nanotubes 5 , nanowires 6 , nanosheets 7 , nanospheres 8 and nanocomposites [9][10][11] . Recently, we have also prepared some nanosized photocatalysts, such as Bi 2 WO 6 superstructures 12,13 , and AgBr-Ag-Bi 2 WO 6 nanojunction system 14 . They always show relatively high photocatalytic activity due to their nanoscaled particle size and large specific surface area. Unfortunately, it is very difficult to recycle these nanosized photocatalysts in practical application (such as degrading organic pollutants in lake and/or river), resulting in second-contamination and limiting their largescale application. The other kind is semiconductor films on the substrates, such as nanoparticles-based composite films on ITO glass 15,16 , nanowires/nanotubes-based film grew on metal foil 17,18 . These film-shaped photocatalysts on the substrate can be easily recycled, but they suffer from the problems, such as relatively low surface area and/ or high production cost. Thus, it is quite necessary to develop novel kind of photocatalysts. Ideal photocatalysts should have a broad range of visible-light response, superior photocatalytic activity, high photostability, low cost and easily recycling characteristics, and etc.It is well known that micro/nano-fibers and nonwoven cloth can be easily prepared via electrospinning technique that represents a simple, cost-effective and versatile method for the large-scale production of fibers. Traditional micro/nano-fibers are polymer or polymer/inorganic composite fibers [19][20][21] . Few kinds of semiconductor nanofibers including TiO 2 22 , Bi 4 Ti 3 O 12 23 , TiO 2 /SnO 2 24 and GaN 25 have been prepared for photocatalysis or photodetector. In photocatalytic application, semiconductor nanofibers have also suffered from the problems such as relatively low surface area and recycle difficulty. It should be noted that macroscale nonwoven cloth are usually composed of polymer or polymer/inorganic composite fibers, and they have already found use in applications (such as drug carriers, tissue engineering and ultrafiltration) and can be easily recycled. If semiconductor nonwoven cloth is composed of nanofibers that are constructed from semiconductor nanoparticles

ta3n5-pt_nonwoven_cloth_with_hierarchical_nanopores_as_efficient_and_easily_recyclable_macroscale_ph

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

<p>with plenty of hierarchical nanopores, it will have both high surface area and easily recycling characteristics for photocatalytic application. These features trigger our interest in the novel concept of developing efficient and easily recyclable macroscale semiconductor photocatalysts with nanostructure.</p><p>Among semiconductor photocatalysts, Ta 3 N 5 with a narrow band gap of approximately 2.1 eV can absorb and utilize a large fraction of visible light up to 600 nm, and Ta 3 N 5 nanomaterials [26][27][28][29] and/or films [30][31][32][33][34][35] have been prepared as visible-light-driven (VLD) photocatalysts. Herein, by using Ta 3 N 5 as a model semiconductor, we report the design and preparation of Ta 3 N 5 -Pt nonwoven cloth that is composed of nanofibers constructed from Ta 3 N 5 nanoparticles, hierarchical nanopores and Pt nanoparticles. The macroscale Ta 3 N 5 -Pt nonwoven cloth exhibits large surface area (23.1 m 2 g 21 ). Furthermore, it can be used as an efficient, stable and easily recyclable macroscale semiconductor photocatalyst with nanostructure, for the degradation of both methylene blue (MB) dye and parachlorophenol (4-CP) under visible light irradiation. This finding promotes the design and development of novel kind of macroscale photocatalysts with nanostructure for practical application, for example, degrading pollutants in lake and/or river.</p><!><p>Synthesis and characterization of the nonwoven cloth. Ta 3 N 5 -Pt nonwoven cloth was prepared by an electrospinning-calcinationnitridation-wet impregnation method, as demonstrated in Figure 1. First step was to prepare PVP/Ta 2 O 5 /Ta(OEt) 4 composite nonwoven cloth, by electrospinning the solution (ethanol-acetic acid mixture (3.351, volume ratio) containing 10 wt% tantalum ethanolate (Ta(OEt) 4 ) and 5 wt% polyvinylpyrrolidone (PVP, M W < 1300000 g mol 21 )) at a high voltage of 15 kV, and followed by the hydrolysis process. The as-prepared PVP/Ta 2 O 5 /Ta(OEt) 4 composite nonwoven cloth is white, and its typical photograph (area: ,10 3 9.5 cm 2 ) is shown in Figure 2a. In fact, in our case, the area of the nonwoven cloth can be easily tuned in a broad range (10 24 , 1 m 2 ) by changing the collecting region of aluminum foil during the electrospinning process. This macroscopic nonwoven cloth is composed of plenty of individual straight fibers with smooth surface and diameters ranging from 250 to 300 nm, as revealed in scanning electron microscopy (SEM) images (Figures 2b, 2c).</p><p>The second step was to calcine the composite nonwoven cloth at 600uC in air for 6 h, for removing polymer component and obtaining inorganic nonwoven cloth based on Ta 2 O 5 fibers. After the calcination process, this Ta 2 O 5 nonwoven cloth still has the macroscopic morphology (Figure 2d) similar to that (Figure 2a) of the as-prepared PVP/Ta 2 O 5 /Ta(OEt) 4 composite cloth, indicating that the calcination process has no obvious adverse effect on the macroscopic morphology. However, after the calcination, there are obvious changes in the microstructure of the cloth. This Ta 2 O 5 nonwoven cloth consists of the pores and bent fibers that interweave and/or stick together (Figures 2e, 2f), which results from the disappearance of PVP component and the high-temperature anneal of Ta 2 O 5 component. Furthermore, the diameters of Ta 2 O 5 fibers shrink to 200-250 nm, and the fibers with rough surface are composed of nanoparticles with diameter of about 10 nm (Figures 2e, 2f).</p><p>The third step was to further nitridize Ta 2 O 5 cloth at 800uC under NH 3 flow (500 mL min 21 ) for 8 h to obtain Ta 3 N 5 cloth. It should be noted that Ta 2 O 5 nonwoven cloth was tailored to ,4.5 3 3 cm 2 to fit the small inter-diameter (,5 cm) of the quartz furnace tube during the nitridation process. Obviously, such Ta 3 N 5 cloth is still freestanding and can be easily transferred and/or recycled for further practical application. Its color turned from white to red-orange, as demonstrated vividly in Figure 2g, indicating the conversion from Ta 2 O 5 to Ta 3 N 5 . SEM images (Figures 2h, 2i) reveal that Ta 3 N 5 cloth is also composed of hierarchical pores (diameter: 0.2-1 mm) and fibers. The diameters of Ta 3 N 5 fibers were reduced to 150-200 nm, and these fibers also interweave and/or stick together. Importantly, Ta 3 N 5 fibers are comprised of plenty of nanoparticles with diameters of ,25 nm and nanopores with diameters of ,15 nm (Figure 2i), probably resulting in high surface area. Further information about Ta 3 N 5 fibers was obtained from the transmission electron microscopy (TEM) images (Figures 2j-l). The TEM images (Figures 2j-l) confirm that Ta 3 N 5 cloth is composed of fibers that are constructed from nanoparticles and nanopores, which agrees well with that revealed by the SEM images. The high-resolution TEM image (Figure 2l) taken from one nanoparticle in the fiber (Figure 2k) shows clear lattice fringes with an interplane spacing of 0.363 nm, which is corresponding to the (110) crystal plane of monoclinic Ta 3 N 5 . It should be noted that when the nitridation temperature was above 900uC, significant collapse of fibers occurred, resulting in the distortion of Ta 3 N 5 cloth (Supplementary Figure S1).</p><p>At last, Ta 3 N 5 cloth was decorated with Pt nanoparticles (,0.5 wt%) by the photocatalytic reduction of H 2 PtCl 6 in methanol aqueous solution under a 300 W xenon lamp light irradiation. The decoration process of Pt has no obvious effects on the shape of cloth, as confirmed by photo (the inset of Figure 2m) and SEM image (Figure 2m). However, from the TEM image (Figure 2n), one can find that there are plenty of nanoparticles with the size of about 2.5 nm on the surface of fibers. The high-resolution TEM image (Figure 2o) shows clear lattice fringes with an interplane spacing of 0.225 nm, which is corresponding to the (111) crystal plane of cubic Pt. Thus, one can confirm the formation of Ta 3 N 5 -Pt nonwoven cloth with well-defined heterostructure.</p><p>The phase and pore structure, and optical characterizations. The phase structure of Ta 3 N 5 -Pt nonwoven cloth was further investigated. Subsequently, the nitrogen adsorption/desorption isotherms of Ta 3 N 5 cloth and Ta 3 N 5 -Pt cloth were investigated (Figure 3b). The Brunauer-Emmett-Teller (BET) surface area of Ta 3 N 5 cloth is calculated to be 22.0 m 2 g 21 . After the deposition of Pt nanoparticles, Ta 3 N 5 -Pt cloth exhibits a slight increase of BET surface area (23.1 m 2 g 21 ). Thus, although both Ta 3 N 5 cloth and Ta 3 N 5 -Pt cloth are macroscale, they have high surface area compared with those of bulk powders (Figure S4) and traditional fibers, resulting from their nanostructure. Moreover, the pore size distributions, which are calculated from the desorption branches, reveal the existence of nanopores in both Ta 3 N 5 cloth and Ta 3 N 5 -Pt cloth (the inset of Figure 3b).</p><p>The nanopores in Ta 3 N 5 cloth have the diameter of about 15 nm, while those in Ta 3 N 5 -Pt cloth have the diameter of about 13 nm, which agrees with that revealed by the SEM and TEM images (Figures 2h-k, 2m). The presence of nanopores in fibers and macro-pores among fibers may greatly improve the physicochemical properties and/or serve as transport paths for small molecules.</p><p>The optical properties of Ta 2 O 5 , Ta 3 N 5 and Ta 3 N 5 -Pt cloths were studied by using an UV-Vis-NIR spectrometer (Figure 3c). The spectrum of Ta 2 O 5 cloth is similar to what has been reported previously for Ta 2 O 5 samples 26 , and it exhibits a short-wavelength absorption edge at approximately 330 nm. Importantly, Ta 3 N 5 cloth shows a large red shift from 330 to 600 nm, due to the band gap narrowing caused by the substitution of N for O atoms 36 , which agrees well with the reported value for the bandgap (Eg < 2.1 eV) of Ta 3 N 5 samples 26,30 . Furthermore, after the decoration of Pt, no obvious change of absorption spectrum has been observed. These facts indicate that both Ta 3 N 5 cloth and Ta 3 N 5 -Pt cloth have a broad region of visiblelight photo-response, and therefore can be expected to act as excellent VLD photocatalysts.</p><p>Photocatalytic activity. In order to investigate the potential of Ta 3 N 5 -Pt cloth as VLD photocatalyst, the photocatalytic activity of macroscopic Ta 3 N 5 -Pt cloth was evaluated by immersing the cloth in the solution containing MB dye or colorless 4-CP as the model pollutant (Figure 4). For comparison, bulk Ta 3 N 5 powder, bulk Ta 3 N 5 powder decorated with Pt nanoparticles (denoted as bulk Ta 3 N 5 -Pt powder), and mesoporous SiO 2 powder decorated with Pt nanoparticles (denoted as SiO 2 -Pt powder) were also prepared and used as the photocatalysts. These bulk powders were also characterized by XRD, SEM, BET, UV-Vis-NIR spectrometer or TEM (Supplementary Figures S2-S6).</p><p>When MB dye was used as the model of organic pollutant, the photocatalytic activity of macroscopic Ta 3 N 5 -Pt cloth was evaluated by immersing the cloth (20 mg, size: ,2.5 3 3.5 cm 2 ) in 60 mL aqueous solution containing 10 mg L 21 methylene blue (MB) dye under visible light irradiation (l . 400 nm). When dissolved in distilled water, MB dye displays a major absorption band centered at 663 nm, which is used to monitor the photocatalytic degradation. With the macroscopic Ta 3 N 5 -Pt cloth as the photocatalyst, the temporal evolution of the absorption spectra of MB is shown in Supplementary Fig. S7. A rapid decrease of MB absorption at wavelength of 663 nm is observed, accompanied with an absorption band shift to shorter wavelengths. The color of MB solution gradually changes from initially blue to transparent as the reaction proceeds (the inset of Fig. S7), indicating that Ta 3 N 5 -Pt cloth exhibits excellent photocatalytic activity for the degradation of MB. For comparison, the photodegradation of MB without photocatalyst (blank test) and with SiO 2 -Pt, bulk Ta 3 N 5 powder, bulk Ta 3 N 5 -Pt powder, or Ta 3 N 5 cloth, was also measured under the other identical conditions, respectively (Figure 5a). The blank test indicates that the degradation of MB is extremely slow without photocatalyst under visible light illumination. By using bulk Ta 3 N 5 powder as the VLD photocatalyst, the photodegradation efficiency of MB can just approach 59.8% after 60 min of reaction. When using Ta 3 N 5 cloth as the photocatalyst, 79.4% of MB is photocatalytically degraded after 60 min. This indicates that Ta 3 N 5 cloth exhibits higher photocatalytic activity than bulk Ta 3 N 5 powder, which can be attributed to its higher BET surface area and hierarchical nanopores. Interestingly, after the decoration with Pt nanoparticles, the Ta 3 N 5 -Pt cloth can degrade 97.2% of MB after 60 min, indicating the highest photocatalytic activity. To investigate the role of Pt nanoparticles in the photocatalytic process, the photocatalytic degradation of MB was conducted in the presence of SiO 2 -Pt powder and bulk Ta 3 N 5 -Pt powder, respectively. Obviously, the SiO 2 -Pt is inactive under visible light irradiation, and the photodegradation of MB is even similar to that of the blank test, which reveals that both SiO 2 and Pt have no photocatalytic activity. However, after the deposition of Pt, bulk Ta 3 N 5 -Pt powder exhibits the improved photodegradation efficiency (83.7%) of MB after 60 min, compared with that (59.8%) of bulk Ta 3 N 5 powder. These facts indicate that Pt nanoparticles can greatly improve the photocatalytic activities of Ta 3 N 5 , which results from the fact that Pt act as electron trap to facilitate the separation of photogenerated electron-hole pairs and promote interfacial electron transfer process 37,38 .</p><p>When colorless parachlorophenol (4-CP) was used as the model of organic pollutant, the photocatalytic activity of macroscopic Ta 3 N 5 -Pt cloth was evaluated by immersing the cloth (20 mg, size: ,2.5 3 3.5 cm 2 ) in 60 mL aqueous solution containing 1.28 mg L 21 4-CP under visible light irradiation (l . 400 nm). 4-CP as a typical pollutant has no photolysis and no visible light absorption characteristics in the photodegradation process. When the macroscopic Ta 3 N 5 -Pt cloth was used as the photocatalyst, the temporal degradation of 4-CP was determined by high-performance liquid chromatography (HPLC) profiles (Supplementary Fig. S8). The peak with retention time of 8.5 min is attributed to the initial 4-CP and is used to monitor the photocatalytic degradation. As the reaction proceeds, the peak decreases rapidly in the reaction, indicating that Ta 3 N 5 -Pt cloth exhibits high photocatalytic activity for the degradation of 4-CP.</p><p>For comparison, the photodegradation of 4-CP without photocatalyst and with SiO 2 -Pt, bulk Ta 3 N 5 powder, Ta 3 N 5 cloth or bulk Ta 3 N 5 -Pt powder, was also measured with otherwise identical conditions, respectively (Figure 5b). The photodegradation of 4-CP without photocatalyst and with SiO 2 -Pt, bulk Ta 3 N 5 powder or Ta 3 N 5 cloth is extremely slow and nearly no 4-CP is degraded after 60 min. Surprisingly, after decoration with Pt nanoparticles, the photocatalytic performances of both bulk Ta 3 N 5 -Pt powder and Ta 3 N 5 -Pt cloth are dramatically improved for the degradation of 4-CP. After 60 min of visible light irradiation, the Ta 3 N 5 -Pt powder exhibits higher photodegradation efficiency of 4-CP (87.2%) than that by Ta 3 N 5 cloth, which can be attributed to the enhanced separation of photogenerated electron-hole pairs in Ta 3 N 5 -Pt heterostructure. With macroscopic Ta 3 N 5 -Pt cloth as photocatalyst, only a 40 min period was required to decompose all the 4-CP in the solution (Supplementary Fig. S8 and Fig. 5b), further demonstrating the highest photocatalytic activity. Compared with bulk Ta 3 N 5 -Pt powder, the outstanding photocatalytic activity for the degradation of 4-CP by Ta 3 N 5 -Pt cloth can be ascribed to its relatively higher BET surface area (23.1 m 2 g 21 ) and special hierarchical structure.</p><p>It is well known that mineralization is the ultimate goal in pollutant treatment. Total organic carbon (TOC) value as an important index for the mineralization of organic species, was studied in the photodegradation of 4-CP (60 mL, 20 mg L 21 ) by 250 mg of Ta 3 N 5 -Pt cloth (Supplementary Fig. S9). It is clear that the TOC concentration of the solution continuously decreases, indicating that 4-CP is steadily mineralized by Ta 3 N 5 -Pt cloth photocatalyst under visible light irradiation. After 120 min of irradiation, the TOC concentration decreases from 10.76 mg L 21 to 3.66 mg L 21 , reaching a high mineralization ratio of 66%. This fact demonstrates that Ta 3 N 5 -Pt cloth can efficiently degrade and mineralize organic pollutants under the irradiation of visible light.</p><p>It has been reported that the conduction and valence band edges of Ta 3 N 5 at pH 0, are at approximately 20.4 V and 11.7 V versus NHE, respectively 39 . Since the redox potential value of photogenerated hole (Q(h 1 )) is approximately equal to that (11.7 V) of the valance band, the Q(h 1 ) is lower than Q(OHN/H 2 O)(12.38 V versus NHE) 14 . As a result, the NOH radicals can not be produced via the direct oxidation of H 2 O molecules by photo-induced holes. The photogenerated electron (Q(e 2 )) is more negative than Q(O 2 / NO 2 2 )(20.33 V versus NHE), which allows the production of NO 2 2 via the reduction of O 2 by conduction band electrons. To confirm this conjecture, the electron spin resonance (ESR) technique (with 5,5-dimethyl-pyrroline N-oxide, DMPO) was used to obtain the information on the active radicals involved in the solution with Ta 3 N 5 cloth irradiated by visible light or un-irradiated. Because NO 2 2 in water is very unstable and undergoes facile disproportionation rather than slow reaction with DMPO 40 , the involvement of NO 2 2 was examined in DMSO in which the DMPO-NO 2 2 has a longer life time 41 . The characteristic peaks of the DMPO-NO 2 2 adducts were observed in DMSO solution with Ta 3 N 5 cloth irradiated by visible light (Supplementary Fig. S10), while no NO 2 2 signal was detected in dark under otherwise identical conditions, which are in good agreement with the previous report 42 . Recently, there are several reports revealed that the NOH can be generated from NO 2 2 with the assistance of the photoinduced electrons [43][44][45] . In our case, to confirm the presence of NOH in the photocatalytic process, the aqueous solution with Ta 3 N 5 cloth irradiated by visible light or in dark was measured by ESR. As shown in Figure 5c, the four characteristic peaks of DMPO-NOH (1525251 quartet pattern) were also observed in aqueous solution with Ta 3 N 5 cloth irradiated with visible light, while no NOH signal was detected in dark under otherwise identical conditions. This fact demonstrates that the NOH can be produced from NO 2 2 , which is similar to the previous reports [43][44][45] . These ESR results confirm that NOH and NO 2 2 were produced in the solution with Ta 3 N 5 cloth under the irradiation of visible light, and they are supposed to finally induce the decomposition of organic pollutants.</p><p>Most importantly, the macroscopic Ta 3 N 5 -Pt cloth (present area: ,4.5 3 2.6 cm 2 ) can be easily transferred and/or recycled in photocatalytic application. To evaluate the stability and reusability of macroscopic Ta 3 N 5 -Pt cloth, a recycling test was performed, as shown in Figure 5d. The photodegradation of 4-CP was monitored for four cycles (each cycle lasted 60 min). After each cycle, the macroscopic Ta 3 N 5 -Pt cloth was taken out and washed with water. Then the cloth was immersed in the same volume (60 mL) of fresh 4-CP solution again. The photocatalytic activity of Ta 3 N 5 -Pt cloth does not significantly decrease in the cycling test and the photodegradation efficiency of 4-CP can still reach 100% for the fourth cycle. Thus, during four cycles, there is no significant loss of photocatalytic activity. The SEM image and the XRD patterns (Supplementary Fig. S11) further confirm that there are no obvious changes in the morphology and the crystalline phase of Ta 3 N 5 -Pt cloth before and after recycling reactions, indicating excellent stability and reusability of Ta 3 N 5 -Pt cloth.</p><!><p>On the basis of the above results and energy band diagram, the photocatalytic process of Ta 3 N 5 -Pt cloth can be proposed, as shown in Figure 4. The photocatalytic activity of macroscopic Ta 3 N 5 -Pt cloth was evaluated by immersing the cloth in aqueous solution containing model pollutant due to its macroscale size (present area: ,4.5 3 2.6 cm 2 ). Ta 3 N 5 with the narrow band-gap (2.1 eV) has a broad range of visible-light photo-response and can exhibit efficient visible-light photoabsorption. The photocatalytic reaction is initiated by the absorption of visible-light photons with energy equal or higher than the band-gap in Ta 3 N 5 semiconductor, which results in the creation of photogenerated holes in its valence band (VB) and electrons in its conduction band (CB). Because of the small particle size of Ta 3 N 5 nanoparticles (,25 nm), the charge carriers can quickly travel to the surface of the catalyst from the interior. Then CB-electrons easily flow into metal Pt through the Schottky barrier because the CB (or the Fermi level) of Ta 3 N 5 is higher than that of the loaded metal Pt, which is consistent with the previous study on electron transfer from semiconductor (such as TiO 2 ) to Pt 37,38 . This process of fast electron transfer contributes to enhancing interfacial charge transfer and realizing the efficient separation of VB-holes and CBelectrons in the heterostructures 37,38 . Thus, plenty of CB-electrons in Pt component are available to reduce O 2 to produce NO 2 2 , which can be further transformed into NOH with the assistance of the photoinduced electrons [43][44][45] . Under successive attacks by NO 2 2 and NOH, MB and 4-CP were effectively photodegraded, as demonstrated in Figures 5a,b and Supplementary Figures S7-S9.</p><p>It is noteworthy that hierarchical pores in Ta 3 N 5 -Pt cloth are supposed to play an important role in this photocatalytic process. As mentioned above, there are plenty of nanopores with diameter of ,13 nm inside Ta 3 N 5 -Pt fibers and micro-pores with sizes of 0.2-1 mm beside Ta 3 N 5 fibers, probably resulting in two positive effects. The one effect is that micro-pores with sizes of 0.2-1 mm increase the photoscattering and absorption of visible light (Supplementary Fig. S12), since the photoabsorption can be enhanced if the nanoarrays are aligned with photonic-crystal microstructures, and/or the faceted end planes of well-shaped crystals serve as good laser-cavity mirrors 46,47 The other results from the fact that the hierarchical combination of smaller nanopores and larger macro-pores can be considered as transport paths 48 . It has been reported that chemical reactions can occur more easily when the transport paths, through which reactant molecules move in or out of the nanostructured materials, are included as an integral part of the architectural design 49 . The textural transport paths have been revealed to have the beneficial effect on photocatalysis 48,50 . We believe that the presence of transport paths in Ta 3 N 5 -Pt cloth also benefits the pollutant molecules to get to the reactive sites on the framework walls of photocatalysts, which results in excellent photocatalytic activity. Furthermore, these transport paths as well as macroscale size probably make Ta 3 N 5 -Pt cloth used as ''microfiltration membrane'' with photocatalytic activity, probably resulting in a very promising application as ''photocatalyst dam'' for the polluted river in the future (Supplementary Fig. S13).</p><p>In summary, macroscopic Ta 3 N 5 -Pt nonwoven cloth with hierarchical nanopores has been synthesized by an electrospinning-calcination-nitridation-wet impregnation method. Such free-standing cloth is composed of nanofibers constructed from Ta 3 N 5 nanoparticles, hierarchical nanopores and Pt nanoparticles. Under visible light illumination, it exhibits excellent photocatalytic activities on MB and 4-CP degradation. Furthermore, it can be easily transferred and/or recycled, with good stability. It should be noted that the present Ta 3 N 5 -Pt cloth is still relatively fragile, further work should be carried out for obtaining Ta 3 N 5 -Pt cloth with better strength and flexibility, and work in this direction is already ongoing. More importantly, this work provides some insight into the design and development of novel, efficient and easily recyclable macroscale photocatalysts with nanostructure, for future practical photocatalytic application, for example, as ''photocatalyst dam'' for the photodegradation of organic pollutants in the polluted river.</p><!><p>Materials synthesis. Synthesis of Ta 3 N 5 nonwoven cloth. At first, 10 wt% Ta(OEt) 4 was dissolved in an ethanol-acetic acid mixture (3.351, volume ratio). Then, 5 wt% polyvinylpyrrolidone (PVP, M W < 1300000 g mol 21 ) was added to the above solution. After vigorously stirring for 24 h, the precursor solution was loaded into a plastic syringe and the feeding rate was kept constant at 0.3 ml h 21 using a syringe pump. A high voltage of 15 kV was applied between the orifice and grounded aluminum foil at a distance of 20 cm. The collected PVP/Ta 2 O 5 /Ta(OEt) 4 composite cloth was calcined at 600uC in air for 6 h to obtain Ta 2 O 5 nonwoven cloth. The Ta 2 O 5 cloth was further nitridized at 800uC under an ammonia flow (500 mL min 21 ) for 8 h to obtain Ta 3 N 5 nonwoven cloth.</p><p>Synthesis of Ta 3 N 5 -Pt nonwoven cloth. Pt (0.5 wt%) was loaded on Ta 3 N 5 nonwoven cloth by the photocatalytic reduction of H 2 PtCl 6 in methanol aqueous solution under a 300 W xenon lamp light irradiation for 4 h.</p><p>Mesoporous SiO 2 was prepared according to the reference 51 . SiO 2 -Pt powder: Pt (0.5 wt%) was loaded on SiO 2 by the photocatalytic reduction of H 2 PtCl 6 in methanol aqueous solution under a 300 W xenon lamp light irradiation for 4 h.</p><p>Bulk Ta 3 N 5 powder was prepared by thermal nitridation of bulk Ta 2 O 5 powder synthesized in our laboratory. Bulk Ta 2 O 5 powder: 2.5 g Ta(OEt) 4 was dissolved in 30 ml absolute alcohol to obtain Ta(OEt) 4 solution, then the ethanolic Ta(OEt) 4 solution was quickly added into 30 ml aqueous solution with pH8 under magnetic stirring, the precipitate was collected by centrifugation, washed with ethanol and distilled water, and dried in an oven at 100uC, then the dried precipitate was calcined at 800uC for 3 h; Bulk Ta 3 N 5 powder: Bulk Ta 2 O 5 powder was nitridized at 850uC under an ammonia flow (500 mL min 21 ) for 15 h to obtain bulk Ta 3 N 5 powder.</p><p>Bulk Ta 3 N 5 -Pt powder. The procedure of Pt loading is the same as that of Ta 3 N 5 -Pt nonwoven cloth except that the Ta 3 N 5 nonwoven cloth was replaced by bulk Ta 3 N 5 powder.</p><p>Characterizations. X-ray diffraction (XRD) measurements were recorded on a D/ max-2550 PC X-ray diffractometer using Cu Ka radiation (l 5 0.15418 nm). The scanning electron microscope (SEM) characterizations were performed on a Hitachi S-4800 field emission scanning electron microscope. The transmission electron microscope (TEM) analyses were performed by a JEOL JEM-2010F high-resolution transmission electron microscope. The optical diffuse reflectance spectrum were conducted on a UV-VIS-NIR scanning spectrophotometer (UV-3101PC, Shimadzu) using an integrating sphere accessory. Nitrogen absorption-desorption measurement were conducted on a Micromeritics ASAP 2020 nitrogen adsorption apparatus (USA). The BET surface area was determined by a multipoint BET method using the adsorption data in the relative pressure (P/P 0 ) range of 0.05-0.3. A desorption isotherm was used to determine the pore size distribution via the Barret-Joyner-Halender (BJH) method, assuming a cylindrical pore model. Electron paramagnetic resonance (EPR) signals of paramagnetic species spin-trapped with 5,5-dimethylpyrroline N-oxide (DMPO) were recorded with a Bruker ESR 300E spectrometer. The irradiation source was a Quanta-Ray Nd:YAG pulsed laser system (l 5 532 nm, 10 Hz). The total organic carbon (TOC) values were detected by a Shimadzu TOC-VCPH total organic carbon analyzer.</p><p>Photocatalytic tests. Photocatalytic activities of the photocatalysts were evaluated by degradation of Methylene Blue (MB) dye and parachlorophenol (4-CP) contaminant in an aqueous solution under visible light irradiation using a 300 W xenon lamp (Beijing Perfect Light Co. Ltd., Beijing) with a cut-off filter (l . 400 nm) as light source. In each experiment, Ta 3 N 5 cloth (20 mg, size: ,2.5 3 3.5 cm 2 ), Ta 3 N 5 -Pt cloth (20 mg, size: ,2.5 3 3.5 cm 2 ), mesoporous SiO 2 -Pt (20 mg), bulk Ta 3 N 5 powder (20 mg) or bulk Ta 3 N 5 -Pt powder (20 mg) as photocatalyst was added into 60 mL of MB aqueous solution (10 mg L 21 ) and 60 mL of parachlorophenol aqueous solution (1.28 mg L 21 ). The temperature of the reaction solution was controlled at 22 6 2uC by cooling water. Before illumination, the suspension was mildly magnetically stirred in the dark for 3 h to ensure that an adsorption/desorption equilibrium was established between the photocatalysts and the target contaminant (MB and 4-CP). When the remaining MB and 4-CP concentration needed to be measured, at given irradiation time intervals (10 min), 3 mL aliquots were collected and centrifuged to remove the remaining solids for analysis. Then, for the photocatalytic test of MB, the absorption UV-vis spectra of the solution were recorded on a U-2910 UV-vis spectrophotometer (Hitachi, Japan). For the photocatalytic test of 4-CP, the 4-CP concentrations in the solutions were analyzed by high-performance liquid chromatography (HPLC) using an Agilent 1100 series (USA) equipped with a diode array detector (DAD) with wavelength set at 280 nm directly after filtration through a 0.22 mm hydrofacies syringe filter. The mobile phase was methanol (80%) and water (20%) and the flow rate was 0.5 mL min 21 ; In the TOC test, 250 mg of Ta 3 N 5 -Pt cloth was added into 60 mL of parachlorophenol aqueous solution (20 mg L 21 ). In the stability and reusability test of the catalyst, four consecutive cycles were tested. The catalysts were washed thoroughly with water and dried after each cycle, and then the cloth was immersed in the same volume (60 mL) of fresh parachlorophenol aqueous solution (1.28 mg L 21 ) again.</p>

Scientific Reports - Nature

The Effect of 5-Alkyl Modification on the Biological Activity of Pyrrolo[2,3-d]pyrimidine Containing Classical and Nonclassical Antifolates as Inhibitors of Dihydrofolate Reductase and as Antitumor and/or Antiopportunistic Infection Agents1a-e

Novel classical antifolates (3 and 4) and 17 nonclassical antifolates (11-27) were synthesized as antitumor and/or antiopportunistic infection agents. Intermediates for the synthesis of 3, 4, and 11-27 were 2,4-diamino-5-alkylsubstituted-7H-pyrrolo[2,3-d]pyrimidines, 31 and 38, prepared by a ring transformation/ring annulation sequence of 2-amino-3-cyano-4-alkyl furans to which various aryl thiols were attached at the 6-position via an oxidative addition reaction using I2. The condensation of \xce\xb1-hydroxy ketones with malonodinitrile afforded the furans. For the classical analogues 3 and 4, the ester precursors were deprotected, coupled with diethyl-l-glutamate, and saponified. Compounds 3 (IC50 = 60 nM) and 4 (IC50 = 90 nM) were potent inhibitors of human DHFR. Compound 3 inhibited tumor cells in culture with GI50 \xe2\x89\xa4 10\xe2\x88\x927 M. Nonclassical 17 (IC50 = 58 nM) was a potent inhibitor of Toxoplasma gondii (T. gondii) DHFR with >500-fold selectivity over human DHFR. Analogue 17 was 50-fold more potent than trimethoprim and about twice as selective against T. gondii DHFR.

the_effect_of_5-alkyl_modification_on_the_biological_activity_of_pyrrolo[2,3-d]pyrimidine_containing

Introduction<!>Chemistry<!>Biological Evaluation and Discussion<!>E. coli DHFR (Table 1)<!>T. gondii DHFR (Table 1)<!>P. carinii DHFR (Table 2)<!>Mycobacterium a\xce\xbdium DHFR (Table 2)<!>Rat liver DHFR (Table 2)<!>Experimental Section<!>1-Hydroxy-2-pentanone (29)<!>2-Amino-4-propyl-furan-3-carbonitrile (30)<!>2,4-Diamino-5-propyl-7H-pyrrolo[2,3-d]pyrimidine (31)<!>Ethyl 4-[2,4-Diamino-5-propyl-7H-pyrrolo[2,3-d]pyrimidin-6-yl)sulfanyl]benzoate (32)<!>4-[2,4-Diamino-5-propyl-7H-pyrrolo[2,3-d]pyrimidin-6-yl)sul-fanyl]benzoic acid (33)<!>Diethyl N-[4-[(2,4-Diamino-5-propyl-7H-pyrrolo[2,3-d]pyri-midin-6-yl)sulfanyl]-benzoyl]-l-glutamate (34)<!>N-[4-[(2,4-Diamino-5-propyl-7H-pyrrolo[2,3-d]pyrimidin-6-yl)sulfanyl]benzoyl]-l-glutamic Acid (3)<!>1-Hydroxy-3-methyl-2-butanone (36)<!>2-Amino-4-isopropyl-furan-3-carbonitrile (37)<!>2,4-Diamino-5-isopropyl-7H-pyrrolo[2,3-d]pyrimidine (38)<!>Ethy l4-[2,4-Diamino-5-isopropyl-7H-pyrrolo[2,3-d]pyrimi-din-6-yl)sulfanyl]-benzoate (39)<!>4-[2,4-Diamino-5-isopropyl-7H-pyrrolo[2,3-d]pyrimidin-6-yl)-sulfanyl]-benzoic acid (40)<!>Diethyl N-[4-[(2,4-Diamino-5-isopropyl-7H-pyrrolo[2,3-d]py-rimidin-6-yl)sulfanyl]-benzoyl]-l-glutamate (41)<!>N-[4-[(2,4-Diamino-5-isopropyl-7H-pyrrolo[2,3-d]pyrimidin-6-yl)sulfanyl]-benzoyl]-l-glutamate (4)<!>2,4-Diamino-5-propyl-6-(2\xe2\x80\xb2,6\xe2\x80\xb2-dichlorophenylsulfanyl)-7H-pyrrolo[2,3-d]pyrimidine (11)<!>2,4-Diamino-5-propyl-6-(2\xe2\x80\xb2,6\xe2\x80\xb2-dimethylphenylsulfanyl)-7H-pyrrolo[2,3-d] pyrimidine (12)<!>2,4-Diamino-5-propyl-6-(phenylsulfanyl)-7H-pyrrolo[2,3-d]py-rimidine (13)<!>2,4-Diamino-5-propyl-6-(4\xe2\x80\xb2-methoxyphenylsulfanyl)-7H-pyr-rolo[2,3-d]pyrimidine (14)<!>2,4-Diamino-5-propyl-6-(2\xe2\x80\xb2,5\xe2\x80\xb2-dimethoxyphenylsulfanyl)-7H-pyrrolo[2,3-d]pyrimidine (15)<!>2,4-Diamino-5-propyl-6-(3\xe2\x80\xb2,4\xe2\x80\xb2-dimethoxyphenylsulfanyl)-7H-pyrrolo[2,3-d]pyrimidine (16)<!>2,4-Diamino-5-propyl-6-(1\xe2\x80\xb2-napthylsulfanyl)-7H-pyrrolo[2,3-d]pyrimidine (17)<!>2,4-Diamino-5-propyl-6-(2\xe2\x80\xb2-napthylsulfanyl)-7H-pyrrolo[2,3-d]pyrimidine (18)<!>2,4-Diamino-5-isopropyl-6-(2\xe2\x80\xb2,6\xe2\x80\xb2-dichlorophenylsulfanyl)-7H-pyrrolo[2,3-d]pyrimidine (19)<!>2,4-Diamino-5-isopropyl-6-(2\xe2\x80\xb2,6\xe2\x80\xb2-dimethylphenylsulfanyl)-7H-pyrrolo[2,3-d] pyrimidine (20)<!>2,4-Diamino-5-isopropyl-6-(phenylsulfanyl)-7H-pyrrolo[2,3-d]pyrimidine (21)<!>2,4-Diamino-5-isopropyl-6-(4\xe2\x80\xb2-methoxyphenylsulfanyl)-7H-pyrrolo[2,3-d]pyrimidine (22)<!>2,4-Diamino-5-isopropyl-6-(2\xe2\x80\xb2,5\xe2\x80\xb2-dimethoxyphenylsulfanyl)-7H-pyrrolo[2,3-d]pyrimidine (23)<!>2,4-Diamino-5-isopropyl-6-(3\xe2\x80\xb2,4\xe2\x80\xb2-dimethoxyphenylsulfanyl)-7H-pyrrolo[2,3-d]pyrimidine (24)<!>2,4-Diamino-5-isopropyl-6-(1\xe2\x80\xb2-napthylsulfanyl)-7H-pyrrolo[2,3-d]pyrimidine (25)<!>2,4-Diamino-5-isopropyl-6-(2\xe2\x80\xb2-napthylsulfanyl)-7H-pyrrolo[2,3-d]pyrimidine (26)<!>2,4-Diamino-5-isopropyl-6-(3\xe2\x80\xb2,4\xe2\x80\xb2-dichlorophenylsulfanyl)-7H-pyrrolo[2,3-d]pyrimidine (27)

<p>Dihydrofolate reductase (DHFR) along with thymidylate synthase (TS) forms part of the system responsible for the synthesis of 2′-deoxythymidine-5′-monophosphate (dTMP), a key component in DNA biosynthesis and cell replication. TS catalyzes the de novo synthesis of dTMP from 2′-deoxyuridine-5′-monophosphate (dUMP). The cofactor, N5,N10-methylene-tetrahydrofolate (N5,N10-CH2-THF), serves as the donor of the methyl group as well as the reductant for this step and is itself oxidized to 7,8-dihydrofolate (7,8-DHF). The recyclization of 7,8-DHF to 5,6,7,8-tetrahydrofolate (5,6,7,8-THF) is catalyzed by DHFR2 for which NADPH acts as the source of the reductant. Thus inhibition of DHFR and/or TS leads to "thymineless death".a</p><p>Pneumocystis jiroνecii(P. jiroνecii) previously known as Pneumocystis carinii(P. carinii)3,4 [Note: P. jiroνecii is the strain that infects humans, while P. carinii refers to the strain that infects rats] and Toxoplasma gondii(T. gondii)3,4 are often fatal opportunistic infections in AIDS patients. Mycobacterium aνium(M. aνium) complex (MAC),3,4 a group of organisms that is responsible for disseminated infections in AIDS patients, additionally decreases the quality of life of patients with AIDS. Several DHFR and TS inhibitors have found clinical utility as antitumor and antiopportunistic agents.5 Classical antifolates like, methotrexate6 (MTX) (Figure 1), raltitrexed,7 and pemetrexed8, are clinically used as antitumor agents. Nonclassical antifolates like trimetrexate (TMQ), pyrimethamine, and trimethoprim (TMP) are clinically used as antiopportunistic infection agents.4</p><p>The combination of a weak DHFR inhibitor (TMP, py-rimethamine), along with a potent dihydropteroate synthase (DHPS) inhibitor (sulfamethoxazole), is currently used to treat infections caused by opportunistic pathogens in AIDS patients.9 However, the combination therapy is successful in only 50-75% of the AIDS population; up to 60% are unable to tolerate the combination therapy due to severe, adverse drug reactions.10 Trimetrexate is coadministered with leucovorin, the classical folate cofactor (6R,6S)-5-formyl-5,6,7,8-THF, which selectively rescues the host cell from the toxicity caused by nonselective, TMQ.11</p><p>Gangjee et al.12 recently reported 1 (Figure 2) as a dual inhibitor of human DHFR (IC50 = 0.21 μM) and human TS (IC50 = 0.54 μM). The 5-CH3 moiety of 1 was incorporated to provide hydrophobic interaction with Val115 in human DHFR. Compound 1 was designed as a nonpolyglutamylatable DHFR inhibitor. However, unexpectedly, 1 had reasonable folyl poly-γ-glutamate synthetase (FPGS) substrate activity. Molecular modeling using SYBYL 6.813 suggested that 1 binds to human DHFR in the normal 2,4-diamino mode (Figure 3) while it could bind to human TS in the flipped mode. In addition, molecular modeling also indicated that the 5-CH3 group in 1 could provide hydrophobic interaction with Trp109 in human TS. Compound 1 was a reasonably potent inhibitor of the growth of human CCRF-CEM leukemia cells in culture with an EC50 value of 190 nM as compared with MTX (EC50 = 12.5 nM). In 11 of the 60 tumor cell lines evaluated at the National Cancer Institute (NCI) preclinical screening program, compound 1 showed GI50 values of ≤ 10−7 M (Table 3). Homologation of the 5-methyl group in 1 to a 5-ethyl group as in 2 (Figure 2) afforded a 3-fold more potent human DHFR inhibitor (IC50 = 0.066 μM).14 Surprisingly, compound 2 was devoid of any significant TS, inhibitory activity (37% inhibition @ >17 μM). However, compound 2 demonstrated increased tumor cell growth inhibitory activities against certain tumor cell lines compared to 1 in the NCI preclinical screening program (Table 3). Thus the size of the alkyl group attached to the 5-position of the 2,4-diamino, pyrrolo[2,3-d]pyrimidine scaffold dictates the activity against DHFR and/or TS as well as tumor cell growth inhibitory potency.</p><p>Molecular modeling (SYBYL 6.91)13 further indicated that the 5-alkyl group could be homologated to a 5-propyl or 5-isopropyl. This homologation could further enhance the van der Waals interaction with Val115 in human DHFR (Figure 4) in the normal DHFR binding mode. To determine the optimum size of the 5-alkyl group for DHFR and/or TS inhibitory activity as well as tumor cell growth inhibitory potency, the 2,4-diamino-5-propyl-6-arylthio-7H-pyrrolo[2,3-d]pyrimidine (3) and 2,4-diamino-5-isopropyl-6-arylthio-7H-pyrrolo[2,3-d]pyrimidine (4) classical analogues (Figure 2) were designed and synthesized.</p><p>The existing regimen used to treat opportunistic infections in AIDS and other immunocompromised patients is suppressive rather than curative and the therapy must be continued indefinitely.3,4 Thus, it is of considerable interest to design single agents that have both the desired selectivity of TMP and the, potency of TMQ. Such agents could be used as single agents to treat opportunistic infections in immunocompromised patients to decrease cost and increase patient compliance. Because patients with AIDS are often infected with multiple opportunistic infections, it is highly desirable to develop single agents that simultaneously target two or more opportunistic pathogen DHFR.</p><p>Gangjee et al.15 also reported nonclassical analogues of 1, including 5 and 6 (Figure 2) as inhibitors of DHFR from opportunistic pathogens. The 5-CH3 moiety was designed to, afford hydrophobic interaction with Ile123 in P. carinii DHFR, Val151 in T. gondii DHFR, and Ile102 in M. avium DHFR on the basis of X-ray crystal structure,16,17 multiple sequence alignment,18,19 and molecular modeling (SYBYL 6.813) studies, respectively. The 5-CH3 group was also suggested to influence the conformations of the 6-arylthio side chain in these inhibitors, thus limiting its flexibility and contributing to the potency of these compounds. Several compounds, including 5 and 6 (Table 2), displayed 10-fold or higher selectivity ratios for T. gondii DHFR and/or M. avium DHFR compared to rat liver (rl), DHFR.15 Compound 6 with a 2′,5′-(OCH3)2 substitution was 16-fold more potent and equally selective compared to TMP against T. gondii DHFR.</p><p>Rosowsky et al.,20 using a different approach, reported compounds 9 and 10 (Figure 2) with a single carbon atom bridge that displayed fair T. gondii DHFR potency and good selectivity. Analogue 9 (IC50 = 0.07 μM) was the most potent in this series against T. gondii DHFR, while analogue 10 was the most selective for T. gondii DHFR compared to rlDHFR with a selectivity ratio of 81.</p><p>A sulfur atom was incorporated in compounds 5 and 6 rather than a carbon atom, as in compounds 9 and 10, to increase the proximity of the 6-arylthio ring to the hydrophobic residues on the pathogen DHFR due to the increased atomic size of the sulfur atom as well as a decrease in the C-S-C angle (98°) compared to a C-C-C angle (109°).15 Compound 6 was 19fold more potent and nearly one-half as selective as the most selective compound (10) of the 6-carbon-bridged analogues. The biological activity of compounds 5 and 6 supported the hypothesis that the 6-arylthio side chain of these compounds indeed interacts more favorably with Phe91 in T. gondii DHFR and Val158 in M. aνium DHFR and that the sulfur bridge increased activity and selectivity. Gangjee et al.14 have also synthesized the ethyl homologues of 5 and 6 with the goal of further increasing the potency and selectivity. Compound 8 (Figure 2), the ethyl homologue of 6, was found to have increased potency and/or selectivity against P. carinii and T. gondii DHFR compared to rlDHFR (Table 2). Similar to their methyl counterparts, the ethyl homologues including 7 and 8 were found to have increased potency and/or selectivity against T. gondii and/or M. aνium DHFR. In most instances, the ethyl homologues tested were found to be more active and/or selective against two or more pathogen DHFR. In an attempt to optimize the size of the 5-alkyl substitution on the potency and selectivity for P. carinii DHFR, T. gondii DHFR, and M. aνium DHFR compounds 11-27 (Figure 2) were also designed and synthesized. Compounds 11-18 contain a 5-propyl group, while compounds 19-27 contain a 5-isopropyl group.</p><!><p>The syntheses of compounds 3 and 11-18 required the synthesis of 2,4-diamino-5-propyl-7H-pyrrolo[2,3-d]pyrimidine, 31 (Scheme 1), while the synthesis of 4 and 19-27 required the synthesis of 2,4-diamino-5-isopropyl-7H-pyrrolo[2,3-d]py-rimidine, 38 (Scheme 2). Taylor et al.21 have reported the synthesis of various 2,4-diamino-5-alkyl-7H-pyrrolo[2,3-d]py-rimidines by a ring transformation/ring annulation sequence of 2-amino-3-cyano-4-alkyl furans. These furans were in turn obtained by the condensation of suitable α-hydroxy ketones with malonodinitrile in the presence of a suitable base such as triethylamine. Gangjee et al.12,14,15,22 and Rosowsky et al.23,24 have also successfully adopted this methodology in their synthesis of pyrrolo[2,3-d]pyrimidine containing antifolates. Extending this general methodology to the synthesis of 31 required the synthesis of 1-hydroxy-2-pentanone, 29 (Scheme 1). Compound 29 was in turn obtained from the commercially available 1,2-pentanediol, 28, by regiospecific oxidation of the secondary alcohol using hexabutyldistannoxane (HBD) and Br2 (Scheme 1).25 Two methods were attempted for the separation of the 1-hydroxy-2-pentanone, 29, from the reaction mixture. The first involved silica gel chromatography on the crude reaction mixture and the second was a direct distillation of the crude reaction mixture. In general, distillation was found to be superior to column chromatography. Condensation of 29 with malonodinitrile using triethylamine as base afforded the 2-amino-4-propyl-furan-3-carbonitrile, 30, in 60% yield. Further, condensation of 30 with guanidine (liberated from guanidine, hydrochloride and NaOMe) afforded 31 in 50% yield. Reaction of 2,4-diamino-5-propyl-7H-pyrrolo[2,3-d]pyrimidine, 31, and ethyl 4,4′-bismercaptobenzoate in EtOH/H2O followed by the addition of I2 at reflux afforded compound 32 in 28% yield.26,27 The disappearance of the 6-vinyl proton at δ 6.38 and the appearance of the characteristic AA′XX′ pattern for the 6-aryl, protons in the 1H NMR spectrum of 32 in DMSO (d6) indicated the success of the oxidative addition reaction.</p><p>Hydrolysis of the ester 32 with aqueous 1N NaOH at 80 °C (24 h) followed by acidification gave the required acid, 33, in 83% yield. Peptide coupling21 of the acid 33 with diethyl-l-glutamate using 2,6-dimethoxy-4-chlorotriazine and N-methyl, morpholine, followed by chromatographic purification afforded the coupled product 34 in 62% yield. The 1H NMR spectrum of 34 in DMSO (d6) revealed the newly formed amide NH proton at δ 8.64-8.67 ppm as a doublet. Hydrolysis of the diester 34 with aqueous NaOH at 0 °C (4 h) and then at room temperature (24 h), followed by acidification, gave the desired compound 3 in 86% yield.</p><p>Similarly, reaction of 31 with appropriately substituted aryl thiols in a mixture of EtOH/H2O (2:1) followed by addition of I2 at reflux as reported previously27 afforded 11-18 in 45%-70% yields. The yields reveal no apparent correlation between the extent of pyrrolo[2,3-d]pyrimidine substitution and the electron-donating or -withdrawing effects of substituents in the thiophenol.</p><p>Analogous to 31, the synthesis of 37 (Scheme 2) required the synthesis of 1-hydroxy-3-methyl-2-butanone, 36. Thiazolium salt-catalyzed benzoin condensation of isopropyl aldehyde 35 with paraformaldehyde catalyzed by N-ethylbenzothiazolium bromide and triethylamine afforded the α-hydroxy ketone, 36, after distillation, in 35% yield.28 Compounds 4 and 19-27 were synthesized as shown in Scheme 2 starting with 36 in essentially the same way as described for 3 and 11-18 in Scheme 1.</p><p>The yields in Scheme 2, as before for Scheme 1, reveal no apparent correlation between the extent of pyrrolo[2,3-d]pyri-midine substitution, and the electronic nature of the substituents in the thiophenol. The lower yields of 18 (Scheme 1) and 19 (Scheme 2) may be a result of unfavorable steric interactions between the bulky 5-isopropyl group in 37 and the 2′,6′-disubstitution present on the thiophenols, which makes 6-sub-stitution more difficult.</p><!><p>Compounds 3, 4, and 11-27 were evaluated as inhibitors of human (h), Escherichia coli (E. coli), and T. gondii DHFR and TS. The inhibitory potency (IC50) values are compared with MTX, pemetrexed, TMP, pyrimethamine, and the previously synthesized 1 and 2 (Table 1). Compounds 3 and 4 are good inhibitors of hDHFR with nanomolar IC50 values and were about 3-fold and 4-fold less potent as hDHFR inhibitors, respectively, compared with MTX and about 25-fold and 17-fold more potent respectively than pemetrexed. Compound 3 was equipotent with the previously synthesized 2 and about 3.5-fold more potent than 1. The biological data of 1-4 indicate that an ethyl, propyl, or isopropyl group at the 5-position are all conducive for potent hDHFR inhibition. The potent hDHFR activity of 2-4 compared to 1 could be attributed to increased hydrophobic interaction of the bulkier alkyl groups in 2-4 with Val115 in hDHFR. The increased activity of 2-4 may also result from favorable orientation of the 6-position thioaryl side chain when bound to hDHFR. Against hTS, 3 and 4 had similar inhibitory potency as MTX but were 4-fold less inhibitory than pemetrexed. Compounds 3 and 4 were about 63-74-fold less potent than 1 as inhibitors of hTS. These results indicate that homologation of the 5-methyl group in 1 to larger alkyl groups as in 2-4 is detrimental to hTS inhibition. This decrease in potency may be due to steric hindrance between the larger alkyl groups in 2-4 and Trp109 in hTS and/or due to unfavorable orientation of the 6-position side chains for interaction with the hTS in the presence of the bulkier 5-alkyl moiety.</p><p>Compound 3 was a poor inhibitor of hTS and E. coli TS (Table 1) but showed moderate inhibition against T. gondii TS (equipotent to pemetrexed). Compound 3 was a good inhibitor of all three DHFR tested. In addition, 3 is also a dual inhibitor of T. gondii DHFR and T. gondii TS. The nonclassical analogues, 11-27 were all poor inhibitors of all three TS tested. They were however reasonably potent inhibitors of E. coli DHFR and T. gondii DHFR. Most of the analogues were weak or poor inhibitors of hDHFR.</p><!><p>In general the 5-isopropyl compounds (19-27) with the exceptions of 19 and 20 were more potent and selective against E. coli DHFR than the corresponding 5-propyl compounds (11-18). In the 5-propyl series, analogue 16 with a 3,4-dimethoxyphenyl side chain was the most potent and selective compound against E. coli DHFR. In the 5-isopropyl series, analogues 21-27 were potent against E. coli DHFR. A number of the nonclassical compounds in the 5-isopropyl series showed good selectivity for E. coli DHFR as compared to hDHFR. Compounds 21-27 were 100-fold to 254-fold more selective for E. coli DHFR than hDHFR. Thus the 5-isopropyl-6-substituted phenyl analogues were reasonably selective for bacterial DHFR.</p><!><p>In general the 5-propyl compounds (13-18) with the exception of 13 were more potent and selective against T. gondii DHFR than the corresponding 5-isopropyl compounds (21-26). In the 5-propyl series, analogue 17 with a 1-naphthyl side chain was the most potent compound against T. gondii DHFR and was 50-fold more potent than TMP, 5-fold less potent than MTX and equipotent with pyrimethamine. Compound 17 was also the most selective compound against T. gondii DHFR with >500-fold selectivity over human DHFR. Thus 17 is 50-fold more potent than TMP and about twice as selective against T. gondii DHFR. Compound 13 with a phenyl side chain was 20-fold more potent than TMP and equally selective against T. gondii DHFR compared with human DHFR. In the 5-isopropyl series, analogue 21 with an unsubstituted phenyl side chain was the most potent analogue against T. gondii DHFR and was 45-fold more potent than TMP and equally selective. Analogue 25 with a 1-naphthyl side chain was 24fold more potent than TMP and equally selective.</p><p>Compounds 3, 4, 11-27, MTX, PYR, and TMP were also assayed against T. gondii DHFR using protocols described for Table 2 (data for T. gondii DHFR not shown), as well as under the conditions described for Table 1. Of the 20 compounds jointly assayed, MTX, 3, and 4 were identified as the three most potent compounds by both protocols; both laboratories also placed TMP, 23, 24, and 27 as among the six least potent compounds. Selectivity could only be directly compared for the compounds in Table 1 that had defined selectivity values. Assays under both sets of conditions identified TMP, 13, and 21 as the most selective compounds in this set of nine compounds and MTX as the least selective. Considering all compounds inde pendently assayed as described in Table 2 for T. gondii, DHFR, the most selective compounds are TMP, 21, 15, 20, 13, 17, and 23; this list is consistent with the results in Table 1, except that the selectivity of compounds 15 and 20 is artificially depressed in Table 1 by the inability to generate full inhibition curves for the human reference enzyme.</p><p>Compounds 3, 4, and 11-27 were also evaluated as inhibitors of P. carinii DHFR, M. avium DHFR, and rlDHFR, which served as the mammalian surrogate under slightly different assay conditions. The inhibitory potency (IC50) values are compared with TMQ, TMP, and the previously synthesized 5-8 (Table 2). Several compounds displayed 10-fold or higher selectivity ratios for M. avium DHFR. Against P. carinii DHFR, in, general, the 5-propyl nonclassical analogues (11-18) were more potent and selective than the corresponding 5-isopropyl analogues (19-27). Against M. aνium DHFR, the nature of the phenyl, substitution along with the 5-alkyl group determined the potency and selectivity of the compound.</p><!><p>Against P. carinii DHFR the most potent analogues bore an unsubstituted phenyl (in 13) or a 4′-methoxyphenyl substitution (in 14) in the 5-propyl (11-18) series. Other substitutions such as a 2,6-dichloro 11, 2,6-dimethyl 12, 3,4-dimethoxy 16, or 2,5-dimethoxy 15 caused a slight drop in activity. The 1-naphthyl substitution in 17 was found to display moderate DHFR inhibitory activity, while a 2-naphthyl substitution in 18 was found to be slightly detrimental for activity compared to 13. In the 5-isopropyl (19-27) series, the most potent analogue contained a 1-naphthyl side chain 25 and displayed submicromolar inhibitory potency. All other substitutions in the side chains as in 19-27, with the exception of 25, displayed micromolar or higher inhibitory potency. Compounds 11-27 were not selective against P. carinii DHFR and increasing the size of the 5-alkyl group did not improve the selectivity, however, it increased the potency of the compounds against P. carinii DHFR compared to the corre sponding methyl (compare 5 with 26 or 6 with 15) and ethyl (compare 7 with 13 or 21) analogues. The biological data of analogues 11-21 indicate that the homologation to a 5-propyl or 5-isopropyl group is conducive for potent inhibition of P. carinii DHFR, however, it does not improve the selectivity.</p><!><p>Analogue 16 con taining a 3′,4′-dimethoxy substitution in the phenyl ring was the most potent and selective analogue in the 5-propyl series. The second best analogue 15 had a 2′,5′-dimethoxy substitution in the side chain. Substitution of the phenyl ring with other substitutions as in 11-14, 17, and 18 resulted in analogues that were considerably less potent and selective. In the 5-isopropyl series, the most potent analogue 25 contains a 1-naphthyl side chain. The 2-naphthyl substituted analogue 26 was 10-fold less potent than 25. The most selective analogue was 23, with a 2′,5′-dimethoxy substitution in the side chain. The electron donating 3,4-dimethoxy substituted analogue 24 was found to be devoid of any selectivity. In sharp contrast, the analogue with electron withdrawing 3,4-dichloro substitution 27 had the second best selectivity. Again, the biological data of 11-27 indicate that both the alkyl group present at the 5-position as well as the substituents present on the 6-position thioaryl side chain play a role in determining the potency and selectivity of the analogues against M. aνium DHFR.</p><!><p>In the 5-propyl series, analogues 13 and 14 with an unsubstituted phenyl and a 4′-methoxy substitution, respectively, were the most potent. Dimethoxy substituted analogues 15 and 16 were 2-fold less potent than the monomethoxy 14. Substitution of the phenyl ring with bulky groups such as 1-naphthyl 17 and 2-naphthyl 18 also resulted in 2-fold less potent compounds compared to 13. However, substitution of the phenyl ring with either a 2′,6′-dichloro 11 or 2′,6′-dimethyl 12 substitution maintained activity compared to 13. In the 5-isopropyl series, the 1-naphthyl substituted analogue 25 was the most potent. The 2-naphthyl substituted analogue 26 was 10-fold less potent than 25. Replacing the 1-naphthyl substituent with an unsubstituted phenyl 21 or substitution of the phenyl ring with various electron donating methoxy (23, 24), methyl (20) or electron withdrawing chloro (19, 27) substitution also afforded analogues that were consider ably less potent.</p><p>Compounds 3 and 4 were selected by the National Cancer Institute (NCI) for evaluation in its in vitro preclinical antitumor screening program.29 The ability of compounds 3 and 4 to inhibit the growth of tumor cells was measured as GI50 values, the, concentration required to inhibit the growth of tumor cells in culture by 50% compared to a control. In 6 of the 60 tumor cell lines evaluated, compound 3 showed GI50 values of <1 × 10−6 M (Table 3). While in only 2 of the 60 tumor cell lines evaluated, compound 4 showed GI50 values of <1 × 10−6 M. It is noteworthy that compound 3 was not a general cell poison but showed selectivity both within a type of tumor cell line and across different tumor cell lines, with inhibitory values which in some instances differed by 10000-fold. In the melanoma LOX IMVI cell line and the renal cancer cell line 786-0, compound 3 displayed GI50 values of ≤1 × 10−8 M. It can be seen from the tumor cell growth inhibitory activity (Table 3) of compounds 1-4 that the tumor cell growth inhibitory potency, in certain instances, were more potent than either their human DHFR and/or human TS inhibitory activity alone (Table 1) and could be the result of a synergistic effect of dual inhibitory activities against TS and DHFR and/or that poly-glutamylation increases inhibitory activity against TS and/or DHFR in tumor cell systems. Against the outgrowth of tumor cells in culture compound, 2 was in general the most potent compound followed by 3 then 1, and the least potent is com pound 4. Though a strict structure-activity relationship cannot be considered for tumor cells in culture the 5-ethyl is clearly superior to a methyl, propyl or isopropyl.</p><p>In summary, homologation of a 5-methyl (compound 1) to a 5-propyl (compound 3) or 5-isopropyl (compound 4) in 2,4-diamino-6-thiobenzoyl-5-alkylpyrrolo[2,3-d]pyrimidines in creases the human DHFR inhibitory activity but is detrimental to the human TS inhibitory activity. We have found that in classical N-[4-[(2,4-diamino-5-alkyl-7H-pyrrolo[2,3-d]pyrimi-din-6-yl)-thio]-benzoyl]-l-glutamic acid containing analogues the size of the alkyl group at the 5-position dictates inhibition of TS and/or DHFR activity as well as tumor cell growth inhibitory activity. The fact that homologated 5-alkyl substit-uents such as ethyl, propyl, and isopropyl are not tolerated by human TS indicates that homologation of the 5-alkyl group beyond a methyl is not conducive for dual human TS-DHFR inhibition in classical 5-alkyl-6-arylthiosubstituted pyrrolo[2,3-d]pyrimidines. Homologation however maintains dual DHFR-TS inhibitory activity against the bifunctional enzyme derived from T. gondii. In the nonclassical series, homologation of the 5-alkyl group is highly conducive for potent inhibition of P. carinii DHFR; however, it does not improve the selectivity of the analogues. Homologation of the 5-alkyl group to a propyl or isopropyl is highly conducive for potent and selective inhibition of T. gondii DHFR compared to human DHFR. The size of the alkyl group present at the 5-position of the pyrrolo[2,3-d]pyrimidine ring system along with the nature of the lipophilic substituents present on the 6-arylthio side chain determines the potency and selectivity against T. gondii DHFR and M. aνium DHFR.</p><!><p>All evaporations were carried out in vacuum with a rotary evaporator. Analytical samples were dried in vacuum (0.2 mmHg) in an Abderhalden drying apparatus over P2O5 at 70 °C. Thin-layer chromatography (TLC) was performed on silica gel plates (What man 250 μM PE SiLG/UV) with fluorescent indicator. Spots were visualized by UV light (254 and 365 nm) or by staining with a solution of KMnO4 in EtOH. All analytical samples were homo geneous on TLC in at least two different solvent systems. Purification by column and flash chromatography was carried out using Merck silica gel 60 (200-400 mesh). The amount (weight) of silica gel for column chromatography was in the range of 50-100 times the amount (weight) of the crude compounds being separated.</p><p>Columns were dry-packed unless specified otherwise. Solvent systems are reported as volume percent of mixture. Melting points were determined on a Mel-Temp II melting point apparatus and are uncorrected. Proton nuclear magnetic resonance (1H NMR) spectra were recorded on a Bruker WH-300 (300 MHz) spectrom eter. The chemical shift (δ) values are reported as parts per million (ppm) relative to tetramethylsilane as internal standard; s = singlet, d = doublet, t = triplet, q = quartet, m = multiplet, br = broad singlet. Elemental analyses were performed by Atlantic Microlab, Inc., Norcross, GA. Elemental compositions were within ± 0.4% of the calculated values. Fractional moles of water or organic solvents frequently found in some analytical samples of antifolates could not be removed despite 24 h of drying in vacuum and were confirmed, where possible, by their presence in the 1H NMR spectrum. High-resolution mass spectra (HRMS), using electron impact (EI), were recorded on a VG Autospec (Fisons Instruments) micromass (EBE Geometry) double-focusing mass spectrometer. All solvents and chemicals were used as received.</p><!><p>To a solution of 1,2 pentanediol, 28 (2.1 g, 20 mmol) and hexabutyldistannoxane (HBD) (15.5 g, 26 mmol) in anhydrous CH2Cl2 (100 mL), Br2 (4.16 g, 26 mmol) solution in CH2Cl2 (10 mL) was added dropwise at room temper ature with stirring under N2 atmosphere. The mixture was stirred for 3 h at room temperature. The solvent was evaporated under reduced pressure, and the resulting oil was distilled under reduced pressure to give 29 as a colorless oil, bp = 68-70° (16 mmHg) [lit.28 bp = 70° (20 mmHg)].</p><!><p>A mixture of malonodinitrile (2.65 g, 40 mmol) and N(C2H5)3 (5.58 mL, 40 mmol) in anhydrous MeOH (120 mL) was added dropwise to a solution of the α-hydroxy ketone 29 (4 g, 40 mmol) in MeOH, and the resulting dark-red solution was stirred at room temperature for 24 h. To this solution was added silica gel (10 g), and the solvent was evaporated to dryness under reduced pressure to afford a dry silica gel plug, which was loaded on top of a wet (hexane) silica gel column and eluted first with hexane and then with 2:1 hexane/EtOAc to afford 1.88 g (60%) of the furan 30 as a red-cream solid: mp 58.8-62.5 °C; TLC Rf = 0.51 (hexane/EtOAc, 2:1). 1H NMR (DMSO-d6): δ 0.85-0.94 (t, 3 H, 4-CH2CH2CH3), 1.47-1.56 (m, 2 H, 4-CH2CH2CH3), 2.21 - 2.26 (t, 2 H, 4-CH2CH2CH3), 6.74 (s, 1 H, 5-CH), 7.20 (s, 2 H, 2-NH2). Anal. calcd for (C8H10N2O) C, H, N.</p><!><p>Amino nitrile furan 30 (1.8 g, 12 mmol) was added to a solution of guanidine hydrochloride (2.5 g, 26 mmol) and NaOMe (1.4 g, 26 mmol) in anhydrous EtOH (100 mL). The resulting dark-red solution was stirred under reflux overnight, during which time it became dark brown. To this solution was added silica gel (5 g), and the solvent was evaporated to dryness under reduced pressure to afford a dry silica gel plug, which was loaded on top of a wet (CHCl3) column and eluted first with CHCl3 and then with a gradient of 1-5% MeOH in CHCl3 to afford 1.15 g (50%) of 31 as a dark-brown solid: mp 215-220 °C; TLC Rf = 0.41 (CHCl3/MeOH, 5:1, with 2 drops of conc NH4OH). 1H NMR (DMSO-d6): δ 0.83-0.89 (t, 3 H, 5-CH2CH2CH3), 1.50-1.60 (m, 2 H, CH2CH2CH3), 2.57-2.62 (t, 2 H, CH2CH2CH3), 5.36 (s, 2 H, 2/4-NH2), 5.91 (s, 2 H, 2/4-NH2), 6.38 (s, 1 H, 6-CH), 10.36 (s, 1 H, 7-NH). Anal. calcd for (C9H13N5·0.1H2O) C, H, N.</p><!><p>To a suspension of 31 (1.1 g, 5.7 mmol) in a mixture of EtOH/H2O (2:1, 75 mL) was added diethyl 4,4′-dithiobis(benzoate) (2.2 g, 6 mmol) and the suspension was heated to 100-110 °C, then I2 (3 g, 12 mmol) was added and the reaction was monitored (TLC) for completion (3 h). To this solution was added excess sodium thiosulfate and the solution was evapo rated to dryness under reduced pressure and the resulting residue was washed with water and air-dried. This residue was then dissolved in MeOH (100 mL) and to this was added silica gel (15 g) and the resulting suspension was evaporated to dryness under reduced pressure to afford a dry silica gel plug that was loaded on top of a wet silica gel (CHCl3) column and eluted first with CHCl3, and then with a gradient of 1-5% MeOH in CHCl3. Fractions containing the desired spot (TLC) were pooled and evaporated to dryness to afford 610 mg (28%) of 32 as a white solid: mp = 260.6-261 °C; TLC Rf = 0.54 (CHCl3/MeOH, 5:1). 1H NMR (DMSO-d6): δ 0.82 (t, 3 H, CH2CH2CH3), 1.29 (t, 3 H, CH2CH3), 1.41 - 1.43 (m, 2 H, CH2CH2CH3), 2.76 (t, 2 H, CH2CH2CH3), 4.27-4.29 (q, 2 H, CH2CH3), 7.09 (s, 2 H, 2/4-NH2), 7.74 (s, 2 H, 2/4-NH2), 7.13-7.16 (d, 2 H, C6H4), 7.84-7.87 (d, 2 H, C6H4), 12.08 (s, 1 H, 7NH). Anal. calcd for (C18H21N5O2S) C, H, N, S.</p><!><p>To a suspension of 32 (330 mg, 0.9 mmol) in EtOH (30 mL) was added aqueous 1N NaOH (12 mL) and the reaction mixture was stirred at 80 °C for 24 h. At this time, TLC indicated the disappearance of the starting ester at Rf = 0.54 (CHCl3/MeOH, 5:l) and formation of one major spot at the origin. The solvent was evaporated to dryness under reduced pressure, and the resulting sodium salt (yellow oil) was dissolved in water (15 mL) and carefully acidified to pH 4 by dropwise addition of 3N HCl. The resulting suspension was filtered and washed carefully with cold water and dried over P2O5 to afford 295 mg (83%) of 33 as a white solid. This was directly used in the next step without further characterization.</p><!><p>To a suspension of the acid 33 (344 mg, 1 mmol) in anhydrous DMF (15 mL) under N2 was added N-methyl morpholine (145 μL, 1.33 mmol) and the resulting suspension was cooled to 0 °C. At this point, 2-chloro-4,6-dimethoxy-1,3,5-benzotriazine (235 mg, 1.33 mmol) was added and the suspension was stirred for 2 h at 0 °C; during this time, it formed a solution. The reaction mixture was again cooled to 0 °C and diethyl-l-glutamic acid (317 mg, 1.33 mmol) was added followed by N-methyl morpholine (145 μL, 1.33 mmol). The solution was slowly allowed to warm to room temperature with stirring and left at room temperature for a total of 24 h. At this time, TLC indicated the formation of one major spot at Rf = 0.58 (CHCl3/MeOH, 5:l). To the resulting solution was added silica gel (5 g), and the DMF was evaporated to dryness at room temperature using an oil pump. The silica gel plug was loaded on a wet (CHCl3) silica gel column and eluted with a gradient of 1-3% MeOH in CHCl3. Fractions containing the desired spot (TLC) were pooled and evaporated to dryness under vacuum to afford 330 mg (62%) of 34 as a white solid: mp 260-260.5 °C; TLC R f = 0.58 (CHCl3/MeOH, 5:1). 1H NMR (DMSO-d6): δ 0.80-0.84 (t, 3 H, CH2CH2CH3), 1.13-1.19 (m, 6 H, CH2CH3), 1.43 - 1.45 (m, 2 H, CH2CH2CH3), 1.98-2.09 (m, 2 H, Glu β-CH2), 2.42-2.50 (t, 2 H, Glu γ-CH2), 2.72 (t, 2 H, CH2CH2CH3), 4.00-4.10 (m, 4 H, CH2CH3), 4.40 (m, 1 H, Glu α-CH), 5.64 (s, 2 H, 2/4-NH2), 6.22 (s, 2 H, 2/4-NH2), 7.04-7.07 (d, 2 H, C6H4), 7.75-7.77 (d, 2 H, C6H4), 8.64-8.67 (d, 1 H, CONH), 11.06 (s, 1 H, 7-NH). Anal. calcd for (C25H32N6SO5) C, H, N, S.</p><!><p>To a suspension of 34 (200 mg, 0.4 mmol) in EtOH (15 mL) was added 1N NaOH (6 mL) at 0 °C and the resulting suspension was stirred at 0 °C (4 h) and then at room temperature for 24 h. At this point, TLC showed the disappearance of the starting ester at Rf =0.58 (CHCl3/MeOH, 5:l) and formation of one major spot at the origin. The solvent was evaporated to dryness under reduced pressure, and the sodium salt (yellow oil) was dissolved in water (5 mL) and the solution was cooled in an ice bath and acidified carefully to pH 4.0 with dropwise addition of 3N HCl. The resulting suspension was frozen using dry ice/acetone and the reaction flask was kept at 5 °C for 24 h and filtered. The residue was washed carefully with cold water and dried over P2O5 to afford 160 mg (86%) of 3 as a white solid: mp 259.5-260 °C. 1H NMR (DMSO-d6): δ 0.74-0.82 (t, 3 H, CH2CH2CH3), 1.43 - 1.45 (m, 2 H, CH2CH2CH3), 1.93-2.08 (m, 2 H, Glu β-CH2), 2.51 (t, 2 H, Glu γ-CH2), 2.99 (t, 2 H, CH2CH2CH3), 4.37 (m, 1 H, Glu α-CH), 5.67 (s, 2 H, 2/4-NH2), 6.26 (s, 2 H, 2/4-NH2), 7.04-7.07 (d, 2 H, C6H4), 7.76-7.78 (d, 2, H, C6H4), 8.51-8.53 (d, 1 H, CONH), 11.09 (s, 1 H, 7-NH), 12.37 (bs, 2 H, COOH). Anal. calcd for (C21H24N6SO5 · 1.0 H2O) C, H, N, S.</p><!><p>In a 1000 mL flask were placed paraformaldehyde (9.45 g, 0.3 mol), 3-ethylbenzothiazolium bromide (7.32 g, 0.03 mol), isobutraldehyde, 35, (27.5 mL, 0.3 mol), anhydrous EtOH (300 mL), and Et3N (4.2 mL, 0.03 mol), and then dry N2 gas was bubbled into the reaction mixture. The mixture was then heated in an oil bath at 60 °C for 72 h, during which time the color of the reaction mixture changed to dark reddish-brown. The solvent was evaporated to dryness under reduced pressure, and to the resulting residue was added EtOAc (20 mL). The resulting suspension was filtered, and the solid was washed repeatedly with EtOAc. The filtrate was evaporated under reduced pressure to a dark-brown oil, which was distilled under low pressure to afford 10.7 g (35%) of 36 as a colorless oil: bp = 65-68 °C (16 mmHg) [lit.28 bp = 65 °C (20 mmHg)]. 1H NMR (CDCl3-d): δ 1.15 - 1.17 (d, 6 H, CH(CH3)2), 2.60-2.69 (m, 1 H, CH(CH3)2), 3.13 (bs, 1 H, OH), 4.32 (s, 2 H, CH2).</p><!><p>A mixture of malonodinitrile (12.55 g, 190 mmol) and Et3N (19.19 g, 190 mmol) in MeOH (220 mL) was added dropwise to a solution of the a-hydroxy ketone 36 (19.4 g, 190 mmol) in MeOH (10 mL) and the resulting solution was stirred at room temperature for 24 h. To this solution was added silica gel (50 g), and the solvent was evaporated under reduced pressure to afford a dry silica gel plug, which was loaded on top of a wet (hexane) silica gel column and eluted first with hexane and then with 2:1 hexane/EtOAc to afford the furan 37 (11.1 g, 73%) as a reddish-brown solid: mp 62-62.5 °C; TLC Rf = 0.64 (hexane/EtOAc, 2:1). 1H NMR (DMSO-d6): δ 1.13 - 1.15 (d, 6 H, CH(CH3)2), 2.59-2.65 (m, 1 H, CH(CH3)2), 6.71 (s, 1 H, 5-CH), 7.23 (s, 2 H, 2-NH2) Anal. calcd for (C8H10N2O) C, H, N.</p><!><p>Amino nitrile furan 37 (1.5 g, 10 mmol) was added to a solution prepared from guanidine hydrochloride (1.43 g, 15 mmol) and NaOMe (0.81 g, 15 mmol) in anhydrous EtOH (100 mL). The resulting dark-red reaction mixture was stirred under reflux for 96 h, during which time it turned dark-brown. To this solution was added silica gel (15 g), and the solvent was evaporated to dryness under reduced pressure to afford a silica gel plug, which was loaded on top of a wet (CHCl3) silica gel column and eluted first with CHCl3 and then with a gradient of 1-5% MeOH in CHCl3 to give 1.15 g (60%) of 38 as a white solid: mp 221 - 222 °C; TLC Rf = 0.38 (CHCl3/MeOH, 5:1). 1H NMR (DMSO-d6): δ 1.17-1.19 (d, 6 H, CH(CH3)2), 3.17 (m, 1 H, CH(CH3)2),5.33 (s, 2 H, 2/4-NH2), 5.90 (s, 2 H, 2/4-NH2), 6.39 (s, 1 H, 6-CH), 10.35 (s, 1 H, 7-NH). Anal. calcd for (C9H13N5) C, H, N.</p><!><p>To a suspension of 38 (1.0 g, 5.2 mmol) in a mixture of EtOH/H2O (2:1, 75 mL) was added diethyl 4,4′-dithiobis(benzoate) (2.2 g, 6 mmol) and the suspension was heated to 100-110 °C, then I2 (3 g, 12 mmol) was added and the reaction was monitored for completion (3 h). To this solution was added excess sodium thiosulfate and the solution was evaporated to dryness under reduced pressure and the resulting residue was washed with water and air-dried. This residue was then dissolved in MeOH (100 mL) and to this was added silica gel (15 g), and the resulting suspension was evaporated to dryness under reduced pressure to afford a dry silica gel plug, which was loaded on top of a wet (CHCl3) silica gel column and eluted first with CHCl3 and then with a gradient of 1-5% MeOH in CHCl3. Fractions containing the desired spot (TLC) were pooled and evaporated to dryness to afford 890 mg (45%) of 39 as a white solid: mp = 263-264.7 °C; TLC Rf = 0.61 (CHCl3/MeOH, 5:1). 1H NMR (DMSO-d6): δ 1.23-1.29 (m, 9 H, CH2CH3 and CH(CH3)2), 3.33-3.41 (m, 1 H, CH(CH3)2), 4.23-4.29 (q, 2 H, CH2CH3), 5.65 (s, 2 H, 2/4-NH2), 6.15 (s, 2 H, 2/4-NH2), 7.05-7.08 (d, 2 H, C6H4), 7.82-7.84 (d, 2 H, C6H4), 11.03 (s, 1 H, 7-NH). Anal. calcd for (C18H21N5O2S·0.4H2O) C, H, N, S.</p><!><p>To a suspension of 39 (530 mg, 1.43 mmol) in EtOH (50 mL) was added aqueous 1N NaOH (20 mL) and the reaction mixture was stirred at 80 °C for 24 h. At this point, TLC indicated the disappearance of the starting ester at Rf = 0.54 (CHCl3/MeOH, 5:l) and formation of one major spot at the origin. The solvent was evaporated to dryness, and the resulting sodium salt (yellow oil) was dissolved in water (15 mL) and carefully acidified to pH 4 by dropwise addition of 3N HCl. The resulting suspension was filtered and washed carefully with cold water and dried over P2O5 to afford 436 mg (90%) of 40 as a white solid. 1H NMR (DMSO-d6): δ 1.24-1.25 (d, 6 H, CH(CH3)2), 3.35 (m, 1 H, CH(CH3)2), 5.60 (s, 2 H, 2/4-NH2), 6.07 (s, 2 H, 2/4-NH2), 6.93-6.95 (d, 2 H, C6H4), 7.75-7.76 (d, 2 H, C6H4), 10.98 (s, 1 H, 7-NH). Anal. calcd for (C16H17N5O2S) MS (EI) calcd m/z = 343.110297; found m/z = 343.109307 (M+).</p><!><p>To a suspension of the acid 40 (300 mg, 0.87 mmol) in anhydrous DMF (25 mL) under N2 was added N-methylmorpholine (145 μL, 1.33 mmol) and the resulting suspension was cooled to 0 °C. At this point, 2-chloro-4,6-dimethoxy-1,3,5-triazine (235 mg, 1.34 mmol) was added and the suspension was stirred for 2 h, during which time it formed a solution. The reaction mixture was again cooled to 0 °C and diethyl-l-glutamate (317 mg, 1.33 mmol) was added followed by N-methylmorpholine (145 μL, 1.33 mmol). The solution was slowly allowed to warm to room temperature with stirring and left at room temperature for a total of 24 h. To the resulting solution was added silica gel (5 g) and the DMF was evaporated using an oil pump. The silica gel plug was loaded on a wet (CHCl3) silica gel column and eluted with a gradient of 1-3% MeOH in CHCl3. Fractions containing the desired spot (TLC) were pooled and evaporated to dryness under vacuum to give 330 mg (70%) of 41 as a white solid: mp 217.6-218 °C; TLC Rf = 0.53 (CHCl3/MeOH, 5:1). 1H NMR (DMSO-d6): δ 1.12-1.19 (m, 6 H, CH2CH3), 1.24-1.27 (d, 6 H, CH(CH3)2), 1.97-2.07 (m, 2 H, Glu β-CH2), 2.39-2.44 (t, 2 H, Glu γ-CH2), 3.99-4.12 (m, 4 H, CH2CH3), 4.40 (m, 1 H, Glu α-CH), 5.63 (s, 2 H, 2/4-NH2), 6.12 (s, 2 H, 2/4-NH2), 7.02-7.05 (d, 2 H, C6H4), 7.74-7.77 (d, 2 H, C6H4), 8.63-8.65 (d, 1 H, CONH), 11.01 (s, 1 H, 7-NH). Anal. calcd for (C25H32N6O5-S ·0.5H2O) C, H, N, S.</p><!><p>To a suspension of 41 (200 mg, 0.37 mmol) in EtOH (15 mL) was added 1N NaOH (6 mL) and the suspension stirred at 0 °C (4 h) and then at room temperature for 24 h. The EtOH was evaporated to dryness under reduced pressure, the yellow oil was dissolved in water (5 mL), and the solution was cooled in an ice-bath and acidified carefully to pH 4.0 with dropwise addition of 3N HCl. This suspension was left at 5 °C for 24 h and filtered. The residue was washed well with water and O(C2H5)2 and then dried over P2O5/vacuum to afford 165 mg (80%) of 4 as a white solid: mp 206.5-207 °C. 1H NMR (DMSO-d6): δ 1.25 - 1.27 (d, 6 H, CH(CH3)2), 1.93-2.06 (m, 2 H, Glu β-CH2), 2.31 - 2.33 (t, 2 H, Glu γ-CH2), 4.36 (m, 1 H, Glu α-CH), 5.78 (s, 2 H, 2/4-NH2), 6.29 (s, 2 H, 2/4-NH2), 7.03-7.06 (d, 2 H, C6H4), 7.76-7.79 (d, 2 H, C6H4), 8.51-8.54 (d, 1 H, CONH), 11.13 (s, 1 H, 7-NH), 12.4 (bs, 2 H, COOH). Anal. calcd for (C21H24N6O5S· 0.2H2O· 1.8C4H10O) C, H, N, S.</p><!><p>To a solution of 31 (300 mg, 1.57 mmol) in a mixture of EtOH/water (2:1, 30 mL) was added 2,6-dichlorophenylthiol (540 mg, 3.00 mmol) and the reaction mixture was heated to 100-110 °C, then I2 (750 mg, 3.00 mmol) was added and the heating continued with stirring for a total of 3 h. To this mixture was added an excess of sodium thiosulfate and the reaction mixture concentrated under reduced pressure. To the resulting residue was added silica gel (10 g) and MeOH (50 mL) and the solution evaporated to dryness under reduced pressure to afford a dry silica gel plug, which was loaded on top of a wet (CHCl3) silica gel column and eluted with a gradient of 1-3% MeOH in CHCl3. Fractions containing the desired spot (TLC) were pooled, and evaporated to dryness. The resulting residue was washed with MeOH, filtered, and dried to yield 385 mg (67%) of 11: mp 238-240 °C; TLC Rf = 0.51 (CHCl3/MeOH, 5:1, with 2 drops of NH4OH). 1H NMR (DMSO-d6): δ 0.76 (t, 3 H, 5-CH2CH2CH3), 1.15 (m, 2 H, 5-CH2CH2CH3), 2.64 (t, 2 H, 5-CH2CH2CH3), 5.54 (s, 2 H, 2/4-NH2), 6.08 (s, 2 H, 2/4-NH2), 7.32-7.35 (m, 1 H, C6H3), 7.48-7.51 (d, 2 H, C6H3), 10.93 (s, 1 H, 7-NH). Anal. calcd for (C15H15N5Cl2S·0.17CHCl3) C, H, N, Cl, S.</p><!><p>Compound 12 was synthesized as described for 11 using 2,6-dimethylphenylthiol (420 mg, 3.00 mmol) and 31 (300 mg, 1.57 mmol): yield 52%; mp 227-230 °C; TLC Rf = 0.55 (CHCl3/MeOH, 5:1, with 2 drops of NH4OH). 1H NMR (DMSO-d6): δ 0.74-0.78 (t, 3 H, 5-CH2CH2CH3), 1.16-1.25 (m, 2 H, 5-CH2CH2CH3), 2.19-2.25 (t, 2 H, 5-CH2CH2CH3), 2.35 (s, 6 H, 2′,6′-diCH3), 5.46 (s, 2 H, 2/4-NH2), 6.00 (s, 2 H, 2/4-NH2), 7.07 (m, 3 H, C6H3), 10.75 (s, 1 H, 7-NH). Anal. calcd for (C17H21N5S·0.6CH3OH) C, H, N, S.</p><!><p>Compound 13 was synthesized as described for 11 using phenylthiol (280 mg, 2.00 mmol) and 31 (200 mg, 1.04 mmol): yield 65%; mp 252.2-252.7 °C; TLC Rf = 0.53 (CHCl3/MeOH, 5:1, with 2 drops of NH4OH). 1H NMR (DMSO-d6): δ 0.81-0.86 (t, 3 H, 5-CH2CH2CH3), 1.43 - 1.45 (m, 2 H, 5-CH2CH2CH3), 2.70-2.75 (t, 2 H, 5-CH2CH2CH3), 5.59 (s, 2 H, 2/4-NH2), 6.15 (s, 2 H, 2/4-NH2), 7.00-7.02 (d, 2 H, C6H5), 7.12-7.14 (m, 1 H, C6H5), 7.24-7.29 (m, 2 H, C6H5), 10.98 (s, 1 H, 7-NH). Anal. calcd for (C15H17N5S·0.2H2O) C, H, N, S.</p><!><p>Compound 14 was synthesized as described for 11 using 4-methoxyphenylthiol (280 mg, 2.00 mmol) and 31 (200 mg, 1.04 mmol): yield 50%; mp 247.9-248.2 °C; TLC Rf = 0.63 (CHCl3/MeOH, 5:1, with 2 drops of NH4OH). 1H NMR (DMSO-d6): δ 0.82-0.87 (t, 3 H, 5-CH2CH2CH3), 1.41 (m, 2 H, 5-CH2CH2CH3), 2.74 (t, 2 H, 5-CH2CH2CH3), 3.69 (s, 3 H, 4′-OCH3), 5.55 (s, 2 H, 2/4-NH2), 6.11 (s, 2 H, 2/4-NH2), 6.85-6.88 (d, 2 H, C6H4), 7.04-7.07 (d, 2 H, C6H4), 10.95 (s, 1 H, 7-NH). Anal. calcd for (C16H19N5OS·0.5H2O) C, H, N, S.</p><!><p>Compound 15 was synthesized as described for 11 using 2,5-dimethoxyphenylthiol (525 mg, 3.00 mmol) and 31 (300 mg, 1.57 mmol): yield 45%; mp 217-218 °C; TLC Rf = 0.56 (CHCl3/MeOH, 5:1, with 2 drops of NH4OH). 1H NMR (DMSO-d6): δ 0.81-0.86 (t, 3 H, 5-CH2CH2CH3), 1.40-1.47 (m, 2 H, 5-CH2CH2CH3), 2.66-2.71 (t, 2 H, 5-CH2CH2CH3), 3.53 (s, 3 H, 2′/5′-OCH3), 3.80 (s, 3 H, 2′/5′-OCH3), 5.59 (s, 2 H, 2/4-NH2), 6.17 (s, 2 H, 2/4-NH2), 5.96 (s, 1 H, C6H3), 6.64-6.67 (d, 1 H, C6H3), 6.89-6.92 (d, 1 H, C6H3), 10.91 (s, 1 H, 7-NH). Anal. calcd for (C17H21N5O2S) C, H, N, S.</p><!><p>Compound 16 was synthesized as described for 11 using 3,4-dimethoxyphenylthiol (350 mg, 2.00 mmol) and 31 (200 mg, 1.04 mmol): yield 65%; mp >230 °C (dec); TLC Rf = 0.53 (CHCl3/MeOH, 5:1, with 2 drops of NH4OH). 1H NMR (DMSO-d6): δ 0.83-0.88 (t, 3 H, 5-CH2CH2CH3), 1.43-1.45 (m, 2 H, 5-CH2CH2CH3), 2.73-2.75 (t, 2 H, 5-CH2CH2CH3), 3.68 (s, 6 H, 3′,4′-diOCH3), 5.59 (s, 2 H, 2/4-NH2), 6.15 (s, 2 H, 2/4-NH2), 6.59-6.62 (d, 1 H, C6H3), 6.82-6.88 (m, 2 H, C6H3), 10.98 (s, 1 H, 7-NH). Anal. calcd for (C17H21N5O2S·0.2CHCl3) C, H, N, S.</p><!><p>Compound 17 was synthesized as described for 11 using 1-naphthylthiol (320 mg, 2.00 mmol) and 31 (200 mg, 1.04 mmol): yield 58%; mp > 255 °C dec; TLC Rf = 0.63 (CHCl3/MeOH, 5:1, with 2 drops of NH4OH). 1H NMR (DMSO-d6): δ 0.80-0.84 (t, 3 H, 5-CH2CH2CH3), 1.43 - 1.45 (m, 2 H, 5-CH2CH2CH3), 2.73-2.75 (t, 2 H, 5-CH2CH2CH3), 5.62 (s, 2 H, 2/4-NH2), 6.19 (s, 2 H, 2/4-NH2), 6.86 (d, 1 H, C10H7), 7.37 (t, 1 H, C10H7), 7.58-7.61 (m, 3 H, C10H7), 7.70 (d, 1 H, C10H7), 7.94 (d, 1 H, C10H7), 11.05 (s, 1 H, 7-NH). Anal. calcd for (C19H19N5S) C, H, N, S.</p><!><p>Compound 18 was synthesized as described for 11 using 2-naphthylthiol (480 mg, 3.00 mmol) and 31 (300 mg, 1.57 mmol): yield 70%; mp > 250 °C (dec); TLC Rf = 0.56 (CHCl3/MeOH, 5:1, with 2 drops of NH4OH). 1H NMR (DMSO-d6): δ 0.80-0.85 (t, 3 H, 5-CH2CH2CH3), 1.42-1.49 (m, 2 H, 5-CH2CH2CH3), 2.74-2.79 (t, 2 H, 5-CH2CH2CH3), 5.60 (s, 2 H, 2/4-NH2), 6.18 (s, 2 H, 2/4-NH2), 7.17-7.20 (d, 1 H, C10H7), 7.40-7.50 (m, 3 H, C10H7), 7.73-7.76 (d, 1 H, C10H7), 7.81 (s, 1 H, C10H7), 7.84-7.85 (d, 1 H, C10H7), 11.06 (s, 1 H, 7-NH). Anal. calcd for (C19H19N5S·0.5H2O) C, H, N, S.</p><!><p>To a solution of 38 (300 mg, 1.57 mmol) in a mixture of EtOH/water (2:1, 30 mL) was added 2,6-dichlorophenylthiol (540 mg, 3.00 mmol) and the reaction mixture was heated to 100-110 °C, then I2 (750 mg, 3.0 mmol) was added and the heating continued with stirring for a total of 2 h. To this mixture was added an excess of sodium thiosulfate and the reaction mixture concentrated under reduced pressure. To the resulting residue was added silica gel (10 g) and MeOH (50 mL) and the solution evaporated to dryness under reduced pressure to afford a dry silica gel plug, which was loaded on top of a wet (CHCl3) silica gel column and eluted with a gradient of 1-3% MeOH in CHCl3. Fractions containing the desired spot (TLC) were pooled and evaporated to dryness. The resulting residue was washed with MeOH, filtered, and dried to yield 70 mg (12%) of 19: mp 230-231 °C; TLC Rf = 0.60 (CHCl3/MeOH, 5:1, with 2 drops of NH4OH). 1H NMR (DMSO-d6): δ 1.10-1.12 (d, 6 H, CH(CH3)2), 5.54 (s, 2 H, 2/4-NH2), 5.87 (s, 2 H, 2/4-NH2), 7.31-7.33 (m, 1 H, C6H3), 7.46-7.49 (d, 2 H, C6H3), 10.90 (s, 1 H, 7-NH). Anal. calcd for (C15H15N5Cl2S·0.3H2O) C, H, N, Cl, S.</p><!><p>Compound 20 was synthesized as described for 19 using 2,6-dimethylphenylthiol (420 mg, 3.00 mmol) and 38 (300 mg, 1.57 mmol): yield 22%; mp 248-248.3 °C; TLC Rf = 0.63 (CHCl3/MeOH, 5:1, with 2 drops of NH4OH). 1H NMR (DMSO-d6): δ 1.11 - 1.14 (d, 6 H, 5-CH(CH3)2), 2.30 (s, 6 H, 2′,6′-diCH3), 5.46 (s, 2 H, 2/4-NH2), 5.79 (s, 2 H, 2/4-NH2), 7.08 (m, 3 H, C6H3), 10.66 (s, 1 H, 7-NH). Anal. calcd for (C17H21N5S·0.2H2O) C, H, N, S.</p><!><p>Compound 21 was synthesized as described for 19 using phenylthiol (330 mg, 3.00 mmol) and 38 (300 mg, 1.57 mmol) except that the compound was washed with hexane and dried. Yield: 45%; mp 228.5-229 °C; TLC Rf = 0.53 (CHCl3/MeOH, 5:1, with 2 drops of NH4OH). 1H NMR (DMSO-d6): δ 1.25 - 1.27 (d, 6 H, 5-CH(CH3)2), 5.39 (bs, 2 H, 2/4-NH2), 5.89 (bs, 2 H, 2/4-NH2), 6.97-7.29 (m, 5 H, C6H5), 10.97 (s, 1 H, 7-NH). Anal. calcd for (C15H17N5S·0.1C6H14) C, H, N, S.</p><!><p>Compound 22 was synthesized as described for 19 using 4-methoxyphenylthiol (280 mg, 2.00 mmol) and 38 (200 mg, 1.04 mmol): yield 40%; mp 288-289 °C; TLC Rf = 0.58 (CHCl3/MeOH, 5:1, with 2 drops of NH4OH). 1H NMR (DMSO-d6): δ 1.26-1.28 (d, 6 H, 5-CH(CH3)2), 3.41-3.48 (m, 1 H, 5-CH(CH3)2), 3.70 (s, 3 H, 4-OCH3), 5.56 (bs, 2 H, 2/4-NH2), 6.02 (bs, 2 H, 2/4-NH2), 6.86-6.89 (d, 2 H, C6H4), 7.02-7.04 (d, 2 H, C6H4), 10.93 (s, 1 H, 7-NH). Anal. calcd for (C16H19N5-SO·0.1H2O) C, H, N, S.</p><!><p>Compound 23 was synthesized as described for 19 using 2,5-dimethoxyphenylthiol (800 mg, 4.00 mmol) and 38 (350 mg, 2.00 mmol): yield 53%; mp 243-244 °C; TLC Rf = 0.60 (CHCl3/MeOH, 5:1, with 2 drops of NH4OH). 1H NMR (DMSO-d6): δ 1.24-1.26 (d, 6 H, 5-CH(CH3)2), 3.53 (s, 3 H, 2′/5′-OCH3), 3.80 (s, 3 H, 2′/5′-OCH3), 5.59 (bs, 2 H, 2/4-NH2), 5.93 (s, 1 H, C6H3), 6.07 (bs, 2 H, 2/4-NH2), 6.63-6.66 (d, 1 H, C6H3), 6.89-6.92 (d, 1 H, C6H3), 10.87 (s, 1 H, 7-NH). Anal. calcd for (C17H21N5SO2) C, H, N, S.</p><!><p>Compound 24 was synthesized as described for 19 using 3,4-dimethoxyphenylthiol (520 mg, 3.00 mmol) and 38 (300 mg, 1.57 mmol): yield 58%; mp 293-293.5 °C; TLC Rf = 0.58 (CHCl3/MeOH, 5:1, with 2 drops of NH4OH). 1H NMR (DMSO-d6): δ 1.27-1.29 (d, 6 H, 5-CH(CH3)2), 3.67 (s, 3 H, 3′/4′-OCH3), 3.69 (s, 3 H, 3′/4′-OCH3), 5.56 (bs, 2 H, 2/4-NH2), 6.01 (bs, 2 H, 2/4-NH2), 6.56-6.59 (d, 1 H, C6H3), 6.79 (s, 1 H, C6H3), 6.87-6.89 (d, 1 H, C6H3), 10.93 (s, 1 H, 7-NH). Anal. calcd for (C17H21N5SO2 ·0.5H2O) C, H, N, S.</p><!><p>Compound 25 was synthesized as described for 19 using 1-naphthylthiol (320 mg, 2.00 mmol) and 38 (200 mg, 1.04 mmol): yield 45%; mp 267-267.5 °C; TLC Rf= 0.60 (CHCl3/MeOH, 5:1, with 2 drops of NH4OH). 1H NMR (DMSO-d6): δ 1.26-1.28 (d, 6 H, 5-CH(CH3)2), 5.63 (s, 2 H, 2/4-NH2), 6.11 (s, 2 H, 2/4-NH2), 6.76-6.78 (d, 1 H, C10H7), 7.37 (t, 1 H, C10H7), 7.59-7.72 (m, 3 H, C10H7), 7.94 (d, 1 H, C10H7), 8.2 (d, 1 H, C10H7), 11.02 (s, 1 H, 7-NH). Anal. calcd for (C19H19N5S) C, H, N, S.</p><!><p>Compound 26 was synthesized as described for 19 using 2-naphthylthiol (640 mg, 4.00 mmol) and 38 (350 mg, 1.83 mmol): yield 40%; mp 247-247.5 °C; TLC Rf = 0.58 (CHCl3/MeOH, 5:1, with 2 drops of NH4OH). 1H NMR (DMSO-d 6): δ 1.27-1.29 (d, 6 H, 5-CH(CH3)2), 5.63 (s, 2 H, 2/4-NH2), 6.12 (s, 2 H, 2/4-NH2), 7.15-7.18 (d, 1 H, C10H7), 7.44-7.47 (m, 3 H, C10H7), 7.74-7.76 (d, 1 H, C10H7), 7.82-7.85 (d, 2 H, C10H7), 11.05 (s, 1 H, 7-NH). Anal. calcd for (C19H19N5S) C, H, N, S.</p><!><p>Compound 27 was synthesized as described for 19 using 3,4-dichlorophenylthiol (540 mg, 3.00 mmol) and 38 (300 mg, 1.57 mmol): yield 37%; mp 244-244.5 °C; TLC Rf = 0.58 (CHCl3/MeOH, 5:1, with 2 drops of NH4OH). 1H NMR (DMSO-d6): δ 1.24-1.27 (d, 6 H, 5-CH(CH3)2), 5.67 (bs, 2 H, 2/4-NH2), 6.17 (bs, 2 H, 2/4-NH2), 6.96 (s, 2 H, C6H3), 7.37 (s, 1 H, C6H3), 11.04 (s, 1 H, 7-NH). Anal. calcd for (C15H15N5SCl2) C, H, N, S, Cl.</p>

PubMed Author Manuscript

Time of day influences the voluntary intake and behavioral response to methamphetamine and food reward

The circadian timing system influences a vast array of behavioral responses. Substantial evidence indicates a role for the circadian system in regulating reward processing. Here we explore time of day effects on drug anticipation, locomotor activity, and voluntary methamphetamine (MA) and food intake in animals with ad libitum food access. We compared responses to drug versus a palatable treat during their normal sleep times in early day (zeitgeber time (ZT) 0400) or late day (ZT 1000). In the first study, using a between-subjects design, mice were given daily 1-h access to either peanut butter (PB-Alone) or to a low or high concentration of MA mixed in PB (MA + PB). In study 2, we repeated the experiment using a within-subjects design in which mice could choose between PB-Alone and MA + PB at either ZT 0400 or 1000. In study 3, the effects of MA-alone were investigated by evaluating anticipatory activity preceding exposure to nebulized MA at ZT 0400 vs. ZT 1000. Time of day effects were observed for both drug and palatable treat, such that in the between groups design, animals showed greater intake, anticipatory activity, and post-ingestional activity in the early day. Furthermore, there were differences among mice in the amount of MA ingested but individuals were self-consistent in their daily intake. The results for the within-subjects experiment also revealed robust individual differences in preference for MA + PB or PB-Alone. Interestingly, time of day effects on intake were observed only for the preferred substance. Anticipatory activity preceding administration of MA by nebulization was also greater at ZT 0400 than ZT 1000. Finally, pharmacokinetic response to MA administered intraperitoneally did not vary as a function of time of administration. The results indicate that time of day is an important variable mediating the voluntary intake and behavioral effects of reinforcers.

time_of_day_influences_the_voluntary_intake_and_behavioral_response_to_methamphetamine_and_food_rewa

1. Introduction<!>2.1. Animals and housing<!>2.2. Preparation of drugs<!>2.3.1. Experimental groups<!>2.3.2. Measures<!>2.3.3. Data analysis<!>2.4. Protocol 2: Study design<!>2.4.1. Experimental groups<!>2.4.2. Measures<!>2.4.3. Data analysis<!>2.5. Protocol 3: Study design<!>2.5.1. Experimental groups<!>2.5.2. Measures<!>2.5.3. Data analysis<!>2.6.1. Experimental groups<!>2.6.2. Measures<!>2.6.3. Data analysis<!>3.1.1. Amount eaten<!>3.1.2. MA intake<!>3.1.3. Activity<!>3.2.1. Preference<!>3.2.2. Amount eaten<!>3.2.3. MA intake<!>3.2.4. First approach<!>3.2.5. Activity<!>3.3.1. Anticipatory activity<!>3.4.1. Serum MA concentration<!>3.4.2. Pharmacokinetic parameters<!>4. Discussion<!>4.1. Time of day effects on drug intake<!>4.2. TOD effects on MA-induced activity<!>4.3. TOD effects on anticipation<!>4.4. TOD effects on reinforcer type<!>4.5. Individual differences in inbred mice<!>5. Overall conclusions

<p>The circadian timing system has a pervasive influence in that it modulates numerous behavioral and physiological responses, including the response to natural and drug reinforcers (Hasler et al., 2012). Indeed, for several types of reinforcers the pharmacological, physiological, and behavioral effects vary as a function of time of administration or availability over a 24-h cycle (Falcon and McClung, 2009; Webb et al., 2009a). These rhythms persist under constant conditions (Terman and Terman, 1975; Kosobud et al., 1998), suggesting that they are under endogenous circadian control by the brain clock located in the suprachiasmatic nucleus of the hypothalamus. Although there have been studies investigating the influence of time of day on the behavioral responses to drugs of abuse, surprisingly methamphetamine (MA) has received limited experimental attention. In humans, time-of-day effects may influence acute subjective, cognitive, and adverse effects of MA.</p><p>Data from participants in a prior experiment in our laboratory suggest that time of day influences of the euphoric effects of MA. When participants received MA at 0115 h, ratings of "good drug effects" were similar across low and moderate doses (5 versus 10 mg) (Hart et al., 2003). In contrast, unpublished data from this experiment reveal that when the same participants received 5 mg MA at 0915, their ratings of "good drug effects" were indistinguishable from ratings for placebo [Fig. S1.a (data) and b (study design)], (Supplementary material; Hart et al., 2003). While this experiment was not designed to examine the influence of time of day, the results do raise a question about how such an effect might influence drug self-administration: a question optimally addressed in studies of laboratory animals.</p><p>Although there have been few studies of time-of day effects of MA in humans (Shappell et al., 1996), diurnal variations in response to amphetamines have been reported in laboratory animals using a variety of procedures including operant avoidance, sensitization, tolerance, general activity, conditioned place preference (CPP), and stereotypic behavior (Arvanitogiannis et al., 2000; Evans et al., 1973; Gaytan et al., 1998a,b; Gaytan et al., 1999; Kuribara and Tadokoro, 1982, 1984; Martin-Iverson and Iversen, 1989; Uchihashi et al., 1994; Urba-Holmgren et al., 1977; Webb et al., 2009b). Overall, these data show greater drug effects around dawn compared to dusk. However, to our knowledge, there has been no prior attempt to explore the impact of diurnal variations on self-administration of amphetamines.</p><p>The goal of the present experiment was to examine time of day effects on reinforcer intake, associated behaviors, and pharmacokinetics. A second goal was to compare time of day effects of two different reinforcers, specifically a palatable treat, peanut butter (PB-Alone), versus drug, methamphetamine (MA) mixed in peanut butter (MA + PB). We also investigated dose–response relationships and individual differences in these behaviors. Finally, we sought to explore changes in these behaviors over time. To investigate these questions, we used a paradigm involving voluntary intake, thereby allowing for simultaneous measurement of anticipatory behaviors, self-administered voluntary intake of drug and/or palatable treat, and locomotor activity. This paradigm is analogous to voluntary human drug use, and does not require surgical implantation of an indwelling catheter for acquisition of self-administration data. Further, because the mice are provided with food ad libitum they have very low activity levels during the day, allowing assessment of responses to reinforcers against low baselines (Mistlberger, 1994; Escobar et al., 2011). In Protocol 1, we used a between-subjects design to compare the behavioral responses to drug and/or palatable treat in the early versus late day. In that work, we noted that marked individual differences in MA intake with self-consistent responses over the course of the experiment. Protocol 2 used a within-subjects design, thereby permitting more detailed measurement of individual differences in intake and time of day effects. To isolate the effects of MA from PB, in Protocol 3 we investigated time of day effects on anticipatory activity associated with nebulized MA. Finally, to assess time of day effects on pharmacokinetic factors, in Protocol 4 MA was injected at several times of day and serum measurement of the drug were taken for the subsequent 4 h.</p><!><p>Adult male C57BL/6 N mice 6 weeks of age, weighing an average of 22 g (range 16°26 g) at the beginning of each experiment were subjects (Charles River, Wilmington, MA). Mice were housed individually in transparent polycarbonate cages (32 × 14 × 13 cm), equipped with a running wheel (diameter, 11 cm) placed in sound attenuating, ventilated chambers (Phenome Technologies Inc. Lincolnshire, IL). The room was maintained at 23 ± 2 °C and 72% humidity. Standard mouse chow(Purina, St. Louis, MO) and water were available ad libitum except as noted. The experimenter changed cages every two weeks, and the data from the 24 h following a cage change were not included in the analyses. For Protocols 1 and 2, body weight was taken on the first and last day of each experiment. For Protocol 3, to measure anticipation to nebulized MA, a skeleton photoperiod with lights on ZT 0000–0030 and ZT 1130–1200 was used to avoid the masking effects of light on activity. Nebulization was performed under dim red light (1 lx) illumination at ZT 0400–0415 or ZT 1000–1015. For all experiments, animals were adapted to a 12:12 light:dark cycle (200 lx), with lights off at zeitgeber time 1200 (ZT 1200) and on at ZT 0000 for 14–16 days before the start of the experiment. Animals were cared for in accordance with the Columbia University Institutional Animal Care and Use Committee and Animal Welfare regulations.</p><!><p>For Protocols 1 and 2, stock solutions of MA hydrochloride (Sigma-Aldrich Inc., St. Louis, MO) were prepared at two concentrations as follows: MA (34 or 68 mg) was added to distilled water (30 ml) to create 1.13 mg/ml and 2.26 mg/ml. PB was commercially available (Jif® Brand, Creamy Peanut Butter). Each animal was assigned its own Petri dish (BD Falcon, 35 × 10 mm tissue culture dish) for the duration of the study. For use, 1.00 ± 0.01 g PB was placed in the center the dish.MA stock solution or water (40 µl of 0, 1.13, or 2.26 mg/ml MA) was mixed thoroughly in the PB to create 45 or 90 µg/g MA + PB or PB-Alone. Ingestion of the entire mixture yields a dose of 2 mg/kg and 4 mg/kg, based on the mean initial body weight of 22 g/mouse. For Protocol 3 the concentration of MA was 0.4 and 1.0 mg/ml. For Protocol 4, body weight was measured immediately prior to injection of 2 mg/kg MA.</p><!><p>Animals were placed in one of four groups (N = 10/grp) differentiated by the time, either early (ZT 0400) or late day (ZT 1000) at which they were given access to either PB-Alone or MA + PB. The training and testing intervals are shown schematically for the ZT 0400 group in Fig. 1A. During an initial 4 day training period, animals were acclimated to food restriction conditions by giving them access to standard chow for 8-h/day from either ZT 0400–1200 or ZT 1000–1800. On days 5–8, animals were provided MA + PB (3 g PB mixed with 45 µg MA) or PB-Alone (3 g PB) for the same 8-h intervals. On days 9–12, they received 1-h access to 1 g of either 45 µg/g MA + PB or PB-Alone, followed by 7-h access to standard chow (ZT 0400–1200 or ZT 1000– 1800). On days 13–23 (Block 1), mice had ad-libitum access to chow, and daily 1-h access to 1 g of 45 µg/g MA + PB or PB-Alone at ZT 0400–0500 or ZT 1000–1100 continued. On days 24–34 (Block 2), the concentration of MA was doubled to 90 µg/g MA + PB, and subsequently returned on days 35–45 (Block 3) to 45 µg/g MA + PB.</p><!><p>The behavioral measures assessed were amount eaten, MA intake, anticipatory activity, activity after ingestion, and total daily activity. For determination of amount eaten, the experimenter was in the room for 20 min for placement and removal of Petri dishes. MA intake is reported in mg/kg body weight, adjusted for interpolated daily individual weight gain. Body weight on the last day of the study (day 45) averaged 28 g (range 24–31 g).Wheel running was monitored continuously using a computer-based data acquisition system, VitalView (Minimitter, Bend, OR, USA) and was quantified in 10 min bins across the 24 h day using Actiview(MiniMitter) and Excel (Microsoft). Activity was normalized within each animal to control for wheel resistance and was calculated by dividing the sum of activity counts in each activity measure by the number of wheel revolutions per bin averaged over 24 h. Anticipatory activity was defined as the average number of wheel revolutions in the 2-h prior to reinforcer access (ZT 0200–0400 or ZT 0800–1000). Activity after ingestion was the average wheel revolutions in the 2 h after the start of food access (ZT 0400–0600 or ZT 1000–1200).</p><!><p>Data was analyzed using a linear mixed model with both fixed and random effects in SAS (SAS Institute Inc., Cary, N.C., USA). One animal died during the first four days of training and data for this animal was not used. After removing outliers (or 2 s.d. from mean), observations from the last 10 days of each 11-day block were averaged for each animal (Total observations = 117). Independent variables included time of day (ZT 0400, ZT 1000), treatment group (PB-Alone,MA + PB), concentration (45, 90 µg/g MA + PB), and block (1, 2, and 3). Animal identity was analyzed as a random effect. Analyses were conducted in two ways: first by time of day, concentration, and block, and second by time of day, treatment group and block. To assess the possibility that the presence of the experimenter influenced the activity of the animals, activity after ingestion was analyzed including and excluding the first 20 min of access (ZT 420–600 and 1020 to 1200). No differences were found. We also examined the relationship between MA intake and activity after ingestion with a correlation using data from all 33 days of the experiment for each animal that received MA + PB (Total observations = 660). Finally, we examined the relationship between time of day and amount of activity after ingestion by analyzing all cases in which animals ingested comparable amounts of MA in early and late time points (0.55–0.75 mg/kg MA; ZT 4, N = 57 and ZT 10, N = 68; independent t test).</p><!><p>In a pilot study, we conducted a taste aversion study to see whether animals would eat MA + PB when PB-Alone was also available. Here, drug-naïve animals that had access to food for 6 h/day were presented with both PB-Alone and MA + PB for 15 min. Mice ate from both dishes, indicating that the taste of MA in PB was not aversive.</p><!><p>Because substantial individual differences among mice were found in the foregoing between subjects study, we next examined individual differences in a within subjects design. Here, mice received 8 days of food restriction training and 16 days of 1-h access to MA + PB and PB-Alone during ad libitum food availability, for a total of 24 days of the experiment. Animals were placed into two groups (N = 10/grp) differentiated by the time of day (ZT 0400 or ZT 1000) at which they were given a choice between one Petri dish containing MA + PB and another containing PB-Alone (Fig. 1B, shows the ZT 0400 group). During the training period, either MA + PB or PB-Alone was presented on alternate days for 1-h daily followed by 7-h access to standard chow. Here, on 1 day, a dish of MA + PB was placed one side of the home cage and on the next day, a dish of PB-Alone was placed on the opposite side, and so on. In order to create distinct environments within the home cage, vertical and horizontal striped paper was taped to the outside of the cage near the placement of each dish with the pattern held constant for each animal and counter-balanced across animals. For the experimental period, animals were given ad libitum access to standard chow and 1-h access to both dishes for 16 days with stripe patterns present, in four 4-day blocks of alternating concentrations of 45 or 90 µg/g MA + PB.</p><!><p>Some of the behavioral measures were identical to the first experiment, as follows: amount eaten (amount of MA + PB and of PB-Alone), MA intake, anticipatory activity, activity after ingestion, and total daily activity. In addition, MA preference and first approach were assessed. MA preference was defined as the percent MA + PB eaten out of the overall amount eaten each day [(MA + PB / (MA + PB + PB-Alone) * 100)]. First approach was defined as the dish eaten from first. The calculation of MA intake was adjusted for individual weight gain. On the last day of the study (day 24), average body weight of the mice was 25 g (range 22 to 27 g).</p><!><p>Data was analyzed using a linear mixed model in SAS with observations from the last 3 days of each 4-day block averaged for each animal, as in the first study. During Block 1 intake was 50% higher than in other blocks, likely due to the preceding period of food restriction, and this interval was omitted from further analysis (Total observations = 45).One animal died during the study and this data was not used. Independent variables included time of day (ZT 0400, ZT 1000), concentration (45, 90 µg/g MA + PB), and block (2, 3, and 4).</p><p>To categorize animals into MA- or PB-preferring groups, we calculated the MA + PB eaten/total amount eaten, for the last 12 days of the study (Blocks 2–4). The data were applied a one sample t-test to compare individuals' average preference to that expected by chance (no-preference or 50%). This yielded four animals with no preference, 8 preferred MA + PB and 7 preferred PB-Alone. Using these categories, we evaluated intake and activity of the MA + PB and PB-Alone preferring groups at each time of day for blocks in which at least 60% of the intake was from one dish. Finally, to explore whether time of day influenced amount of activity after ingestion, we analyzed cases in which meal size was kept within a narrow range — cases in which animals ingested 0.6–0.8 mg/kg MA (ZT 4, N = 20 and ZT 10, N = 13; independent t test). We also examined the relationship between MA intake and activity after ingestion by correlating these measures using data from all 16 days of the experiment for each animal (Total observations = 240).</p><!><p>In a pilot study, we confirmed that nebulized MA increased wheel running activity – measured as in Protocol 1 – (F(2, 13) = 5.846, p = .017: water = 1.94 ± 1.27; 0.4 mg/ml MA = 9.46 ± 3.50; 1 mg/ml MA: 25.98 ± 5.99). A dose of 1 mg/ml nebulized MA was selected as it elicited a similar amount of activity as IP injection of 4 mg/kg MA (1 mg/ml nebulized MA: 25.98 ± 5.99; 4 mg/kg MA (IP): 19.77 ± 4.42).</p><!><p>To measure anticipation of MA, animals were placed in one of four groups (MA: N = 6/grp; Control: N = 3/grp) differentiated by the time of day (ZT 0400, 1000) at which they were administered 1 mg/ml nebulized MA for 15 min at 10:00–10:15 AM for 16 consecutive days. Animals were adapted to the nebulization chamber (chambers: Brain Tree Scientific, Braintree, MA, catalog # MPC-3 AERO; Nebulizer: Briggs Medical Service Company, Waukegan, IL, catalog# 40-370-000) for 4 days at ZT 0400 for 15 min while nebulized with water. Then they were undisturbed for 2 days, received 0.4 mg/ml nebulized MA for 2 days, undisturbed for 2 days, received 1.0 mg/ml MA for 2 days.</p><!><p>Wheel running activity was collected and normalized as described in Protocol 1. Anticipatory activity was defined as the average number of wheel revolutions in the 2-h prior to nebulization (ZT 0200–0400 or ZT 0800–1000).</p><!><p>To test the effect of time of day on anticipatory activity in response to nebulized MA, data was analyzed using a two way ANOVA.</p><!><p>Animals were placed in one of four groups (N = 6/grp) differentiated by the time of day (ZT 0200, 0800, 1400, and 2000) at which they were administered a single injection of 2 mg/kg methamphetamine (IP). Six blood samples were taken: baseline, 15 min, 30 min, 1-h, 2-h, and 4-h post-injection.</p><!><p>Blood samples were collected via the lateral saphenous vein. For each time point, samples were collected into Microvette capillary tubes (Sarstedt, Numbrecht, DE) and allowed to clot. They were then centrifuged (Eppendorf centrifuge 5417r) at 14,000 rpm for 15 min at room temperature. Serum was removed from the centrifuged sample and frozen at −80 °C until analysis. MA levels were quantified in duplicate with sensitivity at 1 ng/ml by commercially available competitive enzyme linked immunosorbent assay (ELISA/EIA) kits from BioQuant, Inc. (San Diego, CA), following manufacturer's instructions. A 1 mg/ml MA standard purchased from Cerilliant (Round Rock, TX) was serially diluted to create standards (ranging from 2 µg/ml to 5 ng/ml) which were used for all analyses. Test samples were balanced across assay plates so that each plate had the same number of animals from each time of administration. Samples within each animal were kept together on a single assay plate.</p><!><p>To assess effects of ZT administration on serum levels of MA, a mixed-factor repeated measures analysis of variance (RMANOVA) was conducted with ZT time of administration as the between-subjects factor and blood sample time as the repeated measure. This analysis was conducted on raw serum [MA] and normalized values for each time-point (percent maximum blood conc.), with no difference in the results. For pharmacokinetic estimates, half-life values were calculated in GraphPad Prism software (San Diego, CA) by fitting the equation for one-phase exponential decay to the data. Area under the curve (AUC), a measure of gross drug effect, was also calculated in GraphPad Prism software using the trapezoid method. To assess whether ZT time of administration affected primary pharmacokinetic measures (half-life, AUC, time to peak serum concentration (Tmax), and peak concentration (Cmax)), data was analyzed using a one-way ANOVA.</p><!><p>There was a substantial effect of time of day, treatment, MA concentration, and block on the amount of PB and MA + PB eaten (Fig. 2A,B; smallest main effect F2,70 = 11.91, p < 0.001). It is noteworthy that in all cases, animals ate more at ZT 0400 than ZT 1000 in both PB-Alone and MA + PB groups (~30 and 50% difference, respectively). The MA + PB groups ate less of the mixture each day than did the PB-Alone groups. In addition, the MA + PB groups ate less of the mixture when the MA concentration was increased and this returned to prior levels when the lower MA concentration was reinstated (Fig. 2B). In contrast, the PB-Alone groups steadily increased their consumption over blocks of trials (Fig. 2A). Interestingly, animals in both groups were self-consistent in amount eaten, with 27% of the variability in the data ascribed to individual differences (t38 = 7.12, p < .0001, estimate = 27.29 ± 3.8).</p><!><p>The amount of MA ingested was significantly influenced by time of day and drug concentration (smallest F value of F2,70 = 4.44, p = 0.015). Animals ingested more MA at ZT 0400 than ZT 1000 (Fig. 2C, 45 µg/g corresponding to 2 mg/kg, t70 = 5.79, p < 0.001; 90 µg/g corresponding to 4 mg/kg; t70 = 9.55 p < 0.001, data not shown) and when the higher concentration was available (ZT 0400; t70 = 16.07, p < .0001; ZT 1000; t70 = 5.87, p < .0001, data not shown). There was a significant interaction between time of day and concentration (F2,70 = 28.01, p < 0.001), with a greater increase in MA intake at ZT 0400 when the higher concentration was introduced. Interestingly, as was the case for PB intake, there were marked individual differences among animals, but mice were self-consistent in MA intake (Fig. 2D) and this was true for each concentration (4 mg/kg; t18 = 6.0 p < 0.001; 2 mg/kg t18 = 5.82, p < 0.001). Individual differences accounted for 62% of the variability in the data (t38 = 12.29, p < .0001, estimate = 62.43 ± 5).</p><!><p>Several aspects of activity were tracked, including overall activity of each group, anticipatory activity prior to reinforcer availability, and activity following reinforcer access. Fig. 3 shows wheel-running activity for representative MA + PB and PB-Alone animals from ZT 0400 and 1000 groups (Fig. 3A–D) and for these groups as a whole (Fig. 3E, F). It is evident in these records that anticipatory activity occurred at ZT 0400 but not at ZT 1000 (Fig. 3E versus F). Furthermore, there was more activity after ingestion at ZT 0400 than ZT 1000 (Fig. 3E versus F).</p><p>There were no significant main effects of time of day, treatment, or concentration on total daily activity (data not shown). On the other hand, anticipatory activity was substantially influenced by time of day and block (Fig. 4A; smallest F value, F2,70 = 3.26, p = 0.04). More specifically, six animals from each group at ZT 0400 exhibited anticipatory activity, while none did so at ZT 1000 (PB-Alone group; t70 = 3.02, p = .003; MA + PB group; t70 = 2.6, p = .01).</p><p>Activity after ingestion was significantly affected by time of day with more activity at ZT 0400 than at ZT 1000 (Fig. 4B; F1,70 = 11.43, p = 0.001). However, this difference only reached significance in the MA + PB group (t70 = 3.11, p = .002). There was a significant correlation between MA intake and activity after ingestion at each time of day (R660 = 0.362, p < .0001), though at ZT 1000, the animals showed relatively little activity after MA treatment. To assess whether this was due to differences in intake, we examined data for days where comparable amounts of MA were ingested at ZT 0400 and 1000. Even when equivalent amounts of MA were ingested, there was more post ingestion activity at ZT 0400 versus 1000 (Fig. 4C right panel; t123 = 2.84, p = .005).</p><!><p>As in Protocol 1, at each concentration individuals were consistent in their preference for either MA + PB or PB-Alone, with the random effect of subjects accounting for 48% of the variability in preference scores (Fig. 5A; t18 = 3.68, p = 0.0019, estimate = 48.2 ± 13.1). Assessment of preferred reinforcer eaten/total eaten indicates that both MA + PB and PB-Alone preferring animals ate a greater percentage of the preferred substance at ZT 0400 than at 1000 (preference by time of day interaction, F1,16 = 6.76, p = 0.019). This difference reached significance only for the MA + PB group at the higher concentration (t16 = 2.14, p = .05, data not shown).</p><!><p>While there was no effect of MA concentration, there were main effects of time of day, preference, and preference by time of day interactions for both intake of PB-Alone or MA + PB (Fig. 5B; smallest F value, F1,16 = 8.64, p < 0.01 and Fig. 5C; smallest F value, F1,16 = 4.37, p = 0.05 respectively). The time of day effect on amount eaten was observed only for the preferred substance; PB-Alone intake was greater at ZT 0400 in the PB-preferring group but not in the MA + PB preferring group (left versus right panel of Fig. 5B; data shown for 2 mg/kg: t16 = 3.98, p = .0007; 4 mg/kg: t16 = 4.02, p = .0007). MA + PB intake was also greater at ZT 0400 only in the MA + PB preferring animals (left versus right panel of Fig. 5C; data shown for 2 mg/kg: t16 = 2.71, p = .014; 4 mg/kg: t16 = 2.43, p = .024).</p><!><p>There was a main effect of preference group on MA intake (Fig. 5D, E; F1,16 = 14.63, p = 0.001). Animals also ingested larger amounts of MA at ZT 0400 than ZT 1000, and MA intake increased with concentration, but these main effects did not reach significance (time of day, F1,16 = 3.85, p = 0.067; concentration, F1,16 = 4.22, p = 0.057). MA intake was significantly higher at ZT 0400 in the MA + PB preferring animals when the higher concentration was available (right panel of Fig. 5E; t16 = 2.57, p = .018).</p><!><p>Animals approached their preferred substance first (F1,16 = 61.57, p < 0.0001).</p><!><p>As evident from the actograms for individual animals (Fig. 6A–D) and in the group data (Fig. 6E, F), anticipatory activity was not seen at either ZT 0400 or 1000. The amount of activity following ingestion was greater at ZT 0400 as in Protocol 1.</p><p>As in Protocol 1 total daily activity was not affected by time of day or preference group (data not shown), but in contrast to the first study, animals did not show anticipatory activity (Fig. 7A).</p><p>As in Protocol 1, in activity after ingestion there were main effects of preference group and preference × time of day; (Fig. 7B; smallest F value was F1,16 = 11.86, p = 0.003). Time of day effects on activity were observed in MA-preferring animals (right panel, Fig. 7B; t16 = 2.7, p = .014), and there was a correlation between amount of MA ingested and amount of activity after ingestion (R240 = 0.264, p < 0.001). Greater activity at ZT 0400 cannot be attributed to greater MA intake as time of day effects were significant when MA intake (0.6 to 0.8 mg/kg) at ZT 0400 and 1000 was taken into account (Fig. 7C; t31 = 2.06, p = .048).</p><!><p>Animals showed anticipatory wheel running before nebulized MA but not water administration (F(1, 17) = 30.418, p < .001). Anticipatory activity was significantly higher at ZT 0400 than at ZT 1000 (Fig. 8; t10 = 1.869, p = .045).</p><!><p>Fig. 9 displays normalized serum [MA] up to 4 h after 2 mg/kg MA administration. A significant, time-dependent decline in [MA] was observed for all groups (F(4, 40) = 22.185, p = 0.000). There was no significant main effect of ZT time-of-day (F(3, 116) = 0.879, p = 0.47) or interaction between ZT and blood sample time.</p><!><p>There were no significant main effects of ZT time of administration on half-life, AUC, Cmax, or Tmax (largest F value, F(3, 116) = 1.440, p = 0.257).</p><!><p>A challenge in the development of animal models of drug use is to assess intake in a way that models human drug use as closely as possible. Some studies of diurnal variations in drug intake and effects involve access to drugs throughout the light/dark cycle (Deneau et al., 1969; Fitch and Roberts, 1993; Hollingsworth and Mueller, 1988; Lynch and Roberts, 2004; Lynch and Taylor, 2004; Martin-Iverson and Iversen, 1989). Restricted access paradigms, such as the one used here, more closely mimics drug availability for recreational human drug users. Furthermore, the present paradigm permits the experimenter to quantify a variety of measures that depend on free movement, and can be used in either a between- or a within-subjects design.</p><p>The present results demonstrate that under controlled conditions, time of day significantly affects reinforcer-related responses, both to palatable food and drug (Protocols 1 and 2) and to drug-alone (Protocols 1, 2, and 3). Further, these differences cannot be attributed to diurnal variations in pharmacokinetics (Protocol 4). Some effects were consistently seen in both the between and within-subjects design. Animals were more responsive to reinforcers in early day (ZT 0400) compared to late day (ZT 1000). Specifically, there were time of day differences in intake, anticipatory activity, and activity after ingestion in both the between (Protocol 1) and the within subjects study (Protocol 2). In the latter, time of day effects in intake were strongest for the preferred substance. Furthermore, robust, self consistent individual differences were observed in both experiments in MA intake and preference.</p><!><p>Diurnal variations in self-administration have been reported previously for another stimulant, cocaine, in restricted access conditions using indwelling catheters. Rats trained to self-administer cocaine show a diurnal variation in bar presses at low (but not at high) concentrations with more responses in mid-day versus mid-night (Baird and Gauvin, 2000). Furthermore, there is a leftward shift in the dose–response curve when access is restricted to mid-day, with peak rates of responding at lower doses. Rhesus monkeys also self-administered more low doses of cocaine in mid-day compared to the other times tested, namely, the transitions from light:dark and dark:light (Negus et al., 1995). The current results, indicating greater methamphetamine intake in the early versus late day, are consistent with these findings on cocaine. Furthermore, animals ingested low to moderate doses of MA in the present experiment (avg dose, Protocol 1: 0.8 mg/kg, Protocol 2: 0.55 mg/kg). Although it is difficult to extrapolate the reinforcing (or other) effects produced by this dosing range in mice to effects produced in humans, it is noteworthy that 0.2–0.8 mg/kg of methamphetamine has been shown to serve as a reinforcer in humans (Hart et al., 2001; Kirkpatrick et al., 2012). Taken together, these studies indicate that time of day modulates drug intake and may be particularly salient at lower doses of stimulants. While the present experimental design counterbalanced visual cues and drug-paired side across animals, it is possible that the mice had a visual pattern or side bias — factors that should be controlled in future studies.</p><!><p>There is a growing body of evidence supporting circadian and time of day effects following administration of stimulants in a variety of activity measures. Daily methamphetamine injections or chronic consumption in water induces marked changes in activity rhythms that occur independently of the central circadian pacemaker in the suprachiasmatic nucleus (Honma and Honma, 2009). However, experiments that utilized temporally restricted administration of lower doses of amphetamines indicate variations in drug effects over the course of the light/dark cycle, though with differences among reports in peak times of drug effects (Evans et al., 1973; Gaytan et al., 1998a, 1999; Kuribara and Tadokoro, 1982; Urba-Holmgren et al., 1977; Webb et al., 2009b), possibly due to differences in drug and dosing regimen, housing conditions, species and behaviors measured. However, overall, these previous publications show greater drug effects around dawn compared to dusk. The present results indicate that in the early day, animals showed MA + PB-induced activity, but in late day, there were no differences between MA + PB versus PB-Alone activity (Figs. 4B and 7B). This is further consistent with prior reports of amphetamine-induced CPP showing a peak in early day and no difference between saline and amphetamine in late day (Webb et al., 2009b).</p><!><p>In addition to voluntary intake and activity after ingestion, the present self-administration method allowed for measurement of anticipatory activity, an appetitive behavior. We found a robust effect of time of day on the development of anticipatory activity. Specifically, most animals in the early day MA + PB and PB-Alone groups exhibited anticipatory activity while none of the animals in the late day groups displayed anticipatory activity. This effect was significant even when we controlled for amount of intake of PB-Alone and MA + PB. This time of day effect was also seen when MA was administered through nebulization. The absence of anticipatory responses in the within subjects study was unexpected. It is not likely due to change in reinforcer intake, as anticipatory activity was not correlated with intake in Protocol 1 and there were no main effects of treatment (PB-Alone versus MA + PB) or self-administered dose of MA in this measure. The absence of anticipation may be due to the shorter training duration in the within subjects study. While food anticipatory activity can take 3–14 days to develop in food-restricted animals (Mistlberger, 1994), reinforcers such as drugs and palatable treats in freely-fed rodents produce relatively weak anticipatory response that develop slowly over several weeks or not at all (Abe and Rusak, 1992; Angeles-Castellanos et al., 2008; Hsu et al., 2010; Mendoza et al., 2005; Mistlberger and Rusak, 1987; Pecoraro et al., 2002; Verwey et al., 2007; Waddington-Lamont et al., 2007; Webb et al., 2009a).</p><p>Anticipatory activity in response to non-nutritive reinforcers is mediated by an SCN-independent oscillator that is theorized to be specific to rewarding stimuli (Webb et al., 2009a). We speculate that this anticipatory activity may be related to anthropomorphic concepts of "craving," specifically, motivational processes associated with positive affective states of craving (for review of craving, see Littleton, 2000; Siegel, 1999). In this view, the current data suggests that craving and/or motivation for reinforcers may be greater in the early day. However, interpretations of anticipatory activity as a measure of craving must be circumscribed, as appetitive measures in both laboratory animals and humans may have limited predictive for actual drug intake (Drummond et al., 2000; Littleton, 2000). In keeping with these findings, anticipatory activity was not correlated with drug intake in the current experiment. As anticipatory activity is a learned response, these time of day effects may be associated with diurnal variation in reward-related learning. This hypothesis is supported by data from preference measures. Early day groups exhibited significantly more extreme preferences than late day groups, possibly because they were better able to distinguish between the two dishes.</p><!><p>Another goal of the experiment was to compare time of day effects in different types of reinforcers. It has been suggested that all rewards engage the mesolimbic dopamine pathway (Wanat et al., 2009). In fact, chronic drug use is sometimes referred to as "hijacking" reward pathways meant to promote evolutionarily adaptive natural rewards. Furthermore, recent data suggests that chronic engagement of natural reinforcers such as highly caloric/palatable foods or sex can cause changes in brain circuitry in reward-related areas and corresponding addiction-like behaviors (Avena et al., 2008; Blum et al., 2000; Johnson and Kenny, 2010; Stice et al., 2010; Trinko et al., 2007). Our results contribute to this literature by indicating that food and drug display similar time of day effects in intake and associated behaviors. Furthermore, within individuals, the early day increase in intake occurred only in the preferred substance, supporting the idea that time of day differences are specifically related to reinforcer value.</p><p>A possible caveat in the experimental design used here is that the MA was provided in a PB mixture possibly masking the distinct effects of these reinforcers. The results indicate that animals responded differently to MA + PB compared to PB-Alone (Protocol 1). The amount of PB eaten increased over days of the study (Fig. 2A), while the amount of MA + PB ingested was stable for each concentration (Fig. 2B). That is, animals reduced the amount of PB eaten when MA was present, and they also regulated consumption according to MA concentration. The amount of MA + PB eaten may have been modulated by the anorexigenic effects of MA, by self-regulation of amount of MA intake, and/or by taste aversion. The present results are consistent with both anorexogenic and/or self-regulatory mechanisms. While MA is known to have a bitter taste, the pilot study negated this notion (Methods, Section 2.4.1), and this was confirmed in experimental data indicating that animals with ad-libitum access to food voluntarily ate MA + PB, even when PB-Alone was also available (Fig. 6C).</p><!><p>Interestingly, although these experiments were conducted in a well documented inbred mouse strain (Zurita et al., 2011; personal communication with Charles Parady, Senior Specialist in Technical Services, Charles River), marked individual differences in preferences were seen. There are a number of possible explanations for robust individual differences including epigenetics, polymorphisms/genetic drift, or gene–environment interactions. Individual differences have been previously reported in inbred mouse strains and this has sparked controversy over possible mechanisms (Blizard et al., 2004, 2005; Griffin et al., 2007; Wahlsten et al., 2006). The mechanism underlying individual differences in inbred mouse strains, particularly with respect to drugs of abuse is a fertile line of study. A caveat in the ingestion study is that non-pharmacological factors such as position bias may account for some of the observed results.</p><!><p>Taken together, the results indicate that there are robust time of day effects in the voluntary intake and behavioral response to food and drug. In Protocols 1 and 2, the greatest intake occurred in the early light phase, regardless of reinforcer type. In parallel, there was more activity in the early day in both protocols. In Protocol 1, both reinforcers elicited an anticipatory response only in the early light phase, and the effect on activity held up when we controlled for meal size. Furthermore, this time of day effect was replicated when MA was administered through nebulization. MA-induced activity was also greater in the early day in both Protocols 1 and 2. When we controlled for MA dose, activity remained greater in the early day indicating greater potency. Finally, pharmacokinetic data indicated that time-of-day effects in behavior cannot be attributed to diurnal variations in the metabolism of MA.</p><p>The effects of the two reinforcers differed in several aspects. There was more post-ingestion activity and less total PB intake following MA ingestion compared to PB-Alone. The amount of PB eaten remained stable at the two different MA concentrations, while the PB intake increased over time in the PB-Alone group. Further, when given a choice, most animals developed a consistent preference for either MA + PB or PB-Alone. Taken together, we conclude that MA + PB intake was regulated differently from PB intake.</p><p>Overall, the present design, allowing the study of temporally restricted, voluntary intake in animals during the inactive phase enables a reasonable parallel to this aspect of human drug use where access to drugs is restricted to the night, or the inactive phase for humans. This has long been considered due to matters of convenience and privacy; however, the present results indicate that drug effects vary by time of day, and, this may importantly modulate drug-taking behaviors and drug effects.</p>

PubMed Author Manuscript

The Pup-Proteasome System of Mycobacterium tuberculosis

Proteasomes are ATP-dependent protein degradation machines present in all archaea and eukaryotes, and found in several bacterial species of the order Actinomycetales. Mycobacterium tuberculosis (Mtb), an Actinomycete pathogenic to humans, requires proteasome function to cause disease. In this chapter, we describe what is currently understood about the biochemistry of the Mtb proteasome and its role in virulence. The characterization of the Mtb proteasome has led to the discovery that proteins can be targeted for degradation by a small protein modifier in bacteria as they are in eukaryotes. Furthermore, the understanding of proteasome function in Mtb has helped reveal new insight into how the host battles infections.

the_pup-proteasome_system_of_mycobacterium_tuberculosis

Introduction<!>Structure of the Mtb 20S Core Particle<!>Chamber of Doom: Core Protease Activity<!>Mpa: Gateway to Doom<!>A Pup-y Tale<!>Lack of a Pupylation Motif<!>Pupylation: An Enzymatic Process That Resembles Glutamine Synthesis<!>Degradation by the Mtb Proteasome: End of the Road\xe2\x80\xa6or Is It?<!>Depupylation: What Goes On, Must Come Off<!>Proteasomes and Pathogenesis<!>Characterization of Proteasome Pathway Mutants<!>Is Mtb Proteasome Protease Activity Necessary for All Phenotypes?<!>Proteasome Function and NO Resistance: An Unsolved Mystery<!>Regulation of Transcription: Meddling with Metals<!>Remaining Questions<!>

<p>Mycobacterium tuberculosis (Mtb) is the causative agent of tuberculosis (TB) and kills nearly two million people every year (http://www.who.int/). The infectious process starts with the inhalation of air-borne droplets containing Mtb bacilli. Bacteria replicate in professional phagocytes in the lungs where they must combat numerous anti-microbial molecules. If the host cannot control the infection, Mtb growth will result in the destruction of lung tissues and, ultimately, the death of the host.</p><p>Despite the astounding mortality caused by TB, most individuals infected with Mtb can control mycobacterial growth for much of their lives. Among the host's arsenal of antimicrobial effectors is nitric oxide (NO), which is produced by activated macrophages and is toxic to numerous microbes [1]. Evidence that supports the notion that NO is critical to controlling Mtb has come from mouse studies. Inactivation of the macrophage associated inducible NO synthase, (iNOS) also known as NOS2, dramatically sensitizes mice to Mtb infections [2]. The cytotoxic effects of NO are likely to be dependent on the formation of highly reactive nitrogen intermediates (RNIs). It is thought that in host cells NO is oxidized to nitrite, which can be protonated to nitrous acid in the phagosomes of activated macrophages. Nitrous acid dismutates to reform NO, which can penetrate bacterial membranes and cell walls to combine with reactive oxygen intermediates (ROIs) such as superoxide to generate peroxynitrite. RNIs and ROIs can induce lethal injuries including DNA and protein damage as well as lipid peroxidation [3, 4].</p><p>Regardless of the apparent protective effects of host-produced NO during Mtb infections, humans, as well as experimentally infected animals, are rarely sterilized of Mtb [1, 2]. This observation was the basis for the hypothesis that Mtb encodes proteins required for resistance to NO toxicity. Due to the emergence of multi-drug resistant (MDR) and extensively-drug resistant (XDR) Mtb strains, researchers around the world are looking for novel ways to target TB. Drugs that inhibit bacterial defenses against mammalian antimicrobial effectors like NO could help the host win the war against this disease.</p><p>It has long been a technical challenge to identify and characterize pathways important for the pathogenesis of Mtb, a highly infectious and slow growing (doubling time ~20 h) Actinobacterium with a high GC-content. Over the last 20 years, improved molecular genetic tools and bio-safety provisions have greatly facilitated studies into understanding the pathogenesis of this challenging organism. With the advent of efficient transposon mutagenesis, it became feasible to perform a screen to identify genes required for NO resistance in vitro. After screening over 10,000 transposon mutants for NO sensitivity, Nathan and colleagues identified five Mtb mutants of the virulent laboratory strain H37Rv with independent insertions in Rv2115c and Rv2097c, two genes that were predicted to be associated with proteasome function [5]. Rv2115c was named mpa (Mycobacterium proteasomal ATPase) due to its high similarity with eukaryotic and archaeal proteasomal ATPases [6]. Mpa is 81 % identical to ARC (AAA ATPase forming ring-shaped complexes) of Rhodococcus erythropolis, the first biochemically characterized bacterial proteasomal ATPase [7]. In contrast to Rv2115c/Mpa, Rv2097c did not exhibit similarity to any known proteins at the time, however, was proposed to participate in proteasomal function, and as such was named paf for proteasome accessory factor (later termed pafA). Importantly, mutations in mpa and pafA severely attenuate Mtb virulence in mice [5].</p><p>Proteasomes are multi-subunit barrel-shaped protease complexes that were first discovered in eukaryotes over 20 years ago [8]. In eukaryotes the 26S proteasome is composed of two functionally distinct sub-complexes: the 20S core particle (CP), required for degradation of the substrate, and a 19S regulatory particle (RP) located at either or both ends of the CP, responsible for substrate unfolding and translocation into the CP [9, 10]. The RP is composed of numerous proteins, the composition of which varies depending on its function. The 19S RP contains 19 subunits, including a ring of six distinct AAA (ATPases associated with different cellular activities) proteins that contact the CP, and non-ATPase subunits, which function in various aspects of substrate recognition and processing [11]. The CP is composed of four stacked rings with catalytic activity located within the central rings. The two inner rings are composed of seven distinct catalytic β-subunits sandwiched between two outer rings composed of seven distinct α-subunits [10]. The β-subunits have several proteolytic activities that allow the proteasome to cleave most types of peptide bonds. Protein fragments are estimated to range in size from 8 to 10 residues [12]. The α-subunit rings form a gated channel that controls the passage of substrates and cleaved peptides, and also serves as a docking surface for protein complexes such as the RP [10, 13].</p><p>Proteasomes are enzymatically and structurally distinct from the ATP-dependent, chambered bacterial proteases ClpP, Lon, FtsH and HslV [14, 15]. The first clue that bona fide proteasomes were present in prokaryotes came from electron microscopy studies on the thermoacidophilic archaeon, Thermoplasma acidophilum, in which CP-like particles were obtained from T. acidophilum lysates [16]. Ultimately, T. acidophilum CPs were purified and crystallized, and shown to be highly similar in structure to eukaryotic CPs [17]. The first bacterial proteasome to be characterized was the R. erythropolis proteasome [18]. Later, 20S proteasomes were characterized from Mycobacterium smegmatis [19], Streptomyces coelicolor [20] and Frankia [21]. Additionally, genomic sequencing revealed the presence of proteasomal genes in the pathogens Mtb [22] and Mycobacterium leprae [23]. Bacterial proteasomes were thought to be confined to Actinobacteria until studies from the Banfield group (reviewed in [24]), discovered two actinomycete-like proteasome genes clusters in a non-culturable Gram-negative bacterium called Leptospirillum. To date, Actinomycetes and Leptospirillum are the only known bacterial lineages with a proteasome system, and may have acquired this protease complex via lateral gene transfer events [24, 25]. In contrast to the eukaryotic CPs, most prokaryotic CPs are composed of homo-heptameric rings; two β-subunits (PrcB) rings, flanked by two α-subunit (PrcA) rings (reviewed in [26, 27]). For the most part, the presence of only one type of β-subunit limits the proteolytic activity of the prokaryotic 20S CP to chymotryptic activity.</p><p>Since the initial identification by Darwin et al. of genes required for NO resistance in Mtb [5], it was later shown (using two separate prcBA mutant Mtb strains) that the CP was also needed for resistance to NO [28, 29]. This provided evidence of a functional link between Mpa, PafA and the CP. However, when compared to the wild type, mpa or pafA Mtb strains, the prcBA mutants grow much more slowly in rich broth (~20–30 % lower optical density at stationary phase) and take longer to form colonies on solid media [28, 29]. This growth defect in these genetically manipulated strains is similar to that observed for wild type Mtb strains treated with a mammalian proteasome inhibitor, N-(4-morpholine)carbonyl-b-(1-naphthyl)-L-alanine-L-leucine boronic acid (MLN-273) or epoxomicin [5]. Collectively, these data suggest that the Mtb CP can also degrade proteins in an Mpa/PafA independent manner; or that the CP has other, possibly protease-independent functions important for growth (to be discussed later). The notion that the Mtb CP is needed for normal growth is supported by a study that attempted to delineate genes essential for Mtb growth in vitro. In this study, Rubin and colleagues found that prcBA, but not mpa or pafA, were required for normal growth in vitro [30], therefore it is, not surprising that prcBA-defective Mtb strains are also highly attenuated in a mouse model of infection [28, 29].</p><p>The identification of a bacterial proteasome associated with a virulence phenotype piqued the interest of numerous laboratories to better characterize this protease. Here, we summarize what is currently understood about the structure and function of the Mtb proteasome and discuss its potential roles in pathogenesis.</p><!><p>The overall architecture of the Mtb 20S CP is similar to the CP of archaeal [31] and eukaryotic [32] proteasomes. All CPs form a barrel-shaped structure consisting of four stacked, seven-subunit rings that are arranged into two central β-rings, flanked by an α-ring at either end. Like other prokaryotic proteasomes, Mtb CPs are arranged into a four-ringed α7β7β7α7 cylinder composed of 14 identical α-subunits and 14 identical β-subunits, ~150 Å in height with a diameter of ~115 Å (Fig. 10.1a, b). The Mtb proteasome shares modest sequence identity with the archaeal CP from Thermoplasma (~32 % identity for both α- and β-subunits) and high identity with the bacterial CP from Rhodococcus (~65 % identity). Despite this, with the exception of helix 2 in both the α- and β-subunits, the three-dimensional structures of all three prokaryotic CPs are virtually super-imposable [33, 34]. Although, only the relative position of helix 2 is altered in the bacterial CPs, this small upward tilt of ~10° creates a wider axial substrate channel in the bacterial CPs when compared to the archaeal CPs [33].</p><p>The proteolytic active sites of the CP are housed in the β-subunits (in eukaryotes only three of the seven β-subunits contain functional sites: β-1, -2, and -5 [35]). These active subunits are synthesized with N-terminal pro-peptides, the autocleavage of which exposes the catalytic nucleophile, an N-terminal threonine (Thr-1) [36]. Similarly, in Rhodococcus, there are two β-subunits (β1 and β2) both of which are translated with long propeptides (65 and 59 residues respectively). In this case, recombinant Rhodococcus α- and β-subunits only assemble into an active proteasomes when all subunits are combined, while separately expressed components remain monomeric [34, 37–39]. Collectively, these findings suggest that the β-subunit propeptide not only facilitates the formation of the first assembly intermediate (the α/β heterodimer) but also shields the catalytic Thr residue during proteasomal assembly, preventing undesired protein degradation [37, 40]. In Thermoplasma, the eight amino acid propeptide of the β-subunit seems to be dispensable for assembly of the CP, as the α-subunits can assemble spontaneously into seven subunit rings when produced in Escherichia coli [41]. In contrast the α-ring appears to serve as a template for assembly of the β-ring in the formation of active CP in Rhodococcus. In Mtb the 56-residue propeptide of PrcB appears to inhibit rather than promote CP assembly [42], as a cryo-EM study of the Mtb half proteasome revealed that the propeptide is located outside of the β-ring rather than between the α- and β-rings [43]. Because assembly of the mature CP is a result of the apposition of two half-proteasomes, it seems that the Mtb β-propeptides, which are auto-catalytically removed, could be a barrier to the assembly process. However, it may be that simply more time is required to overcome this barrier.</p><!><p>Although several prokaryotic CPs characterized so far have significant peptidase activity [17, 40, 44, 45], the in vitro peptidase activity of Mtb CP is relatively low [42]. This low activity suggests that the substrate gate is closed in a manner similar to the eukaryotic CP. In the eukaryotic CP, the substrate entrance to the α-ring is closed to prevent uncontrolled proteolytic activity [46]. In the eukaryotic CP, the gate to the catalytic chamber is blocked by the N-terminal sequences of the seven different α-subunits, which adopt different conformations and seal the entry portal [46]. In Mtb, a high-resolution crystal structure of the CP containing a mutant β-subunit (PrcB T1A, which prevents propeptide cleavage) revealed that the seven identical N-terminal peptides that form the gate were ordered, but exhibit three different conformations to tightly seal its substrate entrance at the seven-fold symmetry axis [43] (Fig. 10.2a, b). Deletion of the N-terminal residues 2–9 of PrcA results in an "open gate" conformation that increases peptidolytic activity to small peptides in vitro [42] (Fig. 10.2c).</p><p>All CPs are N-terminal Thr hydrolases [14, 35, 36]. While most prokaryotic CPs seem to exclusively hydrolyze hydrophobic small synthetic peptide substrates (so-called "chymotryptic" activity), the Mtb CP has broad substrate specificity, targeting not only hydrophobic targets but also basic ("tryptic" activity) and acidic peptides ("peptidyl-glutamyl-peptide-hydrolyzing", "caspase-like" or "post-acidic" activity) [42]. Unlike eukaryotic CPs in which substrate preference is determined by several different β-subunits [35, 47, 48], the Mtb 20S CP contains a single type of β-subunit, raising the question as to how it displays such broad substrate specificity. Structural analysis revealed that the substrate-binding pocket in the Mtb proteasome combines features found in different eukaryotic β-subunits: a hydrophobic upper surface similar to that of formed by a eukaryotic β5 subunit, and a hydrophilic lower surface similar to eukaryotic β1 and β2 subunits [33]. This composite feature of the substrate-binding pocket in the Mtb CP likely accounts for its broad substrate specificity [42].</p><!><p>Like other chambered proteases, proteasomal proteolysis requires an ATP-dependent chaperone to unfold structured proteins for delivery into the proteasome core where they are degraded. In archaea the best-characterized proteasomal ATPase is Methanococcus jannaschii PAN (proteasome activating nucleotidase) [49]. PAN can facilitate the degradation of artificial substrates by CPs [6, 49] but it remains to be determined how PAN recognizes archaeal proteins. Mpa forms homo-hexamers and is homologous to PAN. Like other AAA ATPases, Mpa has characteristic Walker A and B motifs for ATP binding and hydrolysis, respectively [50]. Mpa has relatively low ATPase activity: ATP hydrolysis is about four times slower than that of ARC, the Mpa orthologue in Rhodococcus (Vmax of 62 versus 268 pmol min−1 μg−1) [7, 51], or PAN [49]. As predicted (and will be discussed in detail below), a major function of Mpa is to deliver proteins into the CP for destruction.</p><p>The full-length structure of Mpa is currently unknown, however, crystal structure analysis of partial Mpa polypeptides has yielded highly informative insight into its activity. Mpa and ARC each contain two domains of the oligosaccharide/oligonucleotide-binding (OB) fold in tandem, with the second OB fold appearing to play a major role in Mpa oligomerization [52, 53] (Fig. 10.3a). In contrast, the archaeal PAN has only one OB domain [55], and it is not clear why Mpa and ARC have two. Immediately preceding the OB folds in Mpa is a 75 Å long α-helix [54] (Fig. 10.3b). Remarkably, helices from neighbouring subunit pairs form a coiled coil, thereby reducing the six-fold symmetry at the intermediate OB domain region to three-fold. This structural feature may be important for the specific protein recognition and unfolding activity of this class of ATPases.</p><p>By analogy to the eukaryotic 26S proteasome, the prokaryotic proteasomal ATPases are also expected to physically interact with the CP in order to couple protein unfolding with delivery into the CP. A major distinction between the prokaryotic ATPases and eukaryotic 19S RP is that the 19S binds the CP with an affinity strong enough for the entire 26S complex (19S RP + 20S CP) to be co-purified. In stark contrast, prokaryotic ATPases, despite strong phenotypic associations with CP activity in vivo, only bind CPs either weakly or transiently in vitro [52, 56]. Methanococcus PAN weakly interacts with Thermoplasma CPs [56], and Mpa can directly associate, although flexibly and weakly, with open-gate mutant Mtb CPs [52] and can degrade proteins [57]. Because Mpa only weakly interacts with an altered proteasome in vitro, the precise nature of how any bacterial proteasome interacts with its cognate ATPase in vivo is a mystery. It is notable that proteasomal ATPases from bacteria to mammals contain a "HbYX (hydrophobic amino acid-tyrosine-X) motif" at their C-termini, and this motif is crucial for degradation but not ATPase activity [51, 56, 58, 59]. This motif is needed for PAN-mediated activation of proteolysis [60] and is implicated in proteasome assembly in yeast [59]. Mpa also has a HbYX-like motif, which is critical for the degradation of proteins in Mtb [51, 58]. Like PAN, Mpa lacking this motif is impaired in its interaction with the Mtb 20S CP in vitro [57].</p><!><p>In eukaryotes, proteins that are destined for proteasomal degradation are usually post-translationally modified with the small protein ubiquitin (Ub) (reviewed in [61]). Ubiquitin is synthesized as part of a precursor protein, processed to form a 76 amino acid protein containing a C-terminal diglycine motif (Gly-Gly), with a compact β-grasp fold [62–64]. The C-terminal Gly is subject to a series of reactions that result in the conjugation of Ub to a Lys residue on the target protein [65–67]. In the first step, the C-terminal Gly of Ub is adenylated by an Ub activating enzyme (E1) using ATP. Ub is then transferred to the active site Cys residue on the E1 enzyme. Next, Ub is transferred to an Ub conjugating enzyme (E2) and delivered to an Ub ligase (E3), which catalyzes the formation of an isopeptide bond with a Lys residue on the target protein (Fig. 10.4). In eukaryotes there are numerous E2 activating enzymes and E3 ligases (which contain a variety of substrate binding activities) to provide substrate specificity to the Ub proteasome system (UPS) (reviewed in [68, 69]). In general, proteins that are targeted to the proteasome have several Ub molecules added to the substrate, usually resulting in the formation of polyubiquitin (polyUb) chains [66, 70, 71]. The polyUb chains are recognized by the RP of the proteasome, and removed by proteasome-associated deubiquitinases (DUBs) for recycling of Ub. The deubiquitylated substrate is then delivered to the CP for destruction (reviewed in [11]).</p><p>Despite the presence of archaeal and bacterial CPs that are almost biochemically indistinguishable from the eukaryotic CP, an Ub-like system for the degradation of protein substrates was not found in prokaryotes for many years. A major hurdle at this time was an inability to reconstitute the proteolytic activity of bacterial proteasomes. This strongly suggested that other co-factors were required for proteolysis or that specific model substrates were required. It was speculated that bacterial proteasomal substrates only required intrinsic signals for degradation, as there was no evidence for the existence of any post-translational small protein modifiers in bacteria. To examine this, Darwin and colleagues set out to identify natural Mtb proteasome substrates by comparing the steady-state proteomes of wild type and mpa Mtb strains using two-dimensional polyacrylamide electrophoresis [58]. Although the proteomic profiles of untreated and NO-treated stationary phase cultures showed limited differences between the two Mtb strains, two proteins, FabD (malonyl CoA-acyl carrier protein transacylase) and PanB (ketopantoate hydroxymethyltransferase), accumulate significantly in the mpa strain. Importantly, the transcript levels fabD and panB are nearly identical in both wild type and mpa Mtb strains, which supported the notion that Mpa is required for FabD and PanB turnover, and not fabD or panB expression.</p><p>To further test if FabD and PanB were potential proteasome substrates, Mtb fabD and panB were expressed from a heterologous M. bovis hsp60 promoter in Mtb [72]. In addition to ruling out potential differences in gene expression, this system would also address the possibility that fabD and panB mRNA had differences in translation initiation in the mpa or pafA strains. Each recombinant gene also encoded a FLAG and His6 epitope tag at the N- and C-termini, respectively. Similar to that observed in 2D-PAGE analysis, both FLAG-FabD-His6 and FLAG-PanB-His6 accumulate in the mpa and pafA mutants [58]. Consistently, treatment of wild type Mtb with a eukaryotic proteasome inhibitor stabilized FabD and PanB [58] and deletion of the 20S CP permitted PanB accumulation [29]. Collectively, these data supported the idea that Mpa and PafA were required for the degradation of FabD and PanB by the CP.</p><p>In addition to the identification of FabD and PanB as proteasomal substrates, an unexpected observation was made during these studies: Mpa itself was also identified as a putative proteasomal substrate [58]. Chemical inhibition of the CP results in the accumulation of Mpa in Mtb. This finding was consistent with an earlier observation that mutations in mpa, which disrupt the ATPase activity or the HbYX motif, increase the steady-state levels of Mpa [51]. Similarly, Mpa accumulates in a pafA mutant strain, supporting a role for PafA in substrate degradation [58]. Thus it appears that the proteasome may "cannibalize" its ATPase to regulate its levels and hence activity.</p><p>Despite the identification of these putative substrates, their degradation, using purified Mpa and CP, could not be reconstituted in vitro. Therefore it was proposed that other factors were needed to facilitate proteasomal degradation. To identify these factors a bacterial two-hybrid screen [73] was used to search for proteins that bind to Mpa, with the reasoning that Mpa interacts not only with substrates but also with other proteins that promote proteolysis. From a library of ~100,000 Mtb genomic DNA fragments, Darwin and colleagues identified a protein encoded upstream of the CP genes prcBA, termed Rv2111c [74]. Importantly, recombinant Mpa and Rv2111c interacted in vitro, however, the addition of purified Rv2111c to CP and Mpa was unable to stimulate degradation of FabD.</p><p>The inability to reconstitute proteasomal degradation in vitro suggested that additional co-factors were still needed for proteolysis. Because E. coli does not encode a proteasome system it was reasoned that these co-factors were likely missing and possibly Mycobacterium-specific. A mycobacterial two-hybrid system [75] was thus used to interrogate interactions between proteasome subunits, substrates and the newly identified Rv2111c. In this system, M. smegmatis (Msm), a non-pathogenic relative of Mtb, was used as the host organism to identify protein-protein interactions. An unexpected interaction was detected between the substrate FabD and Rv2111c. To validate the genetic result, recombinant fabD and Rv2111c were co-expressed in Msm and found to co-purify as a heat stable complex. Mass spectrometry revealed that the C-terminal residue on Rv2111c formed an isopeptide bond with the side chain of Lys173 in FabD [74]. However, in contrast to Ub and related modifiers, Rv2111c did not have a C-terminal Gly-Gly motif, but instead has a Gly-Gly-Gln motif. Moreover, the C-terminal Gln was deamidated to Glu, before conjugation to FabD [74]. In a subsequent study, an orthologue of Rv2111c in Msm (MSMEG_3896) was identified as the protein modifier in that species. Thus, these two studies showed that this post-translational protein modification is conserved in both pathogenic and non-pathogenic mycobacteria [74, 76].</p><p>To determine if Rv2111c targeted proteins for degradation, Darwin and co-workers mutated the modified lysine residue (Lys173) in FabD [74]. Consistent with the idea that attachment with Rv2111c was the signal for degradation by the Mtb proteasome, mutagenesis of Lys173 to alanine in Mtb FabD stabilized the protein substrate. Based on its functional similarity to Ub, Rv2111c was named "Pup" for prokaryotic ubiquitin-like protein. Polyclonal antibodies to Pup recognize numerous proteins in Mtb H37Rv demonstrating that "pupylation" is widespread. Immunoblot analysis of the pafA mutant shows no anti-Pup reactive proteins, suggesting that PafA was the only Pup ligase in Mtb [74]. This result was somewhat unexpected as there are several hundred different Ub ligases in eukaryotes. In a subsequent study, Weber-Ban and colleagues demonstrated that PafA and ATP were sufficient to conjugate deamidated Pup to proteasomal substrates in vitro [77]. Interestingly, in contrast to Ub and other Ub-related modifiers, which all form a compact β-grasp fold, Pup is an intrinsically disordered protein with a propensity for helicity [78–80].</p><!><p>Although sequence recognition motifs for Ub ligases and other accessory factors have been identified, there is currently no known sequence motif surrounding the Lys on which Ub is attached (reviewed in [81]). In contrast, SUMO (small ubiquitin-like modifier) often attaches to a Lys residue that is part of a tetrapeptide motif, ΨKxD/E, where Ψ is a large hydrophobic residue and x is any amino acid (reviewed in [67]). The identification of a sequence that could predict pupylation could be useful for understanding how Pup regulates proteins. Therefore, to identify a possible pupylation motif, the "pupylome" was determined by several independent groups by purifying an epitope tagged Pup from Mtb [82] or Msm [83, 84]. In the Mtb study 604 proteins, representing ~15 % of the total predicted proteome, were identified, but only 55 proteins, including Mpa, were confirmed to harbor a site of Pup attachment [82]. In Msm, two independent studies identified 103 and 243 proteins, with 52 and 41 proteins, respectively, having confirmed Pup attachment sites [83, 84]. In all cases, Pup was attached to Lys, and there was little to no evidence of Pup chains, although Pup contains three Lys residues. In one study, Song and colleagues observed pupylation of Lys31 and Lys61 on Pup [84]; however, the authors of this study speculated this might not be physiologically relevant as Pup was overproduced. The authors also noted that pupylation was dynamic and changed depending on the growth condition examined. In all of the "pupylome" studies most, if not all, of the proteins identified are involved in housekeeping functions or stress responses.</p><p>Despite the successful identification of numerous pupylation targets, a motif is yet to be identified. It has however, been speculated that an intrinsic sequence is required to signal pupylation because PafA has a much higher affinity for at least one proteasome substrate, PanB, than for free Lys [85]. Nevertheless, it is hard to imagine how PafA specifically recognizes its targets because so many different proteins can be pupylated. In E. coli, a bacterial species that does not encode a Pup-proteasome system, pupylation can be reconstituted by expressing Pup with a C-terminal Glu (PupGlu) and PafA. Over 50 E. coli proteins can be pupylated using only PafA and a Pup mutant containing a C-terminal Glu (PupGlu), suggesting the notion that an intrinsic Mycobacterium specific sequence is required for pupylation is unlikely [86]. Consistently, PtsI (an E. coli pupylated substrate) was pupylated by a native mycobacterial system when produced in Msm [86]. Thus, signals for PafA target recognition are not expected to be Mycobacterium specific.</p><p>Based on these E. coli studies, it appears that pupylation is partly stochastic. However, it seems unlikely that mere over-expression of pup and pafA could determine the fate of so many proteins. Firstly, not all Lys containing proteins are pupylated, e.g. pupylation of the Mtb protein DlaT (dihydrolipoamide acyltransferase), which contains 27 Lys residues, has not been observed in either Mtb [74] or E. coli [86]. Secondly, not all Lys residues within the target protein are modified. Strikingly, Mtb FabD is preferentially modified on a single Lys residue, Lys173 [74, 82], despite the fact that Mtb FabD contains eight surface exposed Lys residues [87]. Although, two additional Lys residues can, to a lesser extent, be pupylated in Mtb FabD when produced in E. coli [86]. A simple explanation may be that over-production of Mtb FabD in the E. coli system merely gives PafA access to other Lys residues. Alternatively, these data may suggest that other factors regulate how and when FabD is pupylated in mycobacteria. Indeed, Darwin and colleagues speculated that the Lys residues in FabD may be involved in interactions with other enzymes in the fatty acid synthesis II (FASII) pathway [86] protecting them from modification. Interestingly, most enzymes in the FASII pathway, several of which are encoded in an operon with fabD, are pupylation targets [82, 83]. However, under normal culture conditions, only some of the proteins in this pathway appear to be proteasome substrates in Mtb. Recombinant FabG (3-ketoacyl-Acp-reductase), KasA and KasB (3-oxoacyl-Acp-synthases 1 and 2, respectively) do not accumulate in mpa or pafA mutants under routine culture conditions [82]. It is possible that FabG, KasA, and KasB are degraded by the proteasome under different conditions or are degraded more slowly than FabD. Taken together, there may still be additional Mycobacterium specificity factors that regulate pupylation or the delivery of certain pupylated proteins to the proteasome.</p><!><p>At the time of its identification, PafA did not resemble any protein of known function [5]. Shortly after the discovery that PafA was involved in pupylation, Aravind and colleagues performed a detailed bioinformatic analysis that predicted PafA to have structural similarity to glutamine synthetase (GS) and glutamine cysteine synthetase (GCS) [88]. GS catalyzes the formation of Gln from Glu and ammonia, while GCS catalyzes the formation of γ-glutamyl-cysteine from Glu and Cys; both processes occur via a phosphorylated Glu intermediate. It was therefore proposed that the side chain carboxylate group of the C-terminal Glu in Pup would be phosphorylated or "activated" by PafA. This phosphorylated intermediate would then be primed for nucleophilic attack by the ε-amino group of a side chain Lys on a substrate, resulting in an isopeptide bond between C-terminal Glu of Pup and an internal Lys on the substrate. Indeed, as predicted by Iyer et al. [88], Weber-Ban and colleagues could show that Pup ~ substrate conjugates are generated via activation of the carboxylate group on the C-terminal Glu of Pup [77]. Consistently, site directed mutagenesis of residues in PafA predicted to coordinate ATP or Mg2+ disrupted PafA function both in vivo [89] and in vitro [77].</p><p>But how is Pup deamidated prior to activation by PafA? The first evidence came from the Weber-Ban group, who identified a homologue of PafA in H37Rv, (Rv2112c) near the proteasome core genes prcBA [77]. They demonstrated, using purified recombinant proteins, that Rv2112c was responsible for Pup deamidation, rendering Pup competent for ligation to FabD or PanB by PafA. Rv2112c was therefore named Dop for deamidase of Pup [77] (Fig. 10.5). In contrast to PafA, which needs ATP to phosphorylate Pup, Dop can deamidate Pup in the presence of ATP, ADP or non-hydrolyzable ATP analogues, suggesting ATP/ADP are allosteric activators of deamidation. Pupylation could also be achieved, in the absence of Dop, when Pup was replaced with a mutant form of Pup (PupGlu, that does not require deamidation for activation), obviating the need for Dop in vitro [77]. This demonstrated that Dop and PafA catalyze independent reactions: deamidation of Pup and conjugation of Pup, respectively. Curiously some Dop-containing bacteria encode PupGlu, presumably eliminating the need for Dop, which suggests that Dop may play a different role in these organisms. Nevertheless, for mycobacteria, deamidation is required for pupylation to take place as disruption of dop impairs pupylation and proteasomal substrate degradation [89, 90].</p><!><p>How are pupylated proteins recognized by Mpa prior to degradation? In eukaryotes, Ub receptors are present in the RP of the 26S proteasome, ready to receive ubiquitylated proteins (reviewed in [11]). To date, an equivalent Pup receptor has not been found, however, Pup has a strong affinity for Mpa both in vitro and in vivo [74] and pull-down experiments using Pup-decorated beads showed that the N-terminal coiled-coil domain of Mpa interacts with Pup [79]. Unlike Ub, Pup is a mostly unstructured, intrinsically disordered protein [78–80], which raised the question: how does Mpa specifically recognize Pup?</p><p>A series of in vitro and in vivo experiments determined that Pup is a two-part degron where the N-terminal ~30 residues are required for Mpa to start the unfolding process and the C-terminal ~30 residues, which have the propensity for helicity as determined by NMR, are needed to interact with Mpa [57, 91]. Darwin and colleagues showed that the N-terminal half of Pup is essential for degradation, but dispensable for pupylation in vivo [91], suggesting the C-terminal half of Pup is necessary and sufficient for interaction with PafA and Dop. Weber-Ban and colleagues showed that Mpa could unfold green fluorescent protein if fused to Pup, and this required the N-terminal half of Pup [57]. Importantly, this study also showed that a pupylated substrate can be degraded, albeit somewhat slowly, by Mpa and the proteasome. Interestingly, Pup itself is degraded with the substrate in this system, showing removal of Pup is not essential for degradation.</p><p>The molecular details of the interactions required for substrate degradation were ultimately revealed when the three-dimensional structure of the Pup-Mpa complex was solved [54]. Analysis of this complex revealed that the central part of Pup (residues 21–51) becomes ordered upon binding to Mpa [54]. Indeed, the central part of Pup forms an α-helix, using the coiled-coil region of Mpa as a template (Fig. 10.6a). This interaction positions the disordered N-terminus of Pup towards the central channel of the hexameric ATPase, apparently priming the initial threading of Pup into the narrow unfolding pore [54] (Fig. 10.6b). This is consistent with previous studies that suggested the N-terminal half of Pup is needed to facilitate substrate unfolding and degradation [57, 91]. Both hydrophobic and electrostatic interactions are formed between Pup and Mpa, and disruption of either type of interaction abolishes Pup-mediated degradation by the proteasome in Msm [54]. Furthermore, consistent with other AAA proteases, mutagenesis of the conserved hydrophobic "pore loop" (Val342Ala) in Mpa abolishes degradation [52].</p><p>Importantly, although there are three coiled-coils in the Mpa hexamer (each potentially capable of interacting with Pup or a pupylated substrate) it appears that only one Pup associates per Mpa hexamer [54, 78, 79]. This arrangement would prevent multiple substrates from being recruited to the same Mpa hexamer at any one time, and hence would eliminate potential substrate aggregation or jamming at the proteasome.</p><!><p>In the eukaryotic UPS, DUBs play an important role in protein degradation. Some DUBs, by removing Ub, are responsible for reversing the fate of a protein destined for degradation, while other DUBs located at the 19S RP, remove Ub to facilitate degradation (reviewed in [92]). DUBs are also responsible for the recycling of Ub to permit new ubiquitylation reactions. It was therefore predicted that pupylation would also be reversible. In a series of elegant experiments the Pup deamidase (Dop) was ultimately identified as a "depupylase" (DPUP) in both Mtb [91] and Msm [93]. DPUP activity was first demonstrated by Darwin and colleagues, using purified pupylated substrates (Pup ~ FabD and Pup ~ Ino1) with lysates from wild type Mtb. Although wild type Mtb lysates demonstrated DPUP activity, the dop mutant strain did not [91]. These data suggested that Dop was responsible not only for deamidation of Pup, but also for depupylation (Fig. 10.5). Indeed, Dop-mediated DPUP activity was then confirmed in vitro using a variety of pupylated substrates [91, 93]. It was then shown by Weber-Ban and colleagues that Dop cleaves specifically the isopeptide bond between Pup and the proteasomal substrate Lys and that PupGlu is released from the depupylation reaction, suggesting Pup could be recycled [93].</p><p>Dop is a strict isopeptidase because it cannot remove Pup from a longer polypeptide or linear fusion protein [91, 93]. This is in contrast to some Ub processing enzymes (DUBs) that remove Ub from larger polypeptides [64]. In contrast to DUBs, which tend to be either cysteine or zinc metalloproteases (reviewed in [92]), the catalytic motif of Dop is currently unknown; it does not have a nucleophilic Cys in its partially modeled active site [89], and is not known to require zinc for function. Like PafA, Dop is predicted to adopt a GS/GCS fold [88]. Dop binds Pup well in vitro unless Pup's penultimate Gly-Gly motif is mutated, or additional amino acids are added to the C-terminus. Interestingly, however the Gly-Gly motif is not required for substrate attachment [89]. Taken together, it seems that the Gly-Gly motif is important for access to the active site of Dop but not for conjugation to substrates.</p><p>It is challenging to assess the role of Dop's DPUP activity in vivo because Dop is required for Pup deamidation prior to pupylation in Msm and Mtb [89, 93]. Weber-Ban and colleagues fully restored pupylation in an Msm dop mutant by ectopic expression of pupGlu [93]. In striking contrast, however, expression of pupGlu does not restore the pupylome in Mtb [89]. However, the Mtb pupylome can be restored in the dop strain expressing pupGlu if treated with a proteasome inhibitor. This result suggests that Pup, along with its conjugated proteins, is directly degraded by the proteasome in Mtb lacking Dop. Importantly, this observation strongly suggests depupylation is needed to maintain Pup levels for normal pupylation. Thus, it seems likely that a critical function of Dop in Mtb is to act as a DPUP, an activity that is essential for Pup recycling. Furthermore, DPUP activity could also be used to regulate protein stability by altering the fate of a once doomed pupylated substrate. As with PafA and pupylation, it is not known how Dop selects substrates for depupylation. Finally, in bacteria that encode PupGlu, the primary function of Dop must be as a DPUP. It remains to be determined if certain bacteria have evolved to use Pup deamidation as a regulatory step in pupylation. Additionally, we do not yet understand why dop mutants in Msm and Mtb have such different phenotypes.</p><p>A curious observation was made in a study that examined the stability of Ino1 in Msm [94]. Over-expression of pup in wild type Msm results in virtually undetectable levels of endogenous Ino1, presumably due to its accelerated turnover. Interestingly, over-expression of pup in a prcBA mutant results in the accumulation of unpupylated Ino1. This finding suggests that in the prcBA mutant either pupylation of Ino1 is inhibited or depupylation prevents detection of Pup ~ Ino1. In contrast, a follow up study showed that Pup ~ Ino1 accumulates dramatically in an Msm mpa mutant overproducing Pup [91], which led to the hypothesis that Mpa helps to unfold a pupylated substrate in order for it to be depupylated. Consistent with this idea, Weber-Ban and co-workers showed that Mpa increases depupylation of a substrate in vitro [93]. Interestingly, it has been noted that corynebacteria, a distant relative of mycobacteria, encode pup, mpa, pafA, and dop orthologues but do not have proteasomes [90]. It is tempting to speculate that pupylated proteins are degraded by a different protease, or that pupylation-depupylation is involved in regulating protein activity in this bacterial genus.</p><p>The notion that protein unfolding by Mpa is coupled to depupylation poses some challenges to the current model of proteasome-mediated degradation in mycobacteria. It is well established that unfolding of proteasome substrates starts with the engagement of Pup with Mpa [54, 57]. Presumably, Pup is threaded through the channel in Mpa and, as previously shown, can itself be destroyed by the proteasome along with the substrate in vitro [57, 89]. However it also appears that substrates can be depupylated prior to degradation, presumably as they exit from the proximal end of the Mpa hexamer. This scenario implies that Dop interacts with Mpa or substrates at the interface between Mpa and the CP where the unfolded protein is being funnelled into the degradation chamber. One wonders if the conserved, but poorly understood, symmetry mismatch between the six-fold ATPase and the seven-fold CP evolved to prevent tight binding, and allow a gap for the removal of Pup by Dop. Clearly, much needs to be done in order to understand how, when and where Dop coordinates depupylation with degradation.</p><p>In eukaryotes, DUBs, ATPases, ligases and other proteins are associated with the eukaryotic 19S RP to remodel or recycle Ub chains on substrates. The Mycobacterium proteasome system appears to have been streamlined in such a way that Mpa plays multiple roles in the Pup-proteasome pathway by acting as a substrate receptor, unfoldase and a facilitator of depupylation. It remains to be determined if the Mtb proteasome requires additional co-factors to catalyze proteolysis. Because the in vitro degradation rate of a pupylated protein seems unusually slow [57, 91], it seems likely that other undetermined factors may be needed to facilitate degradation in Mtb.</p><!><p>As discussed earlier in this chapter, mpa and pafA were identified as genes required for NO resistance and virulence in an animal model of infection [5]. Later studies (to be discussed below) identified additional components of the Pup-proteasome system that are also needed for Mtb pathogenesis. How does the Mtb proteasome protect against NO toxicity and promote TB pathogenesis? Perhaps the proteasome provides bacteria with a critical pool of amino acids through protein degradation during the chronic phase of infection. Alternatively, the bacterial proteasome may modify the host machinery to its advantage by degrading proteins that alter the recognition of the pathogen by the host's immune system. In the next section we will attempt to address these complex questions by discussing the in vitro and in vivo phenotypes of proteasome associated mutants in more detail.</p><!><p>The mouse model of TB is characterized by two phases, an acute phase, during which Mtb replicates exponentially within the lungs for about 3 weeks and a chronic or latent phase that is brought about by the emergence of acquired cell-mediated immunity. During the chronic phase bacterial numbers are stabilized in the lungs. Eventually, all mice that are experimentally inoculated with wild type Mtb die of TB, in contrast to humans that are naturally infected with Mtb. In a low dose aerosol model of Mtb infection, mice can survive for more than a year before succumbing to TB; in contrast, mice infected with either an mpa or pafA mutant show no symptoms of TB [51, 58]. Similar to the mpa and pafA mutants, an Mtb dop mutant is sensitive to RNI in vitro and severely attenuated in mice [89]. The degree of attenuation in mice (bacterial load and histopathology) is similar among the dop, mpa and pafA mutants, consistent with the notion that pupylation and Mpa-dependent proteolysis are functionally linked [5, 89]. The attenuation of symptoms is likely due to the presence of 100–1,000 times fewer recoverable mutant bacilli in the lungs, spleens and livers during the persistent phase of infection [5]. However, it is also possible that the Mtb proteasome regulates one or more factors that affect the host's response to infection.</p><p>Targeted gene disruptions in the Mtb CP genes dramatically slow Mtb growth on solid media [28, 29]. In C57BL/6 mice infected with either ΔprcBA::hyg or PtetO-prcBA Mtb strains, the number of bacilli recovered from the lungs is approximately 100-fold lower (compared to wild type or non-silenced Mtb) after 3 weeks of infection and continues to decline after this time. This is not completely surprising based on the severe in vitro growth defects associated with Mtb CP mutations. In contrast to the prcBA strains, mpa, pafA and dop Mtb mutants grow more similarly to wild type Mtb in rich broth [5, 89]. These observations suggest that the CP may have critical functions independent of pupylation-dependent proteolysis in Mtb. Interestingly, the mpa, pafA and prcBA-defective Mtb strains are more resistant to hydrogen peroxide than wild type bacteria, suggesting there is an increase in activity or expression of one or more anti-oxidant pathways in the absence of proteasome function [5, 28]. However, it is currently not understood how loss of proteasome activity could lead to increased resistance to ROIs.</p><p>Because proteasome pathway mutants are sensitive to NO in vitro, Nathan and colleagues questioned if mice defective in NO production would be more susceptible to infection with mpa or pafA mutants. In mice and humans NO is produced by three different isoforms of nitric oxide synthase (NOS): endothelial NOS (eNOS), neuronal NOS (nNOS) and inducible or immune NOS (iNOS). iNOS is expressed in activated macrophages and is critical for the control of numerous microbial infections (reviewed in [1]). In comparison to wild type mice, mice genetically inactivated for iNOS (iNOS−/−) or treated with chemical inhibitors of NO production are extremely susceptible to Mtb [2]. Low dose aerosol infection of iNOS−/− mice with wild type Mtb (~200 bacteria/mouse) results in death within 3 months [5, 51, 58]. In contrast, iNOS−/− mice live significantly longer when infected with an mpa or pafA Mtb strain (~200–500 days post-infection) compared to infection with wild type Mtb (~60–80 days post-infection) [51, 58]. Because disruption of iNOS does not fully restore the virulence of the mpa and pafA mutants, it appears that the role of the Mtb proteasome extends beyond protection against RNIs in vivo.</p><p>In another study Bishai and colleagues identified three clones from a collection of random transposon insertion mutants in CDC1551 (a clinical isolate of Mtb) that were consistently smaller than the wild type strain [95]. All three independent mutants contained insertions in MT2175, which is identical to mpa in Mtb H37Rv. The CDC1551 mutants grow slower (doubling time of ~22 h) than the wild type strain (doubling time of 18 h) in standard 7H9 medium and fail to reach the same final culture density as wild type Mtb. Complementation of this CDC1551 mutant strain with mpa restores wild type colony morphology. Infection of BALB/c mice with a CDC1551 mpa mutant results in similar infection profiles as previously observed with the H37Rv mpa mutant. During the chronic phase of infection, bacterial numbers gradually decrease. Mice infected with CDC1551 mpa mutants survive without any signs of disease until 180 days (the latest time point assessed), while in contrast, mice infected with wild type Mtb succumb within 70 days. Lungs of mice infected with CDC1551 mpa mutants have attenuated pathological symptoms, such as less inflammation and fewer granuloma-like foci, and no weight loss compared to mice infected with wild type Mtb. Interferon gamma (IFN-γ) production fails to rise after 3 weeks of infection with the mutant compared to the wild type Mtb strain, hence mpa mutants seem to elicit a milder Th1 immune response in mice [95].</p><!><p>Ehrt and colleagues made the puzzling observation that proteasomes containing a mutation in the active site can complement a prcBA null mutation in Mtb for RNI sensitivity and slow growth, but were unable to rescue impaired bacterial persistence in mice [29]. It is unclear how CPs that are proteolytically inactive could restore certain defects but not others. However, it may be possible that the CP has activities that have not yet been identified by routine biochemical assays. For example, the CP may act as a dock or scaffold for other proteins in order to function in specific stress responses. Taken together, these results suggest that the CP, proteolytically active or not, has a broader role for normal cell growth in virulent mycobacteria compared to its non-pathogenic relative Msm in which the CP appears to be dispensable under all conditions tested so far [19] (K.H.D., unpublished observations). The genome of Msm (7 Mb for Msm mc2155) is considerably larger than that of Mtb (4.4 Mb for Mtb H37Rv) and, unlike Mtb, Msm encodes another ATP-dependent compartmentalized protease (Lon protease) that may be able to compensate for deletion of prcBA in Msm (reviewed in [14]).</p><!><p>Although it is clear that the lack of proteasome function is a disadvantage for Mtb fitness during an infection, it remains to be established how proteasomal proteolysis is linked to pathogenesis. It seems likely that the inability to regulate proteins through degradation compromises bacterial survival when adapting to a new environment, i.e. within activated macrophages. There are several hypotheses that could explain why proteasome function is protective against RNI stress and important for survival in an animal host. Perhaps the simplest explanation is that the proteasome degrades damaged proteins. Oxidative and nitrosative damage of proteins can result in misfolding and aggregation, which is potentially lethal to cells. This damage could possibly be a signal for pupylation. It is also possible that specific accumulated proteasome substrates are particularly dangerous in the presence of NO. For example, iron-sulfur (Fe-S) clusters or copper (Cu) in metal binding proteins can be displaced by NO [96]. The liberation of Fe2+ or Cu+ is highly toxic to the cell as it can catalyze Fenton chemistry, resulting in the production of ROI. The observation that mpa, pafA and prcBA mutants are hyper-resistant to hydrogen peroxide, suggests that other anti-oxidant pathways may already be induced in an attempt to compensate for loss of the Pup-proteasome system. Currently, however, there is no evidence for the presence of increased amounts of metal-binding or damaged proteins in proteasomal degradation-defective mutants treated with NO.</p><!><p>Another potential function of the proteasome is in transcriptional regulation. Almost all (if not all) compartmentalized proteases have been shown to regulate gene expression (reviewed in [97]). A microarray study comparing wild type Mtb with mpa and pafA mutants grown under standard culture conditions revealed that a common set of genes was differentially regulated (Table 10.1) [98]. Notably, none of the identified genes appears to be associated with the NO sensitive phenotype of the mpa and pafA mutants. Among the up-regulated genes in the mpa and pafA mutants were members of the zinc uptake regulator (Zur) regulon. In the presence of Zn, Zur is released from operators in at least three promoters in Mtb, and gene expression is induced [99]. One of the Zur-regulated promoters identified in the microarray drives the expression of the esx-3 (ESAT-6, region 3) operon. The esx-3 locus is, for the most part, essential for the growth of Mtb under normal culture conditions and is proposed to encode a type VII secretion system that is involved in zinc and iron acquisition [100, 101]. In addition, Zur regulates the expression of genes that encode homologues of Zn-binding ribosomal proteins. Ribosomes are comprised of numerous small proteins, several of which bind Zn. Under Zn-limiting conditions, these Zn-binding proteins are thought to be replaced with non-metal binding components [102, 103], allowing the bacteria to gain access to a large pool of zinc. If mutations in mpa or pafA result in deregulation of the Zur regulon in vivo as they do in vitro, these data would suggest that metal homeostasis is critical during infection and an inappropriate increase in expression of the Zur regulon during infection may also have deleterious affects on bacterial survival for other reasons.</p><p>Transcriptome analysis also identified a set of genes repressed in the mpa and pafA mutants that are regulated by copper [98]. Several of these genes form a copper-inducible regulon, which is under the control of RicR (regulated in copper repressor). During copper depleted conditions, RicR represses five promoters that drive the expression of ricR itself, mymT (a copper methallothionein) and several genes of unknown function (Table 10.1). Disruption of ricR results in hyper-resistance of Mtb to normally toxic levels of copper, presumably due to the constitutive expression of one or more copper resistance genes like mymT [96]. It is worth noting that several of the RicR-regulated genes (mymT, lpqS, socAB) are unique to pathogenic mycobacteria, suggesting that copper regulation is important for virulence. Thus, the attenuated phenotype of Pup-proteasome pathway mutants may in part be explained by the incomplete derepression of the RicR regulon during infection. These data support an emerging notion that copper has an important antimicrobial role during TB infection and possibly other infections [96, 98, 104–107].</p><p>It is interesting that two metal-dependent regulons are deregulated in mpa and pafA mutants. In both cases, the mutant bacteria appear to be responding to low metal concentrations. These data also suggest Mtb (and possibly other bacteria) need to adapt to changes in metal homeostasis in the host. As with Zur, it is currently not understood how the proteasome affects the expression of the RicR regulon.</p><!><p>As with other organisms, regulated proteolysis is critical for numerous aspects of TB biology. Mtb possesses a proteasome highly similar to those found in other domains of life, and uses it to resist host derived stresses like NO and regulate pathways that may be needed for pathogenesis. Proteasomal proteolysis is controlled, in part, by pupylation, which is functionally, if not biochemically, similar to ubiquitylation. The characterization of proteasome biochemistry and biology will undoubtedly allow researchers to gain valuable insight into the lifestyle of one of the most successful human pathogens. Among the numerous questions that remain to be asked of the young field of bacterial proteasome biology include:</p><!><p>How are proteins selected for pupylation and depupylation?</p><p>How does Mpa interact with the 20S CP? Why is Mpa itself a proteasome substrate?</p><p>Does pupylation have functions independent of targeting proteins to the proteasome?</p><p>How is proteasome function linked to NO resistance?</p><p>Are misfolded or damaged proteins degraded in a proteasome dependent manner?</p><p>Why and how are the Zur and RicR regulons affected by proteasome activity?</p>

PubMed Author Manuscript

Porphyrin-polymer nanocompartments: singlet oxygen generation and antimicrobial activity

AbstractA new water-soluble photocatalyst for singlet oxygen generation is presented. Its absorption extends to the red part of the spectrum, showing activity up to irradiation at 660 nm. Its efficiency has been compared to that of a commercial analogue (Rose Bengal) for the oxidation of l-methionine. The quantitative and selective oxidation was promising enough to encapsulate the photocatalyst in polymersomes. The singlet oxygen generated in this way can diffuse and remain active for the oxidation of l-methionine outside the polymeric compartment. These results made us consider the use of these polymersomes for antimicrobial applications. E. coli colonies were subjected to oxidative stress using the photocatalyst–polymersome conjugates and nearly all the colonies were damaged upon extensive irradiation while under the same red LED light irradiation, liquid cultures in the absence of porphyrin or porphyrin-loaded polymersomes were unharmed.Graphical abstract Electronic supplementary materialThe online version of this article (10.1007/s00775-017-1514-8) contains supplementary material, which is available to authorized users.

porphyrin-polymer_nanocompartments:_singlet_oxygen_generation_and_antimicrobial_activity

<!>Introduction<!><!>General<!>Synthesis of TPyCP<!>Photocatalysis<!>Kinetics<!>Polymersome preparation<!>Transmission electron microscopy<!>Light scattering<!>Fluorescence spectroscopy<!>Confocal laser scanning microscopy (CLSM)<!>Bacterial assays<!>Live/Dead® staining for microscopy<!><!>Methionine substrate and a comparison with Rose Bengal sensitization<!><!>Methionine substrate and a comparison with Rose Bengal sensitization<!><!>Biological evaluation of TPyCP in solution and within polymersomes<!><!>Live/Dead® staining<!>Conclusions<!>

<p>The structure of Rose Bengal</p><!><p>To improve the effect of photosensitizers in the specific cell compartments where they function and to avoid side effects associated with their presence in other regions, a number of carrier systems have been developed. The use of carriers based on nanoassemblies is very appealing because such assemblies (micelles, nanoparticles, polymersomes) can be chemically designed to possess the necessary sizes and properties to be taken up by cells, protect the photosensitizers and release them only under specific conditions related to the cell compartments where singlet oxygen and related ROS production is required [22, 23].</p><p>The use of nanoassemblies to host photosensitizers (PS) based both on natural and synthetic polymers provides an efficient and safe approach to photodynamic therapy (PDT) [24, 25] or to generate surfaces with antimicrobial properties [26]. As long as the functionality of the PS remains intact within the nanoassemblies during the photoactivation process, the system can reach its full potential and new opportunities for biomedical applications become accessible. Among the different assemblies, polymer vesicles (or polymersomes) are excellent candidates for this purpose. Their properties can be tuned leading to systems that are mechanically more stable than lipid-based compartments (liposomes). Furthermore, if appropriately selected with respect to the chemical nature of the amphiphilic copolymers, polymersomes are biocompatible and may be stimuli-responsive. The hollow spherical architecture of polymersomes permits the encapsulation of hydrophilic compounds and the insertion of hydrophobic compounds into their membrane [27, 28]. An elegant solution to improve the control of the photosensitizers is not to release them at the desired cell compartments [29, 30], but to ensure that they remain encapsulated in polymersomes allowing one to produce ROS 'on demand' upon irradiation [31]. Using this approach, we have shown that encapsulated Rose Bengal conjugated with BSA generates singlet oxygen inside the cavity of polymersomes upon irradiation, and the associated ROS is released in the environment of the polymersomes and is able to induce apoptosis [31, 32]. Reports indicating that porphyrin derivatives have increased intrinsic toxicity upon irradiation in various cell lines [33] encourage the use of nanocarriers to decrease their intrinsic toxicity through encapsulation.</p><!><p>Structure of the photosensitizer TPyCP</p><!><p>1H and 13C NMR spectra were recorded on a Bruker Avance III-400 NMR spectrometer. 1H and 13C NMR chemical shifts were referenced to residual solvent peaks with respect to δ(TMS) = 0 ppm. Solution absorption spectra were recorded using an Agilent Cary 5000 spectrophotometer and ESI mass spectra were recorded using a Shimadzu LCMS-2020 instrument. The LED light source used to activate the photocatalyst was a THORLABS 4-Wavelength High-Power LED Source LED4D067.</p><!><p>The compound was prepared according to literature [34]. The compounds TPyP, ethyl 6-bromohexanoate and l-methionine were purchased from Sigma-Aldrich and used without further purification.</p><!><p>l-Methionine (200 mg, 1.34 mmol) and TPyCP (0.25 mg, 0.165 μmol, 0.013 mol %) were dissolved in D2O (5 mL) in a 5 mL round-bottomed flask. The reaction was repeated using 0.006 mol% catalyst loading: l-methionine (200 mg, 1.34 mmol) and TPyCP (0.12 mg, 0.79 μmol, 0.006 mol%). In each case, the flask was closed with a rubber septum and two needles were inserted through the septum. A pump was connected to one needle to bubble air into the solution; the other was used as an exhaust. The LED lamp was placed so that the light source was perpendicular to the flask ensuring that the whole of the solution was irradiated. Conversion was monitored using 1H NMR spectroscopy.</p><!><p>l-Methionine (63 mg, 0.42 mmol) was dissolved in D2O (5 mL) in a 5 mL round-bottomed flask. Either TPyCP (1.3 mg, 0.84 μmol, 0.2 mol%) or Rose Bengal (4.3 mg, 4.2 μmol, 1 mol%) were added to the solution and the flask was capped with a rubber septum and two needles were inserted through the septum. The flask was connected to a pump to force air in through one needle and the second needle was the outlet. The mixture containing Rose Bengal was irradiated at 505 nm and that with TPyCP at 660 nm. Conversion of l-methionine to l-methionine sulfoxide was monitored using 1H NMR spectroscopy by comparison of the relative integrals of the resonances of the signals for the Me group adjacent to the sulfur atom (δ 2.10 ppm for l-methionine and δ 2.70 ppm for l-methionine sulfoxide). Reactions were repeated with different catalyst loadings as detailed in the text. Kinetic studies were also performed with the TPyCP-loaded polymersomes in three different initial concentrations. These results will be discussed in the next section.</p><!><p>For the formation of the polymersomes, the amphiphilic triblock copolymer PMOXA34–PDMS6–PMOXA34 [35] and prepared as previously described was used. The film rehydration method was followed [36]. Polymer (5 mg) was dissolved in MeOH (1 mL) and dried under vacuum to form a polymer film on the inner bottom surface of a 5 mL glass flask. The polymer film was rehydrated with Tris buffer (50 mM, pH 7.6) at room temperature for 48 h in the dark in the presence or absence of a 50, 100 and 200 μΜ TPyCP solution, respectively. The suspension was then sequentially extruded through 0.2 and 0.1 μm Nucleopore Track-Etch membranes from Whatman using an Avanti Extruder (Avanti Polar Lipids, USA). Any TPyCP left in solution was separated from the polymersomes containing TPyCP by passage through a HiTrap desalting column (Sephadex G-25 Superfine, GE Healthcare, UK) or a 20 cm3 in-house prepacked column (Sepharose 2B, Sigma-Aldrich). The polymersomes obtained were characterized by light scattering measurements (LS) and transmission electron microscopy (TEM).</p><!><p>For visualization, 10 μL of a polymersome solution was negatively stained with 2% aqueous uranyl acetate solution, deposited on a carbon-coated copper grid, and then examined with a transmission electron microscope (Philips Morgani 268 D) operating at 80 kV.</p><!><p>Dynamic (DLS) and static (SLS) light scattering experiments were performed on an ALV (Langen, Germany) goniometer equipped with an ALV He–Ne laser (JDS Uniphase, wavelength λ = 632.8 nm). Polymersome emulsions were serially diluted to polymer concentrations ranging from 5 to 0.325 mg/mL, and measured in 10 mm cylindrical quartz cells at angles of 30°–150° and a temperature of 293 ± 0.5 K. The photon intensity autocorrelation function g 2(t) was determined with an ALV-5000E correlator (scattering angles between 30° and 150°). A non-linear decay-time analysis supported by regularized inverse Laplace transform of g 2(t) (CONTIN algorithm) was used to analyse DLS data. The angle-dependent apparent diffusion coefficient was extrapolated to zero momentum transfer (q 2) using the ALV/Static and dynamic FIT and PLOT 4.31 software. Angle and concentration-dependent SLS data were analysed using Guinier plots. Errors were calculated from the deviation of the fit parameters using the ALV/static and dynamic FIT and PLOT software</p><!><p>The fluorescence measurements were carried out on an LS 55 fluorescence spectrometer from Perkin Elmer with a FL Winlab software. Polymersomes loaded with TPyCP were measured in a 1 cm path length quartz cuvette. A wavelength of 424 nm was used to excite in the Soret band and the emission was monitored at 580 nm. Excitation and emission slits were set at 7.5 nm.</p><!><p>CLSM experiments were performed on a confocal laser scanning microscope (Zeiss LSM 880, Carl Zeiss, Jena, Germany) with an Argon/2 laser (λ = 488 nm, 30 mW, 10% power output, 0.5% transmission) as the excitation source. A main dichromatic beam splitter (HFT 488/543), and a band pass filter (BP 505–530) were used in all experiments. The images were recorded with a water immersion objective (C-Apochromat 40×/1.2 W).</p><!><p>Aliquots of 5 μL taken from a stock E. coli colony were dispersed into 15 mL of lysogeny broth (LB) in a 50 mL falcon tube. The suspension was shaken at 180 rpm in an incubator overnight at 37 °C, with the cup not entirely closed to allow molecular oxygen to diffuse and reach the bacteria. The culture was then concentrated by centrifugation at 10,000g for 10–15 min. The supernatant was removed and the culture was resuspended in 15 mL of phosphate buffered saline solution (PBS), centrifuged again to remove the remaining media and resuspended for the last time in 15 mL of PBS containing 1% tryptic soy broth (TSB). The bacterial culture was serially diluted (10−1–10−5 CFU/10 μL, CFU being colony forming units) in a 24-well plate. In the last liquid culture, which corresponds to the desired dilution, 200 μL of an aqueous TPyCP solution or a solution with polymersomes containing TPyCP were added. The 24-well plate was kept in the dark until the exposure to LED irradiation started. Liquid cultures were taken and placed into LB-Agar plates after 0, 30, 120, 240 and 360 min of irradiation. The plated cultures were incubated overnight at 37 °C and the colonies formed counted. As a control, the same series of experiment were performed in the absence of TPyCP. Furthermore, a so-called dark control was performed simultaneously: an E. coli culture in the presence of TPyCP was kept in the dark by means of aluminum foil wrapping. Three independent experiments were run, and for each three replicates were plated. Both controls were carried out during each experiment. All the E. coli essays were carried out in a sterile environment.</p><!><p>For this assay, the Live/Dead® BacLight (https://assets.thermofisher.com/TFS-Assets/LSG/manuals/mp07007.pdf). Bacterial Viability Kit for microscopy has been used. Briefly, the E. coli bacteria liquid cultures were grown as described previously, with the difference that instead of PBS, aqueous 0.85% NaCl (0.85 g per 100 mL) was used as the suspension buffer. After illumination with a red LED light at 660 nm, the cultures were stained with a 1:1 mixture of SYTO 9 (https://assets.thermofisher.com/TFS-Assets/LSG/manuals/mp07572.pdf), which belongs to the family of SYTO dyes and is a cell-permeant nucleic acid stain, and propidium iodide dyes using 3 μL of 72 nM stain per 1 ml of sample, and then incubated for 10 min at room temperature. Bacteria with intact cell membranes (considered alive) stained fluorescent green, whereas bacteria with damaged membranes (considered dead) stained fluorescent red. The excitation/emission maxima for these dyes are in the range 480–500 nm for SYTO 9 and 490–635 nm for propidium iodide. The background remains virtually non-fluorescent. The results were analysed by CLSM.</p><!><p>a The absorption spectrum of TPyCP (4 μM in D2O). The coloured areas show the overlap of the absorption with the emission bands of the LED used for irradiation. b Expansion of the low energy part of the spectrum</p><p>Excited state energy levels for TPyCP compared to the ground and low-lying excited states of O2. The S1 electronic state can be populated by means of both green and red LED light by virtue of its extended absorption (vibrational structure)</p><p>Conversion of l-methionine to the diastereoisomeric pair of l-methionine sulfoxides. Atom labels are for NMR spectroscopic assignments</p><p>a l-Methionine conversion (given as % on the ordinate) to the diastereoisomeric l-methionine sulfoxides in the presence of 0.06 mol% TPyCP. 470 nm irradiation (blue), 505 nm irradiation (green), 660 nm irradiation (red). b Conversion of a 0.42 mmol l-methionine solution in D2O in presence of 0.2 mol% TPyCP (red) or 1 mol% Rose Bengal (green). Irradiation was performed at 505 nm for Rose Bengal or at 660 nm for TPyCP. Data points (circles) and linear best fit (straight lines), r 2 = 0.9915 for Rose Bengal and 0.9872 for TPyCP</p><!><p>We next carried out a comparative investigation of the kinetics of the selective oxidation of l-methionine using TPyCP or the commercially available Rose Bengal as photosensitizer. The aim of this study was to determine whether use of the TPyCP could be advantageous over the commercial and widely used Rose Bengal. The data in Fig. 3a confirmed that TPyCP acts with equal efficiency upon irradiation at 505 or 660 nm. In contrast, Rose Bengal possesses no absorption at 660 nm and, as expected, no conversion of l-methionine to its sulfoxides was observed upon irradiating at this wavelength. For the kinetic experiments, a D2O solution of l-methionine (0.42 mmol) was irradiated in the presence of TPyCP (0.2 mol%) or Rose Bengal (1 mol%) with λ exc = 660 or 505 nm, respectively. The reaction was monitored by 1H NMR spectroscopy and the results are shown in Fig. 3b. Both photocatalysts lead to complete conversion of l-methionine to a 1:1 ratio of both diastereoisomers of l-methionine sulfoxide, and although Fig. 3b might suggest that the reaction with Rose Bengal is faster; it is significant that TPyCP is present at 0.2 mol% compared to 1 mol% of Rose Bengal.</p><!><p>l-Methionine (270 mM in D2O) conversion for a TPyCP concentration of 150 μM (0.055 mol%) (circles), 67 μM (0.024 mol%) (squares), 34 μM (0.013 mol%) (triangles) and 17 μM (0.006 mol%) (filled red circles) (λ exc = 660 nm)</p><p>Proposed mechanism for the photocatalytic oxidation of methionine (based on scheme in Ref. [40])</p><!><p>The scheme can be summarized, and made more general, by writing Eqs. 1–4 in which P and P* are the porphyrin catalyst in its ground and excited state, respectively, 3O2 and 1O2 are molecular oxygen in its triplet ground and singlet excited state, respectively; S, SOO and SO are defined in Scheme 4.1\documentclass[12pt]{minimal} sepackage{amsmath} sepackage{wasysym} sepackage{amsfonts} sepackage{amssymb} sepackage{amsbsy} sepackage{mathrsfs} sepackage{upgreek} \setlength{\oddsidemargin}{-69pt} egin{document}$${ ext{P}} + hv\mathop o \limits^{{ k_{1 } }} { ext{P}}^{*}$$\end{document}P+hv→k1P∗ 2\documentclass[12pt]{minimal} sepackage{amsmath} sepackage{wasysym} sepackage{amsfonts} sepackage{amssymb} sepackage{amsbsy} sepackage{mathrsfs} sepackage{upgreek} \setlength{\oddsidemargin}{-69pt} egin{document}$${ ext{P}}^{*} +^{3}{ ext{O}}_{2} \mathop o \limits^{{ k_{2 } }} {} ^{1} { ext{O}}_{2}$$\end{document}P∗+3O2→k21O2 3\documentclass[12pt]{minimal} sepackage{amsmath} sepackage{wasysym} sepackage{amsfonts} sepackage{amssymb} sepackage{amsbsy} sepackage{mathrsfs} sepackage{upgreek} \setlength{\oddsidemargin}{-69pt} egin{document}$${ ext{S}} + ^{1} { ext{O}}_{2} \mathop o \limits^{{ k_{3 } }} { ext{SOO}}$$\end{document}S+1O2→k3SOO 4\documentclass[12pt]{minimal} sepackage{amsmath} sepackage{wasysym} sepackage{amsfonts} sepackage{amssymb} sepackage{amsbsy} sepackage{mathrsfs} sepackage{upgreek} \setlength{\oddsidemargin}{-69pt} egin{document}$${ ext{SOO}} + { ext{S}}\mathop o \limits^{{ k_{4 } }} 2 { ext{SO}}$$\end{document}SOO+S→k42SO</p><p>From Eqs. 1–4, we can derive Eqs. 5–9 for species in the catalytic cycle.5\documentclass[12pt]{minimal} sepackage{amsmath} sepackage{wasysym} sepackage{amsfonts} sepackage{amssymb} sepackage{amsbsy} sepackage{mathrsfs} sepackage{upgreek} \setlength{\oddsidemargin}{-69pt} egin{document}$$rac{{{ ext{d}}[{ ext{P}}^{ *} ]}}{{{ ext{d}}t}} = k_{1} - k_{2} [^{3} { ext{O}}_{2} ][{ ext{P}}^{*} ]$$\end{document}d[P∗]dt=k1-k2[3O2][P∗] 6\documentclass[12pt]{minimal} sepackage{amsmath} sepackage{wasysym} sepackage{amsfonts} sepackage{amssymb} sepackage{amsbsy} sepackage{mathrsfs} sepackage{upgreek} \setlength{\oddsidemargin}{-69pt} egin{document}$$rac{{{ ext{d}}[^{1} { ext{O}}_{2} ]}}{{{ ext{d}}t}} = k_{2} [^{3} { ext{O}}_{2} ][{ ext{P}}^{*} ] - k_{3} [{ ext{S}}][^{1} { ext{O}}_{2} ]$$\end{document}d[1O2]dt=k2[3O2][P∗]-k3[S][1O2] 7\documentclass[12pt]{minimal} sepackage{amsmath} sepackage{wasysym} sepackage{amsfonts} sepackage{amssymb} sepackage{amsbsy} sepackage{mathrsfs} sepackage{upgreek} \setlength{\oddsidemargin}{-69pt} egin{document}$$rac{{{ ext{d}}[{ ext{S}}]}}{{{ ext{d}}t}} = - k_{3} [{ ext{S}}][^{1} { ext{O}}_{2} ] - k_{4} [{ ext{SOO}}][{ ext{S}}]$$\end{document}d[S]dt=-k3[S][1O2]-k4[SOO][S] 8\documentclass[12pt]{minimal} sepackage{amsmath} sepackage{wasysym} sepackage{amsfonts} sepackage{amssymb} sepackage{amsbsy} sepackage{mathrsfs} sepackage{upgreek} \setlength{\oddsidemargin}{-69pt} egin{document}$$rac{{{ ext{d}}[{ ext{SOO}}]}}{{{ ext{d}}t}} = k_{3} [^{1} { ext{O}}_{2} ][{ ext{S}}] - k_{4} [{ ext{SOO}}][{ ext{S}}]$$\end{document}d[SOO]dt=k3[1O2][S]-k4[SOO][S] 9\documentclass[12pt]{minimal} sepackage{amsmath} sepackage{wasysym} sepackage{amsfonts} sepackage{amssymb} sepackage{amsbsy} sepackage{mathrsfs} sepackage{upgreek} \setlength{\oddsidemargin}{-69pt} egin{document}$$rac{{{ ext{d}}[{ ext{SO}}]}}{{{ ext{d}}t}} = 2k_{4} [{ ext{SOO}}][{ ext{S}}]$$\end{document}d[SO]dt=2k4[SOO][S]</p><p>Experimentally, it is not possible to detect SOO. It does not accumulate over the course of the reaction and we may therefore, apply the steady-state approximation for this intermediate (Eqs. 10, 11).10\documentclass[12pt]{minimal} sepackage{amsmath} sepackage{wasysym} sepackage{amsfonts} sepackage{amssymb} sepackage{amsbsy} sepackage{mathrsfs} sepackage{upgreek} \setlength{\oddsidemargin}{-69pt} egin{document}$$rac{{{ ext{d}}[{ ext{SOO}}]}}{{{ ext{d}}t}} = k_{3} [^{1} { ext{O}}_{2} ]\left[ { ext{S}} ight] - k_{4} \left[ { ext{SOO}} ight]\left[ { ext{S}} ight] = 0$$\end{document}d[SOO]dt=k3[1O2]S-k4SOOS=0 11\documentclass[12pt]{minimal} sepackage{amsmath} sepackage{wasysym} sepackage{amsfonts} sepackage{amssymb} sepackage{amsbsy} sepackage{mathrsfs} sepackage{upgreek} \setlength{\oddsidemargin}{-69pt} egin{document}$$[{ ext{SOO}}] = rac{{k_{3} [^{1} { ext{O}}_{2} ]}}{{k_{4} }}$$\end{document}[SOO]=k3[1O2]k4</p><p>Equation 11 requires that SOO does not accumulate if k4 ≫ k3. Substituting Eq. 11 into Eq. 7 leads to Eq. 12, and hence, Eq. 13.12\documentclass[12pt]{minimal} sepackage{amsmath} sepackage{wasysym} sepackage{amsfonts} sepackage{amssymb} sepackage{amsbsy} sepackage{mathrsfs} sepackage{upgreek} \setlength{\oddsidemargin}{-69pt} egin{document}$$rac{{{ ext{d}}[{ ext{S}}]}}{{{ ext{d}}t}} = - 2k_{3} [{ ext{S}}][^{1} { ext{O}}_{2} ]$$\end{document}d[S]dt=-2k3[S][1O2] 13\documentclass[12pt]{minimal} sepackage{amsmath} sepackage{wasysym} sepackage{amsfonts} sepackage{amssymb} sepackage{amsbsy} sepackage{mathrsfs} sepackage{upgreek} \setlength{\oddsidemargin}{-69pt} egin{document}$$[{ ext{S}}] = { ext{e}}^{{ - 2k_{3} [^{1} { ext{O}}_{2} ]t}}$$\end{document}[S]=e-2k3[1O2]t</p><p>From Eq. 13, it follows that the consumption of l-methionine depends exponentially the supply of oxygen (represented by k 3) and the concentration of singlet oxygen. The latter dependency can be eliminated if we consider the experimental setup. Triplet oxygen, 3O2, is introduced into the reaction vessel by means of a pump, which keeps the concentration of 3O2 constant at the saturation level of the solvent. In addition, the concentration of the catalyst in the excited state (P*) is also constant. This is due to the fact that the concentration used is such that the absorbance of the reaction mixture at the LED λ exc = 660 nm is 0.09 ([catalyst] = 150 μmol dm−3, path length = 2.5 cm, ε = 200 dm3/mol/cm) which converts to 19% of the excitation light being absorbed as it travels through the reaction flask. It is true that, once excited, P is in the lowest singlet excited state, whereas it is the lowest triplet excited state that is able to transfer energy to molecular oxygen. Due to intersystem crossing (ISC) which is a unimolecular process, P* relaxes to this state, which constitutes the active state of the catalyst. We have no reason to assume any of those processes (light absorption in non-saturated conditions and ISC) are time dependent. Furthermore, the extent of the energy transfer to ground state molecular oxygen is a property of a given couple of molecules, and therefore, time independent as well. The latter two considerations allow us to simplify Eq. 12 by applying the steady-state approximation to 1O2 (Eqs. 14, 15). This species is the product of the reaction between P* and 3O2, concentrations of which are constant over time.14\documentclass[12pt]{minimal} sepackage{amsmath} sepackage{wasysym} sepackage{amsfonts} sepackage{amssymb} sepackage{amsbsy} sepackage{mathrsfs} sepackage{upgreek} \setlength{\oddsidemargin}{-69pt} egin{document}$$rac{{{ ext{d}}[^{1} { ext{O}}_{2} ]}}{{{ ext{d}}t}} = k_{2} [^{3} { ext{O}}_{2} ][{ ext{P}}^{ *} ] - k_{3} [{ ext{S}}][^{1} { ext{O}}_{2} ] = 0$$\end{document}d[1O2]dt=k2[3O2][P∗]-k3[S][1O2]=0 15\documentclass[12pt]{minimal} sepackage{amsmath} sepackage{wasysym} sepackage{amsfonts} sepackage{amssymb} sepackage{amsbsy} sepackage{mathrsfs} sepackage{upgreek} \setlength{\oddsidemargin}{-69pt} egin{document}$$[^{1} { ext{O}}_{2} ] = rac{{k_{2} [^{3} { ext{O}}_{2} ][{ ext{P}}^{ *} ]}}{{k_{3} [{ ext{S}}]}}$$\end{document}[1O2]=k2[3O2][P∗]k3[S]</p><p>Substituting Eq. (15) into Eq. (12) gives Eq. 17 in which [S]0 is the initial concentration on l-methionine. Equation 17 shows a linear relationship between [S] and t. All of the other dependencies are eliminated.16\documentclass[12pt]{minimal} sepackage{amsmath} sepackage{wasysym} sepackage{amsfonts} sepackage{amssymb} sepackage{amsbsy} sepackage{mathrsfs} sepackage{upgreek} \setlength{\oddsidemargin}{-69pt} egin{document}$$rac{{{ ext{d }}[{ ext{S}}]}}{{{ ext{d}}t}} = - 2k_{2} [^{3} { ext{O}}_{2} ][{ ext{P}}^{ *} ]$$\end{document}d[S]dt=-2k2[3O2][P∗] 17\documentclass[12pt]{minimal} sepackage{amsmath} sepackage{wasysym} sepackage{amsfonts} sepackage{amssymb} sepackage{amsbsy} sepackage{mathrsfs} sepackage{upgreek} \setlength{\oddsidemargin}{-69pt} egin{document}$$[{ ext{S}}] = [{ ext{S}}]_{0} - 2k_{2} [^{3} { ext{O}}_{2} ][{ ext{P}}^{ *} ]t$$\end{document}[S]=[S]0-2k2[3O2][P∗]t</p><!><p>TEM micrographs of polymersomes a without TPyCP and b with encapsulated TPyCP. Scale bars are 200 nm</p><p>Light scattering data for supramolecular assemblies without and with TPyCP</p><p>The stated concentration is the initial value</p><p>a Emission spectra of the free TPyCP and TPyCP-loaded polymersomes: free TPyCP 200 μΜ in Tris buffer (magenta line), TPyCP in polymersomes with an initial concentration of 200 μΜ (blue line), TPyCP in polymersomes with an initial concentration of 100 μΜ (green line), TPyCP 50 μΜ in polymersomes with an initial concentration of 50 μΜ (red line), empty polymersomes (black line), second fraction from the purification of 200 μΜ TPyCP in polymersomes (dashed black line). b Integral of the emission spectra for the porphyrin-loaded polymersomes as a function of the initial concentration of TPyCP</p><p>l-Methionine conversion to (R)/(S)-sulfoxide by free TPyCP (67 μΜ in D2O, blue squares), TPyCP-loaded polymersomes with an initial porphyrin concentration of 200 μM (yellow squares), free methionine in solution (red empty triangles) and empty polymersomes incubated with 200 μM TPyCP and then purified (green empty squares) (λ exc = 660 nm)</p><p>a CFU E. coli when the bacteria were irradiated with a LED red source (λ max = 660 nm) and treated with 200 μΜ free TPyCP (purple circles), not treated with TPyCP (circle), and without irradiation in presence of TPyCP (hexagons). b CFU of E. coli treated with: TPyCP-loaded polymersomes irradiated with a LED red source (λ max = 660 nm) (blue squares), TPyCP-loaded polymersomes without irradiation (hexagons) and empty polymersomes irradiated with a LED red source (λ max = 660 nm) (squares)</p><!><p>Similar behaviour was observed for bacteria cultures treated with porphyrin-loaded polymersomes: they were able to drastically reduce CFU values only upon irradiation (Fig. 8b). As expected, in the presence of empty polymersomes no influence on bacterial growth upon irradiation was detected. Furthermore, porphyrin-loaded polymersomes had no effect on the CFU in the dark, which supports an "on demand" functionality: only upon irradiation can the porphyrin-loaded polymersomes produce singlet oxygen inducing significant bacterial inhibition. The porphyrin-loaded polymersomes remain "silent" in the dark. Porphyrin-loaded polymersomes induced a significant decrease of the CFU in a rapid manner.</p><!><p>E. coli bacteria stained with SYTO 9 (considered alive, green) and propidium iodide (considered dead, red) incubated in presence of 200 μΜ free TPyCP after 0 min (a), 30 min (b), 120 min (c), 240 min (d) and 360 min (e) of illumination under red LED light (λ max = 660 nm)</p><p>E. coli bacteria stained with SYTO 9 (considered alive, green) and propidium iodide (considered dead, red) incubated in presence of TPyCP-loaded polymersomes after 0 min (a) 30 min, (b) 120 min, (c) 240 min, (d) and 360 min (e) of illumination under red LED light (λ max = 660 nm)</p><!><p>Of greater significance is what happens when photoactivation of TPyCP takes place within polymersomes (Fig. 10). Similar to the free porphyrin, at the beginning the bacteria are alive (green stain), while as the irradiation time increases the population shifts towards a red stain. After 360 min of irradiation (Fig. 10e), almost all bacteria were apoptotic (red). Once more, as indicated from the study of CFU, only the combination of irradiation and TPyCP is able to damage E. coli. When treated with TPyCP, but kept in dark, and as well as when irradiated, but in absence of TPyCP, the bacteria stay unharmed for 360 min (Figs. S4–S7).</p><!><p>A tetraalkylpyridinium porphyrin TPyCP has been prepared, with the aim of exploiting its light absorption for photosensitized conversion of triplet to singlet oxygen. The presence of the set of four Q bands allows the compound to operate in the deep red, with an excitation wavelength as high as 660 nm. The compound is a photocatalyst for the oxidation of l-methionine and, more importantly, its activity is not diminished by encapsulation in polymersomes. Although the encapsulated TPyCP remains internalized in the polymersome, the small, long-lived and reactive singlet oxygen can diffuse through the membrane and react with external substrate. Encapsulation allows incubating the compound with a bacterial culture, without the drawback of the photosensitizer diffusing in the media. Live bacteria decrease significantly when the TPyCP-loaded polymersomes are irradiated with red light. These promising results prove the antimicrobial activity of TPyCP-polymersome system and make us consider expanding the biological evaluation towards in vitro studies on human cells.</p><!><p>Supplementary material 1 (PDF 2187 kb)</p><p>Photodynamic therapy</p><p>Photosensitizer</p><p>Light scattering</p><p>Transmission electron microscopy</p><p>Laser scanning microscopy</p><p>Colony-forming units</p><p>Confocal laser scanning microscopy</p><p>Lysogeny broth</p><p>Phosphate buffered saline</p><p>Reactive oxygen species</p><p>Tryptic soy broth</p><p>MS electrospray ionization mass spectrometry</p><p>Light-emitting diode</p><p>5,10,15,20-Tetraphenylporphyrin</p><p>5,10,15,20-Tetra(pyridin-4-yl)porphyrin</p><p>5,10,15,20-Tetrakis(1-(6-ethoxy-6-oxohexyl)-4-pyridin-1-io)-21H,23H-porphyrin tetrabromide</p><p>We dedicate this manuscript to our good friend and colleague in Basel, Helmut Sigel, on the occasion of his 80th birthday.</p><p>Electronic supplementary material</p><p>The online version of this article (10.1007/s00775-017-1514-8) contains supplementary material, which is available to authorized users.</p><p>Angelo Lanzilotto and Myrto Kyropoulou have contributed equally to this work.</p>

PubMed Open Access

Diastereoselective Ugi reaction of chiral 1,3-aminoalcohols derived from an organocatalytic Mannich reaction

Enantiomerically pure β-aminoalcohols, produced through an organocatalytic Mannich reaction, were subjected to an Ugi multicomponent reaction under classical or Lewis acid-promoted conditions with diastereoselectivities ranging from moderate to good. This approach represents a step-economical path to enantiomerically pure, polyfunctionalized peptidomimetics endowed with three stereogenic centers, allowing the introduction of five diversity inputs.

diastereoselective_ugi_reaction_of_chiral_1,3-aminoalcohols_derived_from_an_organocatalytic_mannich_

Findings<!><!>Findings<!><!>Findings<!><!>Findings<!><!>Findings<!>

<p>Isocyanide-based multicomponent reactions [1–3], such as the Ugi reaction, were demonstrated to be very useful in the rapid assembly of complex drug candidates [4], introducing three to four diversity inputs. Furthermore, a nearly limitless variety of heterocycles can be accessed through post-condensation transformations [5–7], adding only one to two steps to the synthetic sequence. However, the main drawback of the Ugi reaction is the poor stereochemical control that is typically achieved [8–9], which hampers its utilization in the diversity-oriented or target-oriented synthesis of complex chiral peptidomimetics. No efficient asymmetric catalytic classic Ugi reaction has been reported to date (whereas some success was obtained on simpler variants) [10–12]. On the other hand, diastereoselective reactions using at least one chiral component are troublesome. Chiral isocyanides and chiral carboxylic acids invariably afford nearly 1:1 mixtures. α-Chiral aldehydes have a high tendency to racemize/epimerize [13–14] and additionally, no report of valuable diastereocontrol by them has appeared so far. Successful examples of diastereoselective Ugi reactions have been reported only with chiral amines [15–19] or with chiral cyclic imines (Ugi–Joullié reaction) [20–23], although in the latter case, racemization/epimerization can again be an issue in special cases [24]. However, the use of amines as chiral auxiliaries has been seldom exploited in peptidomimetic synthesis [16,25–26] because the need to remove the auxiliary reduces the number of diversity inputs and increases the number of synthetic steps.</p><p>From the point of view of atom- and step-economy, the use of chiral amines that are retained in the final products will be more valuable [27]. In this case they are not "chiral auxiliaries" and are not removed after the multicomponent reaction, and they contribute to the diversity of the final products. However, the usefulness of this approach relies on an efficient and diversity-oriented preparation of the required amines in high enantiomeric excess.</p><p>Chiral aminoalcohols can be ideal substrates for diastereoselective Ugi reactions: the additional hydroxy group can both help in modulating diastereoselectivity and be employed for post-condensation transformations in order to add further fragments or to form heterocyclic structures. We have previously developed some syntheses of heterocycles through Ugi reactions with 1,2-aminoalcohols followed by nucleophilic substitutions [28], whereas chiral 1,2-aminoalcohols have been proved by Nenajdenko and co-workers to be able to induce good levels of diastereoselectivity in the Ugi reaction [17].</p><p>Our attention was drawn by 1,3-aminoalcohols of general formula 5 (Scheme 1), which can be obtained by List's organocatalytic Mannich-type reaction of aldehydes with N-Boc imines 2 and catalytic L-proline [29–30], followed by reduction of 3 and cleavage of the Boc group.</p><!><p>Overall strategy.</p><!><p>This short and straightforward synthesis allows the introduction of 2 diversity inputs (R1 and Ar), whereas stereochemical diversity can also be explored using D-proline or different, anti-selective organocatalysts.</p><p>We prepared two known carbamoyl sulfones 1 [30–31] and transformed them without isolation of intermediates into a series of five Boc-protected β-aminoalcohols 4a–e (Scheme 2). Using caesium carbonate, carbamoyl sulfones were converted into the corresponding N-Boc-protected imines 2 that were immediately submitted to List's organocatalytic Mannich reaction [29–30]. The resulting aldehydes 3 were not isolated (also in view of their known stereochemical lability) but directly reduced to alcohols 4 [32–33]. Purification was carried out through chromatography and, in some cases, by additional crystallization, affording these key intermediates in high ee and de (syn relative configuration, see Supporting Information File 1).</p><!><p>Boc-protected aminoalcohols used as inputs in a diastereoselective Ugi reaction.</p><!><p>The tert-butyl urethane was then deblocked with trifluoroacetic acid. Neutralization and extraction afforded crude aminoalcohols 5a–e, that were directly employed in the Ugi reaction. We first optimized this step using isobutyraldehyde, 5-chloro-2-thiophenecarboxylic acid and cyclohexyl isocyanide, to give the two diastereomers of compound 6a (Table 1).</p><!><p>Optimization of the synthesis of 6a.</p><p>aOverall yield from aminoalcohol. bRelative configuration not yet determined.</p><!><p>When the reaction was carried out under the classical conditions (using methanol as the solvent), only a moderate diastereoselectivity was achieved (Table 1, entry 1), which could be increased by changing the solvent to trifluoroethanol, especially effective at 0 °C. Considering the recent work by Nenajdenko et al. [17], we explored the usage of Lewis acids in an aprotic solvent in order to further improve the diastereoselectivity. We had anticipated that the binding of the Lewis acid to the free alcohol, followed by intramolecular activation of the aldehyde, would establish a cyclic transition state, thereby enabling better stereocontrol. It is indeed well-known that the Ugi reaction does not proceed in aprotic solvents such as THF at low temperature, and therefore the background, uncatalyzed reaction should not interfere. As shown in Table 1, the best results were achieved by using 1 equiv of zinc bromide (Table 1, entry 8), affording a 10:1 diastereomeric ratio and an excellent overall yield. Other zinc-based catalysts were less efficient, whereas most of the other tested Lewis acids failed to promote the reaction at all. The use of Lewis acids in methanol or trifluoroethanol afforded lower yields with no improvement of diastereoselection. It is worth noting that a 10:1 diastereoselectivity is considered excellent for isocyanide-based multicomponent reactions, due to the very low steric biases of isocyanides.</p><p>We then moved on to establish the scope of the method, varying the Boc-protected aminoalcohol, the carboxylic acid and the isocyanide (see Table 2). For a comparison, we performed all Ugi reactions either under Lewis acid-promoted conditions, or under the classical Ugi conditions (MeOH, rt). The stereochemical results were found to vary remarkably from case to case. While in some instances (products 6b–d) the activation with ZnBr2 brought about an increase of diastereoselectivity, in other combinations of substrates, the outcome was similar (products 6h and 6j) or even better using the "classical" conditions (products 6e, 6f, 6i). However, in all cases, the two diastereomers could be easily separated and the ratio was typically, with few exceptions, around 3:1 to 5:1. As far as the isolated yields were concerned, the Lewis acid-promoted reaction is typically less efficient, especially with aromatic isocyanides or aldehydes (compounds 6e, 6f, 6g). The relative configuration of the major adduct has not yet been unambiguously determined. However, TLC, HPLC, polarimetric and NMR analogies suggest that the main diastereomer was always the same, with one notable exception: product 6f obtained in the absence of Lewis acid. In this case, it was necessary to carry out the reaction in THF/iPrOH because the isocyanide was poorly soluble in MeOH, and thus the unexpected diastereoselectivity inversion might be due to the different solvent and not to the structure of isocyanide.</p><!><p>Scope of the synthesis of Ugi adducts 6.</p><p>aOverall yield from Boc aminoalcohols 4. bRelative configuration not yet determined. A: THF, −38 °C, 1 equiv of ZnBr2; B: MeOH, 25 °C; C: iPrOH/THF 2:1, 25 °C. All reactions carried out for 48 h at 0.1 M concentration of aminoalcohol with 1.00 equiv of aminoalcohol 5, 1.05 equiv of aldehyde, 1.2 equiv of carboxylic acid and isocyanide and 100 mg of powdered 3 Å molecular sieves per mmol of aminoalcohol.</p><!><p>The synthetic route from carbamoyl sulfones 1 to peptidomimetics 6 is quite short: intermediate purification was carried out only at the level of the Boc-protected aminoalcohols 4 and of the final products 6. Thus, this method offers an operationally simple route to enantiomerically pure complex structures like 6, introducing up to five diversity inputs and controlling three stereogenic centers (also thanks to the final chromatography).</p><p>Compounds 6 are endowed with several functionalities that can be exploited for post-Ugi cyclization steps or as a handle for attaching further fragments: the primary alcohol and the secondary amide (which are present in all products), a protected phenol (for compounds 6c, 6i, 6j), and a protected amine (6j). Studies towards this goal are in progress and will be reported in due course.</p><!><p>General remarks, experimental procedures and characterization data; 1H and 13C NMR spectra of new compounds 4 and 6 (major isomer only).</p>

PubMed Open Access

The missing link: allostery and catalysis in the anti-viral protein SAMHD1

Vertebrate protein SAMHD1 (sterile-α-motif and HD domain containing protein 1) regulates the cellular dNTP (2′-deoxynucleoside-5′-triphosphate) pool by catalysing the hydrolysis of dNTP into 2′-deoxynucleoside and triphosphate products. As an important regulator of cell proliferation and a key player in dNTP homeostasis, mutations to SAMHD1 are implicated in hypermutated cancers, and germline mutations are associated with Chronic Lymphocytic Leukaemia and the inflammatory disorder Aicardi–Goutières Syndrome. By limiting the supply of dNTPs for viral DNA synthesis, SAMHD1 also restricts the replication of several retroviruses, such as HIV-1, and some DNA viruses in dendritic and myeloid lineage cells and resting T-cells. SAMHD1 activity is regulated throughout the cell cycle, both at the level of protein expression and post-translationally, through phosphorylation. In addition, allosteric regulation further fine-tunes the catalytic activity of SAMHD1, with a nucleotide-activated homotetramer as the catalytically active form of the protein. In cells, GTP and dATP are the likely physiological activators of two adjacent allosteric sites, AL1 (GTP) and AL2 (dATP), that bridge monomer–monomer interfaces to stabilise the protein homotetramer. This review summarises the extensive X-ray crystallographic, biophysical and molecular dynamics experiments that have elucidated important features of allosteric regulation in SAMHD1. We present a comprehensive mechanism detailing the structural and protein dynamics components of the allosteric coupling between nucleotide-induced tetramerization and the catalysis of dNTP hydrolysis by SAMHD1.

the_missing_link:_allostery_and_catalysis_in_the_anti-viral_protein_samhd1

SAMHD1 structure and catalytic activity.<!>Introduction<!>Regulation of cellular dNTP levels<!>Allostery<!>SAMHD1 domain organisation<!>Nucleotide-dependent tetramerization of SAMHD1 HD domain.<!>Nucleotide-dependent tetramer assembly<!>SAMHD1 tetramerization is essential for catalysis<!>Long-lived, activated state of SAMHD1 corresponds to the homotetramer<!>Thr592 phosphorylation destabilises the tetramer<!>Thr592 phosphorylation modulates SAMHD1 tetramer stability.<!>Thr592 phosphorylation destabilises the tetramer<!>SAMHD1 catalytic site accommodates dNTPs and dNTP analogues.<!>Catalytic site selectivity<!>dNTP geometry in the catalytic site<!>Catalytic mechanism<!>‘Open’ and ‘closed’ conformations of HD domain.<!>‘Open’ and ‘closed’ HD domain conformations<!>Linkage between allosteric and catalytic sites<!>Summary<!>

<p>(A) The dNTP triphosphohydrolysis reaction catalysed by SAMHD1, dGTP is shown as the example substrate. (B) Domain organisation of human SAMHD1, showing the nuclear localisation signal in blue, the SAM domain in dark orange, the HD domain in light orange, the HD motif residues in maroon and phosphorylated residue Thr592 in teal. (C) Left: NMR structure of the human SAMHD1 SAM domain (PDB code: 2E8O residues 34–113). Right: X-ray crystal structure of WT human SAMHD1 HD domain (PDB code: 4BZC, monomer A, residues 115–599) [85]. The structures of the two domains are connected by a short, dotted and grey line. The HD motif-co-ordinated manganese ion is shown as a purple sphere, and the HD motif residues are shown as maroon sticks. (D) A close-up view of the HD motif-co-ordinated manganese ion in the catalytic site. PyMOL was used to prepare the structural figures.</p><!><p>Tight regulation of the cellular dNTP concentration, and of the balance among the individual dNTPs (dATP, dGTP, dCTP and TTP), is required to prevent genomic instability and tumourigenesis [20,21]. SAMHD1 mutations have been shown to alter cellular dNTP concentrations [22,23] and are associated with the development of certain cancers [23–29] (reviewed comprehensively by Mauney and Hollis [30]). Furthermore, SAMHD1 catalytic activity modulates the efficacy of anti-cancer and anti-viral nucleoside analogues [31–40].</p><p>In addition to its dNTP triphosphohydrolase activity, it was reported that SAMHD1 possesses nuclease activity against DNA and RNA substrates [41–45]. However, the observed nuclease activity could not be replicated in other studies [46–48] and has been associated with a contaminant co-purified with recombinantly expressed SAMHD1 [46]. Despite lacking intrinsic nuclease activity, SAMHD1 binds nucleic acids [41,46,48–51], with selectivity for single-stranded over double-stranded oligonucleotides [46,50], and for RNA over DNA [46,49].</p><p>SAMHD1 may utilise dNTP hydrolase activity, nucleic acid-binding or a scaffolding activity in other cellular roles such as in DNA replication [52], DNA repair [53] and inhibiting LINE-1 retrotransposition [54–57]. Germline mutation to SAMHD1 or any of six other genes (TREX1, RNaseH2A, RNaseH2B and RNaseH2C, ADAR1 and IFIH1) can cause the auto-inflammatory condition Aicardi–Goutières Syndrome [30,58–65]. More broadly, SAMHD1 appears to repress immune responses to viral infection and other inflammatory stimuli [52,60,66–68]. However, these mechanisms remain to be elucidated.</p><!><p>The dNTP hydrolase activity of SAMHD1 is essential for both cellular dNTP homeostasis and regulating cell proliferation [24,29,69]. While SAMHD1 catalyses dNTP hydrolysis to reduce the cellular dNTP pool, several enzymes act antagonistically to increase the dNTP pool by catalysing dNTP synthesis either de novo or via salvage pathways. dNTPs are continually synthesised and degraded throughout the cell cycle, with the highest rates of dNTP flux occurring in S-phase [70]. dNTP levels in mammalian cells are ∼10- to 18-fold higher in S-phase than G0/G1 [70–72], and dNTP synthesis must continue during S-phase to complete chromosomal replication [72–75].</p><p>The catalytic activity of SAMHD1 is tightly controlled throughout the cell cycle through a mechanism of phosphorylation and dephosphorylation [3]. In S-phase, SAMHD1 appears to be phosphorylated at residue Thr592 by cyclin-dependent kinase 1 or 2 and cyclin A (CDK1/2-cyclinA) to lower the rate of dNTP hydrolysis [76–79]. At the end of M-phase, SAMHD1 catalytic activity is recovered due to dephosphorylation by phosphatase PP2A-B55α [80]. SAMHD1 expression levels may also vary throughout the cell cycle to further regulate dNTP hydrolase catalytic activity [3,4,30].</p><!><p>In addition to post-translational regulation, SAMHD1 is subject to allosteric regulation by nucleotides to fine-tune its catalytic activity. Allosteric regulation occurs when the binding of a ligand at one site affects the affinity of ligand binding or catalysis at a second site in the same protein. Allosteric effects can be positive, termed 'allosteric activation', or negative, 'allosteric inhibition'. Allostery is observed in many proteins, from single-domains to large multimeric complexes [81–84]. SAMHD1 is allosterically activated by nucleotide binding in two allosteric sites, AL1 and AL2 [5,6,85–88]. However, there is little evidence to suggest allosteric site coordination modifies catalytic site selectivity in SAMHD1 [87]. This review focuses on the mechanism by which allostery regulates human SAMHD1 catalysis and so the experiments described refer to human SAMHD1 studies, unless explicitly stated otherwise.</p><!><p>Human SAMHD1 is a 626-residue protein that contains an N-terminal nuclear localisation signal, 11KRPR14 [89], a sterile-α-motif (SAM) domain (residues 34–113) and an HD catalytic domain (residues 115–599; Figure 1B–D). The SAM domain is an α-helical domain with an unknown function in SAMHD1 [90,91]. The HD catalytic domain is a phosphohydrolase domain that is named after the two pairs of Histidine–Aspartate residues (His167, His206, Asp207 and Asp311) that co-ordinate a metal ion at the catalytic site (Figure 1D) [92]. The HD domain of SAMHD1 is itself sufficient to catalyse the hydrolysis of dNTPs into 2′-deoxynucleoside and triphosphate [3,5,6,51,85,86,93].</p><p>Structural studies on human SAMHD1 have primarily focussed on the HD catalytic domain, as the flexibility of the linker connecting the SAM and HD domains has hindered structural studies on full-length human SAMHD1. More recently, X-ray crystal structures have been determined for mouse SAMHD1 containing both SAM and HD domains [94], providing some insight into how the human SAM and HD domains may interact with one another.</p><!><p>(A) Ordered pathway for tetramer assembly, proposed by Hansen et al. [97], and elucidated structurally using X-ray crystallography [77,78,85–88,94,95,101]. The four protein monomers of SAMHD1 are shown in light orange, grey, maroon and light green. From left to right: The apo SAMHD1 HD domain is in a monomer–dimer equilibrium with GTP (dark blue), which stabilises dimerisation by coordinating AL1 (PDB code: 4RXO) [95]. dATP (light blue) co-ordinates AL2, and a magnesium ion (green sphere) bridges adjacent AL1–AL2 allosteric sites to stabilise the HD domain tetramer (PDB code: 4RXP) [95]. dCTP in the catalytic site is shown in pink. (B) GTP–dATP coordination at adjacent allosteric sites, AL1–AL2, of an HD domain tetramer (PDB code: 4TO0) [87]. GTP in AL1 and dATP in AL2 are shown in stick representation, with Cα atoms shown in yellow. Hydrogen-bonding interactions are represented by dashed lines.</p><!><p>GTP is the likely physiological activator of AL1 [87,97–99], as the cellular concentration of GTP is greater than that of the other guanine-based nucleotides [100], and, unlike dGTP, GTP is not a substrate of SAMHD1 [5]. In AL1, the guanine base of a co-ordinated nucleotide is recognised through a hydrogen-bonding network involving residues Asp137, Gln142 and Arg145 [85,87,95]. AL1 has a preference for GTP ≥ dGTP > ddGTP [77,97,98], due to hydrogen bonds formed between the GTP ribose and the backbone carbonyl of Val117 and the adjacent AL2-co-ordinated dNTP [87,88,95]. The preference in AL1 for GTP > GDP >> GMP [98] is due to extensive salt-bridges formed between the GTP triphosphate, Lys116 of one SAMHD1 monomer, and Arg451 and Lys455 of a second monomer [85,86], explaining how AL1-coordination stabilises dimerisation.</p><p>In AL2, the bulky side chains of Val156 and Phe157 prevent binding by nucleotides functionalised at the 2′-ribose position, such as ribonucleoside-5′-triphosphates (NTPs), although fluoro-substitution at the 2′-proS position is tolerated [38,87,101]. AL2 selects for dNTPs over 2′,3′-dideoxynucleoside triphosphates (ddNTPs) due to two hydrogen bonds formed between the 3′-hydroxyl of a dNTP and the protein backbone of Asn119 and Val156 [85,86]. Salt-bridges between the dNTP triphosphate and residues Arg333, Arg352 and Lys354 further stabilise nucleotide coordination in AL2. Additionally, Lys523 forms salt-bridges with the γ-phosphates of each AL1–AL2 pair to select for triphosphorylated nucleotides in AL1 and AL2.</p><p>All four canonical dNTPs can co-ordinate AL2, with a preference for dATP > dGTP > TTP > dCTP [6,87,88]. The polar side chains of residues Asn119 and Asn358 and several water molecules adapt their hydrogen-bonding network to accommodate all four bases in AL2 [87]. The preference in AL2 for the purine nucleotides dATP and dGTP, over the pyrimidines TTP and dCTP, is due to more extensive cation-π stacking of the larger purine bases with the guanidino side chain of Arg333 [87]. Ji et al. [87] proposed that dATP is the primary activator of AL2, as they observed a stronger salt-bridge formed between the side chains of Arg333 and Glu355 only when dATP is bound in AL2. In contrast, dCTP is a poor AL2-activator of SAMHD1 [88], likely due to the inability of the cytosine base to form a direct hydrogen bond with Asn358.</p><!><p>Extensive experimental evidence supports the hypothesis that SAMHD1 tetramerization is essential for catalysing dNTP hydrolysis [85,86,88,93,97]. Firstly, dGTP, which can co-ordinate AL1, AL2 and the catalytic site, is hydrolysed by SAMHD1 in vitro in the absence of additional nucleotides [5,6,86,93], whereas the three other canonical dNTPs (dATP, dCTP and TTP) are only hydrolysed by SAMHD1 in the presence of AL1-activating GTP or dGTP [5,6,98]. Secondly, point mutations to key residues in AL1 (D137A, Q142A, R145A and R451E), AL2 (R333E) and the dimer–dimer interface (D361K, H364K and R372D) reduced SAMHD1 tetramerization and dNTP hydrolysis in vitro [77,85,86,88,93,95,102]. Finally, N- and C-terminal truncations in constructs 120–626 (Δ115–119) and 115–583 (Δ584–626), respectively, abolished tetramerization and reduced catalytic activity in comparison with constructs 1–626 and 115–626 [5,51,77], demonstrating the importance of tetramerization for catalysis.</p><!><p>Hansen et al. [97] observed that GTP and dNTPs, or dGTP alone, generate a long-lived, activated state of SAMHD1 that corresponds to the SAMHD1 homotetramer. Furthermore, the activated, homotetrameric state of SAMHD1 is not in equilibrium with free GTP or dNTP activators in solution. Strikingly, the SAMHD1 homotetramer persisted in vitro for hours without further exchange of nucleotides in AL1 or AL2 [77,97]. It is proposed that this slow rate of tetramer dissociation, despite activator depletion, enables SAMHD1 to deplete cellular dNTP concentrations to the nanomolar concentrations observed in macrophages and resting CD4+ T-cells.</p><!><p>Phosphorylation of human SAMHD1 residue Thr592 by CDK1/2-cyclinA regulates catalytic activity throughout the cell cycle [76–78]. Phosphorylation of residue Thr592 in vitro or introducing the phosphomimetic mutation T592E reduced tetramerization and catalytic activity [77,78]. The mutation T592E also eliminated the ability of SAMHD1 to restrict HIV-1 infection in macrophage-like PMA-differentiated U937 cells [77]. Similarly, another phosphomimetic mutation, T592D, also impaired the ability of SAMHD1 to block the lytic replication of the Epstein–Barr herpesvirus in producer Akata cells [103].</p><!><p>(A) X-ray crystal structure of SAMHD1 HD domain tetramer (PDB code: 4BZC) [85]. Monomer A is shown in cartoon representation in light orange. Monomers B–D are shown in surface representation in grey. Residues 559–599 are in pink and residues 522–537 are in blue. (B) C-terminal, the α-helical region between residues 559–599, comprising phosphorylated residue Thr592.</p><!><p>Molecular dynamics (MD) simulations by Patra et al. [104] showed that mutation T592E caused minor local perturbations to residues 585–595, but did not affect the integrity of the allosteric or catalytic sites on the timescale modelled. Further analysis of correlated motions across the SAMHD1 tetramer in the MD simulations revealed that the mutation T592E decoupled a signalling pathway between residue Thr592 and the allosteric sites, and increased the dynamic coupling between Thr592 and α-helix 13 (α13; residues 352–375) at the dimer–dimer interface [105]. The authors concluded that phosphorylation of Thr592 may trigger a loosening of the HD domain tetramer.</p><!><p>(A–F) Nucleotide coordination in the catalytic site of the human SAMHD1 HD domain. (A) WT SAMHD1 with dGTPαS (PDB: 4BZC) [85]. (B) H206R/D207N SAMHD1 with dATP (PDB: 4QG1) [88]. (C) H206R/D207N SAMHD1 with TTP (PDB: 4TNZ) [87]. (D) WT SAMHD1 with dCTP (PDB: 4RXR) [95]. (E) WT SAMHD1 with ddGTP (PDB: 5AO1) [77]. (F) H206R/D207N SAMHD1 with gemcitabine-TP (Gem-TP) (PDB: 6DW5) [101]. Nucleotides are shown in stick representation with Cα atoms in pink. Manganese (purple), magnesium (green) and iron (brown) metal ions are represented as spheres.</p><!><p>Base selectivity in the catalytic site is achieved through subtle differences in the hydrogen bonding between a dNTP base, its network of hydrating water molecules and residues Leu150, Tyr374, Gln375, Asn380 and Asp383, which line the catalytic pocket [87,88]. The side chains of residues Leu150, Tyr315 and Tyr374 form a tight-binding pocket around the base and 2′-deoxyribose moieties of a dNTP substrate [38,85,101]. The 3′-hydroxyl group on a dNTP substrate is hydrogen bonded by the polar side chains of residues Gln149 and Asp319 [86,88]. Nucleotide binding in the catalytic site is further stabilised by salt-bridges between the triphosphate and the basic side chains of Arg164, Lys312 and Arg366 [85,88].</p><p>In addition to canonical dNTPs and dUTP, the SAMHD1 catalytic site can co-ordinate and hydrolyse particular dNTP analogues. The poorly hydrolysed analogue ddGTP co-ordinates the SAMHD1 catalytic site, as revealed through crystal structures (Figure 4E) [77]. Knecht et al. [101] solved co-crystal structures in which the catalytic site was occupied by the triphosphorylated forms of anti-cancer drugs cladribine, clofarabine, fludarabine, cytarabine and gemcitabine (Figure 4F), and the anti-viral agent vidarabine. The authors' structural and biophysical studies revealed that the catalytic site could tolerate fluoro- and chloro-substitutions at the carbon-2 position on an adenine base, fluoro- and hydroxyl-substitutions at the 2′-proS ribose position, and a fluoro-group at the 2′-proR position. Such substitutions at the 2′-proS position are tolerated by a compensatory rotation of the ribose moiety of these analogues within the catalytic pocket. Leu150 and Tyr374 side chains prevent bulkier functionalisation at the 2′-proR position, while Tyr315 prevents functionalisation to the 3′-proS position [38,85,101].</p><!><p>Numerous crystal structures have been solved of the SAMHD1 HD domain with dNTPs or dNTP analogues in the catalytic site [77,78,85–88,94,95,101]. Frequently, the inactivating double mutation H206R/D207N has been employed in these studies [78,85,87,88,101]. The H206R/D207N mutation to the HD motif (His167, His206, Asp207 and Asp311) prevents coordination of a metal ion at the HD motif and eliminates catalytic activity in SAMHD1 [85,93]. Metal ion coordination at the HD motif is likely important for catalysis, as a further HD motif mutant, D311A, is also catalytically inactive [5,44,106].</p><p>While it is possible that dNTP or dNTP analogue coordination may be perturbed in crystal structures of SAMHD1 mutant H206R/D207N (Figure 4B,C,F), a similar binding mode is observed for the analogue dGTPαS in a wild-type (WT) catalytic site (Figure 4A) [85]. The consensus between independently reported H206R/D207N-dNTP and WT-dGTPαS structures (Figure 4A–C) [85,87,88] suggests there may be a physiological basis for this nucleotide-binding mode in the catalytic site. Therefore, it could be postulated that these non-catalytically competent SAMHD1-nucleotide structures represent enzyme–substrate complexes prior to catalysis.</p><p>In comparison, a different triphosphate geometry is modelled in the crystal structures of catalytically competent WT–dNTP complexes (Figure 4D) [86,95]. The base and 2′-deoxyribose portions of the dNTP ligands superimpose with those of non-catalytically competent H206R/D207N-dNTP structures. However, the triphosphate moiety is modelled in different configurations. Thus, the WT–dNTP structures may represent intermediate- or product-like states during catalysis.</p><p>Structures of WT SAMHD1 with the poorly hydrolysed analogue ddGTP reveal a further binding mode for the nucleotide in the catalytic site [77], with ddGTP less well buried within the catalytic pocket (Figure 4E). The WT–ddGTP crystal structures reveal a unique substrate-binding mode that may be required for ddGTP hydrolysis, but importantly could represent a nucleotide-bound state along the dNTP substrate-binding pathway of SAMHD1. Further SAMHD1-nucleotide structural studies may be required to elucidate nucleotide-binding modes at various stages of catalysis, including substrate binding, hydrolysis and product release.</p><!><p>The chemical reaction catalysed by SAMHD1 was initially identified through chromatography-based experiments in which dNTP substrates were demonstrated to be hydrolysed directly into 2′-deoxynucleoside and triphosphate products (Figure 1A), rather than by sequential monophosphate cleavages via 2′-deoxynucleoside-5′-diphosphate (dNDP) and 2′-deoxynucleoside-5′-monophosphate (dNMP) intermediates [5,6]. The catalytic mechanism was further investigated using mass spectrometry experiments that determined oxygen from bulk water is incorporated into the triphosphate product, rather than the 2′-deoxynucleoside product, supporting a mechanism of nucleophilic attack on the α-phosphorous that results in cleavage of the α-phosphorous-to-5′-oxygen covalent bond [107].</p><p>In addition to residues in the HD motif, residues His210, Asp218 and His233 have been proposed to be important for catalysis, based on the observation that mutations H210A and H233A disrupt catalysis [88], and on the conservation of these three residues across HD phosphohydrolase domains, including in the homologous protein EF1143 from the bacterium Enterococcus faecalis [6,85,108]. Furthermore, a crystal structure of mutant H210A was found to lack nucleotide coordination in the catalytic site, supporting a function for residue His210 in substrate dNTP coordination [88].</p><!><p>(A) X-ray crystal structure of the HD domain in 'open' conformation co-ordinated to GTP (dark blue) in AL1 (PDB: 4RXO) [95]. The tetramer comprises two dimers, with one formed from the orange and grey monomers, and the second from the maroon and light green monomers. (B) X-ray crystal structure of HD domain in 'closed' tetrameric conformation, with dGTP co-ordinated in AL1, AL2 and the catalytic site (PDB: 4BZB) [85]. dGTP is shown in dark blue (AL1), pale blue (AL2) or pink (catalytic site) and magnesium ions are shown as green spheres. (C) Superposition of an HD domain monomer in 'open' (grey) and 'closed' (orange) conformations. Residues 326–375 and 454–599 differ in their conformation between the 'open' and 'closed' states and are represented by darker shades of grey and orange, respectively. The dGTP- and magnesium-coordination observed in the catalytic site of the 'closed' state is displayed, while the manganese- and phosphate-coordination observed in the 'open' state is not shown. (D) Interactions formed by the α13 helix at the dimer–dimer interface (residues Asn358, Asp361, His364 and Arg372), in the catalytic site (Arg366, Tyr374 and Gln375) and in AL2 (Arg352, Lys354 and Asn358). One monomer is shown in pale orange, and its neighbouring monomer at the dimer–dimer interface is shown in maroon. dGTP in AL1 (dark blue), AL2 (pale blue) and the catalytic site (pink) is shown in stick representation. Magnesium ions are represented by green spheres. Hydrogen bonds are shown as dashed lines.</p><!><p>SAMHD1 HD domain crystal structures with nucleotides simultaneously bound in AL1, AL2 and the catalytic site adopt a so-called 'closed' conformation (Figure 5B) that is more compact about the catalytic site, and ordered to a greater extent, with density observed for the HD domain backbone for all residues between positions 115–599, except for a short loop between residues 278–283 [85]. In 'closed' structures, the HD domains assemble in the crystal lattice into homotetramers that contain D2 dihedral symmetry, whereby the four monomers are related to one another by three 2-fold symmetry axes.</p><p>Structural comparisons suggest that the HD domain must undergo a change in conformation during the dimer-to-tetramer transition to accommodate dNTPs into AL2 and the catalytic site. Secondary structural elements are conserved between dimeric ('open') and tetrameric ('closed') conformations of the SAMHD1 HD domain. However, between dimeric and tetrameric structures, motions of up to ∼5 Å affect tertiary packing in two regions of the protein, between residues 326–375, and 454–599 (Figure 5C) [85,86]. Several residues in these two regions are important for nucleotide coordination in AL1, AL2 and the catalytic site. Therefore, it is likely that these regions have important functions in allosteric regulation in SAMHD1.</p><!><p>As described above, the HD domain conformation varies between dimeric (apo or AL1-occupied) and tetrameric (AL1-, AL2- and catalytic site-occupied) states of SAMHD1. While the majority of residues across the catalytic site do not appear to be significantly structurally perturbed during tetramer assembly, tertiary structural changes alter the positioning of catalytic site residues Arg366 and Gln375 (Figure 5C), which lie on one face of α13 [85,86]. Residues Arg366 and Gln375 are involved in dNTP coordination in the catalytic site of closed, tetrameric SAMHD1 structures, but appear too distal for substrate coordination in open, dimeric structures that lack dNTPs in AL2 and the catalytic site.</p><p>Helix α13, which spans residues 352–375, bridges the catalytic and allosteric sites, and makes important interactions at the dimer–dimer interface (Figure 5D) [86]. Catalytic site residues Arg366 and Gln375 are at the C-terminal end of α13, while residues Arg352, Lys354 and Asn358 at the N-terminal end of α13 are involved in dNTP coordination in AL2. At the dimer–dimer interface, the α13 helix of one monomer interacts with the neighbouring monomer's helix, α13', through a network of salt-bridges and hydrogen bonds involving residues Asn358, Asp361, His364 and Arg372 from both α13 and α13' elements. Thus, α13 appears to be crucial for allosteric regulation, by communicating allosteric site occupancy and tetramerization to the catalytic site, with residues Arg366 and Gln375 supporting substrate dNTP binding once AL2 is occupied and the protein has tetramerized.</p><p>In addition to structural changes in the catalytic site, HD domain tetramerization likely alters protein dynamics. Patra et al. [104,105] explored mechanisms for cross-talk between allosteric and catalytic sites in SAMHD1 using correlation analysis of MD simulations. The authors observed that correlated motions between allosteric and catalytic sites were reciprocated across the HD domain tetramer, revealing both short-range and long-range allosteric signal transduction in SAMHD1. Furthermore, removing dATP from one AL2 site in the SAMHD1 HD domain tetramer significantly reduced the rigidity of the protein around the dATP-occupied catalytic site. In separate MD simulations, Cardamone et al. [109] observed that removing all nucleotides and magnesium ions from the protein tetramer weakened α13–α13' interactions at the dimer–dimer interface, and the AL1 mutation R145E accelerated the destabilisation of the tetramer [110]. Overall, biophysical and computational experiments demonstrate that interactions at the dimer–dimer interface and nucleotide occupancy at the allosteric sites modulate the catalytic site structure and dynamics in order to regulate catalysis.</p><!><p>Allosteric site occupancy and HD domain tetramerization control both the structural integrity and the rigidity of the SAMHD1 catalytic site [85,86,105,109,110]. In the absence of nucleotide coordination in AL1, AL2 and the catalytic site, SAMHD1 exists in a monomer–dimer equilibrium [93,97]. GTP or dGTP binding in AL1 increases the proportion of dimeric SAMHD1 [97] and is necessary for subsequent dNTP binding in AL2 [97]. Changes in tertiary structure and protein dynamics result from the dNTP-induced dimer-to-tetramer transition, including structural perturbations to residues 326–375 and 454–599 [85,86], and changes in the dynamics of catalytic site residues, including His206, Tyr374 and Gln375 [104]. The catalytic site becomes more rigid upon nucleotide-induced tetramerization [104,105], and there appears to be an energetic coupling between nucleotide binding in AL2 and the catalytic site [97]. Subsequent to catalysis, the reaction products, 2′-deoxynucleoside and triphosphate, dissociate from SAMHD1 and the catalytic site of a closed tetramer appears sufficiently accessible for nucleotide exchange to occur without tetramer disassembly. Kinetic experiments demonstrate that the AL1- and AL2-co-ordinated tetramer is a long-lived, activated state, in which the AL1- and AL2-co-ordinated nucleotides are not in exchange with free nucleotides [97]. This is relevant to a cellular environment in which the dNTP pool has been largely depleted. Stable, active SAMHD1 tetramers persist [77,97] and hydrolyse dNTPs to drive the cellular dNTP pool to nanomolar concentrations that are observed in resting cells and are required for the restriction of HIV-1 replication.</p><!><p>SAMHD1 has important anti-viral, anti-cancer and anti-inflammation functions in the cell. SAMHD1 restricts HIV-1 replication in dendritic and myeloid lineage cells. Mutations to SAMHD1 have been identified in hypermutated cancers, and germline mutation to SAMHD1 can cause Chronic Lymphocytic Leukaemia and auto-immune condition Aicardi–Goutières Syndrome.</p><p>The dNTP triphosphohydrolase catalytic function of SAMHD1 is essential for HIV-1 restriction but also for cellular dNTP homeostasis. The catalytic domain of SAMHD1 tetramerizes in a nucleotide-dependent manner, with GTP and dATP coordinating two allosteric sites (AL1 and AL2) per SAMHD1 monomer to stimulate dNTP hydrolysis in the catalytic site. This review combines the results of biophysical, structural and MD studies to present a unified mechanism for the allosteric regulation of catalysis by SAMHD1.</p><p>X-ray crystallographic studies have revealed how nucleotide coordination in AL1 and AL2 stabilises catalytic domain tetramerization. However, it remains unclear how a substrate dNTP is co-ordinated in the WT catalytic site of SAMHD1 prior to catalysis and how SAMHD1 catalyses dNTP triphosphohydrolysis. Further studies are required to elucidate the catalytic mechanism of SAMHD1.</p>

PubMed Open Access

Catalytic Kinetic Resolution of a Dynamic Racemate: Highly Stereoselective \xce\xb2-Lactone Formation by N-Heterocyclic Carbene Catalysis

This study describes the combined experimental and computational elucidation of the mechanism and origins of stereoselectivities in the NHC-catalyzed dynamic kinetic resolution (DKR) of \xce\xb1-substituted-\xce\xb2-ketoesters. Density functional theory computations reveal that the NHC-catalyzed DKR proceeds by two mechanisms, depending on the stereochemistry around the forming bond: 1) a concerted, asynchronous formal (2+2) aldol-lactonization process, or 2) a stepwise spiro-lactonization mechanism where the alkoxide is trapped by the NHC-catalyst. These mechanisms contrast significantly from mechanisms found and postulated in other related transformations. Conjugative stabilization of the electrophile and non-classical hydrogen bonds are key in controlling the stereoselectivity. This reaction constitutes an interesting class of DKRs in which the catalyst is responsible for the kinetic resolution to selectively and irreversibly capture an enantiomer of a substrate undergoing rapid racemization with the help of an exogenous base.

catalytic_kinetic_resolution_of_a_dynamic_racemate:_highly_stereoselective_\xce\xb2-lactone_formatio

Introduction<!>Computational Methods<!>Results and Discussion<!>Conclusion

<p>β-Lactones are highly useful building blocks for the synthesis of target compounds, especially in the area of natural product synthesis.2–13 Catalytic asymmetric methods have provided new approaches to access this valuable strained ring system, and additional selective routes from different substrate classes open new synthetic possibilities.14–19 We recently disclosed the first NHC-catalyzed dynamic kinetic resolution (DKR) reaction that furnishes β-lactones and cyclopentenes in good yields with high stereoselectivities from racemic α-substituted-β-keto esters (eq. 4).20 Here, we report a collaborative computational study of the origins of stereoselectivities and the reaction mechanism. We have discovered how the degree of conjugation to an electrophile controls the stereoselectivity of this reaction, and how the stereochemical environment around the forming bond leads to a divergence in mechanism. In the process, we have also discovered that this reaction is part of an unusual class of DKRs in which the catalyst is responsible for the kinetic resolution that irreversibly traps an enantiomer of a dynamically racemizing substrate in a stereocontrolled manner.</p><p>The conversion of racemic starting materials to enantioenriched products is an ongoing goal in chemical synthesis with significant impact on the production of high value medicinal compounds.21–27 Dynamic kinetic resolutions (DKRs) are one particularly efficient and widely used approach to convert racemic substrates to stereochemically pure products with a theoretical yield of 100%.28–36 During the reaction sequence, a catalyst rapidly racemizes the substrate and stereospecifically transforms one enantiomer of the substrate. The ongoing catalyst driven racemization driven by Le Chatelier's principle eventually leads to the accumulation of a stereochemically pure product. Substituted-β-ketoesters are the archetypal substrate for DKR reactions due to their configurational lability at the α-position (Scheme 1, eq. 1).37 Examples of DKRs with α-substituted-β-ketoesters include several asymmetric hydrogenations (eq. 2),38,39 and a Baeyer-Villiger oxidation (eq. 3).40 In 2007, we reported the NHC-catalyzed desymmetrization of 1,3-diketones, a kinetic resolution process.41 In this process, the chiral NHC-generated enol undergoes selective addition to one of the two ketones, to allow for the formation of enantioenriched lactones and cyclopentenes. Our 2012 report, the title reaction (Scheme 2), is an expansion of this reaction to a dynamic kinetic resolution process.</p><p>N-Heterocyclic carbenes (NHCs) have greatly advanced the fields of organic and inorganic chemistry as ligands42–49 and as catalysts.50–61 These unique Lewis bases have been used to generate acyl anion,62–76 homoenolate,77–94 and enolate equivalents,41,95–105 as well as promote hydroacylation106–111 and an exciting variety of non–Umpolung processes.112–115 These carbene-catalyzed processes have been used to access numerous challenging compound classes with high levels of diastereo- and enantioselectivities. With all of the different reaction manifolds accessed through carbene catalysis, it is interesting to note that before our 2012 report, there had been no previous examples in the literature of NHCs facilitating a DKR.116</p><!><p>The mechanism and origins of stereoselectivity of this reaction were studied using M06-2X117/6-31+G**118,119/PCM(DCE)120//M06-2X/6-31G* as implemented in the Gaussian 09 suite of programs.121 This method has previously been shown by Sunoj to reproduce experimentally observed stereoselectivity in a related NHC process.122 Ethyl groups were modeled as methyl to reduce the degrees of freedom. Manual, exhaustive conformational searches were performed to ensure all relevant intermediates and transition structures were located. Intrinsic reaction coordinates (IRCs) were computed for all transition structures to verify reaction pathways.</p><!><p>Previous computational studies of NHC-catalyzed processes have elucidated the mechanisms, reactivities, and stereocontrol in various NHC-organocatalyzed processes.66,122–136 This study builds on these earlier reports and reveals not only an unusual method of stereocontrol, but also an unprecedented mechanism. The computed catalytic cycle and the reaction coordinate diagram are shown in Figure 1. The attack of the NHC catalyst on the ω-aldehyde, proton transfer, and two tautomerizations lead to the key enolate intermediate IV. The subsequent irreversible aldol cyclization diverges to two pathways, depending on the stereochemistry around the forming bond (Figure 1): 1) For the major (R,S,S)-product, a concerted asynchronous aldol lactonization pathway is operative. This is a formal (2+2) cyclization where the forming alkoxide simultaneously attacks the regenerating adjacent carbonyl, leading directly to the catalyst-lactone adduct VIII. 2) For all minor products, a stepwise spiro-lactonization mechanism is operative, where the aldol irreversibly leads to a spiro compound VIb, the collapse of which leads to the catalyst-lactone adduct VIII. In all cases, the facile dissociation of the NHC catalyst regenerates the catalyst and releases the product lactones IX.</p><p>The discovery of two divergent mechanisms for the aldol-lactonization step contrast to the originally proposed mechanism in two ways: 1) originally, a stepwise mechanism that involves the formation of the zwitterionic aldol adduct VIa was postulated (Figure 1). This is unusual considering how frequently it is invoked and found in related reactions, most recently in the elegant work by Paddon-Row and Lupton.127 Surprisingly, neither this adduct nor the subsequent transition state (VIIa) to form the catalyst-lactone adduct VIII exist on the potential energy surface. All our efforts to locate these structures have led to intermediate VIb and transition state VIIb, respectively. 2) Computations unequivocally reveal that the aldol-cyclization occurs via the enolate rather than the enol. In fact, the aldol-lactonization TS involving the enol does not exist. In line with these computational observations, we observed that Lewis acid and thiourea additives either significantly slowed, or completely shut down the reaction.137</p><p>The aldol transition states (TSs) for all observed products where R1 = Ph are shown in Figure 2. The TS-V-(R,S,S), which leads to the major product, is favored by 2.7 kcal/mol compared to minor enantiomer TS-V-(S,R,R). This compares favorably with experimental enantioselectivity of 2.9 kcal/mol. In contrast, the computed diastereoselectivity is overestimated by ∼1 kcal/mol compared to experiments – major TS-V-(R,S,S) is favored by 2.2 kcal/mol compared to the minor diastereomer TS-V-(S,S,S), while the experimental diastereoselectivity is 1.1 kcal/mol.</p><p>The energetic preference for the major TS-V-(R,S,S) stems from a double activation of the electrophilic ketone. First, there is significant electrostatic stabilization of the developing negative charge on the ketone undergoing nucleophilic attack by a critical C–H⋯O non-classical hydrogen bonding interaction138 from the catalyst pyranyl C–H (indicated by thin green lines, Figure 2). Moreover, the phenyl substituent of the electrophilic ketone is in conjugation with the carbonyl (–0.2°), maximizing the reactivity of the ketone.</p><p>The origin of diastereocontrol arises from the reduced reactivity of the electrophilic ketone in the minor diastereomer TS-V-(S,S,S). The epimer at the ester stereocenter changes the torsion around the forming C–C bond such that although the stabilizing C–H⋯O interaction is maintained, it forces the electrophilic ketone to be twisted out of conjugation (–36°) with the phenyl ring to avoid steric interactions with the catalyst.139</p><p>We have computed a model system to quantify the effect of conjugative electrophilic activation on transition state stabilities. Shown in Table 1 are the energetic penalty from the loss of conjugation in various substituted benzaldehydes by comparing the fully conjugated planar (0° dihedral between the carbonyl and the Ph) with the phenyl twisted out of conjugation to the same degree as found in the minor transition state (34° dihedral average, across TS-V-(S,S,S) involving substrates where R = H, F, and OMe). In the ground state, the loss of conjugation amounts to ∼2 kcal/mol destabilization, regardless of the identity of the phenyl substitution. However, in model transition states of hydride addition to the carbonyl where the hydride approach has been fixed at 2Å, conjugative stabilization was worth significantly more, ∼2-7 kcal/mol, depending on the electronic nature of the aryl substituent (entries 2 and 3, Table 1). This highlights a unique conjugative stabilization effect present only in the transition state but absent in the ground state that is strong enough to effect excellent stereocontrol.</p><p>The importance of this conjugative stabilization may explain why alkyl ketone substrates are not compatible under these reaction conditions.140 This NHC-catalyzed DKR proved to be general for α-substituted-β-ketoesters with aryl ketones. The decreased electron-withdrawing ability of alkyl and alkenyl substrates led to either no reaction or formation of side products.</p><p>The minor enantiomeric product is formed via TS-V-(S,R,R). The pyranyl non-classical hydrogen bonding C–H⋯O interaction controls the enantioselectivity. In the major TS, the pyranyl C–H is sandwiched between the enolate oxygen and the approaching electrophilic carbonyl, stabilizing the developing negative charges. However, approach to the opposite face of the enolate, as in the minor enantiomer pathway TS-V-(S,R,R), precludes stabilization with the pyranyl C–H. Instead, stabilization is realized by weaker alkyl C–H⋯O interactions. The weakness of these interactions is the cause for the destabilization of this transition state. A related investigation of electrostatic control of stereoselectivity in NHC-catalyzed [4+2] annulation reactions has been recently reported by Bode and Kozlowski.132</p><p>After we had computed the reaction coordinate, we questioned whether the NHC catalyst was indeed involved with the epimerization of the α-proton or was simply playing the role of a kinetic resolution catalyst. To test these two possibilities, we carried out a series of deuterium exchange experiments. α-Allylated β-keto ester 30 was dissolved in CD2Cl2/CD3OD mixture (0.07 M).141 Addition of cesium carbonate (30 mol %) as the base led to virtually instantaneous and complete deuterium incorporation at 23 °C as seen by 1H NMR spectroscopy (time = 0). The same experiment performed at −10 °C showed significant deuterium exchange after 5 minutes (Figure 3), and complete exchange after 30 minutes. These results demonstrate that the optimal basic conditions used in our DKR reaction promote extremely rapid keto/enol tautomerization and epimerization of these α-substituted β-keto esters.</p><p>An alternative possibility exists where the NHC azolium salts (pKa ∼17-25)142–152 itself drives the deprotonation.153 While experiments involving preparation of pre-generated carbene using strong bases (LDA or NaH) led to decomposition of starting materials, this possibility cannot be completely excluded.</p><p>We designed a stereodivergent reaction on a racemic mixture (RRM) experiment to verify that the aldol-lactonization process, not the epimerization is rate-limiting. In contrast to a standard kinetic resolution, a divergent RRM converts both enantiomers of a racemic mixture to non-enantiomeric products.154–156 We employed racemic α-disubstituted β-ketoesters, which are configurationally stable (Scheme 4). Complete cyclization to diastereomeric β-lactone products 29a and 29b was achieved in excellent yield (50% maximum theoretical yield for each) and enantioselectivity in 12 hours, considerably slower than the exogenous base-mediated epimerization of the substrate.</p><p>Together these results suggest that the basic conditions needed to generate the activated catalyst additionally promote substrate racemization at a faster rate than the overall DKR reaction process. We suspect that the NHC catalyst simply plays the role of a kinetic resolution catalyst that captures and irreversibly transforms one enantiomer of the substrate. The current DKR process is different from the classical DKR where the catalyst is involved with both the racemization and kinetic resolution; the racemization occurs tangentially to the kinetic resolution (Figure 4). The stark differences in the role of the catalyst prompt us to suggest differentiating these two DKRs. Such processes, in fact, though rarely reported, have been observed by others.157–159 We suspect that this unusual DKR is more common than is currently recognized.</p><!><p>In summary, computations have uncovered the mechanism and origins of stereoselectivity in the first NHC-catalyzed dynamic kinetic resolution of α-substituted-β-keto esters that provide β-lactones in high yields and selectivity. This study has uncovered two new mechanisms for the aldol-lactonization that challenge the currently accepted mechanism: 1) A concerted, asynchronous formal (2+2) aldol-lactonization process that leads to the major product, or 2) a stepwise spiro-lactonization mechanism that traps the forming enolate with the NHC catalyst iminium for all other products. The previously proposed stepwise mechanism, originally proposed by us, also invoked in other related reactions, is not operative.20 Additionally, we have uncovered how conjugative stabilization to the electrophile and C–H⋯O non-classical hydrogen bonds are key to stereocontrol. Finally, we have also discovered that that the current reaction exhibits an unusual DKR, one we coined as non-classical due to the atypical role of the catalyst. The combined experimental and theoretical efforts described in this manuscript have led to discoveries that refine and further distinguish the current understanding of carbene-catalyzed reactions and DKR processes. These advances will continue to be enhanced and employed towards the discovery and advancement of new processes.</p>

PubMed Author Manuscript

mTORC1 signaling promotes limb bud cell growth and chondrogenesis

mTORC1 signaling has been shown to promote limb skeletal growth through stimulation of protein synthesis in chondrocytes. However, potential roles of mTORC1 in prechondrogenic mesenchyme have not been explored. In this study, we first deleted Raptor, a unique and essential component of mTORC1, in prechondrogenic limb mesenchymal cells. Deletion of Raptor reduced the size of limb bud cells, resulting in overall diminution of the limb bud without affecting skeletal patterning. We then examined the potential role of mTORC1 in chondrogenic differentiation in vitro. Both pharmacological and genetic disruption of mTORC1 significantly suppressed the number and size of cartilage nodules in micromass cultures of limb bud mesenchymal cells. Similarly, inhibition of mTORC1 signaling in chondrogenic ATDC5 cells greatly impaired cartilage nodule formation, and decreased the expression of the master transcriptional factor Sox9, along with the cartilage matrix genes Acan and Col2a1. Thus, we have identified an important role for mTORC1 signaling in promoting limb mesenchymal cell growth and chondrogenesis during embryonic development.

mtorc1_signaling_promotes_limb_bud_cell_growth_and_chondrogenesis

Introduction<!>Mouse strains<!>H&E staining and whole-mount in situ hybridization<!>Micromass culture of limb bud mesenchymal cells<!>ATDC5 cell culture<!>Flow cytometry to measure relative cell size<!>Quantitative RT-PCR (qPCR)<!>Alcian Blue Staining<!>Statistical Analysis<!>Deletion of Raptor in prechondrogeic limb mesenchyme reduces cell size<!>mTORC1 signaling promotes chondrogenesis in cell cultures<!>Discussion

<p>Chondrocytes, the cells responsible for generating cartilages, are important cellular components of mammalian limb skeleton (Long and Ornitz, 2013). Formation of chondrocytes from mesenchymal progenitors, known as chondrogenesis, occurs in both physiological and pathological conditions, such as endochondral bone development, fracture healing, and heterotopic bone formation (Long and Ornitz, 2013). In all cases, chondrogenesis begins with condensation of mesenchymal progenitors caused by increased cell-cell contact (Woods et al., 2007). Subsequently, cells in the center of mesenchymal condensation differentiate into chondrocytes, which express cartilage-specific matrix genes, including collagen II and aggrecan. Previous studies have identified important roles of extracellular signaling pathways and transcriptional factors in chondrogenesis (Long and Ornitz, 2013). In particular, Sox9, a HMG-box-containing transcriptional factor, was found to be essential for chondrogenesis to proceed after the condensation stage, although it is dispensable for initiation of mesenchymal condensation (Barna and Niswander, 2007; Lim et al., 2015). Following chondrogenesis, chondrocytes within the cartilage template undergo an initial proliferation phase and then a maturation process to become hypertrophic chondrocytes. The hypertrophic cartilage is eventually removed and replaced by bone during endochondral bone development (Long and Ornitz, 2013).</p><p>Mechanistic target of rapamycin (mTOR), an evolutionarily conserved serine/threonine kinase, functions as a signal integrator for multiple signals, including growth factors, nutrients, and energy (Laplante and Sabatini, 2012; Sengupta et al., 2010b). By doing so, it regulates a variety of cellular processes, such as cell metabolism, growth, differentiation, and survival (Laplante and Sabatini, 2012; Sengupta et al., 2010b). mTOR is present in two functionally distinct protein complexes: mTOR complex 1 (mTORC1) and complex 2 (mTORC2) as a catalytic subunit (Laplante and Sabatini, 2012; Sengupta et al., 2010b). mTORC1 and mTORC2 can be distinguished by their unique components. For example, Raptor and Rictor are only present in mTORC1 and mTORC2, respectively. Ablation of Raptor or Rictor leads to the disruption of mTORC1 or mTORC2 activity both in cell cultures and in animals, indicating that Raptor and Rictor protein are essential for the activity of their respective complexes (Bentzinger et al., 2008; Guertin et al., 2006). Moreover, the complexes have their different downstream effectors (Laplante and Sabatini, 2012; Sengupta et al., 2010b). The two best characterized targets of mTORC1 are p70 S6 kinase (p70S6K) and eukaryotic translation initiation factor 4E binding protein (4EBP1) (Laplante and Sabatini, 2012; Sengupta et al., 2010b).</p><p>mTOR pathways have been shown to play important roles in mammalian limb development. Conditional deletion of mTOR with Prx1-Cre caused severely diminished limbs (Chen and Long, 2014). Moreover, deletion of Raptor by the same Cre line largely recapitulated phenotypes exhibited by mTOR knockout mice (Chen and Long, 2014), whereas similar ablation of Rictor only mildly affected limb growth (Chen et al., 2015), indicating that mTORC1 is the major mechanism by which mTOR kinase regulates embryonic limb development. Mechanistically, mTORC1 pathway promotes embryonic skeletal growth through regulating chondrocyte size, hypertrophy, and matrix production (Chen and Long, 2014). However, it still needs to be determined whether mTORC1 controls limb mesenchymal cell growth and chondrogenic differentiation (chondrogenesis).</p><p>In this study, by employing both genetic and pharmacological approaches to disrupt mTORC1 signaling, we provided direct evidence that mTORC1 promotes limb bud cell growth and chondrogenesis.</p><!><p>Mouse strains used in this study, including Prx1-Cre and Raptorf/f, have been described previously (Logan et al., 2002; Sengupta et al., 2010a) and were purchased from Jackson Laboratory (Bar Harbor, ME). Production of RapCKO (Prx1Cre; Raptorf/f) is as previously described (Chen and Long, 2014). For timed pregnancies, matings were set up in the late afternoon and mice were checked for vaginal plugs early next morning. The noon of the day when a vaginal plug appeared was designated as embryonic day (E) 0.5. Animal studies were approved by the Animal Studies Committee at Washington University.</p><!><p>For histology-based analyses, embryonic limbs were dissected out in PBS, fixed in 10% formalin overnight at room temperature, and then processed for paraffin embedding prior to sectioning at 6 μm thickness. H&E staining was performed on paraffin sections following the standard protocols. Whole-mount in situ hybridization was performed as described previously (Lim et al., 2015).</p><!><p>Limb bud micromass cultures were performed as previously described (Lim et al., 2015). Briefly, limb buds were dissected out from E11.5 mouse embryos, dissociated into single cells, and reconstituted at a density of 2 × 107 cells/ml. 20 μl cells were spotted onto each well of a 12-well plate, and then allowed to attach for 30 minutes before being cultured in standard medium (DMEM containing 10% FBS and 1% penicillin/streptomycin). Cells were then cultured for 6 days with media changed every other day prior to Alcian blue staining. For rapamycin experiments, cells were first cultured in standard media for 2 days, and then cultured in standard media containing either DMSO or 20 nM rapamcyin for additional 4 days.</p><!><p>ATDC5 cells (RIKEN BRC) were maintained in culture medium (DMEM, 5% fetal bovine serum, 1% penicillin/streptomycin). For chondrogenic induction, confluent ATDC5 cells were cultured in differentiation medium supplemented with 50 μg/ml ascorbic acid (Sigma) and 1% ITS premix (Gibco) in the presence of either vehicle (DMSO) or 20 nM rapamycin for indicated times. Media were changed every other day.</p><!><p>For cell size analysis, E11.5 limb buds were dissociated into single cell suspension, and then stained with propidium iodide (PI) solution. The mean forward scatter height (FSC-H) of 20,000 single live cells was determined by flow cytometry. Histogram plots were created using FlowJo software.</p><!><p>Total RNA was extracted from cells using the Qiagen RNeasy kit (Qiagen) following the manufacturer's instructions. 1 μg RNA was reverse transcribed to cDNA using an iScript cDNA synthesis kit (Bio-Rad) according to the manufacturer's instructions. Quantitative RT-PCR was performed with universal SYBR green supermix (Bio-Rad). The following qPCR primers were used in this study: Col2a1 (F: ACTGGTAAGTGGGGCAAGAC, R:CCACACCAAATTCCTGTTCA), Acan (F:CGTGTTTCCAAGGAAAAGGA, R:TGTGCTCGATCAAAGTCCAG), Sox9 (F:AGGAAGCTGGCAGACCAGTA, R:CGTTCTTCACCGACTTCCTC), and Actb (F: GTGACGTTGACATCCGTAAAGA, R: GCCGGACTCATCGTACTCC). The expression levels of Col2a1, Acan, and Sox9 were normalized to Actb.</p><!><p>ATDC5 cells or micromass cultures were rinsed with PBS, fixed in Kahle's fixative for 10 minutes, and then incubated with Alcian blue staining solution (1.0% Alcian blue in 0.1N HCl) for 1 hour at room temperature. Excess stain was washed off with double distilled water.</p><!><p>All quantitative data were presented as mean±standard deviation (S.D.) with a minimum of three independent samples. Statistical significance was determined by two-tailed Student's t-test. P-values less than 0.05 were considered statistically significant.</p><!><p>The mTORC1 pathway is known to be an important regulator of cell size (Fingar et al., 2002). We previously demonstrated that mTORC1 signaling increases the size of chondrocyte, but did not investigate the effect of mTORC1 on the prechondrogenic mesenchymal cells (Chen and Long, 2014). Here, we first compared the size of mesenchymal cells isolated from the limb buds of wild-type versus Prx1-Cre; Raptorf/f (RapCKO) littermate embryos at E11.5 prior to overt chondrogenesis. Flow cytometry of viable limb bud cells revealed a clear leftward shift in the forward scatter height (FSC-H) histogram for the RapCKO cells relative to the control cells, indicating a decrease in cell size of the mutant cells (Fig. 1A). Quantification confirmed that the mean FSC-H value of the mutant cells was decreased by ~8.8% (P<0.001) compared to the control (Fig. 1B). Consistent with the smaller cell size, the limbs of the RapCKO embryos at E11.5 and E12.5 were smaller than those of the control littermates (Fig. 2A–D). At E12.5, the digit arrays exhibited a normal configuration in the RapCKO embryo despite the smaller size. Histological sections confirmed that a smaller humerus was present in the mutant limb (Fig. 2E, F). In keeping with the morphology, whole-mount in situ hybridization in E11.75 embryos detected Sox9 expression in the presumptive humerus of the mutant limb bud, albeit within a smaller domain than normal (Fig. 2G, H). However, in the distal limb mesenchyme, discrete Sox9-expressing domains demarcating digit primordia, as evident in the control embryo at E11.75 (denoted by asterisks), were yet to appear in the RapCKO embryo; this could indicate a delay in chondrogenesis or merely a size reduction of the domains beyond the detection limit (Fig. 2G, H). Overall, mTORC1 signaling in the limb mesenchyme is required for the normal size of both individual cells and the whole limb primordium, but appears to be dispensable for skeletal patterning during embryogenesis.</p><!><p>We next test directly whether mTORC1 signaling regulates chondrogenesis. We first performed micromass culture with primary embryonic cells. When cultured at high density, mouse limb bud mesenchymal cells undergo cellular condensation and subsequent differentiation into chondrocytes to form cartilage nodules, mimicking chondrogenesis in vivo (Dong et al., 2010; Lim et al., 2015). We treated micromass cultures prepared from E11.5 embryos with either vehicle or rapamcyin, a pharmacological inhibitor of mTORC1 signaling. Alcian blue staining detected cartilage nodules in both vehicle- and rapamycin-treated micromass cultures by day 6 in culture (Fig. 3A). However, the number and size of cartilage nodules were significantly reduced in rapamycin-treated cultures (Fig. 3B). Moreover, when micromass cultures were performed with limb bud cells from E11.5 RapCKO versus littermate control embryos, few cartilage nodules were detected in the mutant cells (Fig. 3C, D). Finally, we examined the role of mTORC1 in chondrogenesis from ATDC5 cells, a commonly used mouse cell line derived from teratocarcinoma (Atsumi et al., 1990; Shukunami et al., 1996). The vehicle-treated ATDC5 cells as expectedly formed cartilage nodules progressively from day 7 to 14 in culture. Rapamycin however markedly reduced the number of nodules at both time points (Fig. 4A–F). Molecular analyses with qPCR confirmed that rapamycin greatly suppressed the induction of the cartialge matrix genes Acan and Col2a1 after 7 days of chondrogenic culture (Fig. 4G). Importantly, Sox9, a master regulator of chondrogenesis normally induced by the chondrogenic conditions, was no longer up-regulated in the presence of rapamycin (Fig. 4G). Thus, the studies in vitro support the conclusion that mTORC1 signaling promotes chondrogenesis.</p><!><p>Although prior studies have identified multiple signaling pathways and transcriptional factors regulating proliferation and survival of mesenchymal progenitors as well as their differentiation into chondrogenic lineage (Long and Ornitz, 2013), such knowledge is still incomplete. In particular, the mechanism by which mesenchymal progenitors sense and integrate environmental cues to adjust their growth and chondrogenic differentiation is still not clear. In this study, we demonstrated that the nutrient-sensing mTORC1 pathway is important for limb bud cell growth and chondrogenic differentiation. There remains a pressing need for therapeutically modifying chondrogenesis in a number of disease conditions. In some instances, enhancing chondrogenesis is desirable, such as repairing cartilage damage in osteoarthritis, while in other situations inhibiting chondrogenesis is preferred, such as eliminating heterotopic bone formation. Therefore, gaining a comprehensive understanding of molecular mechanism underlying chondrogenesis is crucial for developing effective treatments for these diseases.</p><p>The present study, for the first time to our knowledge, provided the genetic evidence for an important role of mTORC1 signaling in promoting cell growth and chondrogenesis during early limb development. However, the authors should point out that reduction of cell size in RapCKO limb bud cells may be caused by the delay of the chondrogenesis. But since the majority of limb bud cells in E11.5 are uncommitted mesenchymal cells, the reduction of cell size in RapCKO likely reflects the important role of mTORC1 in regulating limb bud cell growth. The role of mTORC1 signaling in promoting chondrogenesis during early limb development was supported by both in vitro and in vivo data, however, excessive mTORC1 signaling is problematic for cartilage homeostasis (Zhang et al., 2015). For example, in OA cartilages mTORC1 signaling was reported to be hyper-activated, leading to autophagy suppression, subsequently chondrocyte apoptosis and cartilage degradation (Zhang et al., 2015). Surprisingly, inducible cartilage-specific deletion of mTOR in five weeks old mice did not cause any defect in articular cartilage, instead protecting mice from surgically induced osteoarthritis (Zhang et al., 2015). Furthermore, rapamycin treatment of OA chondrocytes increased the expression of anabolic genes including Col2a1 and Acan (Zhang et al., 2015). The previous and the present study together indicate that mTORC1 signaling is differentially required in chondrocyte development versus homeostasis, and that different levels of mTORC1 activity could cause opposite outcomes.</p><p>Rapamycin treatment inhibited the induction of Sox9 expression in ATDC5 cells, indicating that mTORC1 signaling may promotes chondrogenesis in part through regulating Sox9 expression. However, how mTORC1 exactly regulates chondrogenesis is still uncertain. Eukaryotic translation initiation factor 4E (eIF4E)-binding protein 1 (4E-BP1) and S6 kinase 1 (S6K1) are two best known targets of mTORC1 pathway. It will be interesting to determine whether 4E-BP1 or S6K1 or both is the downstream mediator of mTORC1 in chondrogenesis.</p>

PubMed Author Manuscript

DNA Bending Propensity in the Presence of Base Mismatches: Implications for DNA Repair

DNA bending is believed to facilitate the initial recognition of the mismatched base for repair. The repair efficiencies are dependent on both the mismatch type and neighboring nucleotide sequence. We have studied bending of several DNA duplexes containing canonical matches: A:T, G:C, various mismatches: A:A, A:C, G:A, G:G, G:T, C:C, C:T, T:T, and a bis-abasic site: X:X. Free energy profiles were generated for DNA bending using umbrella sampling. The highest energetic cost associated with DNA bending is observed for canonical matches while bending free energies are lower in the presence of mismatches, with the lowest value for the abasic site. In all of the sequences, DNA duplexes bend towards the major groove with widening of the minor groove. For homoduplexes, DNA bending is observed to occur via smooth deformations, whereas for heteroduplexes, kinks are observed at the mismatch site during strong bending. In general, pyrimidine:pyrimidine mismatches are the most destabilizing, while purine:purine mismatches lead to intermediate destabilization and purine:pyrimidine mismatches are the least destabilizing. The ease of bending is partially correlated with the binding affinity of MutS to the mismatch pairs and subsequent repair efficiencies, indicating that intrinsic DNA bending propensities are a key factor of mismatch recognition.

dna_bending_propensity_in_the_presence_of_base_mismatches:_implications_for_dna_repair

INTRODUCTION<!>Molecular Dynamics Protocol<!>Preparation of DNA duplexes<!>Umbrella sampling using DNA bending angle restraints<!>Structural and energetic analysis<!>RESULTS AND DISCUSSION<!>Free energy profile associated with DNA bending<!>Directionality of bending<!>Variations in helicoidal parameters<!>Fluctuations in phosphodiester backbone torsions<!>Conformational space sampled at various bending angles<!>Hydrogen bonding and stacking interactions between base pairs<!>CONCLUSIONS

<p>DNA structure is commonly characterized in terms of its helical form. However, deviations from the regular helix structure are observed during many biological processes, such as gene expression,1,2 DNA repair,3–6 or even under mechanical stress.7 These deformations can range from the wrapping of DNA around histones1,2 to the local bending of DNA during protein-DNA recognition.4–6 One prominent example is the highly bent DNA structure found in complexes with MutS and MutSα, the enzymes responsible for the first echelon of post-replication mismatch repair.3–6 DNA bending is believed to be a key feature by which base mismatches or base insertions/deletions are recognized. One hypothesis suggests that the mismatch repair proteins initially bind to DNA non-specifically and then probe for increased local flexibility in the DNA due to the presence of a mismatch.8–12 The crystal structures of mismatch DNA bound to the mismatch recognition proteins for both prokaryotes13–15 and eukaryotes16,17 show highly specific contacts with the mismatch, involving a conserved motif that is inserted into the minor groove of the DNA. This raises the question of whether DNA mismatch recognition is achieved primarily because of intrinsic DNA properties or through specific protein-DNA interactions at the mismatch site. The broader question is how the mismatches alter the flexibility of DNA duplexes.</p><p>A DNA base mismatch occurs when other base pairs are present in place of the correct canonical matched pairs adenine:thymine (A:T) or guanine:cytosine (G:C). This may result most frequently from nucleotide misincorporation during replication. Interestingly, the mismatch repair system does not repair all types of mismatches with equal efficiency,3–5,18–23 suggesting that the differences in the intrinsic properties of DNA owing to different mismatches may play a role. Several biophysical,24–26 biochemical,27 and molecular biology experiments18–21 have been carried out to characterize the mismatch efficiencies for different mismatch types. It was found that the probability of repair is correlated with both the mismatch and the neighboring nucleotide sequence.18,19,28 A general conclusion from these studies is that, for the methyl-directed mismatch repair system in E. coli, G:T, A:C, G:G, and A:A mismatches are repaired with the highest efficiency, followed by T:T, C:T, and G:A that are repaired with intermediate efficiencies, while C:C mismatches are the least efficiently repaired. The order of affinity for the binding of synthetic DNA fragments containing possible mismatches to E. coli MutS was found to be as: G:T > G:G > A:A ≈ T:T ≈ T:C > C:A > G:A > C:C > G:C from band shift analysis.29 The order of affinity largely matches the repair efficiencies suggesting that the initial binding of DNA to MutS is a major factor. Similar trends are observed for the eukaryotic mismatch repair efficiencies,4,21,30 with the general conclusion that purine:pyrimidine and purine:purine mismatches are repaired more effectively than pyrimidine:pyrimidine mismatches.5,18,31 Biophysical characterizations of DNA duplexes containing mismatches, using gel electrophoresis, NMR, and calorimetric techniques24,25,32,33 furthermore suggest an influence of the neighboring sequence context on the stability of the mismatch pairs. The stability of mismatch-containing DNA in turn is roughly correlated with their observed repair efficiencies. However, these experiments fail to establish a clear mechanistic picture for how exactly the mismatches are identified by the DNA repair enzymes. In particular, the intrinsic bending propensity of mismatch-containing DNA remains unclear, which is the focus of the present work.</p><p>Sequence-induced DNA deformations have been studied previously.34–37 For bent DNA structures,38,39 different types of deformations have been observed including kinked structures,40,41 flipped-out nucleotides,42 and the formation of local bubbles.39,41 However, the inherent flexibility and transient stability of bent DNA structures limit the role of experiments when aiming to characterize such structures in atomistic detail. Instead, molecular dynamics (MD) studies have been used extensively to provide structural details for bent DNA, involving DNA minicircles40,41 and DNA oligomers, such as A-tract DNA43–46 and B-DNA oligomers.44,45,47–49</p><p>A number of studies have previously examined bending in DNA duplexes. Using the helical parameters roll, tilt, and propeller twist to evaluate the magnitude and direction of bending in B-DNA47,48 and A-tracts,43,47 it was observed that the precise magnitude of the bend is sequence dependent and bending is facilitated by a smooth deformation that is induced via rolling of adjacent base pairs. For small DNA minicircles, Lankas et al.40 observed that in addition to smooth deformations, both, local base unpairing and kinking, enhance the DNA flexibility; and kinking remains a feasible route for relaxing the elastic energy in strongly bent DNA giving rise to anharmonic behavior50. Two types of kinks were observed at the positions of high curvature: type I and type II. The type I kink is characterized by the significant bending at a single base pair in the DNA structure51. The type II kink is characterized by the localized melting of the DNA,39,52,53 where the central base pair of three consecutive base pairs is broken and its bases are stacked onto the 5′ neighboring bases of the corresponding strand.</p><p>Curuksu et al.44,45 used a screw-axis orientation approach to induce continuous bending deformations of B-DNA, which is based on the calculation of a set of screw axes of the adjacent base pairs, and restraining the angle between two axes using umbrella sampling. Such controlled bending studies on oligomers with alternating GC and AT sequences and A-tracts45 indicate that while moderate bending occurs mainly through the coupled rolling of adjacent base pairs, strong bending leads to local type II kinks. Adiabatic mapping along the screw-axis orientation for oligomers containing G:A, G:T, C:C mismatches, or an abasic site 44 suggested that a greater variety of bent conformations with different directions and magnitudes of global bending are possible for the G:A and C:C mismatches compared to regular B-DNA. In another approach, Lankas et al. 46 used averaging of two or more base-fixed coordinate frames to quantitate the global bending of A4T4 or T4A4 tracts. They observed that A4T4 tracts are bent significantly more than T4A4 tracts. In a recent study,49 Spiriti et al. used an adaptive umbrella sampling approach to determine the free energy surface of DNA dodecamers using a pseudoroll angle definition to describe bending. In these simulations, DNA bending was studied in both directions at a central base pair step, exploring roll angles between −70° to +70°, with positive and negative roll angles corresponding to bending of the DNA towards the major and minor grooves, respectively. They observed that most of the DNA bending flexibility is due to changes in the roll angles. The key finding from this study was that bending at high roll angles appeared to be facilitated by ionic screening. While the focus of the previous studies has been on bending in regular B-DNA structures,44,45,49 very little is known about bending propensities of mismatch-containing DNA.</p><p>Here, we focus specifically on the intrinsic ability of mismatch-containing DNA to assume bent conformations. Because spontaneous bending of DNA in the absence of proteins is a rare event, we employ umbrella sampling simulations using the bending angle as a restraining degree of freedom, and report relative free energy profiles as a function of the mismatch type. The results are then further interpreted in the context of DNA repair and recognition of mismatch-containing DNA by repair enzymes.</p><!><p>Molecular dynamics simulations were performed with NAMD 2.754 using the CHARMM27 force field for nucleic acids.55,56 Parameters for protonated cytosine (at N3) were taken from supplementary parameters provided in the CYTH residue patch developed by the MacKerell group and distributed with the recent CHARMM c36 and c37 packages. Parameters for protonated adenine (at N1) were developed following the same protocol as the original nucleic acid parameters. Further details are provided in the supplementary material. The CHARMM program (version c36a6)57 was used for the initial setup and to prepare the protein structure files (PSF). Each structure was solvated in a truncated octahedral box of TIP3P water molecules in such a way that the minimum distance between solute images would be 20 Å. This resulted in a box size of approximately (65 Å)3. The negative charge of the overall system was then neutralized by adding sodium ions (28 for the regular DNA duplex and 27 for protonated nucleic bases, corresponding to 0.16 M). All simulations were carried out using periodic boundary conditions. Electrostatic interactions were calculated using the particle-mesh Ewald (PME) summation scheme with the PME grid spaced evenly at 1 Å. The simulations were performed in the NPT ensemble at a constant temperature of 298 K using a Langevin thermostat applied to non-hydrogen atoms only, with a damping coefficient of 5 ps−1. A constant pressure of 1 atm was maintained using a Langevin pressure piston with a piston decay time of 100 fs, a piston oscillation time of 200 fs, and with a piston temperature of 298K. All simulations were performed under holonomic constraints placed on all bonds that included a hydrogen atom, using the SETTLE algorithm58 with a tolerance of 10−8 Å.</p><!><p>A total of 15 pentadecamer DNA duplexes with the sequence 5′-GAACCGCXCGCTAGG-3′/5′-CCTAGCGYGCGGTTC-3′ were considered where the central base pair X8:Y23 was varied to include canonical and mismatched pairings as well as a bis-abasic variant without a base at either X or Y sites. Throughout this paper, each of these DNA oligomers is referred to by the name of its central base pair. For instance, the G:T notation will refer to the DNA heteroduplex consisting of guanine and thymine as the central mismatch base pair. Initial structures were prepared from the crystal structure of the G:T-mismatch DNA in complex with MutSα (PDB ID: 2O8B 16). The bent mismatched DNA fragment was extracted, minimized, solvated in a rectangular (48 × 55 × 70 Å3) box of explicit (TIP3P) water molecules, and then simulated (unrestrained MD, 298 K, Langevin thermostat with damping coefficient of 5 ps−1, 20 ns) in the absence of the protein. This resulted in a slightly curved B-DNA structure with a bending angle of around ~170°. This equilibrated structure was then used as a starting point to model the structures of the other mismatches.</p><p>The mismatches A:A, A:C, G:A, G:G, C:T, C:C, and T:T, as well as canonically paired structures (G:C, A:T) were prepared by mutating the bases using mutateNA. pl from the MMTSB Tool Set.59 For the three purine-purine mismatches A:A, G:A, and G:G, both, anti, anti and anti, syn forms, are possible. NMR studies suggest that only the anti, anti form with transient hydrogen bonds is seen for A:A.24 For G:A60 and G:G base pairs,61 there appears to be a dynamic equilibrium. The anti, anti form is believed to be dominant for G:A,60,62 but the G(anti):A(syn) form is also observed by X-ray analysis.15 In the case of G:G, the G(anti):G(syn) form is believed to be dominant.15,26,62,63 For mismatches C:C and A:C, the protonation states of cytosine and adenine require special attention since the pKa values of protonated A:C and C:C were estimated at 7.2033 and 6.95,64 respectively, which is close to physiological pH.</p><p>To consider alternate mismatch conformations, four additional simulations were set up. One simulation each was run with G(anti):X(syn) (X = G or A), where the equilibrated and solvated structures of DNA duplex containing G(anti):X(anti) were taken as the starting point but with the atoms of the base X rotated around the C1′-N9 axis by 180°. Two simulations were run with protonated pairs C(+):C and A(+):C, which were built from the equilibrated and solvated structures of DNA duplexes containing C:C and A:C but with the extra hydrogen atom added using CHARMM. The four additional systems were minimized in vacuum for 500 steps. Restraints were initially placed on the following bonds: in G(anti):G(syn), N1(GUA8)–O6(GUA23), N2(GUA8)–N7(GUA23); in C(+):C, N4(CYTH8)–N3(CYT23), N3(CYTH8)–O2(CYT23), H3(CYTH8)–O2(CYT23), H41(CYTH8)–N3(CYT23); in A(+):C, N1(ADEH8)–O2(CYT23), N6(ADEH8)–N3(CYT23), H1(ADEH8)–O2(CYT23), H62(ADEH8)–N3(CYT23). Then, the structures were simulated using the setup described above (Langevin thermostat, 5 ps−1 at 298 K, Langevin barostat at 1atm) with force constants that were varied as follows: k = 2 kcal/mol/Å2 for 20 ps, k = 10 kcal/mol/Å2 for 20 ps, k = 20 kcal/mol/Å2 for 50 ps, k = 5 kcal/mol/Å2 for 1.0 ns, k = 1 kcal/mol/Å2 for 0.5 ns, followed by an unrestrained simulation for 0.5 ns. The G(anti):A(syn) simulation created a stable wobble base pair without any restraints and was simply equilibrated with conventional MD for 1.1 ns. All of the hydrogen bonds remained stable in the unrestrained simulations. The relative position of the bases (anti/syn) and the hydrogen bonding patterns of the resulting mismatches adequately represents known experimental structures of matched G:C, A:T, as well as the A(anti):A(anti),24,65,66 G(anti):A(anti),60,62 G(anti):G(anti),24,61,62,67 G(anti):A(syn),15 G(anti):G(syn),15,61–63 A:C,33,68 A(+):C,33,68 C:T,64 C:C,24,64 C(+):C,24,64 T:T,24 and G:T13–16,69 mismatches (Figure 1).</p><p>Finally, a system with an bis-abasic oligomer, where both nucleobases 8 and 23 were completely eliminated and replaced with a hydrogen atom, was set up for comparison of the bending energies. The resulting system was minimized for 150 steps and then equilibrated over 40 ps.</p><p>In order to determine which protonation state is most relevant during bending we compared bending free energy profiles for protonated and unprotonated cases. In the case of A:C, we find that the bending is energetically more favorable for A(+):C compared to A:C (Figure SI1). Given a pKa value of 7.2 for adenine, this suggests that at pH = 7 there may be an equilibrium between A and A(+) for unbent DNA, but upon bending, A(+):C is likely to be the preferred protonation state for the sequence we are considering here. For C:C base pairs, the bending free energy profiles are similar for both C:C and C(+):C (Figure SI1) suggesting that the protonation is secondary to the focus of this paper. Since the pKa is 6.95, we will focus the subsequent analysis on the C:C base pair.</p><!><p>To perform umbrella sampling simulations, a bending angle (ξ) was used as the reaction coordinate. The angle was calculated between the centers of mass of heavy atoms of three nucleotide blocks of the pentadecamer: bases 2–5 / 26–29, 6–10 / 21–25, and 11–14 / 17–20 (Figure 2). The bending angle restraint was applied using the angle component of the NAMD colvar module. The reference bending angle, ξref was changed from 90° to 180° with increments of 5° for a total of 19 windows. All simulations used a simple harmonic biasing potential of the form V(ξ) = ½k(ξ − ξref)2. To prepare the initial structures for each window, the equilibrated starting structures (see preparation above) with a bending angle of 170° were simulated with restraints gradually changing in both directions, towards 180° and towards 90°. Each window was sampled initially for 0.6 ns with a force constant of 0.1 kcal mol−1 degree−2. After the initial umbrella run, an additional 20 ns of production sampling was carried out with a force constant of 0.4 kcal mol−1 degree−2.</p><p>The free energy associated with DNA bending was extracted from the umbrella simulations using the weighted histogram analysis method (WHAM).70,71 WHAM expresses the optimal unbiased probability distribution P(ξ), from a set of simulations performed along the reaction coordinate ξ with biased potentials, V(ξ). The free energy or potential of mean force (PMF) at a given ξ is then obtained as W(ξ) = −kBT ln P(ξ).</p><p>For the calculation of free energy landscapes along additional degrees of freedom that are not part of the biasing function, such as helicoidal parameters, multi-dimensional WHAM analysis was carried out where the first dimension is the bending angle and the additional coordinates were included as additional dimensions. 1D PMFs were generated using in-house generalized WHAM code, based on a formalism described earlier.72,73</p><!><p>The resulting DNA conformations were analyzed in terms of their helicoidal and DNA backbone parameters using 3DNA.74 Averaging of the parameters was done after bias elimination using Boltzmann averaging. A hydrogen bonding analysis was performed using VMD.75 Cluster analysis of conformations, sampled at 4 ps intervals, was performed using the kclust program in the MMTSB Tool Set.59 Structures were grouped based on mutual root mean square deviations (RMSD) of heavy atoms of the central nucleotides 5–11 and 20–26 using a 3.0 Å radius cutoff.</p><p>The van der Waals components of the stacking energies of the nucleobases of the central three base pair segments consisting of nucleotides 7–9 and 22–24 were calculated using CHARMM's interaction energy module (version c36a6).</p><p>Much of the discussion in this paper focuses on the effect of bending on DNA energetics and structure. We have covered here bending angles ranging from 90° to 180° which ranges from highly bent DNA to unbent DNA. In order to separate the effects of different bending regimes we separate the discussion into three ranges that will be referred to throughout the paper: (a) strong bending, corresponding to ξ < 110°; (b) moderate bending, corresponding to 110° ≤ ξ < 130°, and (c) weak bending, corresponding to 130° ≤ ξ ≤ 180°. As can be seen from Table SI1, the moderate bending regime corresponds to the bending angles observed for the DNA when bound to the mismatch recognition proteins.</p><!><p>The primary goals of this study are to examine the energetics as a function of DNA helix bending and to describe the structural perturbations in the DNA that result from bending in the presence of different mismatches. To address these questions, we have performed umbrella sampling simulations of mismatch-containing DNA duplexes along a helix bending reaction coordinate. Eight different mismatches at the central base pair are compared with the two canonical base pairs and bis-abasic DNA. The resulting energetic and structural analysis is presented in detail in the following sections.</p><!><p>The free energy of bending for the DNA duplexes is shown in Figure 3. The profiles adopt an overall similar shape with a global minimum near 160° that corresponds to a slightly bent structure and rising energetic cost of bending towards smaller or larger angles. The shape of the profile is relatively independent of the presence and type of mismatch pairing for bending angles between 130° and 180°. For angles below 130°, differences in the bending free energy become more pronounced. The highest energetic cost for bending DNA is observed for the canonical matches reaching more than 15 kcal/mol at 90°, and the lowest cost is found for the abasic DNA with about 8 kcal/mol at 90°. The energetic cost of bending mismatch-containing DNA is distributed between these two extremes.</p><p>The overall shape of the free energy curves is similar to those obtained by Curuksu et al. using screw axis orientations44,45 and Spiriti et al. using pseudoroll angles,49 but with differences in the magnitude of the bending angle due to different definitions. Large values in our bending reaction coordinate correspond to low positive bending angles in the studies by Curuksu et al.44,45 and Spiriti et al.49 Both of those studies44,45,49 determined the minimum of free energy minima surface at ~10°–20°, which corresponds to ~160°–170° in our reaction coordinate. The free energy costs associated with stronger bending of DNA are higher in our study than those found by Curuksu et al. (they obtained a value of about 5.5 kcal/mol at 90°),44,45 but similar to those obtained from the simulations using pseudoroll angle restraints.49 The difference may be because of sequence effects that play an increasing role as DNA is strongly bent at angles less than 130° or because of differences in force fields and computational methodology.</p><p>A particular interesting observation is that differences between canonical and mismatched base pairs do not become pronounced until bending angles of 120° are reached. Such angles correspond to the bending angles observed for the bent DNA conformations when bound to mismatch recognition proteins (Table SI3). A comparison of the bending free energy values at the bending angles seen in the DNA-mismatch recognition protein crystal structures (110° < ξ ≤ 130°) shows increasing values in the order: X:X < A(+):C < C:C, A:A < C:T < G(anti):G(anti) < G:T< G(anti):A(anti) < G(anti):A(syn) < T:T < G(anti):G(syn), A:T, G:C (Table 1). It is clear that less energetic cost is associated with bending of mismatch-containing DNA in comparison with canonical pairs. For the glycosidic rotamers, the free energies associated with the bending of DNA duplexes with G:A and G:G in the anti, syn conformations are much less favorable than the anti, anti conformers. This suggests that the anti, anti conformations of G:A and G:G may be the relevant conformations in bent DNA. Furthermore, the overall trend in the relative cost of bending DNA for different mismatches is in line with the thermodynamic studies which indicate that (a) purine:pyrimidine and purine:purine mismatches are stable mismatches;3,4,32 (b) pyrimidine:pyrimidine mismatches are destabilizing in all sequence contexts.24,32 However, there is less of a correlation with experimental MutS binding affinities (Table 1) suggesting that bending propensity alone may not be sufficient to explain MutS binding preferences.</p><p>Because our definition of the bending angle is somewhat arbitrary, we have also calculated bending angles using the commonly used DNA analysis software 3DNA and generated the corresponding free energy profiles (see supplementary material, Figure SI2). There is generally a good correlation between our definition and the 3DNA analysis with the overall trends being similar except for a systematic shift in the angles by ~10° due to differences in how DNA bending is defined. Therefore, we believe that our findings are robust irrespective of the exact bending angle definition.</p><!><p>The DNA duplex is free to bend in any direction during our simulations since we only restrain the bending angle but not its direction. In our simulations, all DNA duplexes bend towards the major groove. This is reflected in the DNA groove widths (Figure 4). For weak bending, we observed that the major groove widths of all DNA heteroduplexes are wider than the minor groove widths as expected for canonical DNA structures. As the DNA structure bends more, we observe narrowing of the major grooves and widening of the minor grooves for all of the duplexes (Figure 4). The effect is most pronounced for A(+):C, A:A, G:A, and C:C mismatches where the minor groove is wider than the major groove by 5 to 10 Å. For an ideal B-DNA structure, the minor and major groove widths are about 12 Å and 17 Å, respectively. Structural studies of mismatch DNA -protein complexes13–16 indicate that the mismatch recognition proteins MutS and MSH6 insert a conserved "Phe-X-Glu" motif into the minor groove of DNA. This is accompanied by bending of the DNA towards the major groove side and widening of the minor groove around the mismatch site (Table SI3). In the complex structures, the minor grooves are about 5 Å wider than the major grooves similar to what we observe for strongly bent DNA.</p><p>We have also calculated the curvature of the central seven base pairs from roll and tilt angles (Figure 4). As would be expected, higher curvature values are observed for more highly bent DNA duplexes. The curvature angle is a measure of local DNA bending away from straight DNA (0° degrees curvature). The average values of 50–70° for highly bent DNA therefore suggest that most of the overall DNA bending angle is accounted for by changes in the base pair roll and tilt angles in the three base pairs – the mismatch itself, as well as the pairs before and after the mismatch site.</p><!><p>To analyze the influence of bending on the helical conformations of the DNA duplexes, we have calculated the intra base pair (shear, stretch, stagger, buckle, propeller, and opening) and inter base pair step (roll, tilt, twist, rise, slide, and shift) parameters (Figure 5) using 3DNA. The intra base pair parameters account for the deformations of base pairs, and the inter base pair step parameters account for the alterations in stacking. Local variations are observed for the central 5 base pairs at and around the mismatch site (−2 to +2), with the rest of the duplex conformation showing negligible perturbations. Comparing average values between weakly, moderately, and strongly bent DNA duplexes, we find that for canonical base pairs there are only small changes upon bending except for roll and tilt, which change to accommodate for the overall bending (Figure 4). In the presence of the mismatches, the DNA structure differs significantly from the structure with canonical base pairs for most mismatches, even for weakly bent DNA. Furthermore, DNA bending induces significant changes from the unbent structures for many mismatches. The specific changes in helicoidal parameters upon bending depend strongly on the type of mismatch and it is difficult to discern common trends. For example, the A:A, G:A, and G:G mismatches lead to a greatly increased propeller twist upon bending while C:T and T:T mismatches exhibit significantly altered stagger and buckle upon bending. The only common observations are significant variations in opening and shear parameters of the mismatch pairs upon bending, and a nearly uniform increase in roll angles that is reflective of DNA bending towards the major groove and widening of the minor groove.43,47,48</p><p>Comparing our simulated helicoidal parameters with those observed crystallographically at the mismatch site when heteroduplex is bound to MutS or MutSα proteins, there is close agreement for shear, stretch, and stagger, but buckle and opening values differ significantly (Figure 5A). In the crystal structures of DNA heteroduplexes, the mismatch base orients to form hydrogen bonds with a conserved glutamate residue and the mismatched base pairing opens up, which leads to a significant increase in buckle and opening angle values. For the inter-base pair step parameters (Figure 5B), increased rise, slide, and curvature are observed for the crystal structures at the base pair step (C7X8/Y23G24). No significant variations are observed for the consecutive base pair step (X8C9/G22Y23) parameters. This may be attributed to the fact that in the protein-bound crystal structures, phenylalanine is inserted between Y23 and G24, which leads to a type I kink at the mismatch site and alters the slide, rise, and curvature.</p><!><p>We further analyzed the DNA backbone by calculating ribose pseudorotation and backbone torsion angles (Figure 6, Supplementary Table SI4). At low bending, the sugar puckering conformations are predominantly C2′-endo, as expected from regular B-DNA structures. As the DNA bends, frequent transitions of sugar puckering at the mismatch site are observed from C2′-endo to C3′-endo, primarily for Y23 (Figure 6A, Table SI4). These transitions are observed more frequently for the pyrimidine nucleobases, reflecting previous findings that pyrimidines have a greater tendency towards A-type sugar conformations than purines.76 For the crystal structure bending angle range (ξ = 120°–125°), we observe that for X8, the preferred sugar puckering conformation is C2′-endo and for Y23, the sugar puckers are mostly simulated as O4′-endo. This is in agreement with the crystallographic data where in structures of DNA bound to MutS or MutSα proteins, mismatch pairs have primarily C2′-endo sugars on the plus strand (X8) and O4′-endo sugars on the minus strand (Y23).</p><p>We also analyzed BI/BII transitions that are characterized by the difference in the phosphodiester torsions, ε-ξ, where e is the torsion angle defined by C4′-C3′-O3′-P atoms and ξ is defined by C3′-O3′-P-O5′ atoms. We find that in general, BII conformations are more frequent during strong bending (Figure 6B, Table SI4), as suggested earlier.76 For the mismatches G:T, A(+):C, C:C, G(anti):G(anti), we observed BII conformations for Y23 during low DNA bending. The mismatch pairs observed in the crystal structures of mismatch DNA bound with MutS or MutSα proteins show a preference for the BII conformation at X8 and for BI at Y23. Interestingly, Y23 has a uniformly high propensity for BI for all base pairs between 120° and 140°, close to the bending angle of the DNA when bound to MutS, while BII conformations are sampled increasingly at X8 for angles below 120° for many mismatches.</p><p>The simulations reported here were carried out with the CHARMM27 force field to be compatible with previous simulations of MutS77 and MSH2-MSH6 bound DNA.78,79 Recently, modifications to the CHARMM nucleic acid force field were published to improve the sampling of BI/BII conformations of DNA.80 The modifications increase the amount of BII sampling relative to BI but otherwise have only marginal effects on backbone torsion sampling and essentially no effect on the helicoidal parameters.80 In this study, we focus on the energetics of DNA bending which is driven primarily by changes in roll and tilt angles as discussed above. Therefore, we expect that the updated parameters would have little difference on the bending energetics and overall structural features upon bending. We would expect an overall increase in the sampling of BII conformations, but it is not clear that the updated force field parameters would change the conclusions from our study that DNA bending increases the sampling of BII conformations.</p><!><p>To further characterize the different conformations visited by bent DNA duplexes as a function of mismatches, we pooled the conformations obtained from the simulations at all bending angles, 90° ≤ ξ ≤ 180°, and focused the analysis on the base pairs −3 to +3 around the central base pair. Clustering was then carried out to identify the major conformations that are sampled at different bending angles. The centroids of the clusters are presented as representative structures in Figures SI3, SI4, and SI5 along with their relative occurrences as a function of bending angle. It should be pointed out that the relative populations of different clusters are only qualitatively meaningful since they result from sampling under biased conditions.</p><p>To analyze whether the DNA conformations similar to those observed in the crystal structure are also visited during moderate bending, we calculated RMSD values with respect to the DNA in the MutS crystal structure. The RMSD calculation was carried out for backbone atoms as well as heavy atoms of the central seven base pair fragment of the DNA in sampled conformations with bending angles between 110° ≤ ξ < 130°. The analysis further focused on the structures closest to the crystal structure and the fraction of structures below a certain cutoff (2 Å RMSD for backbone atoms and 4.2 Å for heavy atoms). The results are given in Table 2. We observe that for most of the mismatches, backbone conformations as close to ~1.5 Å are visited during the simulations as indicated by the lowest RMSD values (RBB, Table 2). Interestingly, for certain mismatches such as A:A and G:A, significant populations of backbone conformations (PBB) sample the crystal structure conformation. This suggests that crystal-structure-like conformations are sampled even in the absence of the protein and that once the DNA is bent binding to MutS may involve conformational selection rather than an induced fit mechanism. However, the closest heavy atom RMSD values are relatively large suggesting that the nucleobases do not sample conformations that are consistent with the complex-bound structures and that an induced fit mechanism is required to orient the bases appropriately.</p><p>For most of the mismatches, we observe that type II kinks are found at strong bending angles (ξ ≤ 120°). All type II kinked base-pair steps are associated with highly positive roll angles and highly negative tilt angles (Figure 5B). In crystal structures of protein-DNA complexes of mismatch recognition proteins, a conserved phenylalanine residue stacks between the mismatch base and its 5′ flanking base of minus strand of DNA. The mismatched base is then displaced towards the widened minor groove and type I kinks are observed as the mismatched pair unstacks from 5′ flanking base pair. The type I kinks, similar to those observed in crystal structure and characterized by unstacking of two base pairs, are not observed in our simulations. Presumably, type I kinks would be introduced only in the presence of the protein once the phenylalanine is inserted to stack with the mismatch base.</p><p>In previous studies on DNA bending,45 the presence of kinks was observed to be force field-dependent. For simulations with the parmbsc0 force field,81 only type II kinks were observed during bending simulations.45 However, during simulations with the parm94 force field, which artificially stabilizes the γ trans angles, type I kinks were observed.45 In a separate study, Spiriti et al. used capping potential to exclude the effect of type II kinks and study only the type I kinks with both CHARMM27 and AMBER parmbsc0 force fields.49 Thus some of our conclusions presented here may also depend on the choice of force field.</p><!><p>To further assess the stability of the central base pairs in the presence of mismatches, we have analyzed intra base pair hydrogen bonding for X8:Y23 as a function of DNA bending (Figure 7). In general, the base pairs which form more stable hydrogen bonds have been observed to have higher resistance to bending (Figure 3). An interesting case is observed for G:G, where no consistent hydrogen bonding interaction is present between the guanines (Figure 7) in anti, anti conformer. The representative structures for G:G simulations indicate that in anti, anti conformation, larger disruptions in the DNA structure, also indicated by higher backbone fluctuations (Figure 6), are observed even at weaker bending. G:G pairing in anti, syn conformation on the other hand, stabilizes the DNA duplex with persistent hydrogen bonding interactions.</p><p>For weakly bent DNA, the experimentally known hydrogen bonds (Figure 1) are largely maintained. However, upon bending many of the hydrogen bonds are weakened significantly as bending increases to angles below 120–130° for the mismatches. In contrast, canonical base pairs retain hydrogen bonding until bending angles of about 100°. This suggests that MutS could identify mismatches through weakened hydrogen bonding at bending angles around 120°, allowing for favorable interactions with protein residues in the context of MutS. In fact, a conserved glutamate residue has been identified that interacts in the crystal structures with the mismatch base pair in the minor groove in a position to compete for hydrogen bonding with the mismatch bases.13–16</p><p>Besides hydrogen bonding interactions between the bases, the stacking interactions of the nucleobases also stabilize the DNA duplex. Bending reduces the ability of bases to form optimal stacking interactions and an interesting question is to what extent the energetics of losing such interactions contributes to the cost of bending. We have analyzed the van der Waals contributions to the energy for the nucleobases of the central three base pair fragment (Figure 8) as a measure of stacking between the base pairs −1 to +1 around the mismatch site. For all of the duplexes, the van der Waals energies become more favorable as bending decreases, indicating enhanced stacking interactions among the central base pairs in unbent DNA duplexes. The increase in van der Waals energies upon bending ranges from 2 to 10 kcal/mol, on the same order as the increase in free energy. This suggests that the loss of stacking interactions is a major contributor to the overall increase in free energy upon bending. However, the ordering of different base pairs in terms of bending free energy and in particular the much larger cost of bending DNA with canonical base pairs is not reproduced by the change in van der Waals energies alone.</p><!><p>The results from this study indicate that the free energies associated with the bending of DNA duplex are lower for mismatch containing heteroduplexes than for homoduplexes with canonical base pairs. For homoduplexes, DNA bending is observed to occur via smooth bending of base pair steps, whereas for heteroduplexes, kinks are observed at the mismatch site during strong bending. Our results are in line with the thermodynamic studies3,4,32 which indicate that purine:pyrimidine and purine:purine mismatches are more stable mismatches than pyrimidine:pyrimidine mismatches. Furthermore, the relative ease of bending mismatch containing DNA suggests at least a partial mechanism for discriminating heteroduplex from homoduplex DNA.</p><p>From the PMF profiles associated with DNA bending (Figure 3), it is evident that the bending of either homoduplex or heteroduplex DNA is not a spontaneous process. In the absence of proteins, the global minimum for bending is observed near ~160° and even for the most easily bendable mismatch containing heteroduplex the free energy cost of bending beyond 120° exceeds 3 kcal/mol. This indicates that in order to achieve bent DNA as observed in the crystal structures in complex with MutS or MutSα, the interactions with the proteins are necessary to induce and stabilize the bent conformations.</p><p>Our conformational analysis (Table 2) for structures in the moderate bending regime, corresponding to the structures in complex with MutS, indicates that a fraction of the backbone conformations assumed upon forced bending are quite similar to the crystal structure of protein-bound DNA. Therefore, after DNA is bent by the proteins, the crystallographically observed backbone conformation could be assumed simply through conformational selection. However, the orientations of the base pairs upon bending in the absence of the proteins do not closely match the crystal structure conformation. Hence, an induced fit mechanism is required to assume the protein-bound conformation. In the crystal structures,13–16 there are two key interactions: The mismatched base stacks with the aromatic ring of a conserved phenylalanine, which requires base unstacking and an increase of helical rise. Furthermore, a conserved glutamate forms a hydrogen bond with the N3 atom of a mismatched thymine or the N7 atom of a mismatched purine. We did not gain insight from our simulations whether phenylalanine insertion may be more favorable for heteroduplex DNA. However, a reduction in hydrogen bonding upon bending in our simulations for mismatch-containing DNA suggests that competition for base-base hydrogen bonding by the conserved glutamate residue would provide additional discrimination for mismatch-containing DNA by preferentially stabilizing a complex with heteroduplex over homoduplex DNA.</p><p>An interesting question arises from the anti, anti – anti, syn equilibrium of purine:purine mismatches. In order to form the latter interaction with the glutamate residue, the mismatched purines are observed in syn conformations. On the other hand, our results indicate that the anti, anti conformers of the purine:purine mismatches are relatively easier to bend than the corresponding anti, syn conformers. This suggests that upon initial bending of DNA the purine:purine mismatch may be in the anti, anti conformation but then would have to convert to the anti, syn conformation to assume the crystallographically observed structure. Close inspection of the crystal structures of DNA duplexes consisting of purine-purine mismatch pairs bound to MutS15 shows that there would be enough space for rotation of the purine rings from anti to syn at the mismatch site (Figure 9). Further studies of heteroduplex DNA bound to MutS would be required to examine this idea in more detail.</p><p>In vivo, the mismatch recognition proteins have variable affinities for different mismatches.3,4,6,18–20,22,23 The strongest binding to the MutS is observed for G:T, A:C, and G:G mismatches, whereas weaker binding is observed for C:C mismatch. Correlating our mismatches bending energetics (observed for the crystal structure bending angles) with the binding affinities of MutS to different mismatches (Table 2) and the order of repair efficiencies, we observe that the easily bent mismatches, such as G:T, A:A, G:G, and A:C are good substrates for the MutS recognition, and are repaired efficiently. T:T and G:A mismatches which are relatively resistant to bending are repaired with intermediate efficiencies. This indicates that the intrinsic ease of bending mismatch-containing DNA is likely a key contribution to mismatch recognition by MutS. However, the less efficiently repaired C:C and C:T mismatches do not agree with this trend since their bending free energies are relatively low. This indicates that there are likely additional factors beyond the intrinsic properties of heteroduplex that ultimately determine repair efficiencies.</p><p>Our simulation results combined with existing single molecule8–11 and biochemical data,4–6,12,28 allow us to propose the following mechanism for mismatch recognition by MutS and related homologs: First, MutS scans the DNA duplex via one-dimensional diffusion8–10. While doing so, the protein attempts to induce bending in the DNA. In the presence of canonical base pairs, the energetic cost is high and a stable complex is not formed. However, when a mismatch is encountered, bending the DNA becomes easier and formation of a stable complex stops the diffusion process. Second, upon successful binding to bent heteroduplex DNA more specific interactions involving the "Phe-Xaa-Glu" motif are formed with the DNA. The mismatched base hydrogen bonds with the glutamate; and forms stacking interaction as the phenylalanine is inserted to introduce a type I kink at the mismatch site. These conformations would presumably further stabilize the MutS-heteroduplex DNA complex and allow repair to be initiated. Our previous works on the post-mismatch recognition by MutS or MutSα77,79 proteins suggest that the kink introduced at the mismatch site will destabilize the DNA base pairing at the 5′ adjacent base pair. The opening of the 5′ adjacent base next to the mismatched base was observed to promote the ATP hydrolysis,77 which is associated with the initiation of the sliding-clamp formation.9,10 In this sliding mode, the protein is capable of sliding along the DNA via ATP hydrolysis, and eventually, signals the repair machinery to repair.9,10,12</p><p>Collectively, our results demonstrate that DNA bending is an essential aspect of the mismatch recognition. Here, we have simulated free DNA structures and the next step is to study DNA bending in the presence of MutS to further understand the details of the mismatch recognition process.</p>

PubMed Author Manuscript

Role of the Distal Hydrogen-Bonding Network in Regulating Oxygen Affinity in the Truncated Hemoglobin III from Campylobacter jejuni

Oxygen affinity in heme-containing proteins is determined by a number of factors, such as the nature and conformation of the distal residues that stabilize the heme bound-oxygen via hydrogen-bonding interactions. The truncated hemoglobin III from Campylobacter jejuni (Ctb) contains three potential hydrogen-bond donors in the distal site: TyrB10, TrpG8, and HisE7. Previous studies suggested that Ctb exhibits an extremely slow oxygen dissociation rate due to an interlaced hydrogen-bonding network involving the three distal residues. Here we have studied the structural and kinetic properties of the G8WF mutant of Ctb and employed state-of-the-art computer simulation methods to investigate the properties of the O2 adduct of the G8WF mutant, with respect to those of the wild-type protein and the previously studied E7HL and/or B10YF mutants. Our data indicate that the unique oxygen binding properties of Ctb are determined by the interplay of hydrogen-bonding interactions between the heme-bound ligand and the surrounding TyrB10, TrpG8, and HisE7 residues.

role_of_the_distal_hydrogen-bonding_network_in_regulating_oxygen_affinity_in_the_truncated_hemoglobi

<!>Ctb Sample Preparation<!>Resonance Raman Measurements<!>O2 Dissociation Kinetic Measurements<!>O2 Association Kinetic Measurements<!>Classical Molecular Dynamics<!>Hybrid Quantum Mechanics\xe2\x80\x93Molecular Mechanics Calculations<!>Entropic Contribution Estimation<!>Equilibrium Structural Properties of the O2-Adducts<!>Equilibrium Structural Properties of the CO-Adducts<!>Oxygen Dissociation Kinetics<!>Oxygen Association Kinetics<!>MD Simulations of the Wild-Type Protein<!>MD Simulations of the Distal Ctb Mutants<!>QM-MM Calculations of Oxygen Binding Energies<!>DISCUSSION<!>Structural Basis for the High Oxygen Affinity of Ctb<!>Hydrogen Bonding and the Regulation of Oxygen Affinity<!>Physiological Implications<!>CONCLUSION

<p>Truncated hemoglobins (trHbs) belong to the hemoglobin superfamily of proteins. They are ~20–40 amino acids shorter than conventional globins. Instead of the 3-over-3 α-helical sandwich motif, trHbs adopt a 2-over-2 α-helical structure.1,2 Sequence analysis of more than 200 trHbs indicates that they can be divided into three groups: I, II, and III (sometimes referred as N, O, and P, respectively).2,3 Up to now, most studies on trHbs have been focused on trHb I and II groups of globins,1,4,5 whereas the group III is much less explored. Nonetheless, recently, the trHb III from the foodborne bacterial pathogen Campylobacter jejuni, Ctb, has been structurally and kinetically characterized.4,6–8</p><p>The globin function is typically characterized by its reactivity toward small ligands, such as O2, CO, and NO, which bind to the heme distal site.5,9–11 Among them, O2 is the most abundant and physiologically relevant ligand. In general, Hbs displaying a moderate oxygen affinity act as O2 carrier or storage,1,12 whereas those exhibiting a high oxygen affinity are involved in oxygen chemistry.4,5,13–16 To perform their designated functions, oxygen affinity of Hbs therefore has to be tightly regulated. The structural basis underlying O2 affinity of trHbs has been extensively studied.4–6,11,17–20 These studies show that O2 affinity is mainly modulated by hydrogen-bonding network between heme-bound O2 and amino acid residues located in the distal site of the heme.11,19 Depending on the distal interactions, oxygen dissociation rate (koff) can vary by more than 5 orders of magnitude.21 O2 affinity is also shown to be influenced by the ligand association rate (kon), which is intimately related to the presence of tunnels that facilitate ligand migration from the solvent toward the heme active site.22,23</p><p>Previous oxygen binding studies of trHb I from Mycobacterium tuberculosis (Mt-trHbN), an NO dioxygenase,23,24 showed an unusually high O2 affinity (Kd = 2.3 nM),4 allowing oxygen binding even at low O2 concentrations. The high oxygen affinity of Mt-trHbN is achieved by the presence of two polar residues in the distal site, TyrB10 and Gln E11,13,25 which form hydrogen bonds (HBs) with the heme-bound O2. Mutation of any of these residues to non-HB-forming residues significantly increases koff.26 The other trHb from Mycobacterium tuberculosis, a trHb II (Mt-trHbO), also displays a high O2 affinity (Kd ~ 11 nM).4 Although Mt-trHbO contains three polar residues in the distal cavity, TyrB10, TyrCD1, and TrpG8, which can donate HBs to the bound O2, mutation of TyrB10 or TyrCD1 only slightly alter the koff, whereas the mutation of TrpG8 to Phe increases koff by 3 orders of magnitude.27 Computer simulation studies suggest that the increased koff of the G8 mutant arises from the larger flexibility of the TyrCD1 and LeuE11 side-chain groups.18 As TrpG8 is conserved throughout groups II and III of trHbs,28 these data highlight the pivotal role of TrpG8 in the structural and functional properties of these trHbs.</p><p>Like M. tuberculosis, C. jejuni contains two globins, Cgb and Ctb, which belong to the single-domain Hb (sdHb) and trHb III groups, respectively. Neither globin is essential for bacterial growth under laboratory conditions, yet both are up-regulated in response to nitrosative stress.29,30 Cgb has been suggested to function as a NO dioxygenase, similar to Mt-trHbN.20,31 On the other hand, the physiological function of Ctb is less clear.4,6 When the Ctb gene is knocked out in C. jejuni, the bacterium does not show extra sensitivity to nitrosative stress, but when grown under conditions of high aeration, the bacterium exhibits lowered respiration rates, suggesting a role of Ctb in modulating intracellular O2 flux.30,32</p><p>Ctb contains three polar residues in the distal cavity: TyrB10, HisE7, and TrpG8.8 Previous studies revealed that an intricate HB network involving the three distal polar residues, and the heme-bound ligand in Ctb leads to an extremely high oxygen affinity due to a slow dissociation rate (0.0041 s−1).6 Using CO as a probe, it was deduced that the electrostatic potential surrounding the ligand increases when TyrB10 and/or HisE7 are mutated to nonpolar substituents,6 in sharp contrast to the general trend observed in many other Hbs.4,6 Consistently, either single or double mutation in the TyrB10 and HisE7 residues significantly decreases O2 dissociation rate.6 Despite its importance, the mechanism by which the three distal polar residues control oxygen reactivity in Ctb remains unclear.</p><p>In order to achieve a comprehensive understanding of the structural role of each distal polar residue in controlling oxygen affinity in Ctb, in this work we examined the structural and kinetic properties of the G8WF mutant of Ctb and compared them with those of the previously studied wild-type enzyme and its B10YF and E7HL mutants. The experimental data are complemented with molecular dynamics (MD) simulations,9,11,18,33,34 which have been shown to accurately reproduce the structure and dynamics of ligand-bound heme proteins.10,19,35 In addition, hybrid quantum mechanics–molecular mechanics (QM-MM) were employed to compute the oxygen binding energies of the wild-type and mutant proteins in order to rationalize the experimental data. Our data show that the unusual ligand binding characteristics of Ctb arise from a complex interplay between the structure and dynamics of the three distal polar residues.</p><!><p>The G8WF mutant was prepared using the Stratagene QuikChange site-directed mutagenesis kit. Primers were designed with the codon for the mutated amino acid in the center: forward primer (5′ CACTTAGATCTACCTCCTTTTCCTCAAGAGTTTTTTGAAATTTTTCTAAAACTTTTTGAAGAAAGTTTAAATATAGTTTTAATGAA 3′)- and reverse primer (5′ TTCATTATAAACTATATTTAAACTTTCTTCAAAAAGTTTTAGAAAAATTTCAAAAAACTCTTGAGGAAAAGGAGGTAGATCTAAGTG 3′). PCR was carried out on pBAD/His (Invitrogen) carrying the ctb gene using primers appropriate for the desired mutation. The PCR was then incubated with DpnI, an endonuclease specific for methylated DNA. The remaining nonparental mutated DNA was used to transform Escherichia coli XL1-Blue supercompetent cells. Constructs were checked by sequencing. The G8WF mutant of Ctb was expressed and purified in the same way as for wt Ctb.6,7 To form the CO-bound ferrous complexes, 12C16O (Tech Air, NY) or 13C18O (Icon Isotopes, Summit, NJ) was injected into sodium dithionite-reduced Ctb under anaerobic conditions. The 16O2-bound derivative was prepared by passing the CO-bound protein through a G-25 column to allow the exchange of CO with atmospheric 16O2 and to remove the excess of dithionite. The 18O2-bound sample was prepared by injecting 18O2 into an 16O2-bound protein solution, followed by spontaneous exchange of heme-bound 16O2 with 18O2. The concentration of the protein samples used for the Raman measurements was ~30 μM in 50 mM Tris buffer at pH 7.4 containing 50 μM EDTA.</p><!><p>The resonance Raman (RR) spectra were collected as previously described.6 Briefly, 413.1 nm excitation from a Kr ion laser (Spectra Physics, Mountain View, CA) was focused to a ~30 μm spot on the spinning sample cell. The scattered light was collected by a camera lens, dispersed through a polychromator (Spex, Metuchen, NJ), and detected by a liquid nitrogen-cooled CCD camera (Princeton Instruments, Princeton, NJ). A holographic notch filter (Kaiser, Ann Arbor, MI) was used to remove the laser scattering. Typically, the laser power was kept at ~1 mW and the spectral acquisition time was 60 min. The Raman shift was calibrated by using indene (Sigma) and an acetone/ferrocyanide (Sigma) mixture as the references for the 200–1700 and 1600–2200 cm−1 spectral window, respectively.</p><!><p>The O2 dissociation reaction was initiated by injecting a small volume (~50–60 μL) of concentrated O2-bound protein into a sealed quartz cuvette containing ~850–900 μL of pH 7.4 buffer (50 mM Tris and 50 μM EDTA) with 1 mM CO. The final concentrations of O2-bound protein and free oxygen were ~2–4 and 20 μM, respectively. The reactions were followed at 422 and 405 nm. The dissociation rate constants were calculated based on eq 1:20</p><!><p>The O2 association reaction was measured with a nanosecond laser flash photolysis system (LKS.60 from Applied Photophysics).6 In this system, the 532 nm output (~5 ns, 110 mJ) from a Nd:YAG laser was employed as the photolysis beam. The output from a 150 W xenon arc lamp, at right angles to the photolysis beam, was used as the probe beam. The probe beam passed through a monochromator prior to reaching the quartz cuvette (4 × 10 mm with a 10 mm optical path) containing the sample. The light transmitted through the sample entered a second monochromator, which was synchronized with the first one, and was detected by a photomultiplier tube (1P28 from Hamamatsu Corp.). The signal from the photomultiplier tube was transferred to a digital oscilloscope (Infinium from Agilent Technologies) and then to a personal computer for subsequent analysis. Typically, five or six kinetic traces were averaged to obtain a satisfactory signal-to-noise ratio. The O2 association reaction was initiated by flashing off CO with the photolysis beam in a freshly prepared mixture of a dithionite-free CO-bound complex (with ~0.2 mM CO) in the presence of various concentrations of O2. Under the conditions applied, the rebinding of the CO to the heme iron was much slower than the O2 association reactions, which was monitored at 440 nm following the photolysis.</p><!><p>MD simulations were performed starting from the crystal structure of wild-type (cyanomet form) Ctb (PDB entry 2IG3; monomer A at 2.15 Å resolution)8 with the CN− ligand replaced by O2. The system was immersed in an octahedral box of TIP3P36 water molecules. Simulations were performed under periodic boundary conditions and Ewald sums37 for treating long-range electrostatic interactions. The parm99 and TIP3P force fields, which are mean field, nonpolarizable potentials implemented in AMBER, were used to describe the protein and water, respectively. The heme parameters were obtained using the standard Amber protocol from QM calculations on model systems. These parameters have been successfully employed in previous papers of the group.9–11,19,22–24,37 The nonbonded cutoff used was 9 Å for the minimization and 12 Å for the equilibration and constant-temperature simulations. The temperature and pressure were regulated with the Berendsen thermostat and barostat, respectively, as implemented in AMBER. SHAKE was used to constrain bonds involving hydrogen atoms. The initial system was minimized using a multistep protocol and then heated from 0 to 300 K, and finally a short simulation at constant temperature of 300 K, under constant pressure of 1 bar, was performed to allow the systems to reach proper density. The final structure was used as the starting point for a 50 ns MD simulation at constant temperature (300 K). All mutations were performed in silico.</p><p>The feasibility of the conformational transition in the G8WF mutant was investigated by computing the free energy profile using umbrella-sampling techniques.38 To this end, the distance between the hydroxylic hydrogen of TyrB10 and the imidazole Nδ proton of HisE7 was used as distinguished coordinate. A set of windows of 0.1 Å, simulated for 0.4 ns each, was employed to scan the selected coordinate from 5.6 to 1.7 Å. Two independent sets of 12 windows each were employed to estimate the statistical error.</p><!><p>The initial structures for the QM-MM calculations39 were taken from one representative snapshot of each conformation, which were cooled down slowly from 300 to 0 K and subsequently optimized. The iron porphyrinate plus the O2 ligand and the axial histidine were selected as the QM subsystem, while the rest of the protein and the water molecules were treated classically. All the QM-MM computations were performed at the DFT level with the SIESTA code40 using the PBE exchange and correlation functional.41 The frontier between the QM and MM subsystems was treated using the link atom method42,43 adapted to our SIESTA code. The ferrous pentacoordinated (5c) heme group was treated as a high-spin quintuplet state,44,45 while the ferrous–O2 complex was treated as a low-spin singlet state, as they are known to be the ground states of the system. Further technical details about the QM-MM implementation can be found elsewhere.46</p><p>Binding energies (ΔE) were calculated in selected conformations using eq 2: (2)ΔE=EHeme-O2-(EO2+EHeme) where EHeme–O2 is the energy of the oxy form of the protein, EHeme is the energy of the deoxy form one, and EO2 is the energy of the isolated oxygen molecule. All the O2 binding energies for this work were calculated as stated above. Protein fluctuations influence protein–ligand interactions mainly by the dynamics of the H-bonding network, which may present more than one stable conformation. It is expected that most of protein fluctuation effects are due to these conformations and are of enthalpic nature.</p><p>There is a large amount of experimental evidence of the electric field influence on ligand binding in heme proteins. In this context, bound CO has been extensively used as a probe.47 In order to correlate the simulation results with experimental Raman spectra, we have performed an analysis of the QM-MM optimized structures using Badger's rule,48 which has been successfully applied to ligand bound heme iron systems:49,50</p><p> (3)re=cij(1/ve2/3)+dij where re is the equilibrium C–O distance and νe refers to ν(CO). The empirical parameters cij and dij were obtained from a fit of the computed optimized bond distances versus the experimental frequencies, 97.919 Å/cm2/3 and 0.557 Å, respectively. These empirical values were obtained plotting the RR frequencies vs the C–O bond distance from the QM-MM calculations. The structures for the vibrational calculation were obtained by QM-MM calculations over representative snapshots obtained from MD simulations of the CO complexes.</p><p>Furthermore, the electrostatic potential at oxygen atom from the CO ligand was computed for several snapshots taken from the trajectories of each representative conformation (25 snapshots per nanosecond of MD simulation). Electrostatic potential was computed using the following equation: (4)V(r)=14πε0∑iNqi|r-ri| where ε0 is the permittivity in vacuum (8.8542 × 10−12 F/m), qi is the charge of each atom around the CO (heme atomic charges are not considered),47 r is the position of the oxygen from the CO ligand, and ri is the position of the atoms considered to compute the electrostatic potential.</p><!><p>In this work, we employ a combination of both classical MD simulations to explore the protein conformational space and QM-MM calculations to obtain ligand binding energies with electronic detail. Because of the high computational expense of QM-MM calculations, we performed geometry optimizations in which thermal and entropic effects are not considered. However, on the other hand, in classical MD simulations thermal motions are allowed, and entropic effects are taken into account, since the sampling of different conformations is governed by the actual free energy surface. In order to obtain an estimation of entropic effects on hydrogen-bonding patterns, we have computed entropy, associated with the active site plus the distal aminoacids (TyrB10, HisE7, TrpE15, and TrpG8), by diagonalization of the Cartesian coordinate covariance matrix using the Schlitter method,51 used in previous works of the group.52 Because of sampling issues, entropy estimation is dependent on the time window employed, to compute accurate values requires very extended simulations. So, in the case of the entropy contribution to the different conformations, we are conditioned by the time each one remains stable. In spite the entropy is dependent on the sampling time (t), tending to a limit (S∞) at infinite simulation time.53 Calculated entropies could be fitted well using the empirical expression</p><!><p>Previous RR work demonstrates that the νFe–O2 and νO–O modes of the oxy derivative of wt Ctb appear at 542 and 1133 cm−1, respectively.6 The νO–O mode of heme proteins with His as the proximal ligand is typically silent in RR spectra and was not experimentally observed until recently in several trHbs6,54,55 its appearance has been linked to a dynamic distal HB network.1 Likewise, a low νFe–O2 (<560 cm−1) has been shown to indicate strong HB interactions between the heme bound ligand and its environment.6 As shown in Figure 1a, the G8WF mutant exhibits νFe–O2 and νO–O modes at 552 and 1139 cm−1, respectively, pointing to a strong HB interaction between the ligand and the distal residues. As summarized in Table 1, the wild type and all the distal mutants of Ctb show νFe–O2 frequencies lower than 560 cm−1, consistent with a distal environment involving multiple H-bonding interactions.</p><!><p>To further investigate the ligand–protein interactions in Ctb, CO was used as a structural probe for the distal pocket. Figure 1b shows the RR spectra of the 12C16O and 13C18O derivatives of the G8WF mutant of Ctb. As shown in the 12C16O–13C18O difference spectrum, two sets of νFe–CO/νC–O modes were found with values of 499/1963 and 548/1898, which shift to 487/1871 and 534/1812 cm−1, respectively, upon the substitution of 12C16O with 13C18O. It is widely accepted that, when CO is coordinated to the ferrous heme iron, the Fe–CO moiety may be represented by the resonance forms given in eq 6.56</p><p>As a general rule, a positive polar environment destabilizes form (I) and facilitates iron π back-bonding interaction, leading to a stronger Fe–CO and weaker C–O bond. On this basis, the νFe–CO and νC–O typically follow a well-known inverse correlation as illustrated in Figure 2a.56–59 The location of the νFe–CO/νC–O data point on the inverse correlation line is suggestive of the electrostatic potential around the heme-bound ligand.60 Accordingly, we assigned the two sets of νFe–CO/νC–O modes of the G8WF mutant of Ctb at 499/1963 and 548/1898 cm−1 to an open and a closed conformation, respectively. In the closed conformation (G8C), the heme-bound CO is stabilized by strong HB interactions with its environment, whereas these interactions are absent in the open conformation (G8O).</p><p>In the wt protein, only one set of νFe–CO/νC–O modes at 515/1936 cm−1 was observed (located at the middle of the correlation line in Figure 2a), indicating a distal environment with a medium positive electrostatic potential. The B10YF/E7HL and B10YF mutations cause the data to shift slightly up along the inverse correlation line, suggesting an increase in the distal electrostatic potential. The E7HL mutation, like the G8WF mutation, leads to the presence of two conformations: one (E7C1) looks like G8C, while the other (E7C2) resembles the wt protein. The RR data, together with the kinetic data described below, are summarized in Table 1.</p><!><p>The O2 dissociation reactions were measured by spontaneous exchange of O2 in the O2-bound protein with CO as described in the Materials and Methods section. Figure 2b shows the O2 dissociation kinetics of the wt and mutant Ctb proteins. All the kinetic traces follow single-exponential functions. As listed in Table 1, the dissociation rate of the G8WF mutant, 0.033 s−1, is significantly faster than those of the wt and other distal mutants, manifesting the importance of the TrpG8 residue in stabilizing heme-bound O2 in Ctb. However, the rate is still ~400-fold slower than that of swMb.</p><!><p>The O2 association reaction was measured by flash photolysis of CO from the CO-bound complex in the presence of various concentrations of O2. The kinetic traces obtained for the G8WF mutant are shown in Figure S1 of the Supporting Information. The bimolecular rate constant, kon, obtained from the concentration-dependent plot (Figure 2c) is 1.7 × 106 M−1 s−1, which is 2-fold faster than that of the wt protein6 but 8-fold slower than that of swMb (Table 1). The oxygen affinity (KO2) of the G8WF mutant, calculated based on the ratio of kon versus koff, is 52 μM−1, which is 4-fold lower than the wt protein but is ~40-fold higher than that of swMb. As listed in Table 1, the distal mutations in Ctb cause the oxygen affinity to vary over 3 orders of magnitude, highlighting the importance of the HB interactions involving TyrB10, HisE7, and TrpG8 in controlling ligand binding in Ctb.</p><!><p>To investigate the molecular details governing the ligand–protein interactions in the O2-adducts of Ctb, 50 ns MD simulation was performed on the oxygenated wt protein. Visual inspection of the distal site reveals two distinct conformations, cf1 and cf2 (Figure 3a,b), which interconvert along the time scale of the MD simulation (see Figure 3c).</p><p>In both cf1 and cf2, TyrB10 interacts with the bound oxygen, by forming stable HBs with both the terminal (OT) and proximal oxygen (OP) atoms. In addition, in cf1, the Fe–O–O moiety points toward TrpG8. As a consequence, TrpG8 is strongly hydrogen bonded to the OT atom (2.13 ± 0.27 Å), while the HisE7 is 4.65 ± 0.78 Å away from the OT atom. In cf2, the Fe–O–O moiety points to the HisE7 due to its ~180° rotation along the Fe–OP bond, and a new HB is established between HisE7 and OT (2.21 ± 0.33 Å). In this state, TrpG8 forms a HB with TyrB10, keeping the TrpG8 far from the dioxygen ligand. In general, both cf1 and cf2 are stable and well-sampled during the time course of the simulation; the heme-bound dioxygen ligand always accepts two HBs from the TyrB10 and one additional HB from either TrpG8 or HisE7 (in cf1 and cf2, respectively).</p><p>The deoxy derivative of the wt protein was also examined. The TyrB10 forms an HB with the HisE7 for the majority of the simulation time window (see Figure S2a in Supporting Information). This HB breaks occasionally, and the breaking is associated with the movement of TyrB10 toward the TrpG8 and TyrE15, with an averaged dB10–G8 significantly longer than dB10–E15. In the absence of this HB, HisE7 has no interactions and is able to change the conformation of its side chain in a similar way of the well accepted "E7 gate" in Mb.61,62 This conformation was observed in the X-ray structure.8 Once the O2 binds to the iron, a new HB between TyrB10 and the bound ligand is established, disrupting the TyrB10–HisE7 interaction (see Figure S2 in Supporting Information) and promoting oxygen interaction with TrpG8. During the simulation, several water molecules occasionally enter into the distal pocket and interact with the distal polar residues. This event has been observed in some MD simulations in both the presence and absence of ligand in the distal site. In any case, in the simulations, water molecules used different pathways, and their study is outside the scope of this work.</p><!><p>Like wt protein, the MD simulation of the G8WF mutant shows the interconversion between two conformations that exhibit distinct HB interactions (Figure 4). In cf1, TyrB10 donates a HB to OT (2.08 ± 0.29 Å), while HisE7 forms HBs with both oxygen atoms. In cf2, the side chain of HisE7 rotates ~180° along the Cβ–Cγ bond to accept an HB from TyrB10, preventing both distal residues forming HBs with the heme-bound ligand. In general, the heme-bound dioxygen accepts either 2–3 or 0 HB (Figure 4, a and b, respectively) from the surrounding environment. Extending the MD simulation to 80 ns (see Figure S3 in Supporting Information), the conversion of TyrB10 back to the cf1 conformation was observed during almost 10 ns. Taken together this fact and the results obtained for the wt Ctb, where TyrB10 seems to be the most important residue stabilizing the bound ligand, this behavior suggests that both conformations contribute significantly to the network of HB interactions present in the distal cavity of the G8 mutant oxy derivative. Moreover, the MD simulations indicate that the intercorversion between those conformations occurs in a time scale lower than that required for ligand dissociation. To determine the feasibility of the transition between both conformations, an umbrella sampling calculation was performed. The calculated free energy barrier, around 4 kcal/mol, agrees with the observation of interconversion between those conformations along the time scale of the simulation at room temperature (Figure 4d). In addition, to consider entropic effects in the hydrogen-bonding pattern in the G8WF mutant Ctb, we estimated entropy using the Schliter method using a sampling time of 15 ns. The value obtained (TΔS 0.51 kcal/mol) indicates, as expected, the minor influence of the entropic effects in this particular conformational exchange.</p><p>The MD simulation of the E7HL mutant shows only one stable conformation (Figure 5a), where the OT of the dioxygen ligand is stabilized by two HBs from TrpG8 (2.11 ± 0.29 Å) and TyrB10 (1.89 ± 0.17 Å), similar to the cf1 of the wt protein. As indicated by the asterisks in Figure 5a,b, an additional longer HB between the TyrB10 and the OP of the ligand can be formed transiently, with a mean distance of 2.44 ± 0.24 Å. During the time scale of the simulation, the LeuE7 side chain remains inside the heme pocket as its hydrophobic nature prevents its movement into the solvent. As compared to the wt protein, the data show that HisE7 is important in stabilizing cf2 by positioning the Fe–O–O moiety in an appropriate orientation for interaction with the TyrB10–TrpG8 pair.</p><p>In the B10YF mutant, TrpG8 forms an HB with OT (2.38±0.99 Å) during most of the simulation time (Figure 5c,d). Occasionally, this HB is broken due to rotation of the Fe–O–O moiety toward HisE7, with concomitant formation of a new HB between HisE7 and OT, as indicated by the asterisks. This conformational shift is similar to that observed in the wt protein, although in the absence of TyrB10 cf1 is strongly favored.</p><p>Additional mutation of HisE7 to Leu in the B10YF mutant (leading to the B10YF/E7HL double mutant) does not significantly alter the structural dynamics (Figure 5e,f), indicating that the transient HB interaction between HisE7 and the dioxygen ligand in the B10YF mutant is not critical for ligand escape. This scenario is consistent with the observation that the O2 dissociation rate constants for the B10YF and B10YF/E7HL mutants are very similar.</p><!><p>To estimate the protein–oxygen interaction in each representative Ctb structure obtained from the MD simulations, hybrid QM-MM calculations were performed to calculate the ligand binding energies of the O2-adducts. The results are listed in Table 2. No significant differences were observed for the geometric properties of the Fe–O–O moiety in all the energy-minimized structures of the various derivatives of Ctb. On the other hand, the charge on the heme-bound O2 (qO2) estimated using Mulliken populations varies significantly, ranging from −0.263 to −0.403. Nonetheless, the observed negative changes are consistent with the superoxide character of the bound dioxygen as revealed by the RR data (i.e., the νO–O modes at 1132–44 cm−1). The difference in qO2 indicates that the heme-bound ligand in each protein derivative is differentially polarized by the unique ligand environment with distinct electrostatic potential.</p><p>The calculated O2 binding energies (ΔEO2) range from 27.1 to 38.2 kcal/mol. As expected, ΔEO2 values correlate with the number of HBs to the heme-bound ligand: the larger the number of HBs, the higher the binding energy. In addition, ΔEO2 also reveals subtle differences that arise from the nature and orientation of the involved HBs. For example, comparison of the cf1 and cf2 of the wt proteins (both displaying 3HBs) indicates that the HB donated by TrpG8 is stronger than that donated by HisE7, as suggested by the relatively higher binding energy of cf1 (34.9 vs 31.9 kcal/mol).</p><p>Previous computational studies of heme proteins showed that ΔEO2 is strongly correlated with O2 dissociation rate.9,11 Accordingly, it is reasonable to expect that differences in O2 dissociation in the wt protein and its mutants will be mainly determined by the thermal breakage of protein–oxygen interactions in the heme active site. As listed in Table 2, the ΔEO2 values of all derivatives of Ctb are higher than that reported for Mb (27.0 kcal/mol), except the cf2 of G8WF. The observation is consistent with the lower koff observed in all derivatives of Ctb as compared to Mb. Moreover, the E7HL mutant, which displays the lowest koff value, shows the highest binding energy, when the data of G8WF are excluded. The higher dissociation rate of the G8WF mutant suggests that the protein is in a fast equilibrium between the high (cf1) and low (cf2) binding energy conformations and that the ligand dissociation mainly involves the low binding energy conformation (vide infra). The reported 10-fold increase in koff for the G8WF mutant corresponds to a ΔΔG of about 1 kcal/mol. Our results (binding energies of 34.9 and 31.9 kcal/mol for the wt and 38.2 and 27.1 kcal/mol for the G8WF mutant) are qualitatively consistent with the experimental value, since there is one conformation of the mutant with lower binding energy. However, due to limitations in sampling and force fields, it is not possible to obtain reliable populations necessary to obtain quantitative predictions.</p><p>To correlate the computed values with experiment, C–O frequencies of the conformations obtained from the MD simulations were estimated as well as the electrostatic potential on the oxygen atom of the bound CO ligand (see Table 3). The computed results, both calculated frequencies using eq 3 and electrostatic potential using eq 4, showed a good correlation with experimental data.</p><!><p>Oxygen affinity is the key point in determining the heme protein function. To understand how structure and dynamics regulate O2 affinity, experimental or theoretical studies are commonly used. In this work we show how combining experimental and computer simulation approaches allows us to gain insight into the molecular basis of protein reactivity and structure.</p><!><p>Since its discovery, the structural characterization of Ctb has been focused on the relative contribution of distal residues to ligand stabilization. Initial mutagenesis studies revealed that neither mutation of HisE7 nor TyrB10 to non-HB-forming residues increases the koff.6 The elevated koff for the G8WF mutant (10-fold higher than that for wt) presented herein suggests that TrpG8 provides the major contribution to ligand stabilization. Our RR data show that, in general, the heme-bound O2 in the wt and mutants of Ctb is stabilized by unique HB interactions with its surroundings, as all the νFe–O2 frequencies are <560 cm−1 and the νO–O are Raman-active.</p><p>The MD simulations demonstrate that, in the wt protein, the main residue responsible for oxygen stabilization is TyrB10. Bound oxygen is further stabilized by an additional HB from either TrpG8 (cf1) or HisE7 (cf2), which depends on the orientation of the Fe–O–O moiety (Figure 3a, b). The QM-MM data show that the HB with TrpG8 is stronger than that with HisE7, and therefore mutation of HisE7 leads to a higher binding energy and a lower koff by constraining Ctb to a cf1-like conformation.</p><p>In the O2-adduct of the G8WF mutant, two conformations were observed. In cf1 (Figure 4a), the heme-bound O2 is stabilized by both TyrB10 and HisE7. In cf2 (Figure 4b), a strong HB between TyrB10 and HisE7 is established, preventing either residue from interacting with the ligand, leading to low binding energy. The nature of the two conformers observed in the RR data of the G8WF Ctb CO-adduct (Figure 2a) is consistent with the existence of those conformations: the closed one (cf1), which leads to a strong positive polar electrostatic potential around the heme-bound ligand, and the open conformation (cf2), where the electrostatic potential is less positve. The relatively fast O2 dissociation rate of the G8WF mutant (Table 1) indicates that the dissociation reaction is likely influenced by the presence of the exchange between open and closed conformations, which interconverts in a time scale lower than the dissociation of O2 from the heme.</p><p>It is important to note that under equilibrium conditions the O2-adduct of the G8WF mutant exhibits only one νFe–O2 mode with a relatively low frequency (552 cm−1, see Table 1), indicating the presence of strong HB(s).6,55 Hence, the data suggest that the equilibrium favors the G8C conformer. In contrast, the coexistence of both conformations in the CO-adduct of the G8WF mutant indicates that both G8C and G8O are populated. The population shift toward G8O in the CO-adduct presumably reflects the smaller net partial charge of the heme ligand.</p><p>Also noteworthy is the fact that in the wt protein the CO-adduct exhibits only one conformation with medium polar electrostatic potential (Figure 2a), while the O2-adduct displays two conformations on the basis of MD simulations (Figure 3a, b). QM-MM studies of the O2-adduct show that the energy difference between these two conformations is small, as compared to the G8WF mutant (~3 vs 11 kcal/mol). Moreover, in the wt protein both conformations have the same number of HBs with the ligand, suggesting that the difference in the electrostatic potential surrounding the bound CO in the two conformations of the wt protein is too small to be resolved by the RR studies.</p><p>Even though more accurate entropic data were computed, the results obtained for the G8WF mutant confirm that the sampling of different conformations is governed by the actual free energy surface.</p><p>In the E7HL mutant, the Fe–O–O moiety is preferentially oriented toward TrpG8. The dioxygen ligand is stabilized by HBs donated by TrpG8 and TyrB10, leading to the highest oxygen binding energy and slowest dissociation rate. The HB with TrpG8, however, can be transiently broken due to the rotation of the Fe–O–O moiety. RR studies of the CO-adduct of the E7HL mutant revealed two conformations: E7C1 with a G8C-like high positive polar electrostatic potential and E7C2 with a wt-like medium positive polar electrostatic potential (Figure 2a). The E7C1 likely corresponds to the major conformation observed in the O2-adduct, whereas the E7C2 probably corresponds to the transient conformation, in which the HB donated from TrpG8 to the heme ligand is broken.</p><p>In the B10YF mutant, the O2 dissociation rate is 2-fold higher than that of the wt protein. MD simulations show that, when TyrB10 is mutated to Phe, the dioxygen ligand loses the constraint imposed by TyrB10; the Fe–O–O moiety rotates toward TrpG8, establishing a single yet strong HB with it. As a result, the oxygen binding energy decreases slightly, leading to a small increase in the koff. This is in contrast to Mt-trHbN, where the replacement of TyrB10 with Phe or Leu in Mt-trHbN results in a significant increase in the koff 4 due to the lack of TrpG8.</p><!><p>The coexistence of multiple conformations for the residues in the distal cavity, each characterized by a distinct pattern of HB interaction, creates differences in the local polarity and affects the stabilization of the heme-bound ligand in Ctb. The existence of different HB networks is not a unique propety of Ctb. In fact, the plasticity of HB interactions between the heme ligand, TyrB10, Gln E7, Gln E11, TyrCD1, and/or TrpG8 residues has also been discussed in the single domain Hb from C. jejuni and two trHbs from M. tuberculosis.18 In addition, the interplay between distinct HB interactions between the heme ligand, TyrB10, Gln E7, and ThrE11 in the CerHb from Cerebratulus lacteus nerve tissue and the trHb I from Paramecium caudatum10,63–65 and those between the heme ligand, TyrB10, and HisE7 in soybean leghemoglobin65,66 have all been demonstrated to be important for regulating the oxygen affinities of these Hbs.</p><!><p>The requirement for a high oxygen affinity of Ctb in C. jejuni physiology is at present not understood. However, the microaerophilic lifestyle of this pathogen and a requirement to control intracellular oxygen concentrations and prevent oxidative stress may have resulted in the evolution of a globin with extraordinarily high affinity for oxygen.</p><!><p>Our results show that oxygen affinity of Ctb is intricately regulated by the existence of different hydrogen-bonding interactions between the heme-bound ligand and the three key distal residues: TyrB10, TrpG8, and HisE7. The results also show that the behavior of each residue is affected by the other residues, and therefore the binding affinity is the result of a cooperative property. The accurate description of the behavior of distal residues in all the studied mutants has potential implications for understanding and predicting O2 affinity.</p>

PubMed Author Manuscript

Multifunctional pyrazoline based AIEgens: real-time tracking and specific protein “fishing” of lipid droplets

Despite the rapid development of organic fluorescent probes for bioimaging and biosensing applications, construction of advanced probes with multiple biological functions by precisely integrating different functionalized elements into a single molecule has rarely been reported. In this contribution, a series of multifunctional pyrazoline based fluorescent probes (Pyr-n, n ¼ 1-5) were designed and synthesized by introducing different aromatic moieties into the pyrazoline core. All Pyr probes exhibited the aggregation-induced emission effect. Thanks to the excellent biocompatibility and suitable lipophilicity, the Pyr probes can stain the lipid droplets (LDs) in living cells with high specificity as well as track the lipid metabolism in Zebrafish embryos. The protonation-deprotonation capability of the diethylamino group enables Pyr-5 to reversibly migrate between LDs and mitochondria, and real-time monitor the intracellular pH change in dual-color mode. The mild reaction between the pentafluorophenyl unit and thiol group makes Pyr probes the ideal probes to "fish out" the proteins associated with LDs in living cells.

multifunctional_pyrazoline_based_aiegens:_real-time_tracking_and_specific_protein_“fishing”_of_lipid

Introduction<!>Synthesis and optical properties<!>Crystal analysis and theoretical calculation<!>LD imaging and lipid metabolism tracking<!>LD related protein labeling<!>Conclusions<!>Conflicts of interest

<p>Lipid droplets (LDs), ubiquitous lipid-rich organelles that exist in fungi, plants and animals, play crucial roles in diverse biological processes. 1 It has been shown that LDs regulate the cellular energy storage between surplus and starvation to provide cells a competitive evolutionary advantage. 2 By generating phospholipids and sterols from catabolism or mobilization of lipids, LDs involve in membrane biosynthesis. 3 Using their buffering capacity, LDs protect cells from endoplasmic reticulum stress under excessive amounts of lipids and eventually protect cells against apoptosis or necrosis. 4 Meanwhile, LDs participate in the progress of several human diseases from diabetes, cardiovascular diseases to hepatic steatosis, neurodegeneration, and cancers by controlling the lipid homeostasis. 5 LDs also serve as the sites of storage and metabolism of specic proteins. [6][7][8] For example, by temporarily storing histones, LDs could support the rapid development of Drosophila and Zebrash embryos. By rerouting the peptides or proteins derived from phagocytosed materials, LDs involve in the process of antigen expression from immune cells to T cells.</p><p>Considering the important biological roles of LDs, tracing their dynamic functions is of prime importance.</p><p>The progress of uorescence techniques stimulated the principle, design and synthesis of uorescent probes with capabilities to specically stain subcellular organelles. 9 In particular, some novel uorescent probes for LD imaging have been developed, such as commercial probes BODIPY493/503 and Nile red. 10,11 However, non-specic binding and low signal-to-noise ratio (SNR) resulting from the aggregationcaused quenching effect severely limited the application scope of these conventional probes. 12 The emergence of luminogens with aggregation-induced emission characteristics (AIEgens) provides the optimal choice to overcome the drawbacks of traditional uorescent probes. 13 Beneting from their "lightup" properties that arise from the restriction of intramolecular rotation, AIEgens allow staining of subcellular organelles with high brightness, and have superior photostability, backgroundfree imaging ability and wide range concentration tolerance. 14 Some elegant AIEgens that are favourable for staining LDs have been developed, such as TPE-AmAl, 15 AIP, 16 NAP, 17 and triphenyl amino based probes. 18 Nevertheless, it should be noted that most of the LD specic AIEgens solely focus on staining, and design and construction of simple AIEgens with advanced biological functionalities for tracking the dynamics, metabolism and functionalities of LDs is still rare.</p><p>Herein, we have developed a series of multifunctional uorescent probes (Pyr-1 to Pyr-5) by integrating the pyrazoline core with different functionalized moieties (Scheme 1). The resulting Pyr probes exhibited unique AIE characteristics, large Stokes shis and tunable emission. The excellent biocompatibility as well as suitable lipophilicity with high Clog P values enabled these Pyr probes to display superior LD staining ability in living cells and track the lipid metabolism in Zebrash embryos. Beneting from the protonation of the diethylamino group under acidic conditions, Pyr-5 could reversibly migrate between LDs and mitochondria with dual emission color by precisely responding to the intracellular pH changes within narrow acidbase transition. Moreover, taking advantage of mild reaction conditions between the pentauorophenyl unit and thiol group, Pyr probes could serve as ideal probes to "sh out" the proteins associated with LDs in living cells.</p><!><p>The synthetic routes to Pyr probes are shown in Scheme S1 and S2. † By adapting the convergent "photo-click" reaction of pen-tauorophenyl substituted tetrazoles and styrene derivatives, 19 the pyrazoline based compounds (Pyr-1 to Pyr-5) were facilely obtained with reasonable yield. All Pyr probes were fully characterized by NMR spectroscopy and high-resolution mass spectroscopy, and the data were in good agreement with the target structures.</p><p>The UV-Vis spectra of Pyr probes were studied at the initial stage. As illustrated in Fig. 1A and Table 1, the longest absorption peaks of Pyr-1 to Pyr-5 in acetonitrile varied from 367 to 402 nm, which is probably attributed to the intramolecular charge transfer (ICT) transition from the electron donating part to the electron decient unit. 20 The red-shied absorption peaks from Pyr-1 to Pyr-5 could be caused by the increase of electron donating ability from the 4-methylphenyl to the 4diethylaminophenyl group. When the solutions of Pyr probes were photoexcited, faint or moderate photoluminescence was observed. However, increasing the water fractions (f w ) of mixed solutions enhanced their emission intensity (Fig. 1B and S1 †).</p><p>By taking Pyr-5 as an example, the uorescence intensity remained at a low level when the f w was less than 70%, and a continuous increase of f w led to the emission "light-up" and the emission intensity at 565 nm reached the maximum at an f w of 99% with an approximately 30-fold enhancement compared to that in the solution state (Fig. 1C), demonstrating its typical AIE effect. The existence of nanoaggregates at an f w of 99% was further conrmed by the measurement of dynamic light scattering (Fig. S2 †). It is worth mentioning that the emission intensity of Pyr-4 dropped when the f w was more than 80%, which could be ascribed to the change of morphology and size of nanoaggregates at higher f w . 17 Importantly, all Pyr probes exhibited a large Stokes shi (>133 nm), which shows great advantage for bio-imaging applications due to the reduction of self-absorption. Moreover, the solid-state emission of Pyr-1 to Pyr-5 was tuned from 480 to 558 nm with higher uorescence quantum yields (F f , 5.7-45.2%) than those in the solution state (0.3-34.8%), which further proved their AIE feature.</p><p>The uorescence lifetimes of all Pyr probes both in the solid (2.19-3.91 ns) and solution (1.14-2.26 ns) states were also measured (Table 1). The rate constants for radiative (k r ) and non-radiative decay (k nr ) of Pyr probes in different states were calculated based on the uorescence lifetimes and quantum yields (Table S1 †). Taking Pyr-5 as an example, the k r was 1.639 Â 10 6 s À1 in the solution state, which was less than the k nr value (5.448 Â 10 8 s À1 ). However, the k r increased to 4.346 Â 10 7 s À1 whereas the k nr decreased slightly to 4.238 Â 10 8 s À1 in the solid state. These results implied that suppression of the nonradiative decay was responsible for the enhanced emission in the solid state.</p><!><p>The single crystal structures of these Pyr probes were thoroughly investigated in order to explain the origination of their emission properties from the aspects of molecular conformations and crystal packing modes. To our delight, the single crystals of Pyr-1, Pyr-2 and Pyr-4 were successfully obtained by slowly evaporating mixtures of tetrahydrofuran/n-hexane or dichloromethane/n-hexane containing the probes at room temperature (Table S2 †). As shown in Fig. 2, all crystals adopted a non-planar conguration and the torsion angles between the pyrazoline core and the substituted groups at position 1 were 13.52, 11.57 and 4.62 in the order from Pyr-1, Pyr-2 to Pyr-4, respectively. Additionally, the torsion angles between the pentauorophenyl group at position 3 and the central pyrazoline ring were 13.57, 8. To provide further insight into the relationship between the electronic transitions and special optical behaviour, the groundstate molecular orbital geometries of Pyr probes were optimized using the density functional theory (DFT) method at the B3LYP/ 6-31G basis set (Fig. 3). The calculation results revealed that the highest occupied molecular orbital (HOMO) was primarily delocalized on the aromatic substitutes at position 1 and the central pyrazoline ring, while the lowest unoccupied molecular orbital (LUMO) is mainly distributed over the pentauorophenyl unit at position 3 as well as the pyrazoline core. The separated frontier orbitals suggested the existence of the ICT process in Pyr probes. It is noteworthy that no electron cloud distribution was observed for the aromatic rings at position 5 because of their non-participation in molecular conjugation. Meanwhile, as the electron-donating abilities increased in the order from Pyr-1 to Pyr-5, the energy levels of the LUMO varied from À0.30 to À0.12 while the energy levels of the HOMO increased from À6.40 to À5.65 eV. The energy gaps between the HOMO and LUMO were calculated from 6.10 to 5.53 eV (Pyr-1 to Pyr-5). The better orbital separation degree and reduced energy gap resulted in the red-shied absorption bands from Pyr-1 to Pyr-5.</p><!><p>Considering the hydrophobic characteristic of Pyr probes and the lipophilic environment within LDs, we anticipated that the Pyr probes could be used for LD-targeted imaging. The cell viability studies were carried out at rst by using the 3-( 4 To provide explanation for the high specicity of Pyr probes to LDs, their Clog P values were calculated. Clog P is dened as the calculated log P (n-octanol/water partition coefficient) value and the probes with Clog P > 5 are usually considered to specically stain LDs. 17,22 As shown in Fig. 5A, the Pyr probes displayed Clog P values between 6.516 and 7.820, which were much higher than those of commercial probes BODIPY493/503 (5.028) and Nile red (4.618), and resulted in their selective localization in LDs.</p><p>Besides their excellent biocompatibility and specicity properties, we further evaluated the photostability of these Pyr probes, since superior photostability is always critical for high resolution 3D imaging and time-resolved studies. To our delight, aer 100 scans, the uorescence intensity of Pyr probes remained over 95% (Fig. 5B and C), demonstrating that they possess higher resistance to photobleaching.</p><p>3D imaging is a powerful tool to track the dynamics and functions of subcellular organelles. Generally, 3D imaging is obtained by the reconstruction of multiple Z-stack images, which involves several laser scans of the sample. Therefore, a qualied uorescent probe capable of giving high resolution 3D images is required to emit more photons before photobleaching than 2D imaging and also without sacricing lateral resolution. However, commercial uorescent probes always result in non-specic staining, which leads to the mixed signals of both specic uorescent response and non-specic background noise, further limiting their ability in 3D imaging. Since Pyr probes displayed high brightness as well as excellent photostability, Pyr-2 and Pyr-5 were adopted for 3D imaging of LDs in living cells. Prior to imaging, the nucleus, LDs and cell plasma membrane were stained with Hoechst 33342, Pyr probes and DiD perchlorate, respectively. Then, 18 mm Z-stacking for Pyr-2 (14 mm Z-stacking for Pyr-5) was performed using 1 mm steps, representing 18 (or 14) scans per image. The 3D images exhibited clear and bright green (yellow) spots (Fig. 6), which revealed the great 3D imaging capability of Pyr probes for subcellular LD localization in HeLa cells.</p><p>Another key feature for a qualied subcellular organelle uorescent tracker is wash-free imaging capability, because the wash-free manner can signicantly simplify the bioimaging procedure and avoid interference from the cell morphology during the cell washing process. In view of the AIE characteristics of these Pyr probes, their wash-free staining ability was evaluated. HeLa cells were treated with Pyr probes, BODIPY493/ 503 or Nile red for 30 min and their uorescent images were then collected directly (Fig. 7). The resulting images revealed that even without washing, the LDs can be successfully stained by Pyr probes with excellent image contrast as well as negligible background uorescence. In sharp contrast, the cell images obtained by incubation with BODIPY493/503 or Nile red without washing exhibited strong uorescent signals in the whole cell area. The wash-free imaging ability of Pyr probes with high SNR could be originating from their suitable lipophilicity nature as well as unique AIE characteristics. By virtue of the wash-free imaging ability of Pyr probes, a time-dependent staining was facilely achieved. As shown in Fig. S7, † aer mixing the cell culture with Pyr probes at 37 C for 5 min, the LDs could be visualized for Pyr-5. Upon prolonging the incubation time to 15 min, all Pyr probes gave uorescent signals from LDs, illustrative of their fast staining ability for LDs.</p><p>Zebrash and Drosophila embryos are two kinds of ideal candidates to study lipid dynamics as well as lipid-related diseases, such as diabetes, atherosclerosis and obesity. 7,23 The superior LD staining capabilities of Pyr probes inspired us to explore their application in the tracking of lipid metabolism in living Zebrash and staining of Drosophila embryos. 3 day post fertilization (dpf) Zebrash embryos were incubated with Pyr probes in embryo media for 30 min at 28 C and the images of Zebrash embryos were collected using CLSM (Fig. 8A and S9-S11 †). Intense uorescent signals mainly located in the yolk sac were observed. As the embryonic yolk zone is the sole energy supplier for larval development within the rst week by storing most of the neutral lipids and phospholipids, the above results proved the ability of Pyr probes to stain the yolk lipids. With the absorption of the yolk sac (from 3 dpf to 5 dpf), the uorescence intensity of Pyr probes decreased gradually, which was consistent with the development of the yolk sac. The results showed that the Pyr probes could be used as a probe for tracking the lipid metabolism in living vertebrate organisms. Meanwhile, during the process of imaging, no Zebrash death was observed, suggesting the good biocompatibility of Pyr probes. Similar specic staining of lipids was obtained when the xed Drosophila embryos (stage-13) were employed (Fig. 8B and S12 †). Real-time monitoring of the intracellular pH in dual-color mode</p><p>The reversed transmembrane pH gradient is a key characteristic of cancer cells, and therefore, development of an intracellular pH indicator becomes particularly important for convenient histopathological analysis. 24 Since Pyr-5 contains a diethylamino group, we speculated that it may serve as a sensitive pH indicator. As shown in Fig. 9A, Pyr-5 gave orange emission at 565 nm when the pH value was more than 9. Reducing the pH value to 7, a new blue emission at 480 nm emerged, which further became dominant when the pH value was less than 5. This blue-shied emission was probably due to the reduction of the ICT effect once the diethylamino group of Pyr-5 was protonated. We next evaluated the capability of Pyr-5 for intracellular pH detection, and the cells pre-stained with Pyr-5 were treated with HEPES buffer at different pH values, respectively. As illustrated in Fig. 9C, Pyr-5 presented orange uorescent signals with the targeting of LDs in living cells at pH 7. However, when the cells were incubated with buffer at pH 6, the orange signals became weak whereas blue uorescent signals appeared. And the signals from the blue channel became dominant at pH 5. The emission change of Pyr-5 under different intracellular pH conditions was further conrmed by the in situ lambda scan (Fig. 9B). It is noteworthy that the blue uorescent signals were not located in the LDs. The co-staining experiment with commercial Mito Tracker Deep Red (MTDR) indicated that Pyr-5 mainly accumulated in the mitochondria under pH 5 conditions with the Pearson's coefficient of 0.87 (Fig. 9D). The cationic probes prefer to target the mitochondria through electrostatic interaction because of the negatively charged mitochondrial inner membrane. 25 Therefore, the location of Pyr-5 in mitochondria under acidic conditions could be attributed to the conversion of the probe from neutral to cationic nature aer the molecule was protonated.</p><p>We further tested the real-time migration of Pyr-5 between LDs and mitochondria along with the change of intracellular pH. As shown in Fig. 10, intense orange signals from LDs were observed when the Pyr-5 pre-stained cells were incubated with buffer at pH 7. The culture medium was then replaced with buffer at pH 5 and incubated for 15 min; Pyr-5 released from LDs and migrated to mitochondria with strong blue emission. Subsequently, the stained cells were re-treated with buffer at pH 7 for another 15 min, and the orange signals from LDs recovered while the blue uorescent signals in mitochondria almost disappeared. The above results depicted that Pyr-5 can reversibly monitor the intracellular pH from the aspects of both localization and emission color. To the best of our knowledge, this is the rst example that uorescent probes can migrate between LDs and mitochondria with different emission colors accompanied by a slight intracellular pH change.</p><!><p>Development of chemical strategies for specic modication of proteins associated with LD dynamics and metabolism would enable a new avenue to precisely identify and regulate LD functions. In view of the reactivity between the penta-uorophenyl unit and thiol group through the nucleophilic aromatic substitution (S N Ar) reaction, 26 Pyr probes would be able to "sh out" the proteins associated with biological processes of LDs in native environments by modifying their cysteine (Cys) residues. To implement this possibility, the model reaction of Pyr probes towards N-acetyl (Ac) cysteine was investigated rst (Scheme S3 †). The Pyr probes could smoothly react with N-acetyl (Ac) cysteine in DMF/Tris solution at 37 C for 24 h to generate the desired product with a high yield (>70%). As shown in Fig. S13A and S13B, † absorption and emission wavelengths of the addition product (Pyr-6) between Pyr-1 and N-acetyl (Ac) cysteine red-shied to 400 and 535 nm in DMSO solution compared to those of Pyr-1 (370 and 510 nm). The reduced energy gap of Pyr-6 (5.74 eV) could be responsible for this red-shi of spectra (Fig. S13C †).</p><p>The above results encouraged us to examine whether Pyr probes could be employed to modify the Cys residue of protein. Therefore, bovine serum albumin (BSA), which contains one free thiol (Cys34) and 17 conserved disulphide bonds, was selected as the target protein and Pyr-1 was used as the representative labeling probe. 27 Aer incubating BSA with Pyr-1 (50 equiv.) at different pH values (6.0, 7.2 and 8.6) for 16 h, the reaction mixtures were subjected to sodium dodecyl sulphate polyacrylamide gel electrophoresis (SDS-PAGE) followed by ingel uorescence analysis (Fig. 11B and S14 †). A clear uorescent band around 66 kDa was detected for the reaction at pH 8.6, while weak (or faint) uorescent bands were observed for the reactions at pH 7.2 (or 6.0), which was attributed to the reduction of reactivity of Pyr-1 when the pH decreased from 8.6 to 6.0. Control experiments showed that the reaction did not proceed in the absence of Pyr-1. To prove that the penta-uorophenyl unit can specically react with the Cys-residue, the compound without uoro substitutes (5HPyr-1) was used to test the labeling reaction at pH 8.6. Gratifyingly, the reaction of 5HPyr-1 with BSA did not give any uorescent band compared to Pyr-1 (Fig. 11C and S15 †). Pyr-2 and Pyr-5 were also reacted with BSA and the results indicated that the Cys-residue of BSA can also be conveniently modied by Pyr probes through the S N Ar reaction (Fig. 11D and S16 and S17 †). The protein labeling was further proceeded in living cells using Pyr-1 due to its excellently long-term LD targeting (Fig. S8 †). Aer treating cells with Pyr-1 for 48 h, the cells were lysed and the cell lysis was analysed by SDS-PAGE (Fig. 11E and S18); two major uorescent bands were detected out of multiple other protein bands. However, no uorescent bands were observed in the absence of Pyr-1, indicating that the proteins associated with LDs were successfully labeled by Pyr-1.</p><!><p>In conclusion, a series of multifunctional pyrazoline based uorescent probes (Pyr-n, n ¼ 1-5) have been designed and synthesized. The Pyr probes exhibited a specic AIE feature and tunable emission. Taking advantage of their excellent cellularpenetration ability and biocompatibility, Pyr probes can serve as the ideal uorescent probes for selective LD imaging in living cells. Meanwhile, the tracking of lipid metabolism in Zebrash embryos as well as staining of Drosophila embryos was successfully achieved by employing Pyr probes. Due to the sensitive pH response, Pyr-5 was able to dynamically and reversibly monitor the intracellular pH through both localization and emission color. In view of the reactivity between the pentauorophenyl unit and thiol group, the protein with the Cys residue which is associated with LDs in living cells can be selectively "shed out" by Pyr probes. This intriguing strategy provides a rational platform for the construction of advanced multifunctional uorescent probes for both fundamental and practical applications.</p><!><p>There are no conicts to declare.</p>

Royal Society of Chemistry (RSC)

Human Mitochondrial Chaperone (mtHSP70) and Cysteine Desulfurase (NFS1) Bind Preferentially to the Disordered Conformation, Whereas Co-chaperone (HSC20) Binds to the Structured Conformation of the Iron-Sulfur Cluster Scaffold Protein (ISCU)*

Background: Iron-sulfur cluster biosynthesis involves a scaffold protein (ISCU), cysteine desulfurase (NFS1), chaperone (mtHSP70), and co-chaperone (HSC20).Results: Human mitochondrial ISCU populates structured (S) and disordered (D) conformational states. S interacts preferentially with NFS1 and mtHSP70; D interacts preferentially with HSC20.Conclusion: Shifts in the S ⇄ D equilibrium reveal functional states.Significance: The scaffold protein metamorphic property seen in Escherichia coli is conserved in humans.

human_mitochondrial_chaperone_(mthsp70)_and_cysteine_desulfurase_(nfs1)_bind_preferentially_to_the_d

Introduction<!><!>Introduction<!>DNA Cloning<!>Production of Proteins<!>Buffers<!>Protein Purification<!>NMR Spectroscopy<!>Resonance Assignment of ISCU(N90A)<!>Determination of the Position of the S ⇄ D Equilibrium<!>Circular Dichroism Spectroscopy<!>Size Exclusion Chromatography (Gel Filtration)<!>ATPase Assays<!>In Vitro Fe-S Custer Assembly<!>ISCU Populates Two Conformational States<!><!>ISCU Populates Two Conformational States<!><!>Interaction of ISCU with NFS1 and ISD11<!><!>mtHSP70 Binds Preferentially to the D-state of ISCU<!><!>mtHSP70 Binds Preferentially to the D-state of ISCU<!>HSC20 Preferentially Binds to the S-state of ISCU<!><!>HSC20 Preferentially Binds to the S-state of ISCU<!><!>HSC20 Preferentially Binds to the S-state of ISCU<!>In Vitro Fe-S Cluster Assembly Assay<!><!>ISCU and HSC20 Stimulate the ATPase Activity of mtHSP70<!><!>DISCUSSION<!>

<p>Fe-S clusters are ancient protein prosthetic groups that participate in a wide variety of biological processes, including electron transfer, substrate binding and activation, redox catalysis, DNA replication and repair, regulation of gene expression, and tRNA modification (1–3). In bacteria, three Fe-S cluster biogenetic systems have been discovered: the NIF (nitrogen fixation) system involved in maturation of nitrogenase, the SUF (sulfur mobilization) system encoded by the suf operon and highly active under oxidative stress conditions, and the ISC2 (iron sulfur cluster) system, the general Fe-S cluster biosynthetic pathway (for review, see Refs. 4, 5, and 7–10). Among the three systems, the ISC system is the best-studied and is believed to be the "housekeeping" biosynthetic system (5). The isc operon of Escherichia coli codes for several proteins: a repressor (IscR), a cysteine desulfurase (IscS), a scaffold protein (IscU), a protein proposed to be an alternative scaffold (IscA), a DnaJ-type co-chaperone (HscB), a DnaK-type chaperone (HscA), a ferredoxin (Fdx), and a protein of uncertain function (IscX). By catalyzing the conversion of l-cysteine to l-alanine, the pyridoxal-5′-phosphate-dependent enzyme IscS generates S0, which is transferred to Cys-328 to form a persulfide and then transferred to IscU (11, 12). Iron is added to form a [2Fe-2S] cluster. IscU-[2Fe-2S] then binds to HscB, which targets it to the HscA-ATP complex. In a reaction involving hydrolysis of ATP, the cluster is transferred to an acceptor protein such as apoferredoxin (13, 14). We have shown by NMR studies that E. coli IscU populates two interconvertible conformational states: a more structured state (S) and a partially disordered state (D) (15). The two states play different roles in the cycle of Fe-S cluster assembly and transfer. The D-state is the substrate for IscS (16); the S-state is the form that binds a [2Fe-2S] cluster (17) and binds preferentially to HscB (15, 18). Upon hydrolysis of ATP, HscA binds to the D-state of IscU, ensuing complete release of the cluster to an acceptor protein.</p><p>ISCU-type proteins are highly conserved throughout living systems (Fig. 1A). Mitochondria contain an ISC-type Fe-S cluster assembly and transfer system, parts of which are homologous to the ISC system of prokaryotes (6). In humans it has been proposed that this system involves a scaffold protein (ISCU) (19), a cysteine desulfurase consisting of two subunits, NFS1 (homologous to IscS) and ISD11 (no bacterial homologue found), a DnaJ-type co-chaperone (HSC20), and a DnaK-like chaperone (mtHSP70) (6, 19). These mitochondrial proteins are synthesized in the cytoplasm with N-terminal extensions that facilitate mitochondrial entry and are cleaved off upon protein maturation in mitochondria (20). The mature form of human ISCU, which shares 77% sequence identity with E. coli IscU (Fig. 1A), has been shown to play an important role in cellular iron homeostasis (21). A tissue-specified splicing mutation of human ISCU has been associated with the disease ISCU myopathy (22). It has been proposed that the small protein ISD11 stabilizes the cysteine desulfurase NFS1 (23) and is important for Fe-S cluster assembly and cellular iron homeostasis (24, 25). Although it has been shown in an in vitro assay that the ISCU-NFS1-ISD11 complex can assemble Fe-S clusters (26), little is known about the interactions among these three proteins. Human mtHSP70 (also termed Mortalin, PBP74, GRP75, or HSPA9) appears to be a multifunctional DnaK-type chaperone (27) (unlike E. coli HscA, which is involved only in Fe-S cluster biogenesis). Human mtHSP70 is known to perform other cellular functions, including protein folding, intracellular trafficking, antigen processing, and aging (28, 29). It has been shown that mtHSP70 binds a variety of substrates, including cancer suppressor protein p53 and Parkinson disease-related protein DJ-1, and mtHSP70 has been associated with Alzheimer disease (30, 31). Because mtHSP70 is the only Hsp70 chaperone present in human mitochondria, it has been suggested that it plays a role in the human ISC machinery similar to that of E. coli HscA (32). This hypothesis is supported by the finding that the mtHSP70 homolog in Saccharomyces cerevisiae (Ssc1) participates in Fe-S cluster biosynthesis when the specialized Hsp70 chaperone (Ssq1) is knocked out (33). HSC20 recently has also been found to be involved in human mitochondrial Fe-S cluster biogenesis (34).</p><!><p>Alignment of sequences of IscU homologues and structural representation of ISCU. A, alignment of sequences of IscU homologues. Analysis (15) of a much larger set of aligned sequences than those represented here showed that the residues highlighted in black were identically conserved, those in blue were conserved, and those in yellow were semi-conserved. The conserved cysteine residues are marked with red arrows, and the conserved LPPVK motif recognized by chaperone proteins is boxed in magenta. Mitochondrial proteins contain an N-terminal sequence that targets ISCU to cross the inner mitochondrial membrane. Excluding this region (boxed in red), human ISCU and E. coli IscU share 77% sequence identity. The numbering systems for mature ISCU and E. coli IscU are identical. Abbreviations used are: Av, A. vinelandii; Ec, Escherichia coli; Hi, H. influenzae; Aa, A. aeolicus; Hs, Homo sapiens; Mm, M. musculus; Sc, S. cerevisiae. B, solution structure of Zn2+ bound M. musculus ISCU (PDB code 1WFZ) (47), which shares ∼98% sequence identity with human ISCU, provides insights into the structure of human ISCU. Similar to other IscU homologues, M. musculus ISCU consists of three β-strands and four α-helices. Residues mutated in this study (Asp-39, Asn-90, and His-105) are shown in red stick format. Asp-39 and His-105 are close to the Zn2+ binding site, and Asn-90 is a solvent-exposed polar residue located in a hydrophobic region. All three residues are highly conserved among IscU homologues, and amino acid substitutions at these sites in E. coli IscU have been shown to perturb the position of the D ⇄ S conformational equilibrium (35). Another important residue Trp-76 is shown in red stick format.</p><!><p>Given the sequence similarities of the bacterial and human proteins, it was of interest to determine, whether the human scaffold protein (ISCU) shares the conformational duality of the E. coli ortholog (IscU) and, if so, whether the two states exhibit differential interaction with the cysteine desulfurase, co-chaperone, and chaperone proteins. We report here that the human scaffold protein for Fe-S cluster biogenesis (ISCU), like its E. coli counterpart (IscU), populates two conformational states, a more structured state (S) and a partially disordered state (D). However, the relative population of the S-state of human ISCU is lower (28%) than the corresponding population of E. coli IscU (80%) under comparable solution conditions. Human cysteine desulfurase (NFS1) alone and in the NFS1-ISD11 complex were found to bind preferentially to the D-state of ISCU. However, ISCU does not interact directly with ISD11. The co-chaperone (HSC20) was found to bind preferentially to the S-state of ISCU, whereas the chaperone (mtHSP70) was found to bind preferentially to the D-state of ISCU. HSC20 activated the ATPase activity of mtHSP70, and this activation was greatly increased by the addition of ISCU.</p><!><p>The cDNAs encoding the human mitochondrial proteins ISCU, NFS1, ISD11, mtHSP70, and HSC20 were ordered from the Mammalian Gene Collection (Thermo Fisher; Pittsburgh, PA). The genes coding for bacterial proteins were isolated by PCR directly from E. coli genomic DNA (Sigma). The clones for the mitochondrial proteins were designed to yield the mature protein sequences; the gene sequence coding for N-terminal mitochondrial targeting peptide was excluded (Fig. 1A, red box), and the clones coded for a SUMO fusion containing an N-terminal His-tag. The genes were cloned into the pE-SUMO-Kan vector (Lifesensors; Malvern, PA) by using BsaI and XhoI restriction sites in the PCR gene fragments and vector. DNA primers used in these experiments were ordered through either the University of Wisconsin-Madison Biotechnology Center or Integrated DNA Technologies, Inc. (Coralville, IA). Restriction enzymes were purchased from either Promega (Madison, WI) or New England Biolabs (Ipswich, MA). All PCR DNA primers used for cloning included nine additional base pairs at the 5′ ends upstream of the restriction sites to make digestion more efficient at the termini of PCR products. DNA ligation and construction of the expression plasmid were carried out in 10-μl reactions with a PCR-based ligation thermo cycling program (30 s at 30 °C and 30 s at 10 °C repeated 800 times for 12 h). The ligation reaction was heat-inactivated at 65 °C for 25 min and then used to transform chemically competent 10G cells (Lucigen; Madison, WI), which were plated onto YT plates (containing 50 μg/ml kanamycin) and incubated at 37 °C overnight. An E. coli recombinant colony archive was constructed by picking individual colonies and storing them in 30 μl of 20% sterile glycerol. 3-μl aliquots were removed for use in a PCR colony screen (20-μl reaction using Promega 2× PCR master mix) to identify positive clones. The original PCR primers used for isolating the target genes were employed in this PCR colony screen, and the reactions were analyzed on 1% agarose gels in TAE buffer (40 mm Tris acetate, 1 mm EDTA, pH 8.0). To prepare plasmids for DNA sequence analysis, the E. coli colony glycerol stocks that yielded positive recombinants according to the PCR colony screens were grown overnight at 37 °C in 2–3 ml of CircleGrow® broth (MP Biomedicals; Santa Ana, CA) in the presence of 50 μg/ml kanamycin. All DNA sequencing reactions were carried out on a Bio-Rad Dyad Peltier Thermal cycler at the University of Wisconsin-Madison Biotechnology Center, and SeqMan software (DNASTAR; Madison, WI) was used to analyze and identify targets with the correct DNA sequence.</p><p>We chose sites for production of single amino acid variants of ISCU to match those known to affect the D ⇄ S conformational equilibrium of E. coli IscU (35). The positions of these residues on Mus musculus Zn2+ bound ISCU (a model for the S-state) are shown in Fig. 1B (PDB code 1WFZ). M. musculus ISCU shares 98% sequence identity with human ISCU (Fig. 1A). Genes for these variants were produced by using the Polymerase Incomplete Primer Extension (PIPE) site-directed mutagenesis method (36).</p><!><p>Single colonies containing validated genes for the target proteins (human ISCU variants, NFS1, ISD11, mtHSP70, and HSC20 and E. coli IscS) were picked from the YT or MDAG plates (37) and grown in 1 ml of CircleGrow® broth or YT with 1% glucose (plus appropriate antibiotics) for 1–3 h at 37 °C at 250 rpm and then transferred to 50–100 ml of MDAG medium (with appropriate antibiotics) and grown overnight at 25 °C. For large scale protein production, 1 liter of unlabeled Terrific Broth (TB) auto-inducing medium or M9 isotopic medium was prepared, and 500-ml aliquots were transferred into sterile PET soda bottles (38, 39). Each 500-ml aliquot was inoculated with 10 ml of the overnight MDAG starter culture, and the cell cultures were grown at 37 °C (250–320 rpm) for 2–5 h before dropping the growth temperature to 10–25 °C for 24–36 h. For induction with isopropyl-1-thio-β-d-galactoside (IPTG), the cell cultures were grown to A600 of 1.0–1.5; then the temperature was dropped to 10–25 °C, and IPTG (0.1–0.2 mm final concentration) was added when the temperature had stabilized (after about 15–30 min). For stable isotope labeling, we used an M9-based medium consisting of 100 ml/liter 10× M9 salts (70 g/liter Na2HPO4, 30 g/liter KH2PO4, and 5 g/liter NaCl), 1 ml of 1000× metal mix, 1 ml of vitamin mixture (40), 30 mg/liter thiamine, 0.5 ml of 0.2 m CaCl2 (0.1 mm final concentration), 2–5 drops of sterile antifoam, 2 ml of 1 m MgSO4 (2 mm final concentration), [U-13C]glucose (2–4 g/liter), 15NH4Cl (1 g/liter), plus the appropriate antibiotics (chloramphenicol to 35 μg/ml and kanamycin to 50–100 μg/ml). At the end of cell growth, the cultures were harvested by centrifugation for 30 min at 4000 × g in a centrifuge with a JS-4.0 rotor (Beckman Coulter; Brea, CA). The cell pastes were stored at −80 °C until needed for protein purification.</p><!><p>The composition of the 1st immobilized metal affinity chromatography (IMAC) buffer was 20 mm Tris-HCl, pH 8, 300–500 mm NaCl, 0.1% Nonidet P-40, 1–2 mm β-mercaptoethanol or DTT, 1 mm phenylmethanesulfonyl fluoride (PMSF), 5–10% glycerol, and 5 mm imidazole. The composition of the 2nd IMAC buffer was the same as the 1st IMAC buffer except that it contained 250 mm imidazole. The SUMO-fusion cleavage buffer contained 20 mm Tris buffer at pH 8, 150 mm NaCl, 2 mm DTT (or β-mercaptoethanol), and 5–10% glycerol. The TND buffer consisted of 50 mm Tris-HCl, pH 8, containing 150 mm NaCl, 5 mm DTT, and 0.3% NaN3. The TKDM buffer consisted of 50 mm Tris-HCl, pH 7.5, containing 150 mm KCl, 5 mm DTT, and 10 mm MgCl2.</p><!><p>All E. coli cell pastes were quickly thawed either on ice or at room temperature and then resuspended in 60–70 ml of lysis buffer: 1st IMAC buffer supplemented with Benzonase (Novagen, Millipore; Billerica, MA) or OmniCleave nuclease (Epicenter, Illumina; Madison, WI), rLysozyme (Novagen), RNase (Qiagen; Valencia, CA), and 0.1% Nonidet P-40 (Sigma). To break open the resuspended cells, we used sonication with a total time of 15–30 min at 4 °C, with a duty cycle of 2 s on and 4 s off. Cell lysates were clarified by high-speed centrifugation at 25,000 rpm for 30 min in a centrifuge with a JA 30.5Ti rotor (Beckman Coulter). The clear cell lysate was then treated with 0.1% w/v polyethylene imine (PEI) to precipitate RNA and then was spun again for 30 min at 25,000 rpm and eluted. DTT was added to a level of 2 mm to ensure reduction of cysteines. The cell lysate was refrigerated at 4 °C, and (NH4)2SO4 was added to 70% (w/v) saturation to precipitate total protein and to remove PEI. Then the sample was spun at 25,000 rpm for 30 min. The protein pellet was resuspended in 30–50 ml of 1st IMAC buffer (without NaCl but with 2 mm DTT), and any debris was discarded after a final centrifugation at 25,000 rpm for 30 min. The clarified protein solution was loaded onto a Qiagen Superflow FF or Ni-Sepharose column (GE Healthcare) IMAC resin at 1–5 ml/min. The IMAC column was washed first with ∼10 column volumes of 1st IMAC buffer and second with 5–10 column volumes of wash buffer (1st IMAC buffer + 30 mm imidazole). The target protein was eluted with the 2nd IMAC buffer, and fractions were collected. SDS-PAGE was used to analyze and assess the purity of the eluted target protein in the collected fractions.</p><p>The His-tagged, N-terminal SUMO fusion protein was digested with 0.5 mg of SUMO protease. The reaction was carried out in SUMO-fusion cleavage buffer either by desalting the fusion protein by size exclusion chromatography and adding SUMO protease or, more usually, by dialyzing the fusion protein in the presence of SUMO protease overnight at 4 °C against the cleavage buffer. The cleaved sample was loaded onto a freshly equilibrated subtractive IMAC column, which bound the cleaved His-tagged SUMO domain and allowed the cleaved target protein to pass through to a fraction collector. The purities of the target protein fractions were assessed by SDS-PAGE. Mass spectrometry was used to determined target protein masses and the level of stable isotope incorporation.</p><!><p>For NMR samples the TND and TKDM buffers were modified to contain 10% D2O for the frequency lock. All NMR spectra were collected on 600 MHz (1H) Bruker BioSpin (Billerica, MA) NMR spectrometers equipped with a z-gradient cryogenic probe. All sample temperatures were regulated at 25 °C. NMRPipe software (41) was used to process the raw NMR data, and SPARKY software (42, 43) was utilized to visualize and analyze the processed NMR data.</p><p>1H,15N HSQC spectra of wild-type (WT) and variant ISCU samples were collected with 0.3 mm U-15N-ISCU in TND buffer. To monitor the effects of added unlabeled NFS1 or ISD11, we first collected a 1H,15N HSQC spectrum 0.5 mm U-15N-ISCU in TND buffer. Then an equal volume of 0.5 mm unlabeled NFS1 was first added, and a 1H,15N TROSY-HSQC spectrum was acquired. Next, an equal volume of 0.5 mm unlabeled ISD11 in TND buffer was added, and another 1H,15N TROSY-HSQC spectrum was acquired. Because of the dilution effect, the intensities of the 1H,15N peaks from ISCU diminished by a factor of two after the addition of NFS1 and by a factor of 3 after the addition of ISD11. However, the quantity of interest was the effect on the relative population of the S-state.</p><p>The titrations with mtHSP70 were started by collecting 1H,15N HSQC spectra of samples of 0.5 mm U-15N-ISCU and 0.5 mm U-15N-ISCU(N90A) in TKDM buffer. Then aliquots of unlabeled 0.5 mm mtHSP70 in TKDM buffer were added to each sample, and 1H,15N TROSY-HSQC spectra were acquired. Because of the dilution effect, the intensities of the 1H,15N peaks from ISCU diminished by a factor of two at equimolar ISCU:mtHSP70. Again, the quantity of interest was the effect on the relative population of the S-state.</p><p>The titrations with HSC20 were started by collecting 1H,15N TROSY-HSQC spectra of samples of 0.5 mm U-15N-ISCU or U-15N-ISCU(N90A) in TKDM buffer. 1H,15N TROSY-HSQC spectra were acquired after the addition of aliquots of 0.4 mm unlabeled HSC20 in TKDM buffer. Because of the dilution effect, the intensities of the 1H,15N peaks diminished by a factor of 2.25 at equimolar ISCU:HSC20. However, the quantity of interest was the ratio [S]/([S] + [D]).</p><!><p>We were able to assign NMR signals of ISCU(N90A), a variant that fully populates the S-state. We collected three-dimensional CBCACONH and HNCACB spectra from a sample of ISCU(N90A) labeled uniformly with 13C and 15N and used the data to carry out sequential backbone assignments.</p><!><p>We used the Newton software package (44) to calculate the relative intensities of the Trp-76 cross-peaks from the S- and D-states. Newton carries out fast maximum likelihood reconstruction (FMLR) of two-dimensional NMR signals to provide rigorous signal intensities by fitting their position, amplitude, line width, and phase. The percent of ISCU in the S-state, %S, is given by Equation 1, where [S] is obtained from the intensity of the S peak and [D] is from the intensity of the D peak from Trp-76. We carried out three or more independent FMLR analyses of each spectrum to determine reproducibility and estimate errors (shown as error bars).</p><!><p>The sample buffer used in circular dichroism (CD) experiments contained 20 mm NaH2PO4 and 50 mm NaCl at pH 8. The solutions were placed in 1-mm path length quartz cuvettes. The concentration of ISCU variants was 20 μm. Far-UV CD spectra of ISCU variants were collected with an Aviv 202SF CD spectrophotometer (Aviv Biomedical; Lakewood, NJ) at 25 °C. Secondary structure content was estimated from the CD spectra by using K2D2 software (45).</p><!><p>Analytical gel-filtration studies were conducted with Hi-Load 16/60 Superdex 75 Column (GE Healthcare) at room temperature. To investigate the interaction between HSC20 and ISCU (WT or N90A), a 2:1 (molar ratio) mixture of HSC20:ISCU in TKDM buffer was injected. The protein sample was eluted at 1 ml/min flow rate with TKDM buffer as the elution buffer, and 2 ml fractions were collected using an automatic fraction collector (GE Healthcare). Eluted fractions were analyzed by SDS-PAGE.</p><!><p>ATPase assays were carried out in TKDM buffer containing 0.1 mm ATP. The ATPase activity of mtHSP70 was determined at 25 °C by using an EnzCheck Phosphate Assay kit (Invitrogen) to measure the rate of phosphate release rate as described previously (46).</p><!><p>A published protocol (47) was used to assemble Fe-S clusters in vitro. All samples were prepared in an anaerobic chamber (Coy Laboratory; Farmingdale, NY) filled with 90% N2 gas and 10% H2 gas. The reconstitution mix in TND buffer at pH 7.5 consisted of 50 μm ISCU or ISCU(N90A), 250 μm Fe2(NH4)2(SO4)2, and 1 μm cysteine desulfurase (NFS1, NFS1-ISD11, or E. coli IscS). The reaction was initiated by adding 250 μm l-cysteine into the reconstitution mix to make the final volume equal 1 ml. The reaction was carried out at 25 °C in a 10-mm path length quartz cuvette sealed with a rubber septum. Spectra were collected on a UV-1700 UV-visible spectrophotometer (Shimadzu; Kyoto, Japan) equipped with a temperature control utility. UVProbe 2.21 software (Shimadzu) was used to collect and analyze the data.</p><!><p>Evidence for the structural heterogeneity of ISCU came from 1H,15N HSQC NMR spectra that exhibited two sets of peaks for certain residues. The most prominent of these was the doubled 1H,15N cross-peak from the side chain of Trp-76, the only tryptophan residue in the protein (boxed signals on Fig. 2A). The spectral analysis was clarified by comparison of 1H,15N HSQC spectra of ISCU (Fig. 2A) with those of the four single-site mutants (Fig. 2B). Variants ISCU(D39V) and ISCU(N90A) yielded 1H,15N HSQC spectra with sharp, well dispersed peaks, as expected for a well structured protein, whereas variants ISCU(D39A) and ISCU(H105A) yielded 1H,15N HSQC spectra with broader, poorly dispersed peaks (particularly in the 1H dimension), as expected for a partially disordered protein. The spectrum of wild-type ISCU exhibited both sets of peaks. We thus assigned the sharper set of peaks to the structured state (S) and the broader set of peaks to the partially disordered state (D). Comparison of the Trp-76 peaks from ISCU (Fig. 2A) with those from the variants (Fig. 2C) allowed us to assign signals to the individual states. The %S values for the ISCU variants studied here are collected in Table 1.</p><!><p>NMR evidence that ISCU exists in solution as two slow interchanging conformational states and that the S ⇄ D equilibrium is perturbed by single amino acid substitutions. A, 1H,15N HSQC NMR spectrum of ISCU. Because ISCU contains only one tryptophan (Trp-76), the presence of two cross-peaks in the boxed region and inset indicates the existence of two different conformational states. Assignments of individual peaks to the S- and D-states are indicated. B, 1H,15N HSQC NMR spectra of ISCU variants with shifted S ⇄ D equilibria. Whereas the substitutions D39V and N90A stabilize the S-state, substitutions D39A and H105A stabilize the D-state. C, expansions of the spectra in B show the Trp-76 side chain signals used to quantify the relative populations of the S- and D-states. All NMR spectra were collected at 600 MHz (1H) at 25 °C with solutions at pH 8.0.</p><p>Properties of wild-type (WT) human ISCU and single-site amino acid variants at pH 8 and 25 °C</p><!><p>We used CD spectroscopy to investigate the secondary structure of the ISCU variants. Far-UV (200–260 nm) CD spectra of ISCU variants that stabilize the S-state as shown by NMR exhibited secondary structure, whereas variants that stabilize the D-state as shown by NMR yielded CD spectra that could not be interpreted in terms of secondary structure (Fig. 3; Table 1). The CD spectrum of wild-type ISCU was consistent with the mixed population determined by NMR of ∼28% S-state and 72% D-state. The CD spectra of the structured variants of human ISCU (D39V and N90A) as analyzed by K2D2 software (43) yielded ∼25% α-helix and ∼20% β-strand (Table 1). By comparison, the solution structure of Zn2+-bound M. musculus IscU (PDB code 1WFZ, 10.2210/pdb1wfz/pdb), a model for the S-state, contained ∼41% α-helix and ∼20% β-strand.</p><!><p>Far-UV CD spectra of ISCU variants. All spectra were collected at 25 °C with solutions at pH 8. The CD spectra of ISCU(N90A) (red) and ISCU(D39V) (blue) indicate the presence of secondary structure, but the CD spectra of ISCU(H105A) (black) and ISCU(D39A) (green) suggest that little secondary structure is present. The CD spectrum of wild-type ISCU (magenta) is intermediate between those of the variants stabilizing the S- and D-states.</p><!><p>Two-dimensional 15N TROSY-HSQC NMR spectra of U-15N-ISCU were collected before and after adding a stoichiometric amount of unlabeled NFS1. The %S determined by FMLR analysis of the relative intensities of the Trp-76 peaks assigned to the S- and D-states (Fig. 4, A and B) decreased from ∼27 to ∼7% upon adding 1 eq of NFS1 and remained unchanged after the subsequent addition of 1 eq of ISD11 (Fig. 4D). This result indicates that NFS1 binds preferentially to the D-state of ISCU. The addition of unlabeled ISD11 to U-15N-ISCU led to no changes in the two-dimensional 15N TROSY-HSQC NMR spectra (Fig. 4, C and E), which indicated that ISCU and ISD11 do not interact directly.</p><!><p>Evidence that NFS1 binds preferentially to the D-state of ISCU. A, top panel, two-dimensional 1H,15N HSQC spectrum of U-15N-ISCU. Middle panel, overlay of the two-dimensional 1H,15N TROSY-HSQC spectra of U-15N-ISCU alone (red) and U-15N-ISCU (blue) mixed with a stoichiometric amount of unlabeled NFS1. Bottom panel, overlay of two-dimensional 1H,15N TROSY-HSQC spectra of U-15N-ISCU (red) and the same sample mixed with stoichiometric amounts of unlabeled NFS1 and ISD11 (green). B, expansions show the 1H,15N peaks from the indole ring of Trp-76 of ISCU. Top panel, U-15N-ISCU. Middle panel, U-15N-ISCU mixed with equimolar unlabeled NFS1. Bottom panel, U-15N-ISCU mixed with equimolar unlabeled NFS1 and ISD11. C, top panel, two-dimensional 1H,15N HSQC spectrum of U-15N-ISCU. Bottom panel, two-dimensional 1H,15N HSQC spectrum of U-15N-ISCU mixed with equimolar unlabeled ISD11. All NMR spectra were collected at 600 MHz (1H) at 25 °C with samples at pH 8.0. D and E, %S calculated by FMLR analysis of the intensities of Trp-76 cross-peaks assigned to the S- and D-states under the conditions indicated.</p><!><p>To determine whether mtHSP70, the only Hsp70 chaperone protein in mitochondria (27), interacts with ISCU, we used two-dimensional 15N TROSY-HSQC NMR spectroscopy to follow the titration of U-15N-ISCU with unlabeled mtHSP70. The addition of mtHSP70 led to a progressive decrease in %S by FMLR analysis (Fig. 5C and Table 2). %S decreased from 27% in the absence of mtHSP70 to 1.2% at equimolar ISCU:mtHSP70 (Fig. 5, A and B). The results indicate that mtHSP70 binds preferentially to the D-state of ISCU. Several 1H,15N peaks from ISCU were found to be perturbed in the presence of 0.2 eq of mtHSP70 (Fig. 6A).</p><!><p>Evidence that mtHSP70 binds preferentially to the D-state of ISCU. A, two-dimensional 1H,15N HSQC spectra of U-15N-ISCU (left panel), U-15N-ISCU sample mixed with 0.2 eq of unlabeled mtHSP70 and diluted by a factor of 1.2 (middle panel), U-15N-ISCU mixed with 1.0 eq of mtHSP70 and diluted by a factor of 2 (right panel). B, expansions of the Trp-76 1H,15N cross-peaks from the spectra in A. C, %S calculated by FMLR analysis of the intensities of the Trp-76 cross-peaks assigned to S and D under the conditions indicated. D, two-dimensional 1H,15N TROSY-HSQC spectra of U-15N-ISCU(N90A) (left panel), U-15N-ISCU(N90A) plus 0.2 eq of unlabeled mtHSP70 and diluted by a factor of 1.2 (middle panel), U-15N-ISCU(N90A) mixed with 1.0 eq of mtHSP70 and diluted by a factor of 2 (right panel). E, expansions of the Trp-76 1H,15N cross-peaks from the spectra in D. F, %S calculated by FMLR analysis (44) of the intensities of the Trp-76 cross-peaks assigned S and D under the conditions indicated. G, two-dimensional 1H,15N TROSY-HSQC spectra of U-15N-ISCU(H105A) (left panel), U-15N-ISCU(H105A) mixed with 0.2 eq of unlabeled mtHSP70 and diluted by a factor of 1.2 (middle panel), U-15N-ISCU(H105A) plus 1.0 eq of unlabeled mtHSP70 and diluted by a factor of 2 (right panel). H, expansions of the Trp-76 1H,15N cross-peaks from the spectra in G. I, %S calculated by FMLR analysis of the intensities of the Trp-76 cross-peaks assigned S and D under the conditions indicated. All NMR spectra were collected at 600 MHz (1H) at 25 °C with samples at pH 8.0.</p><p>Effect of added mtHSP70 on the %S of ISCU variants</p><p>Values in the table represent %S (determined by NMR) = ([S]/([S] + [D])) × 100 (%).</p><p>Effect of added mtHSP70 on selected peaks from 1H,15N TROSY-HSQC spectra of U-15N-ISCU and U-15N-ISCU(N90A). A, U-15N-ISCU alone (red) and U-15N-ISCU with 0.2 eq of added unlabeled mtHSP70 (blue). B, U-15N-ISCU(N90A) alone (red) and U-15N-ISCU(N90A) with 0.2 eq of added unlabeled mtHSP70 (blue).</p><!><p>We hypothesized that the interaction between mtHSP70 and ISCU(N90A), which populates mainly the S-state, would be weaker than its interaction with ISCU. As anticipated, FMLR analysis of 15N TROSY-HSQC NMR spectra (Fig. 5, D and E) showed a much smaller effect: %S decreased from 95% in the absence of mtHSP70 to 83% at equimolar mtHSP70:U-15N-ISCU(N90A) (Fig. 5F). The magnitudes of the chemical shift perturbations upon adding mtHSP70 were smaller for U-15N-ISCU(N90A) than for U-15N-ISCU (Fig. 6B).</p><p>We also investigated the interaction between mtHSP70 and ISCU(H105A), a variant that favors the D-state. The %S of U-15N-ISCU(H105A) decreased from 12 to 0.54% upon the addition of equimolar mtHSP70 (Fig. 5, G–I; Table 2). The results again confirm that mtHSP70 preferentially binds to the D-state of ISCU.</p><!><p>Analytical gel-filtration chromatography was employed to investigate the interaction between ISCU and HSC20. Upon elution of a 2:1 (molar ratio) HSC20:ISCU mixture, a peak emerged at ∼70 ml. This elution volume corresponds to the expected molecular mass of the HSC20-ISCU complex (∼37 kDa) (Fig. 7A). SDS-PAGE of the elution fractions confirmed that the peak eluting at 70 ml contained both ISCU and HSC20 (Fig. 7B). We further followed two-dimensional 15N TROSY-HSQC NMR spectra of U-15N-ISCU upon titration with unlabeled HSC20 (Fig. 8, A and B). FMLR analysis showed that %S increased from ∼22 to ∼31% upon the addition of equimolar HSC20 (Fig. 8C). The results indicate that HSC20 binds preferentially to the S-state of ISCU, in analogy to the finding that E. coli HscB binds preferentially to the S-state of IscU (15). By transferring the backbone assignments for ISCU(N90A) to the 1H,15N HSQC spectrum from the S-state of ISCU, we were able to follow chemical shift perturbations ΔδHN (as given by Equation 2) of U-15N-ISCU upon titration with unlabeled HSC20 (Fig. 8, G and I), where ΔδH and ΔδN are the chemical shift changes in the 1H and 15N dimensions, respectively.</p><!><p>Analytical gel filtration and SDS-PAGE show interaction between HSC20 and ISCU variants. A, analytical gel-filtration elution profiles of ISCU alone (blue line), HSC20 alone (red line), and 2:1 mixture of HSC20:ISCU (black line). The peak at ∼70 ml elution volume (indicated by the star) is assigned to the HSC20-ISCU complex. B, SDS-PAGE of the gel-filtration elution fractions collected between 70 and 80 ml from the 2:1 HSC20:ISCU sample. The protein bands at ∼20 and ∼15 kDa correspond to HSC20 and ISCU, respectively. C, analytical gel-filtration elution profiles of ISCU(N90A) alone (blue line), HSC20 alone (red line), and 2:1 mixture of HSC20:ISCU(N90A) (black line). D, SDS-PAGE of the gel-filtration elution fractions collected between 70 and 82 ml from the 2:1 HSC20:ISCU(N90A) sample. The protein bands at ∼20 and ∼15 kDa correspond to HSC20 and ISCU(N90A), respectively.</p><p>Interaction between HSC20 and ISCU. A, two-dimensional 1H,15N TROSY-HSQC spectra of U-15N-ISCU (left panel), U-15N-ISCU in the presence of equimolar unlabeled HSC20 and diluted by a factor of 2.25 (middle panel), overlay of the NMR spectra from the left and middle panels (right panel). B, expansions of the Trp-76 1H,15N peaks from the spectra in A. C, %S calculated from FMLR analysis of the intensities of the Trp-76 cross-peaks assigned S and D under the conditions indicated. D, 1H,15N TROSY-HSQC spectra of U-15N-ISCU(N90A) (left panel), U-15N-ISCU(N90A) in the presence of 1 eq of unlabeled HSC20 and diluted by a factor of 2.25 (middle panel), overlay of the NMR spectra from the left panel and middle panels (right panel). E, expansions of the Trp-76 1H,15N peaks from the spectra in D. F, %S calculated from FMLR analysis of the intensities of Trp-76 cross-peaks assigned S and D under the conditions indicated. G, chemical shift perturbation of ISCU signals (ΔδHN) upon the addition of 1.0 eq of HSC20 plotted as a function of ISCU residue number. Red triangles indicate residues whose NMR peaks were broadened beyond detection upon addition of HSC20. H, chemical shift perturbations (ΔδHN) for residues of ISCU(N90A) upon the addition of 1 eq of HSC20. I, chemical shift perturbations of ISCU signals resulting from HSC20 binding mapped onto the structure of Zn2+ bound M. musculus ISCU (PDB code 1WFZ) (47). Residues with ΔδHN > 0.04 ppm are colored blue; residues whose NMR peaks were broadened beyond detection are colored red. J, chemical shift perturbations of ISCU(N90A) resulting from HSC20 binding mapped on the structure of Zn2+ bound M. musculus ISCU with color coding as in I.</p><!><p>Examples of the several peaks from U-15N-ISCU that exhibited chemical shift perturbation upon binding HSC20 are shown in Fig. 9A. The residues of ISCU showing the largest chemical shift perturbations (Leu-31 and Val-32) (Fig. 8G) correspond to hydrophobic residues on the first β-strand of the three-dimensional structure of M. musculus Zn2+-ISCU (PDB code 1WFZ) (Fig. 8I). We speculate that the first β-strand of ISCU provides the binding interface for the ISCU-HSC20 interaction.</p><!><p>Two-dimensional backbone 1H,15N NMR peaks from U-15N-ISCU variants corresponding to selected residues (Gly-30, Leu-31, Asn-90, and Lys-78, from left to right) that exhibit chemical shift changes upon binding HSC20. A, U-15N-ISCU alone (red) and with 1.0 eq of unlabeled HSC20 (blue). B, U-15N-ISCU(N90A) alone (red) and with 0.5 eq (green) and 1.0 eq (blue) of unlabeled HSC20.</p><!><p>We also investigated the interaction between HSC20 and the S-state favoring ISCU variant ISCU(N90A). Similar to ISCU, analytical gel-filtration results showed the elution of a peak corresponding to the expected mass of the HSC20-ISCU(N90A) complex (∼37 kDa) (Fig. 7C). SDS-PAGE of the eluted fractions confirmed that this peak contained both ISCU and HSC20 (Fig. 7D). Two-dimensional 15N TROSY-HSQC NMR spectra of U-15N-ISCU(N90A) were acquired as a function of added unlabeled HSC20 (Fig. 8, D and E). Several peaks from U-15N-ISCU(N90A) exhibited chemical shift perturbations upon binding HSC20, and examples of these are shown in Fig. 9B. %S increased from ∼94 to ∼99% after the addition of 1 eq of HSC20, confirming that HSC20 binds preferentially to the S-state of ISCU(N90A) (Fig. 8F). Although more peaks of ISCU(N90A) were broadened beyond detection upon binding HSC20 as indicative of a tighter complex, the pattern of chemical shift perturbations (Fig. 8H) was similar to that for ISCU, suggesting that the first β-strand of ISCU(N90A) interacts with HSC20 (Fig. 8J).</p><!><p>To investigate the physiological importance of these findings, we carried out in vitro Fe-S cluster assembly assays. The UV spectra of the assembly mixture collected as a function of time showed the growth of peaks at 456 and 400 nm, which are characteristic for [2Fe-2S] and [4Fe-4S] clusters, respectively (47). This result indicates that ISCU serves as the scaffold protein for both types of Fe-S cluster (Fig. 10A). We found that NFS1 alone could catalyze Fe-S cluster assembly on ISCU and that the addition of ISD11 increased the Fe-S cluster assembly rate (Fig. 10B). We also investigated in vitro Fe-S cluster assembly on ISCU catalyzed by E. coli cysteine desulfurase (IscS) and found that the assembly rate was faster than that catalyzed by an equivalent concentration of NFS1-ISD11 (Fig. 10C). To investigate the effect of the conformational states of ISCU on cluster assembly, we repeated the reaction replacing ISCU by ISCU(N90A), a variant that is primarily in the S-state. Compared with wild-type ISCU, cluster assembly on ISCU(N90A) occurred at a much slower rate (Fig. 10D). The addition of Zn2+, which stabilizes the S-state (data not shown), inhibited the rate of cluster assembly on ISCU (Fig. 10E).</p><!><p>In vitro Fe-S cluster assembly assays. A, UV-visible absorption spectra of ISCU during Fe-S cluster assembly catalyzed by NFS1-ISD11. Spectra were collected at 60-min intervals. Absorption (ABS) at 400 nm and 456 nm are characteristic of [4Fe-4S] and [2Fe-2S] clusters, respectively. B, time course of Fe-S cluster assembly followed at 456 nm. Black line, reaction containing ISCU and catalyzed by NFS1 alone. Red line, reaction containing ISCU and catalyzed by NFS1-ISD11. C, time course of Fe-S cluster assembly followed at 456 nm. Black line, reaction containing ISCU and a catalytic amount of E. coli IscS. Red line, reaction containing ISCU and a catalytic amount of NFS1-ISD11. D, time course of Fe-S cluster assembly followed at 456 nm. Red line, reaction containing ISCU and a catalytic amount of NFS1-ISD11. Black line, reaction containing ISCU(N90A) and a catalytic amount of NFS1-ISD11. E, time course of Fe-S cluster assembly followed at 456 nm in the presence of a catalytic amount of NFS1-ISD11. Red line, reaction containing ISCU. Black line, Reaction containing ISCU and 1 eq of Zn2+.</p><!><p>We found that mtHSP70 exhibited a basal ATPase activity of ∼0.10 ± 0.023 min−1 (Fig. 11A, black) which is lower than that reported for E. coli HscA (∼0.46 min−1) (49). The addition of either 6 μm HSC20 or 15 μm ISCU to 1 μm mtHSP70 increased the basal ATPase activity of mtHSP70 by factors of ∼1.7 and ∼4.5, respectively (Fig. 11A, red and blue). The addition of both 15 μm ISCU and 4 μm HSC20 increased the basal ATPase activity of mtHSP70 by a factor of ∼15 (Fig. 11A, green). Furthermore, we measured the effects of increasing concentrations of HSC20 alone (Fig. 11B), ISCU alone, and HSC20+ISCU (Fig. 11C) on the ATPase activity of 1 μm mtHSP70. Based on the double-reciprocal plot, ISCU alone elicited a maximal stimulation of ∼5 fold, and half-maximal stimulation occurred at ∼1.5 μm ISCU. In the presence of 5 μm HSC20, a maximal stimulation of ∼17 fold was observed upon adding ISCU, and the concentration of ISCU required for half-maximal stimulation was ∼2 μm (Fig. 11D). In the presence of 5 μm HSC20, the addition of 24 μm ISCU(N90A) increased the basal ATPase activity of mtHSP70 only by a factor of ∼7 (Fig. 11E). Based on the double-reciprocal plot, the maximal stimulation was ∼8 fold, and the concentration of ISCU(N90A) required for half-maximal stimulation was ∼3.5 μm (Fig. 11F).</p><!><p>Stimulation of the ATPase activity of mtHSP70 at 25 °C by ISCU and HSC20. A, time course of ATP hydrolysis catalyzed by 1 μm mtHSP70 alone (black line) in the presence of 6 μm HSC20 (red line), in the presence of 15 μm ISCU (blue line), or in the presence of 4 μm HSC20 + 15 μm ISCU (green line). B, ATPase activity of 1 μm mtHSP70 as a function of added HSC20. C, ATPase activity of 1 μm mtHSP70 as a function of added ISCU in the absence of HSC20 (black line) and in the presence of 4 μm HSC20 (red line). D, double-reciprocal plot of the data of C. E, ATPase activity of 1 μm mtHSP70 as a function of added ISCU(N90A) in the presence of 5 μm HSC20. F, double-reciprocal plot of the data of Fig. 11E.</p><!><p>The best studied ISC system for Fe-S cluster biosynthesis is from bacteria. An NMR structure has been determined for the S-state of E. coli IscU (50), and x-ray and NMR structures have been determined for Zn2+ complexes of Haemophilus influenzae IscU, Streptococcus pyogenes IscU (51), and Bacillus subtilis IscU (PDB code 1XJS, 10.2210/pdb1xjs/pdb). X-ray structures have been determined for the Aquifex aeolicus IscU-[2Fe-2S] complex (17), for E. coli IscS (12), for the E. coli IscS-IscU complex (53), for E. coli HscB (54), for the substrate binding domain of E. coli HscA complexed with the IscU recognition peptide ELPPVKIHC (55), and for E. coli IscA (56). In addition, NMR studies of the E. coli ISC system have elucidated the roles of the S and D conformational states of IscU in the cycle of Fe-S cluster assembly and delivery (15, 16, 18, 50, 57). Despite sequence similarities of the homologous human mitochondrial proteins, the only three-dimensional structure determined to date of a protein from the human ISC system is that of HSC20 (human HscB) (58).</p><p>Results presented here show strong parallels between the conformational properties of the human (ISCU) and E. coli (IscU) scaffold proteins as well as their functional properties. Both human ISCU and E. coli IscU can adopt two very different folded conformations; one more structured (S-state) and one partially disordered (D-state) (Fig. 2A). Thus, the scaffold protein can be categorized as a metamorphic protein (59). In both cases, the S ⇄ D equilibrium is affected by single site amino acid substitutions at positions 39, 90, and 105 (16) (Fig. 2, B and C). Human ISCU is much less structured (∼28%S) than E. coli IscU (∼80%S) under similar buffer, pH, and temperature conditions (35). Another interesting difference is that the D39A substitution, which stabilizes the S-state of E. coli IscU (16, 50) and has been found to stabilize Fe-S clusters in Azotobacter vinelandii IscU and Schizosaccharomyces pombe Isu1 (60), was found to favor the D-state of human ISCU. Because we had earlier seen a parallel between the S-state fraction and cluster stability in E. coli IscU (16), this result appears at odds with an earlier study, which reported that human ISCU(D39A) supported cluster formation, whereas ISCU did not (61).</p><p>Prior studies of the human cysteine desulfurase have focused on the NFS1-ISD11 complex, because of difficulties in isolating the two subunits (26, 62, 63). In this study we successfully expressed and purified the two proteins separately. We show here that isolated ISD11 does not interact directly with ISCU (Fig. 4C). On the other hand, isolated NFS1 binds to the D-state of ISCU (Fig. 4A, middle panel) as does the NFS1-ISD11 complex (Fig. 4A, bottom panel).</p><p>Human ISCU contains the conserved 99LPPVK103 motif (Fig. 1A) found in E. coli IscU, which is recognized by the substrate binding domain of E. coli HscA (64). We found that human mtHSP70 interacts with the D-state of ISCU (Fig. 5, A and B). This result is parallel to the interaction between E. coli HscA and the D-state of IscU (18). The addition of 1 molar eq of mtHSP70 shifted the S ⇄ D equilibrium of ISCU completely to the D-state (Fig. 5, A and B). By contrast, variant ISCU(N90A), which has a stabilized S-state, remained primarily in the S-state upon the addition of 1 molar eq of mtHSP70 (Fig. 5, C and D).</p><p>Human HSC20, the putative human homolog of the specialized DnaJ type co-chaperones, has been reported to be involved in human mitochondrial Fe-S cluster biogenesis and mitochondrial iron homeostasis (34). Although human HSC20 and E. coli HscB share high structural similarity, the former contains an extra N-terminal rubredoxin-like domain not present in E. coli HscB or S. cerevisiae Jac1 (58). We found that human HSC20 preferentially binds and stabilizes the S-state of ISCU (Fig. 8, A–F). This result is parallel to the preferential interaction between E. coli HscB and the S-state of IscU (15). Residues on the first β-strand of ISCU, namely Gly-30, Leu-31, and Val-32, exhibited large chemical shift perturbations upon the addition of HSC20 (Fig. 8, G–J). We speculate that these residues are involved in a hydrophobic interaction between ISCU and HSC20.</p><p>The cluster assembly assay indicated that both [2Fe-2S] and [4Fe-4S] clusters could be assembled on ISCU (Fig. 10A) as catalyzed by NFS1 alone or by the NFS1-ISD11 complex (Fig. 10B). The rate with NFS1-ISD11 was ∼27% faster than with NFS1 alone. We found that E. coli IscS also assembled clusters on human ISCU at an even faster rate than with NFS-ISD11 (Fig. 10C). This result is in agreement with a recent finding that the human NFS1-ISD11 complex, in the absence of frataxin or its bacterial homologue CyaY, exhibited lower cysteine desulfurase activity than E. coli IscS (62). As catalyzed by NFS-ISD11, ISCU(N90A), the variant with a stabilized S-state, assembled clusters 2.5 times more slowly than ISCU (Fig. 10D). Similar results have been reported for E. coli IscU variants with stabilized S-states (16). As with the E. coli system (16), Zn2+ can also stabilize the S-state of human ISCU (data not shown), and the addition of Zn2+ was found to inhibit cluster formation (Fig. 10E).</p><p>The ATPase assay showed that mtHSP70 has a basal ATPase activity of ∼0.10 min−1 at 25 °C, which is close to that of S. cerevisiae Ssc1 and E. coli DnaK (∼0.12 min−1) (52) but much lower than that of E. coli HscA (0.46 min−1) (49). The lower ATPase activity of mtHSP70 compared with E. coli HscA can be attributed to the fact that mtHSP70 requires a nucleotide exchange factor (GrpEL1), which catalyzes the exchange of ADP for ATP, to reach maximal ATPase activity (48). The E. coli HscA/HscB chaperone system does not utilize a nucleotide exchange factor (14). ISCU and HSC20 individually enhanced the ATPase activity of mtHSP70 severalfold, and HSC20 plus ISCU together increased the ATPase activity still more (Fig. 11A). The synergic effect of HSC20 and ISCU in stimulating mtHSP70 ATPase activity is similar to that reported for stimulation of the ATPase activity of E. coli HscA by HscB and IscU (46). Unlike the E. coli system, in which the presence of HscB decreased the concentration of IscU required to stimulate the ATPase activity of HscA (46), the concentration of ISCU needed for half-maximal stimulation of mtHSP70 ATPase activity was the same or higher in the presence of HSC20 (Fig. 11, C and D). We found that ISCU(N90A) (95 %S) in the presence of HSC20 was half as effective as wild-type ISCU (28 %S) in stimulating the ATPase activity of mtHSP70 (Fig. 11, E and F), in agreement with our model in which the D-state of ISCU binds preferentially to mtHSP70. Together, these results are consistent with the proposed function of human mtHSP70 and HSC20 as the chaperone and co-chaperone, respectively, for human mitochondrial Fe-S cluster biosynthesis.</p><p>The findings above support a working model for human mitochondrial Fe-S cluster biogenesis (Fig. 12) that is analogous to one proposed for the E. coli system (35). In this model, conversion of ISCU to the D-state when bound to the cysteine desulfurase ensures that its Cys residues are free of metal (e.g. Zn2+) and capable of accepting sulfur. Cluster formation then stabilizes the S-state of ISCU, weakens its interaction with the cysteine desulfurase (NFS1-ISD11), and strengthens its interaction with the co-chaperone (HSC20), which binds preferentially to the S-state and targets the complex to the chaperone (mtHSP70). Then, the attack of an acceptor protein (e.g. apoferredoxin) triggers activation of the ATPase activity of the chaperone leading to conversion of ATP to ADP and a conformational change in the substrate binding domain of the chaperone to the form that binds the D-state of ISCU. The latter interaction ensures irreversible release of the cluster to the acceptor protein. Upon exchange of ADP for ATP (catalyzed by an exchange factor), ISCU is released to resume its S ⇄ D equilibrium.</p><!><p>Working model for human mitochondrial Fe-S cluster biogenesis. 1, apoISCU in D⇄S equilibrium. 2, complex formed between the cysteine desulfurase complex (NFS1-ISD11) and the D-state of ISCU. 3, sulfur delivered to Cys residues of ISCU. 4, addition of iron to form a [2Fe-2S] cluster stabilizes the S-state of ISCU. 5, the co-chaperone (HSC20) binds to holo-ISCU displacing the cysteine desulfurase complex. 6, the J-domain of HSC20 binds to the ATPase domain of the chaperone (mtHSP70), bringing holo-ISCU close to the chaperone. 7, an acceptor protein containing free Cys -SH groups approaches. 8, attack of cysteine residues from the acceptor protein liberates residues of ISCU that bind to the chaperone leading to activation of its ATPase activity. 9, conversion of ATP to ADP leads to a conformational change in the substrate binding domain of the chaperone, which then binds preferentially to the D-state of ISCU releasing the holo-acceptor protein and HSC20. 1, exchange of mtHSP70-bound ADP with ATP (which involves an exchange factor, not shown) leads to the release of ISCU, which resumes its equilibrium between the S- and D-states.</p><p>This work was supported, in whole or in part, by NIGMS, National Institutes of Health Grants U01 GM94622 (Protein Structure Initiative: Biology, Mitochondrial Protein Partnership), 8P41 GM103399 (to the National Magnetic Resonance Facility at Madison, WI), and 3R01GM058667-11S1.</p><p>iron-sulfur cluster</p><p>human scaffold protein</p><p>E. coli scaffold protein for Fe-S cluster biosynthesis</p><p>human mitochondrial DnaJ-type co-chaperone</p><p>E. coli DnaK-type chaperone</p><p>E. coli DnaJ-type co-chaperone</p><p>heteronuclear single quantum correlation</p><p>E. coli cysteine desulfurase</p><p>fast maximum likelihood reconstruction</p><p>small protein component of the human cysteine desulfurase complex</p><p>human mitochondrial DnaK-type chaperone</p><p>human mitochondrial cysteine desulfurase</p><p>transverse relaxation optimized spectroscopy</p><p>immobilized metal affinity chromatography.</p>

PubMed Open Access

Determination of Glucocorticoids in UPLC-MS in Environmental Samples from an Occupational Setting

Occupational exposures to glucocorticoids are still a neglected issue in some work environments, including pharmaceutical plants. We developed an analytical method to quantify simultaneously 21 glucocorticoids using UPLC coupled with mass spectrometry to provide a basis to carry out environmental monitoring. Samples were taken from air, hand-washing tests, pad-tests and wipe-tests. This paper reports the contents of the analytical methodology, along with the results of this extensive environmental and personal monitoring of glucocorticoids. The method in UPLC-MS turned out to be suitable and effective for the aim of the study. Wipe-test and pad-test desorption was carried out using 50 mL syringes, a simple technique that saves time without adversely affecting analyte recovery. Results showed a widespread environmental pollution due to glucocorticoids. This is of particular concern. Evaluation of the dose absorbed by each worker and identification of a biomarker for occupational exposure will contribute to assessment and prevention of occupational exposure.

determination_of_glucocorticoids_in_uplc-ms_in_environmental_samples_from_an_occupational_setting

1. Introduction<!>2.1. General Setting<!>2.2. Samples Collection<!>2.3. Determination of Glucocorticoids in UPLC-MS<!>3. Results<!>4. Discussion<!>5. Conclusions<!>Conflict of Interests<!>

<p>Under physiological conditions, glucocorticoids possess various functions including modulation of carbohydrate, protein, lipid, and nucleic acid metabolism; sharp increases are observed in response to stressors that threaten homeostasis. These properties, and especially the anti-inflammatory activity, are extensively used in pharmaceutical industry to produce several medical formulations.</p><p>This industrial manufacturing could lead to some exposures to glucocorticoids in occupational settings that possibly entail hazards for workers' health, thus far not adequately considered nor thoroughly investigated.</p><p>Some evidence [1–3] pointed out Cushingoid symptoms, adrenocortical insufficiency, and skin disorders due to chronic accidental absorption of glucocorticoids. To our knowledge, from early 80's to date no other study focused on this specific issue and therefore no assessment of environmental monitoring or human biomonitoring was provided despite the wide variety of clinical problems related to therapies with glucocorticoids and the low plasmatic level at which these molecules execute their endocrine action [4].</p><p>Furthermore, occupational exposures generally entail low doses of environmental pollutants assumed for long periods. The effect of such a kind of exposure to glucocorticoids is thus far neglected as well as the simultaneous exposure to different molecules.</p><p>Such occupational exposures start to concern occupational physicians and industrial hygienists, requiring the development of analytical techniques is accurate and reliable.</p><p>In this regard, our survey needed the measurement of 21 different glucocorticoids from air, hand washing, pad-test, and wipe-test (UNI CEN/TS 15279:2006). Thus, an adequate analytical method to quantify simultaneously 21 different compounds of interest using UPLC and tandem mass spectrometry was developed.</p><p>This paper reports the contents of the analytical methodology, along with the results of this extensive environmental and personal monitoring of glucocorticoids, carried out in a pharmaceutical industry plant.</p><!><p>The industrial plant, in which the monitoring was carried out, produces several active pharmaceutical compounds, mainly steroids, antibiotics, and anticancer drugs. Annually, about 100 tons of pharmacologically active raw materials are used in the plant.</p><p>The production is generally made of batch processes, related to the industrial orders, and is carried out in closed-loop systems. Each class of drugs is produced in a specific department of the plant, distinct from the others.</p><p>The glucocorticoids synthesis is made up by several productive steps, namely, the washout of the reactor with water and subsequently with acetone and/or methanol; inerting and anhydrification; loading of the raw materials; specific chemical reactions; precipitation or crystallization of the products; centrifugation; drying; packaging.</p><p>Among those steps, only the drying could expose workers to glucocorticoids, given that the pharmacologically active compounds are manually introduced in the drying ovens and manually extracted from them. In fact, the operators, after cleaning their workplace with a pressure washer and acetone, have to arrange the semifinished products on trays and place them in the drying ovens. After an appropriate time lapse, depending on the glucocorticoids, workers have to extract trays and to stock the products in specific kegs. Finally, the personnel have to clean again their workplace with a pressure washer and acetone.</p><p>To effectively monitor this working environment and workers' exposures, samples of air, hand-washing test, pad-test, and wipe-test were collected to assess the contamination of work surfaces and work clothing, as described as follows. A list of glucocorticoids considered in this study is provided in Table 1.</p><!><p>All samples were collected during 5 days in November 2013, in which beclomethasone dipropionate, mometasone furoate, prednisolone 17-valerate, and 21-acetate were produced.</p><p>Air samples were collected in workers' breathing zone by means of a pump which draws air through a 25 mm diameter fiberglass filter at a flow rate of 1.4 L/min. For each shift, an operator was monitored.</p><p>Moreover, workers' personal exposures were assessed with hand-washing test. This sampling technique consists in hand washing with an appropriate solvent at the end of the work shift and after that gloves have been removed. In particular, 250 mL of 50% ethanol solution have been directly poured on the operators' hands and recollected in a specific container, after some rubbings ("pouring method") [5–7]. The ethanol has been chosen as solvent given its low skin toxicity and its capability to remove even small amounts of not solubilized pollutants by mechanical action. Sampling has been carried out both after the introduction of semifinished products in drying ovens and after the extraction from drying ovens.</p><p>Contamination of work clothes was assessed by means of four pad-tests (gauzes in TNT, 10 × 10 cm). They were clamped on the workers' smock external surface at thorax, back, right forearm, left forearm, right thigh, and left thigh. In addition, a pad-test was fixed on the internal surface of the workers' smock at right forearm and left forearm [8]. At the end of the work shift samples have been removed from the smock, sealed in syringes, and conserved at 4°C until the day of the analysis.</p><p>The assessment of work surfaces' contamination has been carried out with the wipe-test, in which the sampling substratum was provided by four-layer nonwoven wipes of 10 × 10 cm dampened with 2 mL of ethanol-water solution (50 : 50, v/v). The sampling was carried out by wiping all across the selected surfaces, namely, the worktop next to the extractor (20 × 20 cm, 1 sample), the door stops (60 × 15 cm, 8 samples), the door handles (5 samples), the keg cover (20 × 20 cm, 1 sample) and the keg lateral surface (20 × 20 cm, 1 sample). As for the pads, wipes were sealed in syringes and conserved at 4°C.</p><!><p>Standards of the analytes of interest were provided by the producing company with a degree of purity >95%, while the carbamazepine internal standard (degree of purity >98%), the methanol for the LC-MS instrument, and the formic acid (HiPerSolv degree of purity) were provided by Sigma Aldrich (Munich, Germany). Moreover, water for the mobile phase was daily supplied by a MilliporeMilliQ purifier and for the samples' desorption, solvents with purity AnalaR were used.</p><p>The determination of glucocorticoids in wipe-, pad-, and hand-washing tests and air samples was obtained by liquid chromatography associated with mass spectrometry, namely, with Acquity UPLC system coupled with a triple quadrupole Waters TQD mass spectrometer (Waters, Milford, MA, USA). The Mass Lynx 4.1 software oversaw the instrument and TargetLynx software was used for quantification.</p><p>Briefly, after sampling, wipe- and pad-test were desorbed using a 50 mL syringe with 5 mL × 3 of methanol-acetonitrile (50 : 50, v/v), centrifuged, conveniently diluted before injection in UPLC (range of calibration curve: 10–100 μg); hand washing was realized with 250 mL ethanol-water solution (50 : 50, v/v); then an aliquot was simply centrifuged and diluted before injection (range of calibration curve: 0.17–1.75 mg); fiber glass membrane was desorbed with 2 mL of methanol-acetonitrile (50 : 50, v/v), centrifuged, and diluted before injection (range of calibration curve: 1.4–14 ng).</p><p>Chromatographic separation was performed on a UPLC HSS C18 column (2.1 × 50 mm, 1.8 μm) maintained at 30°C and by gradient elution with a mixture containing variable proportions of 0.1% formic acid solution and methanol delivered at the flow rate of 0.5 mL/min. The gradient program was 50% methanol for 0.5 min; was from 50% to 98% methanol in 9.5 min (linear gradient) and was held for 2.5 min to permit the washing of the column; was from 98% to 50% methanol in 1 min (linear gradient) and was held for 1.5 min (column reconditioning); retention time of carbamazepine (internal standard) and other analytes are provided in Table 1.</p><p>For the detection of the 21 glucocorticoids in mass spectrometry, electrospray was operated in positive ion mode and the acquisition was performed in single ion recording (SIR). An example of chromatographic separation is reported in Figure 1. The correspondence between the number of the peaks and the analytes is indicated in Table 1.</p><!><p>The method in UPLC-MS developed for the determination of 21 glucocorticoids simultaneously was suitable and effective for the aim of the study.</p><p>Limits of detection (calculated as signal-to-noise ratio of 3) of different analytes were within the range of 0.1–1.4 μg for wipe-test and pad-tests, 2–23 μg for hand-washing tests, and 0.25–0.52 μg/m3 for air samples, considering a sampling average time of two hours.</p><p>Tables 2 and 3 show results of the determinations in all kinds of samples. These results account for the viability of this method to assess the workers' exposure.</p><p>The environmental pollution due to glucocorticoids seemed to be widespread. The work surfaces and environment were contaminated with several different analytes, whose mean values often exceeded 10 μg and not infrequently overtook 100 μg. Moreover, the maximum value in wipe-test determinations exceeded 1 mg in two cases: beclomethasone dipropionate (3.37 mg, during the day of compound production) and chlormadinone acetate (1.22 mg) (Table 2).</p><p>Air monitoring showed that during the days in which glucocorticoids were processed, the compound in production held the highest value. Beclomethasone dipropionate showed the maximum concentration, reaching 388.2 μg/m3 (Table 2). It is interesting to note that chlormadinone acetate and mometasone furoate reached the concentration of 1.28 μg/m3 and 1.22 μg/m3, respectively, also in a day in which the production was devoted to beclomethasone dipropionate (Table 2).</p><p>The majority of mean values of the hand-washing test were above 10 μg, rising to 220.2 μg and 258.4 μg for desoximetasone and beclomethasone dipropionate, respectively. Monitored workers were occasionally exposed to nearly or over 1 mg of glucocorticoids (0.83 mg and 2.04 mg for beclomethasone dipropionate and desoximetasone, resp.) (Table 2).</p><p>Considering pad-tests on the external surface of the workers' smock, results were especially high for beclomethasone dipropionate, mometasone furoate, and prednisolone 17-valerate 21-acetate, as expected given that they were the compound in production. The mean values of these analytes were almost constantly over 10 μg, rising to 0.13–0.66 mg for beclomethasone dipropionate and mometasone furoate. As expected, minimum values were found for pad-tests clamped on the back of workers' smock. These results are consistent with values from pad-tests placed on the internal surface of the smock, at right and left forearms, although, for these pad-tests, not entirely negligible values were found also for chlormadinone acetate and desoximetasone. Internal pad-tests showed mean values 50–100-fold lower than external one (Table 3).</p><!><p>Our study shows results from an occupational environment monitoring that challenged our laboratory to develop a method to simultaneously quantify in UPLC-MS 21 different glucocorticoids from wipe-tests, pad-tests, hand washing, and fiberglass filters. To our knowledge, in the literature there are no papers that report methods to quantify these analytes with this chromatographic technique in such substrates. In the first chromatographic tests we took into consideration the chromatographic conditions (type of column and mobile phase) reported by some authors [9, 10], although these works deal with determinations of few glucocorticoids in biological matrices, such as urine or hair, for clinical purposes.</p><p>Usually, in the first development phases of an LC-MS method it is essential to consider both the target regarding the range of calibration curve and limit of quantification. These parameters depend on the real concentration of samples. Unfortunately, we had no such literature information and thus it was necessary to make some proofs directly on industrial samples. It was immediately clear that samples were characterized by a high between-samples and between-analytes variability and so the calibration curve range was chosen in such a manner as to avoid signal saturation. Samples with a signal out from this range were opportunely diluted and injected again.</p><p>Initially, carbamazepine (CBZ) was chosen as internal standard given that this compound was not present in the work environment and its hydrophobicity is similar to glucocorticoids. Nevertheless, after the first analyses on real samples it was decided to add CBZ to all samples although its use was avoided for quantifications, due to high glucocorticoids signal variability. It has not been possible to define a dilution factor before first injection in UPLC and, accordingly, the right initial internal standard spike. In addition, to control for possible instrument drift and quantitative performance issues an external standard was regularly injected every 20 samples.</p><p>Due to time constraints described below and the lack of available guidelines, we could not carry out a formal method validation; nevertheless, we verified the method's reliability as described in the following.</p><p>Recovery of the analytes and matrix effect were verified for each kind of sample by calculating the ratio of the peak area in the presence of matrix (wipe-test, hand washing, or membrane) to the peak area in the absence of matrix (pure solution of the analytes diluted in the same mobile phase). This test was carried out for low and high level concentrations. For wipe-test we confirmed this procedure by spiking standards after cleaning different types of surfaces not contaminated by glucocorticoids; similarly, for hand washing we carried out the test using samples taken from subjects not exposed to our analytes.</p><p>Moreover, this survey was planned according to company demand since some cases of fingers' telangiectasia have been observed among workers. Thus, given that results were waited within a brief time, to shorten the method development without threatening the quality of result, we decided not to infuse standards into the mass spectrometer to obtain transitions of each molecule (MRM). After some injections of each single analyte at different cone voltage, we decided to quantify the analytes in SIM (Table 1). We thought that the choice of reducing the method's selectivity did not penalize the analysis of environmental samples, generally cleaner and ore freer from analytical interferences than biological samples. In fact, the analyses carried out on real blank samples (i.e., wipe-test simply obtained after cleaning different types of surfaces or hand washing from subjects not exposed to glucocorticoids) did not show chromatographic interferences with the same RT and the same m/z.</p><p>Besides, the high level of dilution step of the samples necessary to quantify the analytes within the range of calibration curve contributed to obtain samples particularly clean and thus free from interferences.</p><p>Wipe-tests and pad-tests desorption was carried out by using syringes 50 mL, an innovative technique that saves time without affecting the analyte recovery. No difference was observed between wipe- and pad-tests regarding analyte recovery: results were encouraging showing a recovery mean for all analytes of 92.4%. Indeed, the desorption of this kind of samples is usually carried out by submerging the gauze into a container 100 mL with a desorption volume of 20–25 mL (e.g., urine container) and leaving the sample plunged and shacked for at least 30 minutes [11, 12]. Our method requires a desorption volume of 15 mL only, allowing desorption of samples in a few minutes. Furthermore, it was not necessary to introduce a purification step (typically solid phase extraction) because samples were clean enough and concentrated; on the contrary, after desorption, many samples were opportunely diluted and injected in a second time to return to the range of instrument linearity.</p><p>Our survey shows a wide contamination of occupational environment, both for work surfaces and for structures. Also air samples highlight a respiratory exposure for some glucocorticoids (in particular, beclomethasone dipropionate, i.e., the plant most produced glucocorticoid). The smocks are probably a good protective device, although samples on their internal surface at forearms are not negligible for some compounds (Table 3). This is an indirect proof of a skin contact, as well as the results from hand-washing test that underline a certain exposure although workers wear their protective gloves. Nevertheless, results from hand-washing tests represent the most relevant contamination issue from an occupational health point of view. In fact, these workers are daily exposed to an average dose of about 0.8 mg, considering all glucocorticoids. Some of them (beclomethasone dipropionate, desoximetasone) showed single maximum results close to 1 mg (Table 2). For example, assuming that a single application of a beclomethasone cream (0.1%) to hands is about 1 gram, the dose measured on workers' hands is close to the pharmacological one. Furthermore, the effect of multiple contemporary exposures cannot be disregarded, especially when synergic and addictive effects were taken into account.</p><p>In addition, the widespread contamination of work environment and surfaces posits a cross contamination issue of the pharmacological product that has to be dealt with.</p><p>Therefore, we think that environmental monitoring of exposed workers should no longer be neglected and our analytic method provides a basis to develop further surveys for measuring and limiting the exposures.</p><!><p>Our development of an analytical method to simultaneously quantify 21 different glucocorticoids seemed to be effective and reliable, along with the innovative desorption technique using 50 mL syringes. These methods allow us to carry out an extensive (environmental and personal) monitoring in a pharmaceutical plant, showing a widespread contamination that should no longer be neglected. Moreover, to evaluate the real dose absorbed by each worker, it is of particular concern to identify an effective biomarker of occupational exposure.</p><!><p>The authors declare that there is no conflict of interests regarding the publication of this paper.</p><!><p>Example of chromatographic separation related to a single point of calibration curve (50 μg on wipe-test). Numbers on peaks identify glucocorticoids code as showed in Table 1.</p><p>Codes and principal chemical features of glucocorticoids.</p><p>aRetention time.</p><p>bCone Voltage.</p><p>cInternal standard.</p><p>Results of environmental and personal monitoring of glucocorticoids in a pharmaceutical plant (2013).</p><p>aAbsolute µg.</p><p>b µg/m3.</p><p>Results of pad-tests on workers clothes of a pharmaceutical plant (2013). For each sample N = 5. "I" indicates the internal surface of the smocks.</p>

PubMed Open Access

Small-molecule protein tyrosine kinase inhibitors for the treatment of metastatic prostate cancer

The microenvironment is critical to the growth of prostate cancer (PCa) in the bone. Thus, for clinical efficacy, therapies must target tumor\xe2\x80\x93microenvironment interactions. Several protein tyrosine kinases have been implicated in the development and growth of PCa bone metastasis. In this review, specific protein tyrosine kinases that regulate these complex interactions, including PDGFR, the EGFR family, c-Src, VEGFR, IGF-1R, FGFR and c-Met will be discussed, with an emphasis on why these kinases are promising therapeutic targets for metastatic PCa treatment. For each of these kinases, small-molecule inhibitors have reached clinical trials. Current results of these trials and future prospects for the use of tyrosine kinase inhibitors for the treatment of PCa bone metastases are also discussed.

small-molecule_protein_tyrosine_kinase_inhibitors_for_the_treatment_of_metastatic_prostate_cancer

<!>Src family kinases<!>EGFR signaling axis<!>PDGF/PDGFR signaling axis<!>VEGF & VEGF receptors<!>The IGF/IGF-R signaling axis<!>c-Met & HGF<!>The FGF/FGFR signaling axis<!>Emerging targets<!>Future perspective<!>

<p>Protein tyrosine kinases (PTKs) have become major targets in numerous types of solid tumors. These enzymes, which may be classified as receptor or nonreceptor kinases, are frequently mutated, overexpressed (because of gene amplification or increased translation) or increased in specific activity (due to activation or overexpression of 'upstream' molecules that lead to increased kinase activity). Small-molecule inhibitors, such as Gleevec, have shown remarkable success in controlling the earlier stages of chronic myelogenous leukemia, a disease 'addicted' to aberrant expression of the BCR–Abl fusion gene. Targeting Abl is highly successful prior to the very late stage of the disease (blast crisis) developing. Partly based on this paradigm and increasing knowledge of mechanisms by which PTKs are aberrantly activated, numerous small-molecule inhibitors as well as monoclonal antibodies, are now undergoing clinical trials to 'block' signaling from selective PTKs. In some instances, such as mutated EGFR found in a small percentage of lung cancer, and B-Raf mutations in melanoma, treatment with selective inhibitors has led to increased patient survival. Often, however, resistance arises through overexpression of a PTK not targeted by the initial treatment; for example c-Met overexpression, which often occurs in patients treated with targeted therapies to EGFR [1]. Nevertheless, our increasing knowledge of which PTKs may be 'drivers' of tumor progression and which contribute to resistance to both targeted therapies and chemotherapy is leading to better clinical trials that are gradually increasing survival of patients with numerous solid tumors.</p><p>In prostate cancer (PCa), the roles of PTKs in progression, metastasis, and growth at the metastatic site (usually the bone) have also received considerable attention. However, there is little evidence that PCa is 'addicted' to any specific PTK. Rather, the complex interaction between microenvironment and tumor appears to be a major force in metastatic disease [2–5]. The progression of PCa in the bone is in part through the bidirectional interactions between the PCa cells and bone that leads to the vicious cycle, whereby tumor release factors affect bone remodeling, causing growth factors to be released from bone matrix and these bone-derived factors to further activate multiple tyrosine kinases in the tumor. These interactions do not mean that increased expression of specific PTKs is not important in PCa progression and metastatic growth in the bone; rather that the mechanisms by which PTK are involved in PCa progression are greatly influenced by their cognate growth factors in the microenvironment. As examples, the expression of receptor PTKs c-Met and IGF-R are increased in bone metastases, and this overexpression correlates with poor survival, discussed below [6–8], but the ligands for these receptors are also present in the bone, released by tumor/bone interactions (see later) and must be considered when small-molecule inhibitors of these enzymes are used in therapy, as these PTKs affect overlapping pathways. In addition, several PTKs affect androgen receptor (AR) signaling by phosphorylating AR [9]. Thus, the effects of inhibitors on AR must also be considered when PTK inhibitors are used in clinical trials. The following sections will describe the effects of PTK inhibitors in use in clinical trials for metastatic PCa; combination strategies with PTK inhibitors and other signaling inhibitors will also be described. Tyrosine kinase inhibitors (TKIs) that have already been tested clinically will be described first.</p><!><p>The Src family kinases (SFK) comprises nine highly related nonreceptor PTKs (Src, Yes, Fyn, Lyn, Lck, Hck, Fgr, Blk and Yrk) [10]. Src, Lyn, and Fyn have all been demonstrated to play roles in PCa development and/or progression. The archetypal member, Src, was the first oncogene discovered [11], the first to demonstrate that viral oncogenes were derived from normal cellular proto-oncogenes [12] and the first to be demonstrated to have intrinsic PTK activity [13]. The structure of SFK and mechanisms of activation have been described extensively in numerous reviews [14,15]. Src family members are not directly activated by extracellular signals, but are often rapidly activated by binding to activated cellular receptors, including receptor PTKs and GPCRs, integrins and numerous inducers of stress response. As genetic and epigenetic alterations (overexpression of growth factor receptors and their ligands, and activation of integrins as examples) lead directly to Src activation during PCa progression, it is not surprising that SFK activity is increased in progressive stages of PCa [16].</p><p>In addition to Src, two of its related family members have been implicated in PCa, Lyn and Fyn, both of which are also overexpressed in PCa. Lyn is involved in prostate development, and a peptidomimetic inhibitor of Lyn slowed tumor growth in vivo [17]; a result confirmed by stable transfection and expression of a Lyn shRNA [18]. Fyn affects prostate cell proliferation and chemotaxis, especially in response to HGF [19], a growth/migration factor present in the bone microenvironment.</p><p>Ectopic expression of constitutively active mutants of Src, Yes and Fyn in primary prostate cells are able to induce prostate tumors after implantation into mice [20]. Although the mutations used to activate Src have not been demonstrated in PCa, all three kinases were able to induce tumors, with Src the most potent oncogene, followed by Fyn and Lyn. Thus, considerable preclinical evidence supports the use of SFK inhibitors in PCa and especially in PCa bone metastases.</p><p>Currently, four different small-molecule SFK inhibitors have reached clinical trials: dasatinib, saracatinib, bosutinib and KX2–391 [21]. The first three of these inhibitors compete for ATP binding, whereas KX2–391 competes with Src substrate binding. Importantly, all these inhibitors have additional targets to SFKs, and their efficacy may depend, in part, on these additional targets, with potentially unexpected consequences [22]. Considerable work must be performed before an understanding of the relevance of these additional targets to the efficacy of a given SFK inhibitor, but different side effects are observed from different inhibitors, suggesting different 'off-target' effects.</p><p>Of particular interest in PCa, the major defect observed in src−/− mouse is osteopetrosis, a thickening of bones due to defective osteoclast function [23]. Hence, it would be predicted that Src inhibitors would have bone turnover effects in humans. This prediction was borne out by studies of Hannon et al. who tested the Src-selective inhibitor, saracatinib (AZD0530) on bone turnover in a Phase I trial in healthy men [24]. A dose-dependent decrease in bone resorption markers C-telopeptide of type 1 collagen (serum CTX) and N-telopeptide of type 1 collagen (urinary NTX) was observed in men treated with saracatinib relative to untreated men, in accordance with the expectation from the src−/− mouse strain. Similar effects were seen in men with bone-metastatic PCa with the SFK/Abl inhibitor, dasatinib, described below.</p><p>As described above, saracatinib was first tested in healthy men in Phase I trials. More recently, saracatinib was used in a single-agent Phase II trial conducted by the California Cancer Consortium in patients with advanced castrate-resistant PCa (CRPC) [25]. While the drug was generally well tolerated, little clinical efficacy was observed in this study. These results are not unexpected, as considerable preclinical evidence suggests combination therapy will be required for efficacy of SFK inhibitors. Additional clinical trials have used dasatinib either twice daily [26] or, more recently, once a day [27] in CRPC patients with metastasis. The studies showed similar and encouraging results in a subset of patients.</p><p>Another recent study used dasatinib in combination with docetaxel in a Phase I/II [28] was sufficiently promising to be shortly followed by a Phase III trial. A subset of patients in the Phase I/II trial had durable responses of more than three years with no rise in prostate-specific antigen (PSA) yet observed. The Phase III trial is completed at the time of writing, but has yet to be unblinded. Thus, early trials on SFK inhibitors are showing promise. Determining why some patients respond well to SFK inhibitors, whereas others fail to do so remains an important challenge in determining the best utility of these inhibitors.</p><!><p>The EGFR family comprises four structurally related members, EGFR (ERBB1 and HER1); HER2 (Neu and ERBB2); HER3 (ERBB3) and HER4 (ERBB4) [29–32]. EGFR binds a number of ligands including EGF, TGFα and amphiregulin with high affinity. Other factors demonstrated to bind EGFR with lower affinity include betacellulin, heparin-binding EGF and epiregulin. These latter factors are also known to bind HER4. Neuregulins (NRGs) also bind members of the EGFR family, with NRG1 and NRG2 binding both HER3 and HER4, whereas NRG3 and NRG4 bind only HER4. Her2 is not directly bound by a ligand, but can heterodimerize with ligand-bound EGFR, participating in signal transduction. Homo- or heterodimerization is required to stimulate the intrinsic tyrosine kinase activity of the EGFR family, with heterodimers generally more strongly propagating downstream signals [32]. Activation of the EGFR family triggers numerous signaling pathways [33] important in development, proliferation and wound healing.</p><p>Several studies have demonstrated that EGF or TGFα can stimulate osteoclast formation, leading to bone resorption [34–36]. Because of this role in osteoclast function and bone turnover, several inhibitors of EGFR and/or HER2/neu (both small-molecule inhibitors and monoclonal antibodies), including trastuzumab (a monoclonal antibody targeting HER2), gefitinib, also known as Iressa® (a small-molecule inhibitor targeting primarily EGFR), erlotinib (a small-molecule inhibitor targeting primarily EGF-R), pertuzumab (a monoclonal antibody to HER2 targeting a different epitope than trastuzumab) and lapitinib (a pan EGFR family inhibitor) have been used as single agents or in combination with chemotherapy in clinical trials in patients with CRPC [37–43]. Unfortunately, none of these trials has shown much promise [32].</p><p>The only member of the EGFR family quite frequently overexpressed in PCa is HER3 [44,45]. Inhibiting EGFR and HER2 leads to activation of HER3. Specifically, while inhibition of EGFR and HER2 decreased AR transcriptional activity, the remaining AR function was mediated by HER2/HER3 heterodimerization, not through EGFR [46]. Androgen withdrawal in androgen-dependent cell lines has also been shown to lead to increased HER3 expression [47], but inhibition of HER1 and HER2 sensitizes PCa cells to androgen withdrawal by decreasing HER3 expression [48]. Additionally, HER3 is increased in expression by inhibitors of PI3K in PCa cells, leading to upregulation of AR [49]. Biologic functions of signaling through HER3 in PCa and resistance to therapy are being rapidly elucidated. Soler et al. demonstrated that HER3 is required to maintain the motile and invasive phenotype of the prostate tumor cell line, DU-145 [50]. Collectively, these data suggest that HER3 may play an important role in PCa progression.</p><p>The above results suggest that pan EGFR inhibitors might hold promise in CRPC therapy. However, as discussed above, one such inhibitor, lapatinib, was ineffective in clinical trials [43]. In combination with erlotinib, the pan TKI inhibitor, MP470, inhibited growth of prostate tumor cell lines [51]. MP470 is being evaluated in Phase I clinical trials. Other pan EGFR inhibitors are in development, such as AZD8931 [52], but their effectiveness as therapies for PCa remain unknown. In light of our current knowledge, it is likely that pan HER3 inhibitors will be required to target EGFR family signaling, but their potential success will depend on use in combination therapy, especially with inhibitors affecting AR signaling.</p><!><p>The PDGF/PDGFR signaling axis results from complex interactions between several forms of the ligand that bind to their receptors. PDGFs are members of a family of four distinct polypeptides encoded by four different genes. The individual PDGFs are termed PDGF-A, PDGF-B, PDGF-C and PDGF-D. PDGFs function as homo-or hetero-dimers, linked through disulfide bonding. Specifically, five forms of PDGF dimers have been identified: PDGF-AA, PDGF-BB, PDGF-AB, PDGF-CC and PDGF-DD. Two PTK receptors for PDGFs have been identified: PDGFRα (activated by binding PDGF-AA, PDGF-AB, and PDGF-CC) and PDGFRβ (activated by binding PDGF-BB and PDGF-DD) [53]. In addition, PDGFRα/PDGFRβ form heterodimers that can be activated by binding of PDGF-BB, PDGF-AB and PDGF-CC. As with the other receptor tyrosine kinases discussed in this review, ligand binding leads to dimerization of the receptor. Transphosphorylation of the receptor occurs, which both activates the intrinsic tyrosine kinase activity of the receptor and leads to recruitment of signaling proteins including SFK, PI3K, and phospholipase Cγ [54]. These interacting proteins induce signaling cascades that promote proliferation, survival and cell migration, although only PDGFRβ/PDGFRβ homodimers and PDGFRα/PDGFRβ heterodimers are associated with chemotaxis of smooth muscle cells as well as fibroblasts. Crosstalk of PDGFR occurs with other signaling molecules, such as integrins [55]. Primary roles of PDGFs in adults include stimulation of wound healing [56] and regulation of interstitial fluid pressure of tissues [57,58]. Studies of genetically engineered mice have demonstrated that PDGF is important in vessel maturation and recruitment of pericytes to blood vessels, the latter of which is a PDGFRβ-dependent function [58].</p><p>The last of the PDGF family to be discovered, PDGF-D, has been implicated in malignant transformation [59], and is overexpressed in PCa, promoting angiogenesis and invasion by binding PDGFRβ [60,61]. Overexpression of PDGF-D in PC3 PCa cells increased proliferation rates and invasion of these cells, and conditioned medium from PC3 cells overexpressing PDGF-D-induced tube formation in human umbilical vein endothelial cells [62], suggesting that PDGF-D promotes angiogenesis. Using the same system, this group further demonstrated that PDGF-D contributes to epidermal-to-mesenchymal transition (EMT) [63]. Thus, PDGF-D has been suggested as a novel target for PCa [59], although no specific inhibitors have yet to be developed.</p><p>In both primary PCa and bone metastases, PDGF-R is almost universally overexpressed relative to normal tissue, with very high expression in PCa bone metastases [64–66]. Using the metastatic PC3-MM2 cells implanted into the tibia of male nude mice, Uehara et al. demonstrated that treatment with the multikinase inhibitor, imatinib, with or without paclitaxel, decreased tumor proliferation and increased apoptosis [67]. In a second study, this group also demonstrated that intratibial implantation of PC3-MM2 cells that were multidrug resistant and insensitive to paclitaxel and imatinib in vitro were still sensitive to imatinib in vivo with or without paclitaxel, with decreased bone tumor incidence and tumor weight. PDGFR phosphorylation was inhibited in both tumor cells and endothelial cells, the latter of which underwent apoptosis, suggesting that imatinib-targeting of endothelial cells played a major role in the anti-tumor effects observed in this system, implicating the importance of targeting the microenvironment for treatment of PCa bone metastases [68], and that imatinib is important in inhibiting angiogenesis. These studies were, in part, responsible for initiating a Phase I/II study on imatinib, described below.</p><p>Small-molecule inhibitors of PDGFR have been developed and tested in clinical trials. The multi-institutional Phase II study using SU101 for the treatment of hormone-refractory PCa resulted in partial response in one of 19 patients [69]. A Phase I trial in androgen-independent PCa showed promising results based on PSA decline when imatinib mesylate was applied in combination with docetaxel [70]. A randomized study followed testing imatinib plus docetaxel versus docetaxel alone for the treatment of 144 men with progressive CRPC with bone metastases [71]. The results showed no significant difference in times to progression between the two groups of patients [71]. Another clinical trial that tested PDGFR inhibitor in neoadjuvant setting by combining imatinib therapy, docetaxel and hormone ablation in the preoperative setting in high risk localized PCa [72]. No pathological complete remissions were observed. The lack of promise from most of the clinical studies suggests that either PDGFR inhibitors are unlikely to be efficacious for the treatment of PCa or the right combinations of PDGFR inhibitors with other chemotherapy and/or signal transduction inhibitors have yet to be found.</p><!><p>Angiogenesis, the growth of new from existing blood vessels, is critical for the development, growth and metastatic dissemination of tumors [73]. The search for tumor-derived factors contributing to this process led to the identification of VEGF, also known as VEGF-A, as an endothelial mitogen [74]. VEGF belongs to a family of structurally related dimeric proteins that includes VEGF-B, VEGF-C, VEGF-D and VEGF-E, as well as PlGF [75]. VEGF ligands activate angiogenic programs through binding of several receptors, including VEGF receptor (R)-1, -2, -3, and neuropilins (NRP). VEGFR-1 (Flt-1) binds VEGF, VEGF-B, and PlGF. VEGFR-2 (Flk-1/KDR) is expressed on nearly all endothelial cells and binds VEGF, VEGF-C, VEGF-D and VEGF-E to control endothelial cell proliferation, migration and survival. Through binding to VEGF-C and VEGF-D, VEGFR-3 is thought to facilitate the outgrowth of lymphatic vessels. Additionally, NRP1/2 are co-receptors for VEGF. NRP1 binds the VEGF isoform VEGF165 and PlGF, and NRP2 binds VEGF165 and VEGF-C. Unlike other VEGFRs, NRPs lack intracellular signaling domains, and their specific role in angiogenesis is not fully understood [76]. Although VEGF-A binds VEGFR-1 with a higher affinity than VEGFR-2, the biological effects of VEGF-A are thought to be mediated through VEGFR-2. On ligand binding, VEGFR-2 dimerizes, resulting in kinase activation and autophosphorylation of tyrosine residues that leads to the activation of signal-transduction molecules including PI3K, Akt, Ras, Src, phospholipase Cγ and MAPK [77].</p><p>In PCa, several parameters associated with tumor angiogenesis have been found in correlation with Gleason score, stage, metastatic progression, and survival. For example, high microvessel density, vascular size and irregularity are associated with increased long-term risk of death [78]. Moreover, VEGF and VEGF-C have been implicated in disease progression and bone metastasis [79–81], VEGF-C has been linked to lymph node metastasis [82], and elevated VEGF in plasma and urine is an independent indicator of poor prognosis [83,84]. VEGF plays a critical role during bone vascularization [85,86] and formation during normal development and repair [87,88] and the bone-remodeling process that takes place in PCa skeletal metastasis [80]. Moreover, tumor cell-derived VEGF can affect the proliferation and maturation of osteoclasts, osteoblasts, and their precursors in bone. For example, osteoclast precursors express VEGFR-1, which may facilitate their homing to the site of bone resorption and osteoclastogenesis [89,90].</p><p>Most therapeutic efforts directed toward inhibiting the angiogenic processes for the treatment of PCa have focused on the VEGF pathway. Several inhibitors have been tested, including bevacizumab (a monoclonal antibody that blocks human VEGF), aflibercept or VEGF-trap (a fusion protein of the VEGFR extracellular domain and the Fc portion of IgG1), antibodies that block the VEGFR (IMC-1121b), and small-molecule inhibitors of the VEGFR tyrosine kinases, which because of structural similarity, generally inhibit other target kinases, such as the PDGF and c-Kit receptors. Several different multi-targeted TKIs with some selectivity to VEGFR have been evaluated in PCa (including sorafenib, sunitinib, cediranib and pazopanib) [91–93], with the most number of trials using sunitinib. Two Phase II studies in predominantly metastatic castration-resistant patients showed single agent activity of sunitinib [94,95], often discordant with rising PSA levels. However, a Phase I/II trial of sunitinib in combination with docetaxel demonstrated promising results [96]. An independent study of sunitinib plus androgen deprivation in newly diagnosed, nonmetastatic PCa before prostatectomy described two pathologic complete responses after only 3 months of treatment in patients with high-grade disease [97].</p><p>Available data indicate that, in PCa, sunitinib affects endothelium and bone. Similar to other types of cancer, the concentrations of several mediators of angiogenesis including VEGF, soluble VEGFR-2 and soluble VEGFR-3 are significantly modulated upon treatment, consistent with an on-target effect [94,96]. Of relevance, and consistent with osteoclast expression of VEGFR-1 and CSF-1R, both sunitinib targets, serum N-telopeptides and thus osteoclast activity decreased in patients with bone metastasis treated with sunitinib [94]. Unfortunately, the available preliminary results of a large Phase III trial of sunitinib plus prednisone versus placebo plus prednisone in metastatic CRPC patients resistant or intolerant to docetaxel demonstrated no survival advantage for the experimental arm, suggesting that sunitinib may only benefit subpopulations of PCa patients or, alternatively, that redundant mechanisms of communication between tumor cells and stroma other than VEGF need to be simultaneously inhibited [98].</p><!><p>Two insulin-like growth factors (IGF1 and IGF2) have been identified, as have two receptor PTKs (IGF-1R and IGF-2R) to which these growth factors bind. Both IGF1 and 2 bind to IGF1R, whereas IGF2 (but not IGF1) also binds IGF-2R [99]. IGF-1R has emerged as a target in numerous solid tumors, including PCa. IGF-1R is synthesized as a single-chain precursor, which is processed in the Golgi to yield first a heterodimer then heterotetramer (2α and 2β chains) linked by disulfide bonds. IGF-1R can also heterodimerize with the insulin receptor and be activated by insulin [100]. IGF-1R is implicated in proliferation and survival of many tumor types [101], and is overexpressed in PCa [102]; IGF-1R signaling has been linked to PCa risk, and one of its ligands, IGF-1, is also overexpressed in PCa bone metastases [103] where it promotes proliferation and survival of PCa cells [104–108]. Studies with monoclonal antibodies to IGF-1R have implicated this receptor as important in PCa progression in androgen-sensitive as well as resistant tumor models [109]. Five fully humanized monoclonal antibodies to IGF-1R: IMC-A12, AMG 479, MK0646 and AVE1642 have reached clinical trial, but none has been tested specifically against prostate tumors [110].</p><p>Several small-molecule inhibitors of IGF-1R have been developed and are in various stages of preclinical and clinical trials. INSM-18 (nordihydroguaiaretic acid) has been tested in a Phase I/II clinical trial for patients with relapsed PCa [111]. While the drug was well tolerated, only one patient from the 11 evaluated on this trial had a decrease in PSA of more than 50%. A Phase II study on this agent for patients with nonmetastatic hormone-sensitive PCa was stopped early due to no significant PSA decline [112]. Another IGF-1R inhibitor in clinical trial is BMS754807 (with six trials in progress). Safety and tolerability have been established in a recent dose-escalation Phase I trial of BMS-754807 [201], where the maximal tolerated dose had not been reached at the time of this writing [113]. This small-molecule inhibitor has not yet been used specifically in patients with advanced PCa, although preclinical evidence [Dayyani F, Gallick GE, Unpublished Data] suggests it may be promising in combination with Src family inhibition.</p><!><p>c-Met is a surface receptor with intrinsic PTK activity that is expressed mainly in epithelial cells [114]. The receptor is composed of a disulfidelinked heterodimer consisting of a transmembrane β-chain and an extracellular α chain. The structure of c-Met has been extensively described [115,116]. c-Met has only one ligand, HGF, also described as scatter factor. HGF belongs to the plasminogen subfamily of S1 peptidases, although HGF itself lacks protease activity [117]. HGF expression is generally restricted to cells of mesenchymal origin [114,118]. Active HGF results from secretion of an inactive pro-HGF precursor that is subsequently cleaved into an active form consisting of a disulfide linked active α-chain and β-chain molecule [119]. Physiologically, the HGF/c-Met signaling axis regulates critical steps of embryogenesis as well as tissue repair in adult life, promoting cellular proliferation, differentiation, migration and neovascularization. Aberrant activation of the c-Met/HGF pathway in tumor cells enhances their survival, proliferation, invasive growth capabilities and promotes EMT [120].</p><p>A number of lines of evidence demonstrate that the c-Met-mediated signaling pathway is important in PCa progression including local invasion, bone metastasis and castration resistance. c-Met expression is significantly upregulated in the majority of PCa cell lines with highest expression in the more invasive cell lines, and in these in vitro models c-Met contributes both to proliferation and invasion [121]. Additionally, in PCa cell lines, c-Met expression is inversely correlated with AR expression and PSA production, with less differentiated cell lines expressing higher levels of c-Met [122,123].</p><p>In the prostate, c-Met is primarily expressed by the epithelial compartment and HGF by prostate stromal cells [124], indicating a paracrine mechanism of c-Met activation. In prostate tumor tissues, immunohistochemical studies have shown that c-Met expression is increased in primary tumors relative to normal prostate tissue, with further increases observed in bone metastases [7,125]. Both HGF and c-Met are found in the urine of PCa patients, and levels are significantly elevated in patients with metastatic disease compared with patients with localized disease [123,126]. Thus, the c-Met/HGF signaling pathway is important in tumor–microenvironment interactions, regulating PCa invasive and metastatic growth.</p><p>The above studies demonstrate that c-Met is an attractive target for therapeutic intervention of PCa growth and progression. Several strategies have been proposed to interfere with HGF/c-Met signaling including small-molecule TKIs that inhibit c-Met kinase activity and monoclonal antibodies to both HGF ligand and c-Met receptor. Discussion of the latter is beyond the scope of this review.</p><p>Several small-molecule inhibitors that target c-Met have been developed, although none are selective to c-Met alone. These TKIs to c-Met have rapidly reached clinical trial and for most, incomplete information is available as to their eventual success. Below, we describe some of the inhibitors in clinical trial.</p><p>BMS-777607 is a potent, ATP-competitive c-Met inhibitor. Recent in vitro data showed that BMS-777607 inhibits scattering, migration and invasion of prostate tumor cells in doses less or equal to 1 µM. At much higher doses, proliferation is also affected [127]. These data suggest that inhibition of c-Met is more important to invasion and metastasis, rather than local growth of PCa. BMS-77607 is currently undergoing Phase I and II trials in patients with advanced PCa [127,202]. PF-2341066 is another ATP-competitive c-Met inhibitor with additional potent activity against anaplastic lymphoma kinase [128]. PF-2341066 showed a moderate antiproliferative activity against AR positive and negative PCa cell lines, with drug responsiveness inversely associated with AR levels [128]. In preclinical mouse studies, PF-2341066 suppressed the growth of AR positive, androgen-independent prostate cells and showed a synergistic activity with castration therapy against AR positive castration-resistant PCa cells. These studies suggest that PF-2341066 (and perhaps other c-Met inhibitors), are likely to have synergistic activity with androgen-deprivation treatment. PF-2341066 has reached Phase I clinical trial [203]. Foretinib/GSK 1363089 is a multi-targeted TKI whose targets include c-Met, Ron, AXL, VEGFR2 and PDGFR [129]. Although Foretinib has been tested in several solid tumors in Phase I and Phase II studies, no specific studies have focused on PCa [130]. ARQ-197 is a more selective, but less potent TKI that in contrast with other c-Met inhibitors in development, is a non-adenosine triphosphate competitive inhibitor. In preclinical studies, ARQ-197 showed antitumor activity in numerous cell types, including PC3 prostate tumor cells [131]. In PC3 tumors grown in immunocompromised mice, ARQ-197 led to dose-dependent effects on growth inhibition of PC3-derived tumors in immunocompromised mice. These, and other preclinical studies, led to Phase I and II clinical trials in patients with solid tumors who failed first-line treatment. Preliminary results demonstrated anti-tumor activity, as well as inhibition of other signaling enzymes that can be activated by the HGF/c-Met signaling axis, including FAK [132]. The possible efficacy of ARQ-197 on PCa bone metastases has yet to be evaluated.</p><p>XL-184 (cabozatinib) is a potent, ATP competitive inhibitor with selectivity to c-Met and VEGFR-2 [133]. Cabozatinib is also active at higher IC50s against other RTKs including RET, KIT, FLT3 and TIE2. Preclinical evaluation of cabozatinib showed promising antiproliferative and anti-angiogenic properties. Because of the roles of c-Met signaling in bone metastases described above, cabozatinib has been tested as a single agent in a Phase II clinical trial in patients with metastatic CRPC [133,134]. Early results have been exciting. Interim analyses from the clinical trial have shown that 86% of patients with documented bone disease at baseline experienced complete or partial response on bone scans 6 weeks after treatment initiation. Treatment with cabozatinib also led to diminished bone pain in 64% of the patients. As would be expected for an agent that targets a signaling axis important to both tumor cells and their microenvironment, reduction in bone markers including uNTx and serum alkaline phosphatase was observed in approximately 50% of the patients [134]. Although the results are promising, correlation with other markers of disease response is needed to confirm the current observations and drug efficacy in tumor epithelial compartment, and it seems likely the efficacy of XL-184 is due, in part, to inhibition of both c-Met and VEGF-R2. While it is still too early to completely evaluate the results of the trials described above, given the exciting result achieved thus far, trials with c-Met inhibitors are now being rapidly accelerated and are likely to enter the panoply of clinically useful agents that affect tumor–microenvironment interactions in PCa treatment.</p><!><p>FGFs and their receptors (FGFRs) comprise a subfamily of receptor PTKs involved in diverse cellular and developmental processes including proliferation, apoptosis, migration, EMT and angiogenesis [135]. There are 18 different soluble FGF ligands that bind to four transmembrane receptors (FGFR1–4). The FGF receptors have a high degree of sequence homology. Like other receptor tyrosine kinases, the structure of FGFRs includes an extracellular domain that binds FGF ligands, a transmembrane domain, and an intracellular tyrosine kinase domain [136]. Binding of FGFs to FGFRs leads to receptor dimerization, activation, and initiation of an intracellular signaling cascade that triggers key downstream pathways, including Ras/MAPK, PI3K/Akt and STAT. Activated FGFRs recruit FRS2, an adapter protein that binds to the juxtamembrane portion of each receptor and acts as a critical regulatory 'node' for subsequent signaling events through its interactions with other adaptor proteins [137].</p><p>Within normal tissue microenvironments, FGFs are produced and secreted by stromal cells, while FGFRs are expressed on epithelial cells [138]. FGF/FGFR signaling influences complex epithelial–stromal interactions involved during development, including organogenesis of the prostate gland. For example, FGF10 production in the mesenchyme and FGFR2b expression on epithelial cells are essential for prostate formation in rodent models [139,140]. Conversely, experimentally induced overexpression of FGF10 in prostate stroma or FGFR1 in the prostate epithelium produces prostate adenocarcinoma [141,142]. Reflecting its biologic importance, FGF/FGFR signaling is normally tightly regulated, with low basal activity that is transiently induced and/or repressed by a series of feedback loops.</p><p>There is increasing evidence that constitutive activation of the FGF/FGFR axis contributes to human PCa progression [143]. In comparison to normal prostate epithelium, PCa epithelial cells aberrantly overexpress FGF ligands and FGFR1. For example, FGF9 is overexpressed in approximately 40% of primary prostate tumors and in 100% of bone metastases when compared with normal prostate glands [144]. Within bone, PCa cells expressing both FGFs and FGFRs create an autocrine/paracrine feedback loop that subverts normal bone homeostasis in favor of osteoblastic bone formation. Collectively, these results suggest that the FGF/FGFR signaling axis contributes to PCa progression and support the development of treatment strategies that target this important epithelial–stromal interacting pathway. These treatment strategies include monoclonal antibodies against FGFs and FGFRs, FGF 'ligand' traps, and small-molecule TKIs targeting FGFRs [145,146].</p><p>We recently initiated a clinical trial exploring the therapy potential of TKI258 (Novartis), a novel TKI with high specific activity against FGF receptor kinases [204]. In this study, men with CRPC and biopsy-proven bone marrow involvement are receiving therapy with TKI258. Preliminary results suggest clinical activity, with some patients demonstrating reductions in pain, improvements in bone scans, and responses in lymphadenopathy [Corn P et al. Unpublished Data]. Using molecular and pathologic techniques, bone marrow biopsies collected pretherapy and at 8 weeks after initiating treatment will be analyzed for evidence of TKI258-mediated modulation of FGF signaling in both the epithelial and stromal compartments. Results from this study will provide the foundation to develop candidate predictive markers and rational combinations based on targeted inhibition of the FGF/FGFR pathway.</p><!><p>Ack 1 is a large (~143 kDa) nonreceptor tyrosine kinase containing a SAM domain, an SH3 domain, a CRIB domain, tyrosine kinase domain, a proline-rich domain near its C-terminus, and a ubiquitin binding domain. Ack 1 is recruited to the nucleus by several receptor PTKs, and after phosphorylation can translocate to the nucleus [147]. Ack 1 may be involved in nongenomic AR signaling. Ack-1 phosphorylates AR at tyr-267 and 363 within the AR transactivation domain, resulting in AR recruitment to AR responsive elements, leading to expression of AR regulated genes in the absence of androgen [148,149]. Expression of tyrosine phosphorylated (activated) Ack 1 in prostate tumors increases tyrosine-phosphorylated AR at 267. Thus, targeting Ack1 may be a strategy to suppress androgen-independent AR signaling.</p><!><p>Some promising results have been achieved by targeting PTKs that affect prostate tumor/microenvironment interactions. However, all the inhibitors discussed elicit responses only in a subset of patients. As of yet, we do not understand why some patients respond to specific TKIs and others do not. Thus, a challenge in designing appropriate clinical studies with molecular targeting agents is the identification of biomarkers that would predict which subset of patients would respond to a given targeted therapy. Appropriate biomarkers for monitoring PCa progression on any therapeutic regimen are also needed. Were these tools in hand, selection of patients likely to respond to given molecular targeting agents and the evaluation of the outcomes for a given drug would greatly improve.</p><p>PTKs have overlapping functions [33] and multiple tyrosine kinases are frequently activated in PCa bone metastasis. This precludes the success of targeting individual kinases in most patients. It is also likely that prolonged treatment with a TKI will result in activation of nontargeted kinases regulating compensatory pathways that render the tumor resistant to the initial target. Numerous examples of treatment with TKIs leading to activating nontargeted kinases with overlapping functions have been observed in other tumors. Development of markers that would be early predictors of treatment failure would allow earlier use of other targeted inhibitors that may overcome this problem. Finally, some TKIs may work best in the early stage of the metastatic process before there is too much tumor burden in the bone. Further biologic and clinical studies are necessary to clarify these issues. However, it is likely that small-molecule TKIs will play an important role in treatment of PCa metastases in the foreseeable future.</p><!><p>Prostate cancer (PCa) metastasis is a heterogenous disease. Multiple tyrosine kinases are overexpressed and they are activated by the cognate ligands in the microenvironment.</p><p>Tyrosine kinase inhibitors with some selectivity to Src, EGFR, PDGFR, IGFR, FGFR, VEGFR and c-Met have been developed and tested in clinical trials.</p><p>EGFR inhibitors have shown little promise in clinical trials, potentially due to redundancy in signaling through EGFR family.</p><p>The Src inhibitors dasatinib has completed Phase III trials with promising results in a subset of patients in Phase I/II trials in metastatic PCa.</p><p>Inhibitors for IGFR, FGFR and c-Met are being tested clinically. Early studies in c-Met inhibitors showed promising results in a subset of patients with bone metastases.</p><p>Due to innate and acquired resistance to tyrosine kinase inhibitors, the majority of patients with PCa metastasis are unlikely to benefit from inhibitors of individual tyrosine kinases. Therefore, combinations of small-molecule tyrosine kinase inhibitors will play an important role in the treatment of PCa metastases in the foreseeable future.</p><p>A form of cancer that develops in the prostate.</p><p>Compounds that inhibit tyrosine kinase activity.</p><p>Growth factor or hormone receptor that possesses tyrosine kinase activity.</p><p>Primary tumor invasion to bone.</p><p>Non-receptor protein that possesses tyrosine kinase activity.</p><p> Financial & competing interests disclosure </p><p>The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.</p><p>No writing assistance was utilized in the production of this manuscript.</p>

PubMed Author Manuscript

Paired Spectroscopic and Crystallographic Studies of Proteases

The active sites of subtilisin and trypsin have been studied by paired IR spectroscopic and X-ray crystallographic studies. The active site serines of the proteases were reacted with 4-cyanobenzenesulfonyl fluoride (CBSF), an inhibitor that contains a nitrile vibrational reporter. The nitrile stretch vibration of the water-soluble inhibitor model, potassium 4-cyanobenzenesulfonate (KCBSO), and the inhibitor were calibrated by IR solvent studies in H2O/DMSO and the frequency-temperature line-slope (FTLS) method in H2O and THF. The inhibitor complexes were examined by FTLS and the slopes of the best fit lines for subtilisin-CBS and trypsin-CBS in aqueous buffer were both measured to be \xe2\x88\x923.5\xc3\x9710\xe2\x88\x922 cm\xe2\x88\x921/\xc2\xb0C. These slopes were intermediate in value between that of KCBSO in aqueous buffer and CBSF in THF, which suggests that the active-site nitriles in both proteases are mostly solvated. The X-ray crystal structures of the subtilisin-CBS and trypsin-CBS complexes were solved at 1.27 and 1.32 \xc3\x85, respectively. The inhibitor was modelled in two conformations in subtilisin-CBS and in one conformation in the trypsin-CBS. The crystallographic data support the FTLS data that the active-site nitrile groups are mostly solvated and participate in hydrogen bonds with water molecules. The combination of IR spectroscopy utilizing vibrational reporters paired with X-ray crystallography provides a powerful approach to studying protein structure.

paired_spectroscopic_and_crystallographic_studies_of_proteases

INTRODUCTION<!>Synthetic Chemistry<!>Nitrile Stretch Band Assignment of KCBSO.<!>Solvent Sensitivity of Nitrile Stretch of KCBSO.<!>Probing Protease Active Site Local Environments.<!>Protein X-ray Crystallography<!>CONCLUSIONS

<p>Proteases are ubiquitous enzymes that efficiently cleave amide bonds in proteins and peptides and are involved in such diverse biological functions as angiogenesis, apoptosis, blood coagulation, cell differentiation, and immunity.[1,2] From a mechanistic perspective, they are divided into six classes, metalloproteases, serine, cysteine, aspartic, threonine, and glutamic proteases.[3] There are over 500 human proteases and homologs and serine proteases comprise almost one-third of these enzymes.[4] The serine proteases are found in diverse species from mammals to bacteria, and yet many share the serine-histidine-aspartate catalytic triad, making these enzymes a prototypical example of independent convergent evolution.[1] Given their prevalence and importance, proteases have been extensively studied with over 350,000 scientific articles published on them.[3] However, proteases have not been evaluated by nitrile vibrational reporters[5] and the frequency-temperature line-slope (FTLS) method[6–8] to access their active site solvation state. Herein we describe studies of two serine proteases using this method paired with X-ray crystallographic analysis.</p><p>Site-specific vibrational reporters - such as nitriles - have been demonstrated to be effective reporters of local biomolecule environments.[5,9,10] Nitriles are especially effective vibrational reporters due, in part, to the position of the nitrile symmetric stretch vibration in an open region of the IR spectrum, the sensitivity of this vibration to local environment, the relatively strong oscillator strength, the relatively localized transition, and the relatively small size (only two atoms).[5,11,12] This vibrational reporter has been incorporated into peptides and proteins synthetically, genetically, and post-translationally.</p><p>The non-canonical amino acid (ncAA) 4-cyano-L-phenylalanine (pCNF) is a vibrational reporter which has been incorporated into a number of peptides including the MLCK peptide,[13] the villin headpiece subdomain,[14] the human islet amyloid polypeptide,[15] and the amyloid peptide (Aβ16–22),[16] by standard solid-phase peptide synthesis. Genetic incorporation through the amber codon suppression method has also been used to incorporate pCNF into several protein systems including myoglobin,[17] the N-terminal domain of the L9 protein (NTL9),[18] cytochrome c,[19] superfolder green fluorescent protein (sfGFP),[8,20,21] the N-terminal Src homology 3 domain of the murine adaptor protein Crk-II (nSH3),[22] plastocyanin protein (Pc),[23] and the heme nitric oxide and/or oxygen (H-NOX) protein from Caldanaerobacter subterraneus (Cs H-NOX).[10] Finally, nitrile vibrational reporters have also been incorporated post-translationally though the conversion of a cysteine to a thiocyanate (SCN) or by using enzyme inhibitors containing nitriles. For example, Boxer first demonstrated that cysteine residues can be converted into thiocyanates using Ellman's reagent for IR spectroscopic studies of several proteins including ribonuclease S-protein, human aldose reductase (hALR2), and the bacterial photosynthetic reaction center.[24] This method has also been used to study other peptides and proteins.[25–28] Boxer and coworkers also studied the active site environment of hALR2 using two different inhibitors containing nitrile groups. The electrostatic fields of wild type and mutant hALR2 were evaluated using the vibrational Stark effect.[29,30] Hochstrasser and coworkers studied the allosteric hydrophobic binding site of HIV reverse transcriptase with TMC278, a non-nucleoside inhibitor that contains two nitrile groups with two-dimensional IR spectroscopy to measure the dynamics of the interaction of each nitrile group with the enzyme.[31]</p><p>Here we have studied the active site solvation environments of the proteases subtilisin (type VIII from Bacillus licheniformis) and bovine trypsin from Bos taurus pancreas using 4-cyanobenzenesulfonyl fluoride (1, CBSF, Figure 1A), a suicide inhibitor containing a nitrile vibrational reporter (Figure 1) paired with infrared spectroscopy and X-ray crystallography. Subtilisin is a 274-residue serine protease that is 30% helical and 19% β-sheet and trypsin is a 223-residue serine protease that is 10% helical and 34% β-sheet. Both proteases were reacted with CBSF at the active site serine residue. The sensitivity of the nitrile symmetric stretch of water-soluble analogue potassium 4-cyanobenzenesulfonate (2, KCBSO, Figure 1) and CBSF to solvent and temperature were investigated. The solvents were selected to mimic various protein environments and the temperature-dependent measurements provide a calibration for the application of the frequency-temperature line-slope (FTLS) method to more accurately assess the local solvation state.[6–8,10] The FTLS method is based upon the differential temperature sensitivity of the nitrile symmetric stretch when the reporter is involved in hydrogen bonds. The protease-inhibitor complexes were studied by the FTLS method to assess the solvation state of the active sites and these results were correlated with structural data determined from the X-ray crystal structures of the protease-inhibitor complexes.</p><!><p>The synthesis of 2 (KCBSO) was accomplished in good yields by the Sandmeyer reaction of 4-benzenediazonium sulfonate with CuCN generated in situ from CuSO4, sodium ascorbate, and KCN (Scheme 1). This water soluble salt was needed as a model of the temperature-dependence of inhibitor-enzyme complex in an aqueous environment. To confirm the band assignment, the 13C-isotopomer was also prepared by the same reaction but using 13C-labeled Cu13CN. Inhibitor 1 (CBSF) is soluble in THF and thus served as the model for temperature-dependence in a non-hydrogen bonding environment.</p><!><p>The room-temperature IR spectrum of KCBSO dissolved in water shows a single, symmetrical absorbance band centered at 2240.3 cm–1 resulting from the nitrile symmetric stretching vibration (Figure 2). This assignment was based upon the position of the nitrile IR absorbance band of other nitrile-containing molecules, such as 4-cyano-L-phenylalanine (pCNF),[13,20] and confirmed through isotopic editing of the nitrile group. Specifically, a 13C-labeled KCBSO (K13CBSO), was synthesized and the resulting room-temperature spectrum in water is also shown in Figure 2. The IR spectrum of K13CBSO shows a single, symmetric absorbance band centered at 2188.1 cm–1. Thus the absorbance band centered at 2240.3 cm–1 for KCBSO has red-shifted 52.2 cm–1 upon 13C-labeling of the nitrile group. The direction and magnitude of the isotopic shift is consistent with the DFT gas-phase predicted shift of 53.1 cm–1 at the B3LYP/6–311++G(3df,3pd) level for CBSO and is consistent with the experimentally measured isotopic shift of pCNF[20] upon 13C labeling of the nitrile group (53.2 cm–1).</p><!><p>The sensitivity of the nitrile stretching frequency of KCBSO was explored in several solvents selected to mimic various local protein environments. Specifically a series of dimethyl sulfoxide (DMSO) / water mixtures were utilized to represent either solvated or de-solvated local environments similar to previous work on other nitrile containing molecules such as pCNF[20,21] and 2'-azido-5-cyano-2'-deoxyuridine (N3CNdU).[32] The IR spectra of KCBSO in water, DMSO, and three DMSO / water mixtures each show a single band for the nitrile stretch vibration (Figure 3). The nitrile stretch vibration shifts monotonically from 2227.1 cm−1 (pure DMSO) to 2240.3 cm−1 (pure water) as the DMSO percentage decreases (see Figure S1 in the Supporting Information for details of the line shape analysis). This 13.2 cm−1 blue shift is the result of hydrogen bonding between solvent water molecules and the nitrile group of KCBSO, which are absent in the DMSO solution. The direction of this shift is similar to previous literature studies of the effect of hydrogen bonding on the nitrile stretching frequency of pCNF and N3CNdU, although the magnitude of the shift is slightly larger than the observed shift in pCNF (10.9 cm−1),[20] and N3CNdU (12.5 cm−1)[32] suggesting a greater sensitivity of the nitrile stretching frequency of KCBSO. The blue shift in the nitrile stretch frequency going from DMSO to water was also observed for 13C-labeled KCBSO as expected (see Figure S2 in the Supporting Information). The dependence of the nitrile stretch was also measured in methanol for CBSF (see Figure S3 in Supporting Information). CBSF instead of KCBSO was utilized for methanol based upon solubility limitations. As expected based upon literature precedent,[33] the nitrile IR absorbance band of CBSF in methanol consisted of two components where the high frequency component corresponded to the nitrile hydrogen bonding with methanol and the low frequency component was not participating in hydrogen bonding.</p><!><p>The CBSF inhibitor was then reacted with subtilisin or trypsin to probe the active site environment. Upon reaction with the active-site serine the protein-inhibitor complexes subtilisin-CBS and trypsin-CBS are formed along with one equivalent of hydrogen fluoride as a byproduct. Reaction conditions were selected to maximize inhibitor binding to the active site serine and minimize undesired side-reactions of the inhibitor with other residues in the protein. Mass spectral analysis was utilized to optimize this procedure to achieve the optimal balance between these two goals (see Supporting Information Figure S15). The best conditions utilized 1.0 equivalent of CBSF with subtilisin or trypsin in an aqueous buffer (10 mM Hepes, 20 mM KCl, pH 7.5) at room temperature overnight. X-ray crystallography confirmed the preferential binding of the inhibitor to the active site of both proteases (see below). Unfortunately, nonspecific binding of the inhibitor (i.e., binding to other sites) was also observed with ~40% of the sample containing more than one equivalent of CBS bound to the enzyme. This complicates the analysis of the IR data and is one reason why X-ray crystallographic analysis was coupled with the IR analysis.</p><p>The room-temperature IR spectrum of the nitrile stretching frequency of subtilisin-CBS and trypsin-CBS each show a single, symmetrical absorbance band at 2241.4 cm–1 and 2244.3 cm–1, respectively (Figure 4). The similarity of these frequencies with the room-temperature nitrile stretching frequency of KCBSO in water suggests that the nitrile group is involved in hydrogen bonding interactions with either the solvent or neighboring residues. However, surprisingly both of these frequencies are blue shifted from the nitrile stretching frequency of KCBSO in water. Specifically, the nitrile stretch vibration in subtilisin-CBS and trypsin-CBS are blue-shifted 1.1 or 4.0 cm−1, respectively. This blue shift suggests either (1) an altered hydrogen bonding geometry or strength of the nitrile with solvent/residues in the active site compared to the nitrile group of KCBSO dissolved in water and/or (2) a different electrostatic environment of the active-site nitrile group compared to the nitrile group of KCBSO in water. X-ray crystallographic evidence suggests that the hydrogen bonding partner of the nitrile group of the bound inhibitor is likely water molecules (see below). The FWHM of the nitrile IR absorbance band is also ~2 cm−1 larger in subtilisin-CBS and trypsin-CBS than KCBSO in water on average (see Tables S1, S3, and S4 in Supporting Information), which is likely due to specific binding of CBS to the active site and nonspecific binding in ~40% of the sample.</p><p>The solvation environment of the active-site nitrile group was further assessed through temperature dependent IR spectroscopy. Specifically, the frequency-temperature line slope (FTLS) method was utilized where the sensitivity of the nitrile stretching frequency to temperature is correlated to the local environment. In order to interpret the temperature dependence of the active-site nitrile stretching frequency, the temperature dependence of the nitrile stretching frequency of KCBSO in aqueous buffer and CBSF in tetrahydrofuran (THF) were measured (Figure 5). These serve as models of a fully solvated, high dielectric and de-solvated, low dielectric local protein environments, respectively. The nitrile (KCBSO or CBSF) selected for each solvent based upon solubility limitations. THF was selected here instead of DMSO since the lower dielectric of THF is a more apt model of a buried environment in a protein, while DMSO was selected in Figure 3 due to the solubility of KCBSO in water, DMSO, and mixtures thereof.</p><p>Figure 5 shows the temperature dependence of the nitrile stretching frequency of KCBSO (open squares) dissolved in an aqueous buffer (10 mM Hepes, 20 mM KCl, pH of 7.5) and CBSF (open circles) dissolved in THF fit to a straight line. The decreased temperature range employed for the THF measurements compared to water is due to the lower boiling point of THF relative to water. The frequency shifts were referenced to the nitrile stretching frequency recorded at the lowest temperature (20.5 °C) for each solvent. The corresponding temperature dependent IR spectra and the results of the line shape analysis to determine the nitrile stretching frequencies are shown in the Supporting Information (Figures S4 and S5, Tables S1 and S2). The slope of the best fit line for the frequency shift of KCBSO in aqueous buffer was −4.5±0.1×10−2 cm−1/°C, while the slope of the frequency shift of CBSF in THF was −4.4±2.4×10−3 cm−1/°C. The relatively large (an order of magnitude greater) temperature dependence of the nitrile stretching frequency of KCBSO in aqueous buffer is due to the sensitivity of the nitrile stretch vibration to the geometry of hydrogen-bonding interactions between the nitrile and water molecules.[34,35] The nitrile stretching frequency of CBSF in THF is nearly independent of temperature due to the lack of specific interactions between the nitrile and the solvent. The direction and magnitude of these temperature dependent frequency shifts are in agreement with previous studies using pCNF.[7,8]</p><p>Figure 6 shows the temperature dependence nitrile stretching frequency shifts of subtilisin-CBS (open circles) dissolved in an aqueous buffer consisting of 10 mM Hepes and 20 mM KCl at a pH of 7.5 fit to a straight line. These data are the result of the average of two temperature dependent measurements. The temperature range was selected to ensure reversibility of the measurements. Representative temperature dependent IR spectra and the corresponding line shape analysis results are shown in the Supporting Information (Figure S6, Table S3). The slope of the best fit line for the nitrile stretching frequency shift of subtilisin-CBS in aqueous buffer was measured to be −3.5±0.5×10−2 cm−1/°C. This slope is in between the slope of KCBSO in aqueous buffer and CBSF in THF, indicating that the nitrile group of subtilisin-CBS is involved in hydrogen bonds of moderate strength, likely with solvent molecules (see X-ray crystallographic discussion below). Specifically, the magnitude of the subtilisin-CBS line slope is a factor of 1.3 less than the line slope of KCBSO in aqueous buffer and factor of 8.0 greater than the line slope of CBSF in THF. Thus, the IR temperature dependent measurements of subtilisin-CBS suggests that the active-site nitrile is mostly solvated. This FTLS result is a more robust and accurate means of assessing the local solvation environment of the nitrile group compared to using only the room-temperature nitrile stretching frequency of subtilisin-CBS.[6–8,10]</p><p>The corresponding temperature dependent IR spectra, line shape analysis, and temperature dependent nitrile stretching frequency shifts for trypsin-CBS are shown in Supporting Information (Figures S7 and S8, Table S4). The FTLS results showed the same line slope as subtilisin-CBS, suggesting the nitrile group is mostly solvated in the trypsin-CBS complex. However, these trypsin temperature dependent measurements suffered from a lack of reversibility in most trials. The FTLS analysis for trypsin-CBS reported in the supplemental and referenced above corresponds to a set of measurements that demonstrated reversibility. The lack of reversibility for trypsin attributed to the autolytic nature of this protease[36] and mass spectral results show that some free protease is still present when one equivalents of the inhibitor is added to the protease solution. Excess inhibitor was not used for the IR experiments to minimize reaction of CBSF with serine residues other than the active site serine (see below). The autolytic propensity of this residual free trypsin explains the irreversibility in some of the temperature dependent experiments given the time and temperature required for these measurements.</p><!><p>Formation of the subtilisin-CBS complex was confirmed through X-ray crystallography (Figure 7). Three conformations of the active site residue 220 were modelled: two different conformations of the bound inhibitor (CBS-A and CBS-B) and an unreacted serine (SER-C) (Figure 7B). Even with four equivalents of CBSF added prior to crystallization, SER-C was modelled into the active site to account for unreacted subtilisin to resolve –|Fo-Fc| difference density around the sulfonates when only CBS-A and CBS-B were modeled. The presence of unreacted subtilisin under these reaction conditions was confirmed by mass spectrometry. SER-C220 was a likely hydrogen bonding partner along with the carbonyl oxygen of SER124 in coordinating a water residue (HOHC) that would be absent upon inhibitor binding because of steric limitations. Both CBS-A and CBS-B were modelled at sites along the surface of subtilisin. Electron density for the CBS-A conformation indicated the nitrile group was partially solvated and directed towards GLY126, GLY127, and a calcium cation displaying partial occupancy. A potential hydrogen bonding partner along the backbone for CBS-A was the amide hydrogen of a peptide bond between GLY126 and GLY127, however the orientation of the backbone points the hydrogen away from the cyano-group, reducing the potential for interactions. Density 3.8 Å from the nitrile in CBS-A was modeled as a Ca2+ ion at 58% occupancy as the density was spherical, not fully accounted for by a water molecule, and there was an excess amount of Ca2+ in the crystallization condition. CBS-B was modelled with the nitrile directed towards a solvent exposed environment and within hydrogen bonding distance (2.5 Å) of an ordered water molecule (HOHB) (Figure 7B and C). It is possible that this water molecule reflects bulk water interactions with the inhibitor within the IR spectroscopic data. It is important to note that the IR experiments were performed with 1.0 equivalents of CBSF, unlike the 4.0 equivalents used in the crystallography work, to maximum inhibitor binding to the active site while minimizing undesired side reactions of CBSF with other serine residues so that the observed nitrile stretch frequency can be correlated with the active site solvation environment.</p><p>In the subtilisin-CBS structure the histidine (HIS63) of the catalytic triad occupies two conformations: one pointing towards the active site serine and one pointing away (Figure S10). Similar changes induced by inhibitor binding were observed in previous published crystal structures with HIS63 modelled in two conformations.[37,38] Moreover, structural alignments of subtilisin-CBS with subtilisin complexes with phenylmethylsufonyl (PMS) or vinyl-PMS, and with unreacted subtilisin all revealed that the overall tertiary structures were similar (Figure 8A).[37–39] Alignment of the subtilisin-CBS active site with subtilisin-PMS and subtilisin-vinyl-PMS active sites were illustrative in that CBS-A adopted a similar conformation to vinyl-PMS and CBS-B adopted a similar conformation to PMS (Figure 8B). While the two highest occupancy conformations of CBS are modeled, the active site serine:CBS complex maintains a great deal of flexibility and the less favorable and lower occupancy conformations are not explicitly illustrated by the two conformations modeled here. This flexibility at the inhibitor-bound active site is supported by the difference density of the final structure (Figure S9B). An ethylene glycol molecule is modeled adjacent to CBS_B (~4 Å away), however this molecule was introduced in the cryoprotection solvent and was not present in the buffer for IR spectroscopy. Therefore, it did not interrupt the electronics nor solvation dynamics of the active site during the IR experiments.</p><p>Formation of the trypsin-CBS complex was also confirmed through X-ray crystallography (Figure 9). As with subtilisin, four equivalents of CBSF were reacted with trypsin prior to crystallization although the IR experiments were performed with 1.0 equivalents of CBSF to minimize undesired side reactions. The trypsin active site was modelled with a single conformation of the inhibitor, CBS-A, and the native active site serine, SER-B (Figure 9B). Mass spectrometry indicated that under these reaction conditions some unreactive trypsin remained so SER-B was modelled to account for this. The CBS-A molecule is oriented towards the protein surface and accessible to solvents (Figure 9A and C). This orientation is consistent with the IR spectroscopic data that suggested the nitrile group was in a solvated environment (Figure 5 and S8). Structural alignment of the trypsin-CBS structure to similar phenyl-based inhibitors such as benzamide, PMSF, and benzene-boronic acid revealed that the overall tertiary structure of the protein was not significantly perturbed relative to these structures (Figure 10A).[40–42] Alignment of the inhibitors at the active site was similar between the trypsin-CBS and trypsin-PMS, while the trypsin-benzamide structure shows the feasibility for other inhibitor conformations in the active site (Figure 10B). Sulfonated serines were modelled in addition to serine at four solvent accessible serines in the structure (sites 55, 121, 131, and 169) to account for additional 2Fo-Fc electron density and +|Fo-Fc| difference density (Figure S14). The highly solvated nature of these three sites resulted in a conformational flexibility for the CBS molecules modeled at these sites and did not allow for the cyanobenzyl component of the inhibitor to be modelled. The sulfonated serine residues were each of relatively low occupancy (37, 53, 44, and 44% sulfonated serine occupancy for sites 55, 121, 131, and 169, respectively) and are potential sites where reaction with the CBSF inhibitor had occurred, consistent with mass spectrometry results indicate more than one equivalent of CBS is bound to the protein when four molar equivalents of CBSF were used. Overall, the crystal structure of trypsin-CBS suggests that the inhibitor is present in primarily one conformation at the active site and is consistent with the solvent accessibility observed in the IR experiments.</p><!><p>The active sites of two proteases have been studied by paired IR spectroscopic and X-ray crystallographic studies. The active site serines of subtilisin and trypsin were reacted with CBSF, a protease inhibitor that contains a nitrile vibrational reporter. The nitrile stretch vibration of the inhibitor and its water soluble salt (KCBSO) were calibrated by IR solvent studies in H2O/DMSO and the FTLS method in H2O/THF. The nitrile stretch vibration of the subtilisin-CBS and trypsin-CBS complexes were blue-shifted 1.1 or 4.0 cm−1, respectively, compared to the aqueous KCBSO frequency. This suggested a fully solvated environment in both proteases. The inhibitor complexes were examined by FTLS and the slopes of the best fit lines for subtilisin-CBS and trypsin-CBS in aqueous buffer were both measured to be −3.5×10−2 cm−1/°C. These slopes were intermediate in value between that of KCBSO in aqueous buffer and CBSF in THF. Thus, the more robust and accurate FTLS results suggest that the active-site nitriles in both proteases are mostly solvated. X-ray crystal structure of the subtilisin-CBS and trypsin-CBS complexes were solved at 1.27 and 1.32 Å, respectively. The inhibitor was modelled at the active site in two conformations in the subtilisin complex and in one conformation in the trypsin complex. The crystallographic data support the FTLS data that the active-site nitrile groups are mostly solvated and participate in hydrogen bonds with water molecules. The crystal structures are similar to previous structures for subtilisin- and trypsin-inhibitor complexes with low RMSDs (Figures 8 and 10).[39–42] The similarity in the structures indicates that CBS does not perturb the active sites of these proteases, thus the vibrational reporter provides insight on the native protease active sites. The combination of IR spectroscopy utilizing vibrational reporters paired with X-ray crystallography provides a powerful approach to studying protein structure. When the vibrational reporter is attached to an inhibitor this approach becomes a general method to study enzyme active sites. However, the utility of CBSF as an effective reporter of local solvation environments is hampered by its undesired side reactions with non-active site serine residues, although the active site serine reactivity remained highest. This is the result of the electron withdrawing nature of the nitrile group which creates a more reactive species compared to protease inhibitors such as PMSF. Thus, current work is underway using azide or selenocyanate vibrational reporters which should have significantly lower reactivity with non-active site serine residues while still possessing an effective vibrational reporter of local solvation environment. This general methodology pairing IR spectroscopy and X-ray crystallography with a inhibitor modified with a vibrational reporter is also applicable for the study of other proteases including cysteine proteases such as papain.</p>

PubMed Author Manuscript

Benzylation of Nitroalkanes Using Copper-Catalyzed Thermal Redox Catalysis: Toward the Facile C-Alkylation of Nitroalkanes

The C-alkylation of nitroalkanes under mild conditions has been a significant challenge in organic synthesis for more than a century. Herein, we report a simple Cu(I) catalyst, generated in situ, that is highly effective for C-benzylation of nitroalkanes using abundant benzyl bromides and related heteroaromatic compounds. This process, which we believe proceeds via a thermal redox mechanism, allows access to a variety of complex nitroalkanes under mild reaction conditions and represents the first step towards developing a general catalytic system for the alkylation of nitroalkanes.

benzylation_of_nitroalkanes_using_copper-catalyzed_thermal_redox_catalysis:_toward_the_facile_c-alky

<p>Nitroalkanes are ubiquitous reagents in organic synthesis. They are widely used as synthons for heterocycles, as radical precursors, and for heteroatom installation, including carbonyls via the Nef reaction and amines via reduction. Arguably, their most important function is serving as nucleophiles for carbon-carbon (C–C) bond construction. Although a number of reactions involving C–C bond-formation of nitroalkanes are known, including the Henry reaction,1 conjugate additions to α,β-unsaturated carbonyls,1 and palladium-catalyzed allylation2 and arylation reactions,3 the alkylation of nitroalkanes with alkyl halide electrophiles to form a new C–C bond remains a significant challenge. As early as 1908, failures of attempted C-alkylation reactions were first reported.4 In 1949, Hass and Bender reported a detailed investigation that explained the failure; treatment of nitronate anions with a variety of alkyl halides results in O-alkylation, ultimately leading to the formation of carbonyl products via nitronic esters (Scheme 1A).5 Only minor amounts of C-alkylated products are formed. The O-alkylation pathway predominates with both benzylic and aliphatic halides and with nitronate anions derived from nitromethane, primary and secondary nitroalkanes. The one exception is for ortho- or para-nitrobenzyl chloride electrophiles, which were found to react with nitronate anions at carbon.5 This unique reactivity was later shown by Kornblum and co-workers to proceed via a Srn1 pathway triggered by single electron transfer (SET) from the nitronate anion to the highly electron-deficient aromatic ring.6 However, except for this mechanistically isolated case, C-alkylation of simple nitronate anions does not occur with alkyl halide electrophiles.</p><p>Previous systems to achieve C-alkylation of nitroalkanes require either the formation of nitronate dianions at inconveniently low (−90 °C) temperature7 or the use of complex 2,4,6-trisubstituted N-alkyl pyridinium ions as electrophiles (this latter reaction also proceeds via an Srn1 pathway).8 Both of these methods have serious limitations in preparative chemistry. A procedure using readily available alkyl halides to C-alkylate nitroalkanes under mild reactions conditions would greatly expand both the preparation and utility of nitroalkanes in organic synthesis.</p><p>In this communication, we report the first step towards a practical solution to this century-old problem. We disclose the development of conditions for the benzylation of nitroalkanes using electron-rich Cu(I) catalysts (Scheme 1B). These reactions occur at mild temperature (60 °C), employ benzyl bromides and inexpensive precatalysts, and afford high yields. We believe that these reactions proceed via a thermal redox catalysis pathway. Importantly, this process is very general with respect to both the benzyl bromide and nitroalkane, including the use of secondary nitroalkanes. This wide scope allows preparation of a variety of complex nitroalkanes. In addition, this method enables facile preparation of phenethylamine derivatives, which are important medicinal agents.</p><p>In considering means to effect the C-alkylation of nitroalkanes, we were particularly drawn to the potential use of radical chemistry. In addition to the radical pathways elucidated by Kornblum6 and Katritzky,8 photogenerated alkyl radicals, generated via the homolytic fragmentation of mercury- or cobalt-alkyls, have been shown to react with nitronate anions at carbon.9 Although of limited synthetic utility, these reactions demonstrate that radical-anion coupling involving nitronate anions is feasible.</p><p>Simultaneously, we were cognizant of recent work in the area of metal-catalyzed alkylation of carbon nucleophiles using alkyl halides.10 Many of these reactions have been shown to involve radical intermediates. We were particularly drawn to the copper-based catalyst systems used in the mechanistically related Atom Transfer Radical Addition (ATRA) and Atom Transfer Radical Polymerization (ATRP) reactions, in which Cu(I) catalysts initiate radical reactions of substituted alkenes by undergoing a SET reaction with alkyl halides bearing a wide range of radical stabilizing groups.11 Given the propensity of nitronate anions to undergo reactions with radical intermediates, we reasoned that a copper-based catalyst might promote C-alkylation using readily prepared or commercially available alkyl halides via a pathway involving SET followed by radical-anion coupling (Scheme 2).</p><p>We began our investigation by examining the reaction of 1-nitropropane and benzyl bromide in benzene. Under basic conditions in the absence of catalyst, only trace desired 1-phenyl-2-nitrobutane (7) was observed (<5% by NMR). The major product in these reactions was benzaldehyde (12% by NMR, Table 1, entry 1) along with unreacted starting material. Attempts to employ catalysts derived from palladium, cobalt, nickel or iron lead to similar results (not shown). With the use of CuBr, and simple ligands such as PPh3 or bipyridyl 1, a modest increase in the desired product was seen (entries 2 and 3). Interestingly, the neutral polydentate ligands 2a and 2b, which are often very effective ligands in ATRA/ATRP reactions, were less effective (entries 4 and 5).</p><p>In contrast, trans-N,N′-dimethyl-1,2-cyclohexanediamine (3a), a ligand that has been used in copper-catalyzed Goldberg-type reactions12 but not often used in atom-transfer reactions, led to more promising results. Using this ligand, 7 was observed in 45% yield (entry 6). Unfortunately, efforts to optimize this ligand design were unsuccessful. However, during these studies we noted a major byproduct from the reaction was the dibenzylated ligand 3b. Independent preparation of 3b revealed that it was ineffective as a ligand in the catalytic reaction (entry 7).13 Similar results were observed for other tetra-alkyl diamine ligands, leading us to speculate that the protic N–H bond of 3a might be integral to its success in the reaction; we postulated that the active catalyst might arise from deprotonation of the ligand under the reaction conditions leading to the formation of a highly electron-rich Cu(I)-amido species.</p><p>This line of reasoning led us to examine the use of 1,3-diketimine (nacnac) ligands in the reaction. We predicted that the acidic nature of the nacnac backbone would rapidly result in the formation of a neutral Cu(I)-nacnac under the basic reaction conditions.14 Further, we hoped that the steric bulk of the nacnac architecture would prevent competitive alkylation of the ligand. Using nacnac 4, a 64% yield of 7 was observed under the initial screening conditions. Extensive attempts to optimize the reaction through modulation of the nacnac framework proved unsuccessful (see Supporting Information); however further studies revealed a significant effect of the base counter-ion, with sodium proving optimal in terms of yield and ease of use (entry 12 vs. 13).15,16 Non-polar solvents were also generally favored, with hexanes being the most effective in the screening reaction. Using these optimized conditions, the desired 2° nitroalkane could be isolated in 85% yield on a 1 mmol scale (entry 14).17</p><p>The scope of the reaction with respect to benzyl bromide is broad (Table 2). A wide-range of functional groups are tolerated, including fluorides, chlorides, bromides, nitriles, esters, ethers, and tri-fluoromethyl groups. Both electron-rich (13) and electron-poor (14, 20, and 21) benzyl bromides participate in the reaction, and there is remarkably little variance in the yield of product due to the electronic effects of the arene substituent. The reactions of more sterically encumbered benzyl bromides, such as those containing an ortho methyl group (10), and polyaromatic substrates (23) also proceed without incident. Para-nitrobenzyl bromide also reacts to provide the C-alkylated product under the copper-catalyzed reaction conditions (22).5 Finally, bromomethyl-substituted hetereoaromatic compounds also can be used in the reaction. For example, treatment of 2-bromomethylpyridine hydrobromide with 1-nitropropane lead to nitropyridine 24. Other heteroaromatics, including quinolones (25), thiophenes (26), and benzoxazoles (27) are also efficient substrates.18 The reaction was easily scaled; compound 19 was isolated in 82% yield from a 2.5 gram reaction. In all cases, only trace amounts aldehyde (1–5%) were observed. The major byproduct detected (NMR and GC) was the bibenzyl resulting from dimerization of the alkylating reagent.</p><p>The reaction also enjoys wide substrate scope with respect to the nitroalkane (Table 3). Longer aliphatic nitroalkanes, such as nitrohexane, participated in the reaction well (28). Branching beta to the nitro group was tolerated (29). A range of functional groups on the nitroalkane proved compatible with the transformation, including alkenes, esters, amides and acyl-protected alcohols (30–33). All of these reactions proceeded in good yield under the standard reaction conditions or slight modifications thereof. Nitromethane can also be alkylated using this catalyst system in good yield (73%, 34), provided it is used in excess (7.5 equiv). Under these conditions, good selectivity for the monoalkylated product is observed; with less nitromethane double alkylation competes.</p><p>Importantly, secondary nitroalkanes are also tolerated in the reaction. For example, benzylation of 2-nitropropane resulted in a 71% isolated yield of 35 (Table 3). This transformation allows for the direct construction of a fully substituted carbon bearing a nitrogen substituent, which remains a challenging problem in organic synthesis.19 Not surprisingly, this reaction proceeded more slowly than those employing primary nitroalkanes. Interestingly, however, this reaction was very sensitive to the choice of solvent, and cyclohexane provided consistently higher yields than hexanes, which was employed in the other reactions. The reason for this solvent effect is not clear – no additional byproducts, such as reduced starting materials, were detected. Other secondary nitroalkanes can participate in the reaction, including nitrocyclohexane (37) and those bearing functional groups (38).20</p><p>The ability of secondary nitroalkanes to participate in the reaction opens the possibility for sequential alkylation reactions (Scheme 3). For example, as reported above, alkylation of nitropropane with 4-bromobenzyl bromide gave rise to nitroalkane 19 in 82% yield. Subsequent alkylation of that product with methyl 4-(bromomethyl)-benzoate resulted in tertiary nitroalkane 39 in 65% yield. Such sequential alkylation reactions promise the ability to rapidly prepare complex nitroalkanes and amines from very simple starting materials.</p><p>There is clear relevance of the nitroalkane products from the copper-catalyzed benzylation reaction to the preparation of bioactive molecules. Phenethylamines are important medicinal agents, which have found wide use in the treatment of obesity and other metabolic diesases.21 These compounds can be readily prepared from β-phenyl nitroalkanes.1 As an illustration of the utility of our catalytic process, simple hydrogenolysis of nitroalkane 35 provided the tertiary amine phentermine (40) in high yield. Phentermine is a clinically prescribed anorectic (appetite suppressant) for the treatment for obesity.22 It is typically prepared via the Henry reaction of benzaldehyde and 2-nitropropane followed by a multi-step reduction sequence,23 or via a Ritter reaction of the corresponding tertiary alcohol and subsequent hydrolysis,24 both of which require more steps than the sequence reported herein.</p><p>Mechanistically, we postulate that these reactions are proceeding via a thermal redox mechanism involving single electron transfer (SET) from the electron-rich copper catalyst to the benzyl bromide (Scheme 4). Upon loss of halide, this process would generate a neutral benzylic radical, which could undergo coupling with the nitronate anion. Electron transfer from the resulting nitronate radical would regenerate the copper catalyst, closing the catalytic cycle. The observation of bibenzyl side products is consistent with a single electron pathway.25</p><p>In summary, we have developed a catalytic system for the benzylation of nitroalkanes that utilizes readily available benzyl halides and related hetereoaromatic compounds. This protocol addresses a century-old gap in C–C bond construction and provides the first example of alkylation of nitroalkanes using readily available starting materials under mild reaction conditions. This reaction allows for the conversion of simple starting materials to complex nitroalkanes, which are important synthetic intermediates in the preparation of bioactive molecules, such as phenethylamines. The key to this discovery was the identification of a highly electron-rich Cu(I)-nacnac complex, which can be prepared in situ and is capable of facile reduction of the benzyl halide to the corresponding radical. This thermally driven process clearly bears mechanistic resemblance to catalytic photoredox systems, the synthetic utility of which has been elegantly demonstrated by several groups.26,27 Efforts to expand our copper-based system to other types of catalytic redox reactions, as well as to expand the scope of the nitroalkane alkylation chemistry to other classes of alkyl halides, are currently underway in our laboratory.</p>

PubMed Author Manuscript

Divergent Synthesis and Chemical Reactivity of Bicyclic Lactone Fragments of Complex Rearranged Spongian Diterpenes

The synthesis and direct comparison of the chemical reactivity of the two highly oxidized bicyclic lactone fragments found in rearranged spongian diterpenes (8-substituted 6-acetoxy-2,7-dioxabicyclo[3.2.1]octan-3-one and 6-substituted 7-acetoxy-2,8-dioxabicyclo[3.3.0]octan-3-one) are reported. Details of the first synthesis of the 6-acetoxy-2,7-dioxabicyclo[3.2.1]octan-3-one ring system, including an examination of several possibilities for the key bridging cyclization reaction, are described (Schemes 2\xe2\x80\x935). In addition, the first synthesis of 7-acetoxy-2,8-dioxabicyclo[3.3.0]octanones containing quaternary carbon substituents at C6 is disclosed (Scheme 6). Aspects of the chemical reactivity and Golgi-modifying properties of these bicyclic lactone analogs of rearranged spongian diterpenes are also reported. Under both acidic and basic conditions, 8-substituted 2,7-dioxabicyclo[3.2.1]octanones are converted to 6-substituted-2,8-dioxabicyclo[3.3.0]octanones. Moreover, these dioxabicyclic lactones react with primary amines and lysine side chains of lysozyme to form substituted pyrroles, a conjugation that could be responsible for the unique biological properties of these compounds. These studies demonstrate that acetoxylation adjacent to the lactone carbonyl group\xe2\x80\x94in either the bridged or fused series\xe2\x80\x94is required to produce fragmented Golgi membranes in the pericentriolar region that is characteristic of macfarlandin E.

divergent_synthesis_and_chemical_reactivity_of_bicyclic_lactone_fragments_of_complex_rearranged_spon

Introduction<!>Enantioselective Synthesis of t-Bu-MacE<!>Enantioselective Synthesis of 7-Acetoxy-2,8-Dioxabicyclo[3.3.0]octan-3-ones<!>Chemical Reactivity of the 6-Acetoxy-2,7-dioxabicylo[3.2.1]octan-3-one and 7-Acetoxy-2,8-dioxabicylo[3.3.0]octan-3-one Ring Systems<!>Effects of 6-Acetoxy-2,7-dioxabicylo[3.2.1]octan-3-ones and 7-Acetoxy-2,8-dioxabicylo[3.2.1]octan-3-ones on Golgi Organization<!>Conclusions

<p>Skeletal rearrangement and oxidation of spongian diterpene precursors of general structure 1 (Figure 1A) provides a structurally diverse group of marine-derived natural products referred to as rearranged spongian diterpenes.1 These natural products are isolated from sponges and dorid nudibranchs, the latter of which are believed to acquire these diterpenes from sponge sources as a chemical defense mechanism.1 Among the most structurally complex of the rearranged spongian diterpenes is a structurally unique group that contain a polycyclic hydrocarbon fragment joined to an oxidized lactone unit (Figure 1). The biological activity of this group of spongian-derived diterpenes has been characterized to only a limited extent. Antimicrobial,2,3,4 cytotoxic,4 and nematocidal4 activities have been reported, and norrisolide (9) has been shown to induce irreversible fragmentation and delocalization of Golgi membranes throughout the cytosol in human cell lines.5 In addition, we reported in 2010 that macfarlandin E (4, MacE) and a simplified analog 13 (t-Bu-MacE, see eq 1) induce a novel Golgi organization phenotype that is characterized by small, pericentriolar Golgi fragments and blockage of protein transport from the Golgi to the plasma membrane.6</p><p>A central feature of the group of rearranged spongian natural products depicted in Figure 1 is the presence of highly oxygenated and hydrophobic subunits. The oxygenated fragment is structurally diverse and includes monocyclic variants, such as that found in shahamin K (2),7 and more complex dioxabicyclic lactone fragments. The 6-acetoxy-2,7-dioxabicyclo[3.2.1]octan-3-one subunit is particularly rare; at the onset of our studies, it was known only in the diterpene macfarlandin E (4)2,8 and 13 additional rearranged spongian diterpenes such as aplyviolene (3),8 chromodorolide A (5),4a shahamin I (6),9 and norrlandin (7)10 (Figure 1B). The isomeric 7-acetoxy(or hydroxy)-2,8-dioxabicyclo[3.3.0]octan-3-one fragment is somewhat more abundant in rearranged spongian diterpenes,11 as exemplified by dendrillolide A (8),12 norrisolide (9),13 cheloviolenes A (10) and B (11),14 and omriolide A (12)15 (Figure 1C). The substituted dioxabicyclo[3.3.0]octanone ring system of these diterpenes has been the subject of limited synthetic efforts,16 highlighted by two total syntheses of norrisolide.17,18</p><p>Last year we disclosed the first synthesis of the 4,6-diacetoxy-2,7-dioxabicylo[3.2.1]octan-3-one moiety of MacE and evidence obtained by evaluation of related diterpenes and various analogs that the entire oxygenated subunit—including both acetoxy substituents—is required to elicit the MacE Golgi phenotype.6 Additionally, we showed that t-Bu-MacE (13) and its deacetoxy congener 14 are converted to substituted pyrroles such as 15 and 16 in the presence of primary amines under mild conditions (eq 1). This latter finding suggests a functional role for the oxygenated ring system of MacE—formation of a protein-bound pyrrole species—which could be responsible for the unique biological properties of these compounds. (1) </p><p>Although both MacE and norrisolide have pronounced effects on the structure and function of the Golgi, their phenotypes are different: MacE causes the conversion of the Golgi ribbon into membrane fragments that remain in the pericentriolar region, whereas norrisolide induces fragmentation and dispersal of Golgi membranes throughout the cytoplasm.6,7 In addition, the structure-activity relationships reported to date for these two rearranged spongian natural products are quite distinct: the hydrophobic fragment is suggested to be essential for norrisolide's activity,7 whereas the full 4,6-diacetoxy-2,7-dioxabicylo[3.2.1]octan-3-one subunit, and not the hydroazulene fragment, is believed to be essential for eliciting the MacE-Golgi phenotype.6</p><p>This article reports the synthesis of simplified analogs of the oxygenated subunits of MacE (4) and dendrillolide A (8) and a survey of their chemical reactivity and Golgi-modifying properties. We find that both 6-acetoxy-2,7-dioxabicyclo[3.2.1]octan-3-ones and 7-acetoxy-2,8-dioxabicyclo[3.3.0]octan-3-ones react under mild conditions with benzylamine or lysine side chains of lysozyme to form substituted pyrrole products. Moreover, we show that the presence of acetoxy substitution adjacent to the lactone carbonyl in either the bridged or fused dioxabicyclooctanone ring system induces a nearly identical Golgi organization phenotype and greater reactivity with lysine side chains.</p><!><p>As an appropriate initial target for developing a chemical synthesis of the 4,6-diacetoxy-2,7-dioxabicyclo[3.2.1]octan-3-one ring system, we chose t-Bu-MacE (13) which possesses a tert-butyl group in place of the hydroazulene subunit of MacE. Our synthetic approach to t-Bu-MacE was based on the prospect that kinetically controlled cyclization of dialdehyde acid 17 would not form the desired dioxabicyclo[3.2.1]octanone but rather the 2,8-dioxabicyclo[3.3.0]octan-3-one isomer 19 (eq 2). Moreover, we anticipated, and later established experimentally (see below), that the fused isomer also would be favored thermodynamically. Bolstering our expectation that direct cyclization of 17 was unlikely to generate the 2,7-dioxabicyclo[3.2.1]octan-3-one ring system is the conversion outlined in equation 3. In this example, even enforced 1,3-diaxial proximity of the carboxyl nucleophile and the distal aldehyde of precursor 20 did not result in forming the bridged dioxabicyclo product, but rather fused isomer 21.19 (2) (3) </p><p>The analysis that guided our initial synthesis of t-Bu-MacE (13) is summarized in Scheme 1. Late stage Baeyer–Villiger oxidation was anticipated to form the C6-acetoxy substituent from an acetyl precursor.18,20 It was anticipated initially that the C4 acetate could be installed at a late stage as well. Bridged-bicyclic intermediate 22 was seen arising from the cyclic acetal 23, wherein X is a leaving group. Acyclic tricarbonyl intermediate 24 was seen resulting from oxidative cleavage of cyclopentyloxysilane 25, which in turn would arise from facial-selective conjugate addition of a tert-butylcuprate to cyclopentenone 26 in the presence of a silyl electrophile.</p><p>Guided by these considerations, we initiated the synthesis of t-Bu-MacE by targeting cyclopentenone 26. Attempts to access enone 26 from known cyclopentanone 2721 were unsuccessful, as regioselective installation of the double bond proved problematic under a variety of conditions (Scheme 2).22 Consequently, enantiopure cyclopentenone 29, which is readily available in 4 steps from cyclopentadiene, was used as the Michael acceptor.23 The union of silyl ketene acetal 2824 and enone 29 to afford cyclopentenone 30 was realized in good yield in dry DMF in the presence of lithium acetate.25 Earlier attempts to promote this reaction with Lewis acids (SnCl4 or TiCl4 in CH2Cl2) or tris(dimethylamino)sulfonium difluorotrimethylsilicate (TASF) were less successful. After consumption of the starting materials at 0 °C, addition of a slight molar excess of water to the reaction mixture and allowing the reaction to warm to room temperature prior to aqueous work-up delivered the enone product in reproducibly good yield. If this step was omitted, conjugate addition product 31 was obtained in 30–60% yield.26 We speculate that the trimethylsilyl group is only partially transferred in the initial 1,4-addition; addition of water and warming to room temperature is believed to transform any enoxysilane intermediate to the corresponding ketone, allowing enolate equilibration and β-elimination to take place. Cyclopentenone 26 was then obtained in high yield by cleavage of the butane diacetal group with aqueous trifluoroacetic acid and methylation of the resulting carboxylic acid.27</p><p>In five steps, enone 26 was elaborated in good overall yield to tricarbonyl intermediate 24 (Scheme 3). The tert-butyl substituent was first incorporated by stereoselective addition of the cuprate reagent generated from tert-butyllithium and CuCN (2:1 molar ratio) in the presence of tert-butyldimethylsilyl chloride (TBSCl) to provide exclusively trans-substituted cyclopentenyl silyl ether 32.28 At this stage, we needed to transform the methyl ester to an acetyl group without cleaving the enoxysilane. A two-step sequence, proceeding via methyl enol ether intermediate 33, was eventually developed. Although methylenation of 32 with the Tebbe reagent29 proceeded sluggishly, resulting in incomplete conversion, the Takai methylenation conditions developed by Rainer and coworkers produced methyl enol ether 33 in nearly quantitative yield.30 After examining several conditions for hydrolyzing the methyl enol ether, the use of 1.5 equiv of oxalic acid in aqueous i-PrOH at 0 °C was identified as particularly effective in achieving the conversion to methyl ketone 25 with minimal cleavage of the enoxysilane. We ascribe this rare, if not unprecedented, selective acidic cleavage of a methyl enol ether in the presence of a silyl enol ether to steric shielding of protonation of the cyclopentenyl double bond by the two bulky trans-oriented substituents.31 In our initial efforts, we cleaved the double bond of 25 by reaction with OsO4 and NaIO4, followed by addition of trimethylsilyldiazomethane32 to the crude product mixture to provide tricarbonyl product 24 in useful, albeit variable, yields (55–84%).6 To improve the reproducibility of this conversion and avoid the undesirable use of TMSCHN2, an alternate sequence was developed. In this improved procedure, enoxysilane 25 was oxidized with OsO4 (0.05 equiv) and N-methylmorpholine-N-oxide (NMO, 2.0 equiv), followed by cleavage of the resulting α-hydroxyketone with 1.3 equiv of methanolic Pb(OAc)4, a sequence that reproducibly delivered tricarbonyl intermediate 24 in 95% yield.</p><p>As a prelude to examining the pivotal bridging reaction, acyclic intermediate 24 was transformed to several potential tetrahydrofuryl cyclization precursors (Scheme 4). Cleavage of the silyl ether by reaction of 24 with 1.5 equiv of TBAF provided tetrahydrofuryl lactol 34 in moderate yield, which upon careful saponification with 1.5 equiv of NaOH at 0 °C gave the crude carboxylic acid 35 in sufficient purity for subsequent cyclization studies. Attempts to access intermediate 35 more expeditiously from cyclopentenyl precursor 25 by sequential reaction with OsO4/NaIO4 and TBAF were unsuccessful, providing only complex mixtures of products. Reaction of tricarbonyl intermediate 24 with 1.6 equiv of camphorsulfonic acid (CSA) and trimethylorthoformate in methanol at room temperature generated diacetal 36 as a 1:1 mixture of separable methoxy anomers in 65% yield. The α epimer of 36 could be converted to the β epimer by equilibration in methanol in the presence of camphorsulfonic acid, allowing the β-36 to be obtained in 66% overall yield from acyclic precursor 24 after two recycles. Subsequent saponification of the ester and selective cleavage of the dimethyl acetal with aqueous HCl at 0 °C yielded crude tetrahydrofuryl ketoacid 37. As glycosyl fluorides are used extensively as glycosyl donors because of their stability and the mild, orthogonal methods available for their activation,33 hemiacetal 34 was transformed to anomeric fluoride 38 by reaction with 1.6 equiv of diethylaminosulfur trifluoride (DAST) at −78 °C in CH2Cl2. Saponification of this product with 1.5 equiv of 1 N NaOH in MeOH at room temperature provided carboxylic acid 39.</p><p>With cyclization substrates 35, 37, and 39 in hand, their transformation to 2,7-dioxabicyclo[3.2.1]octan-3-one 22 was studied (Table 1). Mitsunobu conditions failed to promote cyclization of hemiacetal 35 (entry 1); however, in the presence of 1 equiv of camphorsulfonic acid (CSA) in chloroform, lactol 35 was converted at room temperature to dioxabicyclooctanone 22 in 54% overall yield from ester precursor 34 (entry 2). In a similar fashion, methoxy acetal 37 was transformed in the presence of 1 equiv of BF3•Et2O to dioxabicyclooctanone 22 in 66% yield (entry 3). Anomeric fluoride 39 cyclized with slightly enhanced efficiency when exposed to 2 equiv of SnCl2 in DMF at room temperature, generating bicyclic lactone 22 in 71% yield (entry 4).34</p><p>With several complementary methods for accomplishing the critical bridging lactonization reaction identified, all that remained was installing the additional oxygen functionality of t-Bu-MacE. The C6 acetoxy substituent was readily introduced by reaction of dioxabicyclooctanone 22 with 4 equiv of trifluoroperacetic acid at room temperature,35 providing the tert-butyl analog of aplyviolene, 6-acetoxy-2,7-dioxabicyclo[3.2.1]octan-3-one (14), in 88% yield. We had hoped that the enhanced acidity of lactones relative to esters would allow the remaining acetoxy substituent to be incorporated at this stage.36 However, all attempts to directly introduce a hydroxyl substituent adjacent to the lactone carbonyl group were unsuccessful. For example, enolization of 14 with 1–2 equiv of LDA or KHMDS, followed by oxidation (Davis oxaziridine37 or O2), silylation, or acylation resulted in either recovered starting material or extensive decomposition.38 (4) </p><p>As a result of our inability to selectively oxidize bicyclic lactone 14, we turned to examine introduction of the α-acetoxy substituent prior to forming the dioxabicyclo[3.2.1]octanone ring system. Hydroxylation of the sodium enolate of the β-methoxy epimer of tetrahydrofuryl acetal 36 with oxaziridine 4039 proceeded smoothly to generate α-hydroxy ester 41 as a single stereoisomer (Scheme 5). The stereoselectivity of this oxidation is rationalized by orientation of the enolate away from the tetrahydrofuran ring and delivery of the electrophile to the face opposite the tert-butyl group. As the hydroxyl substituent of 41 has the opposite relative configuration to the corresponding acetate of MacE, it was inverted by oxidation with Dess–Martin reagent40 and subsequent stereoselective reduction of the ketone product with sodium borohydride at −78 °C. This sequence provided alcohol 42 in 88% overall yield. Saponification of the ester and acidic hydrolysis of the ketal, followed by exhaustive acylation and anhydride methanolysis yielded α-acetoxycarboxylic acid 43. We were delighted to find that tetrahydrofuryl acetal 43, when exposed to 1.1 equiv of BF3•OEt2 at 0 °C, gave rise to 2,7-dioxabicyclo[3.2.1]octan-3-one 44 in 71% overall yield from tetrahydrofuran 42. Baeyer–Villiger oxidation of this product with trifluoroperacetic acid then provided t-Bu-MacE (13) in 90% yield.41 The synthetic sequence outlined in Schemes 2–5 allowed 0.5 g of t-Bu-MacE to be synthesized, enabling the chemical reactivity and biological profile of this compound to be studied in detail.</p><!><p>Several intermediates prepared during the synthesis of t-Bu-MacE provide potential access to related structures in the 2,8-dioxabicyclo[3.3.0]octan-3-one series. We demonstrated this chemistry with the α-methoxy epimer of intermediate 36, thus allowing both epimers of acetal 36 to be utilized (Scheme 6). Sequential ester saponification and ketal hydrolysis cleanly provided tetrahydrofuran α-37 from acetal precursor α-36. Baeyer–Villiger oxidation of this intermediate took place with concomitant cyclization to provide 7-methoxy-2,8-dioxabicyclo[3.3.0]octanone 45 in 82% overall yield for the three steps. Hydrolysis of 45 with dilute HCl yielded a mixture of bicyclic lactols, which upon acetylation delivered crystalline 7-acetoxy-2,8-dioxabicyclo[3.3.0]octanone 46 as a single stereoisomer. The structure and relative configuration of this product was initially secured by 1H NMR NOE analysis and subsequently confirmed by single crystal X-ray diffraction.42</p><p>Analog 49, which possesses an acetoxy group adjacent to the lactone carbonyl, was also prepared. The enolate of bicyclic lactone 45 was generated with NaHMDS in THF at −78 °C and hydroxylated to afford 47 (Scheme 6). Subsequent acetylation of the secondary alcohol gave acetate 48 in low (unoptimized) yield over two steps.43 The methoxy acetal functional group of 48 was hydrolyzed with dilute HCl, followed by acetylation of the hemiacetal product with acetic anhydride and pyridine to provide an inseparable mixture of 4,7-diacetoxy-2,8-dioxabicyclo[3.3.0]octan-3-one 49 and an uncharacterized aldehyde byproduct. Pure dioxabicyclo[3.3.0]octanone 49 was obtained in modest yield from this crude product mixture by sodium chlorite oxidation,44 which allowed for easy removal of carboxylic acid impurities by chromatography. Of note, the syntheses summarized in Scheme 6 are the first of 7-acetoxy-2,8-dioxabicyclo[3.3.0]octan-3-ones possessing a quaternary-carbon substituent at C6, a structural feature found in many rearranged spongian diterpene natural products (see Figure 1).</p><!><p>Our expectation that the 2,7-dioxabicyclo[3.2.1]octan-3-one ring system would be less stable than the isomeric 2,8-dioxabicyclo[3.3.0]octan-3-one ring system was readily confirmed by exposure of dioxabicyclo[3.2.1]octanone 14 to BF3•OEt2 in acetic acid at 0 °C to generate dioxabicyclo[3.3.0]octanone isomer 46 as 3.5:1 mixture of separable acetal epimers, favoring the β-acetoxy isomer (Scheme 7).45 Alternatively, hydrolysis of 14 at room temperature with 1 N HCl in THF gave a mixture of dioxabicyclo[3.3.0]octanone lactol epimers 50,46 which upon acetylation provided the α epimer of dioxabicyclo[3.3.0]octanone 46 exclusively. This sample was identical to the material prepared by the approach outlined in Scheme 6. In a similar fashion, acidic hydrolysis of t-Bu-MacE (13) provided a lactol intermediate, which was identical to the product formed by acidic hydrolysis of 48. Ensuing acetylation of this lactol intermediate gave α-acetoxylactone 49 in 36% overall yield from t-Bu-MacE.</p><p>The transformation of the 2,7-dioxabicyclo[3.2.1]octan-3-one ring system to fused bicyclic lactone products could also be accomplished under basic conditions. In the presence of sodium hydroxide at room temperature, dioxabicylo[3.2.1]octanone 14 yielded a single 2,8-dioxabicylo[3.3.0]octan-3-one lactol product. However, in this case, 1H NMR NOE analysis showed that the tert-butyl group of product 51 resides on the convex face of the 2,8-dioxabicyclo[3.3.0]octanone ring system (Scheme 8). Confirmation that this product resulted from epimerization of the carbon bearing the tert-butyl substituent, and not the methine hydrogens of the ring junction, was obtained by conversion of 51 to the R and S Mosher esters 52. Enhanced Mosher analysis established that 52 possesses the 1R,5R,6S,7R absolute configuration with a slightly eroded enantiomeric purity of 78% ee.47</p><p>The formation of product 51 from dioxabicyclo[3.2.1]octanone precursor 14 requires further comment. Certainly isomer 51 having the tert-butyl group on the convex face of the cis-dioxabicyclo[ 3.3.0]octanone ring system should be more stable than stereoisomer 50. Under the basic reaction conditions (pH ~14), the predominant aldehyde intermediates generated from hydroxide opening of the lactone (or cleavage of the acetate substituent) of precursor 14 would be expected to be 53 and 54 (eq 5). Base-promoted epimerization of the likely predominant species, uncharged 54, would lead to the formation of the observed product 51. The partial erosion of enantiomeric purity observed in the conversion of 14 to 51 indicates that there is some epimerization under these conditions of the central methine carbon of intermediate 53.48 (5) </p><p>To gain further insight into the reactivity of these dioxabicyclic lactone ring systems, we examined the rate of hydrolysis of compounds 13, 14, 46, and 49. Hydrolytic rate was measured by NMR observation of the disappearance of the dioxabicyclic ring system in a pD 8.3 phosphate buffer containing 10% DMSO at 37 °C (Table 2). The presence of an acetoxy group adjacent to the lactone carbonyl results in an enhanced hydrolysis rate of both the substituted dioxabicyclo[3.2.1]octanone and dioxabicyclo[3.3.0]octanone ring systems, with the latter ring system hydrolyzing more rapidly than the former.</p><p>In our initial report, we demonstrated that the reaction of substituted dioxabicyclo[3.2.1]octanones 13 and 14 with benzylamine leads to substituted pyrroles.6 Specifically, we showed that t-Bu-MacE (13) was converted to pyrrole 15 in a mixture of perdeutero-benzylamine (2.5 equiv) and THF-d8, and that dioxabicyclooctanone 14 is transformed to pyrrole 16 in high yield in the presence of 5 equiv of benzylamine in DMSO and water (eq 1). Under these conditions, fragmentation of these precursors undoubtedly generates transient 1,4-dialdehyde intermediates, 17, which undergo Paal–Knorr pyrrole formation (Scheme 9).6 In the case of the pyrrole derived from t-Bu-MacE (13), the oxygen substituent of the acetic acid side chain is exchanged for a benzylamino substituent, likely after formation of the pyrrole by a gramine-type fragmentation/addition pathway.</p><p>The conversion of t-Bu-MacE (13) and analog 14 to pyrrole products was examined in CD3OD to gain insight into the initial step of the pyrrole-forming process under protic conditions (Scheme 10).49 NMR analysis of the reaction of 13, perdeutero-benzylamine (5 equiv), and CD3OD at room temperature indicated that pyrrole 57 was formed along with equimolar quantities of acetic acid and methyl acetate.50 Similarly, when 14 was converted to 58 under identical conditions, methyl acetate was observed.49 The formation of methyl acetate, as well as the exclusive formation of the carboxylates 57 and 58, demonstrates that fragmentation of the 6-acetoxy-2,7-dioxabicyclo[3.2.1]octanones 13 and 14 is initiated by initial reaction of the protic solvent at the anomeric acetoxy substituent.</p><p>Pyrroles are also formed from the reaction of 7-acetoxy-2,8-dioxabicyclo[3.3.0]octan-3-ones with primary amines. For example, 7-acetoxydioxabicyclo[3.3.0]octanone 46 was converted to pyrrole carboxylic acid 59 when exposed at room temperature to 2 equiv of benzylamine in DMSO/H2O (Scheme 11). When this reaction was conducted in methanol, pyrrole methyl ester 60 was observed as the major product. The formation of methyl ester 60 suggests, in contrast to bridged dioxabicyclo[3.2.1]octanone compounds 13 and 14, that fragmentation in this series is initiated by reaction of the protic solvent at the lactone carbonyl group.</p><p>The divergence in fragmentation pathways of the two isomeric bicyclic lactone ring systems (summarized in Scheme 12) is consistent with the reactivity of related simple lactones. Although sixmembered lactones are typically more reactive than their five-membered ring counterparts,51 oxabicyclo[3.2.1]octanone 61 is saponified at a dramatically reduced rate (Figure 2).52 The reduced rate of saponification of lactone 61 is readily ascribed to developing destabilizing syn-pentane interactions during axial approach of hydroxide to the lactone carbonyl group. Similar destabilizing interactions would be involved in the tetrahedral intermediate generated from the addition of nucleophiles to the lactone carbonyl of 6-acetoxy-2,7-dioxabicyclo[3.2.1]octanones 14.</p><p>To ascertain whether the oxygenated bicyclic lactones would react with lysine residues of proteins under biologically relevant conditions, we examined the reactivity of these molecules with hen egg white lysozyme (HEWL).53 In initial experiments, we found that dioxabicyclo[3.2.1]octanones 13 and 14 converted lysine residues of HEWL to pyrrole adducts at room temperature in pH 7 phosphate buffer (Figure 3 and Table 3, entries 1 and 3). Analog 14 provided the pyrrole-3-acetic acid modification (+164 mu), whereas t-Bu-MacE (13) led to the pyrrole-3-hydroxyacetic acid adduct (+180 mu) resulting from solvolytic incorporation of a hydroxyl substituent at the heterobenzylic site (Figure 3). The reaction of dioxabicyclo[3.3.0]octanones 46 and 49 with HEWL at pH 7 and 8 was also examined. As expected, 6-acetoxy-2,7-dioxabicyclo[3.2.1]octan-3-ones and isomeric 7-acetoxy-2,8-dioxabicyclo[3.2.1]octan-3-ones provided identical pyrrole adducts: 14 and 46 (+164 mu), 13 and 49 (+180 mu).54 t-Bu-MacE (13) and 49, which possess an acetoxy substituent adjacent to the lactone carbonyl group, modified lysozyme to a greater extent than the corresponding desacetoxy congeners 14 and 46. Subjecting the modified lysozymes to trypsin digestion, followed by standard MALDI-MS-MS peptide sequencing, revealed that the two most surface-accessible lysines (K-33 and K-97) were the predominant sites of covalent modification.55</p><!><p>Synthetic access to 2,8-dioxabicyclo[3.3.0]octan-3-ones 46 and 49 allowed us to compare the Golgi-modifying properties of these structures with those of isomers in the 2,7-dioxabicyclo[3.2.1]octan-3-one series. Normal Rat Kidney (NRK) cells grown on coverslips were incubated with analogs 46 and 49 for 60 min at 37 °C, followed by examination of the Golgi for the MacE-induced reorganization phenotype by immunofluorescence analysis with antibodies to the known Golgi resident protein, mannosidase-II.56 The 4,7-diacetoxy-2,8-dioxabicyclo[3.3.0]octanone 49 produced a Golgi phenotype indistinguishable from that of MacE or t-Bu-MacE (13), with the conversion of the Golgi ribbon into small fragments that remained localized adjacent to the centrosome (Figure 4). By contrast, 7-acetoxy-2,8-dioxabicyclo[3.3.0]octanone 46, which lacks an acetoxy substituent adjacent to the lactone carbonyl group, did not affect Golgi structure at concentrations of up to 80 µg/mL. As 6-acetoxy-2,7-dioxabicyclo[3.2.1]octanone 14 also did not impact Golgi structure,6 these findings indicate that the formation of pericentrosomal Golgi fragments characteristic of MacE depends on the presence of oxygenation adjacent to the lactone carbonyl group, either in the substituted dioxabicyclo[3.2.1]octanone or dioxabicyclo[3.3.0]octanone ring systems.</p><!><p>The synthesis and initial comparison of the chemical reactivity of two highly oxidized bicyclic lactone fragments found in rearranged spongian diterpenes—8-substituted 6-acetoxy-2,7-dioxabicyclo[3.2.1]octan-3-one and 6-substituted 7-acetoxy-2,8-dioxabicyclo[3.3.0]octan-3-one—are reported. The syntheses of t-Bu-MacE (13) and congener 14 are the first syntheses of the 8-substituted 6-acetoxy-2,7-dioxabicyclo[3.2.1]octan-3-one ring system found in rearranged spongian diterpenes such as macfarlandin E (4) and aplyviolene (3). A late-stage intermediate, 36, was diverted to provide the first synthetic entry to isomeric structures (46 and 49) in the 6-acetoxy-2,8-dioxabicyclo[3.3.0]octanone series.</p><p>Access to these highly oxidized bicyclic lactones allowed the chemical reactivity and Golgi-modifying activity of these ring systems to be studied. Under both acidic and basic conditions, 8-substituted 6-acetoxy-2,7-dioxabicyclo[3.2.1]octan-3-ones are converted to 6-substituted 7-acetoxy-2,8-dioxabicyclo[3.3.0]octan-3-one products. Both dioxabicyclooctan-3-one ring systems are found to react readily with primary amines to form pyrrole products. Of particular significance, lysine side chains of hen egg white lysozyme are converted under physiologically relevant conditions to substituted pyrroles upon exposure to dioxabicyclic lactones 13, 14, 46 and 49. The presence of an acetoxy substituent adjacent to the lactone carbonyl group, in either the bridged or fused dioxabicyclooctanone series, increases the extent of the lysine to pyrrole conversion and is essential for induction of the macfarlandin E Golgi phenotype. These investigations provide a basis for future studies aimed at identifying the biological target(s) of these Golgi-modifying natural products, as well as initial insight into the reactivity of the family of structurally distinctive rearranged spongian diterpenes depicted in Figure 1.</p>

PubMed Author Manuscript

Design Features to Accelerate the Higher-Order Assembly of DNA Origami on Membranes

Nanotechnology often exploits DNA origami nanostructures assembled into even larger superstructures up to micrometer sizes with nanometer shape precision. However, large-scale assembly of such structures is very time-consuming. Here, we investigated the efficiency of superstructure assembly on surfaces using indirect cross-linking through low-complexity connector strands binding staple strand extensions, instead of connector strands binding to scaffold loops. Using single-molecule imaging techniques, including fluorescence microscopy and atomic force microscopy, we show that low sequence complexity connector strands allow formation of DNA origami superstructures on lipid membranes, with an order-of-magnitude enhancement in the assembly speed of superstructures. A number of effects, including suppression of DNA hairpin formation, high local effective binding site concentration, and multivalency are proposed to contribute to the acceleration. Thus, the use of low-complexity sequences for DNA origami higher-order assembly offers a very simple but efficient way of improving throughput in DNA origami design.

design_features_to_accelerate_the_higher-order_assembly_of_dna_origami_on_membranes

Special Issue<!>Introduction<!><!>Introduction<!>Materials and Methods<!>Buffer Compositions<!>Origami Folding and Purification<!>Surface-Immobilization of DNA Origami<!>Supported Lipid Bilayer (SLB) Preparation and Membrane-Tethering of DNA Origami<!>Total Internal Reflection Fluorescence Microscopy<!>DNA-PAINT Microscopy<!>Single Particle Tracking<!>Imaging for Correlation Analysis<!>Fluorescence Image Analysis<!>DNA-PAINT Microscopy<!>Single-Particle Tracking<!>Correlation Analysis of Cross-Linking Kinetics<!>Atomic Force Microscopy<!>Simple DNA Origami Design for Cross-Linking Studies<!><!>Repeat Connectors Are a Viable Option for Superstructure Assembly<!><!>Repeat Connectors Form Stable Superstructures Faster than Scaffold Connectors<!><!>Repeat Connectors Form Stable Superstructures Faster than Scaffold Connectors<!>Quantification and Mechanisms of Assembly Acceleration<!><!>Quantification and Mechanisms of Assembly Acceleration<!>Conclusions<!><!>Author Contributions<!>

<p>Published as part of The Journal of Physical Chemistry virtual special issue "W. E. Moerner Festschrift".</p><!><p>Over the past 15 years, the development of DNA origami technology led to huge advances in the field of structural DNA nanotechnology, as it allows straightforward construction of large and complex nanostructures.1 This is obtained by forcing long single-stranded DNA (ssDNA) "scaffold" strands into programmed conformations using many short "staple" strands. Diverse structures are possible, and multiple site-specific functionalizations can be introduced into a single structure with few-nanometer resolution.2,3 Applications include single-molecule observation of chemical reactions,4 positioning of nanoparticles for nanophotonics,5 design of sensitive and specific biosensors,6 and many others. Recent examples of DNA origami nanostructures designed in our lab include benchmark targets for single-molecule method development,7 curved nanostructures to deform membranes,8 or nanostructures serving as passive cargo to study transport processes in reaction–diffusion systems.9</p><p>The structural complexity allowed by the DNA origami technology is essentially limited by the length of the scaffold strand, typically 7–8 kb bacteriophage genomes. Even with cutting-edge strategies to increase the scaffold length up to 10 kb and modify it for different applications,10,11 it is still challenging to produce DNA origami in sizes above 100 nm with high yield. To arrive at larger structures, the very first publication of the DNA origami technology already introduced the idea of cross-linking origami "monomer" particles into higher-order structures.1 Nowadays, quite large and complex higher-order DNA origami structures ("superstructures") are being used for nanometer-precise positioning of structures over micrometer scales,12,13 molecular "tubing" systems for linear transport of cargo,14 or the encapsulation of cargo that itself is tens of nanometers in diameter.15</p><p>There are multiple strategies for assembling DNA origami superstructures. The most common ones exploit direct DNA–DNA binding, either sticky-end hybridization16 or blunt-end stacking.17 We focus on sticky-end hybridization strategies in the present manuscript: First, as sticky-end hybridization exploits Watson–Crick base pairing, the association is specific and programmable.12 Second, sticky-end hybridization can be induced in a time-controlled manner by first preparing samples from DNA origami monomers and then cross-linking them by adding "connector strands".18 Notably, programmability and time control are in principle also possible with blunt-end stacking but are more restricted.17,19 Sticky-end hybridization is typically performed by two alternative approaches: One option is to directly prepare one origami species with staple strands that are extended with sticky ends binding to sequences in another origami, either directly in the scaffold, or in staple extensions.12,20 Alternatively, to control the timing of association, one can prepare ssDNA stretches on the origami nanostructures and later add separate connector strands to bind and cross-link those ssDNA stretches in situ.16,18 Here we will address the latter strategy (Figure 2), as DNA superstructure assembly with time-controlled onset is valuable for synthetic biology applications, such as mimicking cytoskeleton assembly in order to probe the response of in vitro reconstituted proteins to changes in their environment. Time control is also accessible through photoactivation schemes,21 but this requires additional functionalization of oligomers. We aimed for a radically simple design for time-controlled DNA origami superstructure assembly, avoiding multistep assemblies,12,14 special buffer requirements,19 or non-DNA functionalizations.21</p><p>To allow time-controlled formation of DNA origami superstructures, the effective association rates after reaction initiation should be as high as possible. Past studies of DNA origami superstructures were often quite unsatisfactory in this regard, usually requiring incubation times in the order of 1 h or more,22 up to overnight incubation.16,23 Several ways to accelerate association have been identified. One option is multivalent binding between origami monomers to facilitate nucleation.20,24 Specifically, for origami in 2D systems, increasing DNA origami monomer diffusion coefficients by adding monovalent cations and/or depositing particles on a fluid lipid bilayer rather than on a solid support accelerates assembly.18,19,23 Additional acceleration comes from precisely matched and rigid geometries of the associating staple extensions to accelerate transition from monovalent binding nucleation to multivalent full binding.20 Importantly, at least in solution, association rates for DNA origami dimerization reach values comparable to typical association rates for free DNA oligonucleotides.20 This indicates that increasing effective association rates of the hybridization reaction itself may yield an additional gain in DNA origami superstructure assembly speed. With this idea in mind, we reasoned that recent developments toward increasing hybridization on-rates in DNA point accumulation for imaging in nanoscale topography (DNA-PAINT) microscopy could be transferred to accelerate DNA origami superstructure assembly.25</p><p>DNA-PAINT (Figure 1b) super-resolution microscopy is an implementation of single-molecule localization microscopy (SMLM) in which fluorophore-conjugated "imager strand" oligonucleotides reversibly bind to "docking sites" on the structure of interest. With low concentrations of imager strands, only a sparse random subset of docking sites is labeled at each time point, allowing their imaging in the single-molecule regime. Acquisition of thousands of frames and subsequent emitter point spread function fitting allows reconstruction of a super-resolved map of docking site coordinates.26−28 Recent improvements in DNA-PAINT acquisition speed focus on improved docking site design. Specifically, docking sites with low-complexity sequences, i.e., repeats of a short sequence motif such as [CTC]N, were found to be superior: These offer a large number of overlapping imager strand binding sites and thus increase the effective association rates for imager strand binding.25 The same strategy can also be used in single-particle tracking (SPT) of sparse sets of DNA origami particles.28 In this case, a long docking strand and a high concentration of imager strands yield unusually long tracks due to continuous replacement of bleached imager strands, circumventing photobleaching limitations to track duration.29</p><!><p>Design of DNA origami nanostructure used in this study. (a) Design schematic (elements not to scale). A 24-helix bundle is functionalized with a 36 docking sites for imager strands. Only a subset of these is shown for clarity, the Picasso Design26 schematic in the corner shows the true arrangement. Additionally, the particle is functionalized for membrane binding (orange extensions binding dark-blue "anchor" sequences) and lateral extensions for linear cross-linking (light-blue). (b) DNA-PAINT super-resolution imaging. Imager strands reversibly bind to the docking sites on the particle, successively highlighting them and allowing their super-resolved position determination. (c) Experimental DNA-PAINT data from surface-immobilized DNA origami particles, with arrows shapes clearly resolved on many particles. Inset shows an average image from 32 901 particles.</p><!><p>We thus set out to characterize two different sticky-end-based DNA origami superstructure assembly approaches in a lipid membrane-anchored 2D system. We use fluorescence techniques including single-particle tracking (SPT), DNA-PAINT, and image correlation analysis, complemented by atomic force microscopy (AFM), to characterize the assembly kinetics and the resulting structures. To this end, we employ a simple, stochastically assembling DNA origami superstructure based on rectangular monomers.1,26 We functionalized this DNA origami with staple extensions for cross-linking using low-complexity sequence connector strands to assemble superstructures in situ rather than preforming them in solution. We demonstrate assembly kinetics that are 1 order of magnitude faster than more traditional approaches by using low-complexity sequence connector strands. We discuss effects contributing to the acceleration, in particular the influence of length of the used sticky end. Our results provide useful insights for future experiments that require rapid cross-linking of DNA origami superstructures.</p><!><p>Unless specified otherwise, chemicals were purchased from Sigma-Aldrich/Merck. DNA oligonucleotide sequences can be found in the Supporting Information.</p><!><p>DNA origami folding buffer: 12.5 mM MgCl2, 10 mM tris, 1 mM EDTA, pH 8.0. Buffer A: 100 mM NaCl, 10 mM tris, pH 8.0. Buffer B: 10 mM MgCl2, 5 mM tris, 1 mM EDTA, pH 8.0. Buffer D: 140 mM NaCl, 7.5 mM MgCl2, 20 mM tris, 0.75 mM EGTA, pH 7.6. SLB formation buffer: 150 mM KCl, 5 mM MgCl2, 25 mM tris, pH 7.5. SLB washing buffer: 150 mM KCl, 25 mM tris, pH 7.5. AFM imaging buffer: 40 mM MgCl2, 5 mM tris, pH 7.5.</p><!><p>DNA origami were designed using Picasso Design software,26 and modified using caDNAno.30 Scaffold DNA (p7249, tilibit nanosystems, 10 nM in folding buffer) was mixed with a 10-fold molar excess of unmodified staple strands or staple strands with extensions for tetraethyleneglycol–cholesterol (TEG-chol)-anchoring to membranes. Staple strands with DNA-PAINT docking site extensions, the adapter sequence for the "tracking handle", or A7 cross-linking extensions were added in a 100-fold molar excess. The folding reaction was performed via melting for 5 min at 80 °C and temperature ramping from 60 to 4 °C over 3 h. The folded origami were PEG-purified by two cycles of dilution (1:1 in folding buffer containing additional 15% w/v PEG-8000 (89510) and 250 mM NaCl), centrifugation (30 min, 17 900 rcf, 4 °C), and resuspension (in folding buffer, 30 min, shaking at 30 °C). DNA origami solutions were stored at −20 °C until use. Before use, DNA origami solutions were diluted with dilution factors adjusted differently for different sample types, typically on the order of 1:20 relative to the concentration obtained after PEG purification.</p><!><p>Liquid chambers were assembled from coverslips (22 × 22 mm2, no. 1.5, Marienfeld) and microscopy slides (Menzel-Gläser) using double-sided sticky tape (Scotch Transparent 665, Conrad) as a spacer. Chambers (ca. 20 μL volume) were passivated with biotinylated BSA (A8549; 1 mg/mL in buffer A, 3 min), washed with 40 μL of buffer A, and functionalized with streptavidin (S888, Thermo Fisher, 0.5 mg/mL in buffer A, 3 min). After washing with 40 μL buffer A and 40 μL buffer B, DNA origami were washed in (20 μL, in buffer B, 6 min). After incubation, unbound origami were washed out with 80 μL of buffer B. Finally, samples were washed with 40 μL of imaging solution (buffer D with imager strands and POCT oxygen scavenger) and sealed in an air-tight container with two-component epoxy glue (Toolcraft Epoxy Transparent, Conrad). The POCT oxygen scavenger consisted of 20 μg/μL catalase (P4234), 0.26 μg/μL pyranose oxidase (C40), 1 μg/μL trolox (238813), and 0.8% w/w glucose.</p><!><p>SLBs were formed via vesicle fusion. Lipids dissolved in chloroform were mixed in glass vials, and after solvent evaporation under N2 flow, the lipids were resuspended in SLB formation buffer to 4 μg/μL. The obtained large multilamellar vesicle suspensions were then sonicated (Bransonic 1510, Branson) until the solutions were clear. These small unilamellar vesicle (SUV) solutions were either used immediately or stored at −20 °C and re-sonicated before use. For fluorescence imaging of SLBs, sample chambers were assembled from cut 0.5 mL reaction tubes glued (NOA 68, Norland) onto ethanol- and water-rinsed coverslips and cured under 365 nm UV light exposure for 20 min. Immediately before use, chambers were surface-etched with oxygen plasma (30 s, 0.3 mbar, Zepto, Diener Electronics). Next, 75 μL of diluted SUV suspension (ca. 0.5 μg/μL in SLB formation buffer) were added into prewarmed (37 °C) chambers and incubated for 5 min, during which SLBs formed. After formation, SLBs were washed with 2 mL of SLB washing buffer, followed by 600 μL of buffer B. After the sample cooled to room temperature, the supernatant was replaced with 100 μL of 10 nM TEG-chol anchor oligonucleotide solution (buffer B, 3 min), followed by washing with 200 μL buffer B. Next, 100 μL of DNA origami solution was added (buffer B, 6 min), and the sample was washed with 200 μL of buffer B, followed by 200 μL of buffer D, and finally flushed twice with 200 μL of each imaging solution in buffer D with POCT. SLBs used in fluorescence experiments consisted of DOPC with 1 mol % biotinyl-cap-DOPE (both Avanti Polar Lipids) and 0.01 mol % Atto655-DOPE (ATTO-TEC). The biotin functionalization was not exploited in generating the data shown in this manuscript. SLBs for AFM imaging consisted of DOPC with 0.1 mol % Atto655-DOPE and were prepared on coverslips (22 mm diameter, no. 1, Marienfeld) in dedicated sample chambers for liquid-phase AFM (JPK). Atto655-DOPE was used to locate and quality-check membranes but not for generation of the data shown here. For preparation of SLBs for AFM, the same protocol was followed with the reagent volumes scaled up 2- to 3-fold compared to the chambers used for fluorescence imaging.</p><!><p>Fluorescence microscopy was performed at a custom inverted microscope described in detail in a previous publication.31 Light from a solid-state laser (561 nm, DPSS-System, MPB) was intensity-adjusted using a half-wave plate and a polarizing beam splitter (WPH05M-561 and PBS101, THORLABS). The beam passed through a refractive beam-shaping device (piShaper 6_6_VIS, AdlOptica) to create a flat illumination profile. To achieve evanescent-field illumination, the beam excentrically entered the oil immersion objective lens (100× NA 1.49 UAPON, Olympus). Fluorescence emission was collected by the same objective and filtered through suitable band-pass filters (605/64, AHF Analsentechnik) before detection on a CMOS camera (Zyla 4.2, Andor). During acquisitions, the temperature was stabilized at 23 °C (H101-CRYO-BL, Okolab), and z-positioning of the sample was stabilized via a piezo stage (Z-INSERT100, Piezoconcept and CRISP, ASI). The camera was operated with the open source acquisition software μManager32 and images were acquired with 2 × 2 pixel2 binning and field of view cropping to the central 700 × 700 (prebinned) pixels to achieve an effective pixel width of 130 nm and a field of view matching the circular flat illumination profile ca. 130 μm in diameter.</p><!><p>DNA origami nanostructures were functionalized with 5xR1 docking sites.25 The imaging solution contained 1.25 nM R16nt-Cy3B imager strands. Illumination intensity was set to ca. 30 μW μm–2. A total of 10 000 images per data set were acquired at a frame rate of 20 Hz.</p><!><p>DNA origami with a 20 nucleotide (nt) adapter sequence were deposited on membranes. A [TCT]38 "tracking handle" docking site analogous to that described by Stehr et al.29 was quasi-irreversibly recruited to the origami via the adapter complement: During DNA origami deposition, 10 nM tracking-handle–adapter conjugate were additionally present. To ensure a sparse subset of labeled DNA origami nanostructures suitable for SPT, a low density of tracking-handle-coupled particles was diluted in a 20-fold excess of unlabeled DNA origami particles, i.e., the same DNA origami, except without the adapter sequence. The imaging solution contained 10 nM R5_S28nt-Cy3B imager strands. Illumination intensity was set to ca. 20 μW μm–2. A total of 10 000 images per data set were acquired at a frame rate of 20 Hz.</p><!><p>DNA origami nanostructures were functionalized with 5xR1 docking sites,25 which were quasi-irreversibly labeled through 4 min incubation with 10 nM R118nt-Cy3B. The imaging solution did not contain imager strands. Connector strands were added at 250 nM immediately before start of acquisition (ca. 10 s delay, limited by speed of pipetting and closing of microscope stage incubation chamber). A total of 300 images were acquired at a frame rate of 30 Hz at each time point along the cross-linking observation. The laser was shuttered between observation time points. Illumination intensity was set to ca. 2 μW μm–2.</p><!><p>Processing parameters for all fluorescence experiments are listed in Table S1.</p><!><p>Image stacks were processed using Picasso software.26 Picasso Addon7 was used for automation. The Python software can be found on Github (https://github.com/schwille-paint). The general pipeline started with Picasso Localize to pick and localize emitters, followed by Picasso Render for drift correction (RCC). In the case of biotin/streptavidin-immobilized origami, particles were manually picked in Picasso Render, followed by automated picking of similar particles and drift correction from picked particles. The average image from many immobilized DNA origami nanostructures was created using Picasso's Average3 module.</p><!><p>The analysis pipeline started with localization in Picasso Localize as in the case of SMLM. Subsequent steps used the "SPT" package, which is also available via the above-mentioned GitHub page, for linking of localizations into tracks and mean-squared displacement analysis.</p><!><p>Image stacks were analyzed using a custom Python script, which is included in the Supporting Information. A detailed explanation of the analysis can be found in the Supporting Information, including a description of the simulations performed to test the accuracy of the analysis.</p><!><p>Measurements were performed on a JPK Nanowizard 3. The AFM images were taken in QI (quantitative imaging) mode using BioLever Mini BL-AC40TS-C2 cantilevers (Olympus). The set point force was 0.25–0.35 nN, acquisition speed 66.2 μm s–1, Z-range 106 nm; 10 × 10 μm2 fields of view were acquired with a 15 nm pixel size. Images were first processed in JPKSPM Data Processing (JPK, v6.1.142) performing a line-wise second-degree polynomial leveling followed by another second-degree polynomial leveling with limited data range (0% lower limit, 70% upper limit). Subsequent plane leveling, third-degree polynomial row alignment and scar correction were performed in Gwyddion (v2.58, http://gwyddion.net/).</p><!><p>To study DNA origami cross-linking, we first designed a suitable monomer structure. We reasoned that the use of a well-characterized modular structure would be most convenient and thus opted for a flat rectangular grid origami used in a number of previous single-molecule fluorescence studies.7,25,29,33,34 On this monomer structure, we arranged 36 DNA-PAINT docking sites in the shape of an arrow. This design challenges the resolution in DNA-PAINT imaging and allows reading out the orientation of the origami on the surface (Figure 1). DNA-PAINT imaging of individual DNA origami particles immobilized on a glass surface via biotin–streptavidin anchoring indeed revealed the expected arrow pattern with high yield (Figure 1c).</p><p>We then functionalized the "bottom" side of the origami structure with staple extensions to bind it to supported lipid bilayer membranes (SLBs) via complementary TEG-chol-coupled oligonucleotides. Only two opposing lateral edges of the DNA origami were further functionalized for cross-linking into higher-order assemblies, aiming for linear chains rather than tilings, as the latter might be more difficult to distinuish from unspecific clustering (Figure 2a). In all cross-linking experiments described in this manuscript, each DNA origami edge participating in the association was designed to bind four connector strands. The DNA origami design exposes no blunt ends of DNA duplexes to avoid uncontrolled association via base stacking. Figure 2 gives a schematic summary of the DNA origami cross-linking strategies. One strategy that we employed has been frequently reported before.16,18,23 Here, DNA origami nanostructures are cross-linked via connector strands that are essentially staple strands which incorporate into both monomers simultaneously (Figure 2b). For concision, we will call these "scaffold connectors". The other strategy is to incorporate modified staples into the DNA origami that carry extensions for indirect binding of connector strands to the DNA origami. We reasoned that DNA origami superstructure assembly could be accelerated through a connector strand design analogous to the above-mentioned high-on-rate docking site design25,29,33 used for example in DNA-PAINT, i.e., the use of low-complexity sequences to increase the effective association rate (Figure 2c). We opted for short stretches of a single nucleotide species, specifically A7 as an extreme case of such a low-complexity sequence. The connector strands were simply oligo-T sequences. These connector strands will be referred to as "repeat connectors". We note that we did not optimize our structure for highly specific assembly geometries. Instead, we aimed for a simple system that would serve as a model system for characterizing the assembly process itself. Thus, a stochastically assembling design was chosen in which also the shape of the formed structures would reveal the action of the connector strands in super-resolution imaging. With the basic origami design and cross-linking strategies at hand, we proceeded to create higher-order DNA origami assemblies on fluid membranes.</p><!><p>Schematic of DNA origami cross-linking kinetics on membranes. (a) Cross-linking geometry. Cross-linking sites are distributed on the DNA origami such that linear assemblies are expected, but with repeat connectors branching is also possible. (b) Scaffold connectors directly bind scaffold loops of two DNA origami particle, yielding highly site-specific assembly. (c) Repeat connectors bind the DNA origami indirectly via A7 staple extensions. Depending on the design of the connector strand, many binding reading frames are available for the A7.</p><!><p>We first characterized the structures of our cross-linked DNA origami structures using AFM to confirm the possibility of forming superstructures with desired geometry using repeat connectors. For AFM imaging, we prepared DNA origami samples on fluid SLBs and cross-linked them for 2 h using all-T repeat connectors of different lengths (T14, T20, T40, T60, T80, or a mixture of all of these referred to as TN mix). Before imaging, we exchanged the buffer, increasing the Mg2+ concentration from 7.5 to 40 mM to decrease mobility of the preformed structures for better AFM image quality. When using repeat connectors, ≥40 nt in length, high-quality images showing the expected formation of extended filaments were obtained which agree with the linear assembly geometry dictated by design (compare Figures 3 and 2). However, we saw hardly any differences between different lengths ≥ 40 nt. Small oligomers formed by shorter repeat connector strands yielded lower quality images, suggesting that these led to hardly any superstructure formation within 2 h. In fact, the structures that we obtained with repeat connectors rather looked like unspecific association due to the high Mg2+ concentration (Figure S3). We did see some lateral assembly as well: As all cross-linking staple extensions have the same A7 sequence and only differ by orientation of 3′- or 5′-ends, there is no strict specificity regarding the orientation of neighboring DNA origami monomers within the superstructure. This allows branching of linear assemblies, which leads to the formation of the observed 2-dimensional superstructures. We observed this branching somewhat less frequently when using scaffold connectors, which are site-specific in their binding to DNA origami and thus suppress branching (Figure S3). The presence of some branching even in this setting suggests Mg2+ unspecific association. Overall, the AFM data suggests that using long repeat connectors allows to cross-link DNA origami superstructures efficiently, albeit with trade-offs in specificity. However, there was no obvious difference between the different repeat connectors that efficiently cross-linked the DNA origami structures. In our AFM experiments, pushing of DNA origami structures by the AFM tip forced us to strongly increase the Mg2+ concentration, which led to unspecific association. Thus, at least with lengths ≥40 nt, repeat connectors do facilitate formation of DNA origami superstructures. To characterize the structures in more detail under origami-typical buffer conditions, we employed single-molecule fluorescence imaging.</p><!><p>AFM characterization of DNA origami superstructures, showing conditions which yielded high-quality images. Additional conditions are shown in Figure S3. All images were acquired after 2 h incubation with 250 nM of the specified connector strand. The TN mix is 50 nM each T14, T20, T40, T60, and T80. The color-coded height scale in all panels is 6 nm.</p><!><p>Before acquiring super-resolution images of our samples, we used SPT to characterize particle mobility prior to cross-linking in the imaging buffer used for all following fluorescence microscopy experiments, containing 7.5 mM Mg2+ and 140 mM Na+. SPT showed that our TEG-chol-anchored DNA origami particles diffused freely on the SLBs with a diffusion coefficient of ca. 0.2 μm2 s–1 (Figure S4). However, upon addition of connector strands, we observed a strong decrease in mobility, indicating superstructure formation. A large fraction of particles was practically immobilized 30 min after addition of a mixture of oligo-T connector strands to A7-functionalized origami (Figure S5). We reasoned that these may in fact be sufficiently immobilized for DNA-PAINT-based structural characterization using an accelerated acquisition protocol following Strauss and Jungmann,25 which reduces the acquisition time to ca. 8 min. SMLM has been successfully applied to samples with slow but non-negligible motion such as live cells before, albeit with trade-offs between acquisition time and resolution.35,36</p><p>Even with that accelerated acquisition, we were unable to resolve any structures in DNA-PAINT imaging without cross-linking (Figure S6a). However, we were able to resolve large DNA origami superstructures on the membrane after cross-linking for only 30 min with the TN repeat connector mixture (Figure 4a). Notably, in all our AFM and DNA-PAINT experiments, the connector strand solution had been replaced with connector strand-free imaging buffer before acquisition. This means that the observed assemblies were rather stable and did not undergo rapid dissociation/reassociation dynamics and, in particular, that the assemblies were not dependent on stabilization by the high Mg2+ concentration in the AFM imaging buffer. This confirms that the use of short A7 sticker sequences combined with multivalent cooperative binding is sufficient for association of stable superstructures. In fact, the branching of oligomers seen in AFM and confirmed by SMLM suggests that our A7 cross-linking extensions are too long for efficient "self-healing" of association sites into "ideal" association geometries.12,37 We saw similar results when using scaffold connectors, but much longer incubation times were needed before high-quality imaging was possible: Compare Figure 4b acquired after 20 h to Figure S7 acquired after 2 h. This is in line with previous publications using scaffold connectors to cross-link DNA origami into 2D systems.18,23 Each scaffold connector first needs to bind to its unique binding site on a DNA origami nanoparticle and then to the appropriate binding site on a second particle, requiring the DNA origami monomers to collide in the correct mutual orientation. Even after 20 h, only rather small assemblies were found. Thus, repeat connectors allowed assembly within less than 1 h, while scaffold connectors seemed quite unsatisfying regarding throughput of the experiment.</p><!><p>DNA-PAINT characterization of DNA origami superstructures cross-linked with different connector strands. (a) TN repeat connector mix containing T14, T20, T40, T60, and T80 at 50 nM each (30 min incubation). (b) Scaffold connectors (250 nM total concentration, 20 h). (c) Individual repeat connectors (250 nM, 30 min).</p><!><p>Although the image resolution in DNA-PAINT on membranes was lower than that in the image of origami directly immobilized on glass, we achieved resolution down to the 10 nm scale even on membranes. The resolution was limited by residual motion on the time scale of the acquisition, as demonstrated by the blurred clouds of localizations in various positions of the image. The orientation of some DNA origami monomers within the context of the superstructures was visible in the SMLM images, giving access to some information about the geometry in association. When repeat connectors are used, both parallel and antiparallel arrow orientations in neighboring particles are seen, which is obviously another consequence of the lack of site specificity in repeat connector binding. This is in stark contrast to the images obtained using scaffold connectors, which yield assemblies specifically with parallel orientation (Figure 4b). Notably, DNA-PAINT imaging of DNA origami deposited in a 3-fold higher density, but not exposed to connector strands, yielded low-resolution images of very different structures (Figure S6b). This confirms that despite the compromises in association geometry specificity when using repeat connectors, the retrieved superstructures are products of hybridization-based, connector strand-dependent association.</p><p>Finally, we compared superstructures formed by different lengths of all-T connector strands using DNA-PAINT imaging (Figure 4c). T14 (not shown) or T20 repeat connectors showed almost no cross-linking within 30 min, supporting the idea that assembly seen with AFM was mostly unspecific due to the high Mg2+ concentration. As in AFM, we saw little difference between the different all-T connectors of lengths ≥ 40 nt. From our DNA-PAINT experiments, we could thus confirm the connector strand-driven association of our DNA origami superstructures, and that long repeat connectors yield faster assembly than scaffold connectors. Motivated by these findings, we decided to characterize more quantitatively the differences between assembly kinetics of scaffold and repeat connectors, in order to obtain a mechanistic understanding of these differences.</p><!><p>In the next experiments, we set out to determine characteristic time scales for DNA origami higher-order assembly under different conditions. We opted for an image correlation analysis-based read-out of oligomerization (see Supplementary Note and Figure S1). The calculated correlation parameter, reporting the amplitude of temporal fluorescence fluctuations, increases as the particles associate into higher-order assemblies: Fluorescence fluctuations are larger when few bright particles diffuse through a pixel than many dim ones do. Later, the correlation parameter falls to zero or a low baseline value, as the assemblies become so large that they are essentially immobile during the 10 s observation: Immobile particles yield an approximately constant signal over time (Figure 5a). The correlation analysis was found to be sensitive to oligomerization and immobilization in simulations of different ratios of mono- and oligomers (Figure S2). Additional advantages for long-term observation of the overall evolution of the sample are lower illumination intensities and the fact that in contrast to SMLM and AFM, this analysis captures the entire ensemble of particles rather than selectively showing immobile assemblies. Thus, image correlation analysis provided a convenient aggregate readout for higher-order assembly kinetics, from which we derived characteristic time scales of immobilization as a surrogate for assembly of DNA origami superstructures (Figure 5b). For these experiments, the spatial arrangement of docking sites previously used for DNA-PAINT plays no role (ca. 50 nm pattern width vs ca. 200 nm spatial resolution). Instead, we created bright particles through quasi-irreversible binding of multiple long R118nt-Cy3B imager strands to the full length of the docking site.38 We systematically compared cross-linking by a variety of connector strands under otherwise constant conditions. These included the previously used scaffold connectors with and without short flexible linkers between the binding sites and all-T repeat connectors of lengths 14, 20, 40, 60, and 80 nt. In addition, we included mixtures of repeat connectors of all lengths, but with inserted oligo-C spacers that do not bind the oligo-A extensions, thus tuning the "sticker" length (i.e., number of binding reading frames) without changing the overall length of the connector strands (Figure 2c). The results are compiled in Figure 5b for comparison, but they will now be discussed sequentially.</p><!><p>Correlation analysis of cross-linking kinetics. (a) Illustration and example data of correlation analysis. At the beginning of the experiments, monomers diffuse rapidly, creating moderate fluorescence fluctuations (red image and fluorescence intensity trace). As oligomerization begins, effectively fewer brighter particles are observed, increasing fluctuation amplitudes at unchanged average intensity (blue). As oligomerization progresses, yielding large, immobile particles, fluctuations become negligible (brown). The time traces of correlation parameter change show two examples of traces quite clearly undergoing these phases within observation time, and a buffer-treated negative control. (b) Kinetics of DNA origami higher-order assembly measured through image correlation analysis (mean ± s.d.). See the main text for details about the different conditions. Numbers in parentheses refer to the number of data sets for which an assembly time scale could be fitted compared to the number of data sets acquired for this condition. One of the mock-treated samples did show clear immobilization, which we attribute to unspecific sample degradation.</p><!><p>Assembly kinetics were observed following addition of connector strands for either 24 h (scaffold connectors and negative controls) or 2 h (repeat connectors). Confirming the findings from DNA-PAINT imaging, very long incubation times in the order of 10 h were needed to create fully assembled structures using scaffold connectors. Adding a short flexible linker sequence to the scaffold connectors did not strongly affect the association kinetics. If anything, it slowed down association, which may be explained by the findings of Zenk et al.20 that larger flexibility of connector binding sites can be detrimental to association.</p><p>We then characterized the repeat connectors with total lengths of 14, 20, 40, 60, and 80 nt. First, we looked at cross-linking kinetics for mixtures of repeat connectors with internal oligo-C stretches and terminal oligo-T stickers. Oligo-T sticker lengths varied from 6 nt (shorter than the A7 docking site) to 9 nt (three binding reading frames). Using repeat connector mixtures for cross-linking, we saw a strong acceleration in association kinetics for sticker lengths of ≥7 nt. Within the 2 h acquisition time, we did not see any notable changes in the fluctuation data for 6 nt stickers, and for 7 nt stickers, only one out of three samples showed immobilization. Increasing the oligo-T sticker length at the end of the repeat connectors to 8 or 9 nt yielded robust assembly within <2 h, demonstrating the desired acceleration. These sticker lengths offer 2 or 3 reading frames for the A7 binding partner, respectively, meaning that the data is entirely consistent with our idea of multiple reading frames accelerating binding. Another cause for acceleration is the same effect that is the cause for the reduced orientation specificity observed by nanoscale imaging: Repeat connectors can bind various positions on DNA origami nanoparticles, reproducing the effect of multivalent binding previously reported.20,24 Time-resolved analysis of cross-linking kinetics thus confirms an order-of-magnitude acceleration in assembly dynamics by using our repeat connector strategy, as compared to our scaffold connector strategy.</p><p>Interestingly, no further acceleration of superstructure assembly was seen by using a mixture of all-T connector strands of different lenths ("TN mix" in Figure 5b) compared to those with 8 or 9 nt stickers. We hoped to find an explanation for this effect by comparing different lengths of all-T connector strands. The longer all-T repeat connectors accelerated assembly compared to shorter ones. While Zenk et al.20 argued that increasing connector strand flexibility (i.e., length) can be detrimental to binding, here the increased length comes with an increase in the number of binding sites. We did not see immobilization within 2 h using T14 or T20 connectors. This suggests an explanation for the fact that using a mix of different all-T connectors did not further accelerate assembly relative to connectors with 9 nt stickers: The inefficient T14,20 connectors likely competed with the more efficient T40,60,80 connectors. The low efficiency of T14,20 connectors may be explained by the fact that their short sequences cause A7 docking sites to compete for overlapping binding sites on the same connector strand, which is clearly detrimental for cross-linking. This competition is suppressed in repeat connectors with internal oligo-C stretches and less relevant in long all-T ones.</p><p>Obviously, by comparing scaffold connectors to repeat connectors only consisting of oligo-T stretches, we looked at two extremes in a broad spectrum of thinkable cross-linker designs: one entirely optimized for assembly speed and the other entirely for specificity. Intermediate strategies would allow different trade-offs between these parameters. For example, one could combine oligo-A staple strand extensions with oligo-G staple extensions, creating two orthogonal cross-linking systems. These could also be combined through connectors concatenating oligo-T stretches and oligo-C stretches to link an oligo-A functionalized DNA origami face to an oligo-G functionalized one. This would increase specificity in assembly geometry, unlikely to result in antiparallel association of our DNA origami monomers. Repeats of 2 or 3 nt sequence motifs further increase the number of orthogonal motifs available for cross-linking,25 but the number of binding reading frames will decrease rapidly with increasing motif length. Notably, such 2 nt motifs, albeit without repeats, were used previously to create very large DNA origami superstructures12 with high specificity in assembly geometry. However, this specific formation of large structures required a multistep assembly that is slow and is not easily transferred to the in situ assembly in which we were interested.</p><p>Finally, an additional mechanism that likely contributes to the acceleration of binding using low-complexity sequences is the absence of internal hairpins from oligo-T or A7 sequences. Hairpin formation can strongly reduce effective on-rates.33,34 Due to sequence constraints from direct binding to the scaffold strand, hairpin formation could not be abolished completely in the design of the scaffold connectors used in this study according to the prediction by NUPACK.39 One might thus consider high-complexity, yet hairpin-free, docking site extensions. While sequence design will become very challenging with increasing numbers of desired orthogonal sequences and the speed gain will likely remain modest compared to what our work demonstrates, such an approach remains highly attractive regarding specificity. In any case, our recommendation for designing rapidly cross-linking sequences for DNA origami superstructures is to avoid direct binding of connector strands to the scaffold and instead use staple extensions, designed with the lowest possible sequence complexity sufficient to ensure the required specificity.</p><!><p>In this work, we compared different design features to optimize assembly kinetics of higher-order DNA origami structures. A significant acceleration was achieved by cross-linking DNA origami indirectly via low sequence complexity connector strands binding to staple strand extensions, instead of direct binding of high-complexity sequences to loops in the scaffold DNA. We postulate two effects to contribute to the increased speed: The presence of multiple binding reading frames increases the effective local concentration of binding sites, and thus the effective association rate, and the used low-complexity sequences prevent the formation of hairpins. Using modifications of the strategy will allow multiple orthogonal sequences, increasing association specificity, with some trade-off in experimental throughput. This quite simple and generic approach to accelerate DNA origami superstructure assembly should prove useful to increase throughput of experiments in the field and to benefit experiments that require time-controlled assembly.</p><!><p>Methods, image correlation analysis workflow, simulations to assess sensitivity of image correlation analysis to oligomerization and immobilization, AFM results, example of single particle tracking results for monomeric DNA origami particles on DOPC SLBs, DNA-PAINT imaging of non-cross-linked and incompletely cross-linked DNA origami nanostructures, image analysis parameters, oligonucleotide sequences for DNA origami design, additional references (PDF)</p><p>MATLAB and Python scripts used for simulation and analysis of cross-linking kinetics (ZIP)</p><p>jp1c07694_si_002.pdf</p><p>jp1c07694_si_003.zip</p><!><p>§ Y.Q. and J.-H.K. contributed equally to this work.</p><!><p>Open access funded by Max Planck Society.</p><p>The authors declare no competing financial interest.</p>

PubMed Open Access

Activation of Transient Receptor Potential Ankyrin-1 (TRPA1) in Lung Cells by Wood Smoke Particulate Material

Cigarette smoke, diesel exhaust, and other combustion-derived particles activate the calcium channel transient receptor potential ankyrin-1 (TRPA1), causing irritation and inflammation in the respiratory tract. It was hypothesized that wood smoke particulate and select chemical constituents thereof would also activate TRPA1 in lung cells, potentially explaining the adverse effects of wood and other forms of biomass smoke on the respiratory system. TRPA1 activation was assessed using calcium imaging assays in TRPA1-overexpressing HEK-293 cells, mouse primary trigeminal neurons, and human adenocarcinoma (A549) lung cells. Particles from pine and mesquite smoke were less potent agonists of TRPA1 than an equivalent mass concentration of an ethanol extract of diesel exhaust particles; pine particles were comparable in potency to cigarette smoke condensate, and mesquite particles were the least potent. The fine particulate (PM<2.5 \xce\xbcm) of wood smoke were the most potent TRPA1 agonists and several chemical constituents of wood smoke particulate: 3,5-ditert-butylphenol, coniferaldehyde, formaldehyde, perinaphthenone, agathic acid, and isocupressic acid were TRPA1 agonists. Pine particulate activated TRPA1 in mouse trigeminal neurons and A549 cells in a concentration-dependent manner, which was inhibited by the TRPA1 antagonist HC-030031. TRPA1 activation by wood smoke particles occurred through the electrophile/oxidant-sensing domain (i.e., C621/C641/C665/K710), based on the inhibition of cellular responses when the particles were pre-treated with glutathione; a role for the menthol-binding site of TRPA1 (S873/T874) was demonstrated for 3,5-ditert-butylphenol. This study demonstrated that TRPA1 is a molecular sensor for wood smoke particulate and several chemical constituents thereof, in sensory neurons and A549 cells, suggesting that TRPA1 may mediate some of the adverse effects of wood smoke in humans.

activation_of_transient_receptor_potential_ankyrin-1_(trpa1)_in_lung_cells_by_wood_smoke_particulate

Introduction<!>Chemicals<!>Preparation of WSPM<!>Preparation of Other cdPM<!>Preparation of Particle Treatment Solutions<!>Cloning, Expression, and Site-Directed Mutagenesis<!>Cell Culture<!>Calcium Imaging Assays<!>Mouse Trigeminal Neuron Calcium Imaging Assays<!>Analysis of WSPM Aldehydes and Ketones<!>qPCR Analysis of TRPA1 Expression in A549 cells<!>Statistical Analysis<!>Production and collection of WSPM<!>Comparison of particle induced calcium flux in TRPA1-overexpressing HEK-293 cells<!>Specificity of TRPA1 activation by pine PM in TG neurons and A549 cells<!>Chemicals involved in TRPA1 activation by WSPM<!>Mechanisms of TRPA1 activation by WSPM and potential chemical constituents<!>Discussion

<p>Wood smoke particulate material (WSPM) is a common and often unavoidable combustion-derived indoor and outdoor air pollutant.(1,2) Where primitive wood burning stoves are the mainstay for cooking and heating, long-term exposure to high concentrations of WSPM and other forms of biomass smoke PM (BSPM) has been linked to progressively deteriorating respiratory function, exacerbation of pre-existing respiratory conditions such as asthma and heart disease, increased rates of respiratory infections, the development of chronic obstructive pulmonary disease and emphysema, and premature death.(1–5) Adverse effects on the respiratory system also have been reported by studies of short-term and seasonal elevations in WSPM/BSPM due to forest fires, crop burning, and home heating.(1,2,5)</p><p>While it is clear that wood and biomass smoke particles impact human health, the molecular and cellular processes underlying many of the commonly reported adverse effects are not completely understood. Similar to other combustion-derived particulate materials (cdPM), WSPM is comprised of carbon soot coated with chemicals including polycyclic aromatic hydrocarbons (PAHs), aldehydes and ketones, and other redox active oxygenated hydrocarbons.(6,7) WSPM has been shown to increase the production of pro-inflammatory mediators (e.g., IL-8, TNF-α, others) in lung cells via oxidative stress.(6,8,9) Additionally, WSPM has been shown to promote lipid peroxidation and oxidative damage to DNA in vitro and in vivo.(7,10) However, not all of the biological effects of WSPM can be directly attributed to oxidative stress.</p><p>An alternate mechanism by which WSPM may exert pulmonary toxicity could involve transient receptor potential-ankyrin 1 (TRPA1). TRPA1 is a cation channel that is a molecular sensor for electrophiles and oxidants in the respiratory tract, including H2O2 and hypochlorite,(11,12) aldehydes in cigarette smoke and cigarette smoke condensate (CSC),(13,14) and electrophiles on diesel exhaust particles (DEP).(15,16) TRPA1 is abundantly expressed on C-fibers originating in the trigeminal and vagal ganglia, which innervate the upper airways (i.e., the nose and mouth), conducting airways (i.e., the trachea, bronchi, and terminal bronchioles), and the respiratory bronchioles, alveolar ducts and alveoli. C-fibers frequently "sense" the presence of potentially toxic inhaled irritants and toxicants.(11,12,17) In airway C-fibers, TRPA1 is co-expressed with transient receptor potential vanilloid-1 (TRPV1), calcitonin-gene related peptide, neurokinin A, and substance P. When these neurons are stimulated by agonists of TRPV1 or TRPA1, they decrease respiratory drive, trigger cough and bronchoconstriction, and neurogenic inflammation. TRPA1 is also expressed by non-neuronal cells including human lung adenocarcinoma (A549) and small airway epithelial cells (SAEC), smooth muscle cells, and fibroblasts. When treated with TRPA1 agonists such as CSC, acrolein, allyl-isothiocyanate (AITC), 4-hydroxynonenal (4-HNE), and crotonaldehyde, these cells produce interlekin-8 (CXCL1/KC) and other pro-inflammatory mediators that promote non-neurogenic inflammation.(18,19) In addition to its role as a molecular sensor for pulmonary irritants, TRPA1 has also been shown to play a central role in the development of ovalbumin-induced airway hypersensitivity in mice.(20) These characteristics make TRPA1 an attractive target to study as a potential mediator of WSPM-related health effects.</p><p>The hypothesis of this study was that TRPA1 would be preferentially activated by WSPM in lung cell models, based on similarities in the chemical composition of WSPM to other cdPM (i.e., CSC and DEP) that have been shown to activate TRPA1. Using calcium imaging (Fluo-4 and Fura-2) in TRPA1-overexpressing HEK-293 cells, wild-type and mutant forms of TRPA1, the TRPA1 antagonist HC-030031, and different samples of WSPM and select constituents thereof, it has been demonstrated that TRPA1 is a molecular sensor of WSPM in cells that represent key cellular targets and mediators of pulmonary toxicity for multiple types of particulate materials.</p><!><p>All chemicals were purchased from Sigma-Aldrich (St. Louis, MO), unless otherwise specified. Agathic acid, dihydroagathic acid, isocupressic acid, abietic acid, dehydroabietic acid, and tetrahydroagathic acid were provided by the USDA Poisonous Plants Research Laboratory, Logan, UT.</p><!><p>Size-fractionated WSPM was produced by burning ~10 g fresh Austrian pine or dry mesquite wood in a laboratory furnace and collecting the particulate material (PM) using an Anderson cascade impactor (Supplementary Figure 1). Briefly, an electric furnace (Blue M, New Columbia, PA) was fitted with a 90 × 2.7 cm i.d. steel tube open at one end to allow insertion of wood and intake of combustion air. The tube was heated to 750°C to induce flaming combustion within seconds of inserting split wood, ~3 to 10 mm thick. Downstream of the combustion tube, room air (4:1) was added to dilute and cool the smoke and allow for the condensation of semi-volatile components into/onto PM, mimicking the process that naturally occurs in smoke plumes. PM was collected using a 10-stage Andersen cascade impactor (ThermoAndersen, Smyrna GA), providing size-fractionated PM between 0.49 to 10 μm in aerodynamic diameter. Total flow through the impactor was monitored using flow-meters and manually regulated at 1L/min. The WSPM was a mixture of tar and solid carbonaceous particles (shown in Figure 1), which were recovered by washing the impactor stages with a minimal volume of 100% ethanol. The suspension was then dried under a stream of filtered air and stored at −20°C in the dark.</p><!><p>DEP was collected from an on-road "black smoker" 2004 Ford F350 truck and was extracted with ethanol (DEP-EtOH) to produce an oily, tar-like material enriched in TRPA1 agonists and similar in consistency to the WSPM and CSC used herein, as previously described by our group.(15) CSC was isolated from equilibrated 3R4F reference cigarettes (University of Kentucky Reference Cigarette Program, Lexington, KY) using a single-port smoking machine operated essentially as described by the Massachusetts Standard smoking Regimen, but without blocking the vent holes (Supplementary Figure 2).</p><!><p>Each particle was re-suspended in DMSO to a concentration of 115 mg/mL and subsequently diluted to the final working concentrations in LHC-9 media (Invitrogen). For all cell treatments, the final concentration of DMSO in the treatment solutions was <1% (v/v). For screening TRP channel activation by WSPM, a concentration of up to 2.3 mg/mL was used. For kinetic studies, a concentration of 1.15 mg/mL was used, which allowed for differences in TRPA1 activation (i.e., potency) and kinetics of the calcium flux response to be differentiated between WSPM, CSC, and DEP-EtOH. For mechanistic studies, concentrations of 0.09 mg/mL and 0.19 mg/mL pine and mesquite PM were used, respectively. These concentrations produced robust responses without saturation of the response or overwhelming the capacity of glutathione (GSH) to inhibit the response.</p><!><p>Human TRPA1 and the TRPA1-3CK mutant were cloned and over-expressed in human embryonic kidney (HEK-293) cells, as previously described.(15) Construction of the TRPA1-ST mutant, a loss of function mutant for menthol,(21) was performed using the QuickChange XL site-directed mutagenesis kit (Stratagene, La Jolla, CA). The primers were: TRPA1-S873V/T874L (+) 5′-GTTGGAGGTAATTTTGAAAACTTTGTTGAGGGTTTTAGTTGTATTTATCTTCCTTCTT CTGGCTTTT-3′; and (−) 5′-AAAAGCCAGAAGAAGGAA GATAAATACAACTAAAACCCTCAACAAAGTTTTCAAAATTACCTCCAA C-3′.</p><!><p>Cells were maintained in a humidified cell culture incubator at 37°C with a 95% air:5% CO2 atmosphere. HEK-293 cells (ATCC; Rockville, MD) and human TRPA1 over-expressing HEK-293 cells(15) were cultured in DMEM:F12 media containing 5% fetal bovine serum and 1x penicillin/streptomycin; for TRPA1-overexpressing cells, Geneticin (300 μg/mL) was also included. Cells were sub-cultured using trypsin. Transient transfection of HEK-293 cells with TRPA1 mutant plasmids (i.e., TRPA1-3CK and TRPA1-ST) using Lipofectamine 2000 (Invitrogen) was also performed as previously described.(15,22) Human adenocarcinoma (A549) cells (ATCC; Rockville, MD), were cultured in DMEM containing 5% FBS and 1x penicillin/streptomycin, and were sub-cultured using trypsin.</p><!><p>Cells were plated in 96-well plates (coated with 1% gelatin for HEK-293 cells) and grown to 80–90% confluence before loading with Fluo 4-AM, a fluorescent calcium indicator, using the Fluo-4 Direct assay kit (Invitrogen). The Fluo 4-AM loading solution was diluted 1:1 in LHC-9 (HEK-293 cells) or calcium buffer (1X HBSS, 20 mM HEPES, pH 7.3; A549 cells) and applied to cells for 60 min at 37°C (HEK-293 cells) or room temperature (A549 cells) in the dark. Thirty minutes prior to analysis, the loading solution was replaced with LHC-9 (HEK-293 cells) or calcium buffer (A549 cells), containing 1 mM probenecid and 0.75 mM trypan red (ATT Bioquest). Treatment-induced changes in cellular fluorescence were quantified from fluorescence micrographs or using a NOVOStar fluorescence plate reader (BMG Labtech; Offenberg, Germany), as previously described.(15,22) All agonist/particle treatment solutions were prepared in LHC-9 or calcium buffer at 3X concentration, and added to cells at room temperature. For A549 cells, the TRPA1 antagonist HC-030031 was added the wash buffer 30 min before assaying using the plate reader. All data were corrected for non-specific responses, if any, observed with HEK-293 cells, and then normalized to the maximum attainable change in fluorescence elicited by ionomycin (10 μM). Additional normalization to a maximum stimulatory concentration of AITC (150 μM), a positive control for TRPA1, was performed in selected experiments, as noted in the figure legends.</p><!><p>Experimental procedures were approved by the University of Utah Animal Care and Use Committee. Mouse TG ganglia were isolated from 3-week old C57Bl/6 mice anesthetized with isoflurane and euthanized by cervical dislocation. The TG neurons were cultured as described for mouse dorsal root ganglia (DRG) neurons.(23) Isolated TG were prepared for imaging by loading with Fura 2-AM and assayed as previously described.(15) Cells were imaged and data was obtained using the Meta Imaging Series Metafluor program (Universal Imaging). Agonists and antagonists were sequentially assayed by exchanging the treatment solutions using two pipettes, one to remove and one to add the solutions, with a buffer wash between treatments. Pine PM was prepared in LHC-9 containing 0.2% v/v DMSO and 0.2% v/v ethanol. Following treatment with pine PM (0.023, 0.073, or 0.23 mg/mL), cells were washed, followed by a treatment with 50 μM AITC to identify TRPA1-expressing neurons, washed again, and then treated with KCl (50 mM) to identify all viable neurons. For studies using the TRPA1 antagonist HC-030031, cells were pre-treated with 50 μM HC-030031 for 30 s prior to co-treatment with particles (0.073 mg/mL), AITC, and KCl. Only viable cells responding to KCl were considered in the data analysis and the data are represented as the percentage of cells responding to the agonist relative to KCl.</p><!><p>Formaldehyde, 5-hydroxymethylfurfural, acrolein, acetone, furfural, 4-hydroxybenzaldehyde, vanillin, coniferaldehyde, 2-butanone, 1,2-naphthoquinone, perinaphthenone, benzaldehyde, o-anisaldehyde, and glyoxal (i.e., chemical constituents of cdPM including WSPM) were assayed as (2,4-dinitrophenyl)hydrazone derivatives (i.e., hydrazones) using liquid chromatography-negative ion electrospray ionization tandem mass spectroscopy (LC/MS2), essentially as described.(24,25) LC/MS2 was performed using a Thermo-Finnigan LCQ Advantage MAX ion trap mass spectrometer (Thermo, San Jose, CA). Standards were prepared by reacting 25 mg pure chemical with 3 mL Brady's reagent (37 mg/mL 2,4-dinitrophenylhydrazine in 75% methanol containing 7.5% (v/v) concentrated sulfuric acid). The resulting hydrazone derivatives were collected by centrifugation, washed 3X with 20% (v/v) methanol in water, and dried under a stream of air. The analytes were chromatographically separated using a Gemini C18 HPLC column (150 × 2 mm, 5 μ) eluted at 0.3mL/min with the following stepwise gradient of acetonitrile and water at 40°C: 40→50% acetonitrile from 0 to 26 minutes; 50→100% acetonitrile from 26 to 35 minutes; 100% acetonitrile from 35 to 37.5 minutes. A representative chromatogram showing the precursor-to-product ion transitions used to detect each analyte is provided as Supplementary Figure 3. Analysis of these aldehydes and ketones in WSPM achieved by reacting 115 μg WSPM from each impactor stage fraction (from Figures 1 and 3) with 250 μL Brady's reagent, and processing as described above. For comparison of specific aldehydes and ketones in the various size fractions of pine and mesquite WSPM, peak intensity was used (Supplementary Figure 4). For quantitative analysis of formaldehyde and coniferaldehyde (i.e., TRPA1 agonists), standard curves were used.</p><!><p>Cells were cultured in 25 cm2 flasks and grown to 90% density. Total RNA was extracted from cells using the RNeasy mini kit (QIAGEN, Valencia, CA), and 2.5 μg of the total RNA was converted to cDNA using iScript (BioRad, Hercules, CA). The resulting cDNA was diluted 1:5 for analysis by quantitative real-time PCR (qPCR). qPCR was performed using LightCycler 480 SYBR Green I Master Mix (Roche, Indianapolis, IN) with a Light-Cycler 480 System as previously described.(22) Values for TRPA1 were normalized to β2-macroglobulin (β2M). Primer sequences were: β2M (+) 5′ – GATGAGTATGCCTGCCGTGTG – 3′ and (−) 5′ – CAATCCAAATGCGGCATCT – 3′; human TRPA1 (+) 5′ – TCACCATGAGCTAGCAGACTATTT – 3′ and (−) 5′ – GAGAGCGTCCTTCAGAATCG – 3′.</p><!><p>Values represent the mean ± SEM unless otherwise stated. One-way or two-way ANOVA with post-testing at the 95% confidence interval was used to determine significance, as indicated in each figure legend.</p><!><p>An image showing the size distribution of pine PM collected on the stages of the Anderson cascade impactor plates is provided as Figure 1. The majority of material deposited on stages 5 and 6 corresponding to PM 0.43 to 2.1 μm; the WSPM appears as dark colored material on the circular, silver impactor stage plates. An identical deposition pattern was observed for mesquite PM, except that the PM was less oily (data not shown). Pine and mesquite PM from stage 5 (1.1–2.1 μm) was used for initial TRP channel activation studies due to the relative abundance of material in this fraction.</p><!><p>The concentration-response relationships for TRPA1 activation by various cdPM are shown in Figure 2A. TRPA1-overexpressing HEK-293 cells were treated with pine and mesquite PM, CSC, and DEP-EtOH, which contains TRPA1 agonists,(15) at concentrations of 0, 0.58, 1.15, and 2.3 mg/mL. The rank order for potency was DEP-EtOH > pine PM and CSC > mesquite PM. Kinetic curves comparing the change in intracellular calcium content due to TRPA1 activation by the WSPM (1.15 mg/mL concentration) are shown in Figure 2B. As in Figure 2A, the rank order was the DEP-EtOH > pine PM and CSC > mesquite PM. DEP was the only PM that reached a maximum change in fluorescence (ΔF) during the 72 s measurement period, but pine PM and CSC ultimately produced the same maximum at longer time periods; mesquite did not reach a maximum at the 1.15 mg/mL treatment concentration.</p><p>The relative capacity of the different size fractions of pine PM (0.43 μm to 10 μm) to activate TRPA1 was also evaluated. The change in intracellular calcium resulting from TRPA1 activation in TRPA1-overexpressing HEK-293 cells treated with 0.73 mg/mL PM from the various impactor stages/size fractions are represented in Figure 3. Pine PM <1.1 μm activated TRPA1 equal to a maximum stimulating concentration of the prototypical TRPA1 agonist AITC (150 μM). As particle size increased, the ability to activate TRPA1 decreased. On a per mass basis, PM <1.1 μm was most potent, followed by PM 1.1 to 4.7 μm; PM >4.7 μm did not activate TRPA1. Comparable to the pine PM, mesquite PM <2.1 μm was the most potent and PM > 4.7 μm did not activate TRPA1 (data not shown). WSPM from stage 5 (1.1–2.1 μm) was used for the remainder of the studies due to abundance and relative potency of this fraction.</p><!><p>The selectivity of pine PM for TRPA1 was studied using isolated mouse trigeminal ganglia neurons (TG) as a model for studying interactions between WSPM and TRPA1-expressing sensory neurons in human and animal airways. A cellular response comparable in magnitude to that of the prototypical TRPA1 agonist AITC (50 μM) was observed in neurons treated with pine PM at 0.23 mg/mL (Figures 4A). Responses of TG neurons to the pine PM were inhibited ~90% by co-treatment with the TRPA1 antagonist HC-030031 (50 μM), which was proportionally greater than the ~75% inhibition observed for AITC (Figure 4B).</p><p>The selectivity of pine PM for TRPA1 was also studied using A549 cells as a general model for non-neuronal lung cells that express TRPA1.(18,19) Expression of TRPA1 mRNA by A549 cells was confirmed using quantitative PCR (Figure 5A). Consistent with this result, the TRPA1 agonist AITC (200 μM; EC50~145 μM in A549 cells) and pine PM promoted concentration-dependent calcium flux, which was inhibited by co-treating cells with the TRPA1 antagonist HC-030031 (Figure 5B). However, at 2.3 mg/mL pine PM, HC-030031 was unable to completely inhibit calcium flux, suggesting either a non-specific response of A549 cells to this concentration of pine PM, or the activation of additional WSPM-sensitive calcium channels. TRPV1, M8, and V4 are also expressed by A549 cells, but HEK-293 cells over-expressing these channels did not respond to either pine or mesquite PM at the 2.3 mg/mL concentration (data not shown).</p><!><p>Chemicals representing several major classes of combustion by-products found in wood smoke particles (i.e., fatty acids, aldehydes, ketones, resin acids, furans, etc.) were selected for screening as TRPA1 agonists based on their relative abundance in wood smoke emissions.(26–31) The structures and results for TRPA1 activation by various chemical constituents of WSPM (250 μM) are shown in Table 1. Coniferaldehyde, 3,5-ditert-butylphenol, and perinaphthenone activated TRPA1. However, the aldehydes furfural, 5-hydroxymethylfurfural, glyoxal, 4-hydroxybenzaldehyde, and vanillin did not. The fatty acid, palmitic acid, also failed to activate TRPA1. The resin acids agathic acid and isocupressic acid were TRPA1 agonists, but structurally related abietic acid, dehydroabietic acid, dihydroagathic acid, and tetrahydroagathic acid were not. Isopimaric acid caused extensive calcium flux in both HEK-293 and TRPA1-overexpressing HEK-293 cells and, thus, was not concluded to be a specific TRPA1 agonist.</p><p>Vanillin, 4-hydroxybenzaldehyde, coniferaldehyde, 5-hydroxymethylfurfural, furfural, glyoxal, perinaphthenone and several other aldehydes and ketones reported to be constituents of WSPM (26–31) were verified in the pine and mesquite PM samples as their 2,4-dinitrophenylhydrazone conjugates. The relative abundance of individual aldehydes and ketones from Table 1, those previously reported to be TRPA1 agonists (e.g., formaldehyde and acrolein),(15) and others that are constituents of WSPM, is shown graphically in Supplemental Figure 4 as well as in the chromatograms in Supplemental Figure 5. In general, the content of aldehydes and ketones (i.e., 2,4-dinitrophenylhydrazine-reactive substances) in the pine and mesquite PM increased as the size of the particles decreased. Additionally, pine PM contained greater total quantities of 2,3-dinitrophenylhydrazine-reactive substances, particularly PM>4.7 μm, since hydrazone pellets were not recovered from mesquite PM collected from stages 1 and 2. A quantitative comparison of the TRPA1 agonists coniferaldehyde (Table 1) and formaldehyde,(15) for the various size fractions of pine and mesquite PM is shown in Figure 6. Pine PM contained up to ~700 ± 80 ng/mg coniferaldehyde, which was near the limits of quantification in the mesquite PM. This quantity could produce a concentration of ~4 μM coniferaldehyde in a 1 mg/mL sample of pine PM. Pine PM also contained up to 1,300 ± 200 ng/mg formaldehyde, while mesquite PM contained up to 700 ± 200 ng/mg. These quantities could produce concentrations of ~43 and 23 μM formaldehyde in 1 mg/mL samples of pine and mesquite PM, respectively. Perinaphthenone was inconsistently detected in both the pine and mesquite PM samples. Isopimaric acid, dehydroabietic acid, and abietic acid were also quantified using gas chromatography-mass spectroscopy by the USDA Poisonous Plants Research Laboratory, Logan Utah. In the pine stage 5 sample, isopimaric acid, dehydroabietic acid, and abietic acid were ~270 ng/mg, ~785 ng/mg, and ~7,350 ng/mg, respectively, potentially producing concentrations of ~0.9, 3, and 24 μM in a 1 mg/mL sample of pine PM. However, only trace quantities of isopimaric and abietic acid were present in the same size fraction of mesquite PM. For all WSPM constituents, the values were consistent with prior reports of emission rates from pine and other hardwoods (26–33).</p><!><p>The relative contributions of the electrophile/oxidant-sensing site (i.e., C621, C641, C665 and K710)(34,35) and the menthol-binding site (i.e., S873 and T874)(21) in TRPA1 activation by WSPM and select chemical constituents thereof, was also assessed. Responses of wild-type TRPA1, the TRPA1-C621A/C641A/C655A/K710R (TRPA1-3CK) mutant, and the TRPA1-S873V/T874L (TRPA1-ST) mutant were compared (Figure 7). The TRPA1-3CK mutant was not activated by any of the agonists, suggesting that the TRPA1-3CK protein had limited function. As such, glutathione (GSH) pre-treatment of WSPM and WSPM-associated chemicals was used to selectively inhibit TRPA1 activation by electrophiles. Samples of pine PM, mesquite PM, agathic acid, and 3,5-ditert-butylphenol were prepared at 0.09 mg/mL, 0.19 mg/mL, 75 μM, and 250 μM, respectively, and incubated for 10 min at room temperature, in the presence or absence of GSH (20 mM) in treatment media. Pre-incubation of the pine and mesquite PM with GSH significantly and comparably reduced the activation of both wild-type TRPA1 and the TRPA1-ST mutant (Figure 7), and there was no significant difference between activation of TRPA1 and the TRPA1-ST mutant for either pine or mesquite PM alone. Wild-type TRPA1 and TRPA1-ST mutant activation by agathic acid was also inhibited ~90% by GSH, similar to the WSPM samples. Conversely, GSH did not reduce wild-type TRPA1 activation by 3,5-ditert-butylphenol, but significant (~60–70%) reduction in response was observed with the TRPA1-ST mutant.</p><!><p>Ambient cdPM, particularly WSPM and BSPM, is increasingly being recognized as a cause of many adverse health effects in humans. Increased hospitalization rates due to exacerbation of pre-existing diseases such as asthma, chronic bronchitis, respiratory infection, development of chronic obstructive pulmonary disease, and premature death have been reported by numerous epidemiological studies investigating the effects of PM on human health.(1,2,4,5) The work presented here identifies TRPA1 as a selective molecular sensor for WSPM and chemical constituents thereof, in cells representative of airway sensory neurons, epithelial, and other non-neuronal cells that express TRPA1. Thus, it is proposed that TRPA1 may play an important role in regulating airway cell and respiratory responses to WSPM and related cdPM, highlighting the possibility of therapeutic modulation of TRPA1 to protect unusually sensitive individuals and/or high-risk populations (e.g., asthmatics, children, elderly) from developing adverse effects due to acute high-level and/or chronic exposure to WSPM and similar cdPM.</p><p>Environmental particulate matter is a complex mixture of solids, absorbed gasses, and liquid. The adverse health effects of environmental PM generally correlate with PM2.5,(1,2,5) which is primarily produced from the inefficient combustion of fossil fuels (e.g., gasoline, oil, diesel, coal) and biomass (e.g., wood, grass, dung).(1,2,36,37) The WSPM collected in this study was an oily, tar-like material mostly <2.1 μm in size. The WSPM was also soluble in DMSO, which is a property of environmental WSPM.(1,2,38) WSPM in the environment is generally the result of seasonal wildfires and home heating.(1,2,5,36,37) It has been shown to be stable and capable of transporting long distances from its source while retaining its ability to affect humans.(1,2,36,37) Source apportionment studies of environmental PM, often regardless of the location of PM collection, routinely report the presence of chemical tracers for wood/biomass smoke PM.(1,2,26–31) Thus, it is a possibility that TRPA1 may contribute to the development of adverse health effects associated with episodic high pollution events, particularly if the event is precipitated by a forest or range fire, or weather that increases the use of inefficient wood/biomass fireplaces and stoves.(1,2,5)</p><p>In the respiratory tract, TRPA1 is expressed by C-fibers sensory neurons originating from the trigeminal and vagal ganglia, which when activated result reduce respiratory drive, trigger cough and bronchoconstriction, and the release of substance P, neurokinin A, and calcitonin-gene related peptide, which cause neurogenic inflammation. Mouse TG neurons were used as a general model of TRPA1-expressing primary sensory neurons. It was determined that pine PM activated calcium flux in TG neurons in a TRPA1-dependent manner, as previously shown for DEP(15,16) and CSC.(13) The similarities between DEP, CSC, and WSPM in TG neurons is most likely do to the presence of similar chemical agonists present in these types of cdPM, including chemicals like acrolein, 3,5-ditert-butylphenol, formaldehyde, PAHs, aldehydes and ketones.</p><p>TRPA1 is also expressed by primary human lung cells such as fibroblasts, small airway epithelial, and smooth muscle cells, where activation by TRPA1 agonists has been shown to promote non-neurogenic inflammation.(18,19) Using A549 cells as a model of TRPA1-expressing lung cells, it was also demonstrated that WSPM selectively activated TRPA1 at lower concentrations, which more closely mimic potential human exposure levels, particularly higher levels of exposure that may occur in fire fighters or those living in primitive dwellings where wood- and biomass burning is essential. The latter group of individuals are exposed to high concentrations of cdPM such as WSPM over longer periods of time, thus it is possible that TRPA1 may play an important role in some of the adverse effects on the respiratory system that are most pronounced in these populations.(1,2,5)</p><p>While the collective data presented by this study support a role for TRPA1 in mediating WSPM pulmonary toxicity, literature overwhelmingly implies a mechanism involving oxidant-mediated injury whereby reactive oxygen species and/or redox-cycling molecules present in WSPM promote pulmonary inflammation and injury.(6,7,9,10,39) Although not specifically tested, it is possible that TRPA1 may also be central to the oxidative injury process. In addition to being directly activated by WSPM, several by-products of oxidative damage, including 4-HNE(40) and 4-oxononenal,(17) are TRPA1 agonists. Thus, further research is needed to differentiate the precise contributions of direct versus indirect activation of TRPA1, at multiple levels of exposure, in order to better elucidate the specific mechanisms by which WSPM and similar cdPM ultimately produce short- and long-term effects on human health, particularly respiratory and cardiovascular health.</p><p>Finally, TRPA1 has three distinct mechanisms for activation: The first involves an electrophile/oxidant-sensitive site;(34,35) the second involves a site selective for menthol(21) and structurally similar chemicals (e.g., propofol and 3,5-ditert-butylphenol); and the third involves an undefined mechanosensitive component that has been shown to be responsive to insoluble components of various forms of PM.(15,22) Here, it was shown that WSPM and specific chemical components of WSPM, presumably acting on TRPA1 as a complementary mixture, predominantly activated TRPA1 through the electrophile binding site, with only 3,5-ditert-butylphenol acting through the menthol site. Based on these results, and those previously published for DEP(15) and CSC,(13) it is probable that the majority of cdPM that activate TRPA1 will do so via this mechanism since cdPM generally contain similar types of electrophiles. However, this generalization must be cautiously interpreted. While many cdPM contain the same chemicals,(2,26–31) the chemical composition of all types of cdPM is not identical. Thus, the biological activity of different cdPM in different cell, organ, animal models, or in humans will vary. For example, CSC, DEP, WSPM from two different types of wood, and even different size fractions of PM from a single type of wood, have strikingly different potencies as TRPA1 agonists. These data indicate that small differences in the chemical composition can markedly affect the ability of the aggregate material to activate TRPA1, and presumably other mediators of biological effects. Thus, it is emphasized that, in addition to evaluating the expression of different PM/PM constituent-sensitive ion channels such as TRPA1,(13–16,19) TRPM2,(42) TRPM8,(15) TRPV1,(22,43) and TRPV4 (15,44) in cell, organ, and animal models, and humans, that the chemical composition of the PM being studied, is more carefully assessed, such that more accurate mechanisms of biological activity can be ascertained.</p>

PubMed Author Manuscript

Pethidine dose and female sex as risk factors for nausea after esophagogastroduodenoscopy

Nausea and vomiting after esophagogastroduodenoscopy have not been studied in detail. The aim of this study was to evaluate the risk factors for post-endoscopic nausea. We performed a case-control study at the Toyoshima Endoscopy Clinic. Eighteen patients with post-endoscopic nausea and 190 controls without post-endoscopic nausea were analyzed. We conducted univariate and multivariate logistic regression analyses with respect to patient age; sex; body height; body weight; the use of psychotropic drugs as baseline medications; and the dosing amounts of midazolam, pethidine, flumazenil and naloxone. On univariate analysis, post-endoscopic nausea was significantly related with patient age (odds ratio = 0.946); female sex (odds ratio = 10.85); body weight (odds ratio = 0.975); and the dose per kg body weight of pethidine (odds ratio = 53.03), naloxone (odds ratio = 1.676), and flumazenil (odds ratio = 1.26). On multivariate analysis, the dose per kg body weight of pethidine (odds ratio = 21.67, p = 0.004) and female sex (odds ratio = 13.12, p = 0.047) were the factors independently associated with post-endoscopic nausea. The prevalence of nausea after esophagogastroduodenoscopy was 0.49% (18/3,654). In conclusion, post-endoscopic nausea was associated with the dose of pethidine and female sex.

pethidine_dose_and_female_sex_as_risk_factors_for_nausea_after_esophagogastroduodenoscopy

Introduction<!>Subjects<!>Endoscopic examination<!>Statistical analysis<!>Results<!>Discussion

<p>Esophagogastroduodenoscopy (EGD) is an important medical tool in the screening, diagnosis, and treatment of a variety of gastrointestinal diseases.(1,2) Reported post-endoscopic complications include throat pain, nausea, and headache. Though some these are rare, the rate of post-endoscopic nausea, which is considered to be relatively major among the different post-endoscopic complications, has been reported to be 1.5%.(3) Post-endoscopic nausea could be caused by the use of a peri-endoscopic sedative and analgesic medications, air insufflation, and pharyngeal stimulation.(4) Post-endoscopic nausea is one of the most undesirable complications, and can complicate management after EGD, delaying discharge and recovery. Furthermore, post-endoscopic nausea can lead to refusal to undergo repeat EGD. In rare cases, severe nausea after EGD may require hospitalization. Following surgery with anesthesia, post-operative nausea and vomiting (PONV) affects about 20–40% of patients. Extensive literature about PONV suggests prophylactic strategies and pharmacological management tailored to the patient's risk level.(5) However, no reported study has examined post-endoscopic nausea. Thus, the aim of this study was to evaluate the risk factors associated with post-endoscopic nausea.</p><!><p>We performed a case-control study at the Toyoshima Endoscopy Clinic. Between May 2016 and April 2017, 3,654 patients underwent EGD. Among them, patients with post-endoscopic nausea were enrolled in the current study. The control group included consecutive patients who underwent EGD between April 11, 2017 and April 28, 2017. The diagnostic criteria for post-endoscopic nausea include grade 2–3 nausea or vomiting within 12 h after EGD. Grade 1 nausea is defined as loss of appetite without alteration in eating habits according to common terminology criteria for adverse events (CTCAE). Grade 2 nausea is defined as decreased oral intake without dehydration. Grade 3 nausea is defined as inadequate oral intake with an indication for tube feedings or hospitalization. Grade 4 is defined as life-threatening consequences. Patients who simultaneously underwent EGD and colonoscopy were excluded.</p><p>The following demographic and clinical characteristics were collected from medical records: patient age; sex; body height; body weight; body mass index (BMI); the use of psychotropic drugs as baseline medications; and the administered doses of midazolam, pethidine, flumazenil and naloxone. Written, informed consent was obtained from each patient included in the study. This study was approved by Ethical Review Committee of Hattori Clinic (September 7, 2017), and conformed to the ethical guidelines of the 1975 Declaration of Helsinki as reflected in a prior approval by the institution's human research committee.(6)</p><!><p>EGD was performed by 14 experienced endoscopists. EGD was performed as a screening method during a health evaluation, for follow-up of gastritis and/or gastric tumor, for the examination for abdominal symptoms, to investigate an abnormality of photofluorography, to examine abnormal serum pepsinogen levels, or due to a positive finding of H. pylori antibody. The pharynx of the patients was topically anesthetized with a gargle of lidocaine hydrochloride 2% viscous solution (Xylocaine® Viscous 2%; AstraZeneca Inc., Cambridge, UK) before the EGD.(7) The endoscopists were allowed to use their clinical judgement to decide the amount and type of sedative and analgesic medication and the antagonist—midazolam (0–10 mg), pethidine (0–70 mg), flumazenil (0–0.5 mg) and naloxone (0–0.4 mg)—to be used. Following the EGD, the patients were transferred to the recovery room. All adverse events including nausea and vomiting were evaluated by the recovery room nurse. Patients were requested to return 10 to 14 days later for the explanation of their EGD results and were also interviewed regarding any additional adverse events.</p><!><p>We evaluated the effects of patient age; sex; body weight; BMI; the use of psychotropic drugs as baseline medications; and the dose per kg body weight of midazolam, pethidine, flumazenil, and naloxone on post-endoscopic nausea. The clinical parameters were analyzed via the chi-square or univariate logistic regression analysis. The predictors found to be associated with post-endoscopic nausea on univariate analysis (p<0.1) were subsequently assessed using a multiple logistic regression method to identify independent factors.(8) Patient age, body weight, BMI, and the dose per kg body weight of each drug were included as continuous variables in the univariate and multivariate logistic regression analyses. A p value of less than 0.05 was considered statistically significant. The data were analyzed using the StatMate IV software (ATOMS, Tokyo, Japan).</p><!><p>Of the 25 eligible patients, seven were excluded because they simultaneously underwent EGD and colonoscopy. Of the 274 controls, 84 were excluded because they simultaneously underwent EGD and colonoscopy. Finally, 18 patients with post-endoscopic nausea and 190 controls without post-endoscopic nausea were analyzed.</p><p>Table 1 shows the univariate and multivariate analysis results for post-endoscopic nausea. On univariate analysis, post-endoscopic nausea was significantly related with patient age (odds ratio = 0.946, p = 0.0054), female sex (odds ratio = 10.85, p = 0.022), body weight (odds ratio = 0.975, p = 0.0511), the dose per kg body weight of pethidine (odds ratio = 53.03, p<0.001), the dose per kg body weight of naloxone (odds ratio = 1.676, p<0.001), and the dose per kg body weight of flumazenil (odds ratio = 1.26, p = 0.0374).</p><p>On multivariate analysis, the dose per kg body weight of pethidine [odds ratio = 21.67, 95% confidence interval (CI) = 2.547–184.3, p = 0.005] and female sex (odds ratio = 13.12, 95% CI = 1.035–166.2, p = 0.047) were independently associated with post-endoscopic nausea (Table 2).</p><p>The prevalence of nausea after EGD was 0.49% (18/3,654) in this study. Two patients received only pethidine, 1,735 patients received both pethidine and midazolam, 1,316 patients received only midazolam, and 601 patients received neither pethidine nor midazolam. Among patients who did not receive pethidine, the prevalence was 0.21% (4/1,917). Among those receiving pethidine, the prevalence was 0.81% (14/1,737). The prevalence in patients receiving pethidine was significantly higher than that in patients who did not receive pethidine (p = 0.019).</p><!><p>This is a first report about post-endoscopic nausea. In this study, the dose per kg body weight of pethidine and female sex were found to be independent risk factors for the onset of post-endoscopic nausea.</p><p>Peri-endoscopic sedative and analgesic medications have often been used to provide patient comfort, reduce procedure time, and improve examination quality during EGD.(9–11) Benzodiazepines such as midazolam are the most commonly used sedatives,(12,13) and these are generally given to the patient along with an opiate (pethidine or fentanyl) for synergism.(14,15) Two randomized controlled trials compared sedation with midazolam plus pethidine versus midazolam alone.(16,17) Sedation with midazolam and pethidine led to significantly less retching, which interfered with the procedure, and endoscopists reported favoring the use of both medications over the use of midazolam alone. However, adverse effects of opiates include nausea and vomiting. Opiates mainly inhibit the neurotransmission of pain by binding to specific opioid receptors that are present in the central nervous system and peripheral tissues.(18) Nausea and vomiting resulting from stimulation of the medullary chemoreceptor trigger zone occur in a dose-independent manner.(19) We also found that post-endoscopic nausea was associated with the dose of pethidine.</p><p>In this study, women experienced post-endoscopic nausea more often than men. Silva et al. reported that the risks of postoperative nausea and vomiting were associated with female sex in surgery and general anesthesia settings.(20) The observed sex differences could be explained by the presence of a different socialization process for men and women that influences the willingness to communicate distress.(21) Women report more pain than men,(22) and describe more numerous somatic symptoms than men.(21) Other possible explanations include the interaction between sex hormones and opiates and the hormone fluctuations associated with the menstrual cycle.(23)</p><p>The limitations of this study include its retrospective and case-control design. A follow-up study should be performed prospectively to confirm and clarify the characteristics of nausea and vomiting after EGD.</p><p>In conclusion, we found that post-endoscopic nausea was associated with the dose of pethidine and female sex. Endoscopists should recognize that the use of high-dose opiates in female patients might provoke nausea and vomiting after EGD.</p>

PubMed Open Access

Catalytic two-electron reduction of dioxygen catalysed by metal-free [14]triphyrin(2.1.1)

The catalytic two-electron reduction of dioxygen (O 2 ) by octamethylferrocene (Me 8 Fc) occurs with a metal-free triphyrin (HTrip) in the presence of perchloric acid (HClO 4 ) in benzonitrile (PhCN) at 298 K to yield Me 8 Fc + and H 2 O 2 . Detailed kinetic analysis has revealed that the catalytic two-electron reduction of O 2 by Me 8 Fc with HTrip proceeds via proton-coupled electron transfer from Me 8 Fc to HTrip to produce H 3 Tripc + , followed by a second electron transfer from Me 8 Fc to H 3 Tripc + to produce H 3 Trip, which is oxidized by O 2 via formation of the H 3 Trip/O 2 complex to yield H 2 O 2 . The rate-determining step in the catalytic cycle is hydrogen atom transfer from H 3 Trip to O 2 in the H 3 Trip/O 2 complex to produce the radical pair (H 3 Tripc + HO 2 c) as an intermediate, which was detected as a triplet EPR signal with fine-structure by the EPR measurements at low temperature. The distance between the two unpaired electrons in the radical pair was determined to be 4.9 Å from the zero-field splitting constant (D).

catalytic_two-electron_reduction_of_dioxygen_catalysed_by_metal-free_[14]triphyrin(2.1.1)

Introduction<!>Protonation of HTrip with HClO 4<!>Conclusion<!>General procedure<!>Spectroscopic measurements<!>Kinetic measurements<!>Electrochemical measurements<!>Spectroelectrochemical measurements<!>EPR measurements<!>Theoretical calculations

<p>Utilization of natural energy to produce chemical energy consisting of earth-abundant elements is an essential technology for building a society based on the sustainable use of materials. Hydrogen peroxide (H 2 O 2 ) produced by two-electron reduction of O 2 is a versatile and environmentally benign oxidant, which is widely used on a large industrial scale. 1,2 Furthermore, H 2 O 2 has been proposed as a sustainable energy carrier that can be used in fuel cells, where direct and efficient conversion of chemical to electrical energy is required. [3][4][5] However, the anthraquinone process, currently used to produce H 2 O 2 in industry, requires potentially explosive hydrogen and a noble metal catalyst. 6 Extensive efforts have so far been devoted to provide an alternative way to produce H 2 O 2 photochemically or thermally without the use of noble metal catalysts. [7][8][9][10][11][12][13] In many cases, redox-active transition metal-based complexes such as cobalt, 14-23 iron, [24][25][26][27] and copper complexes, [28][29][30][31] have been employed as O 2 reduction catalysts, because triplet O 2 is inactive towards organic compounds due to spin restriction in the absence of an appropriate catalyst. 32 Recently, nitrogen-doped carbon materials have attracted increasing attention as an efficient metal-free catalyst for the catalytic reduction of O 2 . [33][34][35] However, the catalytic mechanism has yet to be well understood, because few spectroscopic studies to detect reaction intermediates in a catalytic cycle have been performed on heterogeneous systems. In homogeneous systems, reduced avin analogues involved in avoenzymes have so far been known to play a pivotal role in the catalytic reduction of O 2 , which is a key step of biological oxidation. 36,37 In particular, the deprotonated states of reduced avin analogues, which are thermodynamically more able to reduce O 2 via an electron-transfer process, are considered to be a reactive intermediate in the reduction of O 2 . 38 On the other hand, Girault and coworkers recently reported that the free base porphyrin has the ability to catalyse the twoelectron reduction of O 2 using one-electron reductants such as ferrocene at liquid-liquid interfaces. 39 In such systems, although the catalytic mechanism of metal-free organocatalysts has yet to be claried, the oxidation state of the organocatalyst is thought to remain the same during the catalytic reduction of O 2 . Thus, no electron-transfer reduction of organic catalysts has been reported in relation to the catalytic reduction of O 2 .</p><p>In this context, Nocera and coworkers recently reported the stabilization of the peroxide dianion within the cavity of a hexacarboxamide cryptand, 40 where strong hydrogen bond donors are arranged to completely surround the peroxide dianion with a partial positive charge. This result provides support for the proposal that metal-free organocatalysts, which have multiple hydrogen bonding moieties, can efficiently catalyse O 2 reduction.</p><p>We report herein the catalytic two-electron reduction of O 2 by an one-electron reductant, octamethylferrocene (Me</p><!><p>HTrip was protonated by addition of perchloric acid (HClO 4 ) to an air-saturated benzonitrile (PhCN) solution of HTrip. The characteristic absorption bands for HTrip at 524 and 581 nm decreased in intensity, with an increase in the absorption band at 565 nm, exhibiting clean isosbestic points, as shown in Fig. 1a. As can be seen in Fig. 1b, the absorbance change at 565 nm is saturated in the presence of 1 equiv. of HClO 4 . Thus, HTrip is protonated to afford H 2 Trip + , as given by eqn (1).</p><p>The pK a value of H 2 Trip + in PhCN was estimated to be 15.6 from the titration of HTrip with triuoroacetic acid (TFA), as shown in Fig. S1 in the ESI. † The pK a value of H 2 Trip + is slightly larger than that of free base porphyrin analogues. 42 There is no further protonation due to strong repulsion between NH protons in the small macrocyclic ligand, as reported previously. 41</p><p>Electrochemical measurements of HTrip in the presence of HClO 4</p><p>Electrochemical measurements of HTrip were performed in deaerated PhCN containing 0.10 M TBAPF 6 , as shown in Fig. 2.</p><p>A cyclic voltammogram of HTrip exhibits reversible reduction waves at E 1/2 ¼ À1.13 and À1.37 V (vs. SCE), which correspond to the rst and second one-electron reduction of HTrip. The rst one-electron oxidation occurs at E 1/2 ¼ 1.04 V, which is followed by an irreversible oxidation (Fig. 2a). The formation of HTripc À was detected by UV-vis absorption spectra in the electrochemical reduction of HTrip at a controlled potential of À1.25 V vs. SCE in the thin-layer cell, as shown in Fig. S2 in the ESI. † By addition of HClO 4 , the rst reduction potential of HTrip was positively shied from E 1/2 ¼ À1.13 V to À0.31 V (vs. SCE) because of the protonation of HTrip, but the reduction became irreversible (Fig. 2b). In such a case, proton-coupled electron transfer from an electron donor with the one-electron oxidation potential, which is less negative than À0.31 V, to HTrip may be thermodynamically feasible (vide infra).</p><p>Electron-transfer reduction of HTrip in the presence of HClO ). The stoichiometry of the overall reaction is given in Scheme 1.</p><p>The rate of proton-coupled electron-transfer reduction of H 2 Trip + (k et ) to form H 3 Tripc + was determined from the dependence of the observed rate constant (k obs ) on concentrations of Me 8 Fc and HClO 4 , as shown in Fig. 4. The k obs value was determined from the increase in absorbance at 738 nm due to H 3 Trip, which obeyed rst-order kinetics (Fig. S3 in the ESI †). The k obs value increased linearly with increasing concentrations of Me 8 Fc and HClO 4 , as shown in Fig. 5. Thus, the rate of formation of H 3 Trip is given by eqn (2). HClO 4 ][Me 8 Fc] (2)</p><p>Chart 1 Structure of HTrip.</p><p>The rate of formation of Me 8 Fc + in the catalytic reduction of O 2 with excess Me 8 Fc and HClO 4 in Fig. 7b The dependence of the rst-order rate constant for the formation of Me 8 Fc + on the concentrations of HTrip, HClO 4 , Me 8 Fc, and O 2 was examined, as shown in Fig. S14 (in the ESI †), where the rst-order rate constants were determined from the initial slopes of the rst-order plots in order to avoid further complication due to the deactivation of the catalyst during the reactions, as shown in Fig. S15 (in the ESI †). The observed rst-order rate constant (k obs ) was proportional to the concentration of HTrip, whereas the k obs value remained constant irrespective of the concentration of HClO 4 or Me 8 Fc (Fig. 8). Although no degradation of HTrip occurred under the present acidic conditions (Fig. S16</p><p>where k cat is the rate constant of the hydrogen atom transfer from H 3 Trip to O 2 in the H 3 Trip/O 2 complex. Because the concentration of the H 3 Trip/O 2 complex is given by eqn ( 5)</p><p>using the formation constant (K), the initial concentration of HTrip, which is converted to H 3 Trip in the catalytic reaction, and the concentration of O 2 , eqn ( 4) is rewritten as eqn (6).</p><p>This kinetic equation agrees with the experimental observations in Fig. 8. The k cat and K values were determined from the dependence of the catalytic rate on the concentration of O 2 (Fig. 8d) to be 0.5 s À1 and 8.4 Â 10 2 M À1 , respectively. Although the radical pair (H 2 Tripc/HO 2 c) in Scheme 3 cannot be detected during the catalytic reaction, the formation of the radical pair (H 2 Tripc/HO 2 c) was successfully detected by EPR measurements using 1-benzyl-1,4-dihydronicotinamide dimer [(BNA) 2 ] 44 as an electron donor to produce H 3 Trip under photoirradiation at low temperature. The observed EPR spectrum in aerated PhCN in the presence of HClO 4 at low temperature is shown in Fig. 9. A triplet ne structure EPR signal was observed as well as the typical anisotropic signals for HO 2 c with the g || value of 2.0341, and isotropic signals for H 2 Tripc at 2.0030. 45,46 From the zero-eld splitting value (D ¼ 230 G), the distance (r) between two unpaired electrons was determined using the relation D ¼ 27 800/r 3 47 to be 4.9 Å. This distance is consistent with the estimated distance between O 2 and H 3 Trip in the H 3 Trip/O 2 complex by DFT calculations (Fig. 9b).</p><!><p>Metal-free triphyrin acts as an efficient catalyst for the twoelectron reduction of O 2 by Me 8 Fc to produce H 2 O 2 in the presence of HClO 4 in PhCN at 298 K. The rate-determining step (RDS) in the catalytic cycle has been found to be hydrogen atom transfer from H 3 Tip to O 2 in the H 3 Trip/O 2 complex to produce the radical pair (H 3 Tripc + /HO 2 c), which was detected as a triplet species by EPR at 80 K. The distance between the two unpaired electrons (4.9 Å) determined from the zero-eld splitting constant (D) agrees with the distance in the H 3 Trip/O 2 complex calculated by DFT. The present study provides valuable insight into the catalytic mechanism of the two-electron reduction of O 2 with an organic catalyst, and may lead to the development of more efficient metal-free organic catalysts for the selective twoelectron reduction of O 2 to produce H 2 O 2 .</p><!><p>Chemicals were purchased from commercial sources and used without further purication, unless otherwise noted. Perchloric acid (HClO 4 , 70%), triuoroacetic acid (TFA), ferrocene (Fc), and 1,1-dimethylferrocene (Me 2 Fc) were purchased from Wako Pure Chemical Industries Ltd. Octamethylferrocene (Me 8 Fc) and decamethylferrocene (Me 10 Fc) were received from Sigma Aldrich. Fc, Me 2 Fc, Me 8 Fc, and Me 10 Fc were puried by sublimation or recrystallization from ethanol. Benzonitrile (PhCN) used for spectroscopic and electrochemical measurements was distilled over phosphorus pentoxide prior to use. 48 [14]Triphyrin(2.1.1) [HTrip] was synthesized according to the reported procedure. 41 Fe(II)(TMC)(OTf) 2 (TMC ¼ 1,4,8,11-tetramethyl-1,4,8,11-tetraazacyclotetradecane; OTf ¼ CF 3 SO 3 ) was prepared according to a literature method. 43 Tetra-n-butylammonium hexauorophosphate (TBAPF 6 ) was twice recrystallized from ethanol and dried in vacuo prior to use. 1 H NMR spectra (300 MHz) were recorded on a JEOL AL-300 spectrometer at room temperature and chemical shis (ppm) were determined relative to tetramethylsilane (TMS). UV-vis absorption spectroscopy was carried out on a Hewlett Packard 8453 diode array spectrophotometer at room temperature using a quartz cell (light path length ¼ 1 cm).</p><!><p>The amount of hydrogen peroxide (H 2 O 2 ) produced was determined by titration with iodide ion: a dilute CH 3 CN solution (2.0 mL) of the product mixture (50 mL) was treated with an excess amount of NaI, and the amount of I 3 À formed was determined from the absorption spectrum (l max ¼ 361 nm, 3 ¼ 2.8 Â 10 4 M À1 cm À1 ). 49 The formation of H 2 O 2 in the catalytic O 2 reduction with HTrip was again conrmed by the reaction between H 2 O 2 and Fe(II)(TMC)(OTf) 2 to afford the corresponding Fe(IV)-oxo species. The amount of the Fe(IV)-oxo species produced was determined from the absorption spectrum (l max ¼ 820 nm, 3 ¼ 400 M À1 cm À1 ). 43 The turnover numbers (TON ¼ the number of moles of H 2 O 2 formed per mole of HTrip in the catalytic two-electron reduction of O 2 ) were determined from the concentration of produced Me 8 Fc + under catalytic conditions, where stoichiometric production of H 2 O 2 was conrmed by iodometric titration.</p><!><p>Rate constants of oxidation of ferrocene derivatives by O 2 in the presence of a catalytic amount of HTrip and an excess amount of HClO 4 in PhCN at 298 K were determined by monitoring the appearance of an absorption band due to the corresponding ferrocenium ions (Fc + , l max ¼ 620 nm, 3 max ¼ 330 M À1 cm À1 ; Me 2 Fc + , l max ¼ 650 nm, 3 max ¼ 290 M À1 cm À1 ; Me 8 Fc + , l max ¼ 750 nm, 3 max ¼ 410 M À1 cm À1 ; Me 10 Fc + , l max ¼ 780 nm, 3 max ¼ 450 M À1 cm À1 ). 14 At the wavelengths monitored, spectral overlap was observed with H 3 Trip (l ¼ 738 nm (3 ¼ 1.6 Â 10 3 M À1 cm À1 )), H 3 Trip/O 2 (l ¼ 720 nm (3 ¼ 1.2 Â 10 3 M À1 cm À1 )). The concentration of O 2 in an air-saturated PhCN solution was determined to be 1.7 Â 10 À3 M as reported previously. 50 The concentrations of ferrocene derivatives employed for the catalytic reduction of O 2 were much larger than that of O 2 , as O 2 is the rate-limiting reagent in the reaction solution. The PhCN solutions containing various concentrations of O 2 for the kinetic measurements were prepared by N 2 /O 2 mixed gas bubbling using a KOFLOC GASBLENDER GB-3C. Typically, a PhCN stock solution of a ferrocene derivative was added using a</p><!><p>Cyclic voltammetry (CV) measurements were performed on an ALS 630B electrochemical analyser and voltammograms were measured in deaerated PhCN containing 0.10 M TBAPF 6 as a supporting electrolyte at room temperature. A conventional three-electrode cell was used with a glassy carbon working electrode (surface area of 0.3 mm 2 ) and a platinum wire as the counter electrode. The glassy carbon working electrode (BAS) was routinely polished with BAS polishing alumina suspension and rinsed with acetone before use. The potentials were measured with respect to the Ag/AgNO 3 (1.0 Â 10 À2 M) reference electrode. All potentials (vs. Ag/AgNO 3 ) were converted to values vs. SCE by adding 0.29 V. 51 Redox potentials were determined using the relation E 1/2 ¼ (E pa + E pc )/2.</p><!><p>UV-visible spectroelectrochemical experiments were performed with a home-built thin-layer cell (1 mm) that had a light transparent platinum net working electrode. Potentials were applied and monitored with an ALS 730D electrochemical analyser.</p><!><p>EPR spectra were measured on a JEOL X-band EPR spectrometer (JES-ME-LX) using a quartz EPR tube containing a deaerated frozen sample solution at 80 K. The internal diameter of the EPR tube is 4.0 mm, which is small enough to ll the EPR cavity but large enough to obtain good signal-to-noise ratios during the EPR measurements at low temperatures (at 80 K). EPR spectrum of HTripc À produced by the electrochemical reduction of HTrip was measured using a home-built three-electrode quartz EPR tube. Potentials were applied and monitored with an ALS 730D electrochemical analyser. EPR spectra were measured under nonsaturating microwave power conditions. The amplitude of modulation was chosen to optimize the resolution and the signal-to-noise (S/N) ratio of the observed spectra. The g values were calibrated with a Mn 2+ marker.</p><!><p>Density functional theory (DFT) calculations were performed on a 32CPU workstation (PQS, Quantum Cube QS8-2400C-064). Geometry optimisations were carried out using the B3LYP/ 6-31G(d) level of theory 52 for HTripc À , H 2 Trip + , H 3 Trip 2+ , H 3 Tripc + , and [H 3 Trip/O 2 ]. All calculations were performed using Gaussian 09, revision A.02. 53 Graphical outputs of the computational results were generated with the GaussView soware program (ver. 3.09) developed by Semichem, Inc. 54</p>

Royal Society of Chemistry (RSC)

Generalized Indirect Covariance NMR Formalism for Establishment of Multi-Dimensional Spin Correlations

Multidimensional nuclear magnetic resonance (NMR) experiments measure spin-spin correlations, which provide important information about bond connectivities and molecular structure. However, direct observation of certain kinds of correlations can be very time-consuming due to limitations in sensitivity and resolution. Covariance NMR derives correlations between spins via the calculation of a (symmetric) covariance matrix, from which a matrix-square root produces a spectrum with enhanced resolution. Recently, the covariance concept has been adopted to the reconstruction of non-symmetric spectra from pairs of 2D spectra that have a frequency dimension in common. Since the unsymmetric covariance NMR procedure lacks the matrix-square root step, it does not suppress relay effects and thereby may generate false positive signals due to chemical shift degeneracy. A generalized covariance formalism is presented here that embeds unsymmetric covariance processing within the context of the regular covariance transform. It permits the construction of unsymmetric covariance NMR spectra subjected to arbitrary matrix functions, such as the square root, with improved spectral properties. This formalism extends the domain of covariance NMR to include the reconstruction of non-symmetric NMR spectra at resolutions or sensitivities that are superior to the ones achievable by direct measurements.

generalized_indirect_covariance_nmr_formalism_for_establishment_of_multi-dimensional_spin_correlatio

Introduction<!>Theory<!>Materials and Methods<!>Results<!>Discussion and Conclusions

<p>Multidimensional nuclear magnetic resonance (NMR) is a powerful tool for probing molecular connectivity and structure by displaying magnetization transfer between nuclear spins due their magnetic interaction as correlation peaks in a multidimensional spectrum.1 However, multi-dimensional NMR spectra with high resolution and sensitivity require the acquisition of a large number of scans, which is NMR spectrometer time intensive.2 Establishment of direct correlations between insensitive nuclei, such as 13C and 15N, requires particularly long measurement times.3</p><p>Indirect covariance NMR4 offers a linear algebraic approach to establish correlations between pairs of hetero-nuclei that are coupled to a common set of protons. Formally, the indirect covariance transform of the N1 × N2 NMR spectrum X produces the (symmetric) spectrum C=(XXT)1∕2 (where the superscripts T and 1/2 denote the matrix transpose and matrix-square root, respectively). Unsymmetric covariance NMR5-8 generates asymmetric spectra via matrix multiplication of two distinct spectra that share (at least) one common dimension. An example is the multiplication of an 13C-1H HSQC9 with a 1H-1H TOCSY10 to correlate all 1H and 13C nuclei in the same spin system. This reconstructs a 13C-1H HSQC-TOCSY spectrum from two standard 2D experiments without requiring additional measurement time and thereby yields additional 13C, 1H correlations, which can facilitate chemical shift assignment by linking unassigned 13C chemical shifts to already assigned 1H and 13C chemical shifts.6 Hyperdimensional NMR reconstructs high-dimensional spectra, which are often asymmetric, from lower dimensional spectra for the purpose of protein resonance assignment.11,12 COBRA13,14 and Burrow-Owl15 apply linear algebraic spectral manipulations for the same purpose.</p><p>An important property of unsymmetric covariance NMR is that the sensitivity of the covariance spectrum is limited only by the sensitivity of the experiments it combines.16 For example, unsymmetric covariance of an 13C-1H HMBC17 with a 13C-1H HSQC spectrum establishes carbon-carbon correlations with the enhanced sensitivity characteristic of an inverse detected 13C-1H heteronuclear spectra rather than that of a direct detected 13C-13C correlation spectrum.4</p><p>A key difference between symmetric and un-symmetric covariance NMR is the applicability of the matrix-square root transform. The matrix-square root, which minimizes artifacts due to relay effects and chemical shift (near) degeneracy ("pseudo-relay effects")4,18-20 is properly defined only for symmetric and positive semi-definite covariance spectra, e.g. when the product matrix is a regular covariance matrix.</p><p>In this paper, a general approach is presented for constructing a covariance matrix from multiple NMR spectra. Since the standard covariance transform is recovered as a special case when identical spectra are used as input, the generalized covariance matrix formalism reconciles symmetric and un-symmetric covariance processing. The generalized covariance matrix is symmetric, which makes it amenable to the extraction of arbitrary matrix functions, including the matrix-square root and other matrix powers λ. Depending on the types of spectra that are correlated, application of the square root suppresses false positives. It is found that the analysis of the variation of covariance peak intensity as a function of λ is an effective indicator for the identification of false positives in unsymmetric covariance spectra. Covariation of a 13C-1H HMBC with a 1H-1H TOCSY spectrum to obtain reliable 13C,1H correlations not detectable in the HMBC experiment demonstrates the utility of this method. The generalized covariance formalism therefore expands the power of covariance NMR to the reconstruction of non-symmetric spectra.</p><!><p>Unsymmetric indirect covariance NMR5-8 takes an N1,1 × N2 2D spectrum X1 (matrix) and an N1,2 × N2 2D spectrum X2 and 'concatenates' them into a single N1,1 × N1,2 spectrum C via matrix multiplication:(1)C=X1⋅X2TMatrix element Cij of C is a measure of the correlation between the pair (i,j) of spins belonging to the ith row vector of X1 and the jth row vector of X2. Such a correlation either indicates a direct interaction between the two spins, a mutual correlation to a common 3rd spin, e.g. via spin-diffusion in NOESY spectra,18,21 or a pseudo-relay effect due to correlations to different spins with identical chemical shift. In the symmetric case, i.e. X1 = X2, extraction of the matrix-square root effectively reduces both relay and pseudo-relay effects.18,19,22</p><p>Generalized (indirect) covariance (GIC) NMR provides a framework in which unsymmetric covariance spectra are embedded in symmetric covariance spectra amenable to general matrix functions. GIC starts out with the construction of a stacked spectrum from n 2D spectra of dimensions N1i × N2 (i = 1,…,n):(2)S=[X1⋮Xn]A generalized covariance matrix is then defined as(3)C=S⋅ST=[X1⋮Xn]⋅[X1T⋯XnT]=[X1X1T⋯X1XnT⋮⋱⋮XnX1T⋯XnXnT]Because of Parseval's theorem, Eq. (3) yields (up to a constant prefactor) the same result irrespective whether the direct dimensions of X1,…,Xn are in the time domain or in the frequency domain.18 Matrix C is symmetric and semi-positive definite, which permits the straightforward calculation of arbitrary matrix functions, including matrix roots. For n=1, Eq. (3) reduces to the indirect covariance NMR spectrum.4 For n ≥ 2, C contains the unsymmetric covariance matrix given in Eq. (1) as an off-diagonal submatrix. For simplicity, the GIC spectrum from X1 and X2 (n = 2) is denoted by X1*X2 and, when raised to the matrix power λ, by [X1*X2]λ.</p><p>After application of singular value decomposition (SVD) to matrix S of Eq. (2), S = U·D· VT , where U and V are orthogonal matrices and D is diagonal, Eq. (3) becomes(4)C=S⋅ST=(U⋅D⋅VT)(V⋅D⋅UT)=(U⋅D2⋅UT)For the matrix-square root, λ = ½, it follows C0.5 = U·D·UT and for general powers(5)Cλ=U⋅D2λ⋅UTOf practical importance, calculation of a series of spectra with different powers λ of C only requires a single SVD, which makes such calculations efficient.</p><p>The unsymmetric covariance matrix given by Eq. (1) constitutes an off-diagonal submatrix of the generalized covariance matrix C of Eq. (3). The same submatrix of Cλ defines the λth power of the unsymmetric covariance matrix including the matrix-square root of an unsymmetric covariance matrix.</p><p>GIC is applicable to a stack of spectra, X1,…,Xn, as long as each combination of covariance spectra X1X1T,X1X2T,…, gives rise to non-diagonal blocks and thereby expands the block-diagonal parts stemming from the "auto-covariances" XiXiT. GIC can reconstruct any spectrum that factors into individually measurable NMR experiments. For example, a [13C-1H-HMBC*1H-1H-TOCSY]λ covariance spectrum reconstructs a 2D 13C-1H HMBC-TOCSY spectrum while [13C-1H-HMBC*15N-1H-HSQC]λ yields a 2D through-bond 13C-15N correlation spectrum.23 Experiments probing spin-diffusion, relay, or multi-spin correlation effects (NOESY, TOCSY, HMBC) are particularly suitable for GIC analysis due to the analogy between the matrix (square) root operation of covariance NMR and the shortening of the experimental mixing time.18</p><p>In symmetric covariance, the matrix-square root minimizes artifacts due to pseudo-relay effects.18,19,22 Likewise, the square root of the generalized covariance matrix suppresses artifacts in sub-matrices belonging to the unsymmetric covariance spectra. Hence, the intensities of pseudo-relay correlation peaks are systematically weakened by the root operation as compared to the intensities of bona fide signals. Generally, the more rapidly the covariance cross-peak intensity Cij(λ) increases with λ, the less likely is that peak to be a valid signal. Hence, the slope of log[Cij(λ)] as a function of λ serves as a useful metric by complementing signal intensity alone for assessing the veracity of the signal for matrix element (i,j).</p><p>Eq. (5) may be rewritten in terms of matrix elements (where Dk denotes the kth singular value and Uik the ith component of the kth singular vector)(7)Cij(λ)=∑kUik⋅Dk2λ⋅UjkThus the slope of the natural log[Cij(λ)] is(8)∂log[Cij(λ)]∂λ=2Cij(λ)=∑kUik⋅Dk2λ⋅logDk⋅UjkNote that the plot of log[Cij(λ)] is typically a straight line (Fig. 2) and thus the slope given by Eq. (8) is constant over a broad range of λ values.</p><!><p>2D 1H-1H-TOCSY10 (90 ms mixing time using MLEV-17 24) and 13C-1H-HMBC spectra17 were recorded at 18.8 T and 298 K for a mixture of seven common metabolites at natural 13C abundance (D-carnitine, D-glucose, L-glutamine, L-histidine, L-lysine, myo-inositol, and shikimic acid) each at a concentration of 10 mM in D2O. The direct 1H dimension of each spectrum was acquired with 2048 complex points and a spectral width of 8013 Hz. The indirect 1H dimension of the TOCSY was acquired with 1024 complex points and the same spectral with as the direct dimension. The indirect 13C dimensions of the HMBC spectrum was acquired with 1024 complex points and a spectral width of 32206 Hz, respectively.</p><p>Additionally, 2D 1H-1H-TOCSY (50 ms mixing time using DIPSI-2 25) and 13C-1H HMBC spectra were also recorded at 298 K using a sample of the MDM2-binding p53 peptide construct with sequence ETFSDLWKLLPEN, described previously.26 The spectra were acquired with the same spectral widths as above but with half the number of complex points along each dimension, except for the indirect dimension of the TOCSY having only 256 complex points, and with a spectral width of 44643 Hz in the indirect (13C) dimension of the HMBC spectrum.</p><p>All spectra were recorded on a Bruker AVANCE 800 spectrometer equipped with a cryogenic probe and processed in NMRPipe.27 For the HMBC spectra, a magnitude spectrum was calculated after 2D FT.17 All other calculations were performed in Matlab.28</p><!><p>To demonstrate the approach, a generalized indirect covariance (GIC) HMCB*TOCSY spectrum for a 2-component mixture was calculated from a simulated 13C-1H HMBC spectrum (Fig. 1A) and 1H-1H TOCSY spectrum (Fig. 1B) with sharp lines. The mixture consists of two molecules represented by 2 different spin systems: the first has 3 linked 13C,1H pairs X-Y-Z and the second has 2 pairs U-V. To simulate the effects of overlap, the protons of pairs Y and U are assigned degenerate chemical shifts. Related models with different degenerate chemical shifts were explored, but all gave results similar to those reported here. λ = 1 gives rise to a false peak in the generalized indirect covariance spectrum between CX-HV as indicated in (Fig. 1C).</p><p>Figure 2A shows the suppression of the false positive CX-HV peak (red) achieved by varying the exponent λ in Eq. (5). This log-linear plot demonstrates the higher slope (Eq. (8)) associated with the false positive signal (red) relative to the true signals (black).</p><p>Figure 2B shows the analogous plot for a GIC HMBC*TOCSY spectrum derived from experimental 13C-1H-HMBC and 1H-1H TOCSY spectra of a metabolite mixture sample. The false positive signal, which incorrectly correlates a 13C resonance of myoinositol to a 1H resonance of carnitine, exhibits a systematically stronger λ scaling compared to the true positive signals. Its intensity in the λ = 1 covariance matrix lies between the intensities of two true positive signals, a glucose cross-peak and a myoinositol cross-peak, but when λ = 0.5, its intensity is only as high as the weaker of the two true signals and the slope of its intensity build up as a function of λ is higher than the slope of the true signals. The higher slope and weaker intensity at λ = 0.5 provide a signature that this peak is a false positive.</p><p>Fig. 3 demonstrates the preferential suppression of artifact signals via the matrix-square root in two GIC HMBC*TOCSY covariance spectra calculated from two experimental pairs of 13C-1H-HMBC and 1H-1H TOCSY spectra recorded of the metabolite mixture (Fig. 3A,B) and the p53 peptide (Fig. 3C,D). Peak intensity better separates false peaks (red dots) from true peaks (black dots) in the λ = 0.5 spectrum than in the λ = 1 spectrum (Fig. 3A,C and Table 1). However, while intensity in the λ = 1 spectrum alone is a relatively poor indicator of peak veracity, deviations from the trend visible amongst the true peaks in Fig. 3A,C are indicative of peak authenticity: peaks lying on the upper left hand side of the distribution marked by the ellipse, i.e. peaks for which the matrix-square root reduces peak intensity by a large amount, are most likely to be false.</p><p>Plotting the slope (Eq. (8)) versus the intensity at λ = 0.5 also separates true from false peaks (Fig. 3B,D). Peaks characterized by especially high slopes relative to their intensity (above and to the left of the ellipse surrounding most peaks) are most likely to be false. In fact, plotting the slope versus the intensity at λ = 0.5 identifies false peaks more effectively than does plotting intensity at λ = 1 versus that at λ = 0.5.</p><p>The selection procedure can be formalized by applying principal component analysis (PCA) in two dimensions,29 which in good approximation reproduces the ellipses drawn in Fig. 3. The major axis of the ellipse is given by the first principal component and the minor axis by the second principal component. PCA transforms intensity and slope into a new variable pair of independent statistics that is a linear combination of the original pair. The first principal component adjusts peak intensity using slope information, while the second component combines intensity and slope information into a measure of peak quality. Under the assumption that the principal components are Gaussian distributed, the value for the second principal component calculated for a given peak can be transformed into a p-value that quantifies the probability that this peak is real rather than an artifact arising from spurious chemical shift degeneracy.</p><p>The following procedure allows one to edit peaks picked from a GIC derived spectrum: i) perform PCA as described above on (only) the peaks picked in the λ = 0.5 spectrum, ii) reject peaks for which the p-value calculated (as in a one-tailed test) from the second principal component is less than 5%. Application of this procedure cuts the false-positive rate (reported for the λ = 0.5 spectra in Table 1) in half while only rejecting one (p53 peptide) and two (metabolite mixture) true peaks. The peaks plotted in Fig. 3 include only those peaks reported in Table 1 whose line shapes do not qualitatively change as a function of λ as illustrated in Fig. 4. This figure shows a region of the metabolite mixture GIC [HMBC*TOCSY]λ spectrum for different λ values. The unsymmetric covariance spectrum (λ = 1) displays a noise ridge (cross-hatched box) 16 due to the covariance of a signal arising from the carnitine methyl groups with noise. This ridge is suppressed after application of the matrix roots using the GIC formalism.</p><p>The decrease in intensity with decreasing λ for the false positive is again much more pronounced than for the other peaks: relative to the other peaks in panel A, peak (3) is quite strong whereas it is weak relative to the other peaks in panel C and negative in panel D. The slope given by Eq. (8) at λ = 0.5 for this peak is 52 while a slope of 45 is typical for this data set. This peak appears in the upper left of Fig. 3B (encircled in red) outside of the ellipse surrounding true peaks. Due to its high slope and low intensity at λ = 0.5, this peak can be easily identified and eliminated improving the analysis of the GIC HMBC*TOCSY spectrum.</p><p>Application of λ ≤ 0.5 also recovers the splitting present in the direct dimensions of the HMBC and TOCSY spectra of this mixture, which is lost by covariation of the direct dimension in the unsymmetric covariance process. However, the onset of distortions in line-shape (e.g. peak 2 in Fig. 3D) and signal reduction generally preclude the use of very low λ values (λ ≤ 0.25).</p><p>Fig. 5 shows a region of the GIC HMBC*TOCSY spectrum of the p53-peptide. Again, the matrix-square root suppresses a false positive peak and a ridge, demonstrating the applicability of generalized covariance to larger systems, such as peptides. Unlike an experimentally recorded HSQC-TOCSY, the GIC HMBC*TOCSY exhibits correlations connecting quaternary and other non-protonated carbons, such as carbonyl and carboxyl carbons as illustrated in Fig. 6. Thus, GIC provides a powerful representation of spectral information for the resonance assignment of small and large molecules, including peptides.</p><!><p>Many informative spin correlations are not directly accessible by experiment by multidimensional NMR due to measurement and sensitivity considerations. For instance, correlations between insensitive nuclei can often be observed only indirectly, i.e. via correlations between those nuclei via protons. Other spectra, such as heteronuclear NOESY and TOCSY, which contain useful information for resonance assignment and structure determination of complex molecules, are often not collected due to limited sensitivity and spectrometer time constraints. However, unsymmetric covariance NMR can reconstruct heteronuclear TOCSY and NOESY spectra from homonuclear NOESY and TOCSY spectra and common heteronuclear 13C-1H HSQC or HMBC spectra.7</p><p>Similarly, the high-dimensional correlation information required to make chemical shift assignments in polypeptides can often only be practically measured by a series of lower dimensional spectra. A typical manual analysis of NMR spectra establishes higher order correlations via a comparison of strip plots. Visual assessment of a non-vanishing correlation of peaks between slices (strip plots) in two NMR spectra links the spin-systems associated with the strip plots being compared. Automated analysis methods, particularly those for protein backbone assignment,30-37 often work with peak lists rather than with the underlying spectra. However, such methods generally require high quality peak lists that are manually curated. Recently developed methods such as hyperdimensional NMR,11,12 COBRA13 and Burrow-Owl15 use unsymmetric covariance5,7 to automate the traditional manual approach of establishing spin correlations via comparison of strip plots, prior to peak picking. However, the application of such methods can confound downstream analysis due to the presence of spurious correlations between strip plots caused by (near-)degenerate chemical shifts and therefore may benefit from the generalized indirect covariance approach presented here. GIC establishes correlations between spectra rather than peak lists and thereby 'delays' the otherwise iterative and sometimes difficult process of peak picking until true peaks become self-evident.</p><p>The GIC formalism generalizes the use of the matrix-square root for the suppression of relay effects and pseudo-relay effects, originally demonstrated for symmetric covariance NMR spectra,18,19 to unsymmetric covariance spectra.6 Previous work in covariance reconstruction of unsymmetric spectra compared unsymmetric and indirect covariance results in order to identify artifacts in each.20 The generalized covariance matrix (Eq. (3)) presented here computes both unsymmetric and symmetric covariance spectra in the same step. Furthermore, the GIC formalism allows for the extraction of multiple roots in a single covariance calculation. For the examples used here, extraction of the square root via the generalized covariance matrix reduces the false positive count of a HMBC*TOCSY spectrum by about a factor of three. Removal of peaks characterized by weak intensity following extraction of the square root concomitant with a rapid intensity build up with λ further reduces the false positive rate.</p><p>The generalized covariance formalism addresses the issue of false positives in unsymmetric covariance spectra caused by resonance overlap and extends the applicability of unsymmetric covariance NMR to systems with an increased number of signals of greater resonance degeneracy, including complex mixtures, for example of metabolites, and biological macromolecules, such as peptides and proteins. By providing a mechanism to identify false positive correlations, generalized indirect covariance lays a linear-algebraic foundation for the accurate and sensitive identification of spin correlations that are distributed over multiple 2D NMR spectra. The establishment of spin correlations that are not easily experimentally observable via an automated method analogous to the comparison of strip plots, mark a path toward the development of computer-based assignment procedures that are as robust as are the most expert manual analyses of NMR data.</p>

PubMed Author Manuscript

Calcium(<scp>ii</scp>)-catalyzed enantioselective conjugate additions of amines

The direct enantioselective chiral calcium(II)$phosphate complex (Ca[CPA] 2 )-catalyzed conjugate addition of unprotected alkyl amines to maleimides was developed. This mild catalytic system represents a significant advance towards the general convergent asymmetric amination of a,b-unsaturated electrophiles, providing medicinally relevant chiral aminosuccinimide products in high yields and enantioselectivities. Furthermore, the catalyst can be reused directly from a previously chromatographed reaction and still maintain both high yield and selectivity.

calcium(<scp>ii</scp>)-catalyzed_enantioselective_conjugate_additions_of_amines

<!>Conclusions

<p>Chiral amines are a ubiquitous motif in pharmaceuticals and natural products (Fig. 1). 1 The conjugate addition of amine nucleophiles to various a,b-unsaturated systems is a wellestablished transformation to access the corresponding bamino carbonyl products. 2 However, catalytic enantioselective methods for the construction of C-N bonds directly from amines remain a challenge in synthetic organic chemistry. Direct conjugate additions of amines with a,b-unsaturated electrophiles have been shown to proceed at high temperatures and pressures; 3 however the reversibility of the initial attack by the amine eventually leads to racemic products (Fig. 1A). 4 Stoichiometric homochiral lithium amides can be successfully deployed under kinetic control, achieving high yield and selectivity. However, these sensitive, strongly basic reagents are further limited by the need to remove the chiral a-methylbenzyl moiety to carry the products forward to useful targets. 5 To circumvent these issues, current catalytic methods have relied upon the use of non-basic nitrogen nucleophiles as amine surrogates to avoid catalyst poisoning (Fig. 1B), 6 which is common when basic amines are used as reagents in the presence of chiral Lewis or Brønsted acidic catalysts. 7 Therefore, numerous examples of non-basic nitrogen nucleophiles including azides, 8 hydroxylamines, 9 Ofunctionalized carbamates, 10 1,2,4-triazoles, 11 indoles, 12 and anilines 13 have been strategically deployed to avoid Lewis acid complexation, 14 Brønsted acid neutralization, or unselective iminium activation. 10a However, in all of these cases, a protected nitrogen atom is installed which requires multiple steps to elaborate further. Thus, a more convergent approach would be enabled by the direct asymmetric amination of basic primary and secondary amines without the use of protecting groups. Apparently, there are only three examples of catalytic asymmetric amino-conjugate additions that have successfully employed alkyl amines. 15 In 2003, Togni briey explored asymmetric amino-conjugate additions to activated olens as the initial step in a catalytic asymmetric hydroamination reaction catalysed by a novel chiral Ni(II) phosphine complex. 15b Morpholine and piperidine produced modestly enantioenriched products when reacted with methacrylonitrile (69% and 20% ee, respectively), which represents the rst signicant example of an effective, enantioselective intermolecular hydroamination reaction employing alkyl amines. Despite this promising proof-of-concept study, general asymmetric aminoconjugate additions with unfunctionalized/masked amines remain unrealized, which underscores the fundamental challenge associated with the use of highly basic and sterically unencumbered reagents in conjunction with Lewis acidic metal catalysts. In 2015, Huang and co-workers reported an efficient, highly enantioselective conjugate addition of primary alkyl amines to activated b-aryl b-triuoromethyl nitroolens. 15c Unlike most catalytic examples, their strategy uses chiral Brønsted base catalysis. 16 However, a major limitation of this method is the lack of secondary amines as nucleophiles. Additionally, strongly basic and cryogenic conditions are required, which potentially limit the generality of this transformation. 15c Therefore, we sought mild catalytic conditions capable of providing enantioenriched amino-conjugate addition products from a general set of readily available alkyl amines with maleimides, which were chosen as an ideal substrate for catalyst identication and optimization due to their ready availability and excellent conjugate acceptor properties (Fig. 1C). Additionally, enantioenriched aminosuccinimide products serve as an easily functionalized scaffold to generate aminolactams and aminopyrrolidines. 17 Aminosuccinimides and their derivatives are also a common motif in bioactive small molecules, pharmaceuticals, and natural products (Fig. 1D). 18 Access to these products from achiral starting materials can facilitate the rapid generation of diverse small molecule libraries aimed at probing new chemical spaces.</p><p>We began our studies with a reaction between equimolar quantities of N-benzylmaleimide and p-tolylamine. Our primary focus was to enhance the enantioselectivity of the title reaction (Table 1). An initial exhaustive screen of various asymmetric catalyst families including hydrogen bond donors (HBD), metal-TADDOL complexes, metal BINOL-complexes, and chiral phosphoric acids (CPA), identied that CPA A-H possessing 1napthyl substitution at the 3,3 0 -positions has the capability to produce the title compound in modest yield and selectivity (entry 1). Subsequently, we investigated the role of water in the reaction and observed that the addition of 4 Å MS had a moderate but reproducible impact on selectivity (entry 2). We then investigated a wide range of desiccants 9b and found that calcium oxide had a greater than anticipated positive effect on the selectivity of the reaction (entry 3). 19 Additionally, we observed a moderate increase in e.r. over time to 80 : 20 e.r. (entry 4). We therefore hypothesized that the reaction of calcium oxide with A-H led to the formation of a more enantioselective calcium phosphate catalyst. Our hypothesis was enlightened by combining the prior elegant work of Ishihara, Antilla, and Rueping who demonstrated the role of catalytic chiral alkali metal and alkaline earth metal-phosphate salts in various reactions. 20 Thus, we investigated two pre-formed calcium phosphate complexes (entries 5 & 6) and observed that the calcium CPA complex possessing 9-phenanthracenyl substitution on the phosphate 3,3 0 -positions, Ca[B] 2 (Table 1), facilitated the title reaction in 76% yield and 95 : 5 e.r. (entry 6). Strikingly, removal of the 4 Å MS diminished both yield and selectivity (entry 7). Aer investigating selectivity as a function of temperature (entries 8 & 9), we looked at other CPA salts (entries 10-12) and determined that Ca[B] 2 was indeed optimal. We then compared calcium and magnesium phosphate complexes, and demonstrated again that Ca[B] 2 was optimal (entry 13). Furthermore, increasing its concentration to 0.05 M and lowering the catalyst loading to 5 mol% increased the yield to 95% with 94 : 6 e.r. (entries 14 & 15).</p><p>With the optimized conditions in hand, we next investigated the scope of the reaction with a range of aliphatic amines and maleimides (Table 2). Para-substituted primary benzylamines with a range of electron donating and withdrawing groups afforded the conjugate addition products (6-11) in 93 : 7-94 : 6 e.r. and 77-91% yield. Meta-and ortho-substituted benzyl amines afforded 12 and 13 in similar yields and selectivities. The products derived from less sterically bulky amines and linear amines were obtained with lower enantioselectivity (14-17) and moderate yields. In contrast, bulkier amines gave products 18 and 19 in high yield and selectivity. Notably, secondary cyclic amines provided conjugated products 20-24 in 93 : 7-97 : 3 e.r. These substrates would be difficult to access via other methodologies or from an enantiopure amino acid derived starting material. 21 The enantioselectivity for the arylpiperidine-derived 24 uniquely improved at À40 C which was not general for the other substrates.</p><p>Acyclic secondary amines showed the lowest selectivity among the nucleophiles (25). Also, substitutions on the benzyl maleimide were tolerated (26-28). The cross-reaction between piperidine and a substituted benzyl maleimide generated product 29 in good yield and selectivity. N-Phenyl maleimide was a poor substrate with regard to selectivity (74 : 26 e.r.); however, the desired 1,4-addition product 30 was synthesized in 93% yield with no observed 1,2-addition product (a common side-reaction with N-aryl maleimides). 22 Maleimide substrates with smaller appendages were observed to react with lower selectivities (31). The unsubstituted maleimide product 32 was not observed, presumably due to a lack of solubility.</p><p>The title reaction was successfully scaled up by 1000-fold from the initial screening conditions (Scheme 1). Taking into account the observed dependence of enantioselectivity on concentration, the amine nucleophile was added slowly to the other reaction components via a cannula. These conditions afforded 7.15 g of the product (93% yield) in 94 : 6 e.r. The product was successfully recrystallized to >99 : 1 e.r. Additionally, >95% of the catalyst Ca[B] 2 was recovered via column purication. The recovered Ca [B] 2 was subsequently able to reproduce the title reaction without loss of yield or selectivity. The ability to directly recover and reuse Ca[B] 2 from each reaction at >95% efficiency gives this methodology more utility, especially given the high molecular weight of the catalyst.</p><p>In an effort to rationalize the observed enantioselectivity, we obtained X-ray crystal structure and 31 P NMR spectroscopy data for the pre-formed calcium phosphate complex used in our optimization and scope studies (Fig. 2B). 23 Surprisingly, the observed structure shows a 4 : 2 ratio of B to Ca 2+ , not a Ca[B] 2 complex. Additionally, both calcium atoms are coordinatively saturated, with each cation bound to ve molecules of water, which creates a hydrogen-bonding network. Although it is possible that the observed ORTEP structure is the actual catalytic species, we hypothesize that it is more likely a precatalyst is activated in the presence of molecular sieves. This observation is supported by the signicant change in the 31 P NMR spectrum in the presence of 4 Å MS (Fig. 2B). The yield and selectivity of the reaction also diminished in the absence of the 4 Å MS (Table 1, entry 7), which supports that dehydration of the Ca 2 [B] 4 $(H 2 O) 10 complex is necessary. Interestingly, when all the reaction components are present, the 31 P NMR data is reminiscent of the precatalyst (Fig. 2B). This data indicates that the presence of the amine re-establishes the hydrogen-bonding network that is lost upon dehydration of the Ca 2 [B] 4 $(H 2 O) 10 complex. Understanding this Lewis base/Lewis acid interaction between the active catalyst and a coordinated maleimide substrate will require further investigation.</p><p>Based on the obtained spectroscopic data, we hypothesize that the Ca 2 [B] 4 $(H 2 O) 10 complex is activated via dehydration in the presence of 4 Å MS, inducing it to reorganize to form Ca[B] 2 complex 33 (observed by HRMS, see ESI ‡). The extent of dehydration of Ca 2 [B] 4 $(H 2 O) 10 required to form the active catalyst cannot be quantied by these experiments; however, it is reasonable to postulate that the loss of some coordinating water ligands from the Ca 2 [B] 4 $(H 2 O) 10 complex should open up Lewis acidic sites on the calcium atom, which are then able to coordinate the amine nucleophile. Based on structure 33, we propose a model for enantioselectivity, where the si-face of maleimide 5 is blocked, which allows the re-face attack of the amine nucleophile (Fig. 2C).</p><p>Aer exploring the scope of our conjugate addition with a variety of amines and maleimides, we also applied our methodology to the synthesis of 35, a potent novel 5-HT 2A agonist developed by Acadia Pharmaceuticals (Scheme 2). 18c Since the binding affinity of 35 was measured as a racemic mixture, we envisioned that our methodology could readily determine the more active enantiomer. Starting from recrystallized 6, lithium aluminium hydride reduction cleanly produced 34 in 95% yield and >99 : 1 e.r. (Scheme 2). Selective acylation of 34 with 4-methoxyphenylacetic acid produced 35 in 43% yield and >99 : 1 e.r. To further demonstrate the utility of this methodology, we selectively removed the benzylic group on the amine (36) as well as selectively deoxygenate the position adjacent to the amine (37).</p><!><p>In summary, we have discovered an efficient and scalable catalytic asymmetric conjugate addition of unmasked and unfunctionalized amines to maleimides. This process accommodates both primary and secondary amines, which underscores the unusual compatibility of these Lewis basic nucleophiles with the Lewis acidic Ca 2+ complex. Crystallographic studies indicate an initial Ca 2 [B] 4 species is formed through the reaction of a chiral phosphoric acid and calcium(II) methoxide. Further spectroscopic studies indicate that a dynamic process is involved, where molecular sieves are required for the observed reactivity and selectivity, which are thought to play a role in the activation of the catalyst. The addition of amine nucleophiles can re-establish a hydrogenbonding network similar to that found in the hydrated Ca 2 [B] 4 $(H 2 O) 10 complex. Furthermore, although the calcium phosphate catalyst Ca[B] 2 has a relatively high molecular weight, it can be effectively recovered in >95% yield. Future investigations involve continued analysis of the calciumphosphate dynamics and applications of this reaction in the synthesis of bioactive compounds.</p>

Royal Society of Chemistry (RSC)

Neuroprotective Action of Multitarget 7-Aminophenanthridin-6(5H)-one Derivatives against Metal-Induced Cell Death and Oxidative Stress in SN56 Cells

Neurodegenerative diseases have been associated with brain metal accumulation, which produces oxidative stress (OS), matrix metalloproteinases (MMPs) induction, and neuronal cell death. Several metals have been reported to downregulate both the nuclear factor erythroid 2-related factor 2 (Nrf2) pathway and the antioxidant enzymes regulated by it, mediating OS induction and neurodegeneration. Among a recently discovered family of multitarget 7-amino-phenanthridin-6-one derivatives (APH) the most promising compounds were tested against metal-induced cell death and OS in SN56 cells. These compounds, designed to have chelating activity, are known to inhibit some MMPs and to present antioxidant and neuroprotective effects against hydrogen peroxide treatment to SN56 neuronal cells. However, the mechanisms that mediate this protective effect are not fully understood. The obtained results show that compounds APH1, APH2, APH3, APH4, and APH5 were only able to chelate iron and copper ions among all metals studied and that APH3, APH4, and APH5 were also able to chelate mercury ion. However, none of them was able to chelate zinc, cadmium, and aluminum, thus exhibiting selective chelating activity that can be partly responsible for their neuroprotective action. Otherwise, our results indicate that their antioxidant effect is mediated through induction of the Nrf2 pathway that leads to overexpression of antioxidant enzymes. Finally, these compounds exhibited neuroprotective effects, reversing partially or completely the cytotoxic effects induced by the metals studied depending on the compound used. APH4 was the most effective and safe compound.

neuroprotective_action_of_multitarget_7-aminophenanthridin-6(5h)-one_derivatives_against_metal-induc

Introduction<!>Chemistry<!>Synthesis of the Target Compounds APH<!>Metal-Chelating Properties of Compounds APH1–APH5<!><!>Metal-Chelating Properties of Compounds APH1–APH5<!><!>Metal-Chelating Properties of Compounds APH1–APH5<!><!>Metal-Chelating Properties of Compounds APH1–APH5<!>Highest Occupied Molecular Orbital and Lowest Unoccupied Molecular Orbital analysis<!><!>Highest Occupied Molecular Orbital and Lowest Unoccupied Molecular Orbital analysis<!>Computational Prediction of Physicochemical and Absorption, Distribution, Metabolism, Excretion, and Toxicological Properties of APH Compounds<!><!>Computational Prediction of Physicochemical and Absorption, Distribution, Metabolism, Excretion, and Toxicological Properties of APH Compounds<!>Neuroprotective Activity of Compounds APH1–APH5 against Metal-Induced Cell Viability Reduction<!><!>Neuroprotective Activity of Compounds APH1–APH5 against Metal-Induced Cell Viability Reduction<!>NRF2 Pathway Induction by Compounds APH1–APH5<!><!>NRF2 Pathway Induction by Compounds APH1–APH5<!>Antioxidant Activity of Compounds APH1–APH5 against Metal-Induced Lipid Peroxidation<!><!>Antioxidant Activity of Compounds APH1–APH5 against Metal-Induced Lipid Peroxidation<!>Conclusions<!>General Experimental Information<!>Synthesis of 1<!>Synthesis of 2<!>Synthesis of APH<!>7-(Butylamino)-9-phenylphenanthridin-6(5H)-one (APH1)<!>7-((4-Methoxyphenyl)amino)-9-phenylphenanthridin-6(5H)-one (APH2)<!>7-((4-Dimethylaminophenyl)amino)-9-phenylphenanthridin-6(5H)-one (APH3)<!>7-((4-Methoxyphenyl)amino)-9-(thiophen-2-yl)phenanthridin-6(5H)-one (APH4)<!>7-(Phenylamino)-9-(thiophen-2-yl)phenanthridin-6(5H)-one (APH5)<!>Metal-Chelating Study<!>HOMO and LUMO Analysis<!>Physicochemical and Prediction of ADMET Properties Analysis<!>Culture of SN56 Cells<!>Cytotoxic and Neuroprotective Effect on SN56 Cells<!>Real-Time PCR Analysis<!>Protein Expression Analysis<!>Protective Effects on Cell Lipid Peroxidation<!>Statistical Analysis<!><!>Author Contributions<!>

<p>Several neurodegenerative disorders including Alzheimer's disease (AD), Parkinson's disease (PD), Huntington's disease (HD), and amyotrophic lateral sclerosis (ALS), among others, are undergoing a worrying increase in incidence worldwide. Although the ultimate causes of these diseases remain unknown, the possible involvement of environmental factors in their development has been suggested.1−3 The rise in human industrial activity has exponentially increased the population's exposure to heavy metals and their body accumulation, leading to toxic effects.1−4 Brain accumulation of heavy metals, such as aluminum (Al), cadmium (Cd), and mercury (Hg) due to overexposure, or essential metals, including iron (Fe), copper (Cu) and zinc (Zn) due to pathological circumstances, has been correlated with the development of neurodegenerative diseases.1−8 Accumulation of metals in the brain was related with the induction of oxidative stress generation.1,2 Oxidative stress, which results from the mismatch between reactive oxygen species (ROS), reactive nitrogen species (RNS), and antioxidant defense can cause oxidative damage to cell structures and cell death.1,2 The central nervous system (CNS) is the most metabolically active organ, resulting in a high production of superoxide radical anion (O2–), hydroxyl radical (OH•), and hydrogen peroxide (H2O2), thus being particularly susceptible to oxidative stress and ROS damage.1,2 Furthermore, ROS generation was reported to mediate the brain accumulation of oxidized, misfolded, and aggregated proteins such as tau-phosphorylated (p-tau), α-synuclein (α-syn), and Aβ peptides. After their accumulation, they induce neurofibrillary tangles, amyloid plaques, and Lewy bodies, the main hallmarks of AD and PD, neuroinflammation, mitochondrial dysfunction, and finally neuronal cell death.1,2</p><p>Calcium and zinc regulate the activity of matrix metalloproteinases (MMPs), which control a variety of physiological processes. Under pathological conditions, MMPs overexpression or abnormal expression has been related with the induction of brain damage and with neurodegenerative disease, including AD, PD, HD, and ALS.9 Otherwise, heavy metals were also related to the induction of MMPs, which could lead to cytotoxicity.10 The main MMPs present in the brain are MMP-2/3/9/14, which play a key role in the neuroinflammatory processes and brain damage.11,12 Besides, MMPs dysfunction was related with the production of Aβ, p-tau, α-synuclein, and other abnormal proteins related with neurodegenerative diseases, especially MMP-9.9,13−15 Therefore, MMP inhibitors were tested to improve symptoms of different neurodegenerative diseases and to reduce the toxicity associated with metals.16</p><p>In this context, the control of metal accumulation in the CNS and its pathological effects is an important target in the development of new therapeutic agents against neurodegeneration and neurodegenerative disorders. In this sense, we have previously reported the discovery of new multitarget phenanthridin-6(5H)-one derivatives (APH) that are able to inhibit MMP enzymes, reduce oxidative stress, and protect against ROS.17 Although these compounds were designed to present chelating activity, computational studies indicated binding to a distal region at the S1′ site, with no involvement of the catalytic zinc ion.16 Thus, it is necessary to confirm experimentally the role of the Zn2+ ion in MMPs inhibition by these compounds.</p><p>Furthermore, the neuroprotective effects shown for these compounds against oxidative stress and cell death induced by metals could be mediated through a direct ROS scavenger activity. The induction of cellular antioxidant enzymes, which reduce ROS, may also be involved, since they show neuroprotective and antioxidant action against hydrogen peroxide in SN56 cells.17 Nuclear factor erythroid 2-related factor 2 (Nrf2) is a key factor in the transcription regulation of antioxidant cytoprotective enzymes such as superoxide dismutase 1, catalase, glutathione peroxidase, NAD(P)H quinone oxidoreductase 1, and heme oxygenase-1, among others.18−20 Dysfunction of the Nrf2 pathway has been identified, in different neurodegenerative diseases, as a factor of oxidative stress generation.21−23 Besides, many metals were reported to alter the regulation of the Nrf2 pathway, leading to oxidative stress and cell death induction.24−27 Nrf2 levels are regulated by kelch-like ECH-associated protein 1 (Keap1), which under normal conditions binds to Nrf2 in the cytoplasm, promoting its degradation. However, under oxidative stress, Keap1 is inactivated and releases Nrf2, which is transported to the nucleus, inducing the expression of downstream pathway antioxidant enzymes.19,20 Another mechanism to avoid the oxidative stress, and other harmful actions of metals, is to sequester them with chelator compounds. Thus, as these compounds were designed to present chelating activity, they could also protect against metals toxicity through this mechanism.</p><p>In the context of the treatment of neurodegenerative diseases, it is crucial to find new multitarget drugs that reduce the neurotoxic effects of metals that may avoid the onset and progression of these diseases. In this sense, the use of molecules designed to possess antioxidant action and as metal ionophores to sequester, redistribute, and remove metals in the CNS is an attractive therapy to reduce the neurotoxic effect of metals in the brain. To reach this aim, we studied the chelating activity, the action on Nrf2 pathway, and the neuroprotective action against heavy and essential metals of a small library of 7-amino-phenanthridin-6-one derivatives that had been previously characterized as MMP inhibitors and found to exert the most neuroprotection against oxidative stress.</p><!><p>The target compounds APH1–APH5 were synthesized as shown in Scheme 1 by our previously reported method.17 Thus, 4,6-diaryl-5,6-dihydroanthranylate derivatives 1 were obtained by a CAN-catalyzed multicomponent reaction between ortho-nitrochalcone, ethyl acetoacetate, and the suitable primary amines. This was followed by oxidative dehydrogenation of the central ring in the presence of 2,3-dichloro-5,6-dicyano-1,4-benzoquinone (DDQ) and reductive cyclization to give the target compounds APH1–APH5 in good overall yield.</p><!><p>Reagents and conditions: (i) EtOH, CAN (5%) reflux, 16 h; (ii) DDQ (1.5 equiv), toluene, 2 h at room temperature; (iii) EtOH, Pd/C (10%), H2 1 atm, 2 h at room temperature.</p><!><p>N and O electron-donor atoms and carbonyl group assembled to a cyclic system are considered suitable for efficient interactions with metal cations. These binding moieties are present in some biological relevant compounds such as the tetracycline,28 3-hydroxy flavonoids,29 and anthranilic acid,30 all of which have chelating properties. Thus, we designed the APH compounds as bidentate ligands to interact with the metal anions guest via a potential metal-binding site formed by the lactam group and the exocyclic nitrogen atom, hypothesizing that the substituents R on the nitrogen atom at C-7 position could modulate the affinity of the nitrogen atom for metals (Figure 1). For the current study, we selected compounds APH1–APH5 from our library, due to their higher activity on the selected targets (i.e., MMPs inhibition and antioxidant activity), according to our previous studies.17</p><!><p>Hypothesized APH–metal interaction and some therapeutically useful chelating biomolecules.</p><!><p>The ability of the APH1–APH5 to bind Fe2+, Cu2+, Zn2+, Hg2+, Cd2+, and Al3+ was studied by ultraviolet–visible (UV–vis) spectroscopy. This is one of the most widely used techniques to study chelation and to elucidate the stoichiometry of the formed metal complexes.31−34 To this end, the UV–vis spectra of the compounds alone and in the presence of one of the evaluated metals were compared. Blank solutions composed of dimethyl sulfoxide (DMSO) 100 mM phosphate buffer at pH 7.30 showed high absorption values between 190 and 240 nm, and consequently only absorbances from 240 to 480 nm were considered.</p><p>When FeCl2 or CuCl2 were added to individual solutions containing APH1–APH5, and when HgCl2 was added to solutions containing APH3, APH4, or APH5, new optical bands did not appear, and, therefore, the wavelengths corresponding to the maximum absorption bands of the Fe2+/compound, Cu2+/compound, or Hg2+/compound solutions were similar to those of the compounds alone. By contrast, the analysis of solutions at different Fe2+/compound, Cu2+/compound, and Hg2+/compound molar ratios evidenced important differences regarding the intensity of the absorption bands. The observed changes suggest an interaction between the tested compounds and Fe2+, Cu2+, and Hg2+ and, consequently, the formation of iron, copper, and mercury complexes, respectively. However, when the same experiments were performed with ZnCl2, CdCl2, or AlCl3 and solutions containing APH1–APH5, and when HgCl2 was added to solutions containing APH1 or APH2, no differences between spectra of the different Zn2+/compound, Cd2+/compound, Al3+/compound, and Hg2+/compound molar ratios and the compounds alone were observed. This absence of spectral changes suggests the lack of interaction between the tested compounds and Zn2+, Cd2+, Al3+, and Hg2+.</p><p>Due to the high spectral similarity between the compounds and their Fe2+, Cu2+, Hg2+, Zn2+, Cd2+, or Al3+ complexes, derivative spectroscopy was applied in order to enhance differences among spectra, to resolve overlapping bands, to reduce the effects of other absorbing compounds, and to eliminate the background. Once UV–vis spectra were registered, mathematical methods were used to generate the first-order derivative spectra. Our analysis shows excellent results in enhancing differences among spectra, increasing the resolution of the overlapped bands, reducing the effects of other absorbing compounds, and eliminating the background. In fact, as the first derivative spectrum passes through zero at the same wavelength as λmax of the absorbance band, the spectral differences are highlighted. Thus, the analysis of derivative spectra allowed to confirm, through a resolution enhancement effect, that APH1–APH5 were able to interact with Fe2+ and Cu2+. Moreover, it was verified that APH3, APH4, and APH5 were also able to interact with Hg2+ and, thus, to form iron, copper or mercury complexes. Otherwise, no changes were appreciated in Hg2+-APH1 and Hg2+-APH2 and neither in Zn2+, Cd2+, or Al3+with APH1–APH5 from derivative spectra.</p><p>The stoichiometry of the chelated iron, copper, and mercury compounds were estimated by means of the derivative spectra collected under two different conditions: (i) maintaining constant the metal ion concentration (50 μM) and varying the compound concentration and (ii) maintaining constant the compound concentration (50 μM) and varying the metal concentration. The results showed that the five compounds evaluated presented a similar behavior regarding Fe2+ and Cu2+ chelation. Figure 2a,b shows the UV–vis spectra registered for Fe2+-APH4, as an example of metal-compound interaction. Figure 2a evidenced differences between absorbance spectra of the compound alone and in the presence of FeCl2. Nevertheless, differences among the spectra corresponding to different stoichiometries are not properly appreciated. However, as depicted in Figure 2b, first-order derivative absorption spectra revealed some differences depending on the evaluated stoichiometry. The maximum absorption band at 280 nm (Figure 2a) appeared in all stoichiometries, being associated with the negative band at 290 nm as reflected in the derivative spectra (Figure 2b). This band may correspond to the absorption of the iron complex as it increases, as metal concentration reaches its maximum in Figure 2a. Otherwise, a ligand absorption band was observed at 300 nm, which overlaps with the optical band of the Fe2+ complex. This effect is shown in Figure 2b, where the derivative spectra of the compound alone passed through 0 at 300 nm wavelength, in contrast with the different stoichiometries. By analysis of the first-order derivative spectra (Figure 2b), the stoichiometry of the Fe2+-APH4 complex is consistent with a 1:1 Fe2+/compound molar ratio. However, complexes of different stoichiometry could be simultaneously formed.</p><!><p>UV–vis spectra (a) and first-order derivative absorption spectra (b) of APH4 alone and in the presence of FeCl2 in buffer (100 mM phosphate, pH= 7.30) at room temperature. [APH4] = 50 μM and [Fe2+] = 50, 25, 12.5, and 6.25 μM, corresponding to stoichiometries 1:1, 1:2, 1:4 and 1:8, respectively.</p><!><p>Similar observations were made for Fe2+ and Cu2+ with APH1–APH5 and for Hg2+ with APH4 and APH5. The spectra revealed bands that could be associated with the iron complexes at wavelengths of 280 nm (Fe2+-APH1), 270 and 305 nm (Fe2+-APH2), 280 nm (Fe2+-APH3), and 270 nm (Fe2+-APH5), while the copper complexes showed bands at 275 and 300 nm (Cu2+-APH1), 265 nm (Cu2+-APH2), 260 nm (Cu2+-APH3), 265 nm (Cu2+-APH4), and 270 nm (Cu2+-APH5). Moreover, with the mercury complexes bands appeared at 295 and 300 nm (Hg2+-APH4) and 300 and 305 nm (Hg2+-APH5). Once again the 1:1 metal/compound molar ratio seems to be the main common stoichiometry for the formed complexes. However, for Hg2+-APH3, showing an absorption band at 305 nm, the stoichiometry remains unclear.</p><p>Figure 3a,b shows the Cd2+-APH1 spectra, as an example of an noninteracting metal-compound. There were no differences between compound bands (APH1) and metal-compound bands at different molar ratios (1:1–1:4) in the UV–vis spectra (Figure 3a). These differences were not appreciated in the first derivative spectra (Figure 3b). Similar observations were deduced for Zn2+, Cd2+ and Al3+ with APH1-APH5 and for Hg2+ with APH1 and APH2. The rest of UV–vis and first-order derivative absorption spectra data for APH compounds and the metals studies are shown in Figures S3–S7 in the Supporting Information.</p><!><p>UV–vis spectra (a) and first-order derivative absorption spectra (b) of APH1 alone and in the presence of CdCl2 in buffer (100 mM phosphate, pH= 7.30) at room temperature. [APH1] = 50 μM and [Cd2+] = 50, 25, 16.7, and 12.5 μM corresponding to stoichiometries 1:1, 1:2, 1:3 and 1:4, respectively.</p><!><p>In our previous work, we showed that APHs are able to inhibit MMP enzymes and computational studies suggested that this inhibitory activity is not mediated by the chelation of Zn2+ ion at the catalytic site, as is the case for many other MMP inhibitors.17 Our present results show that APH compounds tested were not able to chelate Zn2+ ion, confirming the computational results. Further analysis should be performed to determine the mode of action behind MMPs inhibition by APHs.</p><!><p>The obtained results also show that all five APH compounds are able to bind to Fe2+ and Cu2+, while they are unable to bind to bigger cations such as Cd2+, Al3+, and Hg2+. The Hg2+ binding capacity of APH3–APH5 is probably due to coordination with the nitrogen atom of dimethylamino group for APH3 and with the thiophen sulfur of APH4 and APH5. Furthermore, Zn2+ and Cd2+ ions usually form five-membered complexes with bidentate species, while Fe2+ and Cu2+ preferentially lead to the formation of six-membered metallocycles, consistent with the observed results.</p><p>To corroborate our experimental results of the chelating capacity of APH compounds and to relate these results with the compound structures, we studied highest occupied molecular orbital (HOMO) and lowest unoccupied molecular orbital (LUMO) orbital energies and energy gaps for the APH compounds and for the corresponding APH-Fe complexes (Table S1). Iron was chosen since, together with copper, it was chelated by all APH compounds. Our results for compounds APH1–APH5 show that the proposed metal binding site was accessible (without impediment from the substituent of the exocyclic nitrogen in the most stable conformation calculated) for the five compounds (Figure S1). It was also observed that the HOMO orbital is mainly extended on the alkyl or the π system at the C-7 position. Particularly, for APH3, the charge density is concentrated on the dimethylamino group. Finally, the HOMO–LUMO energy gap (7.049, 7.799, 4.494, 5.982, and 6.643 eV for APH1–APH5, respectively), calculated at Merck molecular force field 94 (MMFF94) level, is compatible with complex formation.</p><p>The HOMO and LUMO levels for the APH compounds complexed with iron showed a reduction of gap energy of 0.917, 0.712, 3.922, 0.145, 0.299 eV for APH1-Fe to APH5-Fe complexes, respectively (Table S1), indicating a high stability for the complexes. The highest reduction of gap energy was observed in AHP4-Fe, showing that this compound is the most stable (Table S1) and therefore that AHP4 has the highest chelating activity, a prediction that correlates well with the fact that this compound presented the best neuroprotective activity as shown below. Taking APH4 as an example, it is apparent that the exocyclic nitrogen gives a higher contribution to the HOMO levels of APH4 than to the HOMO levels of the APH4-Fe complex (Figure 4a,c), which indicates that the nitrogen atom depletes charge from the Fe2+ ions. Additionally, the presence of an aromatic amine at C-7 seems to stabilize the iron complex in comparison to the alkyl amine (see APH1-Fe and APH2-Fe in Table S1), and the amine group on the aromatic ring seems to destabilize the complex compared to the methoxy group (see APH3-Fe in Table S1). Finally, the presence of a thiophene at the C-9 position appears to lead to a smaller gap than the presence of a benzene ring (compare APH4-Fe and APH2-Fe in Table S1).</p><!><p>Optimized molecular structure of the APH4 and APH4-Fe complex (upper) and their corresponding HOMOs (A and C) and LUMOs (B and D). Color scheme: Carbon is gray, hydrogen is white, nitrogen is blue, oxygen is red, and iron is brown. HOMOs and LUMOs are shown in red and blue.</p><!><p>Therefore, although the number of compounds studied is somewhat limited to conclude a structure-complex formation ability relationship, it could be confirmed that their ability to interact with Fe2+ and Cu2+ does not depend on the substituent of 7-amino phenanthridi-6-one skeleton. The computational analysis also revealed that phenyl moieties on the exocyclic nitrogen by the iron complex stabilizes the HOMO charge. Thus, the chelating cavity is suitable for small metal cations.</p><!><p>According to the ADMETsar version 2.0 and swissADME software analysis, as shown in Table 1, the APH compounds satisfied all the values required except for the lipophilicity values,35 which are slightly above the Lipinski limit.</p><!><p>Prediction with ADMETsar version 2.</p><p>Prediction with Swiss ADME.</p><!><p>Based on the results shown in Table S2, it can be assumed that APH compounds have favorable intestinal absorption and Caco2 cell permeability. The CNS distribution properties of our molecules was estimated through the calculation of the brain distribution parameter, which can evaluate the ability of small molecules to cross the blood–brain barrier (BBB) and be distributed into the brain. The predictions of the two methods used were fully consistent for APH1, where both predicted a good BBB penetration ability. For APH2, APH3, APH4 and APH5, ADMETsar version 2.0 showed good BBB penetration data, but swissADME gave borderline negative results. The metabolic properties of these compounds were estimated according to the predicted inhibitory capacity of cytochrome P450 isoforms Cyp1a2, Cyp2c19, Cyp2c9, and Cyp2d6. As an additional piece of information connected to the prediction of pharmacokinetic properties, glycoprotein substrate capacity was assessed, but again the results from the two softwares were not homogeneous. To summarize, it was found that, although the lipophilicity scores may need to be further optimized, the predicted absorption, distribution, metabolism, excretion, and toxicological (ADMET) properties of the APH compounds characterize them as a suitable starting point to develop a leader compound. All the predicted characteristics of APHs in passing the BBB, permeability to Caco-2 cells, interaction of the compound with the cytochrome P450 isoenzymes and P-glycoproteins are shown in Table S2.</p><!><p>APH1-APH5 were selected because they were, out of all the APH compounds tested, the ones with less cytotoxic effects, also presenting the most potent antioxidant and neuroprotective effect against ROS isults.17 Thus, the treatment of SN56 cells with APH1–APH5 at different concentrations (1, 10, 50, 100, 150, and 200 μM) only started to reduce cell viability at 200 μM (Figure S2). Thus, the range of 10–100 μM concentrations was selected, as it has shown no cytotoxicity, to tests APHs neuroprotective effects.</p><p>Treatment of neurons with the heavy (cadmium, mercury, and aluminum) and essential (iron, copper, and zinc) metals' produced a reduction of cell viability. This effect was avoided completely or partially when cells were pretreated with APH1–APH5 in a concentration-dependent way (Figures 3 and 4). The heavy and essential metals studied were reported to induce neuronal cell death,1−8 supporting our results.</p><p>The protection against essential metals started from 10 μM after pretreatment with AHP2, APH3, or APH4. For FeCl2, pretreatment with APH2 started the protection from 50 μM as well as after the pretreatment with APH1 or AHP2 in cells treated with essential metals (Figure 5a–c). Besides, in cells treated with heavy metals, the APH1–APH5 pretreatment started to protect, from 10 μM, against the cell viability reduction observed after heavy metals treatment alone. For CdCl2, pretreatment with APH1 and APH2 and for AlCl3, pretreatment with APH1 started cell death reversion from 50 μM (Figure 5a–c). Pretreatment with compounds APH3 and APH4, at the concentration of 100 μM, protected completely against the cytotoxic effects induced by all metals studied except for FeCl2 (Figures 5 and 6). Besides, pretreatment with compound APH5, at the concentration of 100 μM, only protected completely against the cytotoxic effects induced by ZnCl2, CdCl2, and HgCl2 (Figures 5 and 6).</p><!><p>Cell viability effects after FeCl2 (A) for CuCl2 (B) and ZnCl2 (C) treatment with or without selected APH compounds (10–100 μM) pretreatment. Data represent the mean ± SEM of three independent experiments in triplicate. ***p ≤ 0.001 compared to control; #p ≤ 0.05 and ###p ≤ 0.001 compared to metal treatment.</p><p>Cell viability effects after CdCl2 (A) for HgCl2 (B) and AlCl3 (C) treatment with or without selected APH compounds (10–100 μM) pretreatment. Data represent the mean ± SEM of three independent experiments in triplicate. **p ≤ 0.01 and ***p ≤ 0.001 compared to control; p ≤ 0.05 and ###p ≤ 0.001 compared to metal treatment.</p><!><p>The differences observed in the concentrations from which these compounds start to show protection against the metal-induced cytotoxic effects can be ascribed to the different potency in their different protective mechanisms. These differences may also be produced because each APH compound may induce different mechanisms that mediate the neuroprotection observed. In this sense, our chelating studies suggest that Zn2+, Cd2+, and Al3+ ions are not chelated by any of the APH compounds studied and that Hg2+ is only chelated by APH3, APH4, and APH5 compounds. This fact indicates that the neuroprotection observed against zinc, cadmium, and aluminum by all these compounds or against mercury by compounds APH1 and APH2 is not mediated by their chelating activity and other mechanisms are involved. Besides, the protection observed against iron, and probably copper, at the lowest compound concentrations should be limited, especially for iron. The chelating activity at the higher concentrations used of these metals only blocked a very limited part of the metal pool, suggesting other mechanisms are involved. In this sense, we previously described that these compounds are MMP inhibitors and present antioxidant activity,17 which could participate in these neuroprotective effects. The greater neuroprotective action observed after pretreatment with all metals studied was mediated by compound APH4. This effect is correlated with our previous results on MMPs inhibition and antioxidant effect17 that support that these mechanisms are probably involved in the neuroprotective effects of these compounds.</p><!><p>The treatment of SN56 cells with compounds APH1–APH5 induced a concentration-dependent decrease in the gene and protein expression of Keap1 (Figure 7a,b). A correlated concentration-dependent increase in the gene and proteins expression of Nrf2 factor (Figure 7c,d) was observed. Besides, the gene and protein expression of Gpx and Cat enzymes was increased, in a concentration-dependent way, after the treatment with compounds APH1–APH5 (Figures 8a–c). Thus, APH compounds induce Nrf2 pathway.</p><!><p>Obtained results from Keap1 (A) and Nrf2 (C) gene expression and Keap1 (C) and Nrf2 (D) protein expression after 24 h treatment. Keap1 and Nrf2 gene and protein expressions are compared with controls [cells treated with vehicle were the negative control]. Each bar represents mean ± SEM of three separate experiments from cells of different cultures, each one performed in triplicate and presented as percent untreated control for gene expression and in ng/mg protein form protein expression. ***p < 0.001 significantly different from controls.</p><p>Obtained results from Gpx (A) and Cat (B) gene expression and Gpx (C) and Cat (D) protein expression after 24 h treatment. Gpx and Cat gene and protein expressions are compared with controls [cells treated with vehicle were the negative control]. Each bar represents mean ± SEM of three separate experiments from cells of different cultures, each one performed in triplicate and presented as percent untreated control for gene expression and in ng/mg protein for protein expression. ***p < 0.001 significantly different from controls.</p><!><p>Keap1 protein regulate Nrf2 proteins levels, inducing its degradation under no stress situations, through proteasome degradation. However, after stress situations, like ROS generation, its affinity for Nrf2 protein is reduced, increasing Nrf2 levels. Nrf2 is then translocated to the nucleus where it acts as the master regulator of the expression of different antioxidant enzymes such as Gpx and Cat among others.18−20,36,37Keap1 downregulation was reported to induce Nrf2 protein overexpression and the enzymes regulated by it.38,39 Besides, Keap1 and Nrf2 gene expression was reported to be regulated by different factors like microRNAs (miRNAs) or different transcription factors.38,40 Nrf2 protein levels were shown to be altered by mechanisms independent of Keap1 protein action.36 In this sense, we showed that compounds APH1–APH5 induced the Nrf2 pathway by different mechanisms. On one hand, they induce the Nrf2 pathway by an independent mechanism of Keap1 regulation of Nrf2 proteins degradation, since these compounds act to upregulate the Nrf2 gene expression. On the other hand, compounds APH1–APH5 regulate the Nrf2 pathway by a mechanism dependent on keap1, since they downregulated Keap1 gene expression, contributing to the overexpression of Nrf2 protein and of the Gpx and Cat antioxidant enzymes regulated by it. Further studies are needed to determine the mechanism through which APH compounds mediate the Keap1 and Nrf2 gene expression alteration.</p><p>We previously described that APH compounds present antioxidant activity against ROS-induced lipid peroxidation.17 In the present study, we show that this action is mediated through the induction of Nrf2 pathway, but we cannot discard other mechanisms are involved. Besides, the greater effect on Keap1, Nrf2, Gpx, and Cat expression was mediated by compound APH4 (Figures 7 and 8), which is correlated with the higher antioxidant effect observed for this compound,17 supporting our results.</p><!><p>The heavy (CdCl2) and essential (FeCl2) metals generation of malondialdehyde (MDA) was determined as a marker of oxidative stress damage induced by metals. MDA, a product of the ROS-promoted degradation of arachidonic acid, is the best-known lipid peroxidation marker. Obtained results show that both metals induced lipid peroxidation at the concentrations used (Figure 9a,b). Concentrations used were selected because they were the ones that produce cell damage (data no shown). Several in vitro studies reported that FeCl2 and CdCl2 induced, at these concentrations, oxidative stress and cell death,8,41,42 supporting our results.</p><!><p>Neuroprotective action of selected APH compounds against lipid peroxidation induced by FeCl2 and CdCl2 metals in SN56 cells was measured by MDA assay. (A) MDA content after FeCl2 treatment with or without APH compounds. (B) MDA content after CdCl2 treatment with or without APH compounds. Values are given as mean ± SEM of three separate experiments from cells of different cultures, each one performed in triplicate and presented as percent untreated control. ***p ≤ 0.001 compared to control; ###p ≤ 0.001 compared to metal treatment.</p><!><p>Treatment of SN56 cells with APH compounds alone did not produce any effect on MDA levels (data not shown). Pretreatment with the antioxidant trolox at 100 μM concentration prevented partially the lipid peroxidation induced by cadmium and iron (Figure 9a,b). Pretreatment with compounds APH1, AHP2, APH3, APH4, or APH5, prior to heavy or essential metal treatment, averted lipid peroxidation in a concentration-dependent way, starting from the 10 μM concentration (Figure 9a,b). This protection was complete after pretreatment with APH4 compound at 50 μM concentration in both metals (Figure 9a,b). Finally, pretreatment with compounds APH2, APH3, or APH5, at 100 μM concentration, the same one used for trolox, protected completely against the lipid peroxidation induced by both metals (Figure 9a,b). Thus, compounds APH2–APH5 are more potent than trolox. These data support the antioxidant effect of these compounds not only after H2O2 insults as previously shown17 but also against all insults induced by ROS generated after metals exposure.</p><p>Otherwise, the compound APH4 was not able to protect completely against the cell damage induced by any metal at the concentration of 50 μM (Figures 5 and 6). However, it was able to avoid completely the oxidative stress induced by FeCl2 and CdCl2 (Figure 9a,b). Besides, the compounds APH3, APH4, and APH5, at the concentration of 100 μM, were not able to completely protect against the cell damage induced by FeCl2 (Figure 5a), but they prevented completely the oxidative stress induced by this metal (Figure 9a). These facts indicate that, besides the oxidative stress, additional mechanisms should be involved in the cytotoxicity of these metals. It also suggests that, in addition to the antioxidant effect of these compounds, they mediate their neuroprotective effect by other mechanisms, probably different than MMPs inhibition or the chelating activity depending of the metal. These metals were reported to induce the accumulation and aggregation of amyloid proteins leading to cell death.1−3 Moreover, these metals were related with the alteration of Wnt/β-catenin pathway and glutamatergic system and the induction of neuroinflammation, producing neurodegeneration.1−3,43,44 Studies to elucidate the neuroprotective mechanisms of APH compounds, besides of those commented already, have not been previously performed. However, the parent compound 6(5H)-phenanthridinone was described to present poly(ADP-ribose)polymerase (PARP) inhibitor activity.45 PARP was involved in inflammation induction. Its inhibition produces anti-inflammatory effects,46 and it was also observed that its inhibition increases the NRF2 nuclear and cytosolic levels in rat retinal cells injured by chronic hypoxia/reoxygenation process.47 Thus, AHP1–APH5 compounds could mediate their antioxidant effects as well as other neuroprotective mechanisms through PARP inhibition. Besides, phenanthridin-6(5H)-one derivatives were also shown to act as Wnt/β-catenin signaling pathway agonists.48 Additionally, Wnt/β-catenin signaling pathway activation was reported to possibly induce the NFR2 expression through GSK-3β inhibition.49 Therefore, APH compounds could also block these cytotoxic actions through these mechanisms, contributing together with the antioxidant, chelating, and MMPs inhibition activities to the neuroprotective effects observed against the neurotoxic action of these metals. Further studies are required to determine the rest of the neuroprotective mechanisms of these multitarget compounds that could be used in the treatment of different neurodegenerative diseases.</p><!><p>To summarize, all compounds APH1–APH5 were able to chelate iron and copper ions. Moreover, APH3, APH4, and APH5 were also able to chelate mercury ions. Nevertheless, none of the APH compounds was able to chelate zinc, cadmium, and aluminum. The cytotoxic effects induced by all studied metals were prevented completely or partially depending on the compound used, with the most promising compound APH4. However, it is necessary to corroborate the neuroprotection action in vivo and against other toxic stimuli to confirm APH4 compound as the best APH therapeutic tool.</p><p>Otherwise, the antioxidant effect of the evaluated compounds is mediated through induction of the Nrf2 pathway that leads to overexpression of antioxidant enzymes. Further studies are required to survey the rest of the possible mechanisms involved in the neuroprotective effect of these compounds.</p><p>Consequently, the relevance of our results lies in finding the chelating, antioxidant, and neuroprotective properties of new multitarget 7-aminophenanthridin-6(5H)-one derivatives that could help improve the treatment management of different neurodegenerative diseases.</p><!><p>All reagents were purchased from Sigma-Aldrich and Fluka (Madrid, Spain), and solvents from SDS (Madrid, Spain) were of commercial quality and were used as received. Reactions were monitored by thin-layer chromatography, 0.20 mm silica gel 60 F254 plates (Merck, Madrid, Spain) with fluorescent indicator (SDS CCM221254), and then visualized with an ultraviolet (UV) lamp. Separations by flash chromatography were performed on silica gel (SDS 60 ACC 15 40–63 μm). Infrared spectra were recorded on a PerkinElmer Paragon 1000 FT-IR spectrophotometer (PerkinElmer, Madrid, Spain). NMR spectra were obtained on a Bruker Avance 250 spectrometer working at 250 MHz for 1H and 63 MHz for 13C and operated via the standard Bruker software (Nuclear Magnetic Resonance, Centre for Research Assistance, Complutense University, Madrid, Spain). Chemical shifts are reported in parts per millions relative to tetramethylsilane, and spin multiplicities are given as s (singlet), d (doublet), t (triplet), q (quartet), or m (multiplet).</p><!><p>A solution of ethyl acetoacetate (3 mmol, 1 equiv) and the suitable primary amine (3–3.9 mmol, 1–1.3 equiv) in ethanol (5 mL) was added to cerium(IV) ammonium nitrate (5 mol) and was stirred for 30 min at room temperature. Then, appropriate ortho-nitrochalcone (2.4–3.6 mmol, 0.8–1.2 equiv) was added to the stirred solution, and the mixture was heated to 80 °C overnight. The reaction mixture was dissolved in ethyl acetate (30 mL) and washed consecutively with water and brine, and the organic layer was dried with anhydrous Na2SO4. The solvent was removed under reduced pressure, and the solid was purified by flash column chromatography on silica gel, eluting with a petroleum ether and ethyl acetate mixture (9:1, v/v). Compounds 1 were reported in the literature, and their characterization data were identical to those previously described.17</p><!><p>A solution of appropriate dihydroantranilate 1 (1.5 mmol) and DDQ (1.5 equiv) in 10 mL of toluene was stirred (10 mL) for 2 h at room temperature. Then, the reaction mixture was extracted with ethyl acetate (30 mL × 3), and the combined organic phase was washed with 30 mL water and 30 mL brine solution. The residue was concentrated under vacuum, and the crude was purified by flash chromatography on silica gel with a petroleum ether and ethyl acetate gradient (from 9/1 to 8/2). Compounds 2 were reported in the literature, and their characterization data were identical to those previously described.17</p><!><p>A solution of appropriate meta-terphenyl compound 2 (1 mmol) was dissolved in ethanol (25 mL), and 10% palladium on carbon (Pd/C) (0.1 mmol, 10 mol %) was added. The suspension was stirred at room temperature for 14 h under hydrogen atmosphere, the catalyst was removed by filtration over Celite and washed with dichloromethane, and the solvent evaporated. The crude was purified by column chromatography on silica gel, eluting with a petroleum ether-ethyl acetate gradient (from 8:2 to 1:1). Compounds APH were reported in the literature, and their characterization data were identical to those previously described.17</p><!><p>1H NMR (250 MHz, DMSO-d6) δ 9.46 (t, J = 4.9 Hz, 1H), 8.42 (d, J = 7.9 Hz, 1H), 7.86 (d, J = 6.9 Hz, 2H), 7.74 (s, 1H), 7.57–7.40 (m, 4H), 7.31 (d, J = 7.1 Hz, 1H), 7.21 (t, J = 7.5 Hz, 1H), 6.88 (s, 1H), 3.31–3.24 (m, 2H), 1.75–1.61 (m, 2H), 1.56–1.38 (m, 2H), 0.96 (t, J = 7.3 Hz, 3H).</p><!><p>1H NMR (250 MHz, CDCl3) δ 10.93 (s, 1H), 9.20 (s, 1H), 8.27 (d, J = 7.8 Hz, 1H), 7.73 (d, J = 1.3 Hz, 1H), 7.63 (dd, J = 8.1, 1.5 Hz, 2H), 7.52–7.40 (m, 4H), 7.33–7.27 (m, 3H), 7.26 (d, J = 1.4 Hz, 1H), 7.18 (d, J = 7.9 Hz, 1H), 6.99 (d, J = 8.9 Hz, 2H), 3.88 (s, 3H).</p><!><p>1H NMR (250 MHz, DMSO-d6) δ 8.46 (d, J = 8.1 Hz, 1H), 7.84 (s, 1H), 7.68 (d, J = 6.9 Hz, 2H), 7.59–7.42 (m, 5H), 7.42–7.32 (m, 2H), 7.29–7.17 (m, 3H), 7.11 (s, 1H), 6.80 (d, J = 8.9 Hz, 2H), 2.91 (s, 5H).</p><!><p>1H NMR (250 MHz, DMSO-d6) δ 11.62 (s, 1H), 11.19 (s, 1H), 8.46 (d, J = 7.9 Hz, 1H), 7.94 (s, J = 1.0 Hz, 1H), 7.74 (dd, J = 3.7, 1.0 Hz, 1H), 7.62 (dd, J = 5.0, 1.0 Hz, 1H), 7.50 (t, J = 8.1 Hz, 1H), 7.38–7.23 (m, 4H), 7.21–7.14 (m, 2H), 7.02 (d, J = 8.9 Hz, 2H), 3.79 (s, 3H).</p><!><p>1H NMR (250 MHz, DMSO-d6) δ 11.35 (d, J = 9.0 Hz, 2H), 8.64 (d, J = 8.1 Hz, 1H), 8.12 (d, J = 1.6 Hz, 1H), 7.84 (dd, J = 3.7, 1.2 Hz, 1H), 7.77 (d, J = 8.3 Hz, 1H), 7.72–7.63 (m, 2H), 7.52–7.45 (m, 2H), 7.45–7.37 (m, 3H), 7.29–7.08 (m, 2H).</p><!><p>The chelating studies were performed with a diode array HP8543 UV–vis spectrophotometer (Agilent Technologies), by using a 1.0 cm quartz cuvette and the HP Chemstation software. The absorption spectra were collected at room temperature. Absorbance measurements of the tested compounds alone or in the presence of FeCl2, CuCl2, HgCl2, ZnCl2, CdCl2, or AlCl3 were carried out in 100 mM phosphate buffer solution at pH 7.30 and registered within the 200–600 nm range. In all cases, blank solutions were properly analyzed.</p><p>Stock solutions of Fe2+, Cu2+, Hg2+, Zn2+, Cd2+, and Al3+ (2.5 mM, final concentration) were prepared in Milli-Q water. The metal solutions were stirred until complete dissolution by using an ultrasound probe (Vibracell Sonics), equipped with a 2 mm diameter titanium microtip, at 60% power width during 2 min. Evaluated compounds were dissolved in an aqueous mixture of DMSO 100 mM phosphate buffer pH = 7.30 at ratios 1:2 v/v (0.50 mM, final concentration) and 1:14 v/v (0.10 mM, final concentration) by employing an ultrasonic bath for 10 min. Working solutions of compounds alone and compound-metal complexes were daily prepared by dilution of stock solutions in buffer as required.</p><p>Chelation was detected as a consequence of the changes observed in the absorption spectra after 30 min of incubation. With the aim of estimating the stoichiometry of the compound-metal complexes, a fixed amount of APH1, AHP2, APH3, APH4, or APH5 compounds (50 μM) was mixed with increasing amounts of metal ion (6.25–50 μM). Similarly, a fixed amount of metal (50 μM) was mixed with increasing amounts of APH1, AHP2, APH3, APH4, or APH5 compounds to achieve compound concentrations from 50 to 200 μM. UV–vis spectra were carefully examined in order to evaluate absorption differences and to estimate the ratio metal/compound in the formed complexes. Derivative spectroscopy was also applied to resolve stoichiometry studies.</p><!><p>The ability of a molecule to form complexes can be examined by the molecular orbital frontier approach. This approach is based on the premise that electrons are transferred from the HOMO orbital to the LUMO orbital in the formation of new bonds between the atoms. Thus, HOMO orbital indicates the ability to donate electrons, the LUMO orbitals show the ability to accept electrons of a molecule, and the greater the energy gap is between these HOMO/LUMO orbitals, the higher the molecule polarizability is, which is usually translated into a higher chemical reactivity.50</p><p>Molecular modeling was performed with ChemBio3D software (version 14.0 ultra), and HOMO and LUMO levels were researched using molecular mechanics (MM2) and MMFF94 force field analysis. For the optimized structure, HOMO–LUMO energies were analyzed.</p><!><p>In silico assessment of relevant drug-likeness properties and a series of ADMET properties was carried out using ADMETsar version 2.0 (http://lmmd.ecust.edu.cn/admetsar2) and SwissADME; S (http://www.swissadme.ch/) software.51,52 We calculated the most significant physicochemical properties required for absorption into the central nervous system such as the molecular weight and octanol–water partition coefficient (log P), as other parameters in association with Lipinski's rule of five and the suitability of a drug molecule for oral administration such as H-bond acceptors and H-bond donors.53</p><!><p>The protection induced by APH compounds against metals toxic effects was evaluated using a cholinergic murine neuroblastoma cell line (SN56 cells) derived from the basal forebrain septum. Basal forebrain contains the majority of CNS cholinergic neurons,54 which innervate the hippocampus and frontal cortex, regulating learning and memory process.55,56 Their selective degeneration, as observed in AD and other neurodegenerative diseases, induced cognitive dysfunction.57,58 We chose this cells line since the essential and heavy metals brain accumulation was related with memory and learning alterations,1−3 with some of them reported to induce a more pronounced damage in basal forebrain cholinergic neurons,8,59 so it is a sensitive model to study their toxic effects. Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum, penicillin/streptomycin, 2 mM l-glutamine (Sigma, Madrid, Spain), and 1 mM sodium pyruvate was used to maintain cells at 37 °C and 5% CO2. This medium was changed every 48 h.60 Cells were differentiated via culture for 3 days with 1 mM dibutyryl-cAMP and 1 μM retinoic acid.61,62 Following cells differentiation, there were no differences in cells regarding the periods of treatment. All cells used in these studies showed to be mycoplasma-free using the Look Out Mycoplasma PCR Detection Kit (Sigma, Madrid, Spain).</p><p>Cells were seeded in 6-well plates at a density of 106 cells/well. To determine APH compounds neuroprotective effects against FeCl2 (200 μM), CuCl2 (20 μM), ZnCl2 (250 μM), CdCl2 (100 μM), HgCl2 (30 μM), and AlCl3 (300 μM) cytotoxic effects, cells were treated with the APH1, AHP2, APH3, APH4, or APH5 compounds in concentrations between 10 and 100 μM and with trolox at 100 μM concentration. Besides, to determine the antioxidant mechanisms of APH1, AHP2, APH3, APH4, and APH5 compounds, cells were treated for 24 h with them in concentrations between 10 and 100 μM. At least three replicate wells/treatment were used. A vehicle group was employed in parallel for each experiment as a control.</p><p>APH1–APH5 compounds were selected because they were the APH compounds with less cytotoxic effects and because they presented the most potent antioxidant and neuroprotective effect against ROS stimulus.17 The range of 10–100 μM concentrations was chosen because it is a range of concentrations without cytotoxicity in which it is more probable that their neuroprotective effects are developed. The concentrations of metals used were selected because they were observed to induce cell death (data not shown). The metals used were selected because they were reported to induce neurodegeneration and were mainly associated with neurodegenerative disese.1−3</p><!><p>The 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay, which is based on the cleavage of the yellow tetrazolium salt MTT to purple formazan crystals by mitochondrial dehydrogenase, was used to evaluate SN56 cells viability. The cytotoxic effects of metals on SN56 cells were determined after incubation for 24 h with FeCl2 (200 μM), CuCl2 (20 μM), ZnCl2 (250 μM), CdCl2 (100 μM), HgCl2 (30 μM), and AlCl3 (300 μM). The metal concentrations used were previously tested and observed to induce cell death (data not shown). The neuroprotective effects of compounds APH1, AHP2, APH3, APH4, and APH5 against these metals were evaluated after incubation for 24 h with FeCl2 (200 μM), CuCl2 (20 μM), ZnCl2 (250 μM), CdCl2 (100 μM), HgCl2 (30 μM), and AlCl3 (300 μM) of SN56 cells pretreated with compounds 4 at various concentrations (10–100 μM) for 2 h.</p><p>At the end of the assays, cells were incubated with 100 μL of yellow MTT solution (final concentration 0.5 mg/mL) for 4 h. After 4 h at 37 °C, the medium was removed, and 150 μL DMSO was used to dissolve the formazan reaction product. The formation of solubilized formazan product was measured spectrophotometrically at 570 nm (Fluoroskan Ascent FL Microplate Fluorometer and Luminometer, Thermo Fisher Scientific, Madrid, Spain). Control cells were taken as 100% viability.</p><!><p>The Trizol reagent method (Invitrogen, Madrid, Spain) was used to extract total RNA. A Nanodrop 2000 (Thermo Fisher Scientific, Madrid, Spain) spectrophotometer and an Experion Lab Chip (Bio-Rad, Madrid, Spain) gel were used to determine the final RNA concentration and assess the quality of total RNA samples, respectively. We synthesized the first strand cDNA using a PCR array first strand-synthesis kit (C-02; Super Array Bioscience, Madrid, Spain) with 1000 ng of cRNA, following the manufacturer's instructions, with an added genomic DNA elimination step and external RNA controls. Following reverse transcription, QPCR was performed using prevalidated primer sets (SuperArray Bioscience) for mRNAs encoding Keap1 (PPM34919B), Nrf2 (PPM24614A), Cat (PPM04394C), Gpx (PPM04345E), and Actb (PPM02945B). Actb was used as an internal control for normalization. A CFX96 was used to run reactions with real-time SYBR green PCR master mix PA-012 (SuperArray Bioscience). The thermocycler parameters were established as 95 °C for 10 min, followed by 40 cycles of 95 °C for 15 s and 72 °C for 30 s. We used the Ct (cycle threshold) method to calculate relative changes in gene expression. The expression data are presented as actual change multiples.63</p><!><p>Treated cells were washed with cold PBS, scraped, and harvested. Then, harvested cells were lysed using radioimmunoprecipitation assay (RIPA) reagent (Thermo Fisher, Madrid, Spain) with a protease repressors mixture (Thermo Fisher, Madrid, Spain), following the manufacturer's protocol. Finally, cell lysates were centrifuged at 10,000g during 10 min at 4 °C, and the supernatant was collected. Total protein levels were assayed with a bicinchoninic acid (BCA) assay kit (catalog number 23225; ThermoFisher, Madrid, Spain), according to manufacturer's guideline.</p><p>Keap1, Nrf2, Gpx, and Cat protein expression was determined employing commercial ELISA tests in cell lysate supernatant (catalog numbers MBS7212746, MBS776676, MBS732750, and MBS728474, respectively; MyBiosource, CA, USA) in accordance with producer's protocols. The protein concentrations of each target gene were normalized with the total protein levels measured by BCA kit, to prevent possible interferences with the real value of the targets' protein concentrations measured due to the induction of cell death. Protein levels were expressed in ng/mg protein.</p><!><p>MDA concentration was determined as an indicator of lipid peroxidation products induced by free radicals. Intracellular MDA production was quantified after 24 h exposure to FeCl2 (200 μM) or CdCl2 (100 μM) with or without APH1, AHP2, APH3, APH4, or APH5 compounds (10–100 μM), using a Lipid Peroxidation MDA Assay Kit (Abcam, Cambridge, UK), following the manufacturer's protocol. Briefly, following treatment, we collected 1 × 106 cells and homogenized them on ice in MDA lysis buffer (300 μL) with 3 μL BHT (100×). Then, the mix was centrifuged for 10 min at 13,000g to remove insoluble material. Sample (200 μL) or standard (200 μL of MDA) was mixed with 600 μL of thiobarbituric acid solution, incubated at 95 °C for 50 min, and cooled to room temperature in an ice bath for 10 min. We loaded all samples and standards (200 μL) (duplicate) into a clear 96-well plate, and a microplate reader (Fluoroskan Ascent FL Microplate Fluorometer and Luminometer, Thermo Fisher Scientific, Madrid, Spain) was used to record the absorbance at 532 nm. Concentration of MDA, determined as nmol/mg protein, is presented as percent untreated control.</p><!><p>At least three replicates for each experimental condition were performed, and the presented results were representative of these replicates. Data are presented as means ± standard error of the mean. Comparisons between experimental and control groups were performed by two-ways ANOVA analyses (concentration vs treatment) or one-way ANOVA analyses (analysis of different 7-aminophenanthridin compounds concentrations) followed by the Tukey posthoc test. Statistical difference was accepted when p ≤ 0.05. Statistical analysis of data was carried out using GraphPad Prism 5.01 software.</p><!><p>Figures showing optimized molecular structure of APH1–APH3 and APH5, their iron complex and their corresponding HOMOs and LUMOs; cell viability effects after APH1–APH5 treatment; and APH1–APH5 UV–vis spectra and first-order derivative absorption spectra, alone and in the presence of metals. Tables with results of HOMO/LUMO gap energy analysis and ADMET prediction of APH compounds (PDF)</p><p>cn1c00333_si_001.pdf</p><!><p>∥ These authors contributed equally.</p><!><p>This research was supported by the following grants: CTQ2015-68380-R and RTI2018-097662-B-I00 (from MICINN) to J.C.M. and PR26/16-16B (from UCM) to J.F.G., J.P., and N.R.C.</p><p>The authors declare no competing financial interest.</p>

PubMed Open Access

Valorization of products from grounded-coffee beans

The valorisation of ground coffee beans is discussed in two parts; the first research question relates to the extraction of cold brew from ground coffee beans to provide a healthy cold beverage. Two parameters were investigated: temperature, and the ratio of ground coffee beans to water. This work suggests that cold brew coffee can be extracted between 15 and 20 °C over 2 to 4 h instead of 24 h as outlined in typical cold brew extraction processes. The coffee aroma was the response variable. Part of this investigation was to develop a downstream product from the waste spent grounded coffee bean. This part of the study investigates the production of firelighters from spent ground coffee beans to reduce the impact of dumping significant quantities of spent coffee grounds from coffee houses, restaurants, and baristas on landfill sites, which can lead to environmental problems such as polluting water systems, killing wildlife and disturbing ecosystems. The study used spent ground coffee beans in products such as firelighters to test their efficacy. This application has shown promising results, with the firelighters showing longer burning times for the ignition of log fires while also emitting a gentle, pleasant coffee aroma. Abbreviations CBCCold brew coffee GCB Ground coffee beans SGCB Spent ground coffee beans TDS Total dissolved solids v/v Volume to volume ratio For many people, drinking coffee is part of a daily routine to start the day 1 . Coffee is a comforting beverage that influences mood and satisfies thirst. Coffee consumption in South Africa has increased steadily from 29 760 tonnes in 2012/13 to 55 000 tonnes in 2016/17 2 . A survey by Masterton's coffee showed that consumers want greater varieties of flavour and different experiences from their coffee drinking 3 . Cold-brewed coffee (CBC) could be a pleasant alternative as a cold beverage for summer months launching a new experience for coffee drinkers while still allowing them to enjoy the kick caffeine brings in a hot-brewed coffee drink. Cold-brewed coffee is not iced coffee (which is hot-brewed coffee over ice); CBC is prepared at room temperature over a 12 to 18-h period compared to quick, traditional hot brewing methods 4-6 .Brewing coffee is an extraction process that depends on several factors, such as the ratio of water to ground coffee beans (GCB) 7 , the temperature of the water, the diameter of the ground coffee particles (coarse, medium or powder), and the brewing time. The temperature significantly influences the aqueous solubility of the compounds, and the literature records a significant difference in composition between hot brewing and cold brewing 8 . Much literature has been published detailing the chemistry of hot water brewing, including quantification of the caffeine concentration as a function of the hot water brewing method 9,10 . During the hot brew extraction much more of the caffeine and caffeinic acids are extracted from the coffee bean giving hot brew coffee that rich coffee aroma and bitter taste.Coffee grounds contain a mixture of volatile 11 and non-volatile components, such as various oils, acids, and other aromatic molecules 5 . Collectively, these compounds are referred to as "coffee-soluble" and contribute significantly to the flavour of the coffee. Factors such as temperature affect the solubility and volatility of these compounds 5 . Solubility describes the number of solids dissolved out of the GCB into the water, where volatility refers to their ability to evaporate into the air. It is the increased volatility at higher temperatures that releases the aromatics more easily, giving rise to the enticing aroma of freshly brewed coffee 5 . What cold brewed coffee lacks in temperature, it makes up for in brewing time 4 . Increasing the time of extraction from a few minutes to hours aims to maximize the extraction of the solutes from the GCBs. Other factors, such as oxidation and degradation,

valorization_of_products_from_grounded-coffee_beans

<!>Materials and methods<!>Sample preparation.<!>Cold coffee extraction<!>Pilot-scale demonstration of the CBC extract.<!>Spent ground coffee bean preparation<!>Results<!>Firelighters.<!>Conclusion

<p>can still occur in cold brewed extraction methods, but this happens much more slowly than in the hot brewed method. A decrease in bitterness and acidity are also reported in cold brewing 12 , especially if the CBC is kept cold; in consequence CBC extracts might take on a much sweeter, floral characteristic. Cold brewed coffee has been reported to possess a different flavour profile, with a typical characterization of aroma, acidity, aftertaste, body, flavour and sweetness that differs from other coffee extraction processes 13 .</p><p>The large quantities of coffee consumed worldwide everyday result in immense quantities of spent coffee grounds. The environmental impact of dumping spent ground coffee beans (SCGB) is enormous; hence other the need to find ways to use the SCGB to produce other commercial products 14 .</p><p>Part of this investigation is to develop downstream products that will make use of the spent coffee beans as a more valuable commodity product that can be used by the end user. The second section of the paper describes the production of firelighters from SCGB. Normally, these residues would be dumped as fertilizer to condition soil, or used as a green alternative in construction materials 15 , but other higher value products, such as firelighters or burning bricks for heat generation, are possibilities. The grounds could be used as an exfoliant for cosmetic products, in biofuel 16 , or as a source of extracted antioxidants 17,18 .</p><p>In this research, the authors focused on firelighters as a possible alternative product for the SGCB.</p><p>The objectives of this study were to the development of a range of coffee-based beverages using a cold-brew process, develop quality assurance protocols for the cold-brew extract, evaluate methods to improve the shelf-life of the final beverage, and to beneficiate the coffee grounds resulting from the cold-brew process. The following 3 core objectives will be discuss in this paper; (1) to determine the optimum extraction time and temperature for a specific coffee bean from Rwanda and so extend the consumption of coffee into the hot summer months, (2) to utilize sensational criteria to determine the best dilution of CBC extract with water or as a freeze-dried product to study the best coffee aroma dilution, (3) to utilize the SGCB to manufacture firelighter as an alternative to fossil fuel firelighters.</p><p>The proposed project is illustrated diagrammatically below (Fig. 1):</p><!><p>Raw materials. Roasted coffee beans (medium) of Rwanda obtained from a local vendor, Masterton's Coffee House, Gqeberha (Port Elizabeth), South Africa.</p><!><p>Coffee beans were coarsely ground with a commercial coffee grinder.</p><!><p>The goal of this experiment was to determine the optimum temperature at which CBC should be brewed to yield the most product between 15 and 20 °C. Extractions of the coarse GCB were carried out between 2 and 4 h at ratio of coffee to water of 1/5, 1/7 and 1/12. The effect of these factors was monitored by measuring pH and total dissolved solids (TDS).</p><p>Total dissolved solids. Total dissolved solids were measured according to previously described methods 19 , and the total weight of water mass divided by the GCB weight used for each extraction.</p><!><p>A small-scale pilot demonstration was performed at a sister Technology Station-Agrifoods at Cape Town University of Technology. A reactor was charged with coarse ground Rwandan coffee beans (720 g) to which water (5 L) was added. The mixture was agitated slowing at 150 rpm for 12 h at 20 °C, filtered to remove the spent coffee beans, after which the CBC extract was freeze dried. Freeze drying the product will stabilize the product giving it a longer shelf-life.</p><p>3D-Surface plots. The CBC extractions and the freeze-dried granular products were prepared, and a in house panel consisting of the authors to rate the coffee, according to aroma, acidity, aftertaste, balance, body, and flavour. The tasting method as described in literature 20 was utilized. From these results, 3D surface plots</p><!><p>Spent ground coffee beans (SGCB) were mixed with wax and molasses and placed in a mould to set. Once set, a weight sample was set alight, the burning time was measured and compared to a commercial firelighter of the same weight.</p><!><p>Part 1: CBC extraction. During these experiments, it was noted that the extraction proceeds more rapidly than the 18-h suggested in the literature.8 All the different ratios reached their end point in just under four hours. This means time can be saved and more than one batch can be prepared in a single day 4 .</p><p>To verify the results, two extractions of two hours and four hours each were performed. Table 1 summarizes the results. A 100 ml sample was taken from each of these extractions and freeze-dried. The yield percentage was then calculated and compared to the TDS readings and to previously reported 18-h extractions.</p><p>Figure 2 shows the combined effect of pH results for the three GCB to water ratios at 1/5, 1/7, and 1/12 (volume:mass) mixture ratios respectively. There was no significant difference observed between the three ratios and the coffee extracts. The pH observed was between 4.5 and 5.5 which indicates the caffeic acids present in the coffee extract 21 .</p><p>The TDS results (Fig. 3) show that the ratio between GCB and water became constant after 1 to 1.5 h for all three GCB to water ratios, regardless of the concentration of the GCB in the cold water. As with the pH readings, after 1 to 1.5 h the extraction was complete.</p><p>The CBC extraction samples were subjected to various criteria to grade the CBC extract. A chart was constructed with the following criteria: aroma, flavour, acidity, body, aftertaste, and balance.</p><p>The aroma was chosen because of the effect aroma has on taste. Coffee aroma has several attributes that could enhance the coffee favour, other than the mouthfeel and the sweet, salt, bitter, and sour taste attributes that are perceived by the tongue 22 . The CBC needs to have a good, full-body taste and not be flat or bland. Aroma is detected by two different mechanisms: it is sensed nasally via smelling the coffee, or by retro nasal perception which occurs when the coffee is either present in the mouth or has been swallowed and aromatic volatile compounds drift upward into the nasal passage 23 .</p><p>Acidity has nothing to do with the amount of acid present or the pH; it describes a coffee taste that refers to the fruity, tangy, wine-like flavour that characterizes many Arabica coffees 24 . The acidity becomes more prominent in longer roasting times of GCB.</p><p>"Body" describes the texture of the cold brew coffee; the full-bodied coffee taste refers to the strong, soft, and enjoyable feel of the coffee in the mouth. A coffee's body (light, medium, or full) refers to the depth of the aroma on the palate caused by the amount of dissolved and suspended solids and oils extracted from the coffee grounds. This may range in thickness from a thin, watery texture to a thick, creamy texture 25 . The aftertaste refers to the taste of brewed coffee vapours released after swallowing. Also called "finish", aftertastes can be burnt, chocolatey, spicy, etc.</p><p>A balanced coffee may be complex but does not have any overwhelming flavour or aroma characteristics. Most mixes were made up in a ratio of one teaspoon of coffee to a cup of water. Table 2 summarizes the CBC extract results which revealed that the most favoured test sample seemed to be the roasted coffee bean extract that was extracted in a 1:7 (v/v) CBC: water mixture ratio. Most of the tasters liked the 1:7 (v/v) freeze-dried sample. Sample 4 consisted of the fresh extract CBC in a 1: 2 (v/v) dilutions with cold water. This small population of tasters indicated that the ground roasted coffee bean, coarse particle size in a 1:7 (v/v) ratio to the water extract medium gave the best cold brew extract coffee according to the taste parameters.</p><p>The 3D surface model plots showed that when the sample has a good body, the aroma and flavour improve when compared to diluted samples or weaker ratios of extract coffee to water. Figure 4a shows the 3D surface plot for aroma, flavour, and body.</p><p>From the 3D surface model plot (Fig. 4b) which shows aroma, flavour, and acidity, it is clear that almost all the tasters like the acidity of the cold brew extract as the aroma becomes fuller in the body. Figure 4c is the 3D surface model plot for the comparison between aroma, body, and aftertaste; the comparison of these three parameters shows the tasters agreed that the CBC extract have a good aftertaste and coffee aroma.</p><p>Figure 4d shows the comparison of aroma, acidity and aftertaste. As the acidity increases, so does the aroma, and there were some improvements in the aftertaste of the coffee that was extracted at a 1:7 ratio.</p><p>Freeze drying. These results demonstrate that a 4-h brew yields 18% freeze-dried solid mass, compared to the 16% during a 2-h cold brew extraction. It was observed that the optimum cold brew extraction is in the region of about 4 h compared to the 12-18 h suggested by other sources 8 .</p><p>Temperature determination. By analysing the results of the experiments at 10, 15 and 20 °C we can determine how the temperature affects the cold brewing of coffee. By keeping all factors constant except the temperature changes, we were able to compare the yields of these experiments. From the results, we can conclude that the temperature at which coffee is cold brewed, does not affect the quality of the product yield. This means we can save energy costs by not needing to cool the setup to 10 or 15 °C, but rather brewing at 20 °C. The cold extraction technique was compared to hot extraction at 95 °C. In each study, it can be observed that a smaller amount of TDS was extracted using hot water than extracts from the cold-water extraction. The amount TDS extracted ranged between 1021 and 1050 ppm compared to over 2000 ppm for the cold brew extraction. The pH remained the same at 4.88 for all the extractions, indicating that the caffeinic acid been extracted from the ground coffee beans. The amount of coffee extracted from the GCB during the cold brew extraction process gave a yield of 19% compared to the hot extraction method that resulted in only a 6% freeze-dried solid mass percentage.</p><p>Cold brew extraction. Table 2 summarized seven CBC dilutions and three freeze-dried CBC granular to water ratio samples that were prepared for sensational analysis to identify the best coffee aroma characterization. The characterization and interactions were to taste for acidity, aftertaste, aroma, balance, body, and flavour. The purpose of this test was to establish which product would be best to take to market and in what format. Most commercial coffee sellers prepare the CBC in advance and store it in a keg, in a cool place in order to achieve a good quality brew extract; they then filter the CBC before serving. The authors suggest that, in order to achieve a longer shelf life, freeze-drying the CBC extract would remove caffeinic compounds from the coffee and provide a better, healthier option, rich in caffeine but without the caffeinic aromatics. Table 2 indicates that the CBC extract, in most cases, appears to be flat in comparison to hot brew (traditional way) coffee and scores a 61% for the 1:7 freeze dried product (2.6 g/90 mL). The best overall ratio was the CBC extract dilution of 1:14, further diluted 1:1 with water. This was scored overall as 71%, followed by Sample 7, the 1:7 freeze-dried product, with a score of 67%. The panel enjoyed the overall taste of cold brew coffee and stated that although the body of the CBC is flatter than hot brew coffee extract, the CBC taste is more balanced and has a sweeter taste. Cold brew coffee extract also has a soft coffee aroma and aftertaste. Since the aim of this research was to utilize the entire GCB, it was important to study a case where they could be used as firelighters. There are other uses for spent beans, such as a fertilizer for soil conditioning, or a meat rub before BBQ, and so on. However, some entrepreneurs suggested investigating firelighters. The method was discussed under "Materials and Methods" and the results are given below. Utilization of waste material such as the spent GCB can be used for fertilizer or be used to make firelighters. The following discussion outlines the results obtained from producing firelighters from SGCB. The length of burning time of these firelighters was compared to commercial firelighters; on average they burn for about 9.45 min compared to commercial firelighters which burn for 10.45 min per 20 g.</p><!><p>The firelighters were constructed from three components: SGCB, molasses, and wax. The SGCB provided the filler, the wax the fuel, and the molasses the binder. Two ratios were used for the firelighters 2:1 and 1:1 SGCB to wax. With each ratio, molasses was added to one sample as a binder to see if it changed the way the firelighter burned, so in total, 2 samples of each ratio were prepared. The sample size of the firelighter is relatively big and will be made smaller for commercial purposes. The size of the test firelighters manufactured from the SGCB and the commercial ones was roughly between 20 and 25 g.</p><p>Results. The 20 g, 1:1 SGCB:wax ratio firelighters stayed alight for 9.45 min compared to the 20 g, 2:1 SGCB:wax ratio which burned for 16.05 min. In comparison, a commercial product stayed alight for about 10.45 min.</p><p>Table 3 summarizes the results of the burn test of the commercial firelighters against the laboratory prepared coffee firelighters (Fig. 5). Commercial firelighters only achieved an average burn time of 10 min and 45 s. The 2:1 (m/m) coffee to wax firelighters performed the best and burned for 16 min. The addition of molasses as a binder seems to have increased the burn time by 45 s, but by observation, a binder is not necessary for this ratio because, without the molasses, the moulds kept their shape throughout the burn test.</p><p>The ratio of 1:1 (m/m) did not perform as well as the 2:1 (m/m) ratio, achieving a burn time of only 9 min 45 s. Even with the addition of molasses as a binder, the burn time reached only 10 min. Observation confirmed that the 1:1 ratio contained too much wax, making the binding less firm than that of the 2:1 ratio; thus, as it started to burn, it fell apart very early on, so reducing the burn time.</p><p>The 2:1 ratio seems to be the best firelighter medium, surpassing even the commercial firelighter (Fig. 5).</p><!><p>The research shows that the cold brew coffee could be extracted in four hours with a GCB to water ratio of about 1:7 (m/v) mixture. This ratio produced the best aroma and acidity, giving a fuller bodied texture. Particle size was shown to have no significant effect on the extraction process. www.nature.com/scientificreports/ Future recommendations would be to continue with the 1:7 CBC extract in the water at room temperature but on a slightly bigger scale of about 5-10 L with and without agitation, under air and nitrogen atmosphere. This approach could provide good evidence about the feasibility of the project. Further studies could be done on freeze-drying the CBC extract or investigating suitable packaging materials by making use of sachets under dry conditions or using concentrates.</p><p>The firelighters proved to be a possible alternative use of the SGCB rather than merely disposing of it on a compost heap or using it as fertilizer. Making firelighters adds value to a spent product and could be used to start an entrepreneurial business.</p>

Scientific Reports - Nature

Illuminating the diversity of aromatic polyketide synthases in Aspergillus nidulans

Genome sequencing has revealed that fungi have the ability to synthesize many more natural products (NPs) than are currently known, but methods for obtaining suitable expression of NPs have been inadequate. We have developed a successful strategy that bypasses normal regulatory mechanisms. By efficient gene targeting, we have replaced, en masse, the promoters of non-reducing polyketide synthase (NR-PKS) genes, key genes in NP biosynthetic pathways and other genes necessary for NR-PKS product formation or release. This has allowed us to determine the products of eight NR-PKSs of A. nidulans, including seven novel compounds, as well as the NR-PKS genes required for the synthesis of the toxins, alternariol (8) and cichorine (19).

illuminating_the_diversity_of_aromatic_polyketide_synthases_in_aspergillus_nidulans

INTRODUCTION<!>RESULTS<!>Phylogenetic Relationships of NR-PKSs and Carbon Skeletons of their Products<!>DISCUSSION<!>CONCLUSION<!>Molecular genetic manipulations<!>Fermentation and LC/MS analysis

<p>Fungal natural products (NPs) are an important source of medically important compounds and sequencing projects have revealed that many fungal genomes contain large numbers of clusters of genes that are predicted to encode NP biosynthetic pathways.1,2 The number of predicted NP biosynthetic pathways for each species far exceeds the known number of natural products and there is great variation in NP gene clusters among species. Fungi, thus, have the capacity to produce many more NPs than are currently known and precedence indicates that many of them have remarkably useful medical activities. Most fungal biosynthetic pathways are cryptic, however, producing no products under normal laboratory growth conditions.3,4</p><p>Several approaches have been employed to obtain expression of orphan pathways.3–5 One approach has been to alter genes involved in chromatin packing to induce expression.6 While this has worked to a limited extent, this approach, so far has not lead to the expression of most NP gene clusters. Other approaches such as altering media7 or co-culturing with bacteria8 have produced some successes, but the great majority of NP gene clusters have not responded to these approaches. Clearly, to unlock the treasure house of natural products that fungi can produce, we need to have a more generally successful approach to fungal NP production.</p><p>One promising approach takes advantage of the fact that fungal natural product gene clusters often contain transcription factors that govern expression of all genes in the cluster.9 Replacing the promoters of these transcription factors with regulatable promoters has allowed the induction of expression of three clusters in A. nidulans that encode biosynthetic pathways for aspyridones,9 emodin derivatives,6,10 and asperfuranone11 (Table 1). Since this approach has been successful for us and for others, we wished to determine if this approach was generally applicable. We found, however, that in most cases induction of these transcription factors did not result in induction of useful amounts of NPs. We consequently developed an alternative strategy, completely bypassing normal regulation, in which we directly replaced the promoters of non-reducing polyketide synthase (NR-PKS) genes, key genes of NP biosynthetic pathways as well as other genes required for compound production or release.12,13 Because we did not replace the promoters of genes that may modify the products of the NR-PKS, we did not anticipate that we would obtain production of the final products of the NP biosynthetic pathways. We did, however, hope to identify novel compounds and develop a better understanding of NR-PKS function.</p><p>We report here that this approach is highly successful. It has allowed us to complete the determination of the products of eight NR-PKS genes of A. nidulans. We have discovered seven novel compounds and our findings have allowed us to determine the NR-PKS genes required for synthesis of cichorine (19), a phytotoxin produced by A. nidulans,14 and of alternariol (8),15 an important toxic contaminant of cereals and fruits not previously known to be produced by Aspergillus sp. Our approach should allow the discovery of products of other NP gene clusters, the production of large amounts of NPs from target clusters, and facilitate the prediction of chemical structures of the products of novel NR-PKS identified in fungal genome projects.</p><!><p>We took advantage of progress in generating transforming fragments by fusion PCR and in gene targeting in A. nidulans16,17 to replace the promoters of all putative transcription factors (TFs) associated with NP clusters in A. nidulans that have not previously been studied (Table 1). We replaced them with the regulatable promoter from the alcA gene18 using the strategy shown in Figure 1A. The alcA promoter allows us to repress expression of genes while the strain germinates and grows, and then induce the expression to very high levels. This allows us, in principle, to induce production of compounds that might be toxic to the producing strain. In instances in which the TFs are predicted on the basis of sequence to involve proteins encoded by two genes, we replaced the promoters of both genes. In some cases multiple putative TFs are associated with a single NP cluster and in those cases we replaced the promoters of each TF. In all, we created 33 new combinations of single and multiple TF promoter replacements for NP clusters for which the products are unknown. We then induced expression of the TFs and analyzed organic extracts for production of new NP by HPLC-DAD-MS (Supplementary Figure S1). Among seventeen clusters studied, induction of TFs in nine did not result in production of any detectable new NPs; and in five clusters, induction of one or more TFs resulted in the production of new NPs but in amounts too small to be isolated and structurally characterized by NMR (Table 1 and Supplementary Figure S1). This strategy, thus, does not result in adequate up regulation of most NP clusters that contain transcription factors.</p><p>Clearly, to unlock the treasure house of natural products that fungi can produce, we needed to have a more generally successful approach for fungal NP production. Given the efficiency of gene targeting that is now possible in A. nidulans, we reasoned that the systematic replacement of native promoters of unknown NP genes in A. nidulans with the regulatable alcA promoter would provide expression levels suitable to identify new NPs. To determine if bypassing normal NP regulatory mechanisms was feasible and practical, we focused our attention on NR-PKSs.</p><p>Fungal polyketide synthases (PKS)12,19 are key NP genes and they are abundant in the A. nidulans genome. We analyzed all 29 A. nidulans PKS genes and classified them into 14 NR-PKSs, 13 highly reducing PKSs (HR-PKSs), one HR-PKS lacking an enoyl reductase (ER) domain, and one hybrid PKS-non-ribosomal peptide synthetase, based on phylogeny and their domain architectures (Supplementary Figure S2). As discussed below, we were able to further divide the NR-PKSs into seven groups based on phylogeny, domain structure and known products (Figure 2, Supplementary Table S1, and Figure S3).</p><p>A great deal of work by several labs over many years19 has revealed the chemical products of six A. nidulans NR-PKSs (StcA,20,21 WA,22 MdpG,6,10,14 AptA,23,24 OrsA,6,8,25 and AfoE11) (Figure 2). To determine the products of the remaining eight NR-PKSs, we replaced their native promoters with the alcA promoter18 (Resulting strains are listed in Supplementary Table S3). We carried out the replacements in strains in which the production of the major natural product sterigmatocystin was eliminated.6 This simplifies analysis and we anticipated that, in some cases, elimination of sterigmatocystin production would free up subunits for incorporation into other NPs. Strains were grown in inducing conditions and natural products extracted from the medium or mycelia were subjected to HPLC-DAD-MS metabolite profiling (Figure 3). Gratifyingly, replacement of the promoters of five NR-PKS genes (ANID_06448.1, ANID_08383.1, ANID_00523.1, ANID_07903.1, and ANID_03230.1 using the Broad Institute gene designations (http://www.broadinstitute.org/annotation/genome/aspergillus_group/MultiHome.html) resulted in the production of at least one major compound upon induction of the alcA promoter. We designate these genes pkbA (ANID_06448.1), pkdA (ANID_00523.1), pkeA (ANID_07903.1), pkfA (ANID_03230.1). The gene ANID_08383.1 has been designated ausA previously.26,27 All the major polyketides produced were purified from large-scale cultures and the structures elucidated by spectroscopic methods (Supplementary Figure S6, for details on structural elucidation, see Supporting Information). These compounds are listed in Figure 2 (compounds 1– 7).</p><p>Induction of ANID_07071.1 resulted in the production of several NPs but all in small amounts (Figure 3). ANID_07071.1 does not contain a releasing domain, and, based on previous data,28 we hypothesized that another gene in the cluster is required for product release. ANID_07070.1, which is adjacent to ANID_07071.1, encodes a β-lactamase-type thioesterase that potentially might be involved in product release. We created a strain in which both the promoters of ANID_07071.1 and ANID_07070.1 were replaced with the alcA promoter. As hypothesized, induction of this strain resulted in a dramatically increased yield (Figure 3). We designate ANID_07071.1 pkgA and ANID_07070.1 pkgB. A larger scale induction of an ANID_07071.1 and ANID_07070.1 promoter replacement strain followed by extensive chromatography led to the isolation of two major compounds, alternariol (8),15 an important toxin with antifungal, phytotoxic and anti-cholinesterase activity previously reported from Alternaria sp., and dehydrocitreoisocoumarin (9), as well as four minor compounds including two heptaketide isocoumarins (10 and 11) and two hexaketide isocoumarins (12 and 13). It is interesting to note that ANID_07071.1 and WA both are heptaketide synthases. Instead of the intramolecular Claisen cyclization that occurs in WA which has a functional TE/CLC domain,29 with ANID_07071.1, which lacks a TE/CLC domain, the released heptaketide undergoes lactonization to produce 9, which converts to 8 in vivo (Figure 2).</p><p>The fact that induction of the two remaining NR-PKS genes, ANID_02032.1 and ANID_03386.1, did not produce compounds suggested that coexpression of nearby genes might be necessary to provide specialized starter units for these NR-PKSs.11,30 ANID_02032.1 is located next to an HR-PKS, ANID_02035.1, and ANID_03386.1 is located near ANID_03380.1 and ANID_03381.1, putative α and β subunits of a fatty acid synthase (FAS). We created a strain where both the promoters of the NR-PKS ANID_02032.1 and the HR-PKS ANID_02035.1 were replaced with the alcA promoter. The resulting strain produced at least two UV-Vis detectable compounds upon induction, although in very small amounts (Figure 3). We designate ANID_02032.1 pkhA and ANID_02035.1 pkhB. We were able to isolate one major compound, 2,4-dihydroxy-6-[(3E,5E,7E)-2-oxonona-3,5,7-trienyl]benzaldehyde (14) (Figure 2), from the scale-up culture when we lowered the induction temperature from 37 °C to 30 °C and extended the incubation time from 48 hr to 72 hr (Figure 3). Our data thus indicate that the HR-PKS, ANID_02035.1, produces the octatrienoyl starter that is loaded onto the SAT domain of the NR-PKS, ANID_02032.1 (Figure 2).</p><p>We created a strain where the promoters of ANID_03386.1 and the two FAS subunit genes ANID_03380.1 and ANID_03381.1, were all replaced with the alcA promoter. The two FAS subunits are divergently transcribed and we developed a procedure to replace the promoters of both genes using a single selectable marker (Figure 1B). Induced expression of all three genes led to the isolation of a major product, 2,4-dihydroxy-3-methyl-6-(2-oxoundecyl)benzaldehyde (15), from the mycelium (Figure 3). We also isolated compounds 16 and 17 from the mycelium and 18 from the medium (Figure 2). We designate ANID_03386.1 pkiA, ANID_03380.1 pkiB, and ANID_03381.1 pkiC. These data indicate that a decanoyl starter unit is first synthesized by the FAS (ANID_03380.1 and ANID_03381.1) and then loaded onto ANID_03386.1. Since we did not detect any decanoic acid, the decanoyl starter is likely to be transferred directly from the FAS to the SAT domain of ANID_03386.1, similar to the biosynthesis of norsolorinic acid anthrone.31 Compound 15 is then transaminated and aromatized to become 16, which is then oxidized to generate 17. The minor compound 18, thus, is a shunt product (Figure 2).</p><p>Based on these data, we were able to determine the PKSs required for the production of three A. nidulans natural products, a process that is often difficult and time-consuming. A labeled precursor approach32 has revealed that 3,5-dimethylorsellinic acid (3) (Figure 2) is the precursor of austinol and dehydroaustinol, two meroterpenoids isolated from A. nidulans.23 The fact that 3,5-dimethylorsellinic acid (3) is the product of ANID_08383.1, suggested that this PKS is required for the synthesis of these meroterpenoids. This has recently been confirmed by another group and by us.26,27 We have recently isolated the phytotoxin cichorine (19) (Figure 2) from A. nidulans.14 The chemical structure of cichorine suggested it might be synthesized by modification of the product of ANID_06448.1, i.e. compound 1. We deleted ANID_06448.1, and found that cichorine production was, indeed, eliminated (Supplementary Figure S7). ANID_06448.1 is, thus, the PKS responsible for cichorine biosynthesis.</p><!><p>The genus Aspergillus contains more than 150 species and many of them are potent producers of NP.33 The sequencing of the genomes of members of this genus presents an opportunity to tap this potentially rich source of important compounds, but a key missing element for the many species without developed molecular genetics systems is the ability to correlate NP gene clusters with their products easily. For example, if a species with a sequenced genome, but no molecular genetic system, was found to produce a potentially useful compound, but only at low levels, it would be difficult to exploit the potential utility of the compound. The ability to predict which PKS is essential for the production of the product might allow one to identify the gene cluster responsible for the production of the product and this cluster could be cloned and expressed heterologously to produce the desired compound in larger amounts.</p><p>Our data greatly expand the knowledge of polyketides produced by NR-PKSs in A. nidulans and we wondered whether our data in combination with sequence information might allow us to make useful predictions about the products of NR-PKSs in other Aspergillus species. We reasoned that a phylogenetic analysis of NR-PKSs might be useful because such an analysis includes key information on the presence, absence and order of functional NR-PKS domains. We used a previous phylogenetic analysis of KS domains by Kroken et al. as a starting point for our analyses.34 The analysis of Kroken et al. classified NR-PKSs into three subclades. Subclade I and II NR-PKSs do not contain a methyltransferase (CMeT) domain and are associated with aflatoxin and melanin biosynthesis, respectively. It should be noted that when this initial analysis was made, no product was associated with any of the NR-PKS in subclade III, which are NR-PKSs that contain CMeT domains. Since the product template (PT) domain embedded in the middle of the NR-PKSs has been shown to control regioselective cyclization,35–37 Li et al. further categorized subclade I and II NR-PKSs into five major groups with each group corresponding to a unique product size or cyclization regioselectivity using PT phylogeny.38</p><p>In this study, we have identified the products of eight NR-PKSs in A. nidulans and consequently have greatly increased the number of fungal NR-PKSs for which the products are known. In particular, we have increased the number of NR-PKSs in subclade III with associated products and this has now allowed us to further classify subclade III into groups VI and VII (Supplementary Figure S3). All group VI and VII NR-PKSs analyzed so far produce monocyclic aromatics with C2–C7 cyclization regioselectivity (Figure 2). Noticeably, some CMeT domainless NR-PKSs like ANID_03230.1 and ANID_02032.1 fell into group VII. These NR-PKSs produce monocyclic aromatics without methyl branching on their benzene rings (Figure 2). We further performed phylogenetic analysis of PT domains extracted from the known NR-PKSs and found, interestingly, that the phylogeny obtained using the PT domains is similar to the phylogeny obtained using simply the full length NR-PKS (Supplementary Table S1, Figure S3, and S4). Our analysis showed that NR-PKSs within the same group produce similar polyketides except NR-PKSs belonging to group V, which lacks the product-releasing domain. NR-PKSs in group V seem to produce the most diverse aromatic polyketides. They require separate genes to release their products and they generate multicyclic aromatics with various chain lengths and cyclization regioselectivities. The known NR-PKSs in group V include hepta- (ANID_07071.1 and GsfA39), octa- (ACAS28 and MdpG10), nona- (AptA24 and VrtA39), and decaketide synthases (AdaA24) with C2–C7 (ANID_07071.1), C6-C1 (GsfA), and C6-C11 (ACAS, MdpG, AptA, VrtA, and AdaA) cyclization modes.</p><p>To determine if the phylogenetic analysis we have done in A. nidulans could be applied more broadly to Aspergillus NR-PKSs sequenced so far, we analyzed 71 fungal NR-PKSs available at the Broad Institute Aspergillus Comparative Database (Supplementary Table S2 and Figure 4). Gratifyingly, 69 of 71 NR-PKSs from the Aspergillus Comparative Database fell into these seven groups. The only two outliers are AFL2G_04689 and AO090023000877 which are more similar to MSAS, a PR-PKS from A. terreus.40 Among 51 unknown NR-PKSs analyzed, 48 fell into groups III ~ VII, and the previously underrepresented groups VI and VII contained the greatest number of NR-PKSs. Although currently it is still not possible to predict the exact product of the NR-PKS using bioinformatics analysis alone, our data shows that at least for Aspergillus species genome sequenced so far, 69 of 71 NR-PKSs with canonical domain architectures can be grouped into these seven groups and, with the exception of group V, the polyketide products of the PKSs within each group are structurally similar. This is an important step for genome mining of aromatic polyketides biosynthesis genes in Aspergillus species with sequenced genomes because if a valuable compound is identified, the structure of the compound will, in many cases, narrow the search for NR-PKSs responsible for its production into a particular phylogenetic group containing only a few members.</p><!><p>The approach we have developed has allowed us to complete the determination of the products of aromatic polyketide synthases in A. nidulans, including seven novel compounds (4, 7, 14, 15, 16, 17, and 18). This represents a major step in exploiting the diversity of fungal natural products coded in genomes. Several important conclusions can be drawn from our results. First, it is feasible and practical to investigate A. nidulans natural products by bypassing normal regulation. Indeed, much more time and effort was expended in analyzing products of the expression strains than in creating them. Our efforts have resulted in the discovery that A. nidulans can produce the important toxin alternariol (8), the identification of the PKS responsible for its synthesis, and the identification of the PKS responsible for cichorine (19) biosynthesis. Second, we have shown that serial promoter replacements are feasible, and not prohibitively time consuming, and, thus, that there are no conceptual barriers to up regulating all the genes of target clusters. Serial promoter replacement should allow the overproduction of the final products of target natural product gene clusters (when the promoters of all genes in the cluster are replaced) as well as intermediates in the pathways (when some of the genes in the cluster are replaced). Replacing the promoters of all genes in large clusters would require the development of additional selectable markers or techniques to recycle markers, but this is certainly feasible. This promises to be valuable in translating NP into products. Although we have focused initially on NR-PKSs, the approach should work for other classes of NP biosynthetic pathways and should be applicable to other fungal species with developed molecular genetic systems. Third, our data validate our previous suggestion that if a NR-PKS does not contain a releasing domain, a separate gene in the same cluster encoding a β-lactamase thioesterase is required for releasing the compound.23,28 This is true for three NR-PKSs in A. nidulans, and 23 of the 71 NR-PKSs in the Aspergillus species genomes sequenced so far lack product releasing domains. Seventeen of them have a β-lactamase thioesterase or an esterase gene nearby (Supplementary Table S2). Release of NR-PKS products by β-lactamase thioesterase genes separate from the NR-PKS thus appears to be common in Aspergillus species. Fourth, our data indicate that if an HR-PKS or FAS is present in a cluster with an NR-PKS, it is very likely to provide the starter unit loaded onto the SAT domain of the NR-PKS. This is true for A. nidulans NR-PKSs, and 17 of the 71 NR-PKS in the Aspergillus species genomes sequenced so far are in clusters with an HR-PKS or FAS (Supplementary Table S2). Production of starter units for NR-PKSs by HR-PKSs or FASs thus appears to be a frequent feature of fungal secondary metabolism. Fifth, our data in combination with previous data19,24 indicate that each of the 14 NR-PKSs in A. nidulans produces a unique product (Figure 2). This reveals that the PKS, themselves, generate a great deal of the diversity of fungal natural products. This diversity is then multiplied by the enzymes that modify the PKS products resulting, in principle, in huge numbers of different NPs. Finally, our results have greatly expanded the data on the products generated by fungal NR-PKSs and will facilitate genome mining efforts by narrowing down the number of target genes that need to be experimentally verified.</p><p>In view of previous data,6,9–11 we were surprised that up regulation of transcription factors generally yielded no or inadequate increases in compound production. We do not know why this was the case. It could be due to post-translational down regulation of the activity of the transcription factor as has been demonstrated for the AflR transcription factor,41 to other mechanisms of transcriptional inhibition as has been demonstrated in A. terreus,42 or to other, as yet unknown mechanisms. In any case, simple up regulation of transcription factors associated with SM clusters does not appear to be a generally successful strategy for obtaining expression of SM gene clusters.</p><!><p>We have developed a strategy that allows us to obtain expression of cryptic NP genes by bypassing normal regulatory mechanisms. Using this strategy we have successfully deciphered the products of eight NR-PKSs in A. nidulans. Furthermore, structural information derived from this work has allowed improved prediction of the carbon skeleton of aromatic polyketides from NR-PKS sequences through structure-phylogenetic analysis. This will greatly facilitate the elucidation of NR-PKS biosynthetic pathways in other fungal species.</p><!><p>Replacement of endogenous promoters with the alcA promoter was carried out as shown in Figure 1. Primers used in this study are listed in Supplemental Table S4. Transforming fragments were generated by fusion PCR as described16,17 except that in most cases KOD DNA polymerase (EMD Biosciences) was used instead of Accuprime Taq HiFi because of the stronger proofreading activity of KOD. With KOD enzyme, fusions were carried out at Tm (for the lowest melting temperature primer) + 2 °C and the extension time was 30 seconds per kb of expected fusion product. LO2026 (Supplemental Table S3) was used as a recipient strain for most transformations and the Aspergillus fumigatus pyrG (AfpyrG) and pyroA (AfpyroA) genes were used as selectable markers. For the ANID_03380.1; ANID_03381.1; ANID_03386.1 gene cluster, LO4389 was used as the recipient strain (Supplemental Table S3). Replacing the promoters of ANID_03380.1 and ANID_03381.1 presented a problem because the two genes (which encode fatty acid synthase subunits) are divergently transcribed and the distance between the two coding sequences is only about 600 bp. With our normal promoter replacement procedure, replacement of the promoter of one of the two genes with the alcA promoter would have been straightforward but subsequent replacement of the promoter of the second gene would have resulted in the deletion of the alcA promoter of the first gene. To circumvent this problem, we developed the procedure shown in Figure 1B. Fusion PCR was used to create two transforming fragments. One contained a portion of ANID_03381.1 fused to the alcA promoter and a fragment carrying a portion of the AfpyrG gene. The second contained a portion of the AfpyrG gene and the alcA promoter fused to ANID_03380.1. The two fragments of AfpyrG overlapped by 540 bp. Transformation resulted in ANID_03381.1 and ANID_03382.1 being driven by separate copies of the alcA promoter and fusion of the two portions of AfpyrG creating a full-length, functional copy. The ANID_08383.1 and ANID_06448.1 deletions strains were generated by replacing each gene with the Aspergillus fumigatus pyrG gene in the A. nidulans strain LO2026 (nkuAΔ, stcJΔ).9 The construction of fusion PCR products, protoplast production, and transformation were carried out as described previously,16,17 For the construction of the fusion PCR fragments, two 1,000-bp fragments of genomic A. nidulans DNA, upstream and downstream of the targeted gene, were amplified by PCR. All transformants were verified by diagnostic PCR (Supplemental Figure S5). Genotypes of all strains are given in Table S3 in the Supplemental Information.</p><!><p>For alcA(p) induction, 30× 106 spores were grown in 30 ml liquid LMM medium (15 g/l lactose, 6g/l NaNO3, 0.52g/l KCl, 0.52g/l MgSO4·7H2O, 1.52g/l KH2PO4, 1 ml/l trace element) in 125 ml flasks at 37°C with shaking at 180 rpm and supplemented with uracil (1 g/ l), uridine (10 mM), riboflavin (2.5 mg/l), or pyridoxine (0.5 mg/l) when necessary. Cyclopentanone at a final concentration of 10 mM was added to the medium 18 hr after inoculation. Culture medium was collected 48 hr after cyclopentanone induction by filtration and extracted with the same volume of EtOAc twice. The mycelium collected was soaked in 50 ml of MeOH for one day. After removing the cell debris by filtration, MeOH was collected, concentrated, resuspended in 25 ml of ddH2O, and extracted with the same volume of EtOAc twice. EtOAc from the combined EtOAc layers was evaporated by TurboVap LV (Caliper LifeSciences). The crude extracts were then re-dissolved in 0.5 ml of DMSO:MeOH (1:4) and 10 μl was injected for LC-DAD-MS analysis as described previously.6</p>

PubMed Author Manuscript

Stereoselective peptide catalysis in complex environments – from river water to cell lysates

Many stereoselective peptide catalysts have been established. They consist, like nature's catalysts, of amino acids but have significantly lower molecular weights than enzymes. Whereas enzymes operate with exquisite chemoselectivity in complex biological environments, peptide catalysts are used in pure organic solvents and at higher concentrations. Can a peptide catalyst exhibit chemoselectivity reminiscent of enzymes? Here, we investigated the properties of tripeptide catalysts in complex mixtures in hydrophobic and aqueous solvents. We challenged the catalysts with biomolecules bearing functional groups that could interfere by coordination or reaction with the peptide, the substrates, or intermediates.H-DPro-aMePro-Glu-NHC 12 H 15 emerged through tailoring of the trans/cis ratio of the tertiary amide as a conformationally well-defined tripeptide that catalyzes C-C bond formations with high reactivity and stereoselectivityregardless of the solvent and compound composition. The chemoselectivity of the tripeptide is so high that it even catalyzes reactions in cell lysates. The findings provoke the question of the potential role of peptide catalysis in nature and during the evolution of enzymes.

stereoselective_peptide_catalysis_in_complex_environments_–_from_river_water_to_cell_lysates

Introduction<!>Results & discussion<!>Peptide catalysis in complex mixtures<!>Conformational tailoring of peptide catalyst<!>Peptide catalysis in cell lysate<!>Conclusion

<p>During the past two decades, peptides have been recognized as potent catalysts for different reactions. 1,2 Several of these peptide catalysts feature exquisite levels of stereoselectivity and reactivity. Since they consist, like nature's catalysts, of amino acids but are signicantly smaller, peptide catalysts can be viewed as "mini-enzymes". Yet, whereas enzymes catalyze reactions in aqueous media, most peptidic catalysts operate in organic solvents. ‡ [1][2][3][4][5][6][7][8][9][10][11][12][13] A further marked difference is the environment in which catalysis takes place. Enzymes work in highly complex cellular environmentsreaction media that require exceptional chemoselectivitywhile peptide catalysts are used in well-dened environments consisting only of substrates and products in pure solvents. Moreover, enzymes operate at signicantly lower concentrations than peptide catalysts. We became intrigued by the question of whether a peptide catalyst can have a chemoselectivity reminiscent of enzymes. Can stereoselective peptide catalysis occur in water in the presence of compounds abundant in nature, in complex compound mixtures, and possibly even in cell lysates?</p><p>Herein, we show that the peptide H-DPro-aMePro-Glu-NHC 12 H 15 is so chemoselective that it catalyzes C-C bond formation reactions between aldehydes and nitroolens with exquisite stereoselectivity in complex mixtures, including cell lysates. Key to the performance of this peptide is the aMePro residue that ensures a high trans/cis ratio of the tertiary amide bond regardless of the solvent and compound composition.</p><!><p>Peptide catalysis in water in the presence of biomolecules We used the alkylated tripeptide H-DPro-Pro-Glu-NHC 12 H 15 1 as a starting point for our studies (Fig. 1). This peptide is a stereoselective catalyst for the conjugate addition reaction of aldehydes to nitroolens in water and organic solvents. 13,14 The catalytic reaction proceeds via the formation of an enamine intermediate between the peptide and the aldehyde followed by a C-C bond formation with the nitroolen. [15][16][17][18] Essential for the catalytic efficiency of 1 is a high trans/cis ratio of the DPro-Pro amide bond 19 and the CO 2 H group of glutamate as an intramolecular proton donor. 14 In water, the alkyl chain facilitates the formation of an emulsion and thus the solubility of the otherwise water-insoluble substrates. 13 We began by exploring the effect of compounds common in biological systems on the reactivity and stereoselectivity of peptide 1 (Scheme 1). We thus added amino acids (A, B), mono- and disaccharides (C-E), nucleic acid derivatives (F-H), lipids and terpenes (I-K), vitamins (L, M), compounds of the central carbon metabolism (N, O), and glutathione (P) to the peptidecatalyzed conjugate addition reaction between butanal and (E)-nitrostyrene. In the presence of an equimolar amount of additives A-P with respect to 1 (3 mol%), the catalytic reaction proceeded with comparable diastereo-and enantioselectivity (d.r. 96 : 4, 91% ee) as in their absence (d.r. 98 : 2, 91% ee). Even the reactivity of peptide 1 was not affected by these additives (quantitative conversion of starting material to the product aer 24 h). The only exception was pyruvic acid (O) which caused a drop to $80% conversion. These results are noteworthy since the additives include carboxylic acid and phosphoric acid groups (A, B, H, M-O), metal complexes (M), heteroaromatic groups (B, F-H, M), alcohols (C-E, K-N), and thiols (P). These acidic and basic, H-bond donor and acceptor sites, and the metal complex could interfere by coordination to peptide 1 or the intermediates with the catalytic cycle. Even additives containing carbonyl (C-E), thiol (P), and primary amino groups (A, B, F, G) that could react with the secondary amine catalyst or the aldehyde and nitroolen substrates do not affect the reaction. The observed deactivation by pyruvic acid is likely due to its acidity (pK a ¼ 2.5), resulting in protonation of the amine, the reactive site of 1. Next, we increased the amount of each additive stepwise from an equimolar amount with respect to the peptide catalyst (3 mol%, 13.2 mM) to 10 mol% (44 mM), 50 mol% (220 mM), and 100 mol% (440 mM; Scheme 1). The latter concentrations are comparable to the overall metabolite concentration inside cells ($200 mM). 20 At the highest concentration (100 mol%, 440 mM), the stoichiometry of the additive and the nitroolen substrate is 1 : 1. Even at this concentration, most additives did not affect the catalytic performance of peptide 1 and full conversion; diastereoselectivities between 94 : 6 and 99 : 1, and enantioselectivities between 87% ee and 92% ee were observed. Only citric (N) and pyruvic acid (O) reduced the reactivity (<50% conversion), and phenylalanine (B) and glutathione (P) reduced the diastereoselectivity (d.r. #80 : 20). These ndings highlight the high chemoselectivity of peptide 1 for the conjugate addition reaction, which proceeds efficiently in water even in the presence of compounds that are abundant in nature and contain functional moieties that could interfere with the catalytic cycle.</p><!><p>We next extended our study to peptide catalysis in multicomponent mixtures. Specically, we performed the catalytic reaction in water from different sources (S1-S4), milk and milk substitutes (S5-S9), fruit and vegetable juices (S10-S15), coffee and tea (S16-S20), honey solution (S21), so drinks (S22-S25), alcoholics (S26-S33), vinegar (S34-S35) and olive oil (S36; Scheme 2a). These commercial, readily available "solvents" differ signicantly from each other in polarity, pH, viscosity, and their ingredients and are, thus, a good platform to probe the effect of different environments on peptide catalysis.</p><p>Control experiments showed that the aldehyde and nitro-olen do not react with each other in any of these mixtures in the absence of peptide 1. In the presence of 1, the conjugate addition product formed quantitatively in all solvents but with different stereoselectivity. The diastereomeric ratio syn/anti ranged from 93 : 7 in vodka to 98 : 2 in green tea, the enantioselectivity from 71% ee in honey solution to 90% ee in olive oil (Scheme 2a, blue bars). Hence, the complex environments affected the enantioselectivity of H-DPro-Pro-Glu-NHC 12 H 25 1. Whereas these results show that peptide catalysis is possible in complex mixtures, we sought an analog of 1 that would allow for Scheme 2 (a) Catalysis in complex mixtures S1-S36 with peptides 1 and 2. (b) Trans/cis ratio of 1a and 2a in different solvents.</p><p>consistently high product formation and stereoselectivity regardless of the environment.</p><!><p>We recently found that the stereoselectivity of the parent, nonalkylated peptide H-DPro-Pro-Glu-NH 2 1a correlates with the trans/cis conformer ratio of its DPro-Pro amide bond (Scheme 2b). 19 Hence, we suspected that the reduced stereoselectivity in some "solvents" arose from lower trans/cis ratios in these complex environments. Therefore, we sought an analog of 1 with a high trans/cis ratio regardless of the environment. Previously developed catalyst analogs bearing pipecolic acid or g-substituted Pro derivatives in place of Pro in the middle position have a higher trans/cis ratio and thus higher stereoselectivity than 1a in apolar solvents. 19,21 In polar solvents (e.g., H 2 O, MeOH, DMSO, DMF), these analogs have, however, lower trans/cis ratios and poorer stereoselectivities than 1a. 19 Thus, we needed an analog of 1 with a residue in the middle position that "locks" the peptide in the trans conformation. We envisioned that a-L-methylproline (aMePro), a proline derivative known to favor the trans conformer of Xaa-Pro bonds, [22][23][24][25] could ll this need. Based on the conformational analysis of peptide 1a, 16 we reasoned that the additional methyl group at C a should not affect the conformation of the peptide catalyst. Indeed, a crystal structure of the TFA-salt of H-DPro-aMePro-Glu-NH 2 (2a) was almost identical to that of the parent peptide 1a (see ESI † p. S10). Both peptides adopt a b-turn conformation with a H-bond between the C]O of DPro and the C-terminal amide. The similarity of 1a and 2a was further conrmed by NMR spectroscopy, as indicated by the coupling constants and nuclear Overhauser effects (Scheme 2b, right, and ESI † p. S9).</p><p>Since the trans/cis ratio of 2a is difficult to monitor in complex mixtures, we initially used polar and apolar, protic and aprotic organic solvents to probe the effect of the aMePro residue on the conformation and the catalytic properties of the peptide catalyst (Scheme 2b, middle). Peptide 2a has in all of these solvents a trans/cis ratio of $95 : 1 whereas the trans/cis ratio of peptide 1a varies with the solvent (80-98% of trans conformer). As a result, the stereoselectivity of 2a is higher than that of 1a in all solvents (see ESI † p. S11).</p><p>Building on these data, we prepared the alkylated analog H-DPro-aMePro-Glu-NHC 12 H 25 2 and evaluated its catalytic properties in the different "solvents" S1-S36 (Scheme 2a). Peptide 2 provided the g-nitroaldehyde with diastereoselectivities of 87 : 13-95 : 5 and enantioselectivities of 96% or 97% ee in each of these environments (Scheme 2a, orange bars). The data show that the "locked" trans amide bond of 2 enables peptide catalysis in any environment with uniformly high enantioselectivity. These results are noteworthy since the tested environments have signicantly different polarities (e.g., olive oil and milk versus water and alcoholics), viscosities (e.g., honey versus apple juice), pH (e.g., coke and vinegar versus tap water), and contain numerous compounds, including carbohydrates, lipids, peptides, and proteins, that could interfere or inhibit peptide catalysis. The data thus evidence the high chemoselectivity of the peptide catalyst for the conjugate addition reaction.</p><!><p>Encouraged by these data, we challenged peptide 2 to catalyze reactions in cell lysates (Scheme 3). We used lysates with a total protein concentration of 5.3 g mL À1 from human hepatoma cells (HepG2), an immortal cell line that secrets a variety of major plasma proteins. Under standard conditions (880 mM aldehyde, 440 mM nitroolen, 13.2 mM 2), complete conversion of the nitroolen was observed, and the conjugate addition product was isolated in 97% yield with a stereoselectivity of d.r. 94 : 6 and 95% ee. Thus, peptide 2 withstands even the multicomponent mixture of a cell lysate with numerous compounds that could either react or coordinate to the catalyst, the substrates, or the reaction intermediates. Enzyme concentrations range between micro-and millimolar inside cells. 20 Therefore, we reduced the reaction concentration by a factor of 10 and 100. Even at the lowest concentration, peptide 2 (132 mM) provided the product with an 81% yield and only slightly lower stereoselectivity (d.r. 85 : 15, 93% ee). These data show that the peptide is so reactive and chemoselective that stereoselective catalysis takes place in a highly competitive environment at a concentration typical for enzymes.</p><!><p>The study put forth the rst examples of stereoselective peptide catalysts that are sufficiently chemoselective to operate in complex mixtures, including cell lysates. The conformationally tailored peptide catalyzes C-C bond formation in both, hydrophobic and aqueous environments with remarkable stereoselectivity, even at concentrations typical for enzyme catalysis. Biomolecules that could interfere through coordination or covalent reaction with the peptide, the catalyst intermediates, or the substrates did not affect the peptide catalysis.</p><p>Of note, the key to the exquisite performance of the peptide catalyst in any environment was the tuning of the trans/cis amide bond ratio through substituents at proline, a concept that nature uses to regulate protein folding and stability as well as enzyme activity. [26][27][28][29][30] The results illustrate that a short-chain peptide can be a potent catalyst in any environment, including those that are hallmarks of naturally evolved enzymes. The ndings show the utility of peptide catalysts for organic synthesis and provoke the question of the potential role of small catalytic peptides in the evolution of enzymes. Proteomic analyses have revealed that the cells of microbes as well as those of humans contain a large number of peptides. 31 The function of many of these peptides is unknown. The presented data spark the question of whether some of these peptides catalyze biochemical reactions, orchestrating an underworld of clandestine organocatalysis that yet needs to be discovered?</p>

Royal Society of Chemistry (RSC)

Optimization of Enzymatic Antibody Fragmentation for Yield, Efficiency, and Binding Affinity

Enzymatic antibody fragmentation has been well studied for various hosts and isotypes, but fragmentation patterns also vary unpredictably by clone, and optimizing Fab or F(ab\xe2\x80\x99)2 production by trial and error consumes large quantities of antibodies. Here, we report a systematic strategy for optimizing functional F(ab\xe2\x80\x99)2 production via pepsin digestion from small quantities of IgG. We tested three key parameters that affect fragmentation: pH, enzyme concentration (% pepsin w/w), and reaction time, and found that pH had the greatest impact on fragmentation yield and efficiency. We then developed a systematic approach to obtaining acceptable yields, digestion efficiency, and binding affinity. Three case studies are described to illustrate the approach. We anticipate that this work will provide a quick and cost-effective method for researchers to produce antibody fragments from whole IgG, avoiding haphazard trial and error.

optimization_of_enzymatic_antibody_fragmentation_for_yield,_efficiency,_and_binding_affinity

Introduction<!>Fragmentation yield and efficiency varies by clone under identical conditions<!>Fragmentation efficiency and yield are most sensitive to pH<!>Identification of acceptance criteria<!>Case Studies: On reaching acceptable yield, efficiency, and function in real life<!>Case Study 1:<!>Case Study 2:<!>Case Study 3:<!>Summary: Suggested optimization procedure<!>Conclusions<!>Filter Preparation procedure<!>Antibody fragmentation procedure<!>SDS PAGE and analysis of data<!>ELISA for affinity

<p>Antibody fragments have enjoyed widespread use in a variety of disciplines ranging from bioanalytical immunoassays to pharmacological research. Their small size and lack of an Fc-mediated binding region allow for better penetration into tissue and specificity for numerous applications.1–4 There are two primary ways to produce such fragments: digestion with an enzyme such as papain for Fab fragments or pepsin for F(ab')2 fragments,5 or expression of a recombinant antibody fragment. The latter requires either knowledge of the sequence of the antibody, or a lengthy process of production from a host animal and RNA transfer to E. coli through a bacteriophage.6 This work focuses on enzymatic digestion, as it rapidly produces fragments from commercially available antibodies.</p><p>Although enzymatic antibody fragmentation with pepsin has been studied since the 1960's and 70's, most laboratories still must optimize the fragmentation of each particular antibody by trial and error. Commercial kits for fragmentation are widely available, and these offer suggested digestion conditions based on reported trends for isotype and host species.7–12 However, reports of fragmentation as a function of clonal differences are anecdotal, which can necessitate optimization by the user for each clone. Furthermore, the success criteria for fragmentation may vary based on the intended use of the fragments. Yield, purity, or efficiency of the reaction are frequent criteria. Another is the retention of binding affinity; depending on the reaction conditions, enzymatic digestion can impair function through aggregation or denaturation.10 There is no one optimal procedure for fragmentation reactions because each antibody and reaction presents unique challenges and goals.</p><p>A systematic approach to antibody fragmentation is needed to accelerate progress in laboratories working to generate their own antibody fragments. Such a procedure may prevent optimization from devolving into a series of trial and error experiments, which can quickly become cost prohibitive due to high price of antibodies. In this paper we report the development of such an approach for fragmentation by pepsin, designed to identify a set of conditions that meets the acceptance criteria for yield, digestion efficiency, and function as quickly as possible. We focused on digestion by pepsin to generate F(ab')2 fragments, which offer a higher avidity than single Fab fragments. We systematically determined the importance of pH, amount of enzyme, and digestion time in producing functional fragments from small (50 – 1000 ug) quantities of commercially available antibodies. Three case studies were used to illustrate the method and to identify key factors to be aware of when optimizing.</p><!><p>Here we focused on fragmentation of rat IgG antibodies because of their widespread use to detect antigens in murine tissues. Like other species of antibodies, rat IgGs differ in their sensitivity to pepsin depending on isotype (IgG 2c > 2b > 2a > 1), with less sensitive isotypes requiring longer digestion times, more enzyme, or lower pHs.8 However, these isotypic trends may not always translate to individual clones or preparations of antibody. Clonal factors such as primary sequence can determine how the antibody unfolds under acidic conditions, which may impact the efficiency of pepsin cleavage.13 In addition, the degree of glycosylation can vary based on host, isotype, and recombinant culture conditions, and the sterics of the glycans can alter the interaction between an antibody and the active site of pepsin.14</p><p>To illustrate this variation, we digested multiple rat IgGs with pepsin under identical conditions. Four clones used commonly for immunological experiments, spanning three isotypes, were selected for study: three clones of rat IgG1 (HRPN: isotype control, R4-6A2: anti-mouse IFN-γ, and XMG1.2: anti-mouse IFN-γ), a rat IgG2a (RA3-6B2: anti-mouse/human CD45R/B220), and a rat IgG2b (GK1.5: anti-mouse CD4). We tested a fragmentation condition that is typically recommended by commercial vendors: 1 mg/mL antibody at pH 4.4, using 125 % w/w pepsin for 2 hours.</p><p>Fragmentation followed patterns typical for most antibodies (Fig 1a). Residual unfragmented IgG appeared in the SDS-PAGE at ~ 140–160 kDa. Most of the antibodies ran as a doublet of peaks (except clone GK1.5), possibly due to multiple conformational states of the proteins (Fig 2b). This may be a result of only partially denaturing the samples with SDS and not performing a heat treatment and disulfide reduction of samples before gel analysis. Doublets were observed in undigested samples as well (data not shown). For most clones, the F(ab')2 fragment appeared close to the canonical molecular weight of 90 – 100 kDa, with a partially digested product (likely loss of a single Fc heavy chain) at ~ 125 kDa. Even this could vary, however, as seen with clone GK1.5, whose F(ab')2 appeared at ~ 80 kDa. The identity of the F(ab')2 bands were confirmed in one of two ways: 1) treatment of the sample with 2-MEA reducing agent, which eliminated the F(ab')2 band by reducing cysteine disulfides connecting the two Fab' regions, resulting in a 40–50 kDa band indicative of single Fab' fragments, or 2) passing the sample through an anti-Fc-γ affinity column, which removed incompletely digested products and retained the F(ab')2 band (Fig. S1).</p><p>The SDS-PAGE data was used to quantify the yield (Eq. 1, Fig. 1d) of F(ab')2, digestion efficiency (Eq. 2), and purity (Eq. 3) of the reaction: (1)% yield of F(ab′)2=mol of F(ab′)2 producedmol of starting IgG×100 (2)% F(ab′)2 Efficiency=F(ab′)2Undigested (150 kDa)+ Partially Undigested (125 kDa)+F(ab′)2(100 kDa)×100 (3)% F(ab′)2 Purity=F(ab′)2all products×100 Efficiency is distinct from purity as a readout, as efficiency reports on the conversion of the intact IgG to F(ab')2, whereas purity is concerned with the ratio of F(ab')2 to all fragments produced. Efficiency was defined as the percent of F(ab')2 product out of undigested and partially undigested products and purity was defined as the percent of F(ab')2 out of the total products, including all of the smaller fragments (Fc fragments and other small over-digested/degraded fragments). Thus, conditions that have a high digestion efficiency could potentially have a significantly lower purity due to the inclusion of smaller fragments (Fig 1e–f). Whether purity or efficiency is the more relevant metric will depend on the intended use of the product. Here, we used efficiency because our goal was primarily to remove the Fc region from the antibody, and smaller fragments would not affect our downstream applications.</p><p>Under these conditions, both the yield and efficiency of digestion varied more than 10-fold between clones and isotypes of rat IgG. Even amongst the three IgG1s, F(ab')2 yield ranged from 4.0 to 42 %, and efficiency between 9.2 and 99 %. These data clearly demonstrate the need for tailored reaction conditions for each antibody clone if high yield, efficiency, or purity is desired.</p><!><p>There are multiple variables that affect the efficiency and yield of an enzymatic fragmentation reaction, including reaction time, temperature, pH, concentrations of antibody and enzyme, and mixing during reaction. For these experiments, we focused on three variables that had a significant impact in prior studies: pH, ratio of enzyme to antibody (% w/w), and reaction time.10 The reaction time and the ratio of enzyme (pepsin) concentration to substrate (antibody) concentration drive the reaction based on enzyme kinetics, whereas pH has a more indirect effect by impacting enzymatic activity and protein unfolding. Pepsin is most active at a pH of 2, and gradually loses activity as pH is increased until being irreversibly denatured at a pH of 6.15 In addition, it is thought that as pH decreases, some antibodies may slightly unfold or denature, increasing the accessibility of the antibody to the pepsin's active site.</p><p>Because the three selected variables are not independent, small trial-and-error-based changes in reaction conditions can easily result in under or over-shooting the desired efficiency, with a trade-off in yield. For example, the optimal enzyme concentration likely increases at higher pH as the enzyme loses activity, so changing only the concentration or only pH is not ideal. In this situation, a systematic Design of Experiments approach proves especially useful to quickly identify the most critical parameters.16 Rather than vary one variable at a time, multiple variables are tested simultaneously to identify critical variables and co-dependencies.</p><p>Here, we used this systematic approach to determine which variable had the greatest effect on the % yield of the F(ab')2 fragment and efficiency of the reaction. We fragmented a rat IgG1 isotype control (HRPN), an inexpensive and readily available antibody, and simultaneously tested all three variables with two levels each (23 = 8 combinations of conditions). The array of conditions tested was representative of oft-suggested conditions for antibody fragmentation: pH at 2.8 or 4.4, pepsin concentration at 1 or 20 %, and reaction time at 2 or 6 hr. The antibody concentration was kept constant at 1 mg/mL, and all reactions were conducted at 37 °C with continuous gentle agitation.</p><p>Over this range of conditions, a change in pH had the greatest impact, followed by change in concentration of pepsin (Fig. 2). Digestion at a low pH of 2.8 produced the highest yields and efficiency of HRPN F(ab')2, reaching 99 % yield and 100 % efficiency at the optimal condition (Table 1). While it is possible that varying the pepsin concentration or time over a greater range (e.g. 1 % – 200% pepsin or 1 h – 24 h) may increase the impact of those variables, higher concentrations and times can lead to over-fragmentation and/or degradation of the antibody.17</p><!><p>Practically speaking, for most experiments, optimizing for the highest possible yield and efficiency is unrealistic and unnecessarily laborious. It is more useful to define in advance the "acceptance criteria" – the minimum values of yield and efficiency (and function) that are considered sufficient. The values of acceptance criteria will depend on the application. The requirement for % yield is likely driven by cost considerations and the quantity of the fragment that is required for downstream use. The required purity and efficiency of the reaction depends on both the intended use and how easily the products can be purified by subsequent processing. Finally, the requirement for functionality (i.e. binding affinity) depends on whether the fragment will be used for antigen binding, versus simply being sequenced.</p><p>Whether reaction efficiency is a priority likely depends largely on the available options for downstream purification. When working with small quantities of antibody (50 μg per reaction), isolation of functional F(ab')2 fragments from other fragments may be challenging. Microscale size exclusion chromatography is the best option, but requires specialized equipment.18 For many antibodies, affinity-based techniques such as proteins A and G, which bind to antibody Fc regions, may prove a suitable means to remove residual unfragmented or partially fragmented products from the mixture.19 For others, however, this approach is not viable. For example, protein A has little to no affinity for rat IgG. Protein G does have some affinity for rat IgGs, especially rat IgG2a, but it binds both the Fc and Fab regions of the antibody, compromising its use for purification.19–21 Thus, for easily purified antibodies, especially those available at large scale, a mediocre efficiency (~50–80%) may be acceptable, but for others a much higher efficiency is needed.</p><!><p>As a series of case studies, this approach to optimizing fragmentation was applied to three rat IgG's (Table 1). We intended to use the F(ab')2 fragments as detection reagents in applications such as flow cytometry, ELISA, and immunofluorescent staining of tissue.3 Acceptance criteria were determined based on this set of applications as follows. Given that the selected clones were available in bulk from commercial vendors, we determined that a yield of > 50 % was sufficiently cost effective to accept. However, for some of the intended applications, residual Fc regions would be detrimental. Because of the difficulty of further purifying rat F(ab')2 fragments, we prioritized reaction efficiency over yield, and initially required a efficiency of > 90%. Finally, because we intended to use the fragments for antigen binding, it was also critical that the binding affinity remain high, ideally within 2-fold of the IC50 of the original antibody as measured by an affinity ELISA. In each case, we first screened conditions for yield and efficiency, then confirmed the binding affinity of the product.</p><!><p>First, we sought to produce F(ab')2 fragments from anti-mouse CD4 (clone GK1.5), which is a rat IgG2a antibody. CD4 is a canonical phenotypic marker for helper T cells in flow cytometry and immunofluorescence. Based on the results shown above, in which pH had the most significant impact on the fragmentation, the pH was varied in three levels from 2.8, the optimal value for HRPN, and the traditional pH 4.4, which is recommended in many commercial kits. The % pepsin was varied simultaneously in two levels from 1 to 10%. Reaction time was held constant at a convenient value (2 hr).</p><p>Once again, we observed that pH had a substantial impact on yield and efficiency (Figure 3a–b), especially at lower pepsin concentration. Pepsin concentration also had a substantial impact, especially at higher pH, where it was presumably less active. Fortunately, one of the tested conditions (pH 3.5, 10% pepsin, 2 hours) surpassed the acceptance criteria for yield and efficiency (Table 1), and thus further optimization was not required. The F(ab')2 product from that condition was subsequently tested for affinity, and was found to have comparable affinity to its whole antibody precursor (~15 nM IC50 for both; Figure 3c). Thus, this condition was accepted for further use. This case study demonstrates the usefulness of screening multiple pHs over a limited range of pepsin percentage and reaction time, which can quickly identify a near optimal condition that can be further refined if needed.</p><!><p>We found that for some antibodies, despite modulating pH, time, and pepsin, it was difficult to obtain high yield and efficiency simultaneously. For example, this phenomenon occurred with fragmentation of anti-mouse IFN-γ (clone R4-6A2), a rat IgG1 antibody. IFN-γ is a cytokine, i.e. a protein involved in extracellular signaling between immune cells, that is critical for adaptive immunity. In attempting to optimize this antibody, we tested pHs of 3.5, 4.0, and 5.0 with a wide range of pepsin concentrations and reaction times, in an extensive series of optimization experiments. We observed that a pH of 5.0 produced the greatest yields, and at lower pH such as pH 2.8 the antibody denatured and aggregated (data not shown). At pH 5.0, yields were never above 50 %, and increasing the efficiency by increasing the pepsin concentration or time often came with a further tradeoff in yield due to over-fragmentation (Figure 4a,b). Given our acceptance criteria and priorities outlined above, the most suitable condition was pH 5.0 with 100 % pepsin w/w for 24 hours, which had the highest efficiency but the lowest yield (Table 1). When the F(ab')2 fragment from that condition was tested for affinity, its IC50 was comparable to the affinity of whole (unfragmented) antibody (Figure 4c), so this condition was accepted for further use. This case study demonstrates the need to compromise between acceptance criteria for some antibodies.</p><!><p>As a final case, we generated F(ab')2 fragments from anti-mouse IFN-γ (clone XMG1.2), a rat IgG1 antibody. This antibody forms an ELISA pair with R4-6A2 for detection of IFN-γ. Based on the results above, we initially screened several conditions at pH 2.8 and 4.4 (Figure 5a,b). Digestion at pH 2.8 resulted in denaturation and aggregation of antibody and thus was unable to be quantified. A pH of 4.4 produced high efficiency with 50% pepsin, but yield was lower than desired. Therefore, in an effort to increase yield, we kept the 50% pepsin concentration, raised the pH to 5.0, and screened reaction time as a third variable (Figure 5d,e). This increased the yield with minimal loss of efficiency. Interestingly, in both scenarios, the apparent affinity of the F(ab')2 fragment was ~50-fold higher than that of the intact antibody (Figure 5c,f). One potential explanation may be that the removal of the Fc region increases the flexibility of the hinge between the Fab arms, reducing steric strain and improving bivalent binding.22 In summary, this case study indicates that even small modulations in pH can increase yield, especially for antibodies that aggregate at low pH, and that F(ab')2 fragments can sometimes have improved affinity over their parent IgG.</p><!><p>Given the wide range of possible conditions for fragmentation, it is reasonable to ask where one should start when working with a new antibody. If resources are abundant, then a thorough initial screen could be performed over a wide range of conditions (Figure 6), using Design of Experiments to reduce the number of conditions tested. However, often both time and reagents are precious. If a simple first experiment is desired, based on our experience with the antibodies above, we suggest the following four conditions: pH (2.8 and 4.4) x pepsin (10 % and 50 %), with a reaction time of 2 hours. Afterwards further optimization can be performed if the acceptance criteria are not met, using the results of the first screen to point to the next set of conditions. After screening, once a condition with suitable yield, efficiency, and affinity is found, further minor changes to the condition can be made to "dial in" on the exact specifications desired. For example, one could increase digestion time by 10–30 minutes to boost digestion efficiency by 5–10 % depending on exact conditions used.</p><!><p>In this work we have outlined a simple yet systematic strategy to reach desired levels of yield, efficiency, and function of F(ab')2 fragments from small quantities of IgG, including analysis of affinity after enzymatic digestion. We determined that optimization over a range of pHs has the largest impact, likely because it directly impacts the activity of pepsin and the steric constraints of interaction with the active site of the enzyme. pH screening may be followed by or conducted in parallel with optimizing the ratio of pepsin to the antibody. Reaction time had only a small impact and did not always require optimization. This process can be re-iterated until desired yield, efficiency, and function of F(ab')2 fragments is achieved. This general outline for fragmentation has been performed successfully for several antibodies used routinely in our lab. We envision that this work will provide a quick and cost-effective guide for researchers to produce in-house generated antibody fragments, without the need for recombinant expression or haphazard trial and error approaches.</p><!><p>The following procedures used Amicon Ultra 0.5-mL 50-kDa molecular weight cut off (MWCO) centrifuge filters (Fisher Scientific, Waltham, MA) for concentration and buffer exchanges of antibody samples. In preliminary work, we found that PEO-coating was critical to prevent loss of protein to the filter when working with small quantities of antibody; we estimated that up to 10 μg of antibody was lost to an unblocked filter. Therefore, the Amicon filters were coated with 100 kDa polyethylene oxide (PEO) (Fisher Scientific) before use.23 PEO solution (50 mg/mL) was centrifuged in the filter at 14,000 x g for 3 min, then manually rinsed out with DI water. The filters were then centrifuged with DI water for 14,000 G for 3 min, then inverted and centrifuged at 1,000 G for 1 min before use.</p><!><p>The antibody clones used were HRPN, R4-6A2, XMG1.2, GK1.5 (Bio X Cell, West Lebanon, NH), and RA3-6B2 (BioLegend, San Diego, CA). Antibodies were buffer exchanged to the desired pH by using a PEO-treated 50-kDa MWCO centrifuge filter to put them into 100 mM formic acid buffer (pH 2.8 – 3.5) or 20 mM sodium acetate buffer (pH 4.0 – 5.0). These buffers were made in house with formic acid or sodium acetate in DI water and adjusting the pH with HCl or NaOH. All reagents were from Fisher Scientific unless otherwise noted.</p><p>Antibodies were fragmented using immobilized pepsin on 6 % agarose resin (pepsin resin, Fisher Scientific). This pepsin product was reported to have 1–1.5 mg/mL enzyme in a mix containing glycerol storage buffer. The same batch of pepsin was used for all samples (LOT: SB246536, activity 10,012 Units/mL). To prepare the pepsin resin for use, it was removed from storage buffer, pipetted into a 0.5 mL Peirce spin column (Fisher Scientific), centrifuged for 30 s at 100 G, and resuspended in the desired pH buffer (i.e. formic acid buffer or acetate buffer). Next, the immobilized pepsin and the buffer-exchanged antibody solution were mixed together so that the final concentration of antibody was 1 mg/mL. The sample was placed in an incubator at 37 °C on a shaker, shaking gently for the duration of the reaction. After the reaction, fragmented antibody sample was separated from immobilized pepsin using a 0.5-mL Peirce spin column, which allows the antibody solution to pass through while retaining the resin-bound enzyme above the column. The immobilized pepsin resin was rinsed several times with phosphate buffered saline (PBS, no calcium or magnesium). The sample (eluent from the first and subsequent rinses) was collected, then buffer exchanged into PBS and concentrated using a PEO-coated 50-kDa MWCO centrifuge filter. F(ab')2 fragments were confirmed (see Supplementary Fig. S1) via reduction of cysteine disulfide bonds by beta mercaptoethylamine (2-MEA, Fisher Scientific) or affinity purification using anti-Fcγ antibody (Jackson ImmunoResearch, West Grove, PA). Antibody samples were stored at 4°C in PBS until analysis.</p><!><p>Antibody fragmentation samples were analyzed using 4–12 % Bis Tris NuPAGE SDS PAGE (Fisher Scientific) with MOPS buffer (Fisher Scientific). To preserve the structure of antibodies and F(ab')2 during analysis, the samples were not reduced (no beta-mercapto ethanol) and were not heat-treated, unlike standard treatment prior to SDS PAGE. In addition to the fragmentation samples and controls, a standard curve (1 mg/mL – 0.1 mg/mL) of corresponding whole antibody was also run on the gel. A Precision Plus molecular weight ladder (BioRad, Hercules, CA) was included in lane 1. Samples were prepared according to vendor specifications using Lithium dodecyl sulfate (LDS) sample buffer 4X (Fisher Scientific) and DI water. Sample loading was calculated to never exceed 1.5 μg/lane. SDS PAGE was run at 175 V constant for 52 min, which was long enough for the solvent front to reach the end of the gel. Gels were stained with Commassie R-250 (Fisher Scientific) as per vendor instructions.</p><p>Images of gel were collected on a ChemiDoc XRS+ (BioRad), and images were analyzed using Image Lab v5.2.1 (BioRad). Using Image Lab, the integrated stain intensity (area under the curve, AUC) was found for each protein band, and background subtraction and peak selection were performed uniformly across all samples within the same gel (Figure S2). Using the standard curve and molecular weight ladder, mass and molar concentrations of each band were calculated, as well as % yield (Eq 1). Digestion efficiency (Eq 2) and purity (Eq 3) of F(ab')2 were calculated using peak areas. When samples were run in duplicate on the gel, the average AUC was used for analysis.</p><!><p>An ELISA was performed to measure the affinity of antibody fragments for their target protein antigen, as compared to the affinity of whole antibody precursors, by following the procedure outlined in published protocols.24,25 In this assay, a fixed concentration of antibody was equilibrated with antigen in solution, with the antigen at concentrations ranging from below to above the IC50 of binding, and then the mixture was added to a plate coated with the antigen. Antibody that bound to the plate was detected using a labelled secondary antibody. This assay design produces a negatively-sloped sigmoidal curve whose midpoint reports the IC50 of the antibody-antigen interaction. IC50 is expected to be similar to the Kd for the low coating concentrations of antigen used here (SI methods). The detailed procedure was as follows:</p><p>F(ab')2 samples from fragmentation were visually confirmed to be > 98% free of products that contained an Fc region by non-reducing SDS-PAGE before affinity analysis (i.e., > 98 % efficiency). The antigens used were recombinant mouse interferon gamma (IFN-γ, Peprotech, Rocky Hill, NJ) and soluble mouse CD4 protein-his tag (Sino Biological, Wayne, PA). The general procedure was as follows: Serial dilutions of antigen (0 – 256.4 nM, in 2-fold or 5-fold dilutions) were prepared in 1 % BSA and 0.05 % Tween-20 (Fisher Scientific) in PBS (block solution), and then mixed 1:1 v/v with a constant concentration (6.67 nM final concentration) of antibody in block solution (mixed sample). Mixed samples were stored overnight at 4 °C. Separately, the antigen was coated onto a high-binding plate (Corning Costar 96 well 1/2 area, #3690; Fisher Scientific) in PBS overnight at 4°C (IFN-γ at 1–3 μg/mL, CD4 0.2 μg/mL), after which the wells were blocked for 1 hour with block solution. The mixed samples were added to the plate and incubated for 2 hours, then washed. Horseradish peroxidase enzyme (HRP) - conjugated goat anti-rat F(ab')2 secondary antibody (Jackson ImmunoResearch) was added to the plate at 0.5 μg/mL in block solution. All washing steps were performed with 0.05% Tween-20 in PBS. Plates were developed using TMB substrate (Fisher Scientific) and absorbance values were read at 450 nm on a plate reader (CLARIOstar; BMG LabTech, Cary, NC). Sample curves were fit in GraphPad Prism 6 with a sigmoidal 4 parameter curve (Eq 4), where X is log of antigen concentration, Y is absorbance, min and max are the plateaus of the sigmoidal curve on the Y axis, and HillSlope describes the steepness of the slope.</p>

PubMed Author Manuscript

Engineered Synthetic Polymer Nanoparticles as IgG Affinity Ligands

A process for the preparation of an abiotic protein affinity ligand is described. The affinity ligand, a synthetic polymer hydrogel nanoparticle (NP), is formulated with functional groups complementary to the surface presentation of the target protein. An iterative process is used to improve affinity by optimizing the composition and proportion of functional monomers. Since the polymer NPs are formed by a kinetically driven process, the sequence of functional monomers in the polymer chain is not controlled; only the average composition can be adjusted by the stoichiometry of the monomers in the feed. To compensate for this the hydrogel NP is lightly crosslinked resulting in chain flexibility that takes place on a sub millisecond time scale allowing the polymer to \xe2\x80\x9cmap\xe2\x80\x9d onto a protein surface with complementary functionality. In this study, we report a lightly crosslinked (2%) N-isopropyl acrylamide (NIPAm) synthetic polymer NP (50~65 nm) incorporating hydrophobic and carboxylate groups, binds with high affinity to the Fc fragment of IgG. The affinity and amount of NP bound to IgG is pH dependent. The hydrogel NP inhibits protein A binding to the Fc domain at pH 5.5, but not at pH 7.3. A computational analysis was used to identify potential NP-protein interaction sites. Candidates include a NP binding domain that overlaps with the protein A-Fc binding domain at pH 5.5. The computational analysis supports the inhibition experimental results and is attributed to the difference in the charged state of histidine residues. Affinity of the NP (3.5~8.5 nM) to the Fc domain at pH 5.5 is comparable to protein A at pH 7. These results establish that engineered synthetic polymer NPs can be formulated with an intrinsic affinity to a specific domain of a large biomacromolecule.

engineered_synthetic_polymer_nanoparticles_as_igg_affinity_ligands

INTRODUCTION<!>Synthesis and screening of NP-IgG affinity<!>Optimization of NP composition and evaluation of NP binding specificity<!>Effect of pH and buffer on NP-Protein interactions<!>Inhibition of protein A-Fc binding by NP7<!>NP7 Inhibition of protein A binding to IgG<!>Computational study of the Fc surface and analysis of potential NP7 binding sites<!>Modeling of potential NP7 binding sites at neutral pH<!>Modeling of potential NP7 binding sites at low pH<!>Affinity of NP7 to the Fc domain<!>CONCLUSION<!>

<p>Nanomedicine is driven by the premise that discrete synthetic nanoparticles (NPs) can be formulated to target specific proteins, cells or organs. NP targeting coupled with function (drug delivery, imaging, diagnostics, concentration, isolation and purification) provides opportunities for transformative approaches to therapeutics, diagnostics and biomacromolecule isolation and purification. This is a vibrant area of research with recent successes that include therapeutic reagents,1,2 drug delivery vehicles,3–5 sensors,6–8 toxin neutralization9–11 and enzyme inhibition.12,13 NP specificity for target biomolecules is most often accomplished by the attachment of affinity ligands, including antibodies. The need for a comprehensive collection of affinity agents for proteins has been heightened by National Institutes of Health's (NIH's) broad initiative to obtain multiple capture agents for all proteins in the proteome.14</p><p>Recombinant antibodies are the current gold standard of affinity agents and it is likely they will play a dominant role for the foreseeable future. However, antibodies are not without some limitations. For example, the cost of developing new protein capture agents is high. The time required for discovery of an effective antibody can also be lengthy. Some proteins may not function for all intended applications. These and related issues create practical challenges to formulating a comprehensive set of antibody target capture reagents. In addition to antibodies, alternative technologies that include peptides, peptide mimics and aptamers offer promising opportunities to expand the candidate pool of protein capture reagents.15–17 Considering the range of targets and uses, it is likely that a combination of approaches will be needed to generate a comprehensive resource.</p><p>We have been developing an alternative approach for protein and peptide capture agents. Our strategy takes cognizance of the fact that protein-protein interaction surfaces span hundreds of square angstroms.18 Affinity arises from the cumulative effect of individually weak interactions that include van der Waals, hydrogen bonding, and electrostatic interactions. Our capture agent, a synthetic polymer hydrogel, is formulated with functional groups complementary to protein domains or peptide targets. We then use an iterative process to improve affinity to a target peptide or protein by optimizing the composition and proportion of functional monomers. Since the polymer NPs are formed by a kinetically driven process, the sequence of functional monomers in the polymer chain is not controlled; only the average composition of the polymer can be adjusted by the stoichiometry of the monomers in the feed. However, to compensate for this the hydrogel NP is lightly crosslinked (~2%) resulting in considerable chain flexibility that takes place on a sub millisecond time scale19. This allows the polymer to "map" onto a protein surface with complementary functionality compensating in part for the lack of sequence and topological control of the synthetic polymer NP.</p><p>Our previous efforts focused on synthetic polymer NPs with antibody-like affinity and selectivity to a toxic peptide, melittin. Polymer NPs with low nanomolar affinity and high selectivity were developed and were shown to function by neutralizing the peptides toxicity in vitro and in vivo.10,20 The present study describes an important step beyond peptide recognition and capture, specifically, progress in developing a synthetic polymer NP that binds to a specific targeted domain of a large protein.</p><p>The protein target of this study is the 150 kDa protein immunoglobulin G (IgG). IgG is the workhorse protein for research, diagnostics and increasingly, therapeutic applications.21,22 IgGs are composed of 4 protein chains, 2 identical heavy chains of approximately 50 kDa and 2 identical light chains of about 25 kDa. The tetramer has two identical halves that form a forked Y shape. Each end of the fork contains an identical antigen-binding site (the variable regions). The structure is given in Figure 1, which also highlights heavy and light chains. The base or conserved region of the structure corresponds to the Fc fragment (Figure 1) and can be cleaved from IgG and studied independently. Purification of recombinant IgG from cellular extract is often achieved by a multi step process that includes chromatographic separation using protein affinity ligands that strongly and specifically bind to the conserved Fc domain. Protein A or G are the most common affinity ligands used for purification of IgG. Protein A and G themselves are products of recombinant technology and must be expressed and purified. It is not surprising therefore that considerable effort has been expended in the design of alternatives to protein A/G. These alternatives have included synthetic ligands,22,23 DNA/RNA aptamers,16 peptides,15, 17 and synthetic polymers.24,25 The utility of the Fc domain as a target for protein capture dictated its choice for our study.</p><p>A small library of multifunctional polymer NPs containing varying amounts of hydrophobic, hydrophilic and charged monomers were screened for IgG affinity and the "hits" were taken into a second round of NP synthesis, tuning the composition of critical monomers to arrive at NPs with high affinity for IgG. A more focused effort followed to establish that the dominant NP-protein interaction is with the Fc domain of IgG. This effort was coupled with analysis of the Fc protein surface to identify potential Fc-NP binding domains and their response to changes in pH. Finally a NP-protein A competition was performed to help identify the location of the dominant NP-IgG interaction.</p><!><p>A small library of multifunctional hydrogel NPs (NP1-6) were synthesized by precipitation polymerization.10,26,27 NIPAm was the core monomer in combination with various amounts of AAc, APM, and TBAm, as negatively-charged, positively-charged and hydrophobic functional monomers respectively. Bis, 2 mol%) was incorporated as a crosslinker (Figure 2). NP size was established by dynamic light scattering (DLS, Table 1). Table S1 summarizes the feed ratio and reaction conditions used to prepare the NP library. DLS was also used in the initial screen of NP-IgG interactions in water28,29. An association between a NP and IgG (10~12 nm)28 should result in an observable increase in particle size. From this screen, NP6 incorporating both TBAm and AAc monomers showed a significant increase in size following addition of IgG (Table 1). NP1 and 4 also exhibited an increase in size upon addition of IgG.</p><p>The interactions of NP1, 4 and 6 with Fc-orientated IgG were evaluated by QCM in PBS (35 mM, pH= 7.3 with 150 mM NaCl). IgG orientation was achieved by first immobilizing its antigen, TNFα, on the QCM surface. The surface was then functionalized with humanized monoclonal IgG (anti-TNFα antibody).30 Since TNFα binds the Fab fragment of IgG, this orientates the Fc domain into the bulk solution making it accessible to NPs. A 0.1% BSA solution was subsequently added to reduce non-specific binding. To confirm the Fc domain was accessible from solution, the prepared surface was exposed to Au NPs coated with an antihuman IgG which specifically binds to the Fc domain of IgG. The strong signal confirmed that the Fc domain was accessible to solution affinity ligands (Figure S1), validating the use of this method to screen NP affinity for the Fc domain of IgG. The binding of NP1, 4 and 6 to the oriented IgG is summarized in Figure 3. NP 6 showed greater frequency change than NP1 and 4. The NPs are comprised of functional groups that could participate in hydrogen bonding, hydrophobic and electrostatic interactions with proteins. This preliminary result suggests that a combination of these functional groups is necessary for binding.</p><!><p>In an effort to maximize NP affinity to IgG, an expanded pool of NPs was prepared. One series contained greater amounts of AAc than NP6 (NP7 and NP8, Table 1 and S2). However, efforts to prepare NPs with greater amounts of TBAm were not successful due to the colloidal instability of the resulting materials. The binding of NP1, 6, 7, 8 to the oriented IgG is summarized in Figure 3. The frequency change of NP7 (20% AAc, 40% TBAm) was two to three times higher than other NPs. This NP was taken on for further experiments.</p><p>In addition to IgG affinity a preliminary screen for protein specificity was undertaken. The interaction of BSA, TNFα and hemoglobin against NP7 was studied in PBS (35 mM, pH 7.3 with 150 mM NaCl). NP7 had little interaction with these proteins (Figure 4), suggesting that in the above diagnostic, the observed NP affinity for IgG was primarily associated with the oriented Fc domain of IgG. In addition, multivalency of the interaction between NP7 and IgG was established by observing changes in particle size by DLS before and after the addition of IgG to an aqueous solution of NPs. At a molar ratio of IgG to NP of 1.25, aggregates formed (Figure S2). We interpret this observation as multivalency of both IgG and the NPs.</p><!><p>The preceding experiments provided evidence for a strong interaction between NP7 with the Fc domain of IgG. The composition of protein and functional polymer NP surfaces renders these interactions susceptible to binding conditions such as pH and ionic strength. The response to these variables can provide information about the nature of the binding.31–35 The influences of pH and ionic strength on the NP7-Fc interaction are summarized in Figure 5. In phosphate buffer (pH 7.3, without NaCl), an increase in phosphate concentration from 2 mM to 35 mM resulted in a three fold increase in the amount of protein bound to NP7 (Figure 5a, note that pH of the solution did not change even in the low buffer concentration). This implies that under the specified conditions the dominant contribution to binding is from hydrophobic interactions.32 A comparison of Figures 5a and b, reveals the binding between the Fc domain and NP7 is at least twice as strong at pH 5.5 than pH 7.3 in 20 mM phosphate buffer (without NaCl). At pH 7.3, exposed lysine (pKa 10.79) and arginine residues (pKa 12.48) are positively charged but at pH 5.5 and below histidine (pKa 6.04) is also positively charged. A mixture composed mainly of IgG1 and IgG2 are used in these studies. Assuming the five exposed histidine residues on the Fc domain are positively charged at or below pH 5.5, the human IgG1 Fc domain is estimated to have a net charge of +1 at neutral pH but +6 at or below pH 5.5 (see table S3). Similarly, the human IgG2 Fc domain has a net charge of 0 at neutral pH, but a net charge of +6 at or below pH 5.5. On the basis of simple electrostatic arguments, NP7 is expected to have a stronger interaction for the Fc domain at pH 5.5. Some support for this comes from the effect of ionic strength on binding. It was noted that the binding amount at pH 5.5 decreased as the NaCl salt concentration of 35 mM phosphate buffer was raised from 0 mM to 30 mM. This response is typical of a binding with contributions from electrostatic interactions.34,35</p><!><p>Protein A binds to the Fc domain of IgG. This strong and specific affinity is utilized for antibody purification. An X-ray crystal structure of the protein A-Fc complex identifies the protein-protein interaction site and the specific interactions that are responsible for binding.36 In the crystal, fragment B of protein A forms two important contacts with the Fc domain, one predominantly hydrophobic and the second is mainly polar. While our intention was not to directly mimic protein A-Fc binding, we anticipated that knowledge of the strength and location of this interaction would be helpful in analyzing the results of the NP binding studies (vide infra). To this end a competitive inhibition experiment between protein A and NP7 to the Fc domain of IgG under various buffer conditions was carried out. Figure 6 shows the experimental design and results of the competition at pH 7.3 and 5.5. The QCM cell surface was functionalized with Fc fragments. NPs were introduced into the cells at pH 5.5 and 7.3, respectively, until saturation of the signal was observed. Protein A was then injected into the cells and the response to the added protein was monitored. A control experiment was performed following the same procedure except the NPs were omitted. The results show that at pH 7.3, NP7 has little effect on protein A binding to the Fc domain. In contrast, at pH 5.5, preincubation of the Fc domain with NP7 inhibits protein A binding (~ 97% inhibition, Figure 6b column 4). Non-specific binding between NP7 and protein A at pH 7.3 and 5.5 were also considered. We attribute the difference in association between protein A and NP7 to the change in net charge of the NP/protein A binding domain of Fc due to the status of protonation of histidine residues (vide infra). Furthermore, the experimental results indicate that NP7 and protein A can compete for the same binding site at pH 5.5.</p><!><p>The preceding results suggest that NP7 competes with protein A at a discrete Fc domain at pH 5.5. To establish if this result is also observed with an intact IgG, a procedure similar to that described in Figure 6a was carried out only the Fc domain was replaced with IgG in the immobilization step (Experimental Methods). NP7 was injected into the cell with immobilized IgG until signal saturation was observed. A solution of protein A was subsequently injected. The control experiment consisted of the same procedure except the addition of NPs was omitted. At pH 7.3, there was no difference between cells with or without NP7. The results are quite similar to the previous experiment with the Fc fragment, no inhibition (competition) between NP7 with protein A at pH 7.3. In contrast, at pH 5.5, there is little interaction between protein A and Fc after IgG was treated with NP7 (Figure 7, cyan, column 4). We conclude that the NP inhibits protein A binding at pH 5.5. This phenomenon is analogous to the pH-dependent interaction of the neonatal Fc receptor (FcRn) to IgG.37–40 It exhibits high-affinity binding at pH 6.0–6.5 but weaker binding at pH 7.0–7.5. FcRn binds maternal IgG from ingested milk in the gut (pH 6.0–6.5) and delivers it to the bloodstream of the newborn (pH 7.0–7.5). In a process, called transcytosis, IgG proteins are transported across the gut epithelium to enter the bloodstream of the neonate. Many studies38–40 have shown that the pH response is due to histidine residues on the Fc domain. Both the FcRn and protein A binding sites are at the interface between the CH2–CH3 domains (vida infra) of IgG. It is also known that fragment B of protein A inhibits FcRn binding to IgG.41 This indicates there is overlap in the binding domains of FcRn and protein A to the Fc domain of IgG. In addition, research has shown that two histidines39,41 (Fc residues 310 and 433) are responsible for the pH dependent binding, and they are also located on fragment B of the protein A-IgG binding interface. The similar pH response of NP7 and FcRn binding to the Fc domain supports the hypothesis that the pH-dependent stems from interactions with His residues.</p><!><p>To understand the origins of the inhibition of protein A-Fc binding by NP7, the binding surface of the protein A-Fc complex was analyzed from the X-ray crystal structure of the complex. The analysis identifies the protein-protein interaction site and the specific residues that are responsible for binding.36 The interaction occurs in the hinge region between the CH2 and CH3 domains. Upon binding protein A, 660 Å2 of SASA on Fc is covered. Figure 8 presents various views of the Fc dimer with the atoms involved in binding protein A shown in cyan (an atom in Fc was defined as in contact with protein A if it was within 5 Å of any heavy atom of protein A). The experimental structure of protein A bound to Fc (PDB ID: 1FC2) was used to calculate the contacting atoms, and the images in Figure 8 were generated from the apo structure (PDB: 1HZH) using UCSF Chimera.42 Using the 5 Å distance for defining contact, 77 heavy atoms in Fc are in contact with protein A. The interface between protein A and Fc is stabilized mainly by hydrophobic interactions, with relatively few polar contacts.18 The hinge region between the CH2 and CH3 domains of Fc also contains ample polar groups allowing for high affinity binding to biochemically diverse surfaces.</p><p>A computational study of the Fc surface was performed to search for likely NP7 binding regions to assist in interpreting the pH dependence of the protein A-NP competition study. In addition to hydrogen bond donor and acceptor groups, NP7 is composed of hydrophobic and negatively charged groups, thus, the Fc protein surface was scanned for complementary regions (hydrophobic and positively charged) as described in the Experimental Methods. The calculated charge and hydrophobicity scores for all Fc residues are presented in Table S4. To distill the surface region calculations to a single value, the combined score of a surface region was defined as: 2 × charge score + hydrophobic score. The combined score was set to 0 for surface regions with charge score • 0. The SASA, hydrophobic score, charge score, and combined score for all 97 surface regions are presented in Table S5.</p><!><p>The surface region with the highest combined score (+8.8) at neutral pH is centered on Thr289 in the CH2 domain of Fc. This region has SASA of 669 Å2, it includes 5 positively charged residues (Lys274, Lys288, Lys290, Lys292, and Lys301) and only one negatively charged residue (Glu272), giving it a charge score of +4. The hydrophobicity score is +0.8 and the partially exposed hydrophobic residues in the region include: Phe275, Val284, Ala287, Val303, and Val305. Figure 9 depicts a view of the Fc dimer, highlighting the centers of the surface regions with the highest combined scores. Figure 9a depicts all surface regions with a combined score > 0 at neutral pH in green, with increased brightness corresponding to higher values. Notice that, in addition to the brightest green surface (Thr289) many other residues on the CH2 domain are also green. It is likely that these residues in CH2 form a contiguous interaction interface with NP7. These results indicate that protein A (shown as purple ribbons) and NP7 are unlikely to compete for the same binding site at neutral pH. This is consistent with the results of the inhibition study.</p><!><p>The surface region with the highest combined score (+9.4) at low pH is centered on Met252 in the hinge region between the CH2 and CH3 domains of Fc. This region has SASA of 812 Å2, it includes 5 positively charged residues (Lys246, Lys248, Arg255, His310, and His435) and two negatively charged residue (Asp249 and Glu380), giving it a charge score of +3. The hydrophobicity score is +3.4 and exposed hydrophobic surface includes the fully exposed side-chain of Ile253 and the partially exposed side-chains of Met252 and Leu314. Figure 9b depicts all surface regions with a combined score > 0 at low pH in green with increased brightness corresponding to higher values. The surface region centered on Met252 appears in bright green in the area covered by protein A. In addition to Met252, several other residues in the protein A binding region also appear in green. It is likely that these residues in the hinge region form a contiguous interaction interface with NP7. These results suggest that protein A and NP7 are likely to compete for the same binding site at lower pH and are consistent with the data obtained from the binding studies.</p><!><p>To calculate the binding affinity of IgG and NP7, IgG was immobilized on a QCM cell surface and NPs were injected in aliquots until signal saturation was achieved (Figure 10). A frequency change of approximately 3000 Hz was observed corresponding to almost complete coverage of the surface by NP7.43 Washing the cells resulted in a slight diminution of frequency (less than 2%). A solution of the Fc fragment was then added to the cells in aliquots and the resulting change in frequency was monitored. The data was fitted to a Langmuir isotherm from which a Kd of 8.5 nM for the NP-Fc fragment was obtained. A second method where the Fc fragment was immobilized was also used to obtain Kd (Figure S6). The molecular weight of NP7 obtained by static light scattering is 1.02 × 104 kDa (Figure S9). A calculated apparent dissociation constant of 3.5 nM was obtained by nonlinear fitting of the binding isotherm to a Langmuir isotherm. The two dissociation constants measured by these independent experiments are in good agreement. The dissociation constant of protein A is 10 nM17 at pH 7. These results independently indicate that NP7 has a comparable affinity to IgG at pH 5.5 as protein A at pH 7. These results suggest that synthetic polymers with a composition of NP7 have the potential to be used as an affinity media for IgG.</p><!><p>An iterative process was used to identify synthetic polymer NPs (50~65 nm) composed of 40% TBAm, 20% AAc, 2% Bis and 38% NIPAm that binds to the Fc fragment of IgG. The binding was found to be pH sensitive. The binding amount of the synthetic polymer NPs to Fc at pH 5.5 is twice that at pH 7.3 in 20 mM phosphate buffer. The difference in binding is attributed to the difference in the charged state of histidine residues of Fc. At pH 5.5, NP7 can inhibit protein A binding to the Fc domain. However, there was no such competition observed at pH 7.3. A computational analysis identified potential binding domains. At pH 5.5, a binding domain of NP7 overlaps with the protein A-Fc binding domain. In addition, the affinity of NP7 at pH 5.5 is comparable to protein A at pH 7. We have also presented preliminary evidence that NP7 can capture the whole IgG. At pH 5.5, NP7 competes with the binding of protein A to IgG in a similar fashion to that of the Fc fragment. These results suggest that synthetic polymer NPs can be engineered to have an intrinsic affinity to a specific domain of a large biomacromolecule.</p><!><p> ASSOCIATED CONTENT </p><p> SUPPORTING INFORMATION </p><p>Supporting Information Available: Experimental methods for NPs preparation, confirmation of Fc-orientation immobilization, model calculation, competitive binding of protein A to Fc, binding isotherm of NP 7 to the Fc. This material is available free of charge via the Internet at http://pubs.acs.org.</p>

PubMed Author Manuscript

Design and synthesis of new piperidone grafted acetylcholinesterase inhibitors

Alzheimer\xe2\x80\x99s disease (AD) is a neurodegenerative disorder affecting 35 million people worldwide. A common strategy to improve the well-being of AD patients consists on the inhibition of acetylcholinesterase with the concomitant increase of the neurotransmitter acetylcholine at cholinergic synapses. Two series of unreported N-benzylpiperidines 5(a\xe2\x80\x93h) and thiazolopyrimidines 9(a\xe2\x80\x93q) molecules were synthesized and evaluated in vitro for their acetylcholinesterase (AChE) inhibitory activities. Among the newly synthesized compounds, 5h, 9h, 9j, and 9p displayed higher AChE enzyme inhibitory activities than the standard drug, galantamine, with IC50 values of 0.83, 0.98, and 0.73 \xce\xbcM, respectively. Cytotoxicity studies of 5h, 9h, 9j, 9n and 9p on human neuroblastoma cells SH-SY5Y, showed no toxicity up to 40 \xce\xbcM concentration. Molecular docking simulations of the active compounds 5h and 9p disclosed the crucial role of \xcf\x80-\xcf\x80-stacking in their binding interaction to the active site AChE enzyme. The presented compounds have potential as AChE inhibitors and potential AD drugs.

design_and_synthesis_of_new_piperidone_grafted_acetylcholinesterase_inhibitors

<p>Alzheimer's disease (AD) is a prevalent, irreversible neurodegenerative disorder that affects more than 35 million people worldwide. It is estimated that AD will be the cause of more than 700,000 deaths in 2016, making it one of the leading causes of death in the United States.1 Despite the tremendous efforts in search for novel and potent therapeutic agents, the pipeline of available treatments is low.2,3 The etiology of AD is not completely understood. However, the appearance of extracellular β-amyloid plaques and the imbalance of acetylcholine (ACh) neurotransmitters (due to the severe loss of cholinergic neurons) are two major hallmarks of the disease.4,5 The cholinergic system is severely affected in AD patients, leading to a reduced concentration of the neurotransmitter acetylcholine (ACh) at the synaptic clefts.6 Acetylcholinesterase (AChE), the enzyme responsible for degradation and regulation of acetylcholine in the human body, has been actively investigated as a target for AD treatments.7 Previous reports suggest that increasing acetylcholine levels by inhibiting AChE activity partially restores the substantial impairment of memory and cognitive dysfunctions.8 In fact, the main AD treatments are cholinesterase inhibitors, such as Donepezil, that work by increasing acetylcholine concentration at synaptic clefts.9,10 Therefore, compounds that can inhibit the action of AChE could rival Donepezil as promising AD treatments.</p><p>The active site of AChE is located at the bottom of a 20 Å long, narrow gorge comprised of five regions: the catalytic triad (composed of Glu327, His440 and Ser200), the oxyanion hole (Gly121, Gly122 and Ala201), the choline binding site (Trp84 and Phe330), the acyl binding pocket (Phe288 and Phe290), and the peripheral anionic site (PAS) (Trp279, Phe330 and Tyr70).11–14 The PAS of AChE catalyzes a series of reactions on the Aβ fibrils that lead to the formation of β-sheets with high aggregating potential.15–17 AChE inhibitors that bind to the PAS of the enzyme decrease aggregation rates of amyloid peptides and facilitate their clearance when compared to inhibitors that only bind the catalytic triad of the enzyme.17–19 The AChE active site is highly hydrophobic20 and it is hypothesized molecules with multiple aromatic cores can bind to it more effectively21,22 due to hydrophobic interactions.23,24 Studies also showed that compounds having α, β-unsaturated moiety or the thiazolopyrimidine core structure are capable of inhibiting AChE.23,25–27 Moreover, the N-benzylpiperidine moiety is a structure found in donepezil, a FDA-approved drug for treatment of mild to moderate AD.28 The crystal structure of AChE in complex with donepezil, revealed a π-π stacking as well as hydrogen bonding interaction of N-benzylpiperidine to the active site of AChE.29 Inspired by these precedents, and the characteristics of the enzyme's binding pocket, we used computer modeling to design potential inhibitors of the AChE enzyme. In this study, we report the synthesis, AChE inhibitory activity, and toxicity towards human neuroblastoma SH-SY5Y cells of these new N-benzylpiperidines and thiazolopyrimidines derivatives. Finally, the plausible binding interactions of the most active derivatives were investigated using molecular docking.</p><p>Molecular modeling techniques were employed for designing the potential cores of various inhibitors. The energy minimizations of the different molecules lead us to two potential cores; 5 and 9. The in silico studies showed the critical role of (i) the benzylamine moiety in 5(a–h) and (ii) sulfur atom in 9(a–q) to form favorable π-stacking and hydrogen bonding interactions to the active site residues of AChE enzyme, respectively.</p><p>The desired molecules can be prepared from the same precursor molecule 3. The common α,β-unsaturated ketone precursors (3) were prepared by the Claisen-Schmidt condensation of 4-piperi-done hydrochloride (1) with aromatic aldehydes (2) in the presence of HCl in acetic acid (Scheme 1).30 The reaction of the 3 with benzyl bromides (4) under basic conditions provided N-benzylpiperidines 5(a–h). Also, the unsaturated ketone 3 was reacted with thiourea (6) in presence of sodium ethoxide24 to yield 7. These compounds were reacted with α-Bromoacetophenones 8 (a–d) in refluxing ethanol for 12 h to afford thiazolopyrimidinium salts 9(a–q).</p><p>The synthesis of the thiazolopyrimidinium molecules (9) can yield two possible isomers (see SI). Thus, we prepared 9q and studied its spectroscopic properties to assign the correct stereochemistry (Fig. 1). In the 1H NMR spectrum of 9q,31 the one hydrogen singlet at 6.27 ppm is due to H-5 and assigned to the carbon peak at 63.2 ppm using HMQC. This proton showed HMBC correlations to C-6 at 42.8 ppm besides showing correlation with C-5a and C-3 at 106.6 and 140.0 ppm, respectively. From HMQC, the doublets at 3.38 and 4.08 ppm (J =16.5 Hz) are assigned to H-6a and H-6b whilst the other doublets at 4.27 and 4.42 ppm (J =14.5 Hz) can be attributed to H-8a and H-8b. The singlet at 7.43 ppm is due to H-12 and H-2, which showed HMBC correlation to C=N at 163.8 ppm. The aromatic protons for ring A showed up as two singlets at 6.01 and 6.40 ppm in proton NMR having HMQC correlation with corresponding carbons at 102.0 and 112.3 ppm in 13C NMR. On the same basis, the protons of ring B were elucidated as two singlets at 6.52 and 6.55 ppm in 1H NMR, having correlation to carbons at 112.3 and 101.6 ppm through HMQC. The 1H and 13C chemical shifts as well as selected ROESY and HMBC correlations of 9q are depicted in Figs. 1 and 2a/b, respectively. The infrared spectroscopy (IR) for 9q showed three medium peaks at 3469 cm−1, 1654 cm−1 and 755 cm−1 specific for free NH, C=N and C–S moieties, respectively. This data helped us to conclude the spatial arrangement of the atoms in 9q (and all the other analogues).</p><p>All the newly synthesized N-benzylpiperidines 5 and thiazolopyrimidinium bromide salts 9 were evaluated for their inhibitory activity against AChE (Table 1) using the Ellman's method,32 which uses acetylthiocholine and 5,5′-dithio-bis-2-nitrobenzoic acid (DTNB). In presence of AChE enzyme, the reaction produces the yellow compound 5-thio-2-nitrobenzoate. AChE inhibitors decrease the rate of the reaction resulting in a less colored solution, the absorption difference is correlated to inhibitory activity (see SI). Compound 9p showed the highest inhibitory activity among both series, with an IC50 value of 0.73 μM, nearly two fold better than the standard drug (IC50 = 1.31 μM). Other active compounds are: 5h, 9h, 9n, and 9j with IC50 values of 0.83, 0.98, 1.01 and 1.09 μM. Compounds bearing 2-naphtyl moiety (except for 5f and 9d), showed good IC50 values suggesting the importance of π-π interactions. Based on the above results, it can be deduced that those compounds having extended aromatic systems on either R1 or R3 positions, may display more van der Waals interactions on the active site of AChE enzyme. It was also noticed that compounds 5h and 9p, bearing bromine atoms, showed the highest inhibitory potencies among their series. This may suggest an electronic effect due to the halide. Studies are undergoing to decipher the role electronically diverse substituents have on AChE inhibitory activity.</p><p>The cytotoxicity of the most active inhibitors, namely 5h, 9h, 9j, 9n and 9p, was evaluated on the neuronal cell line SH-SY5Y using the MTT assay, which measures the cell metabolic activity. The assay was performed after 24 h incubation of neuronal cells with concentrations ranging from 0.2 to 40 μM. The molecules showed significant cell viability at this range of concentrations (Fig. 3).</p><p>The most active compounds from each series (5h and 9p) were docked into the groove of the AChE enzyme. Compound 5h (Fig. 4) mainly displayed interaction to the amino acids composing the PAS of the enzyme. As predicted, the main interactions occur between Tyr 70, Trp 279 with the 2-naphthyl moiety of 5h (distances of 3.69 Å and 3.36 Å). The B ring of the molecules also interacts with Phe 330 and Tyr 334 (3.31 Å and 3.67 Å). Compound 9p also sits on the PAS of the AChE enzyme (Fig. 5). The amide nitrogens of Phe 288 and Arg display a H-bond type interaction to the sulfur of the thiazoline ring with contact distance of 2.02 Å and 2.33 Å. Phe 331 displays a strong "T-shaped" stacking interaction to ring A at the left side of the molecule (3.31 Å). Trp 279 shows a "T-shaped" stacking with ring B of the molecule (3.49 Å). Finally, there is a potential cation-π interaction between the piperidine ring and Tyr 334 (3.58 Å).</p><p>In conclusion, two series of piperidone grafted heterocyclic compounds were prepared employing efficient synthetic methodologies in pure form and were tested for their inhibitory activity against AChE enzyme. Compound 5h and 9p displayed the highest inhibitory activities among their series with IC50 values of 0.73 and 0.83 μM, respectively. The most active compounds did not show cell toxicity at the tested concentrations. Molecular modeling disclosed binding interaction patterns to the active site of AChE suggesting the positive effect of aromatic core multiplicity.</p>

PubMed Author Manuscript

Is Nostoc H-NOX an NO sensor or Redox Switch?

Nostoc sp (Ns) H-NOX is a hemeprotein found in symbiotic cyanobacteria, which has ~35% sequence identity and high structural homology to the \xce\xb2 subunit of soluble guanylyl cyclase (sGC), suggesting a NO sensing function. However, UV-Vis, EPR, NIR MCD, and ligand binding experiments with ferrous and ferric Ns H-NOX indicate significant functional differences between Ns H-NOX and sGC. (1) After NO binding to sGC, the proximal histidine dissociates from the heme iron, causing a conformational change that triggers activation of sGC. In contrast, formation of pentacoordinate (5c) NO heme occurs to only a limited extent in Ns H-NOX, even at > 1mM NO. (2) Unlike sGC, two different hexacoordinate (6c)-NO complexes are formed in Ns H-NOX with initial and final absorbance peaks at 418 nm and 414 nm, and the conversion rate is linearly dependent on [NO] indicating that a second NO binds transiently to catalyze formation of the 414 nm species. (3) sGC is insensitive to oxygen, and ferric sGC prepared by ferricyanide oxidation has a 5c high-spin heme complex. In contrast, Ns H-NOX autooxidizes in 24 hr if exposed to air and forms a 6c ferric heme complex indicating a major conformational change after oxidation and coordination by a second histidine side chain. Such a large conformational transition suggests that Ns H-NOX could function as either a redox or a NO sensor in the cyanobacterium.

is_nostoc_h-nox_an_no_sensor_or_redox_switch?

<!>Materials<!>Protein expression and purification<!>Pyridine hemochromogen assay<!>Determination of binding equilibrium constant by optical titration<!>Stopped-flow experiments<!>Laser flow-flash kinetic measurements<!>EPR spectrometry<!>Near infrared magnetic circular dichroism (NIR MCD)<!>Far-UV CD<!>Heme and protein quantitation<!>Autooxidation of Ns H-NOX<!>Spectroscopic and kinetic characterization of CO binding<!>Characterization of NO binding mechanism<!>EPR characterization of resting Ns H-NOX and its complex with NO<!>Identification of axial ligands in oxidized H-NOX<!>Near Infrared MCD (NIR MCD)<!>Protein secondary structure changes with heme redox state transition<!>Characterization of cyanide and azide binding to ferric Ns H-NOX<!>Comparisons between the ligand binding properties of sGC and Ns H-NOX<!>Structural interpretations<!>Physiological function of H-NOX<!>

<p>The chemical versatility of heme for reversible coordination of small molecule substrates/activators and redox reactions has been adopted by living cells in all kingdoms for a variety of physiological functions, including gas storage and transport, electron transfer, enzyme catalysis, sensing of diatomic gases, monitoring cellular redox potential, and intra- and intercellular signaling (1–3). With the advance of efficient genomic and proteomic screening in the last decade, increasing numbers of heme proteins involved in O2, CO, and NO sensing have been identified (4–7). Such signal transduction heme proteins are frequently expressed as two-component systems, usually including one heme-sensor and one transducer (or effector) component, sometimes as domains on the same polypeptide or as separate gene products. Signaling transducer proteins or domains include kinases, phosphodiesterases, chemotaxis machinery and DNA-binding proteins (4, 5, 8, 9).</p><p>Soluble guanylyl cyclase (sGC) is the major nitric oxide (NO) sensing heme protein for mammals, and its enzyme activity is increased ≥100-fold upon NO binding to the heme containing subunit (10, 11). Determination of the exact structural mechanism for the coupling between NO binding to the heme in the β subunit and subsequent activation of cyclase activity in the partner domain remains a challenging problem, in part because there are no crystal structures of either the heme domain or full-length mammalian sGC. However, structural and mechanistic studies of homologous proteins can provide useful insights into the activation mechanism of human sGC and other sensors in response to NO. Genomic screening has identified several microbial heme protein sensor domains, called H-NOX (Heme-Nitric oxide and OXygen binding) or SONO (Sensor Of NO), that have significant sequence identity with the human sGC β subunit and are potential model systems for studying the kinetics and structural changes associated with ligand binding to the heme domain of sGC (8, 12, 13). The crystal structures of sensor proteins from Thermoanaerobacter tengcongensis (Tt H-NOX) and Nostoc sp (Ns H-NOX) have been determined and provided mechanistic insights into the reactions of sGC with NO (12–14). Ns H-NOX appears to be a better sGC model than Tt H-NOX because of its higher sequence identity (33% vs. 18%) and similar apparent in vivo function (12, 14). The Tt H-NOX domain appears to be an oxygen sensor, whereas Ns H-NOX, like sGC, does not appear to bind O2 directly but is capable of binding both NO and CO. The key residues responsible for ligand discrimination by sGC and homologues have been shown to locate primarily in the distal heme-binding pocket and are significantly closer in identity between Ns H-NOX and sGC than for any other bacterial homolog (12).</p><p>We have examined in detail the kinetics of NO, CO, and O2 binding to Ns H-NOX using a variety of spectroscopic, rapid mixing and flash photolysis methods to determine the functional similarity of this bacterial protein to sGC. Ns H-NOX is capable of binding NO and CO rapidly, but the binding mechanisms, particularly for NO, show differences from those observed for sGC. Most significantly Ns H-NOX forms a predominantly 6-c NO heme complex whereas sGC immediately forms the 5-c state resulting from cleavage of the proximal histidine, which is required for activation of the catalytic domain. The heme group in Ns H-NOX autooxidizes, albeit slowly over hours, whereas sGC is insensitive to oxygen for days at a time. The latter in vitro observation contrasts with long-standing evidence that sGC undergoes oxidation in vivo during certain disease states (15, 16). Oxidation of sGC results in loss of activity and proteosomal degradation, however, recent heme-independent activators that target oxidized sGC have been shown to restore sGC activity in vivo (17). These observations warrant a reexamination of redox regulation of sGC and H-NOX enzymatic activity previously deemed an artifact.</p><p>Our results explore and attempt to explain the differences between sGC and Ns H-NOX, shed light on the function of Ns H-NOX, and extend the context in which we not only can explore sGC function but also new functions in a wider scope. In agreement with and analogous to the wide range of roles played by heme protein-sensors, it is likely that, in addition to sensing NO, Ns H-NOX also monitors the general redox potential of the bacterium's environment.</p><!><p>Carbon monoxide and nitric oxide gas were from Matheson-TriGas and nitrous and nitric acid was removed from the latter by passing the gas through a NaOH trap. Potassium ferricyanide, sodium hydrosulfite (~85%), sodium nitrite (97%), sodium azide, potassium cyanide, imidazole, bovine heart cytochrome c, deuterium water, dithiothreitol (DTT), β-mercaptoethanol (β-BME), isopropyl-1-thio-β-D-galactopyranoside (IPTG), δ-Aminolevulinic acid, phenylmethyl-sulfonyl fluoride (PMSF), glycerol, EDTA were purchased from Sigma-Aldrich. Bicinchoninic acid reagents (BCA reagents) are from CalBiochem. 1M cyanide solution, ~pH 9, was prepared by titration of potassium cyanide by 6N HCl at 0 °C on ice. Other chemicals are all reagent grade.</p><!><p>Ns H-NOX 1–183 in pET22b was expressed in Rosetta DE3 pLysS cells (Novagen). Freshly transformed colonies were grown to an OD of 0.4–0.5 in Terrific Broth with appropriate antibiotics at 37°C. Growth temperature was adjusted to 30 °C and cultures were induced by addition of IPTG to 20 µM and δ-Aminolevulinic acid to 2 mM followed by incubation overnight. Cells were harvested by centrifugation at 5000 × g. Pellets were resuspended in 50 mM NaCl H-NOX Buffer (20 mM TRIS, 5 mM β-BME, 2 mM PMSF) and the suspensions were flash frozen in liquid nitrogen and stored at −80 °C.</p><p>Cells were thawed and frozen 3 times; solid DTT was added to 100 mM; the suspension was sonicated until homogenous; and the final sample centrifuged at 20,000 × g for 20 min at 4°C. Ns H-NOX protein was purified as described previously (14). The cleared H-NOX lysate was purified sequentially over 1) Hi-trap, 2) Superdex 200, 3) Mono Q, 4) Superdex 75 columns11. Peak fractions with A420/A280 ratio greater than 2.0 were pooled and concentrated. Protein concentration was determined by bicinchoninic acid, or BCA method (18), using 1 mg/ml human serum albumin as standard.</p><!><p>Heme content was determined by the formation of pyridine hemochromogen as previously described (19). In brief, the spectrum of ~ 100 µg isolated fresh ferrous Ns H-NOX or oxidized sample containing 0.15 M NaOH, 1.8 M pyridine was recorded and then few grains of solid dithionite were added. The total heme content was determined from the reduced minus oxidized difference spectrum of bis-pyridine heme using ΔA556−538 nm = 24 mM−1 cm−1. A metmyoglobin standard (A409 = 157 mM−1cm−1) was run in parallel as a control.</p><!><p>Equilibrium dissociation constants were estimated as the ligand concentration at the half maximal absorbance in titration experiments which monitored the fractional absorbance difference between the initial ligand free Ns H-NOX and the sensor fully saturated by the ligand. The binding isotherm between delta absorbance and ligand concentration constructed by optical titration with serial additions of the ligand were fit to a single-binding site hyperbolic function: ΔAi = ΔAmax [L]i/ (Kd + [L]i), where ΔAi is the absorbance difference after the ith addition of ligand, ΔAmax is the maximal absorbance change when the protein was saturated with ligand, Kd is the equilibrium dissociation constant, and [L]i is the free ligand concentration after the ith addition. Volume changes due to ligand addition were corrected for and were no more than 5%. All titrations were conducted in 50 mM HEPES, pH 7.7 containing 10% glycerol and 0.1 M NaCl.</p><!><p>In most cases association and dissociation rate constants, kon and koff, respectively, were determined using rapid mixing measurements in an Applied Photophysics model SX-18MV stopped-flow instrument with a rapid-scan diode-array accessory (20). Kinetic measurements were normally conducted under pseudo first-order conditions with ligand concentration at least an order of magnitude greater than that of the protein. Ligand binding to the ferric sensor was done aerobically at 1:1 (V/V) equal mixing. For anaerobic experiments with NO and CO binding, the fluid channels in the stopped-flow apparatus were incubated with a dithionite solutions for several hours and then rinsed with nitrogen-saturated buffer. Reactants were rendered anaerobic by 5 cycles of vacuum/argon replacement in a tonometer. The reaction time courses were analyzed by nonlinear regression to single or multiple exponential functions. Estimated kon and koff values were derived from the slope and y-intercept, respectively, in plots of kobs versus ligand concentrations.</p><!><p>The kinetics of NO binding to Ns H-NOX were measured by flow-flash methods as described previously by Salter et al. in studies with Cerebratulus lacteus hemoglobin (Hb) (21). An anaerobic solution of reduced protein in complex with carbon monoxide (0.1 mM CO) was rapidly mixed at 1:1 ratio with a buffer containing 2 mM NO. Then, ~50 ms after flow stopped, the CO complex of the Ns H-NOX was photolyzed by a 0.5 µs excitation laser pulse at 577 nm (phase-R 2100 dye laser). The kinetics of NO association to the reduced protein was monitored at 436 nm and 20 °C. Rapid mixing of solutions was controlled by a Bio-Logic stopped-flow module SFM 400 (Molecular Kinetics, Inc. Indianapolis, IN).</p><!><p>EPR spectra were recorded at liquid helium or liquid nitrogen temperatures on a Bruker EMX spectrometer (22). Data analysis and spectral simulations were conducted using WinEPR and SimFonia programs furnished with the EMX system. The conditions for liquid nitrogen EPR measurements were: frequency, 9.29 GHz; modulation amplitude, 2 G; modulation frequency, 100 kHz; and time constant, 0.33 s. Liquid helium EPR conditions were the same, except with frequency, 9.61 GHz; and modulation amplitude, 2 or 10 G. The microwave power dependence was fit by nonlinear regression to the equation: log (S/P1/2) = −b/2 log (P1/2 + P) + b/2 log(P1/2) + log K, where P is the power, S is the peak to trough amplitude of the EPR signal, P1/2 is the power at half-saturation, and b and K are floating parameters with b = 1 for non-homogeneous saturation (23).</p><!><p>MCD spectra between 800 and 2000 nm were acquired in a JASCO J730 using a 150 W tungsten-halogen lamp and a liquid nitrogen cooled high sensitivity InSb detector equipped with a 1.4 T electromagnet. CD wavelength and sensitivity was calibrated by 1:1 mixture of 0.24 M NiSO4 and 0.36 M K+,Na+−d-tartrate. All measurements were performed at room temperature (24 °C) at a bandwidth of 10 nm, 1 s time constant, 0.5 nm step resolution from 2000 to 800 nm at 200 nm/min scan speed. MCD calculation from H+ (CD + MCD) and H− (CD – MCD) and adjusted for molar delta absorption coefficient: ΔA(M cm T)−1 was done by the Spectral Analysis software that came with the J730 system provided by JASCO.</p><!><p>CD spectra between 190 to 290 nm were recorded in a JASCO J815 using a 150 W xenon lamp. All measurements were performed at room temperature (24 °C) at a spectral bandwidth of 1 nm, 0.5 s time constant, 0.2 nm step resolution at 200 nm/min scan speed. 12 repetitive scans with 2 mm path quartz cuvette were recorded for each protein sample as well as the buffer control.</p><!><p>Total heme concentration of isolated Ns H-NOX protein was quantified by pyridine hemochrome assay. A fresh ferrous Ns H-NOX sample yielded a molar absorbance coefficient of 170 mM−1cm−1 at 429 nm. The molar absorption coefficient for the ferric Soret peak at 414 nm was determined to be 136 mM−1cm−1. Protein concentration determined by the BCA method and combined with the pyridine hemochromogen assay gave a heme stoichiometry of 1.05 ± 0.13 (n = 3) per protein monomer. The ratio of the 414/280 nm for Fe(III) Ns H-NOX is 2:1 and the ratio of 429/280 for the ferrous form is 3:1. The purified protein appears to be > 90% homogeneous based on SDS PAGE analysis and Superdex 75 chromatographic analysis.</p><!><p>Heterodimeric full length sGC is relatively insensitive to autooxidation in the presence of air or 1 atm oxygen (24). When we placed an aerobic solution of the Ns H-NOX sensor, which had been stored aerobically in 5 mM DTT, in a tonometer and purged the gas space with argon, there was a ~3 nm red-shift of the Soret peak from 426 to 429 nm, suggesting that the original sample was partially oxidized and then was re-reduced anaerobically with DTT. To examine autooxidation more directly, a reduced preparation of Ns H-NOX with a 429 nm Soret peak was gel filtered to remove all DTT from the storage buffer, and then monitored for oxidation by recording its optical spectral at 24 °C over a period of two days. A Soret band shift from 429 nm to 414 nm occurs slowly with a discrete isosbestic point at 419 nm, indicating conversion from Fe(II) to Fe(III) with autooxidation at a rate ≈ 0.05 h−1. This value is similar to the rate of autooxidation of oxymyoglobin (MbO2) and about 4-fold greater than that for human oxyhemoglobin (HbO2) (25, 26) (Fig. 1). Thus the Ns H-NOX heme iron is capable of donating electrons to O2, albeit very slowly. However, there is no observable Fe(II)O2 intermediate during this process, suggesting that the affinity of Ns H-NOX for O2 must be very poor, i.e. P50 ≫ 1000 µM, presumably due to a very large dissociation rate constant.</p><!><p>We characterized CO and NO binding to Ns H-NOX for direct comparison with the ligand binding properties of sGC (27–29). CO binding to ferrous Ns H-NOX shifts the Soret peak from 429 nm to 422–423 nm (14) and is a simple one-step reaction with no spectral intermediates when examined by rapid scanning, stopped-flow spectrophotometry. Time courses at 420 nm were measured under pseudo first-order reaction conditions using 5 µM ferrous Ns H-NOX and varying [CO] from 25 to 166 µM at 24 °C (Fig. 2a). Observed rates were obtained by fitting to one-exponential expressions, and kobs depends linearly on [CO], with a calculated second-order rate constant = 2 – 3 × 106 M−1s−1 based on data for two separately expressed and purified Ns H-NOX samples (Fig. 2a, Inset). The total absorbance changes obtained with seven different CO concentrations were the same indicating that complete saturation with ligand was achieved and that the Kd for CO binding is < 25 µM. The y-axis intercept of the kobs versus [CO] gave an unusually high value of 23 s−1 for the CO dissociation rate constant (koff), which when combined with an association rate constant of 2.9 µM−1s−1, gives an estimated Kd for CO binding of ≤ ~8 µM.</p><p>We attempted to measure the CO dissociation rate constant more directly by mixing the CO form with excess NO, which is normally the best method to measure koff for the CO complexes of hemeproteins. When the CO complex of Ns H-NOX is mixed with 1 mM NO in the stopped-flow diode-array spectrophotometer (Fig. 2b), a moderately rapid decrease in absorbance at 422 nm occurs and is accompanied by a blue shift of the Soret peak to 418 nm, indicative of displacement of CO and formation of a six coordinate NO complex. On longer time scales, additional absorbance decreases at both 422 and 418 nm, and a further blue shift of the NO-heme peak to 414 nm are observed. Thus, the absorbance decrease at 422 nm is biphasic. The large (80%) initial phase occurs with a first order rate of 3.6 s−1, which is independent of [NO]. Surprisingly, the rate of the second, slower phase shows a linear dependence on [NO]. Similar rates and dependences on [NO] were obtained from global analysis of complete sets of spectra from the diode array system (not shown), and all of these data are similar to the slow phases observed for excess NO binding to the reduced protein (Fig. 3).</p><p>The initial dominant phase for the NO displacement reaction is first order and represents the conversion from Fe(II)CO to the 6c Fe(II)NO complex (418 nm). The secondary slower and smaller phases appears to represent interconversion of a 6c NO bound H-NOX with a peak at 418 to another 6-c NO complex with a peak at 414 nm and a small amount of 5c-NO complex, as indicated by the shoulder at 399 nm in the final spectrum (Fig. 2b). Regardless of the interpretation of the slow phase, the fast phase provides a direct measure of CO dissociation with koff = 3.6 s−1. Re-binding of excess CO to these altered NO complexes is unlikely because the overall equilibrium affinity of Ns H-NOX for NO is 4 orders of magnitude greater than that of CO (Table 1). The estimated CO dissociation rate from the fast NO displacement phase, koff = 3.6 s−1, can be used as the Y-intercept in the linear fit for plots of kobs versus [CO] in the simple CO binding data shown in Fig. 2a, inset, and the resultant slope gives an association rate constant for CO binding, kon, equal to 3.0 µM−1s−1. The ratio of these values, 3.6 s-1/3.0 µM−1s−1 provides an estimate of 1.2 µM for the Kd for CO binding to Ns H-NOX, which is compatible with equilibrium titration data and previous measurements for H-NOX proteins.</p><!><p>NO binding to Ns H-NOX heme was studied using rapid mixing stopped-flow and flow-flash laser photolysis approaches to examine all time scales. Rapid-scan diode array measurements of NO binding to Fe(II) Ns H-NOX were conducted at low [NO] using either 1:1 protein:NO mixtures or a 5-fold excess of ligand. When only a stoichiometric amount of NO is present, the reaction shows a single rapid phase, with a Soret peak shift from 429 nm to 418 nm and an isosbestic point at 422 nm. Most of the reaction, > 90%, was lost in the 1.5-ms dead-time of the stopped-flow apparatus (Fig. 3a, top). At a 5-fold excess of NO, the initial rapid phase is complete in the dead time, and a second slower phase occurs on milliseconds to second time scales, involving a blue shift in the Soret band from 418 nm to 414 nm. Eventually, two bands form at 414 and 399 nm on seconds to minute time scales and the 399 nm band indicates formation of significant amount of a 5c NO-heme complex (Fig. 3a, bottom). Most of these secondary absorbance changes appear to represent a transition from one 6c-NO complex with a "normal" peak at 418 nm to a second 6c-NO complex with a peak at 414 nm. The rate of decay of the initial 6c-NO complex to the species with a peak at 414 nm depends linearly on [NO] (Fig. 3b, inset), with a slope equal to ~2.0 µM−1s−1 and a y-intercept close to 0. The rate-limiting step for this interconversion of 6c NO-heme complexes appears to require transient binding of a second NO molecule.</p><p>This second transition is followed by a third process, which is first order and involves a slow biphasic conversion of the 6c-NO complexes to 5c-NO complexes, with first order rate constants equal to ~7 and 0.3 s−1, respectively (Fig. 3b). Only ~40% conversion to a 5c-NO complex with a peak at 399 nm is observed even at [NO] ≈ 1000 µM, and the rate of conversion to the 5c species does not appear to depend on [NO] (Fig. 3b, longer time scale).</p><p>In order to determine the association rate constant for the formation of the first 6c-NO complex accurately, we performed flow flash experiments. The CO form of ferrous Ns H-NOX was rapidly mixed with NO then photolyzed with a 0.5 µs laser pulse at 577 nm to generate transiently the unliganded ferrous form of Ns H-NOX. The binding of NO to unliganded H-NOX was then monitored at 436 and 418 nm under pseudo first-order conditions (i.e. [NO]≫[heme]). The dependence of the observed rates for these ultrafast phases on [NO] gave an association rate constant equal to 200 – 400 µM−1s−1, at 24 °C (Fig. 4a), which indicates an open, highly accessible active site. Thus, the initial NO binding step is very rapid and almost diffusion controlled, but in the presence of excess NO this species is relatively short-lived and decays to a second 6c NO complex with a peak at 414 nm.</p><p>Because the second Ns H-NOX 6c-NO complex appears to require additional NO, we attempted to generate only the first 6c-NO complex with a Soret peak at 418 nm by adding a slightly higher than stoichiometric amount of NO anaerobically and then measuring the rate of NO dissociation from this initial complex by mixing it with 0.5 mM CO containing 25 mM dithionite as an NO scavenger. The high concentration of dithionite scavenges any free NO preventing its rebinding, and the rate limiting step for the formation of the CO complex is the initial rate of NO dissociation from the 6c complex. The formation of the CO complex was measured at 423 nm and indicated displacement of bound NO (Fig. 4b). Time courses measured at either 423 or 400 nm were biphasic and gave similar first order rate constants, 0.05 and 0.007 s−1, respectively for the prominent fast and minor slow phases. Assuming that koff ≈ 0.05 s−1 and kon ≈ 300 µM−1s−1 apply for the formation of the initial 6c-NO complex, the apparent Kd for first NO binding step would be ~0.17 nM, which is roughly 100-fold higher than the Kd for NO binding to most mammalian Hbs and Mbs (30).</p><!><p>The EPR spectrum of the purified H-NOX in the presence of DTT is shown in Fig. 5. The sensor as isolated is mostly ferrous, but some oxidized high and low-spin heme signals are visible at ~ g=6 and ~2.3, respectively, together with some nonspecific iron signals at g= 4.3. The ferric signals are probably a consequence of slow oxidation during storage in air, even in the presence of DTT. During autooxidation in air, the two heme signals increase substantially with more than half the integrated intensity coming from the low-spin Fe(III) form. The high-spin signal around g=6 is heterogeneous and has both rhombic and axial components. The low-spin heme signals at g=3.00 and 2.26 are more visible, and the third g component predicted to be at g=1.37 is not quite visible because it is very broad (31). Further treatment of H-NOX by ferricyanide to oxidize the sensor completely increases the amplitudes but does not change the relative amounts of the high and low spin components. The proportion of the high-spin to low-spin heme in the ferric Ns H-NOX changes from batch to batch of purified protein with the low-spin species dominating and in some cases accounts for essentially 100 percent of the sensor as estimated by the MCD signal amplitude of the Soret band and ligand-to-metal charge-transfer band around 600 nm (Tsai et al., unpublished results) (32). Addition of excess imidazole (50 mM) to the ferric sensor abolished the high-spin heme signals and led to a homogeneous low-spin 6c bis-imidazole heme complexes with three principal g values at 2.94, 2.31 and 1.42 (Fig. 5).</p><p>When 80 µM ferrous Ns H-NOX was mixed with a stoichiometric amount of NO, only a low-spin NO-heme complex EPR was observed in the g = 2 region, without any other observable paramagnetic species (Fig. 5, bottom spectrum). The reduced NO-heme spectrum is dominated by a 6c low-spin signal with 3 × 3 hyperfine/super hyperfine features contributed by the two axial nitrogen nuclei, one from the N atom of bound nitric oxide and the other from the Nε atom of the proximal histidine (Fig. 6, bottom solid spectrum). The EPR spectrum can be simulated by two nitrogen nuclei with 24 G isotropic hyperfine (nitric oxide N), 6 G isotropic super hyperfine splitting (histidine Nε) , a three principal g values of 2.080, 1.975 and 2.003 for gx, gy and gz, respectively, and an additional gy' component (Fig. 6, bottom dached spectrum). The optimal hyperfine splittings for the gy and gx obtained by simulation are 10 and 12 Gauss, respectively. The intrinsic line width of the gy component is relatively large in comparison to the other two g components. This 6c NO heme complex forms very rapidly at ratios of heme to NO of 1:1 (data not shown) and 2:1 and is stable for extended periods of time under anaerobic conditions.</p><p>The spectrum for the 1:1 the wild type Ns H-NOX NO-heme complex acquired at 10K has better resolution in the hyperfine and super hyperfine feature than that acquired at 110K but otherwise is very similar and typical for a 6c-NO/histidine heme complex. A microwave power dependence study conducted at 110K yielded a half saturation power > 100 mW if the data are treated as homogeneous broadening for a powdered sample (Fig. 6, Inset). This result indicates the presence of a fast energy relaxation mechanism. The spectrum acquired at 4 mW was used for spin quantification against a 1mM CuSO4 and indicated that the amount of NO complex formed at 1:1 or 1:2 ratio is ≥ 90% of the expected amount of spin based on the protein concentration.</p><p>When a large excess (1 atm) of NO gas is added directly delivered to an anaerobic wild type Ns H-NOX sample, the final NO-heme complex is ~ 40% 5c NO-heme based on the amount of 3-line hyperfine spectra that is present in the EPR spectrum. This observation corroborates earlier electronic spectral data (14) and our interpretation of the initial and final species in our optical stopped-flow data for the reaction of NO with reduced H-NOX (Fig. 3a).</p><!><p>The "truth diagram analysis" developed by Blumberg and Peisach (33) is useful for determining the potential axial heme ligands for the low-spin ferric heme H-NOX complexes by examining the correlation between heme rhombicity and the distal ligand tetragonal field strength. Analysis of the native low-spin ferric heme and imidazole derivative signals of H-NOX indicates that the native oxidized sensor has both proximal and distal imidazole ligands because the correlations between rhombicity and field strength are located in the B zone of the Truth Diagram (Fig. 7, grey circles) (23).</p><p>This conclusion seems supported by the almost super-imposable optical spectra of oxidized Ns H-NOX and its imidazole complex (data not shown). Both have a Soret peak at 414 nm. However, there is no distal histidine side chain located near the heme iron atom in the crystal structure of ferrous Ns H-NOX. The closest distal histidine, His150, has its Nε 17.6 Å from the heme iron (14). The correlations in the Truth diagram analysis show that the oxidized Ns H-NOX sensor also falls in the center of the narrower "C" zone, which overlaps that for the "B" or bis-His region. The "C" zone is for hemeproteins with methionine/histidine ligand pairs, which are commonly found in c type hemeproteins. There are two methionine residues, Met1 and Met144 (Fig. 8) within 8 Å distance from the heme in the distal portion of the heme pocket of ferrous Ns H-NOX (PDB identifier 2O09). The orientation of Met144 seems more favorable than Met1 for coordination of the heme iron but is still ~7.6 Å from the heme iron. However, M144I H-NOX EPR and NIR MCD spectra are almost identical to those of the wild type sensor, excluding M144 as the axial ligand in the ferric H-NOX (Figs. 9 and 10). Thus, regardless of the exact interpretation of the Truth Diagram EPR analysis, there must be a substantial protein conformational change that enables a direct coordination of either His150 or the sulfur atom of Met1 to the heme iron after oxidation. The hypothesis of Met-heme-His coordination is supported by the similar rhombic low-spin EPR signal of ferric cytochrome c, which has Met/His ligand pair (Fig. 9). In addition to the low-spin heme signals, the ferric Ns H-NOX sample also has significant amounts of rhombic high-spin heme at g=6 and some non-specific iron at g= 4.3, just as that seen in Fig. 5. In contrast, the imidazole ferric Ns H-NOX complex shows a homogeneous low-spin species, indicating that the heterogeneous EPR spectrum of the ferric Ns H-NOX probably originates from the difference of distal ligand field strength between imidazole and methionine and resulted in a mixture of 5c-heme-His and 6c His-heme-His or Met-heme-His species.</p><!><p>The ligand to metal charge-transfer (LMCT) MCD signal in the near IR region of the low-spin ferric hemeproteins is another useful diagnostic tool for axial ligand identity. This method complements EPR spectrometry and often enables unambiguous assignment of heme ligands for low-spin ferric heme complexes (34). Met-His ligand pairs have unique NIR MCD signals in the 1700–1900 nm region, whereas the His-His ligand pairs show a signal in the 1450–1600 nm range. A ~65 µM oxidized Ns H-NOX sample prepared in D2O buffer was used for NIR MCD measurement in parallel with ~500 µM cytochrome c, containing an authentic Met/His ligand pair (Fig. 10).</p><p>The peak of cyt c is at 1732 nm, corresponding to Met/His heme ligation, as previously published (34). The peak of Ns H-NOX MCD is at 1584 nm. There are several possible ligand combinations for this charge-transfer energy range including His/His, His/Lys, and His−/Met pairs (34). The His-Lys pair is not supported by the EPR data. The g values observed for the typical model hemeproteins containing His/Lys ligand pairs fall into the "CN−" zone of the Truth Diagram (Fig. 7)(35, 36). In addition, the closest lysine residue in Ns H-NOX, Lys7, is >15 Å from the heme iron (14). As with the ε-amino N atom of Lys, the α-amino group of Met1 cannot be an axial ligand candidate either because of the expected position of its g-values in the "H" region of the EPR "Truth Diagram" (37). The His−-Met combination is supported by the NIR MCD data, but the location of low-spin heme with such a ligand pair in the Truth Diagram is predicted to be in the "H" zone due to the additional charge on the histidine (37). The possibility of having the pyrrole N atom of the indole side chain of W74 as axial ligand is not high even though this atom is located only 4Å away from the iron atom. There are no existing examples of a model heme or hemeprotein complex with an indole N atom as an axial ligand, presumably due to steric hindrance of the Cz atom of the benzene ring as well as the high pKa of the indole proton. Thus, the only axial ligand combination of the Ns H-NOX low-spin ferric complex which satisfies both the EPR and NIR MCD characterizations is a His/His pair.</p><!><p>If coordination by a distal histidine side chain occurs in the ferric state, then oxidation of H-NOX must induce a significant conformational change to bring His150 or another histidine side chain into the active site. To examine this possibility, we recorded Far-UV CD spectra for the ferric and ferrous forms of unliganded Ns H-NOX. As shown in Fig. 11, both reduced and oxidized sensors have well defined α helical secondary structure. The ferric versus ferrous CD difference spectrum shows a peak at 218, a trough at 196 nm, and a zero crossover at 207 nm, indicating a loss of α-helix structure upon oxidation that could allow bis-histidyl coordination of the heme iron.</p><!><p>Optical titration of ~3 µM ferric Ns H-NOX with cyanide showed a simple hyperbolic binding isotherm. The Soret peak shifts from 414 nm to 418 nm and the maximal change occurs at 406 nm. The dissociation constant, Kd, for cyanide binding obtained by non-linear fitting to Eq. 1 is 0.65 mM at 24 °C (Fig. 12A, black circles). A similar titration with azide yielded an apparent Kd of 0.54 mM (Fig. 12A, red circles). Both of these values are large compared to Hbs and Mbs indicating significant hindrance to binding, presumably because of the need to displace one of the bis-histidyl ligands.</p><p>Time courses for azide binding to ferric Ns H-NOX were examined by stopped-flow absorbance measurements under pseudo first order conditions. The observed rate constants were roughly independent of [N3−] in the range from 31.2 to 250 mM and equal to ~ 0.015 s−1. These results indicate that the rate-limiting step for azide binding is first order dissociation of one of the endogenous axial ligands in the low-spin ferric complex (Fig. 12B, red circles). Cyanide binding kinetics are biphasic with a fast phase accounting for ~40% of the total absorbance changes. The fast phase shows a clear hyperbolic dependence on [CN−], whereas the slow phase shows little dependence on ligand concentration and exhibits a first order limiting rate ≈ 0.13 s−1 (Fig. 12B, blue circles). The observed rate for the fast phase of cyanide limited off to a value of 1.8 s−1 at [CN−] ≥ 100 mM. For both phases, a first order process limits cyanide binding and presumably represents dissociation of one of the endogenous ferric axial ligands. The two phases for cyanide binding could reflect the heterogeneity seen in the EPR signals, which is not manifest for the more slowly reacting azide ligand or it could reflect a more complex reaction mechanism involving a bis-cyanide intermediate. However, the key observation is that both azide and cyanide binding are limited by dissociation of one of the internal bis-axial ligands in ferric H-NOX.</p><!><p>Despite significant sequence homology in the distal heme pocket between sGC β subunit and the Ns H-NOX (12, 14), our current study has revealed six significant differences between the ligand binding properties of these two heme proteins, which may indicate an alternative physiological function for Ns H-NOX.</p><p>(1) Ns H-NOX forms a relatively stable 6c-NO complex at 1:1 and 1:2 heme:NO ratios. A 5c-NO complex is formed at very high [NO], as was measured earlier, but the extent is ≤ 40% (14). In contrast, when one equivalent of NO is mixed with sGC, a transient 6c-NO complex is formed, but isomerizes rapidly and completely to the final equilibrium form of the enzyme, which is a 5c NO-heme complex with a Soret peak at 399 nm and a classical three-line EPR spectrum indicating only one axial N atom (38). In the presence of one equivalent of NO, Ns H-NOX only forms a 6c-NO complex with a peak at 418 nm and a nine line EPR spectrum indicating two axial N atoms bound to the reduced iron (Figs 3–6). Very high concentrations of NO are required to convert some of the reduced Ns H-NOX to a 5c NO complex.</p><p>(2) Two different 6c-NO complexes are observed when Ns H-NOX is mixed with excess NO. The first species has the same spectral characteristics (i.e. Soret peak at 418 nm) as the initial transient hexacoordinate NO-heme-His complex observed for sGC and is formed very rapidly (k'on ≈ 300 µM−1s−1) in a simple bimolecular association reaction. The second Ns H-NOX species is unique, with a peak at 414 nm but a typical nine line NO-heme-His EPR spectrum. Unexpectedly, the rate of formation of this second, 6c species exhibits a linear dependence on [NO]. It seems likely that a transient bis-NO-heme intermediate is involved in the formation of the second 6c-NO complex. This transition probably involves a significant conformational change, perhaps mimicking that required for the binding of the second axial protein ligand to the oxidized H-NOX. One possible mechanism is: where ON-heme-N* represents the 414 nm species with the altered conformation with either an altered His105 conformation or the second axial ligand with an N atom, i.e. the His150 imidazole base. Applying steady-state assumptions for the pentacoordinate NO-heme and bis-NO-heme complexes, the observed rate for this second phase is: Equation 1ksecondary(obs)=k−Hisk′NO,2[NO]kN*−NOk+Hisk−NO,2+k+HiskN*−NO+k′NO,2[NO]kN*−NO At low [NO], the k'NO,2[NO]kN*−NO term in the denominator can be neglected, the observed rate of the slow phase is predicted to be linearly dependent on [NO] Equation 2ksecondary(obs)=k′NO,2[NO]K−HiskN*−NO(k−NO,2+kN*−NO) Thus, the apparent second order rate constant is determined by the product of the equilibrium constant for proximal histidine dissociation from the initial 6c complex (K−His), the bimolecular rate constant for the binding of the second NO, k'NO,2, and the fraction of bis-NO-heme complexes that either lose the second NO or re-react with a protein histidine side chain to make a more stable complex, (i.e., kN*−NO/(kN*−NO + kNO,2)). This mechanism rationalizes the linear dependence of the rate of the second phase on [NO] with a y-intercept ≈ 0 (Fig. 3b) and explains why the 418 nm to 414 nm spectral transition only occurs in the presence of excess NO.</p><p>(3) Conversion of the Ns H-NOX 6c-NO complex to a 5c-NO complex is a slow biphasic process, not dependent on [NO] and is only partially complete (≤ 40%). In contrast, conversion from the initial 6c NO-heme-His complex to a 5c NO-heme species occurs rapidly and completely for sGC, even at 1:1 NO to protein (Martin, E., Berka, V. and Tsai, A.-L., unpublished data). In addition, the 6c to 5c conversion depends linearly on [NO] for sGC (28, 38), implying a mechanism similar to that in Scheme I for the formation of the final 6c His-heme-NO complex of Ns H-NOX. A possible mechanism for formation of 5c NO-heme in sGC is: Heme-NO* represents the altered and activated 5c conformation of the enzyme, perhaps with NO on the proximal side of heme. Again, the second NO is acting catalytically to facilitate formation of the new 5 coordinate complex, and the observed rate for formation of pentacoordinate sGC is also predicted to depend on [NO] as shown below and observed experimentally (27, 28) Equation 3k5c(obs)=k−Hisk′NO,2[NO]k−NO*k+Hisk−NO,2+k+Hisk−NO*+k′NO,2[NO]k−NO*≈k′NO,2[NO]K−His(k−NO*k−NO,2+k−NO*)atlow[NO] The activated state for Ns H-NOX appears to be the hexacoordinate ON-heme-N* species in Scheme I, with the nitrogen base being unknown, but presumed to be a histidyl side chain, possibly the distal amino acid, His150. Although Ns H-NOX does not form a fully 5c NO-heme species, a major conformational change in the protein does appear to be induced by transient binding of a second NO as is observed for sGC, and perhaps the conformation change in the activated, six coordinate ON-heme-N* state of H-NOX is more similar to that observed in sGC than is apparent from spectral analyses of the iron complexes.</p><p>(4) Ns H-NOX heme is sensitive to atmospheric oxygen and autooxidizes over a period of 24 hours, but no Fe(II)O2 intermediate is ever observed. In contrast, full-length sGC is inert to reactions with O2 in air or at 1 atm of the pure gas. However, truncated β1 corresponding to the H-NOX domain readily autooxidizes at rates of 4.3 hr−1 and 0.18 hr−1 at 37 °C, respectively, for ferrous β1(1–194) and β2(1–217) mutant sGC (39). These autooxidation rates are much greater than those of mammalian HbO2, MbO2 and reduced Ns H-NOX, indicating that the deleted ~400 residues of sGC β-subunit or perhaps the alpha subunit protects the ferrous heme from autooxidation (24).</p><p>(5) The majority of the heme in oxidized Ns H-NOX is 6c low-spin rather than the pure high-spin 5c form found for sGC (14, 38, 40). There appears to be reversible ligation and dissociation of an endogenous distal ligand, perhaps His150, during redox cycling in Ns H-NOX. Similar experiments with sGC indicate the proximal His remains bound in both the oxidized and reduced forms, but reversible dissociation of the proximal histidine does occur after NO binding to the ferrous form of sGC to create the activated 5c NO complex..</p><p>(6) Ns H-NOX reacts 100-fold more rapidly with CO than sGC and has a much higher affinity for this ligand, implying that the iron atom is more reactive in the bacterial sensor (41). However, the bimolecular rate constants for NO binding to form the first 6c NO-heme-His complexes are nearly identical, very large (100–300 µM−1s−1), and probably diffusion controlled, implying little resistance to ligand movement into both proteins.</p><p>NO binding to Ns H-NOX was examined previously in different way by Boon et al. (42). In their study, the sensor protein was first treated with excess of NO releaser (NONOate), diethylamine and then gel-filtered to remove all side products of NONOate and free NO. Spectral studies and dissociation kinetics were assessed using this NO complex of H-NOX (42). Optical and resonance Raman spectra indicated that both 5c and 6c NO complexes were present (42) and the amount of 5c component (40–50%) was similar to what we observed in the presence of excess amount of NO (Fig. 3a, lower panel). However, in our study when a stoichiometric amount of NO was added to Ns H-NOX only a 6c NO complex with a Soret absorption peak at 418 nm is formed and is stable for several hours. Thus, Ns H-NOX adopts different coordination geometries at 1:1 versus excess NO levels. This observation is most readily interpreted using the multiple-step mechanism for NO binding shown in Scheme I. In principle, this simple hexacoordinate ON-heme-His complex should equilibrate to the final ON-heme-N*/ON-heme mixture, but catalytic amounts of excess NO are needed to facilitate this transition on normal time scales after preparation. Although Boon et al. removed free NO from their final sample via gel filtration, excess NO was almost certainly present during the prolonged incubation of the NONOate with the sensor protein.</p><p>Boon et al. also reported that they were unable to measure NO dissociation from Ns H-NOX using either excess CO (plus 30 mM dithionite) or HbO2, whereas the same methods enabled successful determinations for NO dissociation rate constant for other two H-NOX proteins (42). In contrast, we were able to measure NO dissociation from the initial 6c NO complex of Ns H-NOX (Fig. 4b). Thus, the equilibrium mixture of 6c and 5c species for the NO H-NOX complex prepared in the presence of excess NO appears to have much lower NO dissociation rate constants. However, a more careful study of NO dissociation is needed to resolve the exact kinetic properties of the initial 6c, 418 nm; the intermediate 6c, 414nm; and the 5c, 399 nm NO H-NOX complexes.</p><!><p>We had previously observed that NO binding to Ns H-NOX yields only ~40% population of 5c NO-heme compared to sGC, and proposed a possible structural basis for this difference (14). The Ns H-NOX heme-pocket residues Met144 and Trp74 are larger than the spatially analogous isoleucine and phenylalanine residues in sGC and might be inhibiting stable formation of a 5c NO-heme complex. We subsequently generated M144I and the W74F mutants of Ns H-NOX, engineering the heme pocket to be similar to sGC. Both single mutations resulted in larger populations of 5c NO-heme complexes, as indicated by their optical spectra (14). To extend these observations, we prepared the NO complex of reduced M144I H-NOX (generated either by adding excess NO or reacting sodium nitrite with dithionite), and as shown in Fig. 6 (top solid spectrum), this mutant in excess NO exhibits a pure low spin 5c NO-heme EPR spectrum, which is characteristic of only a single axial N atom with pronounced 3 hyperfine lines with a splitting constant of 17 G caused by the nitrogen nucleus of the proximal ligand, His105. This spectrum is very similar to that observed for the NO complex of the reduced sGC (38). Indeed, the EPR spectrum can be properly simulated by parameters as a nearly axial low-spin 5c NO complex (Fig. 6, top dashed spectrum). This effect led us to speculate instead that Ns H-NOX might be more sensitized to a transition to 5c NO-heme once it is bound to its effector domain. Replacement of Met144 with Ile does appear to make the behavior of Ns H-NOX closer to that of sGC in terms of NO binding and perhaps sensing in the ferrous state. Thus, introduction of a methionine in place of the isoleucine in the Ns H-NOX may be an adaptation to a different functional role, such as sensing the redox potential of the environment.</p><p>A "heme pivoting" model for sGC was proposed to interpret 5c NO-heme formation, activation of the cyclase, and other fast relaxation events after NO binding to the heme sensor domain (14, 43). In the case of Ns H-NOX, oxidation produces a 6c low spin complex that appears to require a substantially larger conformational change than that postulated "heme pivoting and bending" mechanism of sGC, where side chain motions of < 2 Å are proposed to occur in response to gaseous ligand binding (14). The large motions triggered by oxidation are required for the coordination of all the possible axial ligands of ferric Ns H-NOX, and particularly for His150. His150 is the closest histidine residue to the heme in the crystallographic structure of the ferrous H-NOX but is still 17.6 Å away from the iron atom (14). The heme moiety is located in the interface of the proximal αββαββ domain and the distal α-helical domain, and movement of a His150 to become an axial ligand would require "melting" of a significant percentage of α-helical secondary structure. This interpretation is consistent with the loss of ellipticity at 222 nm seen after oxidation of H-NOX (Fig. 11). Attempts to crystallize ferric H-NOX have not been successful perhaps reflecting the conformational flexibility required for bis-histidyl coordination.</p><p>Our current EPR, NIR MCD analysis and the published crystallographic data (14) for Ns H-NOX show little 5c NO-heme formation at low [NO], implying an activation mechanism different from that for sGC. However, as suggested in Schemes I and II and observed experimentally, the secondary spectral and conformational changes following formation of the initial 6c NO-heme-His complex in both proteins depends linearly on [NO] and appear to be catalyzed by transient formation of NO-heme-NO complexes (28). Thus, it is possible that the early stage protein conformational changes of sGC and Ns H-NOX are similar and driven by this unusual bis-NO heme chemistry.</p><p>The structure of the second 6c-NO complex in Ns H-NOX with an absorption peak at 414 nm is not resolved, but both the optical and EPR characterizations provided useful clue about its structure. Although the Soret peak is at the same wavelength, it is not the ferric form of the sensor, because the lineshape of the peaks in the visible region and near UV region are different (compare Figs. 1 and 3a), and this 414 nm NO-heme species has a large g=2 EPR signal with typical hyperfine nine line features indicative of ferrous ON-heme-His complexes. It cannot be a bis-nitrosyl ferrous heme intermediate, which is an EPR-silent diamagnetic species. The initial 6c NO-heme-His complex in Ns H-NOX is also well defined spectrally by a nine-line g=2 EPR signal and is identical to the EPR spectrum of the initial complex formed in sGC. In the latter case, this 6c signal changes to a three-line g=2 signal characteristic of a 5c NO-heme complex even at 1:1 NO/Heme. The very similar EPR spectra of the NO complex of Ns H-NOX containing 1:1 and 2:1 NO strongly support that these two 6c NO complexes, one with a peak at 418 and the other at 414 nm, have His-heme-NO coordination with the former having the proximal His105 and the latter perhaps the distal His150 as the heme axial ligands, i.e., the mechanism proposed in Scheme I.</p><p>A second NO molecule has also been postulated to bind to the sGC heme iron as a proximal ligand or at a non-heme site during the activation process (43–46). However, a site other than the heme iron is hard to justify for the fast NO binding rate because cysteine nitrosothiol formation will not occur anaerobically, and even the nitrosation reaction mediated by oxygen is very slow (47). There is no additional metal center adjacent to the heme iron that could be a second NO target. In our view, the NO concentration dependence of the 6c to 5c transition in sGC is most readily explained by formation of a transient ON-heme-NO complex as shown in Scheme II, and a similar bis-NO heme complex appears to be required for the transition from the 418 to 414 nm 6c NO-heme-His species in Ns H-NOX. Finally, the formation of a bis-axial ligand complex clearly occurs when H-NOX is oxidized and must involve substantial conformational changes regardless of the identification of the second ligand. Thus, hexacoordination of H-NOX is clearly a key part of the triggering mechanism for signaling by this sensor and can be produced by either NO binding or oxidation.</p><!><p>The biological function of H-NOX in Nostoc sp remains unresolved because the "effector" component for this sensor has not been found. This symbiotic cyanobacterium can perform both anaerobic nitrogen fixation and oxygen evolving photosynthesis (48). These two mechanisms of energy metabolism are not compatible, and the bacteria have developed specialized cells called heterocysts to isolate the machinery for nitrogen fixation from oxygen-evolving photosynthesis. Special mechanisms must have evolved in these organisms to protect against oxidative and light stress.</p><p>Proteomic analysis of the soluble fraction of the Nostoc sp has revealed many, perhaps redundant proteins that appear to sense and respond to the environmental changes and stresses (48). Multiple copies of superoxide dismutase, catalase, peroxiredoxin and peroxidase are expressed to metabolize superoxide and peroxides. Several glutathione metabolizing enzymes are present in abundance to maintain the steady-state level of cellular glutathione and regulate the overall cellular redox potential. Most of these major protein components appear to have evolved to counter oxidative stress. Thus, in addition to binding diatomic gases, Ns H-NOX might be sensing redox state changes and triggering its transducer to achieve feedback regulation of the cellular redox potential. Such a function appears to occur for the H-NOX sensor from Shewanella oneidensi (49). This H-NOX is coupled to a histidine kinase to sense environmental changes in oxygen levels. Since NO-bound H-NOX cannot be readily oxidized, retained NO binding capability can further serve to isolate nitrogen fixation and oxygen-evolving machinery by firmly setting this particular effector switch to "off" at high NO concentrations. Oxidation of sGC under pathophysiological conditions leads to ubiquitin-dependent degradation and may be a mechanism for deactivation of sGC (16). The switch between oxidative degradation of sGC and NO activation may have evolved from the proposed switch between NO sensing and redox sensing that we suggest may occur for Ns H-NOX.</p><!><p>This work is supported by NIH grant HL088128 (A.-L. Tsai), HL075329 (F. van den Akker), NIH grant GM 84348 (M. Fabian), and NIH grants HL047020 and GM035649 and grant C0612 from the Robert A. Welch Foundation (J. S. Olson).</p><p>nitric oxide</p><p>carbon monoxide</p><p>Heme-Nitric oxide and OXygen binding</p><p>Nostoc sp H-NOX</p><p>Thermoanaerobacter tengcongensis H-NOX</p><p>Soluble guanylyl cyclase</p><p>dithiothreitol</p><p>mercaptoethanol</p><p>isopropyl-1-thio-β-D-galactopyranoside</p><p>phenylmethyl-sulfonyl fluoride</p><p>hemoglobin</p><p>myoglobin</p><p>oxyhemoglobin</p><p>oxymyoglobin</p><p>circular dichroism</p><p>Magnetic Circular Dichroism</p><p>near infrared MCD</p><p>electron Electron Paramagnetic Resonance spectroscopy</p><p>ferrous heme</p><p>ferric heme</p><p>five coordinate NO-heme complex</p><p>six coordinate NO-heme complex</p><p>NO releaser</p><p>Kinetics of autooxidation of ferrous Ns H-NOX. 6 µM of ferrous Ns H-NOX in the absence of DTT was placed in the HP8453 diode array spectrophotometer to follow the optical changes up to 2 days at 24 °C. The kinetics of ΔA409−443 was fit to one-exponential function to obtain the rate of autooxidation (Inset).</p><p>A: Kinetics of CO binding to the ferrous Ns H-NOX. Ferrous Ns H-NOX, 5 µM, prepared in an anaerobic tonometer was reacted with CO in a stopped-flow at 24 °C. CO was prepared in anaerobic buffer at concentrations varying from 25 to 166 µM. One-exponential change of absorbance at 420 nm was used to determine the observed rates and used for calculation of the 2nd-order on rate constant (Inset). Two sets of experiments using two different batches of sensor protein were conducted (red and black circles). Standard deviation of at least triplicates at each concentration was shown by the significant bar. B: Determination of the dissociation rate constant of CO binding. CO-ferrous Ns H-NOX complex was pre-formed by mixing 4 µM sensor with 1 mM CO in an anaerobic glass tonometer, and then reacted with 1 mM NO solution. The CO-dissociation rate was followed by diode array or single wavelength stopped-flow at 422 nm (Inset) for 8s data collection at 24 °C. Kinetic data were fit to 2- exponential function.</p><p>A: Optical characterization of Ns H-NOX binding to stoichiometric and excess amount of NO. UV-Vis spectral changes recorded by anaerobic rapid-scan diode array stopped-flow within 1 s or 2 min for 5 µM Ns H-NOX reaction with 1:1 and 5:1 ratio of NO, respectively, at 24 °C. B: Kinetics of Ns H-NOX binding with excess NO under pseudo first-order conditions. 5 µM ferrous Ns H-NOX was mixed with > 10 × of NO at different concentrations and the kinetics was recorded by single wavelength stopped-flow at 428 nm at 50 ms/ 5s split time mode to cover the triphasic changes. Data were fit by 3-exponential function and the first fast phase that show [NO] dependence was further analyzed to obtain the 2nd order binding rate constant as the secondary plot (Inset). Average of three shots or more with the standard deviation plotted as the error bar for each NO concentration in two different sets of experiments (blue and red symbols). The linear regression (red dash) and the slope and y-intercept obtained are that with one of the two sets of data.</p><p>Determination of the association (A) and dissociation (B) rate constants of NO binding to Ns H-NOX for formation of the 1st 6c NO complex by flow-flash and stopped-flow. A: CO complex of Ns H-NOX was first prepared with 50 µM sensor protein and 100 µM CO anaerobically in a syringe. This complex was briefly mixed with 1 mM NO in the stopped-flow for 50 ms and a 0.5 µs 577 nm dye laser was applied to flash off the bound CO. Subsequent NO association kinetics was monitored at both 436 (red trace) and 418 nm (blue trace). Black lines are one-exponential fittings for both data. B: NO complex of Ns H-NOX was first prepared anaerobically by mixing 5 µM sensor protein with 10 µM NO in a glass tonometer. This mixture was then mixed with a mixture of 1 mM CO and 25 mM dithionite and aged for 50 ms in a rapid scan stopped-flow and the electronic absorption spectra at 3 ms (black) and 270 s (blue) were shown together with that at zero time. Single wavelength kinetic data at 428 and 400 nm with opposite amplitude changes are shown in the Inset and fit to 1-exponential function to obtain the rate constant. All reactions were conducted at 20 °C.</p><p>EPR spectra of Ns H-NOX at different redox state and coordination. EPR spectra are presented (from top to bottom) for 80 µM isolated ferrous Ns H-NOX, 55 µM ferric Ns H-NOX prepared by ferricyanide oxidation and gel-filtration chromatography, 50 µM ferric Ns H-NOX mixed with 50 mM imidazole, and ferrous NO complex formed with 160 µM NO. EPR conditions were: 4 mW, 9.602 gHz, time constant: 0.33s, modulation amplitude: 10.9G and 10K.</p><p>EPR characterization of the ferrous NO complex of Ns H-NOX. 70 µM ferrous Ns H-NOX was reacted with 2:1 NO by injecting 2 mM NO-saturated buffer under anaerobic condition and manually frozen for EPR assessment (bottom solid trace). The hyperfine splitting by NO nitrogen nucleus and super hyperfine splitting by the proximal histidine were indicated for the gz component as well as the isotropic hyperfine splitting for the gy and gx component optimized via spectral simulation to the dominant 6c low-spin NO complex (bottom dashed trace). An EPR for 60 µM ferrous NO complex of M144I sensor dominated by 5c NO complex was juxtaposed for comparison (top solid trace). The three g values and the associated hyperfine parameter values for a simulated 5c-NO complex spectrum (top dashed trace) that closely matches the acquired spectrum are also indicated for comparison with those obtained for the wild type sensor. Progressive power saturation at LN temperature, the value of half-saturation power, P1/2, and value of b by fitting the data to equation (2) are also shown for the wild type sample (Inset). EPR conditions were the same as described in Fig. 5 except 1.87 G modulation amplitude and 9.28 gHz frequency.</p><p>Truth diagram analysis of the low-spin ferric Ns H-NOX and its imidazole complex. Correlation between the heme rhombicity and the axial ligand strength of many low-spin heme complexes (various symbols) are mapped in six different zones. Complexes in five of them have histidine as proximal ligand and with cyanide (zone CN−), methionine (zone C), histidine (zone B), histidinate/azide (zone H), and hydroxide/phenolate (zone O) as the distal ligand. Those hemeprotein containing a cysteine thiolate proximal ligand including P450, chloroperoxidase and nitric oxide synthase and their derivatives fall in the P zone. This diagram was modified from the original Blumberg Peisach convention (33, 51). The grey solid circles are the ferric Ns H-NOX and its imidazole derivative.</p><p>Candidate axial heme ligand for ferric Ns H-NOX. Potential heme-ligand residues nearby heme iron and at the distal heme side of the wild type ferrous Ns H-NOX are displayed. Distances between two candidate methionine residues (Met144 and Met1), N-terminal amino group, Trp74 indole nitrogen and the heme iron (orange ball) are given. The second closest histidine, His150 and the closest lysine, Lys7 and the proximal heme ligand, His105, are also shown in stick model. Heme porphyrin is shown as red stick model. This is a RASMOL representation of PDB identifier: 2O09.</p><p>Similarity of EPR spectrum between Ns H-NOX and cytochrome c. EPR of 65 µM ferric Ns H-NOX (red), 30 µM M144I Ns H-NOX (blue), and 500 µM bovine cytochrome c (black) are directly compared for the position of the principal g values of the low-spin rhombic heme. The additional features including the high-spin heme at g = 6 region and the non-specific iron at g = 4.3 and organic radical at g = 2 present in the Ns H-NOX are not present in cytochrome c. The two broad features highlighted by the red stars are artifacts of the resonator. The EPR amplitude for the M144I mutant has been adjusted for its concentration difference from the wild type sample.</p><p>Identification of the axial heme ligands of ferric Ns H-NOX by NIR MCD. Same three samples used for EPR measurements in Fig. 9 were used in the NIR MCD measurements in the range from 800 to 2000 nm at 24 °C. All samples were prepared in D2O 50 mM HEPES, pH 7.4 to minimize the NIR MCD overtone background signals from the water solvent (34). MCD conditions are described in the Methods and the spectrum for cytochrome c was obtained as one scan while that of wild type and M144I Ns H-NOX were an average of 8 scans. The wavelengths of the main transitions are labeled for each sample. These marker wavelengths (nm), or energy levels are used to identify the axial ligand pairs from the correlation diagram composed for a series of low-spin hemeproteins with two axial ligands (34).</p><p>Far-UV CD of the ferric and ferrous Ns H-NOX. Both the ferrous and ferric sensor proteins were prepared as 1 µM in 2 mM HEPES containing 2 mM NaCl and 0.1 % glycerol, pH 7.7 by either dithionite or ferricyanide pre-treatment, respectively, and then gel-filtered in the same HEPES buffer to remove excess reductant or oxidant. CD measurements were conducted as described in the Experimental Procedures. The data shown for the ferric (black), ferrous (red) sensor are 12-scan average corrected for background signal originated from buffer alone and the difference spectrum (blue) is obtained by subtracting the CD spectrum of the ferrous sensor from that of the ferric form.</p><p>Equilibrium and kinetic binding of cyanide and azide to the ferric Ns H-NOX. To build binding isotherm for cyanide and azide binding (A), 2.9 µM ferric Ns H-NOX was titrated by aliquots of 0.2 M cyanide to the final concentration of 5 mM (black circles); and 5.8 µM ferric sensor protein was titrated with 1M azide to 3 mM final concentration (red circles), respectively, to reach saturation. Maximal absorbance changes at 307 – 406 nm for cyanide heme complex formation and 377 – 417 nm for azide heme complex formation at different ligand concentrations were fit by single binding site saturation hyperbola (equation 1) (solid lines). The volume change by addition of either ligand was less than 2.5% and was not corrected further for calculating the free ligand concentrations. For kinetic binding measurements (B), 2.9 µM ferric Ns H-NOX was mixed with a series of ligand concentrations at least 10 fold excess to the protein and the absorbance changes at 430 nm in a stopped-flow. The observed rates obtained by single (for azide, red circles) and double (for cyanide, black and blue circles) exponential function were plotted against ligand concentrations. Solid line is the fit to hyperbolic function to show [CN−]- dependent saturation represented by equation 1. Error bars are the standard deviation of triplicate shots. Reaction temperature for both equilibrium and kinetic binding experiments was 24 °C.</p><p>Ligand binding parameter values for the ferrous and ferric sGC and Ns H-NOX.</p><p>calculated as kOFF/ kON using values from ref. (29, 50), N/A - not applicable, N/D - not determined.</p><p>Unless otherwise specified, the experimental temperature was 20 – 25 °C.</p><p>The dissociation constant for the 2nd step, k−2.</p><p>Biphasic kinetics.</p>

PubMed Author Manuscript

Iron-Catalyzed Cross-Coupling of Unactivated, Secondary Alkyl Thio Ethers and Sulfones with Aryl Grignard Reagents

The first systematic investigation of unactivated aliphatic sulfur compounds as electrophiles in transition metal-catalyzed cross-coupling are described. Initial studies focused on discerning the structural and electronic features of the organosulfur substrate which enable the challenging oxidative addition to the C(sp3)\xe2\x80\x93S bond. Through extensive optimization efforts, an Fe(acac)3-catalyzed cross-coupling of unactivated alkyl aryl thio ethers with aryl Grignard reagents was realized, in which a nitrogen \xe2\x80\x9cdirecting group\xe2\x80\x9d on the S-aryl moiety of the thio ether served a critical role in facilitating the oxidative addition step. In addition, alkyl phenyl sulfones were found to be effective electrophiles in the Fe(acac)3-catalyzed cross-coupling with aryl Grignard reagents. For the latter class of electrophile, a thorough assessment of the various reaction parameters revealed a dramatic enhancement in reaction efficiency with an excess of TMEDA (8.0 equiv). The optimized reaction protocol was used to evaluate the scope of the method with respect to both the organomagnesium nucleophile and sulfone electrophile.

iron-catalyzed_cross-coupling_of_unactivated,_secondary_alkyl_thio_ethers_and_sulfones_with_aryl_gri

Introduction<!>1. Existing Methods for Constructive Elaboration of Unactivated C(sp3)\xe2\x80\x93SPh Bonds<!>2.2. Cross-Coupling of Organosulfur Electrophiles: State of the Art<!>2.3. Challenges<!>2.4. Objectives of this Study<!>Results<!>1. Cross-Coupling of Alkyl Aryl Thio Ethers. 1.1. Initial Studies<!>1.2. Survey of Metal Salts as Catalysts<!>1.3. Evaluation of Reaction Solvent<!>1.4. Evaluation of Reaction Temperature<!>1.5. Evaluation of Ligands and Additives<!>1.6. Evaluation of the S-Aryl Group<!>1.7. Preparative Scale Cross-Coupling<!>2. Cross-Coupling of Alkyl Phenyl Sulfones<!>2.1. Orienting Experiments<!>2.2. Survey of Metal Salts as Catalysts<!>2.3. Evaluation of Reaction Solvent<!>2.4. Evaluation of Amine Additives<!>2.5. Reaction Scope. 2.5.1. Nucleophile<!>2.5.2. Sulfone Substrate<!>3. Mechanistic Investigations. 3.1. Stereochemical Course of the Cross-Coupling of Alkyl 2-Pyridyl Thio Ether 3d<!>3.2. Steric Course of the Cross-Coupling of Secondary Alkyl Phenyl Sulfones<!>1. Cross-Coupling of Alkyl Aryl Thio Ethers. 1.1. Effect of the S-Aryl Group<!>1.2. Speculation on the Mechanism<!>2. Cross-Coupling of Alkyl Phenyl Sulfones. 2.1. Speculation on the Mechanism<!>2.2. Scope of the Nucleophile<!>2.3. Scope of the Sulfone Substrate<!>Conclusion and Outlook<!>General Procedure 1 for Cross-Coupling of Secondary Alkyl Phenyl Sulfones<!>2. Preparation of Authentic Samples<!>Preparation of (rac)-(3-Iodobutyl)benzene (48)<!>Preparation of (rac)-1,1\xe2\x80\xb2-(1-Methyl-1,3-propanediyl)bisbenzene (4)<!>Preparation of (rac)-4-Phenyl-2-butanol (28)<!>Preparation of (rac)-(3-Bromobutyl)benzene (29)<!>Preparation of (l)- and (u)-1,1\xe2\x80\xb2-(3,4-Dimethyl-1,6-hexanediyl)bisbenzene (49)<!>3. Preparation of Thio Ether Substrates<!>Preparation of (rac)-(4-Phenylbutan-2-ylthio)benzene (3a)<!>Preparation of (rac)-4-Trifluoromethyl(4-phenylbutan-2-ylthio)benzene (3b)<!>Preparation of (rac)-4-Phenylbutane-2-thiol (7)<!>Preparation of (rac)-2,3,4,5,6-Pentafluoro(4-phenylbutan-2-ylthio)benzene (3c)<!>Preparation of (rac)-2-(4-Phenylbutan-2-ylthio)pyridine (3d)<!>Preparation of (rac)-3,5-Bis(trifluoromethyl)(4-phenylbutan-2-ylthio)benzene (3e)<!>Preparation of (rac)-3-(4-Phenylbutan-2-ylthio)pyridine (3f)<!>Preparation of (rac)-2-(4-Phenylbutan-2-ylthio)pyrimidine (3g)<!>Preparation of (rac)-2-(4-Phenylbutan-2-ylthio)benzo[d]oxazole (3h)<!>Preparation of (rac)-1-Phenyl-5-(4-phenylbutan-2-ylthio)-1H-tetrazole (3i)<!>Preparation of (rac)-8-(4-Phenylbutan-2-ylthio)<!>Preparation of (rac)-10-(4-Phenylbutan-2-ylthio)benzo[h]quinoline (3k)<!>Preparation of (rac)-2-(4-Phenylbutan-2-ylthio)aniline (50)<!>Preparation of (rac)-N,N-Dimethyl-2-(4-phenylbutan-2-ylthio)aniline (3l)<!>Preparation of (rac)-2-(4-Phenylbutan-2-ylthio)benzaldehyde (51)<!>Preparation of (rac)-N,N-Dimethyl-1-[2-(4-phenylbutan-2-ylthio)phenyl]methanamine (3m)<!>4. Cross-Coupling of Thio Ether 3d<!>Preparation of (rac)-1-Methoxy-4-(4-phenylbutan-2-yl)benzene (8)<!>5. Preparation of Alkyl Phenyl Sulfone Substrates<!>Preparation of (rac)-[(4-Phenylbutan-2-yl)sulfonyl]benzene (9)<!>Preparation of (rac)-4-(4-Methoxyphenyl)-2-butanol (52)<!>Preparation of (rac)-4-(3-Bromobutyl)-1-methoxybenzene (53)<!>Preparation of (rac)-1-Methoxy-4-[3-(phenylthio)butyl]benzene (54)<!>Preparation of (rac)-1-Methoxy-4-[3-(phenylsulfonyl)butyl]benzene (14)<!>Preparation of (Phenylsulfonyl)cyclohexane (18a)<!>Preparation of N-Benzylpiperidin-4-yl 4-Toluenesulfonate (56)<!>Preparation of N-Benzyl-4-(phenylthio)piperidine (57)<!>Preparation of N-Benzyl-4-(phenylsulfonyl)piperidine (18b)<!>Preparation of (1l,2l,4u)-2-(Phenylthio)bicyclo[2.2.1]heptane (18c)<!>Preparation of 2-(Phenylsulfonyl)tricyclo[3.3.1.13,7]decane (18d)<!>Preparation of (rac)-1-(4-Methoxyphenyl)-2-propanol (59)<!>Preparation of (rac)-4-(2-Bromopropyl)-1-methoxybenzene (60)<!>Preparation of (rac)-1-Methoxy-4-[2-(phenylthio)propyl]benzene (61)<!>Preparation of (rac)-1-Methoxy-4-[2-(phenylsulfonyl)propyl]benzene (18f)<!>Preparation of (rac)-2-[2-(Phenylsulfonyl)propyl]-1,3-dioxolane (18g)<!>Preparation of (rac)-N-benzyl-4-[1-(phenylsulfonyl)ethyl]piperidine (18i)<!>Preparation of 2-(Phenylsulfonyl)-1-propanol (18l)<!>Preparation of 1-(Phenylsulfonyl)decane (21)<!>Preparation of (rac)-[(2-Methyl-4-phenylbutan-2-yl)sulfonyl]benzene (25)<!>A. Scope of Nucleophile<!>Preparation of (rac)-1-Methoxy-4-(3-phenylbutyl)benzene (16a)<!>Preparation of (rac)-1-Methoxy-4-[3-(4-tolyl)butyl]benzene (16b)<!>Preparation of (rac)-1-Methoxy-4-[3-(3-tolyl)butyl]benzene (16c)<!>Preparation of (rac)-1-Methoxy-4-[3-(4-trimethylsilylphenyl)butyl]benzene (16d)<!>Preparation of (rac)-1-Methoxy-4-[3-(3-trimethylsilylphenyl)butyl]benzene (16e)<!>Preparation of 1-Isopropoxy-3-[4-(4-methoxyphenyl)butan-2-yl]benzene (16f)<!>Preparation of (rac)-1-Methoxy-4-[3-(4-biphenyl)butyl]benzene (16g)<!>Preparation of (rac)-1-Methoxy-4-[3-(2-naphthyl)butyl]benzene (16h)<!>B. Scope of Sulfone Substrate<!>Preparation of 1-Cyclohexyl-3-isopropoxybenzene (19a)<!>Preparation of N-Benzyl-4-phenylpiperidine (19b)<!>Preparation of (1l,2l,4u)-2-(3-Isopropoxyphenyl)bicyclo[2.2.1]heptane (19c)<!>Preparation of 2-(3-Isopropoxyphenyl)tricyclo[3.3.1.13,7]decane (19d)<!>Preparation of (rac)-1-Isopropoxy-3-(4-phenylbutan-2-yl)benzene (19e)<!>Preparation of (rac)-1-Methoxy-4-(2-phenylpropyl)benzene (19f)<!>Preparation of (rac)-2-(2-Phenylpropyl)-1,3-dioxolane (19g)<!>Preparation of (rac)-4-(1-Phenylethyl)cyclo-1-hexanone Ethylene Ketal (19h)<!>Preparation of (rac)-N-Benzyl-4-(1-phenylethyl)piperidine (19i)<!>Preparation of (rac)-N-Benzyl-3-phenylpyrrolidine (19j)<!>Reaction of (u)-2-Phenyl-3-(phenylsulfonyl)tetrahydropyran (18k)<!>Reaction of 1-(Phenylsulfonyl)decane (21)<!>Reaction of (rac)-[(2-Methyl-4-phenylbutan-2-yl)sulfonyl]benzene (25)<!>7. Preparation and Cross-Coupling of Enantiopure Substrates<!>Preparation of ( S)-(3-Bromobutyl)benzene (29)<!>Preparation of (R)-2-(4-Phenylbutan-2-ylthio)pyridine (3d)<!>Reaction of (R)-2-(4-Phenylbutan-2-ylthio)pyridine (3d)<!>Preparation of (R)-(4-Phenylbutan-2-ylthio)benzene (3a)<!>Preparation of (R)-[(4-Phenylbutan-2-yl)sulfonyl]benzene (9)<!>Reaction of (R)-[(4-Phenylbutan-2-yl)sulfonyl]benzene (9)<!>8. Preparation of Grignard Reagents<!>Representative Procedure 1: Preparation of 4-Methoxyphenylmagnesium Bromide<!>Representative Procedure 2: Preparation of 3-thienylmagnesium bromide (15p)

<p>The formation of carbon-carbon bonds via the union of unactivated aliphatic electrophiles with organometallic reagents under transition metal catalysis has long been regarded as one of the most challenging classes of cross-coupling reaction.1 Principal difficulties include the relatively slow oxidative addition of alkyl electrophiles to transition metal centers2 (at least via 2-electron pathways most common with palladium), the proclivity of the derived alkyl metal intermediates to decompose via rapid β-hydride elimination, and the slower reductive elimination of C(sp3) versus C(sp2) moieties.3 Although these hurdles are not necessarily insurmountable under palladium catalysis,4 complexes based on first-row transition metals such as nickel,5 cobalt,6 iron7,8 and copper5b,9 are fast emerging as the most efficient catalysts for such transformations.</p><p>In continuation of our longstanding research program on the "Lewis base activation of Lewis acids",10 we have recently reported the first catalytic, enantioselective thiofunctionalizations of unactivated alkenes, including sulfenoetherification11 (Scheme 1, Eq. 1) and carbosulfenylation12 (Scheme 1, Eq. 2) protocols. During the course of these studies, we became keenly aware of a relative dearth of methods for the constructive elaboration of the C(sp3)–SPh motif in the thio ether products into C(sp3)–C bonds, and considered the potential for cross-coupling reactions analogous to those developed for other unactivated aliphatic electrophiles (e.g. halides, sulfonates). We describe herein our studies on the use of unactivated, secondary alkyl sulfur electrophiles in transition metal-catalyzed cross-coupling, culminating in a reaction protocol for the cross-coupling of unactivated, secondary alkyl phenyl sulfones with aryl Grignard reagents under iron catalysis.</p><!><p>The use of unactivated13 C(sp3)–SPh bonds as a locus for C(sp3)–C bond-formation is rare in chemical synthesis, at least in a direct fashion, and most manipulations of C(sp3)–SPh bonds serve either to remove an unwanted thio ether group by desulfurization,14 or transform it to a more versatile functional group such as an alkene (via elimination of the sulfoxide15) or a carbonyl compound/equivalent (Pummerer rearrangement16). Although alkyl phenyl thio ethers can be converted to the corresponding alkyllithium species by means of reductive lithiation, followed by trapping with carbon electrophiles, this method is incompatible with substrates containing β-heteroatoms (with the exception of β-lithiooxy groups) due to rapid β-elimination.17 Additionally, the scope of carbon electrophiles able to react productively with alkyllithium reagents is relatively narrow, being mostly limited to simple alkylations and additions to carbonyl compounds. Although oxidation of an alkyl phenyl thio ether to the corresponding sulfone followed by α-alkylation and desulfurization18 is a possible alternative, this strategy suffers from the same general limitations.</p><p>Another strategy to effect the conversion of C(sp3)–SPh to C(sp3)–C bonds would be to generate a carbon-centered radical from the thio ether and trap it with radicophilic alkenes – a process which should be tolerant of β-heteroatomic groups.19 Although it is certainly possible to generate carbon-centered radicals from unactivated C(sp3)–SPh bonds with Bu3SnH, the process is considerably less facile than for the corresponding bromides, iodides, or selenides.20 In most cases this procedure is limited to simple reductions, which are often slow (the reduction of i-PrSPh with Bu3SnH in refluxing benzene required 110 h).21 As with alkyl halides, the rate of radical formation from alkyl thio ethers increases with the stability of the incipient carbon radical, in the order 3°>2°>1°>Me.20,21 Notably, only a handful of examples of the intramolecular olefinic trapping of radicals generated from unactivated alkyl thio ethers are on record21,22 and, to the best of our knowledge, no examples of similar intermolecular reactions are known – a probable consequence of the low rate of radical generation from unactivated thio ethers relative to undesired side reactions with the radicophilic alkene trap.23</p><!><p>Although halides and sulfonates have traditionally served as the electrophilic coupling partners in transition metal-catalyzed cross-coupling reactions, a resurgence in the use of organosulfur compounds in cross-coupling has sparked new interest in this long overlooked class of electrophiles.24 The first use of organosulfur coupling partners is found in the work of Takei25 and Wenkert26 on nickel-catalyzed Kumada-Corriu cross-coupling of aryl and alkenyl sulfides in the early 1970's through the mid-1980's. Since these early contributions, the cross-coupling of aryl, heteroaryl, alkenyl, alkynyl, allyl, benzyl, and acyl C–S electrophiles has recently undergone a renaissance, and sulfides, sulfoxides, sulfones, sulfoximines, and sulfonium salts have all been exploited as competent electrophiles.27 Organomagnesium, organozinc,28 organotin,29 and organoboron30 reagents have all been successfully employed as nucleophiles, and catalysis by both first-row (nickel, cobalt, iron) and second-row (palladium, rhodium) transition metals has been achieved.</p><p>In the context of the present work, it should be noted that methods for the nickel- or iron-catalyzed cross-coupling of both aryl sulfones25a,31 (Scheme 2, Eq. 1) and alkenyl sulfones32 (Scheme 2, Eq. 2) with alkyl and aryl Grignard reagents are already on record. Moreover, allylic sulfones have also served as competent electrophiles in copper-33 and iron-catalyzed34 displacements with Grignard reagents (Scheme 2, Eq. 3), nickel-35 and palladium-catalyzed36 alkylations with stabilized enolates, Lewis acid-mediated substitutions with organoalanes,37 and alkylations with lithium dialkylcuprates.38 Recently, Li and co-workers described an isolated example of a nickel-catalyzed cross-coupling of a benzylic sulfone with MeMgBr (Scheme 2, Eq. 4), as well as the nickel-catalyzed cross-coupling of α-keto sulfones with Grignard reagents (Scheme 2, Eq. 5).39 Although not discussed by the authors, the coupling of the α-keto sulfones is likely proceeding by initial enolization of the substrate by the basic Grignard reagent (pKa PhCOCH2SO2Ph = 11.4 in DMSO40), so that the β-oxido vinyl sulfone is the active electrophile; this would also account for the lack of Grignard reagent addition to the keto group.</p><p>However, despite extensive studies performed on the cross-coupling of C(sp2)–S electrophiles and, to a lesser extent, on allylic,26e,33–38,41 benzylic,39,42 or α-carbonyl-activated39,43 C(sp3)–S electrophiles, the use of unactivated C(sp3)–S electrophiles remains uncharted territory. To the best of our knowledge, only two reports of the use of simple alkyl sulfur electrophiles in transition metal-catalyzed cross-coupling are known, and in neither case is the reaction believed to occur by oxidative addition of the metal to the C(sp3)–S bond. Firstly, Vogel and Volla have described an iron-catalyzed desulfinylative cross-coupling of alkyl sulfonyl chlorides with Grignard reagents, in which the oxidative addition step is believed to occur at the S–Cl bond, followed by an extrusion of SO2 gas (Scheme 3, Eq. 1).44 Secondly, Nakamura and co-workers have developed a nickel-catalyzed alkenylative cross-coupling of alkyl phenyl thio ethers with Grignard reagents, in which the oxidative addition is thought to occur at the S–Ph bond (Scheme 3, Eq. 2).45 To date there have been no reports on a cross-coupling of unactivated C(sp3)–S electrophiles in which oxidative addition occurs to the C(sp3)–S bond – a necessary requirement for the alkylative cross-coupling of alkyl aryl organosulfur compounds.</p><!><p>Beside the usual difficulties encountered with alkyl electrophiles, the cross-coupling of alkyl sulfur electrophiles presents two additional challenges regarding the crucial oxidative addition step: (1) compared to alkyl halides, the C(sp3)–S bond is relatively unpolarized (χS–χC = 0.0346) with a higher energy σ*C–S orbital, and (2) the divalent sulfur atom necessarily bears two C–S bonds which must be distinguished in the oxidative addition step. Point (2) is of particular concern, as documented examples of oxidative addition of low-valent transition metals to alkyl aryl thio ethers47 and sulfones31a,b involve selective insertion into the C(sp2)–S bond, such that the compounds behave as aryl, and not alkyl, electrophiles.</p><!><p>The principal objectives of the current study are: (1) to discern the structural/electronic features of the unactivated alkyl sulfur electrophile (e.g. aryl group, sulfur oxidation level) that best facilitate the difficult oxidative addition to the C(sp3)–S bond, (2) to identify a suitable metal pre-catalyst for which the active, low-valent catalytic species produced in situ will undergo selective oxidative addition to the C(sp3)–S bond, and not the C(sp2)–S bond, of an alkyl aryl electrophile, (3) to deduce the stereochemical course of the reaction as an insight into the nature of the oxidative addition step, and (4) to establish the scope and limitations of the reaction from the point of view of the alkyl sulfur electrophile and the nucleophile, particularly with respect to the tolerance of β-heteroatomic groups on the substrate.</p><!><p>To address objective (1), initial studies focused on the use of alkyl aryl thio ether substrates bearing electron-poor aryl groups, in an effort to polarize the C(sp3)–S bond and facilitate oxidative addition. With regard to objective (2), a first-row transition metal catalyst was sought in order to encourage oxidative addition via a 1-electron as opposed to a 2-electron pathway.48 With the knowledge that low-valent nickel species undergo oxidative addition to alkyl aryl thio ethers at the undesired C(sp2)–S bond,47 iron salts were selected as catalysts for the initial experiments. PhMgBr was chosen as the nucleophile on the basis of its successful use in a large number of iron-catalyzed cross-couplings of alkyl halides.</p><!><p>Orienting experiments employed alkyl aryl thio ethers 3a-d as the substrates, PhMgBr (2.18 equiv,49 1.09 M solution in THF) as the nucleophile, FeCl3 (10 mol %) as the catalyst, and THF as the reaction solvent (Table 1). Each experiment was conducted in a GC vial under an argon atmosphere (without stirring) at room temperature for 1 h, followed by heating at 50 °C for a further 1 h. Although several different ligands were surveyed (dppm, dppe, PCy3, TMEDA, 2,2′-bipy, SIi-Pr•HCl (1,3-diisopropylimidazolidinium chloride)), only the results for TMEDA (12 mol %) as the ligand are reported (the results with the other ligands were almost identical within experimental error; see Supporting Information). The product distribution in each case was assessed by GC analysis, employing tetradecane (0.5 equiv) as an internal standard, and the observed retention times of products 4-7 were compared to those of their authentic samples.50 Whereas phenyl thio ether 3a and 4-trifluoromethylphenyl thio ether 3b were unreactive, pentafluorophenyl thio ether 3c and 2-pyridyl thio ether 3d both gave detectable (3-4%) amounts of the desired product 4, albeit with relatively larger amounts of undesired alkene 5 and alkane 6. Biphenyl, resulting from oxidative homocoupling of PhMgBr, was observed in all cases.51 Because of difficulties in quantifying unreacted pentafluorophenyl thio ether 3c by GC, 2-pyridyl thio ether 3d was selected as the substrate for further optimization studies.</p><p>A brief survey of established reaction conditions used for the iron-catalyzed cross-coupling of alkyl halides with Grignard reagents was next conducted,8a,e,f but all of these returned mainly starting material 3d and gave product 4 in a mere 1-6% yield (GC). The first sign of promise came with a slight modification (extended reaction time) of the protocol reported by Hayashi et al. for the cross-coupling of unactivated alkyl bromides with aryl Grignard reagents, which employs Fe(acac)3 (5 mol %) as the catalyst and 2.0 equiv of the requisite aryl Grignard reagent in Et2O at reflux.8b Thus, the treatment of 2-pyridyl thio ether 3d with PhMgBr (2.0 equiv, 2.91 M solution in Et2O) and Fe(acac)3 (5 mol %) in Et2O at reflux for 18 h led to a 24% conversion of 3d to afford 4 as the major product in 13% yield (GC), in addition to 5, 6, and 7 as minor products (Table 2, entry 1). Further optimization of the amounts of PhMgBr and Fe(acac)3 showed that 4.0 equiv of PhMgBr and 30 mol % of Fe(acac)3 led to the formation of desired product 4 in 49% yield (GC) (Table 2, entry 13), and these conditions were selected as an appropriate starting point for a more focused optimization study.</p><!><p>A variety of metal salts as catalysts were next surveyed, using n-Bu2O as the reaction solvent (due to volatility issues with Et2O on a small scale) (Figure 1). For consistency with previous reactions carried out in Et2O, a temperature of 45 °C was employed in all cases. Interestingly, iron salts proved uniquely effective in this transformation, and Ni(acac)2 or Co(acac)x (x = 2,3) produced only traces (2-3%) of 4, in addition to significant quantities of alkyl thiol 7 derived from C(sp2)–S bond cleavage. Notably, Ru(acac)3 was completely ineffective in the reaction. Of all the iron salts tested, Fe(acac)3 provided the highest yield (58% by GC) of 4, and consequently this metal salt was employed as the catalyst in all further studies.</p><!><p>The effect of the reaction solvent on the Fe(acac)3-catalyzed cross-coupling of thio ether 3d with PhMgBr (4.0 equiv, 2.91 M solution in Et2O) was next assessed, with all reactions conducted at ambient temperature for purposes of operational simplicity (Figure 2). Although the reaction generally performed well in dialkyl ether solvents (Et2O, n-Bu2O, i-Pr2O, MTBE, CPME), the use of more strongly coordinating THF proved detrimental, affording alkene 5 as the major product. Interestingly, a similar deleterious effect of THF on the Fe(acac)3-catalyzed cross-coupling of aryl Grignard reagents with unactivated alkyl bromides was noted by Hayashi et al.8b Ethers bearing more than one donor oxygen atom such as dioxane, DMM, DME, and diglyme were also poor reaction media (particularly for the latter two solvents, for which the reaction was largely suppressed). Notably, NMP, which is reportedly a beneficial co-solvent for other iron-catalyzed cross-couplings,52 strongly inhibited the reaction when employed as a co-solvent (9%) with n-Bu2O. Of all the solvents tested, CPME (cyclopentyl methyl ether)53 was by far the most effective, delivering 4 in 63% yield (GC), and it was thus selected as the (bulk) reaction solvent of choice for all further optimization. Additionally, in the knowledge that THF is clearly detrimental to the reaction, solutions of PhMgBr in Et2O were used exclusively for further studies.</p><!><p>The effect of the reaction temperature was also briefly investigated, employing CPME as the reaction solvent (Table 3). At 0 °C, the reaction failed to reach completion after 18 h, and a significant quantity (42%) of starting material 3d remained (Table 3, entry 1). Alternatively, at 45 °C, almost all of the starting material was consumed (93% conversion) but the yield of 4 was only 44% (GC) and the amounts of undesired products 5-7 increased, relative to the reaction run at room temperature (Table 3, c.f. entries 2 and 3). Accordingly, room temperature was maintained as the temperature of choice for further studies.</p><!><p>As tertiary amines such as TMEDA,8a,e,f,i,k,o,w,54 Et3N,8f DABCO,8f and HMTA (hexamethylenetetramine)8k,l have proven to be effective ligands in the iron-catalyzed cross-coupling of alkyl halides, the effect of amines on the cross-coupling reaction of 2-pyridyl thio ether 3d was next assessed (Figure 3).</p><p>Unfortunately, no further enhancement in the yield of product 4 was obtained for any of the amine ligands surveyed, and in many cases the amine proved detrimental to reactivity.55</p><p>Phosphine,8h,j,v-x phosphite,8h and NHC8h,a′ ligands have also found application in the iron-catalyzed cross-coupling of alkyl halides, and so a selection of these ligands was next assessed. Unfortunately, bidentate phosphines (dppm, dppe, dppp, dppf, DPEphos [bis(2-diphenylphosphino)phenyl ether]) and P(OPh)3 strongly suppressed the reaction, and both monodentate phosphines (PPh3, PCy3, t-Bu XPhos [2-di-tert-butylphosphino-2′,4′,6′-triisopropylbiphenyl]) and NHC ligands (SIi-Pr•HCl [1,3-diisopropylimidazolidinium chloride], IMes•HCl [1,3-bis(2,4,6-trimethylphenyl)-3H-imidazol-1-ium chloride]) led to incomplete conversion and an attendant increase in the amount of alkene 5 (see Supporting Information). Other additives employed in iron-catalyzed cross-coupling including LiCl,8t CsF (as a fluoride source56), and 4-fluorostyrene8t were also briefly evaluated but gave no enhancement.</p><!><p>With an optimized protocol in place for the Fe(acac)3-catalyzed cross-coupling of 2-pyridyl thio ether 3d with PhMgBr, the next phase focused on a reevaluation of the effect of the S-aryl group on the efficiency of the reaction. Thus, a variety of alkyl aryl thio ethers 3b-m bearing a diverse range of aryl groups were combined with PhMgBr (4.0 equiv, either 2.74 M or 3.04 M solution in Et2O) and Fe(acac)3 (30 mol %), in CPME at ambient temperature (Figure 4). For thio ethers 3b, 3c, and 3e bearing simple fluorinated phenyl groups, conversion of starting material was low (≤10%) and the desired product 4 was obtained in yields of only 6-10% (GC). In view of the relatively efficient cross-coupling of 2-pyridyl thio ether 3d, additional thio ethers 3f-m bearing proximal nitrogen atoms were also tested under the cross-coupling conditions. Although none outperformed 2-pyridyl thio ether 3d, all of the nitrogen-containing substrates 3f-m proved superior to the thio ethers 3b, 3c, and 3e bearing fluorinated phenyl groups, affording 4 in yields ranging from 20-59% (GC). In several cases, the alkyl thiol 7 derived from undesired cleavage of the C(sp2)–S bond was formed as a significant by-product.</p><!><p>With the optimization phase complete, the cross-coupling of 2-pyridyl thio ether 3d was executed on preparative scale (1.0 mmol) to obtain a yield of the isolated, cross-coupled product. To facilitate chromatographic separation of the product from the biaryl by-product (derived from homocoupling of the Grignard reagent), PhMgBr was replaced with 4-methoxyphenylmagnesium bromide as the nucleophile. Thus, treatment of 3d with 4-methoxyphenylmagnesium bromide (4.0 equiv, 2.17 M solution in Et2O) and Fe(acac)3 (30 mol %) in CPME at ambient temperature afforded 8 in 55% isolated yield (Scheme 4).</p><!><p>Although the cross-coupling of alkyl thio ethers developed thus far could potentially find use in the arylative functionalization of alkyl 2-pyridyl thio ethers generated by atom transfer radical additions of PTOC derivatives (PTOC = [(1H)-pyridine-2-thione]oxycarbonyl),57 we were motivated to remove the somewhat limiting requirement for a 2-pyridyl group on sulfur. Consequently, our attention turned to alkyl phenyl sulfones as alternative alkyl electrophiles, as these are readily accessible via oxidation of the corresponding thio ethers. Moreover, the phenyl sulfonyl group possesses a rich chemistry as an anion-stabilizing group in organic synthesis, facilitating alkylations, conjugate additions, and cycloadditions, as well as other useful C–C bond-forming transformations.58 Although a 2-step alkylation-desulfonylation sequence has long been employed as a strategy to effect the net replacement of a sulfonyl group with an alkyl group,59 the similar introduction of an aryl group cannot be achieved via this strategy. The obvious dilemma with alkyl sulfones as electrophiles for Kumada-type cross-coupling reactions is the possibility for competing deprotonation of the acidic α-protons flanking the −SO2Ph group [pKa PhSO2Et = 31.0 in DMSOM40] by the strongly basic Grignard reagents.60</p><!><p>To ascertain the viability of alkyl phenyl sulfones as alkyl electrophiles for cross-coupling, sulfone 9 was subjected to the reaction conditions previously optimized for 2-pyridyl thio ether 3d. Gratifyingly, 90% conversion of 9 occurred to give the desired product 4 in 48% yield (GC), in addition to alkene 5 in 19% yield (GC), as the major products (Scheme 5).50</p><!><p>To determine whether or not Fe(acac)3 is the most efficient catalyst for the cross-coupling of sulfone 9, a variety of metal salts as catalysts were next surveyed (Figure 5). As for 2-pyridyl thio ether 3d, iron salts proved uniquely effective in this transformation, and nickel, cobalt, copper or ruthenium based catalysts gave little or no desired product 4. Fe(acac)3 proved optimal and was thus employed as the catalyst in all further studies.</p><!><p>The influence of the reaction solvent was also reinvestigated and, as for 2-pyridyl thio ether 3d, CPME was optimal (Figure 6). Notably, THF again proved detrimental, significantly promoting the formation of alkene 5.</p><!><p>Although the inclusion of ligands proved ineffective for the earlier cross-coupling of 2-pyridyl thio ether 3d, a brief study of amine additives on the cross-coupling of sulfone 9 was next conducted. N-Methylmorpholine 10, pyridine 11, TMEDA 12, and PMDETA (N,N,N′,N′,N′′-pentamethyldiethylenetriamine) 13 were selected as representative monodentate (aliphatic and aromatic), bidentate, and tridentate amine ligands, respectively. Thus, sulfone 9 was combined with PhMgBr (4.0 equiv, 2.87 M solution in Et2O) and Fe(acac)3 (30 mol %) in CPME at ambient temperature in the presence of the corresponding amine additive (at loadings of 1.0, 2.0, 5.0, and 10.0 equiv (w.r.t. 9)) (Figure 7).</p><p>Although N-methylmorpholine 10 and pyridine 11 had little influence, regardless of loading, the effect of TMEDA 12 as an additive was striking. At lower loadings (1.0-2.0 equiv), the addition of TMEDA proved notably deleterious, promoting the formation of alkene 5. However, the reaction began to recover at 5.0 equiv of TMEDA, and at 10 equiv of TMEDA the yield of product 4 increased dramatically to 84% (GC). Satisfyingly, the yield of undesired alkene 5 was also reduced to only 2% (GC). In contrast to TMEDA, the use of tridentate amine PMDETA 13 led to complete suppression of the reaction.</p><!><p>Following some final optimization experiments to lower the amounts of Fe(acac)3, PhMgBr, and TMEDA (to 20 mol %, 3.0 equiv, and 8.0 equiv, respectively) (see Supporting Information), the scope of the cross-coupling with respect to the Grignard nucleophile 15 was next established, using sulfone 14 (1.00 mmol) as a representative substrate (Figure 8). In all cases, the Grignard reagent was employed in Et2O solution. Other than PhMgBr (15a, 74% yield of 16a), the reaction was also tolerant of electron-neutral aryl Grignard reagents bearing 4- or 3-substitution, including methyl groups (15b-c, 60-64% yield) or trimethylsilyl groups (15d-e, 56-60% yield). Notably, the trimethylsilyl moieties in products such as 16d and 16e are highly versatile handles for various ipso-functionalizations, including oxidation,61 halogenation,62 borylation,63 sulfonylation,64 nitration,64a acylation,64b or even recently developed gold-catalyzed alkylation65 or arylation66 protocols. A 3-isopropoxyphenylmagnesium bromide nucleophile (15f)67 also participated, affording the corresponding product 16f in 70% yield; the isopropyl group can readily be removed to reveal the phenol from such products if desired.68 Both 4-biphenyl (15g) and 2-naphthyl (15h) Grignard reagents also proved to be competent nucleophiles in the cross-coupling.69 The reaction did however prove highly sensitive to steric effects, as both the 2-methyl (15i) and 4-(tertbutyldimethylsilyloxy)methyl (15j) substituted phenylmagnesium bromides returned unreacted starting material 14. Moreover, Grignard reagents bearing electron-withdrawing (15k-m) or strongly electron-donating (15n-o)70 4-substituents failed to react. The origin of these steric and electronic effects is not clear at the present time, especially as several of the unreactive Grignard reagents (15i and 15k-m) have been used successfully in similar iron-catalyzed couplings of alkyl halides.8a-c,m,t,u,a′ Although alkyl (15r-u), alkenyl (15v), and alkynyl (15w) Grignard reagents were also briefly investigated, these gave either no reaction (15r-t and 15w) or led to predominant or even exclusive formation of the β-hydride elimination product 17 (15u and 15v).</p><!><p>The scope of the reaction with respect to the secondary alkyl phenyl sulfone coupling partner 18 was next assessed (Table 4). Although the functional group compatibility is inherently restricted due to the use of Grignard reagents at ambient temperature, tertiary amino groups (18b and 18i) and acetals (18g and 18h) were well tolerated. As evidenced by the cross-coupling of substrates 18c-d and 18 h-i, branching at the adjacent carbon(s) does not impede the reaction. Notably, with norbornyl substrate 18c, the reaction proceeded with excellent diastereoselectivity (98:2 exo:endo). Homobenzylic sulfone 18f was cross-coupled uneventfully, despite the anticipated potential for β-hydride elimination to generate a conjugated alkene. Unfortunately, β-heteroatoms on the sulfone substrate were poorly tolerated (18j-l), although pyrrolidine 18j did afford a low yield (25%) of the desired product, 9j. In the case of tetrahydropyran substrate 18k, the reaction returned a ~4:1 mixture of starting material 18k to 4,5-diphenylpent-4-en-1-ol 20 (configuration undetermined).</p><p>To examine whether primary sulfones were viable substrates for the cross-coupling reaction, 21 was subjected to the optimized reaction conditions with PhMgBr as the nucleophile. However, the reaction did not prove synthetically useful, and led to incomplete (80%) conversion of 21 to give product 22 in only 27% yield (GC), along with trace amounts (<5%) of alkene 23 and alkane 24, accounting for <45% of the mass balance (Scheme 6). The remainder of the mass comprised a complex mixture of unidentified products which were insufficiently volatile to detect under the GC conditions employed.</p><p>Similarly, to determine whether tertiary sulfones are competent electrophiles, 25 was subjected to the optimized reaction conditions, again with PhMgBr as the nucleophile. Under these conditions, ~70% conversion of 25 occurred to give a 87:13 mixture of alkenes 26 and 27, respectively, and no peaks consistent with the desired cross-coupled product were observed by 1H NMR spectroscopy (Scheme 7).</p><!><p>To ascertain the stereochemical course of the cross-coupling reaction of alkyl 2-pyridyl thio ether 3d, an enantioenriched sample of (R)-3d (98.8:1.2 er) was prepared in a two-step bromination-displacement sequence from commercially available (R)-4-phenylbutan-2-ol 28. Following subjection to the optimized reaction conditions with 4-methoxyphenylmagnesium bromide as the nucleophile, product 8 was isolated in 53% yield and found to be racemic (Scheme 8).</p><!><p>The steric course of the cross-coupling of secondary alkyl phenyl sulfones was next assessed. An enantiopure sample of sulfone (R)-9 (>99.5:0.5 er) was first prepared from bromide (S)-29 via a thiolate displacement-oxidation sequence. Cross-coupling of sulfone (R)-9 with 3-isopropoxyphenylmagnesium bromide under the optimized conditions then gave 19e in 52% isolated yield as a racemic mixture (Scheme 9).</p><!><p>The nature of the S-aryl group of alkyl aryl thio ether substrates 3 proved critical in enabling oxidative addition of the low-valent iron species to the C(sp3)–S bond. The original hypothesis supposed that electron-deficient S-aryl groups may facilitate the oxidative addition step by polarizing the C(sp3)–S bond; however, thio ethers 3b, 3c, and 3e bearing simple fluorinated phenyl groups on sulfur showed little reactivity under the optimized conditions (≤10% conversion of 3 to give 4 in yields of 6-10% by GC). Unexpectedly, substrate 3d bearing a 2-pyridylthio group underwent efficient cross-coupling under the same conditions (63% yield of 4 by GC). This result suggests the possibility of an oxidative addition of the C(sp3)–S bond to the iron center which may be assisted by coordination of the 2-pyridyl group of 3d to the metal, rendering the process pseudointramolecular. Notably, the 2-pyridyl group has been employed in a similar role for the oxidative addition of unactivated C(sp3)–O bonds of alkyl 2-pyridyl ethers to ruthenium71 or iridium72 complexes. In addition, a 2-pyrimidyl group on sulfur was beneficial in the iron-catalyzed cross-coupling of alkenyl thio ethers with aryl Grignard reagents.73 Interestingly, thio ethers 3g and 3i-k bearing proximal pyridyl-type nitrogen atoms on the S-aryl moiety also showed significant competence in the cross-coupling with PhMgBr (48-59% yield of 4 by GC). Even the electron-rich thio ethers 3l and 3m bearing proximal dimethylamino groups proved relatively efficient as electrophiles (30-37% yield of 4 by GC), at least when compared to thio ethers 3b, 3c, and 3e bearing simple fluorinated phenyl groups (6-10% yield of 4 by GC). Notably, 3-pyridyl thio ether 3f, in which the pyridyl nitrogen atom is less well disposed to steer the iron center toward the C(sp3)–S bond, displayed significantly lower reactivity than its 2-pyridyl counterpart 3d (26% of 4 for 3f versus 63% for 3d). Taken together, these results lend credence to the notion that the "directing effect" of a proximal nitrogen atom on the S-aryl group is indeed key to reaction efficiency.</p><!><p>Any mechanistic proposal for iron-catalyzed cross-coupling reactions of Grignard reagents must be tempered with the caveat that the oxidation states of the low-valent catalytic species are obscure in many cases, and are likely to be dependent on the precise reaction conditions, as well as the presence of ligands or other additives.7 On the basis of extensive studies of characterized, low-valent organoiron species, Fürstner has proposed that an Fe(0)/Fe(–II) catalytic manifold is likely operative in low temperature cross-coupling reactions employing alkyl Grignard reagents bearing β-hydrogens.8m,74 However, in the case of aryl Grignard reagents, there is no conclusive evidence that oxidation states as low as Fe(–II) are kinetically accessible from Fe(II) or Fe(III) pre-catalysts under the conditions generally employed in preparative cross-coupling reactions. An Fe(I)/Fe(III) catalytic cycle, originally proposed by Kochi on the basis of by-product analysis and EPR measurements,75 has been suggested to be the most plausible pathway (at least under the specific reaction conditions) in the cross-coupling of alkyl halides76 and aryl halides77 with alkyl Grignard reagents, and benzyl halides with arylzinc reagents.78 An Fe(II)/Fe(III) catalytic cycle for the cross-coupling of aryl Grignard reagents with alkyl halides in the presence of TMEDA has also found experimental support in studies of isolated organoiron species.54 This catalytic manifold has similarly been invoked in related couplings of arylborates8r and alkynyl Grignard reagents8v on the basis that nucleophile homocoupling products were not observed when using Fe(II) pre-catalysts.8r,v Although soluble Fe(0) species have seldom been postulated as the active agents in iron-catalyzed couplings of aryl Grignard reagents with alkyl halides,8k,79 Bedford and co-workers have clearly demonstrated that Fe(0) nanoparticles are produced from the reduction of FeCl3-dpph or FeCl3-PEG (PEG = polyethylene glycol) with 4-tolylmagnesium bromide, and that they are catalytically active in the cross-coupling of alkyl halides.8g However, the possibility that the nanoparticles served merely as a reservoir for catalytically active, soluble Fe(0) species could not be ruled out. Similarly, Krafft and Holton have shown that the addition of 3.0 equiv of MeMgBr to FeCl3 in Et2O leads initially to a finely divided black powder (speculated to be "Me2Fe(II)Ln") which, over the course of 1 h at room temperature, undergoes decomposition to Fe(0) (as characterized by elemental analysis and titration with potassium chromate).80 Similarly, solutions of Fe(III) salts including FeCl3 and Fe(acac)3, in THF solution at room temperature are reduced to Fe(0) nanoparticles upon treatment with various Grignard reagents (3.0 equiv), including EtMgBr and PhMgBr.81</p><p>On the basis of these observations, and the fact that reaction mixtures in this study rapidly turn black on addition of the aryl Grignard reagent, it is proposed that the Fe(acac)3 pre-catalyst in the cross-coupling of thio ether 3d is initially reduced by the aryl Grignard reagent to Fe(0) nanoparticles (Scheme 10).82 The Fe(0) species 30, whether part of a nanocluster or a soluble, mononuclear Fe(0) complex, may then react further with the arylmagnesium reagent to give a catalytically active Fe(0)(aryl)n ferrate species 31. Notably, a homoleptic Fe(0) ferrate species [Ph4Fe(0)][Li(OEt2)]4, prepared from the reaction of FeCl3 with excess PhLi in Et2O at low temperature, has been isolated and characterized by X-ray crystallography, and shown to be active in the reduction of N2, possibly by pre-complexation of the π-acceptor N2 molecule to the iron center.83 Following the generation of the catalytically active Fe(0)(aryl)n ferrate species 31, coordination of the thio ether substrate 3d through the pyridyl nitrogen atom may give adduct 32. Oxidative addition of the C(sp3)–S bond to the iron center (vide infra) could then generate Fe(II) species 33, which would suffer reductive elimination to afford the product 34 and furnish an Fe(0)(aryl)n–1 species 35. Finally, arylation of 35 by the aryl Grignard reagent would regenerate the active catalyst species 31.</p><p>The catalytic cycle in Scheme 10 does not address the elementary steps in the crucial oxidative addition, and there are several scenarios by which this may proceed (Scheme 11). One possibility, depicted in route (a), is that a directed C(sp3)–H bond activation may occur from adduct 32 to furnish a cyclometalated species 36. Following α-elimination of the thiolato group to produce an iron carbene 37, a 1,2-hydride shift could give the alkyliron intermediate 33, which is primed for reductive elimination to the product 34. An analogous pathway is probably operative in the (formal) oxidative addition of the unactivated C(sp3)–O bonds of alkyl 2-pyridyl ethers to ruthenium71 or iridium,72 and finds its origin in studies on the activation of the C(sp3)–O bond of anisole with iridium pincer complexes.84 An alternative scenario is an inner-sphere electron transfer from the iron center to (presumably) the π* orbital of the pyridine ring within adduct 32, initiating C(sp3)–S bond cleavage to give a transient alkyl radical 38 which undergoes recombination with the iron center (route (b)), affording the same alkyliron intermediate 33 as route (a). Alternatively, the alkyl radical 38 could potentially attack the arene ring attached to iron to generate cyclohexadienyl radical 39, followed by rearomatization by loss of the metal fragment (route (c)).8r,x,a′ There is a general consensus, based on stereochemical studies and "radical clock" experiments, that the oxidative addition of alkyl halides with lowvalent iron species follows a radical-based oxidative addition pathway. However, our finding that the cross-coupling of enantiopure (R)-3d proceeds with complete racemization does not allow distinction between any of the aforementioned pathways, although it does argue against a concerted oxidative addition step, which should be highly enantiospecific.85</p><p>A final point to make concerns catalyst deactivation, and this is likely the origin of the rather high loading of Fe(acac)3 (30 mol%), as well as the large excess of Grignard reagent (4.0 equiv) required in the cross-coupling of 3d. Oligomerization of low-valent iron has been proposed as a possible deactivation pathway in iron-catalyzed cross-coupling,75c and it has been suggested that more reactive electrophiles (i.e. faster oxidative addition) or appropriate stabilizing ligands/additives can serve to minimize the unproductive aggregation of the lowvalent iron species.77 In the present case, it may be that the relatively unreactive nature of thio ether 3d (with respect to alkyl halides) and the lack of a stabilizing ligand/additive may serve to enable catalyst deactivation.</p><!><p>A plausible catalytic cycle for the Fe(acac)3-catalyzed cross-coupling of sulfone 9 with an aryl Grignard reagent in the presence of excess TMEDA is depicted in Scheme 12. On the basis of studies by Nakamura and co-workers,54 it is possible that reduction of the Fe(acac)3 pre-catalyst with an aryl Grignard reagent in the presence of TMEDA gives (TMEDA)Fe(II)aryl2 40 as the initial low-valent iron species. Similarly, Sen and co-workers have observed (TMEDA)Fe(II)Bn2 as an intermediate during coupling reactions of benzyl halides mediated by [CpFe(0)(COD)][Li(TMEDA)].86 It should be noted that many other examples of bidentate ligands, often nitrogen-based, are known to stabilize Fe(II)alkyl2 species against reductive elimination to Fe(0).87,88 In the present case, the requirement for such a large excess (8.0 equiv) of TMEDA in the cross-coupling of sulfone 9, as well as notable reactivity differences with certain aryl Grignard reagents, suggests that the catalytic cycle deviates from that put forward by Nakamura for the coupling of alkyl halides. It is thus proposed that (TMEDA)Fe(II)aryl2 40 undergoes reversible association with a second molecule of TMEDA to generate a more electron-rich species trans-(TMEDA)2Fe(II)aryl2 41. Similar di-complexes of Fe(II) with TMEDA are known: trans-[FeCl2(TMEDA)2] has been characterized by X-ray crystallography and shown to be unstable with respect to the binuclear complex [{FeCl(TMEDA)}2(μ-Cl)2] 45, except in the presence of an excess of TMEDA.89 Similarly, cis-(2,2′-bipy)2Fe(II)Et2 has been isolated and characterized by X-ray crystallography.87b,90</p><p>Coordinatively-saturated complex 41 could then engage sulfone 9 by an outer-sphere electron transfer (presumably to the π* orbital of the SO2Ph moiety) to generate a transient radical anion that collapses to alkyl radical 38 and a phenylsulfinate anion; this radical could then attack the arene ligand on iron in an ipso-substitution reaction8r,x,a′ to afford a cyclohexadienyl radical 43, which would then expel the iron fragment 44 and furnish the product 34. A very similar catalytic cycle has been proposed by Nakamura and co-workers in their iron-catalyzed cross-coupling of halohydrins with aryl aluminum reagents: specifically, the ferrate intermediate 47 was suggested to be the active species which transfers an electron to the halide,8x and this intermediate is directly analogous to 41 in this catalytic cycle. This mechanistic proposal is consistent with the fact that enantiopure (R)-9 undergoes cross-coupling with 3-isopropoxyphenylmagnesium bromide to give 19e as a racemic mixture.</p><p>The requirement for such a large excess (8.0 equiv) of TMEDA in the cross-coupling reaction is intriguing. Although TMEDA is a crucial additive in some previously reported iron-catalyzed cross-couplings of alkyl halides with Grignard reagents,8a,e,f,i,k,o,w,54 its role in the reaction is not always clear. The large excess required in this case is likely a consequence of interaction with PhMgBr as well as the iron species, and PhMgBr with TMEDA in THF-d8 has been shown to afford a mixture of PhMgBr(TMEDA), Ph2Mg(TMEDA), and MgBr2(TMEDA)n(THF)2–n (n = 1, 2).54 According to the current mechanistic hypothesis, the catalytically active intermediate 41 can be generated only in the presence of excess (i.e. free) TMEDA and, for this to be possible, the PhMgBr (and Ph2Mg and MgBr2) must first be saturated with TMEDA.</p><!><p>Although the cross-coupling reaction was successfully performed with a variety of electron-neutral 3- and 4-substituted aryl Grignard reagents 15a-h, the use of sterically-encumbered nucleophiles 15i and 15j or those bearing electron-withdrawing (15k-m) or strongly electron-donating (15n-o) 4-substituents led to no reaction (Figure 10). The catalytic cycle proposed in Scheme 12, does not allow a rationalization of these particular results. However, it is clear that an intermediate like 41 derived from non-aromatic Grignard reagents could not react productively with alkyl radical 38, which may be the reason that alkyl (15r-t), alkenyl (15v), and alkynyl (15w) nucleophiles do not react. In the case of alkyl Grignard reagents bearing β-hydrogens, reduction of the iron to oxidation states as low as Fe(–II) becomes possible,8m,74 and this may be why EtMgBr (15u) leads to complete consumption of the sulfone 14, albeit to give the β-hydride elimination product 17.</p><!><p>Although the cross-coupling reaction proved applicable to a variety of unactivated, secondary alkyl phenyl sulfones (Table 4, entries 1-10), the reaction was largely unsuccessful with substrates bearing β-heteroatom substituents (Table 4, entries 10-13). Only in the case of pyrrolidine sulfone 18j could any cross-coupling product be isolated (25% of 19j). For tetrahydropyran sulfone 18k, the major product, which was not isolated in pure form, was 4,5-diphenylpent-4-en-1-ol 20 of unassigned configuration. This compound presumably arises from an E1cB elimination process followed by cross-coupling of the resultant vinyl sulfone with PhMgBr.32</p><p>With respect to the poor cross-coupling efficiency and low mass balance in the reaction of primary sulfones such as 21, it is likely that α-deprotonation of the primary sulfone (which is both kinetically and thermodynamically more acidic than a similar secondary sulfone) by the PhMgBr is occurring,60 followed by oxidation of the resultant carbanion by Fe(III) to give a variety of possible products, including vinyl sulfones or dimeric products such as vicinal disulfones or symmetrical alkenes91 (Scheme 13). These initial sulfone by-products could then undergo further reaction (e.g. cross-coupling) under the reaction conditions, accounting for the complex mixture. This undesired α-deprotonation process may also account for the incomplete mass balance (88%) observed in the cross-coupling of secondary sulfone 9 under the optimized conditions (see data in Supporting Information), implying that this side reaction is operative, albeit to a much lesser extent, in the coupling of secondary alkyl sulfones.</p><!><p>This study chronicles the first attempts to systematically explore the potential of unactivated aliphatic sulfur compounds as electrophiles in transition metal-catalyzed cross-coupling. The first phase of the investigation focused on discerning the structural and electronic features of the alkyl sulfur substrate which enable the difficult oxidative addition to the C(sp3)–S bond, in an iron-catalyzed cross-coupling of unactivated alkyl aryl thio ethers 3 with aryl Grignard reagents. Through extensive optimization efforts, a critical role of a nitrogen "directing group" on the S-aryl moiety of thio ethers 3 was uncovered which served to facilitate the crucial oxidative addition step. Thus, employing 2-pyridyl thio ether 3d as the electrophile, PhMgBr as the nucleophile, and Fe(acac)3 as the catalyst, the first example of the cross-coupling of an unactivated alkyl aryl thio ether was achieved. In addition, alkyl phenyl sulfones were found to be effective electrophiles in the Fe(acac)3-catalyzed cross-coupling with aryl Grignard reagents. A thorough assessment of the various reaction parameters revealed a dramatic enhancement in reaction efficiency with an excess of TMEDA (8.0 equiv). The optimized reaction protocol was used to evaluate the scope of the method with respect to both the Grignard nucleophile and sulfone electrophile.</p><p>Although the motivation behind this work was principally the development of a new cross-coupling process, it was essential to draw upon existing mechanistic studies of the iron-catalyzed cross-coupling of alkyl halides to present plausible (albeit speculative) reaction mechanisms and catalytic cycles for the new reactions. In light of the myriad challenges accompanying the study of processes mediated by low-valent iron species, a deeper understanding of the mechanistic underpinnings of the reactions described herein would necessarily be the focus of an independent investigation, and future efforts are currently directed toward this goal.</p><!><p>An oven-dried, 25-mL, one-necked, round-bottomed flask was charged with the requisite sulfone (1.00 mmol, 1.0 equiv) and Fe(acac)3 (70.6 mg, 0.20 mmol, 20 mol %) in a glove box then was sealed with a rubber septum and removed from the box. Outside of the glove box, a 25-mL Schlenk flask equipped with a stirrer bar, rubber septum and argon inlet was evacuated, flame-dried, left to cool under vacuum, and flushed three times with argon. TMEDA (930 mg, 1.20 mL, 8.00 mmol, 8.0 equiv) was added via syringe to the Schlenk flask and stirring was commenced. The round-bottomed flask containing the sulfone and Fe(acac)3 was charged with CPME (4.0 mL) then sonicated until homogeneous. The clear red solution was then transferred via cannula to the Schlenk flask holding the TMEDA, and the residual material was rinsed across with further portions of CPME (6.0 mL). The requisite aryl Grignard reagent (solution in Et2O, 3.00 mmol, 3.0 equiv) was then added by syringe over ca. 30 sec. During addition, the color of the solution changed from red to pale yellow to brown, but remained clear throughout, and no visible deposits were formed on the edges of the flask. After stirring for 18 h at rt, 1 M aq. HCl (10 mL) was added in one portion and the mixture was filtered through a pad of Celite (5 g) in a 40 mm Ø, porosity 3, sintered funnel under House vacuum. EtOAc (2 × 5 mL) was used to rinse any residual material though the Celite pad. The filtrate was transferred to a separatory funnel and the layers were separated. The organic layer was washed with 1 M aq. HCl (2 × 10 mL) and the combined aqueous layers were extracted with EtOAc (2 × 10 mL). The combined organic layers were dried (MgSO4), filtered and concentrated in vacuo (50 °C, ca. 5 mmHg).</p><!><p> </p><!><p>A 25-mL, one-necked, round-bottomed flask equipped with a stirrer bar, rubber septum and argon inlet was charged with 28 (98%, 613 mg, 627 μL, 4.00 mmol), triphenylphosphine (1.59 g, 6.00 mmol), imidazole (413 mg, 6.00 mmol) and CH2Cl2 (8.0 mL), and stirring was commenced. The resultant mixture was then cooled in an ice/water bath and iodine (1.52 g, 6.00 mmol) was added in one portion. The mixture was allowed to warm to rt over 2 h and pentane (25 mL) was added. The mixture was then rinsed through a pad of neutral alumina (5 g) using minimal pentane and concentrated in vacuo (50 °C, ca. 5 mmHg) to give a colorless oil (1.05 g). Purification via short-path distillation at reduced pressure (0.5 mmHg) gave 48 as a clear, colorless oil (773 mg, 74%). The 1H NMR spectroscopic data and boiling point matched that for alternative preparations.92 Data for 48: bp: 104-105 °C (0.5 mmHg).</p><p> </p><!><p>This preparation is based on a previously reported method for the iron-catalyzed cross-coupling of aryl Grignard reagents with alkyl halides.93 In a glove box, FeCl3 (16.2 mg, 10 mol %) was added to a flame-dried, 20-mL scintillation vial, which was then sealed with a rubber septum and removed from the box. CH2Cl2 (2.0 mL) and TMEDA (11.6 mg, 15.0 μL, 0.10 mmol) were added sequentially via syringe to give a rust-colored suspension of black FeCl3, and the solvent was removed in vacuo (50 °C, ca. 5 mmHg) to give a rust-colored solid residue. Et2O (3.0 mL) was then added to give a suspension of the latter solid. 48 (260 mg, 178 μL, 1.00 mmol) was added via syringe, followed by dropwise addition of PhMgBr (2.70 M in Et2O, 742 μL, 2.00 mmol). A significant exotherm occurred and some of the Et2O evaporated. The reaction mixture was transferred via syringe to a flame-dried, 10-mL, round-bottomed flask equipped with a stirrer bar and reflux condenser and the residual mixture was rinsed across with an additional portion of Et2O (0.5 mL). The mixture was then heated at reflux for 30 min, then quenched by addition of H2O (5 mL). The mixture was rinsed through a pad of Celite (5 g) using minimal EtOAc and the layers were separated. The aqueous layer was extracted with EtOAc (2 × 5 mL) and the combined organic extracts were dried (MgSO4), filtered and concentrated in vacuo (50 °C, ca. 5 mmHg) to give a yellow oil (223 mg). Purification via flash column chromatography (30 g of SiO2, 30 mm Ø, hexane, ca. 5-mL fractions) gave 4 as a clear, colorless oil (121 mg, 57%). The 1H NMR spectroscopic data matched that for alternative preparations.94 Data for 4: GC: tR 3.82 min.</p><p> </p><!><p>A 1-L, one-necked, round-bottomed flask equipped with a stirrer bar and rubber septum was charged with 4-phenylbutan-2-one (25.0 g, 165 mmol, 1.0 equiv) and MeOH (400 mL), and stirring was commenced. The mixture was cooled to 4 °C in an ice/water bath then sodium borohydride (6.88 g, 182 mmol, 1.1 equiv) was added portionwise (the internal temperature did not exceed 17 °C). The resultant turbid, colorless mixture was stirred in the ice/water bath for 20 min, then allowed to warm to rt over 65 h. The mixture was then concentrated in vacuo (50 °C, ca. 5 mmHg) and partitioned between EtOAc (200 mL) and H2O (200 mL). The layers were separated and the aqueous layer was extracted with EtOAc (2 × 100 mL), then the combined organic extracts were dried (MgSO4), filtered, and concentrated in vacuo (50 °C, ca. 5 mmHg) to give a cloudy, colorless oil (25.3 g). Purification via short-path distillation at reduced pressure (0.5 mmHg) gave 28 as a clear, colorless oil (24.3 g, 98%). The 1H NMR spectroscopic data and boiling point matched that for alternative preparations.95 Data for 28: bp: 77-78 °C (0.5 mmHg) [lit.96 75 °C (0.3 mmHg)].</p><p> </p><!><p>Bromine (11.6 g, 3.70 mL, 72.0 mmol, 1.2 equiv) was added dropwise from a 5-mL measuring cylinder by Pasteur pipette to a stirred suspension of triphenylphosphine (19.08 g, 72.0 mmol, 1.2 equiv) in CH2Cl2 (200 mL) in a 1 L, single-necked, round-bottomed flask equipped with a stirrer bar and cooled in an ice/water bath (open to air). The flask was then sealed with a rubber septum and purged with argon via an inlet needle. After stirring the resultant pale yellow suspension for 15 min, a solution of 28 (9.01 g, 60.0 mmol, 1.0 equiv) and imidazole (4.95 g, 72.0 mmol, 1.2 equiv) in CH2Cl2 (100 mL) was added via cannula over ca. 5 min. The cooling bath was removed and the reaction mixture was allowed to warm to rt over 27 h. The mixture was then filtered through a 40 mm Ø, porosity 3, sintered funnel under House vacuum and carefully concentrated in vacuo to leave a yellow oil residue (i.e. avoiding precipitating the phosphorus-containing residues at this point). A stirrer bar was added to the residue and a wide-neck plastic funnel was added to the neck of the flask, and rapid stirring was commenced. Pentane (300 mL) was quickly added in one portion to precipitate the phosphorus-containing residues as a fine white solid. The mixture was filtered through a 40 mm Ø, porosity 3, sintered funnel under House vacuum and concentrated in vacuo (50 °C, ca. 5 mmHg) to give a clear, colorless oil (14.26 g). Purification via short-path distillation at reduced pressure (0.5 mmHg) gave 29 as a clear, colorless oil (11.58 g, 91%). The 1H NMR spectroscopic data and boiling point matched that for alternative preparations.97 Data for 29: bp: 59-60 °C (0.5 mmHg) [lit.96 60-65 °C (0.2 mmHg)].</p><p> </p><!><p>This preparation is based on a previously reported method for the iron-catalyzed homocoupling of alkyl bromides.98 A flame-dried, 50-mL, one-necked, round-bottomed flask equipped with a stirrer bar, rubber septum and argon inlet was charged with magnesium turnings (389 mg, 16.0 mmol, 2.0 equiv), Fe(acac)3 (Strem, 99%, 57.1 mg, 0.16 mmol, 2 mol %), and THF (24.0 mL). 29 (1.70 g, 1.38 mL, 8.00 mmol, 1.0 equiv) was then added via syringe and the resultant mixture was stirred at rt. After ca. 15 min the solution changed color from red to black. After stirring for 1 h 40 min, the mixture was filtered through a pad of Florisil (5 g) in a 40 mm Ø, porosity 3, sintered funnel using minimal EtOAc and concentrated in vacuo (50 °C, ca. 5 mmHg) to give a yellow oil (973 mg). Purification via flash column chromatography (50 g SiO2, 30 mm Ø, hexane, ca. 5-mL fractions) gave a ~1:1 mixture of (l)- and (u)-49 as a clear, pale yellow oil (177 mg, 17%). Data for (l)- and (u)-49: 1H NMR (500 MHz, CDCl3) 7.31-7.24 (for both diastereoisomers: m, 4 H each), 7.22-7.14 (for both diastereoisomers: m, 6 H each), 2.71-2.41 (for both diastereoisomers: m, 4 H each), 1.71-1.33 (for both diastereoisomers: m, 6 H), 0.92 (for one diastereoisomer: d, J = 6.3 Hz, 6 H), 0.86 (for one diastereoisomer: d, J = 6.2 Hz, 6 H); 13C NMR (125 MHz, CDCl3) 143.4, 143.4, 128.7, 128.7, 128.6, 128.6, 125.9, 125.9, 37.4, 37.2, 36.6, 35.3, 34.4, 34.3, 16.7, 14.7; MS: (EI+, 70 eV) 266.2 (M+, 8), 91.1 (C7H7+, 100), 65.1 (16); HRMS (EI+, double focusing sector field) calcd for C20H26: 266.2035, found: 266.2031; TLC Rf 0.38 (hexane) [KMnO4]; GC: first diastereoisomer, tR 5.14 min (47%); second diastereoisomer, tR 5.16 min (53%).</p><!><p> </p><!><p>A 500-mL, one-necked, round-bottomed flask equipped with a stirrer bar, water-jacketed reflux condenser, and argon inlet was charged with 29 (6.39 g, 30.0 mmol, 1.0 equiv), thiophenol (3.41 g, 3.18 mL, 30.0 mmol, 1.0 equiv), potassium carbonate (8.29 g, 60.0 mmol, 2.0 equiv), and acetone (150 mL), and stirring was commenced. The resultant mixture was heated at reflux for 24 h, and was then allowed to cool to rt. The mixture was filtered through a 40 mm Ø, porosity 3, sintered funnel under House vacuum and concentrated in vacuo (50 °C, ca. 5 mmHg) to give a pale yellow oil (7.69 g). Purification via short-path distillation at reduced pressure (0.5 mmHg) gave 3a as a clear, colorless oil (7.07 g, 97%). The 1H NMR spectroscopic data and boiling point matched that for alternative preparations.99 Data for 3a: bp: 136-137 °C (0.5 mmHg) [lit. 124-126 °C (0.1 mmHg)]; GC: tR 4.66 min.</p><p> </p><!><p>A 25-mL, one-necked, round-bottomed flask equipped with a stirrer bar, water-jacketed reflux condenser, and argon inlet was charged with 29 (213 mg, 1.00 mmol, 1.0 equiv), (4-trifluoromethyl)thiophenol (184 mg, 141 μL, 1.00 mmol, 1.0 equiv), potassium carbonate (276 g, 2.00 mmol, 2.0 equiv), and acetone (5.0 mL), and stirring was commenced. The resultant mixture was heated at reflux for 16 h, and was then allowed to cool to rt. The mixture was filtered through a 40 mm Ø, porosity 3, sintered funnel under House vacuum and concentrated in vacuo (50 °C, ca. 5 mmHg) to give a yellow-orange oil (323 mg). Purification via flash column chromatography (20 g SiO2, 20 mm Ø, 95:5, hexane/toluene, ca. 5-mL fractions) gave 3b as a clear, colorless oil (292 mg, 94%). Data for 3b: 1H NMR (500 MHz, CDCl3) 7.49 (d, J = 8.1, 2 H), 7.37-7.27 (m, 4 H), 7.24-7.15 (m, 3 H), 3.36-3.28 (m, 1 H), 2.87-2.73 (m, 2 H), 2.03-1.82 (m, 2 H), 1.38 (dd, J = 6.7, 1.8 Hz, 3 H); 13C NMR (125 MHz, CDCl3) 141.5, 141.5, 129.9, 128.7, 128.7, 128.2 (q, J = 32.7 Hz), 126.3, 125.8 (q, J = 3.8 Hz), 124.4 (q, J = 272 Hz), 41.6, 38.4, 33.3, 21.2; IR (neat) 3085 (w), 3064 (w), 3027 (w), 2963 (m), 2926 (m), 2860 (w), 1607 (s), 1496 (m), 1454 (m), 1401 (m), 1377 (w), 1326 (s), 1165 (s), 1124 (s), 1095 (s), 1063 (s), 1030 (w), 1013 (m), 948 (w), 914 (w), 826 (m), 779 (w), 747 (m), 699 (m), 593 (w); MS (EI+, 70 eV) 310.1 (M+, 17), 132.1 (36), 117.1 (27), 91.1 (C7H7+, 100), 65.1 (18); HRMS (EI+, double focusing sector field) calcd for C17H17F3S: 310.1003, found: 310.1007; TLC Rf 0.36 (99:1, hexane/EtOAc) [KMnO4]; GC tR 4.50 min.</p><p> </p><!><p>A 200-mL, one-necked, round-bottomed flask equipped with a stirrer bar, water-jacketed reflux condenser, and argon inlet was charged with 29 (2.56 g, 12.0 mmol, 1.0 equiv), potassium thioacetate (4.20 g, 36.0 mmol, 3.0 equiv), and DMF (30 mL), and stirring was commenced. The resultant mixture was heated at 100 °C for 17 h, and was then allowed to cool to rt. 4 M aq. NaOH (9.0 mL, 36.0 mmol, 3.0 equiv) was added and stirring was continued for a further 32 h, then the mixture was cooled in an ice/water bath and 3 M aq. H2SO4 was added to pH 2. The mixture was then partitioned between EtOAc (75 mL) and H2O (300 mL) and the layers were separated. The aqueous layer was extracted with EtOAc (2 × 75 mL) and the combined organic extracts were washed with brine (4 × 50 mL), then dried (MgSO4), filtered, and concentrated in vacuo (50 °C, ca. 5 mmHg) to give a dark brown oil (2.90 g). Purification via flash column chromatography (30 g SiO2, 30 mm Ø, 98:2, hexane/EtOAc, ca. 5-mL fractions) gave 7 as a clear, yellow oil (1.23 g, 62%). Data for 7: 1H NMR (500 MHz, CDCl3) 7.35-7.29 (m, 2 H), 7.26-7.20 (m, 3 H), 2.99-2.90 (m, 1 H), 2.86-2.70 (m, 2 H), 1.99-1.80 (m, 2 H), 1.54 (d, J = 6.5 Hz, 1 H, SH), 1.41 (d, J = 6.8 Hz, 3 H); 13C NMR (125 MHz, CDCl3) 141.8, 128.7, 128.7, 126.2, 42.8, 35.3, 33.9, 26.1; IR (neat) 3084 (m), 3061 (m), 3026 (m), 2956 (m), 2922 (m), 2860 (m), 1603 (m), 1496 (m), 1453 (m), 1376 (m), 1030 (m), 747 (m), 699 (s); MS (EI+, 70 eV) 166.1 (M+, 19), 132.1 (33), 117.1 (72), 105.1 (11), 91.1 (C7H7+, 100), 77.0 (17), 65.1 (32), 63.1 (12), 61.1 (16), 51.0 (19); HRMS (EI+, double focusing sector field) calcd for C10H14S: 166.0816, found: 166.0815; TLC Rf 0.38 (99:1, hexane/EtOAc) [KMnO4]; GC tR 2.58 min.</p><p> </p><!><p>A 25-mL, one-necked, round-bottomed flask equipped with a stirrer bar, rubber septum, and argon inlet was charged with 7 (748 mg, 4.50 mmol, 1.0 equiv), hexafluorobenzene (846 mg, 523 μL, 4.50 mmol, 1.0 equiv), potassium carbonate (746 mg, 5.40 mmol, 1.2 equiv), and DMF (11.3 mL), and stirring was commenced. The resultant mixture was stirred at rt for 4.5 h, then sat. aq. NH4Cl (100 mL) was added and the mixture was extracted with EtOAc (3 × 30 mL). The combined organic extracts were washed with brine (4 × 50 mL), then dried (MgSO4), filtered, and concentrated in vacuo (50 °C, ca. 5 mmHg) to give a yellow oil (1.76 g). Purification via flash column chromatography (50 g SiO2, 30 mm Ø, 98:2, hexane/EtOAc, ca. 5-mL fractions) gave 3c as a yellow oil which solidified on standing (931 mg, 62%). To obtain an analytical sample, a 433 mg portion of the above material was dissolved in hexane (3.0 mL) in a 20-mL scintillation vial, sealed with a screw top cap and left in the freezer at –20 °C overnight to give white, needle-like crystals. The crystals were collected via filtration through filter paper in a Hirsch funnel under House vacuum, and were then crushed with a glass rod and dried in vacuo (0.05 mmHg) to give a white, crystalline solid (297 mg, 69% mass return). Data for 7: mp 64-65 °C (hexane); 1H NMR (500 MHz, CDCl3) 7.33-7.27 (m, 2 H), 7.24-7.17 (m, 3 H), 3.44-3.35 (m, 1 H), 2.87-2.76 (m, 2 H), 1.97-1.81 (m, 2 H), 1.32 (d, J = 6.7 Hz, 3 H); 13C NMR (125 MHz, CDCl3) 148.7-146.4 (m100) 141.4, 128.7, 128.6, 126.3, 114.3 (m), 44.0, 39.0, 33.2, 21.7; IR (CHCl3 mull) 3086 (w), 3063 (w), 3027 (m), 2962 (m), 2925 (m), 2861 (w), 1603 (w), 1496 (m), 1458 (s), 1377 (m), 1245 (m), 1030 (m), 955 (s), 813 (m), 747 (m), 698 (m); MS (EI+, 70 eV) 332.1 (M+, 7), 199.0 (15), 117.1 (10), 91.1 (C7H7+, 100), 65.1 (15); HRMS (EI+, double focusing sector field) calcd for C16H13SF5: 332.0658, found: 332.0662; TLC Rf 0.20 (99:1, hexane/EtOAc) [KMnO4].</p><p> </p><!><p>A 500-mL, one-necked, round-bottomed flask equipped with a stirrer bar, water-jacketed reflux condenser, and argon inlet was charged with 29 (8.52 g, 40.0 mmol, 1.0 equiv), 2-mercaptopyridine (4.54 g, 40.0 mmol, 1.0 equiv), potassium carbonate (11.06 g, 80.0 mmol, 2.0 equiv), and acetone (200 mL), and stirring was commenced. The resultant mixture was heated at reflux for 4 h, and was then allowed to cool to rt. The mixture was filtered through a 40 mm Ø, porosity 3, sintered funnel under House vacuum and concentrated in vacuo (50 °C, ca. 5 mmHg) to give a yellow oil (9.97 g). Purification via short-path distillation at reduced pressure (1.2 mmHg) gave 3d as a clear, yellow oil (9.19 g, 94%). To obtain an analytical sample, a 1.75 g portion of the above material was purified on an automated flash column chromatography platform [40 g SiO2 cartridge, hexane (1 CV) then 100:0→70:30, hexane:CH2Cl2 (10 CV) then 70:30, hexane:CH2Cl2 (10 CV), 40 mL min–1 flow rate, 8 mL fractions] to give 3d as a clear, colorless oil (1.67 g, 95% mass return). Data for 3d: bp 127-130 °C (1.2 mmHg); 1H NMR (500 MHz, CDCl3) 8.45-8.42 (m, 1 H), 7.46 (td, J = 7.7, 1.9 Hz, 1 H), 7.32-7.27 (m, 2 H), 7.24-7.14 (m, 4 H), 6.97 (ddd, J = 7.3, 4.9, 1.0 Hz, 1 H), 4.04-3.94 (m, 1 H), 2.88-2.75 (m, 2 H), 2.12-1.91 (m, 2 H), 1.48 (d, J = 6.8 Hz, 3 H); 13C NMR (125 MHz, CDCl3) 159.6, 149.7, 142.2, 136.1, 128.7, 128.6, 126.1, 123.1, 119.6, 39.7, 38.8, 33.6, 21.7; IR (neat) 3084 (w), 3063 (w), 3043 (w), 3026 (w), 2994 (w), 2957 (m), 2922 (m), 2858 (w), 1602 (w), 1578 (s), 1556 (m), 1495 (m), 1453 (s), 1413 (s), 1373 (w), 1353 (w), 1280 (w), 1242 (w), 1178 (w), 1147 (w), 1125 (s), 1089 (w), 1043 (w), 1030 (w), 984 (w), 958 (w), 912 (w), 757 (s), 724 (m), 699 (s), 619 (w); MS (EI+, 70 eV) 243.1 (M+, 16), 182.1 (67), 152.1 (49), 117.1 (28), 111.0 (53), 106.1 (18), 91.1 (C7H7+, 100), 78.0 (37), 77.0 (11), 67.1 (23), 65.1 (29), 51.0 (24); HRMS (EI+, double focusing sector field) calcd for C15H17SN: 243.1082, found: 243.1077; TLC Rf 0.34 (90:10, hexane/EtOAc) [KMnO4]; GC tR 4.75 min.</p><p> </p><!><p>This preparation is based on a previously reported method for the palladium-catalyzed cross-coupling of alkyl thiols with aryl halides.101 1-Bromo-3,5-bis(trifluoromethyl)benzene (299 mg, 176 μL, 1.00 mmol, 1.0 equiv), i-Pr2NEt (142 mg, 192 μL, 1.10 mmol, 1.1 equiv) and 7 (166 mg, 170 μL, 1.00 mmol, 1.0 equiv) were added sequentially to a stirred solution of Pd2(dba)3 (9.2 mg, 0.01 mmol, 1 mol %) and dppf (11.1 mg, 0.02 mmol, 2 mol %) in toluene (1.0 mL) in an oven-dried, one-piece, 5-mL, round-bottomed flask/water-jacketed reflux condenser equipped with a stirrer bar, rubber septum, and argon inlet. The resultant orange solution was heated to reflux for 3 h, and was then allowed to cool to rt. Brine (5 mL) was added and the mixture was extracted with EtOAc (3 × 10 mL) then dried (MgSO4), filtered, and concentrated in vacuo (50 °C, ca. 5 mmHg) to give a brown oil (410 mg). Purification via flash column chromatography (10 g SiO2, 20 mm Ø, 98:2, hexane/EtOAc, ca. 5-mL fractions) gave 3e as a clear, pale yellow oil (353 mg, 93%). Data for 3e: 1H NMR (500 MHz, CDCl3) 7.73-7.65 (m, 3 H), 7.34-7.27 (m, 2 H), 7.25-7.14 (m, 3 H), 3.37-3.29 (m, 1 H), 2.88-2.75 (m, 2 H), 2.03-1.85 (m, 2 H), 1.39 (d, J = 6.7 Hz, 3 H); 13C NMR (125 MHz, CDCl3) 141.1, 140.0, 132.3 (q, J = 32.9 Hz), 128.8, 128.6, 126.4, 123.3 (q, J = 273 Hz), 42.5, 38.2, 33.3, 21.1; IR (neat) 3087 (m), 3065 (m), 3028 (m), 2967 (m), 2927 (m), 2862 (m), 1602 (m), 1497 (m), 1455 (m), 1378 (m), 1352 (s), 1277 (s), 1182 (s), 1135 (s), 1030 (w), 881 (m), 843 (m), 825 (m), 747 (m), 713 (m), 699 (m), 681 (m); MS (EI+, 70 eV) 378.1 (M+, 49), 132.1 (38), 117.1 (31), 91.1 (C7H7+, 100); HRMS (EI+, TOF) calcd for C18H16SF6: 378.0877, found: 378.0873; TLC Rf 0.49 (99:1, hexane/EtOAc) [KMnO4].</p><p> </p><!><p>This preparation is based on a previously reported method for the palladium-catalyzed cross-coupling of alkyl thiols with aryl halides.101 3-Bromopyridine (160 mg, 99 μL, 1.00 mmol, 1.0 equiv), i-Pr2NEt (142 mg, 192 μL, 1.10 mmol, 1.1 equiv) and 7 (166 mg, 170 μL, 1.00 mmol, 1.0 equiv) were added sequentially to a stirred solution of Pd2(dba)3 (9.2 mg, 0.01 mmol, 1 mol %) and dppf (11.1 mg, 0.02 mmol, 2 mol %) in toluene (1.0 mL) in an oven-dried, one-piece, 5-mL, round-bottomed flask/water-jacketed reflux condenser equipped with a stirrer bar, rubber septum, and argon inlet. The resultant orange solution was heated to reflux for 3 h, and was then allowed to cool to rt. Brine (5 mL) was added and the mixture was extracted with EtOAc (3 × 10 mL) then dried (MgSO4), filtered, and concentrated in vacuo (50 °C, ca. 5 mmHg) to give a brown oil (0.28 g). Purification via flash column chromatography (10 g SiO2, 20 mm Ø, 80:20, hexane/EtOAc, ca. 5-mL fractions followed by 20 g SiO2, 20 mm Ø, 90:10, hexane/EtOAc, ca. 5-mL fractions) gave 3f as a clear, orange-yellow oil (227 mg, 93%). Data for 3f: 1H NMR (500 MHz, CDCl3) 8.73 (br s, 1 H), 8.58 (br s, 1 H), 7.67 (d, J = 7.9 Hz, 1 H), 7.36-7.15 (m, 6 H), 3.24-3.15 (m, 1 H), 2.88-2.75 (m, 2 H), 1.99-1.80 (m, 2 H), 1.34 (d, J = 6.7 Hz, 3 H); 13C NMR (125 MHz, CDCl3) 152.9, 148.0, 141.5, 140.0, 132.7, 128.7, 128.6, 126.3, 124.2, 43.1, 38.4, 33.3, 21.5; IR (neat) 3060 (m), 3026 (m), 2960 (m), 2922 (m), 2858 (m), 1602 (m), 1559 (m), 1495 (m), 1457 (m), 1402 (m), 1375 (m), 1317 (m), 1109 (m), 1018 (m), 799 (m), 749 (m), 700 (m); MS (EI+, 70 eV) 243.1 (M+, 46), 132.1 (35), 117.1 (31), 91.1 (C7H7+, 100); HRMS (EI+, TOF) calcd for C15H17NS: 243.1082, found: 243.1081; TLC Rf 0.31 (80:20, hexane/EtOAc) [KMnO4]; GC tR 4.92 min.</p><p> </p><!><p>This preparation is based on a previously reported method for the palladium-catalyzed cross-coupling of alkyl thiols with aryl halides.101i-Pr2NEt (142 mg, 192 μL, 1.10 mmol, 1.1 equiv) and 7 (166 mg, 170 μL, 1.00 mmol, 1.0 equiv) were added sequentially to a stirred solution of 3-bromopyrimidine (164 mg, 99 μL, 1.00 mmol, 1.0 equiv), Pd2(dba)3 (9.2 mg, 0.01 mmol, 1 mol %) and dppf (11.1 mg, 0.02 mmol, 2 mol %) in toluene (1.0 mL) in an oven-dried, one-piece, 5-mL, round-bottomed flask/water-jacketed reflux condenser equipped with a stirrer bar, rubber septum, and argon inlet. The resultant orange solution was heated to reflux for 3 h, and was then allowed to cool to rt. Brine (5 mL) was added and the mixture was extracted with EtOAc (3 × 10 mL) then dried (MgSO4), filtered, and concentrated in vacuo (50 °C, ca. 5 mmHg) to give a brown oil (0.44 g). Purification via flash column chromatography (20 g SiO2, 20 mm Ø, 90:10, hexane/EtOAc, ca. 5-mL fractions) gave 3g as a clear, orange oil (193 mg, 79%). Data for 3g: 1H NMR (500 MHz, CDCl3) 8.50 (d, J = 4.8 Hz, 2 H), 7.33-7.25 (m, 2 H), 7.25-7.16 (m, 3 H), 6.94 (t, J = 4.8 Hz, 1 H), 3.98-3.89 (m, 1 H), 2.89-2.77 (m, 2 H), 2.14-1.94 (m, 2 H), 1.50 (d, J = 6.9 Hz, 3 H); 13C NMR (125 MHz, CDCl3) 173.0, 157.4, 142.0, 128.7, 128.6, 126.1, 116.5, 40.4, 38.4, 33.6, 21.5; IR (neat) 3060 (m), 3026 (m), 2959 (m), 2923 (m), 2858 (m), 1602 (m), 1565 (s), 1546 (s), 1495 (m), 1454 (m), 1381 (s), 1254 (m), 1191 (s), 1030 (m), 980 (m), 798 (m), 773 (s), 748 (s), 699 (s), 628 (m); MS (EI+, 70 eV) 244.1 (M+, 22), 183.1 (100), 153.0 (21), 140.0 (12), 132.1 (14), 117.1 (43), 113.0 (29), 107.1 (12), 91.1 (C7H7+, 61); HRMS (EI+, TOF) calcd for C14H16N2S: 244.1034, found: 244.1036; TLC Rf 0.45 (80:20, hexane/EtOAc) [KMnO4].</p><p> </p><!><p>50-mL, onenecked, round-bottomed flask equipped with a stirrer bar, water-jacketed reflux condenser, and argon inlet was charged with 29 (852 mg, 4.00 mmol, 1.0 equiv), 2-mercaptobenzoxazole (637 mg, 4.00 mmol, 1.0 equiv), potassium carbonate (111 mg, 8.00 mmol, 2.0 equiv), and acetone (20.0 mL), and stirring was commenced. The resultant mixture was heated at reflux for 3 h, and was then allowed to cool to rt. The mixture was filtered through a 40 mm Ø, porosity 3, sintered funnel under House vacuum and concentrated in vacuo (50 °C, ca. 5 mmHg) to give an orange solid (1.15 g). Purification via flash column chromatography (30 g SiO2, 30 mm Ø, 97:3, hexane/EtOAc, ca. 10 mL fractions) gave 3h as a clear, colorless oil (696 mg, 70%). To obtain an analytical sample, a 677 mg portion of the above material was purified on an automated flash column chromatography platform [40 g SiO2 cartridge, hexane (1 CV) then 100:0→70:30, hexane:CH2Cl2 (10 CV) then 70:30, hexane:CH2Cl2 (10 CV), 40 mL min–1 flow rate, 8 mL fractions] to give 3h as a clear, colorless oil (643 mg, 95% mass return). Data for 3h: 1H NMR (500 MHz, CDCl3) 7.64 (d, J = 7.7 Hz, 1 H), 7.46 (d, J = 7.9 Hz, 1 H), 7.35-7.20 (m, 7 H), 4.03-3.94 (m, 1 H), 2.93-2.80 (m, 2 H), 2.23-2.03 (m, 2 H), 1.62 (d, J = 6.8 Hz, 3 H); 13C NMR (125 MHz, CDCl3) 164.7, 151.9, 142.3, 141.4, 128.7, 128.7, 126.3, 124.5, 124.1, 118.7, 110.1, 43.2, 38.6, 33.5, 22.1; IR (neat) 3084 (w), 3062 (w), 3026 (m), 2964 (w), 2925 (m), 2859 (w), 1602 (w), 1497 (s), 1472 (m), 1453 (s), 1376 (w), 1354 (w), 1339 (w), 1282 (w), 1238 (s), 1213 (s), 1180 (w), 1129 (s), 1094 (s), 1030 (w), 1002 (w), 924 (w), 807 (m), 743 (s), 699 (s), 623 (w); MS (EI+, 70 eV) 283.1 (M+, 24), 264.0 (23), 222.1 (68), 219.0 (55), 151.0 (40), 131.0 (31), 122.0 (37), 117.1 (17), 91.1 (C7H7+, 100), 69.0 (38), 65.0 (10); HRMS (EI+, TOF) calcd for C17H17NOS: 283.1031, found: 283.1031; TLC Rf 0.32 (95:5, hexane/EtOAc) [KMnO4]; GC tR 6.12 min.</p><p> </p><!><p>A 50-mL, one-necked, round-bottomed flask equipped with a stirrer bar, water-jacketed reflux condenser, and argon inlet was charged with 29 (852 mg, 4.00 mmol, 1.0 equiv), 1-phenyl-1H-tetrazole-5-thiol (727 mg, 4.00 mmol, 1.0 equiv), potassium carbonate (111 mg, 8.00 mmol, 2.0 equiv), and acetone (20.0 mL), and stirring was commenced. The resultant mixture was heated at reflux for 27 h, and was then allowed to cool to rt. The mixture was filtered through a 40 mm Ø, porosity 3, sintered funnel under House vacuum and concentrated in vacuo (50 °C, ca. 5 mmHg) to give a cloudy, white oil (1.44 g). Purification via flash column chromatography (30 g SiO2, 30 mm Ø, 90:10, hexane/EtOAc, ca. 5-mL fractions) gave 3i as a clear, colorless oil (1.14 g, 92%). Data for 3i: 1H NMR (500 MHz, CDCl3) 7.62-7.53 (m, 5 H), 7.33-7.27 (m, 2 H), 7.24-7.18 (m, 3 H), 4.14-4.05 (m, 1 H), 2.87-2.75 (m, 2 H), 2.23-2.01 (m, 2 H), 1.59 (d, J = 6.8 Hz, 3 H); 13C NMR (125 MHz, CDCl3) 154.1, 141.1, 134.0, 130.4, 130.0, 128.8, 128.6, 126.4, 124.3, 44.6, 38.5, 33.4, 21.7; IR (neat) 3061 (m), 3026 (m), 2965 (m), 2926 (m), 2859 (m), 1597 (s), 1498 (s), 1454 (s), 1386 (s), 1316 (m), 1277 (m), 1238 (s), 1177 (m), 1159 (m), 1088 (s), 1074 (s), 1056 (m), 1030 (m), 1014 (s), 979 (m), 914 (m), 761 (s), 698 (s); MS (EI+, 70 eV) 310.1 (M+, 25), 249.1 (100), 132.1 (59), 117.1 (80), 91.1 (C7H7+, 98), 77.0 (26), 65.0 (23); HRMS (EI+, TOF) calcd for C17H18N4S: 310.1252, found: 310.1252; TLC Rf 0.28 (90:10, hexane/EtOAc) [KMnO4].</p><p> </p><!><p>Error! Bookmark not defined. (3j). This preparation is based on a previously reported method for the palladium-catalyzed cross-coupling of alkyl thiols with aryl halides.101i-Pr2NEt (142 mg, 192 μL, 1.10 mmol, 1.1 equiv) and 7 (166 mg, 170 μL, 1.00 mmol, 1.0 equiv) were added sequentially to a stirred solution of 8-(trifluoromethanesulfonyloxy)quinoline102 (277 mg, 1.00 mmol, 1.0 equiv), Pd2(dba)3 (9.2 mg, 0.01 mmol, 1 mol %) and dppf (11.1 mg, 0.02 mmol, 2 mol %) in toluene (1.0 mL) in an oven-dried, one-piece, 5-mL, round-bottomed flask/water-jacketed reflux condenser equipped with a stirrer bar, rubber septum, and argon inlet. The resultant orange solution was heated to reflux for 3 h, and was then allowed to cool to rt. Brine (5-mL) was added and the mixture was extracted with EtOAc (3 × 10 mL) then dried (MgSO4), filtered, and concentrated in vacuo (50 °C, ca. 5 mmHg) to give a brown syrup (450 mg). Purification via flash column chromatography (15 g SiO2, 20 mm Ø, 88:12, hexane/EtOAc, ca. 5-mL fractions) gave 3j as a clear, yellow-green syrup (268 mg, 91%). To obtain an analytical sample, a 188 mg portion of the above material was purified on an automated flash column chromatography platform [24 g SiO2 cartridge, hexane (1 CV) then 100:0→80:20, hexane:CH2Cl2 (18 CV), 35-mL min –1 flow rate, 8 mL fractions] to give 3j as a clear, pale yellow syrup (117 mg, 62% mass return), in addition to a portion of 3j slightly contaminated with a yellow-colored impurity (52.7 mg). Data for 3j: 1H NMR (500 MHz, CDCl3) 8.98 (dd, J = 4.2, 1.7 Hz, 1 H), 8.14 (dd, J = 8.3, 1.7 Hz, 1 H), 7.59 (dd, J = 7.7, 1.6 Hz, 1 H), 7.47-7.38 (m, 3 H), 7.34-7.28 (m, 2 H), 7.26-7.21 (m, 3 H), 3.64-3.55 (m, 1 H), 2.98-2.84 (m, 2 H), 2.22-2.11 (m, 1 H), 2.07-1.96 (m, 1 H), 1.52 (d, J = 6.7 Hz, 3 H); 13C NMR (125 MHz, CDCl3) 149.5, 146.3, 141.9, 137.9, 136.7, 128.8, 128.7, 128.6, 126.7, 126.2, 125.9, 124.5, 121.8, 38.8, 38.5, 33.5, 21.0; IR (neat) 3081 (w), 3059 (m), 3024 (m), 2959 (m), 2923 (s), 2859 (m), 1603 (m), 1593 (m), 1556 (m), 1490 (s), 1455 (s), 1419 (w), 1374 (m), 1359 (m), 1301 (m), 1214 (m), 1178 (w), 1129 (w), 1068 (w), 1029 (w), 984 (s), 914 (w), 820 (s), 789 (s), 749 (s), 700 (s), 657 (s), 571 (w); MS (EI+, 70 eV) 293.1 (M+, 14), 260.1 (18), 232.1 (33), 202.1 (100), 189.1 (50), 161.0 (46), 156.1 (14), 129.1 (16), 116.1 (18), 91.1 (C7H7+, 29); HRMS (EI+, TOF) calcd for C19H19NS: 293.1238, found: 293.1237; TLC Rf 0.18 (90:10, hexane/EtOAc) [KMnO4].</p><p> </p><!><p>This preparation is based on a previously reported method for the palladium-catalyzed cross-coupling of alkyl thiols with aryl halides.101i-Pr2NEt (142 mg, 192 μL, 1.10 mmol, 1.1 equiv) and 7 (166 mg, 170 μL, 1.00 mmol, 1.0 equiv) were added sequentially to a stirred solution of 10-bromobenzo[h]quinoline103 (258 mg, 1.00 mmol, 1.0 equiv), Pd2(dba)3 (9.2 mg, 0.01 mmol, 1 mol %) and dppf (11.1 mg, 0.02 mmol, 2 mol %) in toluene (1.0 mL) in an oven-dried, one-piece, 5-mL, round-bottomed flask/water-jacketed reflux condenser equipped with a stirrer bar, rubber septum, and argon inlet. The resultant orange solution was heated to reflux for 3 h, and was then allowed to cool to rt. Brine (5-mL) was added and the mixture was extracted with EtOAc (3 × 10 mL) then dried (MgSO4), filtered, and concentrated in vacuo (50 °C, ca. 5 mmHg) to give a brown syrup (0.40 g). Purification via flash column chromatography (10 g SiO2, 20 mm Ø, 90:10, hexane/EtOAc, ca. 5-mL fractions followed by 20 g SiO2, 20 mm Ø, 95:5, hexane/EtOAc, ca. 5-mL fractions) gave 3k as a cloudy, colorless syrup (283 mg, 82%). To obtain an analytical sample, a 207 mg portion of the above material was purified via flash column chromatography (20 g SiO2, 20 mm Ø, 60:40, hexane/toluene, ca. 5-mL fractions) to give 3k as a cloudy, colorless syrup (173 mg, 83% mass return). Data for 3k: 1H NMR (500 MHz, CDCl3) 9.15 (dd, J = 4.4, 1.8 Hz, 1 H), 8.18 (dd, J = 8.0, 1.8 Hz, 1 H), 7.80 (d, J = 8.8 Hz, 1 H), 7.69 (d, J = 8.8 Hz, 1 H), 7.68 (d, J = 7.6 Hz, 1 H), 7.57-7.50 (m, 2 H), 7.45 (d, J = 7.8 Hz, 1 H), 7.37-7.32 (m, 2 H), 7.30-7.23 (m, 3 H), 3.56-3.47 (m, 1 H), 3.02-2.85 (m, 2 H), 2.34-2.24 (m, 1 H), 2.05-1.95 (m, 1 H), 1.58 (d, J = 6.6 Hz, 3 H); 13C NMR (125 MHz, CDCl3) 147.8, 146.6, 142.0, 138.8, 135.7, 135.3, 128.9, 128.9, 128.7, 128.7, 127.5, 127.4, 126.2, 125.8, 124.3, 124.2, 121.0, 39.3, 38.0, 33.8, 20.3; IR (neat) 3084 (w), 3060 (w), 3042 (w), 3023 (w), 2957 (m), 2922 (m), 2860 (w), 1621 (w), 1601 (w), 1583 (m), 1556 (s), 1494 (m), 1454 (m), 1438 (m), 1412 (m), 1394 (m), 1374 (w), 1327 (w), 1289 (w), 1190 (w), 1151 (w), 1140 (w), 1105 (w), 1052 (w), 1052 (w), 1030 (w), 1013 (w), 928 (m), 910 (w), 886 (w), 833 (s), 821 (m), 757 (m), 719 (s), 699 (m), 647 (m); MS (EI+, 70 eV) 343.1 (M+, 8), 252.1 (14), 210.0 (C13H8NS+, 100), 166.1 (12); HRMS (EI+, TOF) calcd for C23H21NS: 343.1395, found: 343.1397; TLC Rf 0.19 (95:5, hexane/EtOAc) [KMnO4].</p><p> </p><!><p>A 50-mL, one-necked, round-bottomed flask equipped with a stirrer bar, water-jacketed reflux condenser, and argon inlet was charged with 29 (852 mg, 4.00 mmol, 1.0 equiv), 2-aminothiophenol (506 mg, 0.43 mL, 4.00 mmol, 1.0 equiv), potassium carbonate (111 mg, 8.00 mmol, 2.0 equiv), and acetone (20.0 mL), and stirring was commenced. The resultant mixture was heated at reflux for 3 h, and was then allowed to cool to rt. The mixture was filtered through a pad of Celite (5 g) in a 40 mm Ø, porosity 3, sintered funnel under House vacuum using minimal EtOAc and then concentrated in vacuo (50 °C, ca. 5 mmHg) to give an orange oil (1.20 g). Purification via flash column chromatography (25 g SiO2, 20 mm Ø, 90:10, hexane/EtOAc, ca. 9 mL fractions) gave 50 as a clear, yellow-orange oil (882 mg, 86%). Data for 50: 1H NMR (500 MHz, CDCl3) 7.39-7.34 (m, 1 H), 7.32-7.25 (m, 2 H), 7.23-7.17 (m, 4 H), 7.16-7.11 (m, 2 H), 4.36 (br s, 2 H), 3.13-3.03 (m, 1 H), 2.87-2.72 (m, 2 H), 2.01-1.77 (m, 2 H), 1.27 (d, J = 6.6 Hz, 3 H); 13C NMR (125 MHz, CDCl3) 149.3, 142.0, 137.5, 130.2, 128.7, 128.7, 126.1, 118.5, 117.1, 115.1, 43.3, 38.6, 33.5, 21.6; IR (neat) 3465 (m), 3365 (m), 3061 (m), 3024 (m), 2921 (m), 2858 (m), 1604 (s), 1495 (m), 1477 (s), 1446 (s), 1373 (m), 1307 (s), 1250 (m), 1157 (m), 1140 (m), 1028 (m), 748 (s), 699 (s); MS (EI+, 70 eV) 257.1 (M+, 97), 125.0 (100), 91.1 (C7H7+, 91), 80.1 (17), 65.0 (12); HRMS (EI+, TOF) calcd for C16H19NS: 257.1238, found: 257.1238; TLC Rf 0.37 (90:10, hexane/EtOAc) [KMnO4].</p><p> </p><!><p>A 25-mL, one-necked, round-bottomed flask equipped with a stirrer bar, rubber septum, and argon inlet was charged with 50 (257 mg, 1.00 mmol, 1.0 equiv), glacial acetic acid (0.29 mL, 5.00 mmol, 5.0 equiv), formaldehyde (37% in H2O, 0.30 mL, 4.00 mmol, 4.0 equiv), and CH2Cl2 (4.0 mL), and stirring was commenced. After 5 min, sodium triacetoxyborohydride (1.12 g, 5.00 mmol, 5.0 equiv) was added portionwise, and the resultant mixture was stirred at rt for 3.5 h. CH2Cl2 (6 mL) and sat. aq. NaHCO3 (10 mL) were then added and the layers were separated. The organic layer was washed with sat. aq. NaHCO3 (2 × 10 mL) and the combined aqueous layers were extracted with CH2Cl2 (2 × 10 mL). The combined organic extracts were washed with brine (20 mL) then dried (MgSO4), filtered, and concentrated in vacuo (50 °C, ca. 5 mmHg) to give a yellow oil (0.32 g). Purification via flash column chromatography (10 g SiO2, 20 mm Ø, 97:3, hexane/EtOAc, ca. 5-mL fractions) gave 3l as a clear, pale yellow oil (233 mg, 82%). Data for 3l: 1H NMR (500 MHz, CDCl3) 7.32-7.25 (m, 2 H), 7.23-7.10 (m, 5 H), 7.09-7.04 (m, 1 H), 6.99-6.93 (m, 1 H), 3.44-3.35 (m, 1 H), 2.88-2.73 (m, 2 H) overlapping 2.77 (s, 6 H), 2.06-1.96 (m, 1 H), 1.92-1.82 (m, 1 H), 1.37 (d, J = 6.7 Hz, 3 H); 13C NMR (125 MHz, CDCl3) 153.0, 142.0, 131.5, 129.5, 128.8, 128.6, 126.3, 126.1, 123.6, 119.6, 44.6, 39.7, 38.5, 33.5, 21.1; IR (neat) 3084 (m), 3058 (m), 3025 (m), 2938 (m), 2857 (m), 2825 (m), 2778 (m), 1602 (m), 1581 (m), 1495 (m), 1477 (m), 1453 (m), 1373 (m), 1316 (m), 1266 (m), 1188 (m), 1157 (m), 1123 (m), 1094 (m), 1062 (m), 1044 (m), 944 (m), 758 (m), 731 (m), 699 (m), 675 (m); MS (EI+, 70 eV) 285.2 (M+, 75), 252.2 (32), 240.1 (15), 224.1 (14), 194.1 (36), 179.1 (20), 164.1 (93), 153.1 (100), 136.0 (55), 122.0 (17), 109.0 (20), 91.1 (C7H7+, 74), 77.0 (13), 65.0 (16); HRMS (EI+, TOF) calcd for C18H23NS: 285.1551, found: 285.1549; TLC Rf 0.49 (95:5, hexane/EtOAc) [KMnO4].</p><p> </p><!><p>This preparation is based on a previously reported method for the palladium-catalyzed cross-coupling of alkyl thiols with aryl halides.101 2-Bromobenzaldehyde (83.1 mg, 52 μL, 0.44 mmol, 1.0 equiv) and i-Pr2NEt (62.6 mg, 84 μL, 0.48 mmol, 1.1 equiv) were added sequentially to a stirred solution of 7 (73.2 mg, 0.44 mmol, 1.0 equiv), Pd2(dba)3 (4.0 mg, 4 μmol, 1 mol %) and dppf (4.9 mg, 9 μmol, 2 mol %) in toluene (0.5 mL) in an oven-dried, one-piece, 5-mL, round-bottomed flask/water-jacketed reflux condenser equipped with a stirrer bar, rubber septum, and argon inlet. The resultant orange solution was heated to reflux for 3 h, and was then allowed to cool to rt. Brine (5 mL) was added and the mixture was extracted with EtOAc (3 × 10 mL) then dried (MgSO4), filtered, and concentrated in vacuo (50 °C, ca. 5 mmHg) to give a brown oil (134 mg). Purification via flash column chromatography on an automated flash column chromatography platform [24 g SiO2 cartridge, 100:0→92:8, hexane/EtOAc (10 CV), 35 mL min–1 flow rate, 8 mL fractions] gave 51 as a clear, pale yellow oil (111 mg, 93%). Data for 51: 1H NMR (500 MHz, CDCl3) 10.56 (s, 1 H), 7.89 (dd, J = 7.7, 1.6, 1 H), 7.48 (ddd, J = 8.0, 7.2, 1.6, 1 H), 7.40 (dd, J = 8.0, 1.1, 1 H), 7.37-7.33 (m, 1 H), 7.33-7.28 (m, 2 H), 7.24-7.17 (m, 3 H), 3.33-3.23 (m, 1 H), 2.89-2.75 (m, 2 H), 2.07-1.97 (m, 1 H), 1.95-1.86 (m, 1 H), 1.37 (d, J = 6.7 Hz, 3 H); 13C NMR (125 MHz, CDCl3) 192.1, 141.4, 140.5, 136.1, 134.1, 132.4, 130.6, 128.7, 128.7, 129.5, 126.3, 43.1, 38.4, 33.4, 21.1; IR (neat) 3084 (w), 3061 (w), 3025 (m), 2960 (m), 2924 (m), 2857 (m), 2735 (w), 1690 (s), 1602 (w), 1586 (m), 1557 (w), 1495 (m), 1456 (m), 1397 (w), 1376 (m), 1287 (w), 1260 (m), 1194 (m), 1127 (w), 1113 (w), 1060 (w), 1030 (w), 843 (w), 824 (m), 750 (m), 699 (m), 634 (w); MS (EI+, 70 eV) 270.1 (M+, 20), 252.1 (40), 210.1 (11), 148.0 (15), 137.0 (59), 109.0 (53), 91.1 (C7H7+, 100), 65.0 (20); HRMS (EI+, TOF) calcd for C17H18OS: 270.1078, found: 270.1072; TLC Rf 0.26 (95:5, hexane/EtOAc) [KMnO4].</p><p> </p><!><p>A 25-mL, one-necked, round-bottomed flask equipped with a stirrer bar, rubber septum, and argon inlet was charged with 51 (216 mg, 0.80 mmol, 1.0 equiv), glacial acetic acid (ca. 2 drops), dimethylamine (2.0 M in THF, 640 μL, 1.28 mmol, 1.6 equiv), and 1,2-dichloroethane (2.7 mL), and stirring was commenced. Sodium triacetoxyborohydride (268 mg, 1.20 mmol, 1.3 equiv) was added portion wise, and the resultant mixture was stirred at rt for 24 h. CH2Cl2 (6 mL) and sat. aq. NaHCO3 (10 mL) were then added and the layers were separated. The aqueous layer was extracted with CH2Cl2 (2 × 10 mL) and the combined organic extracts were dried (MgSO4), filtered, and concentrated in vacuo (50 °C, ca. 5 mmHg) to give a yellow oil (0.33 g). Purification via flash column chromatography (10 g SiO2, 20 mm Ø, 95:5, hexane/EtOAc to EtOAc, ca. 5-mL fractions) gave impure 3m as a clear, yellow oil (141 mg), and a second flash column chromatography (10 g SiO2, 20 mm Ø, 94:5:1 CH2Cl2/MeOH/Et3N, ca. 5-mL fractions) failed to increase the purity. However, a third flash column chromatography (10 g SiO2, 20 mm Ø, 99:1, CH2Cl2/Et3N, ca. 5-mL fractions) gave 3m as a clear, yellow oil (120 mg, 50%). Data for 3m: 1H NMR (500 MHz, CDCl3) 7.40-7.36 (m, 1 H), 7.33-7.25 (m, 3 H), 7.22-7.15 (m, 5 H), 3.57 (s, 2 H), 3.31-3.22 (m, 1 H), 2.87-2.73 (m, 2 H), 2.27 (s, 6 H), 2.04-1.94 (m, 1 H), 1.90-1.81 (m, 1 H), 1.33 (d, J = 6.7 Hz, 3 H); 13C NMR (125 MHz, CDCl3) 141.9, 140.2, 136.0, 131.4, 130.2, 128.7, 128.6, 127.5, 126.4, 126.1, 62.1, 45.6, 42.5, 38.6, 33.4, 21.2; IR (neat) 3083 (m), 3059 (m), 3025 (m), 2969 (m), 2939 (m), 2923 (m), 2853 (m), 2814 (m), 2768 (m), 1693 (m), 1681 (m), 1603 (m), 1588 (m), 1495 (m), 1463 (m), 1454 (m), 1373 (m), 1359 (m), 1251 (m), 1173 (m), 1148 (m), 1096 (m), 1064 (m), 1029 (m), 846 (m), 747 (m), 698 (m); MS (EI+, 70 eV) 299.2 (M+, 30), 284.1 (11), 238.2 (19), 208.1 (36), 166.1 (C9H12NS+, 100), 152.1 (15), 132.1 (13), 123.0 (16), 91.1 (C7H7+, 53); HRMS (EI+, TOF) calcd for C19H2NS: 299.1708, found: 299.1699; TLC Rf 0.16 (97:2.7:0.3, CH2Cl2/MeOH/aq. NH3) [KMnO4].</p><!><p> </p><!><p>An oven-dried, 20-mL scintillation vial was charged with 3d (243 mg, 1.00 mmol, 1.0 equiv) and Fe(acac)3 (106 mg, 0.30 mmol, 30 mol %) in a glove box then was sealed with a rubber septum and removed from the box. Outside of the glove box, a 25-mL Schlenk flask equipped with a stirrer bar, rubber septum and argon inlet was evacuated, flame-dried, left to cool under vacuum, and flushed three times with argon. The vial containing 3d and Fe(acac)3 was charged with CPME (4.0 mL) then sonicated until homogeneous. The clear red solution was then transferred via cannula to the Schlenk flask, and the residual material was rinsed across with further portions of CPME (6.0 mL). 4-Methoxyphenylmagnesium bromide (2.17 M in Et2O, 1.84 mL, 4.00 mmol, 4.0 equiv) was then added by syringe over ca. 2 min. During addition, the color of the solution changed from red to black, and small clusters of black solid could be seen forming during addition. Visible black deposits were also visible at the top of the solution. After stirring for 18 h at rt, 1 M aq. HCl (10 mL) was added in one portion and the mixture was filtered through a pad of Celite (5 g) in a 40 mm Ø, porosity 3, sintered funnel under House vacuum. EtOAc (2 × 5 mL) was used to rinse any residual material though the Celite pad. The filtrate was transferred to a separatory funnel and the layers were separated. The organic layer was washed with 1 M aq. HCl (2 × 10 mL) and the combined aqueous layers were extracted with EtOAc (2 × 10 mL). The combined organic extracts were then dried (MgSO4), filtered and concentrated in vacuo (50 °C, ca. 5 mmHg) to give a pale green residue comprising mainly a white solid (1.40 g). Purification via flash column chromatography (40 g SiO2, 30 mm Ø, 85:15, hexane/toluene, ca. 5-mL fractions) gave a cloudy, colorless oil (173 mg). Further purification via flash column chromatography (C18-reversed phase silica gel, 20 × 160 mm, MeOH, ca. 2.5-mL fractions, loaded with minimal MeCN for solubility reasons) gave a clear, pale yellow oil (149 mg). Several further purifications via flash column chromatography (C18-reversed phase silica gel, 20 × 160 mm, 98:2 MeOH/H2O, ca. 2.5-mL fractions, loaded with minimal MeCN for solubility reasons) gave 8 as a clear, colorless oil (133 mg, 55%). Data for 8: 1H NMR (500 MHz, CDCl3) 7.31-7.23 (m, 2 H), 7.21-7.10 (m, 5 H), 6.91-6.84 (2 H, m), 3.81 (s, 3 H), 2.75-2.62 (m, 1 H), 2.58-2.44 (m, 2 H), 1.96-1.82 (m, 2 H), 1.26 (d, J = 7.0 Hz, 3 H); 13C NMR (125 MHz, CDCl3) 158.1, 142.9, 139.6, 128.7, 128.5, 128.2, 125.9, 114.0, 55.5, 40.5, 38.9, 34.2, 23.0; IR (neat) 3083 (w), 3061 (w), 3026 (w), 2999 (w), 2955 (w), 2927 (w), 2867 (w), 2856 (w), 2834 (w), 1610 (m), 1583 (w), 1511 (s), 1496 (m), 1454 (m), 1374 (w), 1300 (m), 1246 (s), 1177 (m), 1034 (m), 829 (m), 808 (w), 747 (m), 699 (m); MS (EI+, 70 eV) 240.2 (M+, 43), 135.1 (C9H11O+, 100), 105.1 (14), 91.1 (C7H7+, 39), 77.0 (11); TLC Rf 0.30 (80:20, hexane/toluene) [KMnO4]; Analysis C17H20O (240.34). Calcd: C, 84.96; H, 8.39%. Found: C, 84.81; H, 8.26%.</p><!><p> </p><!><p>A 250-mL, onenecked, round-bottomed flask equipped with a stirrer bar and rubber septum was charged with 3a (6.06 g, 25.0 mmol, 1.0 equiv), ammonium molybdate tetrahydrate (3.09 g, 2.50 mmol, 10 mol %), and MeOH (63 mL), and stirring was commenced. The mixture was cooled to 0 °C in an ice/water bath then hydrogen peroxide (30% in H2O, 11.3 g, 10.2 mL, 100 mmol, 4.0 equiv) was added dropwise via a syringe pump over 1 h (the internal temperature did not exceed 8 °C). The resultant turbid, pale yellow mixture was stirred in the ice/water bath for 30 min, then allowed to warm to rt over 1 h, during which time the yellow color intensified. The mixture was then cooled to 1 °C in an ice/water bath and sat. aq. Na2SO3 (30 mL) was added dropwise via a syringe pump over 30 min (the internal temperature did not exceed 15 °C). Starch-iodide paper was used to confirm that no oxidant remained. EtOAc (100 mL) and H2O (100 mL) were then added and the layers were separated. The aqueous layer was extracted with EtOAc (2 × 50 mL) and the combined organic extracts were dried (MgSO4), filtered, and concentrated in vacuo (50 °C, ca. 5 mmHg) to give a cloudy, colorless syrup (6.88 g). Purification via flash column chromatography on an automated flash column chromatography platform [120 g SiO2 cartridge, hexane (1 CV) then 100:0→60:40, hexane/EtOAc (9 CV), 85 mL min–1 flow rate, 24 mL fractions] gave 9 as a clear, colorless syrup (6.20 g, 90%). The 1H NMR spectroscopic data matched that for alternative preparations.104</p><p> </p><!><p>A 500-mL, one-necked, round-bottomed flask equipped with a stirrer bar and rubber septum was charged with 4-(4-methoxyphenyl)-2-butanone (10.0 g, 55.0 mmol, 1.0 equiv) and MeOH (140 mL), and stirring was commenced. The mixture was cooled to 0 °C in an ice/water bath then sodium borohydride (2.29 g, 60.5 mmol, 1.1 equiv) was added portionwise over ca. 25 min (the internal temperature did not exceed 8 °C). The resultant turbid, colorless mixture was stirred in the ice/water bath for 25 min, then allowed to warm to rt over 15 min. The mixture was then concentrated in vacuo (50 °C, ca. 5 mmHg) and partitioned between EtOAc (50 mL) and H2O (100 mL). The layers were separated and the aqueous layer was extracted with EtOAc (2 × 50 mL), then the combined organic extracts were dried (MgSO4), filtered, and concentrated in vacuo (50 °C, ca. 5 mmHg) to give a cloudy, colorless oil (10.8 g). Purification via short-path distillation at reduced pressure (0.01 mmHg) gave 52 as a clear, colorless oil (9.70 g, 98%). The 1H NMR spectroscopic data matched that for alternative preparations.105 Data for 52: bp: 114-116 °C (0.01 mmHg).</p><p> </p><!><p>Bromine (10.2 g, 3.3 mL, 63.6 mmol, 1.2 equiv) was added in one-portion from a 5-mL measuring cylinder to a stirred suspension of triphenylphosphine (16.85 g, 63.6 mmol, 1.2 equiv) in CH2Cl2 (175 mL) in a 1 L, single-necked, round-bottomed flask equipped with a stirrer bar and cooled in an ice/water bath (open to air). The flask was then sealed with a rubber septum and purged with argon via an inlet needle. After stirring the resultant pale yellow suspension for 15 min, a solution of 52 (9.55 g, 53.0 mmol, 1.0 equiv) and imidazole (4.37 g, 63.6 mmol, 1.2 equiv) in CH2Cl2 (90 mL) was added via cannula over ca. 10 min. The cooling bath was removed and the reaction mixture was allowed to warm to rt over 17 h. The mixture was then filtered through a 40 mm Ø, porosity 3, sintered funnel under House vacuum and carefully concentrated in vacuo to leave a yellow oil residue (i.e. avoiding precipitating the phosphorus-containing residues at this point). A stirrer bar was added to the residue and a wide-neck plastic funnel was added to the neck of the flask, and rapid stirring was commenced. Pentane (265 mL) was quickly added in one portion to precipitate the phosphorus-containing residues as a fine white solid. The mixture was filtered through a 40 mm Ø, porosity 3, sintered funnel under House and concentrated in vacuo (50 °C, ca. 5 mmHg) to give a clear, colorless oil (15.24 g). Purification via short-path distillation at reduced pressure (0.01 mmHg) gave 53 as a clear, colorless oil (11.10 g, 86%). The 1H NMR spectroscopic data matched that for an alternative preparation of the (R)-enantiomer.106 Data for 53: bp: 98-100 °C (0.01 mmHg).</p><p> </p><!><p>A 500-mL, onenecked, round-bottomed flask equipped with a stirrer bar, water-jacketed reflux condenser, and argon inlet was charged with 53 (7.29 g, 30.0 mmol, 1.0 equiv), thiophenol (3.41 g, 3.18 mL, 30.0 mmol, 1.0 equiv), potassium carbonate (8.29 g, 60.0 mmol, 2.0 equiv), and acetone (150 mL), and stirring was commenced. The resultant mixture was heated at reflux for 38 h, and was then allowed to cool to rt. The mixture was filtered through a pad of Celite (5 g) in a 40 mm Ø, porosity 3, sintered funnel under House vacuum using minimal EtOAc and then concentrated in vacuo (50 °C, ca. 5 mmHg) to give a clear, pale yellow oil (8.68 g). Purification via short-path distillation at reduced pressure (bp 157-158 °C, 0.01 mmHg) gave a clear, colorless oil (7.53 g). Further purification via flash column chromatography (200 g SiO2, 70 mm Ø, hexane then 60:40, hexane/toluene, ca. 24 mL fractions) gave 54 as a clear, colorless oil (7.06 g, 86%). To obtain an analytical sample, a 157 mg portion of the above material was purified via bulb-to-bulb distillation at reduced pressure (10–5 mmHg) to give a clear, colorless oil (154 mg, 98% mass return). Data for 54: bp: 150 °C ABT (10–5 mmHg); 1H NMR (500 MHz, CDCl3) 7.41-7.36 (m, 2 H), 7.32-7.27 (m, 2 H), 7.26-7.22 (m, 1 H), 7.11 (d, J = 8.4 Hz, 2 H), 6.85 (d, J = 8.0 Hz, 2 H), 3.81 (s, 3 H), 3.26-3.17 (m, 1 H), 2.82-2.70 (m, 2 H), 1.98-1.88 (m, 1 H), 1.87-1.76 (m, 1 H), 1.34 (dd, J = 6.7, 0.9 Hz, 3 H); 13C NMR (125 MHz, CDCl3) 158.1, 135.4, 133.9, 132.2, 129.6, 129.0, 126.9, 114.0, 55.5, 42.7, 38.7, 32.5, 21.5; IR (neat) 3071 (m), 3057 (m), 3030 (m), 3000 (m), 2955 (s), 2926 (s), 2858 (m), 2833 (m), 1879 (m), 1611 (s), 1583 (s), 1511 (s), 1479 (s), 1438 (s), 1374 (m), 1300 (s), 1246 (s), 1177 (s), 1113 (m), 1091 (m), 1068 (m), 1037 (s), 895 (m), 823 (s), 744 (s), 692 (s); MS (EI+, 70 eV) 272.0 (M+, 35), 162.1 (69), 147.0 (31), 121.0 (C8H9O+, 100), 109.0 (10), 83.9 (11), 77.0 (13); TLC Rf 0.24 (60:40, hexane/toluene) [KMnO4]; Analysis: C17H20OS (272.41). Calcd: C, 74.96; H, 7.40%. Found: C, 75.03; H, 7.49%.</p><p> </p><!><p>A 250-mL, one-necked, round-bottomed flask equipped with a stirrer bar and rubber septum was charged with 54 (6.81 g, 25.0 mmol, 1.0 equiv), ammonium molybdate tetrahydrate (3.09 g, 2.50 mmol, 10 mol %), and MeOH (63 mL), and stirring was commenced. The mixture was cooled to 0 °C in an ice/water bath then hydrogen peroxide (30% in H2O, 11.3 g, 10.2 mL, 100 mmol, 4.0 equiv) was added dropwise via a syringe pump over 1.5 h (the internal temperature did not exceed 4 °C). The resultant turbid, pale yellow mixture was stirred in the ice/water bath for 30 min, then allowed to warm to rt over 1 h, during which time the yellow color intensified. The mixture was then cooled to 1 °C in an ice/water bath and sat. aq. Na2SO3 (35 mL) was added dropwise via a syringe pump over 1 h (the internal temperature did not exceed 12 °C). Starch-iodide paper was used to confirm that no oxidant remained. EtOAc (120 mL) and H2O (120 mL) were then added and the layers were separated. The aqueous layer was extracted with EtOAc (2 × 50 mL) and the combined organic extracts were dried (MgSO4), filtered, and concentrated in vacuo (50 °C, ca. 5 mmHg) to give a cloudy, colorless syrup (7.62 g). Purification via flash column chromatography (200 g SiO2, 70 mm Ø, 80:20, hexane/EtOAc then 70:30, hexane/EtOAc, ca. 24 mL fractions) gave 14 as a clear, colorless oil (7.58 g, 100%). Data for 14: 1H NMR (500 MHz, CDCl3) 7.83 (d, J = 9.1 Hz, 2 H), 7.63 (t, J = 7.4 Hz, 1 H), 7.56-7.50 (m, 2 H), 7.01 (d, J = 8.5, 2 H), 6.80 (d, J = 8.5 Hz, 2 H), 3.76 (s, 3 H), 3.06-2.97 (m, 1 H), 2.79-2.70 (m, 1 H), 2.56-2.47 (m, 1 H), 2.31-2.22 (m, 1 H), 1.73-1.62 (m, 1 H), 1.29 (d, J = 6.9 Hz, 3 H); 13C NMR (125 MHz, CDCl3) 158.3, 137.4, 133.9, 132.3, 129.5, 129.3, 129.2, 114.2, 59.3, 55.5, 31.8, 31.1, 13.5; IR (neat) 3062 (w), 3030 (w), 2953 (m), 2935 (m), 2865 (w), 2835 (w), 1611 (m), 1583 (m), 1513 (s), 1461 (m), 1446 (s), 1420 (w), 1380 (w), 1303 (s), 1246 (s), 1178 (s), 1145 (s), 1085 (s), 1034 (s), 999 (w), 929 (w), 902 (w), 821 (m), 764 (m), 729 (s), 693 (s), 660 (w), 637 (w), 620 (w), 593 (s), 566 (m); MS (ESI) 327.1 ([M+Na]+, 100), 322.1 ([M+NH4]+, 32), 305.1 ([M+H]+, 98), 163.1 (22), 143.0 (17), 121.1 (C8H9O+, 25); HRMS (ESI, TOF) calcd for C17H21O3S: 305.1211, found: 305.1214; TLC Rf 0.48 (70:30, hexane/EtOAc) [KMnO4].</p><p> </p><!><p>A 500-mL, one-necked, round-bottomed flask equipped with a stirrer bar, water-jacketed reflux condenser, and argon inlet was charged with cyclohexyl bromide (8.24 g, 50.0 mmol, 1.0 equiv), thiophenol (5.68 g, 5.29 mL, 50.0 mmol, 1.0 equiv), potassium carbonate (13.82 g, 100.0 mmol, 2.0 equiv), and acetone (250 mL), and stirring was commenced. The resultant mixture was heated at reflux for 24 h, and was then allowed to cool to rt. The mixture was filtered through a pad of SiO2 (5 g) in a 40 mm Ø, porosity 3, sintered funnel under House vacuum using minimal EtOAc and then concentrated in vacuo (50 °C, ca. 5 mmHg) to give a clear, yellow oil (6.02 g). Purification via short-path distillation at reduced pressure (bp 98-100 °C, 0.5 mmHg) gave 55 contaminated with thiophenol (~5%) as a clear, colorless oil (3.33 g). Further purification via flash column chromatography [200 g basic alumina (Brockmann grade 1), 50 mm Ø, hexane, ca. 10 mL fractions] gave 55 still contaminated with thiophenol (~5%) as a clear, colorless oil (3.23 g). The 1H NMR spectroscopic data for 55 matched that for an alternative preparation.107 A 25-mL, one-necked, round-bottomed flask equipped with a stirrer bar and rubber septum was then charged with 55 (577 mg, approx. 3.00 mmol, 1.0 equiv), ammonium molybdate tetrahydrate (371 mg, 0.30 mmol, 10 mol %), and MeOH (8.5 mL), and stirring was commenced. The mixture was cooled in an ice/water bath then hydrogen peroxide (30% in H2O, 1.36 g, 1.23 mL, 12.0 mmol, 4.0 equiv) was added dropwise via syringe over ca. 10 min. The resultant turbid, pale yellow mixture was stirred in the ice/water bath for 40 min, then allowed to warm to rt over 1 h, during which time the yellow color intensified. The mixture was then cooled in an ice/water bath and sat. aq. Na2SO3 (3.5 mL) was added dropwise via syringe over ca. 5 min. Starch-iodide paper was used to confirm that no oxidant remained. EtOAc (10 mL) and H2O (10 mL) were then added and the layers were separated. The aqueous layer was extracted with EtOAc (2 × 10 mL) and the combined organic extracts were dried (MgSO4), filtered, and concentrated in vacuo (50 °C, ca. 5 mmHg) to give a cloudy, colorless syrup (651 mg). Purification via flash column chromatography (40 g SiO2, 30 mm Ø, 75:25, hexane/EtOAc, ca. 5-mL fractions) gave 18a as a clear, colorless syrup which solidified on standing to a white, crystalline solid (611 mg, approx. 91%). The 1H NMR spectroscopic data and melting point matched that for an alternative preparation.108</p><p> </p><!><p>A 200-mL, onenecked, round-bottomed flask equipped with a stirrer bar, rubber septum, and argon inlet was charged with N-benzyl-4-hydroxypiperidine (5.00 g, 26.1 mmol, 1.0 equiv), Et3N (7.94 g, 10.9 mL, 78.4 mmol, 3.0 equiv), and CH2Cl2 (65 mL), and stirring was commenced. The mixture was cooled to 0 °C in an ice/water bath then 4-toluenesulfonyl chloride (5.98 g, 31.4 mmol, 1.2 equiv) was added portionwise over ca. 10 min (the internal temperature did not exceed 1 °C). The resultant mixture was stirred in the ice/water bath for 30 min, then allowed to warm to rt over 21 h. Sat. aq. NaHCO3 (100 mL) was then added and the layers were separated. The organic layer was washed with sat. aq. NaHCO3 (2 × 50 mL) and the combined aqueous layers were extracted with CH2Cl2 (2 × 50 mL). The combined organic extracts were then dried (MgSO4), filtered, and concentrated in vacuo (50 °C, ca. 5 mmHg) to give an orange syrup (10.55 g). Purification via flash column chromatography on an automated flash column chromatography platform [120 g SiO2 cartridge, 100:0→40:60, hexane/EtOAc (1 CV) then i-PrOH, 85 mL min–1 flow rate, 24 mL fractions] gave impure 56 as a clear, orange syrup which solidified on standing to a yellow, crystalline solid (5.28 g), in addition to mixed fractions, returned starting material and other unidentified products. Attempted purification of a 100 mg portion of the impure product via bulb-to-bulb distillation at reduced pressure (10–5 mmHg) led to decomposition to a black tar (ca. 150 °C ABT). The remainder of the impure 56 was purified via recrystallization from 90:10, hexane/toluene (10 mL) in a 20-mL scintillation vial. The crystals were collected via filtration through filter paper in a Hirsch funnel under House vacuum, washed with a minimal amount of cold (–78 °C) hexane and then crushed with a glass rod and dried in vacuo (0.05 mmHg) to give a cream-colored, crystalline solid (2.92 g). This material was subsequently combined with a second crop (997 mg) and third crop (535 mg) (both of comparable purity to the first crop according to 1H NMR spectroscopic analysis) to give 56 as a cream-colored, crystalline solid (4.45 g, 50%). To obtain an analytical sample, a 1.03 g portion of the above material was dissolved in EtOH (4.0 mL) in a 20-mL scintillation vial, sealed with a screw top cap and left in the freezer at –20 °C overnight. The resultant crystals were collected via filtration through filter paper in a Hirsch funnel under House vacuum, washed with cold (0 °C) EtOH (2.0 mL) and then crushed with a glass rod and dried in vacuo (0.05 mmHg) to give 56 as a white, crystalline solid (301 mg, 29% mass return). Data for 56: mp 66-67 °C (EtOH); 1H NMR (500 MHz, CDCl3) 7.78 (d, J = 8.3 Hz, 2 H), 7.35-7.20 (m, 7 H), 4.58-4.47 (br m, 1 H), 3.46 (s, 2 H), 2.69-2.55 (br m, 2 H), 2.43 (s, 3 H), 2.31-2.11 (br m, 2 H), 1.87-1.71 (m, 4 H); 13C NMR (125 MHz, CDCl3) 144.8, 138.3, 134.8, 130.0, 129.2, 128.5, 127.8, 127.4, 79.2, 63.0, 50.3, 31.8, 21.9; IR (CHCl3 mull) 3086 (w), 3063 (w), 3029 (w), 2952 (w), 2811 (w), 2773 (w), 1598 (w), 1494 (w), 1454 (m), 1398 (w), 1357 (s), 1306 (w), 1291 (w), 1267 (w), 1253 (w), 1211 (w), 1188 (m), 1176 (s), 1141 (w), 1132 (w), 1120 (w), 1097 (m), 1072 (w), 1037 (w), 1028 (w), 1019 (w), 998 (m), 945 (m), 910 (m), 870 (m), 845 (m), 814 (m), 796 (w), 741 (m), 699 (m), 680 (m), 670 (m), 617 (w), 573 (m), 555 (m); MS (ESI) 346.1 ([M+H]+, 100), 192.0 (26); TLC Rf 0.50 (50:50, hexane/EtOAc) [KMnO4]; Analysis C19H23NO3S (345.46). Calcd: C, 66.06; H, 6.71; N, 4.05%. Found: C, 65.80; H, 6.70; N, 4.15%.</p><p> </p><!><p>A 25-mL, one-necked, round-bottomed flask equipped with a stirrer bar, air condenser, and argon inlet was charged with 56 (1.04 g, 3.00 mmol, 1.0 equiv), thiophenol (511 mg, 476 μL, 4.50 mmol, 1.5 equiv), and DMF (8.0 mL), and stirring was commenced. Sodium hydride (60% in mineral oil, 240 mg, 6.00 mmol, 2.0 equiv) was added portionwise and the resultant mixture was heated at 70 °C for 16 h, and was then allowed to cool to rt. The mixture was then partitioned between EtOAc (10 mL) and H2O (40 mL) and the layers were separated. The aqueous layer was extracted with EtOAc (2 × 10 mL) and the combined organic extracts were washed with H2O (5 × 40 mL), then dried (MgSO4), filtered, and concentrated in vacuo (50 °C, ca. 5 mmHg) to give a clear, yellow oil (927 mg). Purification via flash column chromatography (40 g SiO2, 30 mm Ø, 85:15, hexane/EtOAc, ca. 5-mL fractions) gave impure 57 as a clear, orange oil which solidified on standing to an orange, crystalline solid (635 mg). Further purification was performed via recrystallization from hexane (4.0 mL) in a 20-mL scintillation vial. The crystals were collected via filtration through filter paper in a Hirsch funnel under House vacuum, washed with cold (–78 °C) hexane (2.0 mL) and then crushed with a glass rod and dried in vacuo (0.05 mmHg) to give a cream-colored, crystalline solid (557 mg). Further purification was performed via recrystallization from MeOH (4.0 mL) in a 20-mL scintillation vial. The crystals were collected via filtration through filter paper in a Hirsch funnel under House vacuum, washed with cold (–78 °C) MeOH (2.0 mL) and then crushed with a glass rod and dried in vacuo (0.05 mmHg) to give a white, crystalline solid (471 mg). This material was subsequently combined with a second crop (63.4 mg) (of comparable purity to the first crop according to 1H NMR spectroscopic analysis) to give 57 as a white, crystalline solid (534 mg, 63%). Data for 57: mp 83-84 °C (MeOH); 1H NMR (500 MHz, CDCl3) 7.46-7.41 (m, 2 H), 7.36-7.23 (m, 8 H), 3.52 (s, 2 H), 3.17-3.04 (br m, 1 H), 2.94-2.79 (m, 2 H), 2.18-2.03 (m, 2 H), 2.02-1.92 (m, 2 H), 1.78-1.64 (m, 2 H); 13C NMR (125 MHz, CDCl3) 138.6, 134.8, 132.5, 129.4, 129.1, 128.4, 127.3, 127.1, 63.4, 53.3, 44.8, 32.8; IR (CHCl3 mull) 3058 (w), 2955 (m), 2920 (m), 2850 (w), 2787 (m), 2750 (m), 2717 (w), 2691 (w), 2671 (w), 1585 (m), 1493 (m), 1480 (s), 1450 (m), 1435 (m), 1388 (w), 1354 (m), 1339 (m), 1302 (m), 1271 (w), 1258 (w), 1222 (w), 1211 (w), 1198 (w), 1185 (w), 1166 (w), 1140 (m), 1130 (m), 1091 (m), 1069 (w), 1026 (m), 996 (m), 973 (w), 902 (w), 890 (w), 802 (m), 769 (w), 731 (s), 698 (s), 689 (m); MS (EI+, 70 eV) 283.1 (M+, 28), 174.1 (74), 91.1 (C7H7+, 100); TLC Rf 0.28 (80:20, hexane/EtOAc) [KMnO4]; Analysis C18H21NS (283.43). Calcd: C, 76.28; H, 7.47; N, 4.94%. Found: C, 75.98; H, 7.59; N, 5.01%.</p><p> </p><!><p>A 50-mL, one-necked, round-bottomed flask equipped with a stirrer bar and rubber septum was charged with 57 (562 mg, 1.98 mmol, 1.0 equiv), MeOH (8.0 mL), and THF (6.0 mL), and stirring was commenced. The mixture was cooled in an ice/water bath then a solution of Oxone [49.5% KHSO5, 1.83 g, 5.95 mmol (of KHSO5), 3.0 equiv (of KHSO5)] was added dropwise via Pasteur pipette over ca. 5 min. The resultant turbid, white mixture was then allowed to warm to rt over 19 h. H2O (10 mL) was added and the mixture was filtered through a 40 mm Ø, porosity 3, sintered funnel under House vacuum, and the solid residue rinsed with EtOAc (2 × 10 mL). The filtrate was transferred to a 250-mL separatory funnel and the layers were separated. The aqueous layer was extracted with EtOAc (2 × 20 mL) and the combined organic extracts were washed sequentially with sat. aq. NaHCO3 (30 mL) and brine (30 mL), then dried (MgSO4), filtered, and concentrated in vacuo (50 °C, ca. 5 mmHg) to give a clear, yellow syrup (495 mg). 2 M aq. NaOH (10 mL) was added to the combined aqueous layers, which were then extracted with EtOAc (3 × 10 mL) and dried (MgSO4), filtered, and concentrated in vacuo (50 °C, ca. 5 mmHg) to give a clear, colorless syrup (186 mg). The two portions of organic material were combined to give a clear, yellow syrup (682 mg). Purification via flash column chromatography (40 g SiO2, 30 mm Ø, 60:40, hexane/EtOAc, ca. 5-mL fractions) gave a clear, colorless syrup (559 mg). Further purification was performed via recrystallization from MeOH (4.0 mL) in a 20-mL scintillation vial. Following brief cooling in an ice/water bath, the crystals were collected via filtration through filter paper in a Hirsch funnel under House vacuum, washed with cold (–78 °C) MeOH (2.0 mL) and then crushed with a glass rod and dried in vacuo (0.05 mmHg) to give 18b as a white, crystalline solid (345 mg, 55%). Data for 18b: mp 87-88 °C (MeOH); 1H NMR (500 MHz, CDCl3) 7.89-7.84 (m, 2 H), 7.68-7.63 (m, 1 H), 7.59-7.52 (m, 2 H), 7.32-7.20 (m, 5 H), 3.46 (s, 2 H), 3.00-2.94 (m, 2 H), 2.90 (tt, J = 12.3, 3.7 Hz, 1 H), 2.01-1.88 (m, 4 H), 1.78-1.67 (m, 2 H); 13C NMR (125 MHz, CDCl3) 138.1, 137.1, 133.9, 129.4, 129.3, 129.2, 128.5, 127.4, 62.9, 62.2, 52.4, 25.7; IR (CHCl3 mull) 3062 (w), 3026 (m), 2953 (m), 2806 (m), 2762 (m), 1585 (w), 1494 (m), 1467 (w), 1447 (s), 1394 (w), 1366 (m), 1342 (m), 1304 (s), 1273 (s), 1233 (m), 1145 (s), 1086 (s), 1028 (m), 994 (m), 932 (w), 909 (w), 877 (w), 813 (m), 752 (s), 721 (s), 690 (s), 667 (m), 646 (w), 617 (s), 602 (s), 565 (s); MS (EI+, 70 eV) 315.1 (M+, 9), 174.1 (58), 120.1 (16), 110.0 (11), 91.1 (C7H7+, 100), 82.1 (22), 77.1 (26), 65.1 (13), 51.0 (20); HRMS (EI+, double focusing sector field) calcd for C18H21NO2S: 315.1293, found: 315.1288; TLC Rf 0.16 (60:40, hexane/EtOAc) [KMnO4].</p><p> </p><!><p>A 50-mL, onenecked, round-bottomed flask equipped with a stirrer bar and rubber septum was charged with (1l,2l,4u)-2-(phenylthio)bicyclo[2.2.1]heptane109 (462 mg, 2.26 mmol, 1.0 equiv), ammonium molybdate tetrahydrate (280 mg, 0.23 mmol, 10 mol %), and MeOH (6.5 mL), and stirring was commenced. The mixture was cooled in an ice/water bath then hydrogen peroxide (30% in H2O, 1.03 g, 0.92 mL, 9.05 mmol, 4.0 equiv) was added dropwise via syringe over ca. 15 min. The resultant turbid, pale yellow mixture was stirred in the ice/water bath for 40 min, then allowed to warm to rt over 1 h, during which time the yellow color intensified. The mixture was then cooled in an ice/water bath and sat. aq. Na2SO3 (3.3 mL) was added dropwise via syringe over ca. 6 min. Starch-iodide paper was used to confirm that no oxidant remained. EtOAc (10 mL) and H2O (10 mL) were then added and the layers were separated. The aqueous layer was extracted with EtOAc (2 × 10 mL) and the combined organic extracts were dried (MgSO4), filtered, and concentrated in vacuo (50 °C, ca. 5 mmHg) to give a cloudy, colorless syrup (568 mg). Purification via bulb-to-bulb distillation at reduced pressure (10–5 mmHg) gave a clear, colorless syrup which solidified on standing to a white, crystalline solid (521 mg). Further purification was performed via recrystallization from MeOH (4.0 mL) in a 20-mL scintillation vial, which was sealed with a screw top cap and left in the freezer at –20 °C for ca. 30 min. The resultant crystals were collected via filtration through filter paper in a Hirsch funnel under House vacuum, washed with cold (–78 °C) MeOH (1.0 mL) and then crushed with a glass rod and dried in vacuo (0.05 mmHg) to give a white, crystalline solid (311 mg). This material was subsequently combined with a second crop (120 mg) (of comparable purity to the first crop according to 1H NMR spectroscopic analysis) to give 18c as a white, crystalline solid (431 mg, 81%, >99:1 dr). Data for 18c: mp 81-82 °C (MeOH); 1H NMR (500 MHz, CDCl3) 7.90-7.84 (m, 2 H), 7.64-7.59 (m, 1 H), 7.57-7.50 (m, 2 H), 2.98 (dd, J = 8.5, 6.2 Hz, 1 H), 2.67-2.63 (br m, 1 H), 2.39-2.34 (br m, 1 H), 2.05-1.97 (m, 1 H), 1.83-1.76 (m, 1 H), 1.61-1.45 (m, 3 H), 1.20-1.09 (m, 3 H); 13C NMR (125 MHz, CDCl3) 139.3, 133.6, 129.4, 128.6, 66.9, 39.0, 36.3, 36.3, 32.8, 30.0, 28.3; IR (CHCl3 mull) 3062 (w), 2962 (m), 2872 (m), 1585 (w), 1478 (w), 1446 (m), 1325 (m), 1303 (s), 1272 (m), 1243 (w), 1205 (w), 1146 (s), 1086 (s), 1071 (w), 1047 (w), 1024 (w), 998 (w), 957 (w), 922 (w), 905 (w), 878 (w), 843 (w), 788 (w), 758 (m), 721 (s), 694 (s), 670 (w), 611 (s), 559 (s); MS (CI) 237.0 ([M+H]+, 31), 171.0 (26), 143.0 (80), 95.0 (C7H11+, 100); TLC Rf 0.35 (80:20, hexane/EtOAc) [KMnO4]; Analysis C13H16O2S (236.33). Calcd: C, 66.07; H, 6.82%. Found: C, 65.79; H, 6.86%.</p><p> </p><!><p>A 25-mL Schlenk flask equipped with a stirrer bar, rubber septum, and argon inlet was evacuated and flame-dried, then left to cool to rt and flushed with argon three times. Magnesium turnings (233 mg, 9.60 mmol, 1.2 equiv) were quickly added against a backflow of argon, followed by iodine (several crystals), and THF (1.0 mL), and stirring was commenced. Meanwhile, an oven-dried, 20-mL scintillation vial was charged with 2-bromoadamantane (1.76 g, 8.00 mmol, 1.0 equiv) and was then sealed with an inverted rubber septum and purged with argon. THF (1.0 mL) was then added via syringe and a small portion of the resultant solution was transferred via cannula to the Schlenk flask containing the magnesium and iodine, in order to initiate the reaction. Once the reaction mixture had decolorized (several minutes), THF (8.0 mL) was added to the vial containing the 2-bromoadamantane and the resultant solution was added dropwise via cannula to the Schlenk flask over ca. 15 min. A water-jacketed reflux condenser was added and the reaction mixture was heated at reflux for 1 h, then allowed to cool to rt. The mixture was cooled in an ice/water bath and a solution of diphenyl disulfide (1.68 g, 7.60 mmol, 0.95 equiv) in THF (5.0 mL) was added dropwise via cannula over ca. 5 min, then the reaction was allowed to warm to rt over 24 h, with stirring. 1 M aq. HCl (10 mL) was added, followed by EtOAc (30 mL) and H2O (30 mL), and the layers were separated. The aqueous layer was extracted with EtOAc (2 × 30 mL) and the combined organic extracts were dried (MgSO4), filtered, and concentrated in vacuo (50 °C, ca. 5 mmHg) to give a suspension of white solid in a clear, yellow oil (1.97 g). Purification via flash column chromatography [100 g basic alumina (Brockmann grade 1), 30 mm Ø, hexane, ca. 10 mL fractions] gave a ~1:1 mixture of 58 and diphenyl disulfide as a clear, yellow oil (680 mg), in addition to fractions containing a mixture of 58, diphenyl disulfide, and thiophenol. The latter fractions were concentrated in vacuo (50 °C, ca. 5 mmHg) and purified via flash column chromatography [100 g basic alumina (Brockmann grade 1), 30 mm Ø, hexane, ca. 10 mL fractions] to give a ~1:1 mixture of 58 and diphenyl disulfide as a clear, yellow oil (193 mg). Attempted purification of the combined material (873 mg) via preparative, radial, centrifugally accelerated, thin-layer chromatography on a Harrison Chromatotron (4 mm SiO2 plate, hexane, ca. 5-mL fractions) proved unsuccessful in removing the diphenyl disulfide contaminant. Thus, the ~1:1 mixture of 58 and diphenyl disulfide (873 mg) was carried forward in the next step. The 1H NMR spectroscopic data for the 58 present in the mixture matched that for a pure sample prepared via an alternative procedure.110 A 100-mL, one-necked, round-bottomed flask equipped with a stirrer bar and rubber septum was then charged with the ~1:1 mixture of 58 and diphenyl disulfide (873 mg), ammonium molybdate tetrahydrate (989 mg, 0.80 mmol), and MeOH (25.0 mL), and stirring was commenced. The mixture was cooled to 0 °C in an ice/water bath then hydrogen peroxide (30% in H2O, 3.63 g, 3.27 mL, 32.0 mmol) was added dropwise via a syringe pump over 1 h (the internal temperature did not exceed 1 °C). The resultant turbid, pale yellow mixture was stirred in the ice/water bath for 30 min, then allowed to warm to rt over 2 h, during which time the yellow color intensified. The mixture was then cooled to 0 °C in an ice/water bath and sat. aq. Na2SO3 (20 mL) was added dropwise via a syringe pump over 1 h (the internal temperature did not exceed 9 °C). Starch-iodide paper was used to confirm that no oxidant remained. The mixture was filtered through a 40 mm Ø, porosity 3, sintered funnel under House vacuum and EtOAc (40 mL) and H2O (40 mL) were then added, and the layers were separated. The aqueous layer was extracted with EtOAc (2 × 20 mL) and the combined organic extracts were dried (MgSO4), filtered, and concentrated in vacuo (50 °C, ca. 5 mmHg) to give almost pure 18d as a white solid (571 mg). The material was purified via recrystallization from MeOH (5.0 mL) in a 20-mL scintillation vial, which was sealed with a screw top cap and left at rt overnight. The resultant crystals were collected via filtration through filter paper in a Hirsch funnel under House vacuum, and washed with cold (0 °C) MeOH (2.0 mL) to give a white, crystalline solid (493 mg). A second recrystallization of this material under the same conditions using MeOH (4.0 mL) gave crystals which were crushed with a glass rod and dried in vacuo (0.05 mmHg) to give 18d as a white, crystalline solid (448 mg, 21% based on diphenyl disulfide as the limiting reagent from the first step). Data for 18d: mp 135-136 °C (MeOH); 1H NMR (500 MHz, CDCl3) 7.90-7.85 (m, 2 H), 7.65-7.59 (m, 1 H), 7.57-7.51 (m, 2 H), 3.15-3.10 (br m, 1 H), 2.63-2.53 (m, 2 H), 2.39-2.32 (br m, 2 H), 1.98-1.81 (m, 4 H), 1.76-1.69 (br m, 2 H), 1.64-1.53 (m, 4 H); 13C NMR (125 MHz, CDCl3) 139.1, 133.6, 129.3, 128.6, 69.4, 39.2, 37.4, 31.5, 28.2, 27.7, 27.0; IR (CHCl3 mull) 2918 (s), 2853 (m), 1583 (w), 1472 (m), 1451 (m), 1409 (w), 1357 (w), 1342 (w), 1321 (m), 1300 (s), 1288 (s), 1240 (m), 1221 (m), 1180 (w), 1166 (w), 1148 (s), 1114 (m), 1101 (m), 1086 (m), 1076 (m), 1036 (w), 1022 (w), 998 (w), 981 (w), 967 (w), 939 (w), 903 (w), 828 (m), 784 (w), 766 (m), 756 (m), 722 (m), 695 (m), 666 (m), 623 (w); MS (CI) 277.3 ([M+H]+, 27), 135.2 (C10H15+, 100); TLC Rf 0.43 (80:20, hexane/EtOAc) [KMnO4]; Analysis C16H20O2S (276.39). Calcd: C, 69.53; H, 7.29%. Found: C, 69.28; H, 7.37%.</p><p> </p><!><p>A 50-mL, one-necked, round-bottomed flask equipped with a stirrer bar and rubber septum was charged with 4-methoxyphenylacetophenone (575 mg, 3.50 mmol, 1.0 equiv) and MeOH (8.8 mL), and stirring was commenced. The mixture was cooled to 0 °C in an ice/water bath then sodium borohydride (146 mg, 3.85 mmol, 1.1 equiv) was added portion wise over ca. 5 min. The resultant turbid, colorless mixture was stirred in the ice/water bath for 10 min, then allowed to warm to rt over 30 min. The mixture was then concentrated in vacuo (50 °C, ca. 5 mmHg) and partitioned between EtOAc (10 mL) and H2O (20 mL). The layers were separated and the aqueous layer was extracted with EtOAc (2 × 10 mL), then the combined organic extracts were dried (MgSO4), filtered, and concentrated in vacuo (50 °C, ca. 5 mmHg) to give a cloudy, colorless oil (624 mg). Purification via bulb-to-bulb distillation at reduced pressure (0.05 mmHg) gave 59 as a clear, colorless oil (568 mg, 98%). The 1H NMR spectroscopic data matched that for alternative preparations.111 Data for 59: bp 140 °C ABT (0.05 mmHg).</p><p> </p><!><p>Bromine (578 mg, 186 μL, 3.60 mmol, 1.2 equiv) was added via syringe to a stirred suspension of triphenylphosphine (954 mg, 3.60 mmol, 1.2 equiv) in CH2Cl2 (10 mL) in a 50-mL, single-necked, round-bottomed flask equipped with a stirrer bar and cooled in an ice/water bath (open to air). The flask was then sealed with a rubber septum and purged with argon via an inlet needle. After stirring the resultant pale yellow suspension for 15 min, a solution of 59 (499 mg, 3.00 mmol, 1.0 equiv) and imidazole (248 mg, 3.60 mmol, 1.2 equiv) in CH2Cl2 (5 mL) was added via cannula over ca. 10 min. The cooling bath was removed and the reaction mixture was allowed to warm to rt over 4 h 20 min. The mixture was then filtered through a 40 mm Ø, porosity 3, sintered funnel under House vacuum and carefully concentrated in vacuo to leave a yellow oil residue (i.e. avoiding precipitating the phosphorus-containing residues at this point). A stirrer bar was added to the residue and a wide-neck plastic funnel was added to the neck of the flask, and rapid stirring was commenced. Pentane (15 mL) was quickly added in one portion to precipitate the phosphorus-containing residues as a fine white solid. The mixture was filtered through a 40 mm Ø, porosity 3, sintered funnel under House vacuum and concentrated in vacuo (50 °C, ca. 5 mmHg) to give a clear, colorless oil (536 mg). Purification via bulb-to-bulb distillation at reduced pressure (160 °C ABT, 0.05 mmHg) gave 60 contaminated with (E)-1-(4-methoxyphenyl)prop-1-ene112 (~5%) as a clear, colorless oil (469 mg). Data for 60: 1H NMR (500 MHz, CDCl3) 7.13 (d, J = 8.5 Hz, 2 H), 6.86 (d, J = 8.5 Hz, 2 H), 4.31-4.23 (m, 1 H), 3.81 (s, 3 H), 3.18 (dd, J = 14.1, 6.9 Hz, 1 H), 3.02 (dd, J = 14.1, 7.2 Hz, 1 H), 1.69 (d, J = 6.6 Hz, 3 H); 13C NMR (125 MHz, CDCl3) 158.7, 130.9, 130.5, 114.0, 55.5, 51.4, 46.9, 25.8; MS (EI+, 70 eV) 230.0 ([81Br]M+, 8), 228.0 ([79Br]M+, 8), 149.1 (C8H9O+, 13), 121.1 (C8H9O+, 100); HRMS (EI+, double focusing sector field) calcd for C10H13O79 Br: 228.0150, found: 228.0147.</p><p> </p><!><p>A 25-mL, one-necked, round-bottomed flask equipped with a stirrer bar, water-jacketed reflux condenser, and argon inlet was charged with 60 (462 mg, approx. 2.00 mmol, 1.0 equiv), thiophenol (227 mg, 212 μL, 1.00 mmol, 1.0 equiv), potassium carbonate (553 mg, 4.00 mmol, 2.0 equiv), and acetone (10.0 mL), and stirring was commenced. The resultant mixture was heated at reflux for 14 h, and was then allowed to cool to rt. The mixture was filtered through a 40 mm Ø, porosity 3, sintered funnel under House vacuum and concentrated in vacuo (50 °C, ca. 5 mmHg) to give a colorless oil containing a 84:16 ratio of product:starting material. Resubjection of this material to the above reaction conditions for a further 15 h gave a clear, yellow oil (761 mg). Purification via flash column chromatography (40 g SiO2, 30 mm Ø, hexane then 70:30, hexane/toluene, ca. 5-mL fractions) gave a clear, colorless oil (443 mg). Further purification via bulb-to-bulb distillation at reduced pressure (0.03 mmHg) gave 61 as a clear, colorless oil (421 mg, approx. 80%). Data for 61: bp 175 °C ABT (0.03 mmHg); 1H NMR (500 MHz, CDCl3) 7.47-7.42 (m, 2 H), 7.36-7.30 (m, 2 H), 7.28-7.23 (m, 1 H), 7.14-7.09 (m, 2 H), 6.88-6.83 (m, 2 H), 3.81 (s, 3 H), 3.49-3.40 (m, 1 H), 3.04-2.95 (m, 1 H), 2.68-2.59 (m, 1 H), 1.25 (d, J = 6.7 Hz, 3 H); 13C NMR (125 MHz, CDCl3) 158.1, 135.5, 132.2, 131.5, 130.4, 129.1, 127.0, 114.0, 55.5, 44.9, 42.5, 20.4; IR (neat) 3072 (w), 3057 (w), 3031 (w), 3001 (w), 2958 (w), 2925 (w), 2864 (w), 2834 (w), 1612 (w), 1583 (w), 1513 (m), 1480 (w), 1472 (w), 1463 (w), 1454 (w), 1439 (w), 1373 (w), 1301 (w), 1249 (m), 1177 (w), 1113 (w), 1091 (w), 1036 (w), 814 (w), 745 (w), 692 (w); MS (EI+, 70 eV) 258.1 (M+, 31), 149.1 (C10H13O+, 30), 137.0 (C8H9S+, 100), 121.1 (C8H9O , 74), 109.0 (15), 91.0 (12), 77.1 (14); TLC Rf 0.39 (60:40, hexane/toluene) [KMnO4]; Analysis C16H18OS (258.38). Calcd: C, 74.38; H, 7.02%. Found: C, 74.56; H, 6.94%.</p><p> </p><!><p>A 25-mL, one-necked, round-bottomed flask equipped with a stirrer bar and rubber septum was charged with 61 (394 mg, 1.52 mmol, 1.0 equiv), ammonium molybdate tetrahydrate (188 mg, 0.15 mmol, 10 mol %), and MeOH (4.5 mL), and stirring was commenced. The mixture was cooled in an ice/water bath then hydrogen peroxide (30% in H2O, 691 mg, 622 μL, 6.09 mmol, 4.0 equiv) was added dropwise via syringe over ca. 14 min. The resultant turbid, pale yellow mixture was stirred in the ice/water bath for 35 min, then allowed to warm to rt over 1 h, during which time the yellow color intensified. The mixture was then cooled in an ice/water bath and sat. aq. Na2SO3 (2.3 mL) was added dropwise via syringe over ca. 7 min. Starch-iodide paper was used to confirm that no oxidant remained. EtOAc (10 mL) and H2O (10 mL) were then added and the layers were separated. The aqueous layer was extracted with EtOAc (2 × 10 mL) and the combined organic extracts were dried (MgSO4), filtered, and concentrated in vacuo (50 °C, ca. 5 mmHg) to give a cloudy, colorless syrup (477 mg). Purification via bulb-to-bulb distillation at reduced pressure (10–5 mmHg) gave 18f as a clear, colorless syrup (438 mg, 99%). Data for 18f: bp 200 °C ABT (10–5 mmHg); 1H NMR (500 MHz, CDCl3) 7.94-7.90 (m, 2 H), 7.68-7.63 (m, 1 H), 7.60-7.54 (m, 2 H), 6.99 (d, J = 8.7, 2 H), 6.80 (d, J = 8.7 Hz, 2 H), 3.75 (s, 3 H), 3.35 (dd, J = 13.5, 3.1 Hz, 1 H), 3.26-3.16 (m, 1 H), 2.47 (dd, J = 13.5, 11.5 Hz, 1 H), 1.13 (d, J = 6.8 Hz, 3 H); 13C NMR (125 MHz, CDCl3) 158.7, 137.4, 134.0, 134.0, 130.3, 129.4, 129.0, 114.3, 62.0, 55.5, 34.7, 12.9; IR (CHCl3 mull) 3063 (w), 3031 (w), 2994 (w), 2935 (m), 2875 (w), 2836 (w), 1611 (s), 1584 (m), 1513 (s), 1446 (s), 1421 (w), 1377 (w), 1303 (s), 1249 (s), 1202 (w), 1179 (s), 1145 (s), 1116 (m), 1086 (s), 1070 (m), 1033 (s), 999 (w), 911 (w), 865 (w), 847 (m), 817 (s), 776 (m), 759 (m), 731 (s), 692 (s), 594 (s), 579 (m), 556 (s); MS (ESI) 345.1 (60), 313.0 ([M+Na]+, 100), 149.1 (C10H13O+, 40); TLC Rf 0.17 (80:20, hexane/EtOAc) [KMnO4]; Analysis: C16H18O3S (290.38). Calcd: C, 66.18; H, 6.25%. Found: C, 66.43; H, 6.39%.</p><p> </p><!><p>A 25-mL Schlenk flask equipped with a stirrer bar, rubber septum, and argon inlet was evacuated and flame-dried, then left to cool to rt and flushed with argon three times. Ethyl phenyl sulfone (511 mg, 3.00 mmol, 1.0 equiv) was quickly added against a backflow of argon, followed by THF (6.0 mL), and stirring was commenced. The resultant solution was cooled in a dry ice/acetone bath then BuLi (2.38 M in hexanes, 1.26 mL, 3.00 mmol, 1.0 equiv) was added dropwise via syringe, causing a color change from colorless to yellow. After 15 min, 2-bromomethyl-1,3-dioxolane (517 mg, 317 μL, 3.00 mmol, 1.0 equiv) was added in one portion via syringe and the resultant mixture was allowed to warm to rt over 20 h. 1 M aq. HCl (3 mL) and EtOAc (12 mL) were then added sequentially and the layers were separated. The aqueous layer was extracted with EtOAc (2 × 6 mL) and the combined organic extracts were dried (MgSO4), filtered, and concentrated in vacuo (50 °C, ca. 5 mmHg) to give a brown oil (792 mg). Purification via flash column chromatography (40 g SiO2, 30 mm Ø, 70:30, hexane/EtOAc, ca. 5-mL fractions then 20 g SiO2, 20 mm Ø, 70:30, hexane/EtOAc, ca. 5-mL fractions) gave a clear, pale yellow oil (472 mg). Further purification via preparative, radial, centrifugally accelerated, thin-layer chromatography on a Harrison Chromatotron (4 mm SiO2 plate, 65:35, hexane/EtOAc, ca. 5-mL fractions) gave a clear, colorless oil (407 mg). Further purification via bulb-to-bulb distillation at reduced pressure (10–5 mmHg) gave 18g as a clear, colorless oil (390 mg, 51%). Data for 18g: bp 175 °C ABT (10–5 mmHg); 1H NMR (500 MHz, CDCl3) 7.88-7.83 (m, 2 H), 7.66-7.60 (m, 1 H), 7.57-7.51 (m, 2 H), 4.96-4.91 (m, 1 H), 3.95-3.75 (m, 4 H), 3.34-3.23 (m, 1 H), 2.32-2.23 (m, 1 H), 1.78-1.65 (m, 1 H), 1.32 (d, J = 6.9 Hz, 3 H); 13C NMR (125 MHz, CDCl3) 137.0, 134.0, 129.4, 129.3, 102.2, 56.6, 33.7, 14.2; IR (neat) 3064 (w), 2981 (m), 2938 (m), 2888 (s), 2254 (w), 1728 (w), 1585 (w), 1478 (m), 1447 (s), 1407 (m), 1362 (w), 1303 (s), 1247 (m), 1214 (m), 1140 (s), 1084 (s), 1025 (s), 999 (m), 961 (s), 916 (m), 853 (w), 826 (m), 769 (s), 735 (s), 692 (s), 635 (m), 595 (s), 578 (s); MS (CI) 256.9 ([M+H]+, 10), 115.0 (C6H11O2+, 100); HRMS (Cl , double focusing sector field) calcd for C12H17O4S: 257.0848, found: 257.0852; TLC Rf 0.35 (60:40, hexane/EtOAc) [KMnO4].</p><p> </p><!><p>A 100-mL Schlenk flask equipped with a stirrer bar, rubber septum, and argon inlet was evacuated and flame-dried, then left to cool to rt and flushed with argon three times. Ethyl phenyl sulfone (1.02 g, 6.00 mmol, 1.5 equiv) was quickly added against a backflow of argon, followed by THF (18.0 mL), and stirring was commenced. The resultant solution was cooled in an ice/water bath then BuLi (2.38 M in hexanes, 2.52 mL, 6.00 mmol, 1.5 equiv) was added dropwise via syringe, causing a color change from colorless to yellow. After 30 min, a solution of 56 (1.38 g, 4.00 mmol, 1.0 equiv) in THF (6.0 mL) was added dropwise via cannula over ca. 5 min, causing a color change from yellow to orange. The resultant mixture was allowed to warm to rt over 26 h, then H2O (30 mL) and EtOAc (30 mL) were then added sequentially and the layers were separated. The aqueous layer was extracted with EtOAc (2 × 30 mL) and the combined organic extracts were dried (MgSO4), filtered, and concentrated in vacuo (50 °C, ca. 5 mmHg) to give a yellow oil (1.85 g). Purification via flash column chromatography (100 g SiO2, 55 mm Ø, 90:10→70:30→50:50→0:100, hexane/EtOAc, ca. 24 mL fractions) gave a cream-colored solid (768 mg). Further purification of this material was performed via recrystallization from 60:40, hexane:CH2Cl2 (ca. 6 mL) in a 20-mL scintillation vial. The crystals were collected via filtration through filter paper in a Hirsch funnel under House vacuum, then crushed with a glass rod and dried in vacuo (0.05 mmHg) to give a white, crystalline solid (679 mg). Further purification of this material was performed via recrystallization from MeOH (5.0 mL) in a 20-mL scintillation vial. The crystals were collected via filtration through filter paper in a Hirsch funnel under House vacuum, washed with cold (0 °C) MeOH (2.0 mL) and then crushed with a glass rod and dried in vacuo (0.05 mmHg) to give 18i as a white, crystalline solid (561 mg, 41%). Data for 18i: mp 123-124 °C (MeOH); 1H NMR (500 MHz, CDCl3) 7.89-7.85 (m, 2 H), 7.66-7.61 (m, 1 H), 7.58-7.52 (m, 2 H), 7.33-7.27 (m, 4 H), 7.27-7.21 (m, 1 H), 3.53-3.41 (m, 2 H), 2.99-2.86 (m, 3 H), 2.26-2.16 (m, 1 H), 2.03-1.92 (m, 2 H), 1.92-1.84 (m, 1 H), 1.61-1.43 (m, 3 H), 1.21 (d, J = 7.1 Hz, 3 H); 13C NMR (125 MHz, CDCl3) 138.7, 138.5, 133.8, 129.4, 129.4, 128.8, 128.4, 127.2, 64.2, 63.5, 54.0, 53.7, 34.8, 31.4, 27.0, 9.8; IR (CHCl3 mull) 3061 (w), 3027 (w), 2940 (m), 2803 (m), 2758 (w), 2723 (w), 1584 (w), 1494 (w), 1446 (m), 1394 (w), 1381 (w), 1367 (w), 1343 (w), 1304 (s), 1214 (w), 1149 (s), 1086 (m), 1071 (w), 1043 (m), 1027 (w), 1011 (w), 999 (w), 985 (w), 911 (w), 831 (w), 786 (w), 764 (m), 734 (s), 698 (m), 646 (w), 592 (s); MS (EI+, 70 eV) 343.2 (M+, 10), 202.2 (C14H20N+, 24), 161.1 (31), 110.0 (12), 105.0 (19), 91.1 (C7H7+, 100), 77.1 (26), 51.0 (15); TLC Rf 0.29 (EtOAc) [KMnO4]; Analysis C20H25NO2S (343.48). Calcd: C, 69.93; H, 7.34; N, 4.08%. Found: C, 70.11; H, 7.37;N, 4.17%.</p><p> </p><!><p>A 50-mL Schlenk flask equipped with a stirrer bar, rubber septum, and argon inlet was evacuated and flame-dried, then left to cool to rt and flushed with argon three times. Ethyl phenyl sulfone (1.02 mg, 6.00 mmol, 1.0 equiv) was quickly added against a backflow of argon, followed by THF (12 mL), and stirring was commenced. The resultant solution was cooled in a dry ice/acetone bath then BuLi (2.38 M in hexanes, 2.52 mL, 6.00 mmol, 1.0 equiv) was added dropwise via syringe, causing a color change from colorless to yellow. After 15 min, paraformaldehyde (901 mg, 317 μL, 30.0 mmol, 5.0 equiv) was added in one portion against a backflow of argon and the resultant mixture was allowed to warm to rt over 13 h. 1 M aq. HCl (6 mL) and EtOAc (24 mL) were then added sequentially and the layers were separated. The aqueous layer was extracted with EtOAc (2 × 12 mL) and the combined organic extracts were dried (MgSO4), filtered, and concentrated in vacuo (50 °C, ca. 5 mmHg) to give a dark pink oil (1.17 g). Purification via flash column chromatography (40 g SiO2, 30 mm Ø, 70:30, hexane/EtOAc, ca. 5-mL fractions then 20 g SiO2, 20 mm Ø, 50:50, hexane/EtOAc, ca. 5-mL fractions) gave a clear, colorless oil (329 mg). Further purification via preparative, radial, centrifugally accelerated, thin-layer chromatography on a Harrison Chromatotron (4 mm SiO2 plate, 50:50, hexane/EtOAc, ca. 5-mL fractions) gave a clear, colorless oil (300 mg). Further purification via bulb-to-bulb distillation at reduced pressure (10–5 mmHg) gave 18l as a clear, colorless oil (274 mg, 23%). Data for 18l: bp 175 °C ABT (10–5 mmHg); 1H NMR (500 MHz, CDCl3) 7.89-7.84 (m, 2 H), 7.69-7.64 (m, 1 H), 7.60-7.54 (m, 2 H), 3.92 (dd, J = 12.5, 6.8 Hz, 1 H), 3.78 (dd, J = 12.5, 4.0 Hz, 1 H), 3.31-3.22 (m, 1 H), 2.95 (br s, 1 H), 1.23 (d, J = 7.1 Hz, 3 H); 13C NMR (125 MHz, CDCl3) 137.1, 134.3, 129.5, 129.0, 61.8, 61.8, 11.5; IR (neat) 3499 (br, OH), 3064 (w), 2983 (w), 2940 (m), 2882 (w), 1584 (w), 1303 (s), 1220 (m), 1143 (s), 1084 (s), 1045 (s), 999 (m), 982 (m), 929 (w), 865 (m), 765 (m), 733 (s), 690 (s), 665 (m), 646 (m), 596 (s); MS (EI+, 70 eV) 200.0 (M+, 2), 170 (10), 142.0 (38), 125 (16), 94.0 (17), 78.1 (100), 59.0 (100); HRMS (El+, double focusing sector field) calcd for C9H12O3S: 200.0507, found: 200.0509; TLC Rf 0.30 (50:50, hexane/EtOAc) [KMnO4].</p><p> </p><!><p>A 50-mL, one-necked, round-bottomed flask equipped with a stirrer bar and rubber septum was charged with 1-bromodecane (790 mg, 741 μL, 3.50 mmol, 1.0 equiv), benzenesulfinic acid sodium salt (689 mg, 4.20 mmol, 1.2 equiv), and DMF (12.0 mL), and stirring was commenced. The resultant mixture was stirred at rt for 14 h, then H2O (60 mL) and EtOAc (20 mL) were added and the layers were separated. The aqueous layer was extracted with EtOAc (2 × 20 mL) and the combined organic extracts were washed with H2O (5 × 40 mL), then dried (MgSO4), filtered, and concentrated in vacuo (50 °C, ca. 5 mmHg) to give a clear, colorless oil (948 mg). Purification via flash column chromatography (20 g SiO2, 20 mm Ø, 90:10 hexane/EtOAc, ca. 5-mL fractions) gave 21 as a clear, colorless oil (507 mg, 51%). The 1H NMR spectroscopic data matched that for alternative preparations.113</p><p> </p><!><p>A 100-mL Schlenk flask equipped with a stirrer bar, rubber septum, and argon inlet was evacuated and flame-dried, then left to cool to rt and flushed with argon three times. A solution of 9 (960 mg, 3.50 mmol, 1.0 equiv) in THF (10.0 mL) was added via cannula, followed by additional THF (23.0 mL), and stirring was commenced. The resultant solution was cooled to –78 °C in a dry ice/acetone bath then BuLi (2.57 M in hexanes, 1.36 mL, 3.50 mmol, 1.0 equiv) was added dropwise via syringe, causing a color change from colorless to yellow. After 1 h, iodomethane (745 mg, 327 μL, 5.20 mmol, 1.5 equiv) was added dropwise via syringe over ca. 1 min, causing a color change from yellow to colorless after a further ca. 1 min. The resultant mixture was allowed to warm to rt over 20 h. Sat. aq. NH4Cl (10 mL) and H2O (10 mL) were then added sequentially and the layers were separated. The aqueous layer was extracted with EtOAc (2 × 10 mL) and the combined organic extracts were washed with a 2:1 mixture of brine and sat. aq. Na2S2O3 (30 mL), then (dried (MgSO4), filtered, and concentrated in vacuo (50 °C, ca. 5 mmHg) to give a clear, yellow oil which solidified on standing to a white crystalline solid (1.09 g). Purification was performed via recrystallization from MeOH (3.0 mL) in a 20-mL scintillation vial. The crystals were collected via filtration through filter paper in a Hirsch funnel under House vacuum, washed with cold (–78 °C) MeOH (2.0 mL) and then crushed with a glass rod and dried in vacuo (0.05 mmHg) to give 25 as a white, crystalline solid (853 mg, 85%). The 1H NMR spectroscopic data and melting point matched that for alternative preparations.114</p><!><p> </p><!><p>Following General Procedure 1, 14 (304 mg, 1.00 mmol, 1.0 equiv), PhMgBr (3.12 M in Et2O, 962 μL, 3.00 mmol, 3.0 equiv), Fe(acac)3 (70.6 mg, 0.20 mmol, 20 mol %), TMEDA (930 mg, 1.20 mL, 8.00 mmol, 8.0 equiv), and CPME (10.0 mL) were reacted to give an orange oil (386 mg). Purification via preparative, radial, centrifugally accelerated, thin-layer chromatography on a Harrison Chromatotron (4 mm SiO2 plate, 80:20, hexane/toluene, ca. 5-mL fractions) gave a clear, colorless oil (187 mg). Further purification via bulb-to-bulb distillation at reduced pressure (10–5 mmHg) gave 16a as a clear, colorless oil (177 mg, 74%). The 1H NMR spectroscopic data matched that for alternative preparations.115 Data for 16a: bp 100 °C ABT (10–5 mmHg).</p><p> </p><!><p>Following General Procedure 1, 14 (304 mg, 1.00 mmol, 1.0 equiv), 4-tolylmagnesium bromide (1.14 M in Et2O, 2.63 mL, 3.00 mmol, 3.0 equiv), Fe(acac)3 (70.6 mg, 0.20 mmol, 20 mol %), TMEDA (930 mg, 1.20 mL, 8.00 mmol, 8.0 equiv), and CPME (10.0 mL) were reacted to give a suspension of solid in an orange oil (1.08 g). Purification via flash column chromatography (40 g SiO2, 30 mm Ø, 100:0→75:25, hexane/toluene, ca. 5-mL fractions) gave a clear, colorless oil (161 mg). Further purification via bulb-to-bulb distillation at reduced pressure (10–5 mmHg) gave 16b as a clear, colorless oil (152 mg, 60%). Data for 16b: bp 100 °C ABT (10–5 mmHg); 1H NMR (500 MHz, CDCl3) 7.17-7.09 (m, 4 H), 7.07 (d, J = 8.5 Hz, 2 H), 6.83 (d, J = 8.5 Hz, 2 H), 3.80 (s, 3 H), 2.75-2.65 (m, 1 H), 2.53-2.42 (m, 2 H), 2.36 (s, 3 H), 1.95-1.81 (m, 2 H), 1.27 (d, J = 7.0 Hz, 3 H); 13C NMR (125 MHz, CDCl3) 157.8, 144.6, 135.6, 134.9, 129.5, 129.3, 127.2, 113.9, 55.5, 40.5, 39.2, 33.2, 22.9, 21.3; IR (neat) 3094 (w), 3004 (m), 2955 (s), 2925 (s), 2857 (m), 2834 (m), 1612 (m), 1583 (m), 1512 (s), 1455 (m), 1374 (m), 1299 (m), 1245 (s), 1176 (m), 1116 (m), 1038 (s), 817 (s), 750 (w), 722 (m), 702 (w); MS (EI+, 70 eV) 254.1 (M+, 71), 135.0 (44), 121.0 (C8H9O+, 100), 105.0 (27), 91.0 (27), 77.0 (18); TLC Rf 0.28 (70:30, hexane/EtOAc) [KMnO4]; Analysis C18H22O (254.37). Calcd: C, 84.99; H, 8.72%. Found: C, 84.81; H, 8.89%.</p><p> </p><!><p>Following General Procedure 1, 14 (304 mg, 1.00 mmol, 1.0 equiv), 3-tolylmagnesium bromide (2.21 M in Et2O, 1.36 mL, 3.00 mmol, 3.0 equiv), Fe(acac)3 (70.6 mg, 0.20 mmol, 20 mol %), TMEDA (930 mg, 1.20 mL, 8.00 mmol, 8.0 equiv), and CPME (10.0 mL) were reacted to give an orange oil (600 mg). Purification via flash column chromatography (40 g SiO2, 30 mm Ø, 100:0→90:10, hexane/toluene, ca. 5-mL fractions) gave a clear, yellow oil (170 mg). Further purification via bulb-to-bulb distillation at reduced pressure (10–5 mmHg) gave 16c as a clear, colorless oil (163 mg, 64%). Data for 16c: bp 100 °C ABT (10–5 mmHg); 1H NMR (500 MHz, CDCl3) 7.27-7.72 (m, 1 H), 7.12-7.03 (m, 5 H), 6.85 (d, J = 8.7 Hz, 2 H), 3.82 (s, 3 H), 2.76-2.67 (m, 1 H), 2.57-2.44 (m, 2 H), 2.39 (s, 3 H), 1.99-1.84 (m, 2 H), 1.30 (d, J = 7.0 Hz, 3 H); 13C NMR (125 MHz, CDCl3) 157.9, 147.6, 138.1, 134.9, 129.5, 128.5, 128.1, 126.9, 124.3, 113.9, 55.5, 40.5, 39.6, 33.3, 22.8, 21.8; IR (neat) 3100 (m), 3027 (m), 3006 (m), 2956 (s), 2926 (s), 2857 (m), 2834 (m), 1609 (s), 1584 (m), 1511 (s), 1489 (m), 1456 (s), 1374 (m), 1299 (m), 1245 (s), 1176 (s), 1113 (m), 1038 (s), 880 (m), 828 (s), 785 (s), 750 (m), 704 (s); MS (EI+, 70 eV) 254.1 (M+, 78), 135.0 (34), 121.0 (C8H9O+, 100), 105.0 (35), 91.0 (23), 77.0 (14); TLC Rf 0.25 (70:30, hexane/toluene) [KMnO4]; Analysis C18H22O (254.37). Calcd: C, 84.99; H, 8.72%. Found: C, 84.73; H, 8.85%.</p><p> </p><!><p>An oven-dried, 25-mL, one-necked, round-bottomed flask was charged with 14 (304 mg, 1.00 mmol, 1.0 equiv) and Fe(acac)3 (70.6 mg, 0.20 mmol, 20 mol %) in a glove box then was sealed with a rubber septum and removed from the box. Outside of the glove box, a 25 mL Schlenk flask equipped with a stirrer bar, rubber septum and argon inlet was evacuated, flame-dried, left to cool under vacuum, and flushed three times with argon. TMEDA (930 mg, 1.20 mL, 8.00 mmol, 8.0 equiv) was added via syringe to the Schlenk flask and stirring was commenced. The round-bottomed flask containing 14 and Fe(acac)3 was charged with CPME (4.0 mL) then sonicated until homogeneous. The clear red solution was then transferred via cannula to the Schlenk flask holding the TMEDA, and the residual material was rinsed across with further portions of CPME (6.0 mL). 4-(Trimethylsilyl)phenylmagnesium bromide (1.52 M in Et2O, 1.97 mL, 3.00 mmol, 3.0 equiv) was then added by syringe over ca. 30 sec. During addition, the color of the solution changed from red to pale yellow to brown, but remained clear throughout, and no visible deposits were formed on the edges of the flask. After stirring for 18 h at rt, H2O (10 mL) was added in one portion and the mixture was filtered through a pad of Celite (5 g) in a 40 mm Ø, porosity 3, sintered funnel under House vacuum. H2O (2 × 10 mL) and EtOAc (3 × 10 mL) were used to rinse any residual material though the Celite pad. The filtrate was transferred to a separatory funnel and the layers were separated. The aqueous layer was extracted with EtOAc (2 × 20 mL) and the combined organic layers were dried (MgSO4), filtered and concentrated in vacuo (50 °C, ca. 5 mmHg) to give a brown oil (822 mg). Purification via flash column chromatography (40 g SiO2, 30 mm Ø, 70:30, hexane/toluene, ca. 5-mL fractions) gave a clear, colorless oil (235 mg). Further purification via flash column chromatography (C18-reversed phase silica gel, 20 × 160 mm, MeOH, ca. 5-mL fractions, loaded with minimal MeCN for solubility reasons) gave a clear, colorless oil (185 mg). Further purification via bulb-to-bulb distillation at reduced pressure (10–5 mmHg) gave 16d as a clear, colorless oil (174 mg, 56%). Data for 16d: bp 130 °C ABT (10–5 mmHg); 1H NMR (500 MHz, CDCl3) 7.53 (d, J = 7.7 Hz, 2 H), 7.26 (d, J = 7.7 Hz, 2 H), 7.11 (d, J = 8.5 Hz, 2 H), 6.87 (d, J = 8.5 Hz, 2 H), 3.83 (s, 3 H), 2.81-2.72 (m, 1 H), 2.59-2.48 (m, 2 H), 2.03-1.87 (m, 2 H), 1.33 (d, J = 7.0 Hz, 3 H), 0.33 (s, 9 H); 13C NMR (125 MHz, CDCl3) 157.9, 147.3, 137.8, 134.9, 133.8, 129.5, 126.8, 114.0, 55.5, 40.4, 39.6, 33.3, 22.6, –0.7; IR (neat) 3065 (w), 3030 (w), 3008 (w), 2955 (m), 2930 (m), 2870 (w), 2855 (w), 2833 (w), 1611 (w), 1600 (w), 1584 (w), 1512 (m), 1455 (w), 1398 (w), 1299 (w), 1246 (m), 1176 (w), 1116 (w), 1039 (w), 838 (m), 819 (m), 755 (w), 725 (w), 693 (w), 640 (w), 562 (w); MS (EI+, 70 eV) 312.2 (M+, 36), 297.2 (21), 267.1 (11), 177.1 (13), 161.1 (65), 135.1 (26), 121.1 (C8H9O+, 100), 119.0 (12), 91.1 (22), 77.1 (21), 73.1 (59), 59.1 (14). HRMS (El+, double focusing sector field) calcd for C20H28OSi: 312.1910, found: 312.1904; TLC Rf 0.33 (70:30, hexane/toluene) [KMnO4].</p><p> </p><!><p>An oven-dried, 25-mL, one-necked, round-bottomed flask was charged with 14 (304 mg, 1.00 mmol, 1.0 equiv) and Fe(acac)3 (70.6 mg, 0.20 mmol, 20 mol %) in a glove box then was sealed with a rubber septum and removed from the box. Outside of the glove box, a 25-mL Schlenk flask equipped with a stirrer bar, rubber septum and argon inlet was evacuated, flame-dried, left to cool under vacuum, and flushed three times with argon. TMEDA (930 mg, 1.20 mL, 8.00 mmol, 8.0 equiv) was added via syringe to the Schlenk flask and stirring was commenced. The round-bottomed flask containing 14 and Fe(acac)3 was charged with CPME (4.0 mL) then sonicated until homogeneous. The clear red solution was then transferred via cannula to the Schlenk flask holding the TMEDA, and the residual material was rinsed across with further portions of CPME (6.0 mL). 3-(Trimethylsilyl)phenylmagnesium bromide (2.10 M in Et2O, 1.43 mL, 3.00 mmol, 3.0 equiv) was then added by syringe over ca. 30 sec. During addition, the color of the solution changed from red to pale yellow to brown, but remained clear throughout, and no visible deposits were formed on the edges of the flask. After stirring for 18 h at rt, H2O (10 mL) was added in one portion and the mixture was filtered through a pad of Celite (5 g) in a 40 mm Ø, porosity 3, sintered funnel House vacuum. H2O (2 × 10 mL) and EtOAc (3 × 10 mL) were used to rinse any residual material though the Celite pad. The filtrate was transferred to a separatory funnel and the layers were separated. The aqueous layer was extracted with EtOAc (2 × 20 mL) and the combined organic layers were dried (MgSO4), filtered and concentrated in vacuo (50 °C, ca. 5 mmHg) to give a brown oil (786 mg). Purification via flash column chromatography (40 g SiO2, 30 mm Ø, 70:30, hexane/toluene, ca. 5-mL fractions) gave a clear, colorless oil (220 mg). Further purification via flash column chromatography (C18-reversed phase silica gel, 20 × 160 mm, MeOH, ca. 5-mL fractions, loaded with minimal MeCN for solubility reasons) gave a clear, colorless oil (202 mg). Further purification via bulb-to-bulb distillation at reduced pressure (10–5 mmHg) gave 16e as a clear, colorless oil (186 mg, 60%). Data for 16e: bp 130 °C ABT (10–5 mmHg); 1H NMR (500 MHz, CDCl3) 7.43 (dt, J = 7.2, 1.2 Hz, 1 H), 7.39 (br s, 1 H), 7.37 (dd, J = 7.6, 7.2 Hz, 1 H), 7.26 (dt, J = 7.6, 1.5 Hz, 1 H), 7.13-7.09 (m, 2 H), 6.90-6.86 (m, 2 H), 3.84 (s, 3 H), 2.81-2.73 (m, 1 H), 2.56-2.51 (m, 2 H), 2.03-1.89 (m, 2 H), 1.34 (d, J = 7.0 Hz, 3 H), 0.34 (s, 9 H); 13C NMR (125 MHz, CDCl3) 157.9, 146.7, 140.6, 134.9, 132.6, 131.3, 129.5, 128.1, 127.6, 114.0, 55.5, 40.5, 39.7, 33.3, 22.8, –0.7; IR (neat) 3028 (w), 2995 (w), 2955 (m), 2932 (m), 2870 (w), 2855 (w), 2833 (w), 1611 (w), 1583 (w), 1511 (m), 1458 (w), 1406 (w), 1373 (w), 1299 (w), 1246 (m), 1176 (w), 1121 (w), 1039 (w), 861 (m), 837 (m), 794 (w), 752 (m), 706 (w), 621 (w), 560 (w); MS (EI+, 70 eV) 312.2 (M+, 34), 297.2 (14), 161.1 (53), 135.1 (20), 121.0 (C8H9O+, 100), 119.0 (13), 91.1 (20), 77.0 (17), 73.1 (78), 59.1 (11); TLC Rf 0.40 (70:30, hexane/toluene) [KMnO4]; Analysis C20H28OSi (312.52). Calcd: C, 76.86; H, 9.03%. Found: C, 77.13; H, 9.07%.</p><p> </p><!><p>Following General Procedure 1, 14 (304 mg, 1.00 mmol, 1.0 equiv), 3-isopropoxyphenylmagnesium bromide (1.55 M in Et2O, 1.94 mL, 3.00 mmol, 3.0 equiv), Fe(acac)3 (70.6 mg, 0.20 mmol, 20 mol %), TMEDA (930 mg, 1.20 mL, 8.00 mmol, 8.0 equiv), and CPME (10.0 mL) were reacted to give an orange oil (842 mg). Purification via flash column chromatography (40 g SiO2, 30 mm Ø, 100:0→80:20→50:50, hexane/toluene, ca. 5-mL fractions) gave 16f as a clear, pale orange oil (123 mg), in addition to a mixture of 16f and 3,3′-diisopropoxybiphenyl as a clear, colorless oil (138 mg). The mixed fractions were further purified via flash column chromatography (20 g SiO2, 20 mm Ø, 50:50, hexane/toluene, ca. 2.5-mL fractions then 20 g SiO2, 20 mm Ø, 60:40, hexane/toluene, ca. 2.5-mL fractions) to give 16f as a clear, colorless oil (97 mg). The combined portions of 16f (220 mg) were further purified via bulb-to-bulb distillation at reduced pressure (10–5 mmHg) to give 16f as a clear, colorless oil (209 mg, 70%). Data for 16f: bp 135 °C ABT (10–5 mmHg); 1H NMR (500 MHz, CDCl3) 7.24-7.18 (m, 1 H), 7.09-7.03 (m, 2 H), 6.84-6.80 (m, 2 H), 6.80-6.71 (m, 3 H), 4.56 (hept, J = 6.1 Hz, 1 H), 3.79 (s, 3 H), 2.72-2.63 (m, 1 H), 2.53-2.42 (m, 2 H), 1.95-1.79 (m, 2 H), 1.36 (d, J = 6.1 Hz, 6 H), 1.26 (d, J = 6.9 Hz, 3 H); 13C NMR (125 MHz, CDCl3) 158.2, 157.9, 149.4, 134.9, 129.5, 129.5, 119.7, 115.3, 113.9, 113.0, 69.9, 55.5, 40.4, 39.7, 33.2, 22.7, 22.4; IR (neat) 3031 (m), 2974 (s), 2931 (s), 2870 (m), 2834 (m), 1609 (s), 1582 (s), 1512 (s), 1484 (s), 1453 (s), 1383 (m), 1372 (m), 1246 (s), 1177 (s), 1156 (m), 1137 (m), 1117 (s), 1038 (s), 999 (m), 973 (m), 873 (m), 822 (m), 777 (m), 701 (s); MS (ESI) 321.2 ([M+Na]+, 22), 316.2 ([M+NH4]+, 6), 299.2 ([M+H]+, 100), 257.2 ([M–C3H6+H]+, 96); TLC Rf 0.09 (70:30, hexane/toluene) [KMnO4]; Analysis C20H26O2 (298.42). Calcd: C, 80.50; H, 8.78%. Found: C, 80.48; H, 8.88%.</p><p> </p><!><p>Following General Procedure 1, 14 (304 mg, 1.00 mmol, 1.0 equiv), 4-biphenylmagnesium bromide (1.42 M in Et2O, 2.11 mL, 3.00 mmol, 3.0 equiv), Fe(acac)3 (70.6 mg, 0.20 mmol, 20 mol %), TMEDA (930 mg, 1.20 mL, 8.00 mmol, 8.0 equiv), and CPME (10.0 mL) were reacted to give a dark orange oil (914 mg). Purification via flash column chromatography (40 g SiO2, 30 mm Ø, 80:20, hexane/toluene, ca. 5-mL fractions) gave a clear, pale yellow oil (204 mg). Further purification via flash column chromatography (C18-reversed phase silica gel, 20 × 160 mm, MeOH, ca. 5-mL fractions, loaded with minimal MeCN for solubility reasons) gave a clear, colorless oil (174 mg). Further purification via bulb-to-bulb distillation at reduced pressure (10–5 mmHg) gave 16g as a clear, colorless oil (159 mg, 50%). Data for 16g: bp 180 °C ABT (10–5 mmHg); 1H NMR (500 MHz, CDCl3) 7.68-7.63 (m, 2 H), 7.62-7.58 (m, 2 H), 7.51-7.46 (m, 2 H), 7.41-7.35 (m, 1 H), 7.35-7.30 (m, 2 H), 7.14-7.09 (m, 2 H), 6.90-6.84 (m, 2 H), 3.83 (s, 3 H), 2.86-2.76 (m, 1 H), 2.61-2.49 (m, 2 H), 2.05-1.89 (m, 2 H), 1.36 (d, J = 6.9 Hz, 3 H); 13C NMR (125 MHz, CDCl3) 157.9, 146.8, 141.4, 139.1, 134.8, 129.5, 129.0, 127.8, 127.4, 127.3, 114.0, 55.5, 40.5, 39.3, 33.3, 22.8; IR (neat) 3056 (m), 3028 (m), 3000 (m), 2956 (m), 2929 (m), 2869 (m), 2855 (m), 2834 (m), 1611 (m), 1583 (m), 1512 (s), 1486 (m), 1454 (m), 1408 (w), 1374 (w), 1346 (w), 1299 (m), 1244 (s), 1177 (m), 1118 (w), 1075 (w), 1037 (m), 1008 (m), 837 (m), 765 (m), 733 (m), 697 (m), 573 (w), 559 (w); MS (EI+, 70 eV) 316.2 (M+, 63), 181.1 (88), 178.1 (25), 165.1 (39), 152.1 (24), 135.1 (36), 121.0 (C8H9O+, 100), 115.1 (12), 103.1 (11), 91.0 (31), 77.0 (37), 65.1 (13), 51.0 (10); TLC Rf 0.22 (70:30, hexane/toluene) [KMnO4]; Analysis: C23H24O (316.44). Calcd: C, 87.30; H, 7.64%. Found: C, 87.54; H, 7.70%.</p><p> </p><!><p>Following General Procedure 1, 14 (304 mg, 1.00 mmol, 1.0 equiv), 2-naphthylmagnesium bromide (1.61 M in Et2O, 1.86 mL, 3.00 mmol, 3.0 equiv), Fe(acac)3 (70.6 mg, 0.20 mmol, 20 mol %), TMEDA (930 mg, 1.20 mL, 8.00 mmol, 8.0 equiv), and CPME (10.0 mL) were reacted to give a sticky, orange solid (1.3 g). Purification via flash column chromatography (40 g SiO2, 30 mm Ø, 100:0→80:20, hexane/toluene, ca. 5-mL fractions then 40 g SiO2, 30 mm Ø, 100:0→80:20 hexane/toluene, ca. 5-mL fractions then 20 g SiO2, 20 mm Ø, 80:20, hexane/toluene, ca. 2.5-mL fractions) gave a clear, colorless oil (142 mg). Further purification via flash column chromatography (C18-reversed phase silica gel, 20 × 160 mm, MeOH, ca. 5-mL fractions, loaded with minimal MeCN for solubility reasons) gave a clear, colorless oil (140 mg). Further purification via bulb-to-bulb distillation at reduced pressure (10–5 mmHg) gave 16h as a clear, colorless oil (128 mg, 44%). Data for 16h: bp 180 °C ABT (10–5 mmHg); 1H NMR (500 MHz, CDCl3) 7.89-7.83 (m, 3 H), 7.68 (s, 1 H), 7.54-7.45 (m, 2 H), 7.44-7.40 (m, 1 H), 7.12-7.08 (m, 2 H), 6.88-6.84 (m, 2 H), 3.82 (s, 3 H), 2.98-2.89 (m, 1 H), 2.60-2.47 (m, 2 H), 2.11-1.95 (m, 2 H), 1.40 (d, J = 6.9 Hz, 3 H); 13C NMR (125 MHz, CDCl3) 157.9, 145.0, 134.8, 133.9, 132.5, 129.5, 128.3, 127.9, 127.8, 126.1, 126.0, 125.6, 125.4, 114.0, 55.5, 40.3, 39.8, 33.3, 22.8; IR (neat) 3052 (m), 3027 (w), 3006 (m), 2956 (m), 2930 (m), 2867 (m), 2855 (m), 2833 (m), 1632 (w), 1611 (m), 1583 (w), 1511 (s), 1455 (m), 1440 (m), 1378 (m), 1320 (w), 1299 (m), 1244 (s), 1177 (m), 1127 (w), 1111 (w), 1037 (m), 950 (w), 890 (w), 855 (m), 819 (m), 747 (m), 702 (w), 660 (w), 621 (w), 564 (w); MS (EI+, 70 eV) 290.2 (M+, 27), 156.1 (100), 141.1 (39), 135.1 (12), 128.1 (27), 121.0 (C8H9O+, 61), 91.1 (22), 77.0 (26), 65.1 (10); TLC Rf 0.24 (70:30, hexane/toluene) [KMnO4]; Analysis C21H22O (290.40). Calcd: C, 86.85; H, 7.64%. Found: C, 87.08; H, 7.88%.</p><!><p> </p><!><p>Following General Procedure 1, 18a (224 mg, 1.00 mmol, 1.0 equiv), 3-isopropoxyphenylmagnesium bromide (2.24 M in Et2O, 1.34 mL, 3.00 mmol, 3.0 equiv), Fe(acac)3 (70.6 mg, 0.20 mmol, 20 mol %), TMEDA (930 mg, 1.20 mL, 8.00 mmol, 8.0 equiv), and CPME (10.0 mL) were reacted to give a yellow oil (712 mg). Purification via flash column chromatography (40 g SiO2, 30 mm Ø, 95:5, hexane/toluene, ca. 5-mL fractions) gave a clear, colorless oil (213 mg). Further purification via flash column chromatography (C18-reversed phase silica gel, 20 × 160 mm, MeOH, ca. 2.5-mL fractions, loaded with minimal MeCN for solubility reasons) gave a clear, colorless oil (152 mg). Further purification via bulb-to-bulb distillation at reduced pressure (0.03 mmHg) gave 19a as a clear, colorless oil (145 mg, 67%). Data for 19a: bp 75 °C ABT (0.03 mmHg); 1H NMR (500 MHz, CDCl3) 7.19 (app t, J = 7.8 Hz, 1 H), 6.80-6.74 (m, 2 H), 6.73-6.68 (m, 1 H), 4.60-4.50 (hept, J = 6.1 Hz, 1 H), 2.53-2.39 (m, 1 H), 1.93-1.70 (m, 5 H), 1.49-1.17 (m, 5 H) overlapping 1.34 (d, J = 6.1 Hz, 6 H); 13C NMR (125 MHz, CDCl3) 158.1, 150.1, 129.4, 119.4, 115.0, 112.9, 69.8, 44.9, 34.7, 27.2, 26.5, 22.4; IR (neat) 3029 (m), 2975 (s), 2925 (s), 2851 (s), 1600 (s), 1580 (s), 1489 (s), 1447 (s), 1382 (m), 1371 (m), 1350 (m), 1316 (m), 1287 (s), 1255 (s), 1225 (m), 1181 (m), 1156 (s), 1118 (s), 1017 (m), 999 (m), 977 (s), 919 (m), 872 (m), 829 (m), 774 (m), 751 (m), 698 (s); MS (EI+, 70 eV) 218.2 (M+, 83), 176.1 ([M–C3H6]+, 100), 161.1 (19), 147.1 (16), 133.1 (56), 120.1 (76), 108.1 (79), 91.1 (26), 77.0 (17); TLC Rf 0.34 (90:10, hexane/toluene) [KMnO4]; Analysis C15H22O (218.33). Calcd: C, 82.52; H, 10.16%. Found: C, 82.64; H, 10.28%.</p><p> </p><!><p>An oven-dried, 20-mL scintillation vial was charged with 18b (315 mg, 1.00 mmol, 1.0 equiv) and Fe(acac)3 (70.6 mg, 0.20 mmol, 20 mol %) in a glove box then was sealed with a rubber septum and removed from the box. Outside of the glove box, a 25-mL Schlenk flask equipped with a stirrer bar, rubber septum and argon inlet was evacuated, flame-dried, left to cool under vacuum, and flushed three times with argon. TMEDA (930 mg, 1.20 mL, 8.00 mmol, 8.0 equiv) was added via syringe to the Schlenk flask and stirring was commenced. The vial containing 18b and Fe(acac)3 was charged with CPME (4.0 mL) then sonicated until homogeneous. The clear red solution was then transferred via cannula to the Schlenk flask holding the TMEDA, and the residual material was rinsed across with further portions of CPME (6.0 mL). PhMgBr (2.76 M in Et2O, 1.09 mL, 3.00 mmol, 3.0 equiv) was then added by syringe over ca. 20 sec. During addition, the color of the solution changed from red to pale yellow to brown, but remained clear throughout, and no visible deposits were formed on the edges of the flask. After stirring for 18 h at rt, 1 M aq. HCl (10 mL) was added in one portion and the mixture was filtered through a pad of Celite (5 g) in a 40 mm Ø, porosity 3, sintered funnel under House vacuum. EtOAc (2 × 5 mL) was used to rinse any residual material though the Celite pad. The filtrate was brought to pH 9 by addition of 5% aq. NaOH and was then transferred to a separatory funnel and the layers were separated. The aqueous layer was extracted with EtOAc (2 × 10 mL) and the combined organic extracts were dried (MgSO4), filtered and concentrated in vacuo (50 °C, ca. 5 mmHg) to give a dark red oil (397 mg). Purification via flash column chromatography (40 g SiO2, 30 mm Ø, 98:1.8:0.2 CH2Cl2:MeOH:aq. NH3, ca. 5-mL fractions) gave an 87:13 mixture of 19b:18b as a clear, orange oil (254 mg). Further purification via flash column chromatography [C18-reversed phase silica gel, 20 × 160 mm, MeOH, ca. 5-mL fractions then C18-reversed phase silica gel, 20 × 160 mm, MeOH, ca. 2.5-mL fractions (loaded with minimal MeCN in both cases for solubility reasons)] gave 19b as a clear, pale yellow oil (165 mg). Further purification via bulb-to-bulb distillation at reduced pressure (10–5 mmHg) gave 19b as a clear, colorless oil (149 mg, 59%). Data for 19b: bp 130 °C ABT (10–5 mmHg); 1H NMR (500 MHz, CDCl3) 7.40-7.17 (m, 10 H), 3.58 (s, 2 H), 3.03 (d, J = 11.3 Hz, 2 H), 2.50 (tt, J = 8.1, 8.0 Hz, 1 H), 2.19-2.02 (m, 2 H), 1.91-1.73 (m, 4 H); 13C NMR (125 MHz, CDCl3) 146.8, 138.7, 129.6, 128.7, 128.5, 127.3, 127.2, 126.4, 63.8, 54.6, 43.0, 33.8; IR (neat) 3083 (m), 3060 (m), 3026 (m), 3002 (m), 2933 (s), 2874 (m), 2848 (m), 2798 (s), 2755 (s), 2719 (m), 2693 (m), 2677 (m), 1601 (m), 1493 (s), 1465 (m), 1452 (s), 1392 (m), 1365 (m), 1341 (m), 1313 (m), 1263 (m), 1197 (m), 1145 (m), 1125 (m), 1068 (m), 1028 (m), 991 (m), 970 (m), 907 (m), 824 (m), 785 (m), 756 (m), 737 (s), 698 (s), 647 (m); MS (EI+, 70 eV) 251.2 (M+, 100), 174.1 (20), 160.1 (30), 91.1 (C7H7+, 65); TLC Rf 0.19 (98:1.8:0.2 CH2Cl2:MeOH:aq. NH3) [KMnO4]; Analysis C18H21N (251.37). Calcd: C, 86.01; H, 8.42; N, 5.57%. Found: C, 85.93; H, 8.50; N, 5.73%.</p><p> </p><!><p>Following General Procedure 1, 18c (>99:1 dr, 236 mg, 1.00 mmol, 1.0 equiv), 3-isopropoxyphenylmagnesium bromide (2.24 M in Et2O, 1.34 mL, 3.00 mmol, 3.0 equiv), Fe(acac)3 (70.6 mg, 0.20 mmol, 20 mol %), TMEDA (930 mg, 1.20 mL, 8.00 mmol, 8.0 equiv), and CPME (10.0 mL) were reacted to give a yellow oil (800 mg). Purification via flash column chromatography (40 g SiO2, 30 mm Ø, 95:5, hexane/toluene, ca. 5-mL fractions) gave a clear, colorless oil (240 mg). Further purification via flash column chromatography (C18-reversed phase silica gel, 20 × 160 mm, MeOH, ca. 2.5-mL fractions, loaded with minimal MeCN for solubility reasons) gave a clear, colorless oil (152 mg). Further purification via bulb-to-bulb distillation at reduced pressure (0.03 mmHg) gave 19c as a clear, colorless oil (143 mg, 62%, 98:2 dr). Data for 19c: bp 110 °C ABT (0.03 mmHg); 1H NMR (500 MHz, CDCl3) 7.22-7.15 (for both diastereoisomers: m, 1 H each), 6.82-6.66 (for both diastereoisomers: m, 3 H each), 4.60-4.49 (for both diastereoisomers: m, 1 H each), 3.22-3.16 (for minor diastereoisomer: m, 1 H), 2.70 (for major diastereoisomer: dd, J = 9.1, 5.6 Hz, 1 H), 1.99-1.91 (for minor diastereoisomer: m, 1 H), 1.79-1.72 (for major diastereoisomer: m, 1 H), 1.69-1.12 (for both diastereoisomers: m, 13 H each); 13C NMR (125 MHz, CDCl3) For major diastereoisomer: 158.1, 149.6, 129.3, 119.6, 115.4, 112.4, 69.8, 47.6, 43.1, 39.4, 37.0, 36.4, 30.9, 29.2, 22.4. For minor diastereoisomer (1 peak obscured in aliphatic region): 157.9, 145.7, 129.0, 120.8, 116.5, 112.6, 69.9, 46.3, 42.8, 40.8, 37.8, 34.5, 30.4, 23.3; IR (neat) 3026 (w), 2951 (s), 2870 (s), 1606 (s), 1580 (s), 1487 (s), 1454 (m), 1383 (m), 1371 (m), 1334 (m), 1310 (m), 1250 (m), 1181 (m), 1157 (m), 1136 (m), 1118 (s), 1000 (m), 991 (m), 953 (m), 874 (m), 839 (w), 818 (w), 776 (m), 721 (m), 697 (m); MS (EI+, 70 eV) 230.2 (M+, 43), 188.1 ([M–C3H6]+, 52), 120.1 (44), 108.1 (100), 91.1 (C7H7+, 12); TLC Rf 0.33 (90:10, hexane/toluene) [KMnO4]; Analysis C16H22O (230.35). Calcd: C, 83.43; H, 9.63%. Found: C, 83.53; H, 9.65%.</p><p> </p><!><p>Following General Procedure 1, 18d (276 mg, 1.00 mmol, 1.0 equiv), 3-isopropoxyphenylmagnesium bromide (2.24 M in Et2O, 1.34 mL, 3.00 mmol, 3.0 equiv), Fe(acac)3 (70.6 mg, 0.20 mmol, 20 mol %), TMEDA (930 mg, 1.20 mL, 8.00 mmol, 8.0 equiv), and CPME (10.0 mL) were reacted to give a yellow oil (827 mg). Purification via flash column chromatography (40 g SiO2, 30 mm Ø, 95:5, hexane/toluene, ca. 5-mL fractions) gave a clear, pale yellow oil (225 mg). Further purification via flash column chromatography (C18-reversed phase silica gel, 20 × 160 mm, MeOH, ca. 2.5-mL fractions, loaded with minimal ~1:1 MeCN:CH2Cl2 for solubility reasons) gave a clear, colorless oil (203 mg). Further purification via bulb-to-bulb distillation at reduced pressure (10–5 mmHg) gave 19d as a clear, colorless oil (179 mg, 66%). Data for 19d: bp 150 °C ABT (10–5 mmHg); 1H NMR (500 MHz, CDCl3) 7.25-7.20 (m, 1 H), 6.96-6.89 (m, 2 H), 6.74-6.69 (m, 1 H), 4.55 (hept, J = 6.1 Hz, 1 H), 2.99-2.95 (br s, 1 H), 2.48-2.42 (br s, 2 H), 2.04-1.83 (m, 7 H), 1.82-1.74 (m, 3 H), 1.59-1.52 (m, 2 H), 1.34 (d, J = 6.1 Hz, 6 H); 13C NMR (125 MHz, CDCl3) 158.2, 149.5, 129.2, 119.4, 115.5, 112.2, 69.9, 47.1, 39.4, 38.1, 32.3, 31.4, 28.3, 28.1, 22.4; IR (neat) 3077 (w), 3026 (w), 2974 (s), 2904 (s), 2848 (s), 1604 (s), 1578 (s), 1487 (s), 1467 (m), 1450 (s), 1382 (m), 1371 (m), 1354 (m), 1331 (m), 1289 (m), 1258 (s), 1180 (m), 1157 (m), 1136 (m), 1118 (s), 1069 (w), 1001 (m), 980 (m), 965 (m), 948 (m), 918 (w), 874 (m), 783 (m), 768 (m), 756 (m), 695 (m); MS (EI+, 70 eV) 270.2 (M+, 37), 228.2 ([M–C3H6]+, 100), 107.1 (11); TLC Rf 0.28 (90:10, hexane/toluene) [KMnO4]; Analysis C19H26O (270.41). Calcd: C, 84.39; H, 9.69%. Found: C, 84.53; H, 9.67%.</p><p> </p><!><p>Following General Procedure 1, 9 (274 mg, 1.00 mmol, 1.0 equiv), 3-isopropoxyphenylmagnesium bromide (2.24 M in Et2O, 1.34 mL, 3.00 mmol, 3.0 equiv), Fe(acac)3 (70.6 mg, 0.20 mmol, 20 mol %), TMEDA (930 mg, 1.20 mL, 8.00 mmol, 8.0 equiv), and CPME (10.0 mL) were reacted to give a yellow oil (812 mg). Purification via flash column chromatography (40 g SiO2, 30 mm Ø, 90:10, hexane/toluene, ca. 5-mL fractions) gave a clear, colorless oil (147 mg). Further purification via flash column chromatography (C18-reversed phase silica gel, 20 × 160 mm, MeOH, ca. 2.5-mL fractions, loaded with minimal MeCN for solubility reasons) gave a clear, colorless oil (140 mg). Further purification via bulb-to-bulb distillation at reduced pressure (10–5 mmHg) gave 19e as a clear, colorless oil (142 mg, 53%) which was contaminated with ~5% of a compound tentatively assigned as 3-isopropoxybiphenyl. Data for 19e: bp 125 °C ABT (10–5 mmHg); 1H NMR (500 MHz, CDCl3) 7.30-7.11 (m, 6 H), 6.82-6.71 (m, 3 H), 4.56 (hept, J = 6.1 Hz, 1 H), 2.74-2.64 (m, 1 H), 2.60-2.47 (m, 2 H), 1.99-1.83 (m, 2 H), 1.36 (d, J = 6.1 Hz, 6 H), 1.27 (d, J = 7.0 Hz, 3 H); 13C NMR (125 MHz, CDCl3) 158.3, 149.3, 142.9, 129.6, 128.7, 128.5, 125.9, 119.7, 115.4, 113.2, 69.9, 40.2, 39.9, 34.3, 22.7, 22.4; IR (neat) 3084 (w), 3062 (w), 3026 (m), 2975 (m), 2927 (m), 2870 (w), 1601 (m), 1582 (m), 1485 (m), 1454 (m), 1383 (m), 1372 (m), 1348 (w), 1334 (w), 1312 (w), 1286 (m), 1253 (m), 1180 (m), 1157 (m), 1137 (m), 1118 (m), 1030 (w), 999 (w), 973 (m), 874 (w), 777 (w), 748 (m), 699 (s); MS (EI+, 70 eV) 268.2 (M+, 53), 170.1 (21), 164.1 (46), 122.1 (100), 107.0 (22), 91.1 (C7H7+, 35), 77.0 (14); HRMS (ESI, TOF) calcd for C19H24O: 268.1827, found: 268.1828; TLC Rf 0.19 (90:10, hexane/toluene) [KMnO4].</p><p> </p><!><p>Following General Procedure 1, 18f (290 mg, 1.00 mmol, 1.0 equiv), PhMgBr (2.76 M in Et2O, 1.09 mL, 3.00 mmol, 3.0 equiv), Fe(acac)3 (70.6 mg, 0.20 mmol, 20 mol %), TMEDA (930 mg, 1.20 mL, 8.00 mmol, 8.0 equiv), and CPME (10.0 mL) were reacted to give an orange oil (455 mg). Purification via flash column chromatography (40 g SiO2, 30 mm Ø, 75:25, hexane/toluene, ca. 5-mL fractions) gave a clear, colorless oil (152 mg). Further purification via bulb-to-bulb distillation at reduced pressure (10–5 mmHg) gave 19f as a clear, colorless oil (142 mg, 63%). Data for 19f: bp 100 °C ABT (10–5 mmHg); 1H NMR (500 MHz, CDCl3) 7.31-7.25 (m, 2 H), 7.21-7.15 (m, 3 H), 7.01-6.97 (m, 2 H), 6.80-6.76 (2 H, m), 3.78 (s, 3 H), 3.00-2.91 (m, 1 H), 2.88 (dd, J = 13.4, 6.4 Hz, 1 H), 2.71 (dd, J = 13.4, 8.2 Hz, 1 H), 1.23 (d, J = 6.9 Hz, 3 H); 13C NMR (125 MHz, CDCl3) 158.0, 147.3, 133.2, 130.3, 128.5, 127.3, 126.2, 113.8, 55.4, 44.4, 42.3, 21.4; IR (neat) 3060 (m), 3026 (m), 2999 (m), 2958 (m), 2927 (m), 2833 (m), 1610 (m), 1583 (m), 1509 (s), 1493 (m), 1452 (m), 1374 (m), 1300 (m), 1246 (s), 1177 (m), 1111 (m), 1037 (m), 1013 (m), 817 (m), 783 (m), 761 (m), 699 (s); MS (EI+, 70 eV) 226.1 (M+, 46), 121.1 (C8H9O+, 100), 105.1 (23), 77.0 (24); TLC Rf 0.42 (70:30, hexane/toluene) [KMnO4]; Analysis C16H18O (226.31). Calcd: C, 84.91; H, 8.02%. Found: C, 84.68; H, 8.14%.</p><p> </p><!><p>An oven-dried, 25-mL, one-necked, round-bottomed flask was charged with 18g (256 mg, 1.00 mmol, 1.0 equiv) and Fe(acac)3 (70.6 mg, 0.20 mmol, 20 mol %) in a glove box then was sealed with a rubber septum and removed from the box. Outside of the glove box, a 25-mL Schlenk flask equipped with a stirrer bar, rubber septum and argon inlet was evacuated, flame-dried, left to cool under vacuum, and flushed three times with argon. TMEDA (930 mg, 1.20 mL, 8.00 mmol, 8.0 equiv) was added via syringe to the Schlenk flask and stirring was commenced. The round-bottomed flask containing 18g and Fe(acac)3 was charged with CPME (4.0 mL) then sonicated until homogeneous. The clear red solution was then transferred via cannula to the Schlenk flask holding the TMEDA, and the residual material was rinsed across with further portions of CPME (6.0 mL). PhMgBr (2.78 M in Et2O, 1.08 mL, 3.00 mmol, 3.0 equiv) was then added by syringe over ca. 20 sec. During addition, the color of the solution changed from red to pale yellow to brown, but remained clear throughout, and no visible deposits were formed on the edges of the flask. After stirring for 18 h at rt, H2O (10 mL) was added in one portion and the mixture was filtered through a pad of Celite (5 g) in a 40 mm Ø, porosity 3, sintered funnel under House vacuum. H2O (2 × 10 mL) and EtOAc (3 × 10 mL) were used to rinse any residual material though the Celite pad. The filtrate was transferred to a separatory funnel and the layers were separated. The aqueous layer was extracted with EtOAc (2 × 20 mL) and the combined organic layers were dried (MgSO4), filtered and concentrated in vacuo (50 °C, ca. 5 mmHg) to give a red oil (408 mg). Purification via flash column chromatography (40 g SiO2, 30 mm Ø, 95:5, hexane/EtOAc, ca. 5-mL fractions) gave a clear, yellow oil (108 mg). Further purification via preparative, radial, centrifugally accelerated, thin-layer chromatography on a Harrison Chromatotron (4 mm SiO2 plate, 25:75, hexane:CH2Cl2, ca. 5-mL fractions) gave a clear, yellow oil (102 mg). Further purification via bulb-to-bulb distillation at reduced pressure (0.03 mmHg) gave 19g as a clear, colorless oil (98.9 mg, 51%). The 1H NMR spectroscopic data matched that for an alternative preparation of (R)-19g.116 Data for 19g: bp 75 °C ABT (0.03 mmHg).</p><p> </p><!><p>An oven-dried, 25-mL, one-necked, round-bottomed flask was charged with 18h (310 mg, 1.00 mmol, 1.0 equiv) and Fe(acac)3 (70.6 mg, 0.20 mmol, 20 mol %) in a glove box then was sealed with a rubber septum and removed from the box. Outside of the glove box, a 25-mL Schlenk flask equipped with a stirrer bar, rubber septum and argon inlet was evacuated, flame-dried, left to cool under vacuum, and flushed three times with argon. TMEDA (930 mg, 1.20 mL, 8.00 mmol, 8.0 equiv) was added via syringe to the Schlenk flask and stirring was commenced. The round-bottomed flask containing 18h and Fe(acac)3 was charged with CPME (4.0 mL) then sonicated until homogeneous. The clear red solution was then transferred via cannula to the Schlenk flask holding the TMEDA, and the residual material was rinsed across with further portions of CPME (6.0 mL). PhMgBr (2.78 M in Et2O, 1.08 mL, 3.00 mmol, 3.0 equiv) was then added by syringe over ca. 30 sec. During addition, the color of the solution changed from red to pale yellow to brown, but remained clear throughout, and no visible deposits were formed on the edges of the flask. After stirring for 18 h at rt, H2O (10 mL) was added in one portion and the mixture was filtered through a pad of Celite (5 g) in a 40 mm Ø, porosity 3, sintered funnel under House vacuum. H2O (2 × 10 mL) and EtOAc (3 × 10 mL) were used to rinse any residual material though the Celite pad. The filtrate was transferred to a separatory funnel and the layers were separated. The aqueous layer was extracted with EtOAc (2 × 20 mL) and the combined organic layers were dried (MgSO4), filtered and concentrated in vacuo (50 °C, ca. 5 mmHg) to give a red oil (426 mg). Purification via flash column chromatography (40 g SiO2, 30 mm Ø, 95:5, hexane/EtOAc, ca. 5-mL fractions) gave a clear, pale yellow oil (214 mg). Further purification via preparative, radial, centrifugally accelerated, thin-layer chromatography on a Harrison Chromatotron (4 mm SiO2 plate, 25:75, hexane:CH2Cl2, ca. 5-mL fractions) gave a clear, pale yellow oil (185 mg). Further purification via bulb-to-bulb distillation at reduced pressure (10–5 mmHg) gave 19h as a clear, colorless oil (169 mg, 69%). Data for 19h: bp 110 °C ABT (10–5 mmHg); 1H NMR (500 MHz, CDCl3) 7.32-7.27 (m, 2 H), 7.22-7.14 (m, 3 H), 3.98-3.89 (m, 4 H), 2.54-2.45 (m, 1 H), 1.96-1.89 (m, 1 H), 1.83-1.76 (m, 1 H), 1.71-1.63 (m, 1 H), 1.54 (td, J = 13.2, 4.2 Hz, 1 H), 1.50-1.38 (m, 3 H), 1.37-1.30 (m, 1 H), 1.27 (d, J = 7.0 Hz, 3 H), 1.24-1.13 (m, 1 H); 13C NMR (125 MHz, CDCl3) 147.2, 128.4, 127.7, 126.1, 109.2, 64.4, 64.4, 45.4, 43.1, 34.8, 34.8, 29.0, 27.8, 19.4; IR (neat) 3082 (w), 3060 (w), 3025 (m), 2943 (s), 2876 (s), 1603 (w), 1582 (w), 1493 (s), 1449 (s), 1375 (s), 1359 (m), 1337 (m), 1285 (m), 1256 (w), 1224 (m), 1199 (m), 1175 (m), 1153 (s), 1095 (s), 1035 (s), 1003 (w), 996 (m), 932 (s), 904 (s), 816 (w), 761 (s), 702 (s), 663 (w), 575 (w); MS (EI+, 70 eV) 246.1 (M+, 18), 105.1 (11), 99.0 (100), 86.0 (12); HRMS (El+, double focusing sector field) calcd for C16H22O2: 246.1620, found: 246.1623; TLC Rf 0.36 (90:10, hexane/EtOAc) [KMnO4].</p><p> </p><!><p>An oven-dried, 20-mL scintillation vial was charged with 18i (343 mg, 1.00 mmol, 1.0 equiv) and Fe(acac)3 (70.6 mg, 0.20 mmol, 20 mol %) in a glove box then was sealed with a rubber septum and removed from the box. Outside of the glove box, a 25-mL Schlenk flask equipped with a stirrer bar, rubber septum and argon inlet was evacuated, flame-dried, left to cool under vacuum, and flushed three times with argon. TMEDA (930 mg, 1.20 mL, 8.00 mmol, 8.0 equiv) was added via syringe to the Schlenk flask and stirring was commenced. The vial containing 18i and Fe(acac)3 was charged with CPME (4.0 mL) then sonicated until homogeneous. The clear red solution was then transferred via cannula to the Schlenk flask holding the TMEDA, and the residual material was rinsed across with further portions of CPME (6.0 mL). PhMgBr (2.76 M in Et2O, 1.09 mL, 3.00 mmol, 3.0 equiv) was then added by syringe over ca. 20 sec. During addition, the color of the solution changed from red to pale yellow to brown, but remained clear throughout, and no visible deposits were formed on the edges of the flask. After stirring for 18 h at rt, 1 M aq. HCl (10 mL) was added in one portion and the mixture was filtered through a pad of Celite (5 g) in a 40 mm Ø, porosity 3, sintered funnel under House vacuum. EtOAc (2 × 5 mL) was used to rinse any residual material though the Celite pad. The filtrate was brought to pH 9 by addition of 5% aq. NaOH and was then transferred to a separatory funnel and the layers were separated. The aqueous layer was extracted with EtOAc (2 × 10 mL) and the combined organic extracts were dried (MgSO4), filtered and concentrated in vacuo (50 °C, ca. 5 mmHg) to give a dark red oil (393 mg). Purification via flash column chromatography (40 g SiO2, 30 mm Ø, 98:1.8:0.2 CH2Cl2:MeOH:aq. NH3, ca. 5-mL fractions) gave a clear, orange oil (287 mg). Further purification via flash column chromatography (C18-reversed phase silica gel, 20 × 160 mm, i-PrOH, ca. 2.5-mL fractions, loaded with minimal MeCN for solubility reasons) gave 19i contaminated with traces of terminal and internal alkene by-products as a clear, yellow oil (256 mg). A dihydroxylation protocol was next performed to convert the alkene impurities to more readily separable diols. The residue was dissolved in acetone (2.0 mL) in a 20-mL scintillation vial and N-methylmorpholine N-oxide (164 mg, 1.20 mmol) and osmium tetraoxide (4% in H2O, 0.13 mL, 0.02 mmol) were added sequentially. The resultant dark yellow-orange, biphasic mixture was sealed with a screw cap and stirred at rt (under air) for 45 h. Sat. aq. Na2SO3 (0.5 mL) was then added and the resultant mixture was stirred vigorously at rt for 30 min. H2O (10 mL) was added and the mixture was extracted with EtOAc (3 × 10 mL). The combined organic extracts were dried (MgSO4), filtered through a pad of Celite (4 g) in a 40 mm Ø, porosity 3, sintered funnel under House vacuum, and then concentrated in vacuo (50 °C, ca. 5 mmHg) to give a black oil (271 mg). Purification via flash column chromatography (C18-reversed phase silica gel, 20 × 160 mm, i-PrOH, ca. 2.5-mL fractions, loaded with minimal MeCN for solubility reasons) gave an orange-brown oil (158 mg). Further purification via flash column chromatography (20 g SiO2, 20 mm Ø, 98:1.8:0.2 CH2Cl2:MeOH:aq. NH3, ca. 5-mL fractions) gave a clear, yellow oil (130 mg). Further purification via bulb-to-bulb distillation at reduced pressure (10–5 mmHg) gave 19i as a clear, colorless oil (121 mg, 43%). Data for 19i: bp 160 °C ABT (10–5 mmHg); 1H NMR (500 MHz, CDCl3) 7.32-7.21 (m, 7 H), 7.19-7.10 (m, 3 H), 3.52-3.43 (br s, 2 H), 2.94 (d, J = 11.5 Hz, 1 H), 2.81 (d, J = 11.5 Hz, 1 H), 2.49-2.40 (m, 1 H), 1.99-1.74 (m, 3 H), 1.43-1.09 (m, 4 H) overlapping 1.24 (d, J = 7.0 Hz, 3 H); 13C NMR (125 MHz, CDCl3) 146.8, 138.7, 129.5, 128.4, 128.3, 127.8, 127.1, 126.1, 63.7, 54.3, 54.2, 45.8, 42.7, 31.1, 30.4, 19.2; IR (neat) 3082 (m), 3060 (m), 3026 (s), 3000 (m), 2937 (s), 2906 (s), 2875 (s), 2849 (s), 2799 (s), 2753 (s), 2723 (m), 2692 (m), 2677 (m), 1601 (m), 1493 (s), 1452 (s), 1366 (s), 1342 (m), 1313 (m), 1298 (m), 1274 (m), 1262 (m), 1148 (s), 1120 (m), 1074 (m), 1049 (m), 1028 (m), 1012 (m), 996 (m), 984 (m), 970 (m), 907 (m), 845 (m), 793 (m), 761 (s), 737 (s), 699 (s); MS (EI+, 70 eV) 279.2 (M+, 65), 202.2 (14), 188.1 (14), 174.1 (21), 172.1 (14), 159.1 (20), 146.1 (12), 120.1 (20), 105.1 (21), 91.1 (C7H7+, 100), 77.0 (10); TLC Rf 0.14 (98:1.8:0.2 CH2Cl2:MeOH:aq. NH3) [KMnO4]; Analysis C20H25N (279.42). Calcd: C, 85.97; H, 9.02; N, 5.01%. Found: C, 85.93; H, 8.96; N, 5.17%.</p><p> </p><!><p>An oven-dried, 25-mL, one-necked, round-bottomed flask was charged with 18j (225 mg, 1.00 mmol, 1.0 equiv) and Fe(acac)3 (70.6 mg, 0.20 mmol, 20 mol %) in a glove box then was sealed with a rubber septum and removed from the box. Outside of the glove box, a 25-mL Schlenk flask equipped with a stirrer bar, rubber septum and argon inlet was evacuated, flame-dried, left to cool under vacuum, and flushed three times with argon. TMEDA (930 mg, 1.20 mL, 8.00 mmol, 8.0 equiv) was added via syringe to the Schlenk flask and stirring was commenced. The flask containing 18j and Fe(acac)3 was charged with CPME (4.0 mL) then sonicated until homogeneous. The clear red solution was then transferred via cannula to the Schlenk flask holding the TMEDA, and the residual material was rinsed across with further portions of CPME (6.0 mL). PhMgBr (2.78 M in Et2O, 1.08 mL, 3.00 mmol, 3.0 equiv) was then added by syringe over ca. 20 sec. During addition, the color of the solution changed from red to pale yellow to brown, but remained clear throughout, and no visible deposits were formed on the edges of the flask. After stirring for 18 h at rt, 1 M aq. HCl (10 mL) was added in one portion and the mixture was filtered through a pad of Celite (5 g) in a 40 mm Ø, porosity 3, sintered funnel under House vacuum. EtOAc (2 × 5 mL) was used to rinse any residual material though the Celite pad. The filtrate was brought to pH 9 by addition of 2 M aq. NaOH and was then transferred to a separatory funnel and the layers were separated. The aqueous layer was extracted with EtOAc (2 × 10 mL) and the combined organic extracts were dried (MgSO4), filtered and concentrated in vacuo (50 °C, ca. 5 mmHg) to give a red oil (305 mg). Purification via flash column chromatography (40 g SiO2, 30 mm Ø, 92:7.2:0.8 CH2Cl2:MeOH:aq. NH3, ca. 5-mL fractions) gave a clear, orange oil (53.9 mg). Further purification via bulb-to-bulb distillation at reduced pressure (0.03 mmHg) gave 19j as a clear, colorless oil (40.0 mg, 25%). The 1H NMR spectroscopic data matched that for alternative preparations.117 Data for 19j: bp 75 °C ABT (0.03 mmHg).</p><p> </p><!><p>Following General Procedure 1, 18k (>99:1 dr, 302 mg, 1.00 mmol, 1.0 equiv), PhMgBr (2.76 M in Et2O, 1.09 mL, 3.00 mmol, 3.0 equiv), Fe(acac)3 (70.6 mg, 0.20 mmol, 20 mol %), TMEDA (930 mg, 1.20 mL, 8.00 mmol, 8.0 equiv), and CPME (10.0 mL) were reacted to give a ~4:1 mixture of 18k:20 as an orange oil (484 mg). Purification via flash column chromatography (40 g SiO2, 30 mm Ø, 75:25, hexane/EtOAc, ca. 5-mL fractions) gave an impure sample of 20 as a pale yellow oil (3.7 mg) and an impure sample of 18k as a white solid/yellow oil (181 mg). Data for 20: 1H NMR (500 MHz, CDCl3) (selected peaks) 6.75 (s, 1 H), 3.60 (t, J = 6.4 Hz, 2 H), 2.84-2.78 (m, 2 H), 1.72-1.65 (m, 2 H); MS (EI+, 70 eV) 238.1 (M+, 76), 205.1 (35), 194.1 (82), 178.1 (58), 165.1 (34), 152.1 (16), 147.1 (100), 105.0 (45); HRMS (EI+, TOF) calcd for C17H18O: 238.1358, found: 238.1359; TLC Rf 0.32 (70:30 hexane:EtOAc) [KMnO4].</p><p> </p><!><p>A 12 × 75 mm test tube, equipped with a stirrer bar, was oven-dried, transferred to a glove box, charged with 21 (28.2 mg, 0.10 mmol, 1.0 equiv) and Fe(acac)3 (7.1 mg, 0.02 mmol, 20 mol %), and then sealed with a rubber septum and electrical tabe before being removed from the glove box. Outside the glove box, the test tube was charged with TMEDA (93.0 mg, 120 μL, 0.80 mmol, 8.0 equiv), tetradecane (9.9 mg, 13 μL, 0.05 mmol, 0.5 equiv), and CPME (1.0 mL) via syringe, and stirring was commenced. PhMgBr (3.06 M solution in Et2O, 98 μL, 0.30 mmol, 3.0 equiv) was added via syringe, causing a color change from red to pale yellow to brown/black. After 18 h, the reaction was quenched by addition of MeOH (0.3 mL) via syringe. A 50 μL aliquot of the organic layer was then transferred via syringe to a fresh GC vial and diluted with EtOAc (1 mL) for analysis. According to GC analysis, the reaction had proceeded to 80% conversion to give 22 (27%), 23 (≤5%), and 24 (≤5%). No other products were detected under the conditions of the run.</p><p> </p><!><p>Following General Procedure 1, 25 (288 mg, 1.00 mmol, 1.0 equiv), PhMgBr (2.60 M in Et2O, 1.15 mL, 3.00 mmol, 3.0 equiv), Fe(acac)3 (70.6 mg, 0.20 mmol, 20 mol %), TMEDA (930 mg, 1.20 mL, 8.00 mmol, 8.0 equiv), and CPME (10.0 mL) were reacted to give a 87:13 mixture of 26:27 in addition to unreacted 25 (~30%) as an orange oil (456 mg). Purification via flash column chromatography (40 g SiO2, 30 mm Ø, hexane, ca. 5-mL fractions) gave an 80:20 mixture of 26:27 contaminated with hexane and biphenyl as a pale yellow oil (40.4 mg, ~28%). The 1H NMR spectroscopic data for the 26118 and 27119 present in the mixture matched that for alternative preparations.</p><!><p> </p><!><p>Bromine (1.26 g, 0.40 mL, 7.83 mmol, 1.2 equiv) was added dropwise via syringe to a stirred suspension of triphenylphosphine (2.07 g, 7.83 mmol, 1.2 equiv) in CH2Cl2 (23 mL) in a 1 L, single-necked, round-bottomed flask equipped with a stirrer bar and cooled in an ice/water bath (open to air). The flask was then sealed with a rubber septum and purged with argon via an inlet needle. After stirring the resultant pale yellow suspension for 15 min, a solution of (R)-28 (99% ee, 1.00 g, 6.52 mmol, 1.0 equiv) and imidazole (538 mg, 7.83 mmol, 1.2 equiv) in CH2Cl2 (10 mL) was added via cannula over ca. 5 min. The cooling bath was removed and the reaction mixture was allowed to warm to rt over 17 h. The mixture was then filtered through a 40 mm Ø, porosity 3, sintered funnel under House vacuum and carefully concentrated in vacuo to leave a yellow oil residue (i.e. avoiding precipitating the phosphorus-containing residues at this point). A stirrer bar was added to the residue and a wide-neck plastic funnel was added to the neck of the flask, and rapid stirring was commenced. Pentane (33 mL) was quickly added in one portion to precipitate the phosphorus-containing residues as a fine white solid. The mixture was rinsed through a pad of SiO2 (7 g) in a 40 mm Ø, porosity 3, sintered funnel under House vacuum using pentane (3 × 15 mL) and the filtrate was concentrated in vacuo (50 °C, ca. 5 mmHg) to give a clear, colorless oil (1.31 g). Purification via bulb-to-bulb distillation at reduced pressure (0.05 mmHg) gave (S)-29 as a clear, colorless oil (1.25 g, 90%).120 The spectral data matched that for (rac)-29. Data for (S)-29: bp 90 °C ABT (0.05 mmHg); Opt. Rot. [α]D25=+77.4 (c = 0.87, CHCl3).</p><p> </p><!><p>A 15-mL, one-necked, round-bottomed flask equipped with a stirrer bar, water-jacketed reflux condenser, and argon inlet was charged with (S)-29 (213 mg, 1.00 mmol, 1.0 equiv), 2-mercaptopyridine (113 mg, 1.00 mmol, 1.0 equiv), potassium carbonate (276 mg, 2.00 mmol, 2.0 equiv), and acetone (5.0 mL), and stirring was commenced. The resultant mixture was heated at reflux for 1 h, and was then allowed to cool to rt. The mixture was filtered through a 40 mm Ø, porosity 3, sintered funnel under House vacuum and concentrated in vacuo (50 °C, ca. 5 mmHg) to give a yellow oil (305 mg). Purification via flash column chromatography (10 g SiO2, 20 mm Ø, 95:5, hexane/EtOAc, ca. 3 mL fractions) gave (R)-3d as a clear, colorless oil (235 mg, 96%, 98.8:1.2 er). The 1H NMR spectroscopic data matched that for (rac)-3d. Data for (R)-3d: Opt. Rot. [α]D25=+38.4 (c = 1.16, CHCl3); SFC (R)-3d, tR 5.7 min (98.8%); (S)-3d, tR 6.5 min (1.2%) (Chiralpak AD, 5% MeOH in CO2, 2.0 mL min–1, 220 nm, 40 °C).</p><p> </p><!><p>A 4.0-mL screw-top vial containing (R)-3d (98.8:1.2 er, 72.9 mg, 0.30 mmol, 1.0 equiv) was taken into a glove box and charged with Fe(acac)3 (31.7 mg, 0.09 mmol, 20 mol %) then was sealed with a rubber septum and removed from the box. Outside of the glove box, a 10-mL Schlenk flask equipped with a stirrer bar, rubber septum and argon inlet was evacuated, flame-dried, left to cool under vacuum, and flushed three times with argon. The vial containing (R)-3d and Fe(acac)3 was charged with CPME (1.0 mL) then sonicated until homogeneous. The clear red solution was then transferred via cannula to the Schlenk flask, and the residual material was rinsed across with further portions of CPME (2.0 mL). 4-Methoxyphenylmagnesium bromide (2.17 M in Et2O, 552 μL, 1.20 mmol, 4.0 equiv) was then added by syringe over ca. 1 min. During addition, the color of the solution changed from red to opaque black, and small clusters of black solid could be seen forming during addition. Visible black deposits were also visible at the top of the solution. After stirring for 18 h at rt, 1 M aq. HCl (3 mL) was added in one portion and the mixture was filtered through a pad of Celite (5 g) in a 40 mm Ø, porosity 3, sintered funnel under House vacuum. EtOAc (2 × 5 mL) was used to rinse any residual material though the Celite pad. The filtrate was transferred to a separatory funnel and the layers were separated. The organic layer was washed with 1 M aq. HCl (2 × 3 mL) and the combined aqueous layers were extracted with EtOAc (2 × 5 mL). The combined organic extracts were then dried (MgSO4), filtered and concentrated in vacuo (50 °C, ca. 5 mmHg) to give a pale orange residue comprising mainly a white solid (236 mg). Purification via flash column chromatography (20 g SiO2, 20 mm Ø, 80:20, hexane/toluene, ca. 5-mL fractions) gave a colorless oil (54.2 mg). Further purification via flash column chromatography (C18-reversed phase silica gel, 20 × 160 mm, 98:2 MeOH/H2O, ca. 2 mL fractions, loaded with minimal MeCN for solubility reasons) gave 8 as a clear, colorless oil (38.0 mg, 53%, 50.5:49.5 er). The 1H NMR spectroscopic data matched that for 8 prepared from (rac)-3d. Data for 8: SFC first enantiomer, tR 11.7 min (49.5%); second enantiomer, tR 12.8 min (50.5%) (Chiralcel OB, 5% MeOH in CO2, 1.0 mL min–1, 220 nm, 40 °C).</p><p> </p><!><p>A 15-mL, one-necked, round-bottomed flask equipped with a stirrer bar, water-jacketed reflux condenser, and argon inlet was charged with (S)-29 (213 mg, 1.00 mmol, 1.0 equiv), thiophenol (114 mg, 106 μL, 1.00 mmol, 1.0 equiv), potassium carbonate (276 mg, 2.00 mmol, 2.0 equiv), and acetone (5.0 mL), and stirring was commenced. The resultant mixture was heated at reflux for 19 h, and was then allowed to cool to rt. The mixture was filtered through a 40 mm Ø, porosity 3, sintered funnel under House vacuum and concentrated in vacuo (50 °C, ca. 5 mmHg) to give a colorless oil (246 mg). Purification via preparative, radial, centrifugally accelerated, thin-layer chromatography on a Harrison Chromatotron (1 mm SiO2 plate, 90:10, hexane/toluene, ca. 5-mL fractions) gave (R)-3a as a clear, colorless oil (202 mg, 83%, 98.6:1.3 er). The 1H NMR spectroscopic data matched that for (rac)-3a. Data for (R)-3a: Opt. Rot. [α]D25=−2.1 (c = 10.5, CHCl3); SFC (R)-3a, tR 6.5 min (98.6%); (S)-3a, tR 7.4 min (1.3%) (Chiralpak AD, 1.5% MeOH in CO2, 2.0 mL min–1, 220 nm, 40 °C).</p><p> </p><!><p>A 10-mL, one-necked, round-bottomed flask equipped with a stirrer bar and rubber septum was charged with (R)-3a (195 mg, 0.81 mmol, 1.0 equiv), ammonium molybdate tetrahydrate (99.6 mg, 0.08 mmol, 10 mol %), and MeOH (2.5 mL), and stirring was commenced. The mixture was cooled in an ice/water bath then hydrogen peroxide (30% in H2O, 366 mg, 329 μL, 3.22 mmol, 4.0 equiv) was added dropwise via syringe over ca. 2 min. The resultant turbid, pale yellow mixture was stirred in the ice/water bath for 40 min, then allowed to warm to rt over 1 h, during which time the yellow color intensified. The mixture was then cooled in an ice/water bath and sat. aq. Na2SO3 (1.5 mL) was added dropwise via syringe over ca. 2 min. Starch-iodide paper was used to confirm that no oxidant remained. EtOAc (10 mL) and H2O (10 mL) were then added and the layers were separated. The aqueous layer was extracted with EtOAc (2 × 10 mL) and the combined organic extracts were dried (MgSO4), filtered, and concentrated in vacuo (50 °C, ca. 5 mmHg) to give a cloudy, colorless syrup (0.2 g). Purification via bulb-to-bulb distillation at reduced pressure (10–5 mmHg) gave (R)-9 as a clear, colorless syrup (204 mg, 92%, >99.5:0.5 er). The 1H NMR spectroscopic data matched that for (rac)-9. Data for (R)-9: bp 170 °C ABT (10–5 mmHg); Opt. Rot. [α]D25=+8.7 (c = 1.86, CHCl3); SFC (R)-9, tR 10.2 min (>99.5%); (S)-9, tR 13.1 min (<0.5%) (Chiralcel OB, 7.5% MeOH in CO2, 2.0 mL min–1, 220 nm, 40 °C).</p><p> </p><!><p>Following General Procedure 1, (R)-9 (>99.5:0.5 er, 173 mg, 0.63 mmol, 1.0 equiv), 3-isopropoxyphenylmagnesium bromide (2.24 M in Et2O, 842 μL, 1.89 mmol, 3.0 equiv), Fe(acac)3 (44.4 mg, 0.13 mmol, 20 mol %), TMEDA (585 mg, 755 μL, 5.03 mmol, 8.0 equiv), and CPME (6.3 mL) were reacted to give a yellow oil (562 mg). Purification via flash column chromatography (30 g SiO2, 30 mm Ø, 90:10, hexane/toluene, ca. 5-mL fractions) gave a clear, colorless oil (105 mg). Further purification via flash column chromatography (C18-reversed phase silica gel, 20 × 160 mm, MeOH, ca. 2.5-mL fractions, loaded with minimal MeCN for solubility reasons) gave a clear, colorless oil (91.7 mg). Further purification via bulb-to-bulb distillation at reduced pressure (10–5 mmHg) gave 19e as a clear, colorless oil (92.4 mg, 55%, 50.7:49.3 er) which was contaminated with ~5% of a compound tentatively assigned as 3-isopropoxybiphenyl. The 1H NMR spectroscopic data and boiling point matched that for 19e prepared from (rac)-9. Data for 19e: bp 125 °C ABT (10–5 mmHg); SFC first enantiomer, tR 7.5 min (49.3%); second enantiomer, tR 7.9 min (50.7%) (Chiralcel OD, 5% MeOH in CO2, 2.0 mL min–1, 220 nm, 40 °C).</p><!><p>Grignard reagents 15a-h,i,k-m,q were prepared from the corresponding aryl bromides and magnesium turnings in Et2O (see "Representative Procedure 1" below). Grignard reagents 15j,n-p,v could not be prepared directly using magnesium turnings in Et2O, and were instead prepared from the corresponding aryl/alkenyl bromides by lithium-bromine exchange with t-BuLi followed by transmetalation with MgBr2 (see "Representative Procedure 2" below). Phenylacetylenylmagnesium bromide 15w was prepared by deprotonation of phenylacetylene with EtMgBr in Et2O. Grignard reagents 15r-u were commercially available as solutions in Et2O from Aldrich and were used as received. Titration of the Grignard reagents was carried out using the protocol reported by Watson and Eastham.121 THF was typically added as a co-solvent in these titrations to ensure homogeneity and, in some cases, to give a stronger color to the solution than Et2O alone.</p><p> </p><!><p>An oven-dried, 50-mL, three-necked, round-bottomed flask equipped with a stirrer bar, an oven-dried water-jacketed reflux condenser, two rubber septa, and an argon inlet (at the top of the condenser) was assembled under a flow of argon and charged sequentially with magnesium turnings (729 mg, 30.0 mmol, 1.2 equiv), Et2O (2.0 mL), and a few crystals of iodine, and stirring was commenced. A small portion of neat 4-bromoanisole (from 4.72 g, 3.16 mL, 25.0 mmol, 1.0 equiv) was then added via cannula from an oven-dried, 25-mL, single-necked, round-bottomed flask under argon. Once the reaction had initiated (signified by decoloration and bubbling), Et2O (8.0 mL) was added to the 25-mL flask containing the 4-bromoanisole and the resultant solution was added dropwise via cannula over ca. 20 min to the reaction. The mixture was then heated at reflux for 1 h, then stirring was ceased and the mixture was allowed to cool to rt. The brown, supernatant solution (~9 mL) was then transferred via cannula to a 50-mL, plastic centrifuge tube capped with an inverted rubber septum, under argon. The solution was then centrifuged at 3220 rpm for 10 min. The clear, dark yellow-brown, supernatant solution was then transferred via cannula to an oven-dried, 25-mL Schlenk flask under argon. Based on the titration protocol reported by Watson and Eastham,121 a 350 μL aliquot of the solution of 4-methoxyphenylmagnesium bromide was added to a stirred solution of 1,10-phenanthroline (ca. 1-2 mg) in 2:1 Et2O:THF (3.0 mL) in an oven-dried, 25-mL, three-necked, round-bottomed flask under argon. The resultant deep burgandy solution was titrated against s-BuOH (1.00 M solution in xylenes), with the end point indicated by a sudden color change from deep burgandy to clear yellow. The solution of 4-methoxyphenylmagnesium bromide was 2.17 M.</p><p> </p><!><p>A 100-mL Schlenk flask (marked at 8 mL volume) was equipped with a stirrer bar and and water-jacketed reflux condenser and oven-dried, then assembled under a flow of argon (via an inlet at the top of the condenser) and a septum was placed on the remaining neck of the Schlenk, along with an exit needle. Magnesium turnings (217 mg, 8.91 mmol, 1.1 equiv) were then added against a backflow of argon, and the apparatus was allowed to cool to rt, at which point the exit needle was removed and the argon flow was reduced. The flask was charged sequentially with benzene (2.2 mL) and Et2O (6.7 mL) then 1,2-dibromoethane (1.61 g, 743 μL, 8.51 mmol, 1.05 equiv) was added dropwise to the flask via syringe over ca. 20 min (with cooling in an ice/water bath as necessary to prevent thermal runaway). Once addition was complete, the ca. 1 M solution was stirred for a further 30 min and then left to stand at rt (the solution was clear and colorless aside from residual magnesium). Meanwhile, a 100-mL, single-necked, round-bottomed flask equipped with a stirrer bar and rubber septum was flame-dried whilst being purged with argon via an inlet and exit needle. Once the flask had cooled to rt, it was charged with a solution of 3-bromothiophene (1.32 g, 759 μL, 8.10 mmol, 1.0 equiv) in Et2O (16.0 mL), and then cooled to in a dry ice/acetone bath, with stirring. t-BuLi (1.62±0.03 M in pentanes, 10.0 mL, 16.2 mmol, 2.0 equiv) was transferred to a flame-dried, 10-mL volumetric flask under argon via syringe and then added dropwise via cannula to the solution of 3-bromothiophene over ca. 10 min. Once added, the mixture was stirred in the dry ice/acetone bath for a further 30 min. The clear, colorless solution of 3-thienyllithium was then removed from the cooling bath and added dropwise via cannula to the previously prepared solution of MgBr2 in Et2O/benzene at rt. The resultant homogeneous solution was stirred in a hot water bath under a strong argon flow to reduce the solvent volume to ca. 8 mL (as marked on the Schlenk flask). On doing so, the lithium salts precipitated as a fine white solid and the supernatant, pale yellow solution was transferred via syringe to an oven-dried, 25-mL Schlenk flask under argon. Based on the titration protocol reported by Watson and Eastham, a 500 μL aliquot of the solution of 15p was added to a stirred solution of 1,10-phenanthroline (ca. 1-2 mg) in 1:1 CPME/THF (6.0 mL) in an oven-dried, 25-mL, three-necked, round-bottomed flask under argon. The resultant red-orange solution was titrated against s-BuOH (1.00 M solution in xylenes), with the end point indicated by a sudden color change from red-orange to clear yellow. The solution of 15p was 0.95 M.</p>

PubMed Author Manuscript

HIV-1 subtype influences susceptibility and response to monotherapy with the protease inhibitor lopinavir/ritonavir

ObjectivePI susceptibility results from a complex interplay between protease and Gag proteins, with Gag showing wide variation across HIV-1 subtypes. We explored the impact of pre-treatment susceptibility on the outcome of lopinavir/ritonavir monotherapy.MethodsTreatment-naive individuals who experienced lopinavir/ritonavir monotherapy failure from the MONARK study were matched (by subtype, viral load and baseline CD4 count) with those who achieved virological response (‘successes’). Successes were defined by viral load <400 copies/mL after week 24 and <50 copies/mL from week 48 to week 96. Full-length Gag–protease was amplified from patient samples for in vitro phenotypic susceptibility testing, with susceptibility expressed as fold change (FC) relative to a subtype B reference strain.ResultsBaseline lopinavir susceptibility was lower in viral failures compared with viral successes, but the differences were not statistically significant (median lopinavir susceptibility: 4.4 versus 8.5, respectively, P = 0.17). Among CRF02_AG/G patients, there was a significant difference in lopinavir susceptibility between the two groups (7.1 versus 10.4, P = 0.047), while in subtype B the difference was not significant (2.7 versus 3.4, P = 0.13). Subtype CRF02_AG/G viruses had a median lopinavir FC of 8.7 compared with 3.1 for subtype B (P = 0.001).ConclusionsWe report an association between reduced PI susceptibility (using full-length Gag–protease sequences) at baseline and subsequent virological failure on lopinavir/ritonavir monotherapy in antiretroviral-naive patients harbouring subtype CRF02_AG/G viruses. We speculate that this may be important in the context of suboptimal adherence in determining viral failure.

hiv-1_subtype_influences_susceptibility_and_response_to_monotherapy_with_the_protease_inhibitor_lopi

Introduction<!><!>Amplification of full-length Gag–protease<!>PI susceptibility and infectivity assays<!>Statistical analysis<!>Results<!><!>Discussion<!>Funding<!>Transparency declarations<!>Supplementary data

<p>Efficacious, long-term HAART treatment entails both a high financial cost and a risk of significant side effects, and exploration of alternative treatment regimens is necessary. Simplification strategies have been investigated to reduce the number of antiretroviral drugs required as part of treatment regimens without compromising treatment efficacy. Boosted PI (bPI) monotherapy has been studied in a small number of randomized trials, with poorer outcomes compared with standard triple therapy when used as initial therapy in the MONARK study1,2 or as second-line therapy after an NNRTI first-line failure in resource-limited settings.3,4 When used as a maintenance regimen in patients with a substantial period of viral suppression, bPI monotherapy may be considered as a treatment option to reduce exposure to antiretrovirals and NRTI-associated toxicities with the possibility of re-intensification when needed.5–8</p><p>Based on an estimated 10 million HIV-1-infected patients on antiretroviral therapy9 and a virological failure rate of 20%,10 around 2 million patients worldwide qualify for PI-based second-line treatment, most likely lopinavir boosted with ritonavir. Widespread use of this class of drug raises some concerns, given our poor understanding of virological failure. Indeed, the majority of patients experiencing failure of first-line combined ART with ritonavir-boosted PIs do so without evidence of major PI-associated drug resistance by standard methods.11,12 Studies have provided evidence for the role of Gag in PI susceptibility13–15 and have been reviewed.16 In addition, the inclusion of full-length patient-derived Gag alongside its co-evolved protease in in vitro phenotypic assays has been shown to affect PI susceptibility in both treatment-naive and treatment-experienced patients.17–19 More recently, Env has been implicated in drug resistance following PI exposure.20</p><p>MONARK investigated lopinavir/ritonavir monotherapy in comparison with lopinavir/ritonavir plus two NRTIs at treatment initiation in treatment-naive patients, with the monotherapy arm showing higher rates of virological failure.1,2 HIV-1 subtypes B and CRF02_AG were the dominant subtypes, and a higher failure rate was observed in the latter.21,22 This subtype is prevalent in West Africa and is also present in migrant HIV-1-infected populations in Europe, where subtype B otherwise dominates. Given the evidence for the importance of the inclusion of full-length Gag alongside its co-evolved protease in phenotypic assays and our previous study demonstrating reduced phenotypic susceptibility in patients experiencing virological failure in the MONARK trial, we hypothesized that Gag–protease-mediated susceptibility would differ between patients failing therapy and those with viral suppression, as well as between divergent strains (subtypes).23 This study sought to investigate the viral determinants of treatment failure in the MONARK lopinavir/ritonavir monotherapy arm trial by studying co-evolved, full-length Gag–protease from patients achieving virological response and experiencing virological failure in phenotypic PI susceptibility assays.</p><!><p>Characteristics of study participants and virus isolates from participants</p><p>VL, viral load.</p><!><p>Full-length Gag–protease was amplified from patient samples as previously described.24 Clonal sequencing of up to 10 viral variants for each sample was performed. The variant that most closely represented the consensus was taken forward for phenotypic testing. Protease sequences were analysed for PI resistance mutations using the Stanford Resistance Database (http://hivdb.stanford.edu/) (Table S1).25</p><!><p>PI susceptibility and infectivity were determined using previously described single-cycle assays.17,19,24 Briefly, 293 T cells were co-transfected with p8.9NSX+-derived test vector containing patient gag–protease, pMDG expressing vesicular stomatitis virus envelope glycoprotein (VSV-g) and pCSFLW expressing the firefly luciferase reporter gene with HIV-1 packaging signal. To determine PI susceptibility, transfected cells were seeded with serial dilutions of lopinavir and harvested pseudovirions used to infect fresh 293 T cells. To determine strain infectivity, transfected cells were seeded in the absence of drug. Infectivity was monitored by measuring luciferase activity 48 h after infection. All experiments were performed in duplicate.</p><!><p>We compared baseline characteristics of viral failures and successes using the Mann–Whitney rank sum test for continuous variables and Fisher's exact test for categorical variables. For analyses of lopinavir susceptibility, we defined each individual's susceptibility as the geometric mean of the two measurements. We compared lopinavir susceptibility between viral failures and successes by the Mann–Whitney rank sum test, which is robust for data that are not normally distributed. To check the sensitivity of our results to the choice of analytical method, we repeated the analyses using conditional logistic regression on the matched participants using viral failure or success as the outcome variable and log lopinavir fold change (FC) as the exposure. These sensitivity analyses gave concordant results.</p><!><p>Analyses involved 16 participants, of whom 8 were cases (failures) and 8 were controls (successes). Ten participants harboured HIV subtype CRF02_AG or G and six harboured subtype B, with subtypes split evenly between cases and controls. No patient harboured virus with major or minor resistance mutations to lopinavir, although a number of polymorphisms were present in protease, some of which are considered consensus amino acid positions in subtype CRF02_AG and G viruses (Table S1). Baseline CD4 and viral load were marginally higher in cases, but these differences were not statistically significant (Table 1).</p><!><p>Box plots of phenotypic assay FC in EC50 of lopinavir (LPV) by treatment outcome.</p><p>Box plots of phenotypic assay FC in EC50 of lopinavir (LPV) by subtype.</p><p>Scatter plot of FC in EC50 of lopinavir (LPV) versus (a) replicative capacity and (b) baseline viral load.</p><!><p>A previous analysis of the MONARK study suggested an association between the HIV-1 subtype AG (circulating in West Africa) and poorer outcome of lopinavir/ritonavir monotherapy, although there were confounding factors such as treatment adherence.22 The evaluation of a dual regimen of tenofovir plus lopinavir/ritonavir as first line in West Africa in the DAYANA trial also showed poor virological responses relative to standard NNRTI-based regimens,26 while the non-inferiority of another dual regimen, lopinavir/ritonavir plus lamivudine compared with a lopinavir/ritonavir-based triple regimen was demonstrated in other settings in the GARDEL trial.27 Here, we performed a case–control study within MONARK to test the hypothesis that baseline susceptibility, as determined by a full-length cognate Gag–protease assay, is correlated with outcome. In the primary analysis, in which subtype B and CRF02_AG/G were analysed together, we found a non-significant association between lopinavir susceptibility and failure (P = 0.14). In a subtype CRF02_AG/G-specific analysis, we found a significant difference between the two groups despite a relatively small sample size. Furthermore, subtype CRF02_AG/G viruses are substantially less susceptible to lopinavir than subtype B and this may have contributed to the findings of the MONARK investigators that subtype CRF02_AG patients were more likely to experience virological failure than those harbouring subtype B.</p><p>To date, phenotypic analysis using the commercial Phenosense assay (using a patient-derived protease with a subtype-mismatched Gag) had been performed on viruses derived from patients experiencing virological failure in the lopinavir/ritonavir monotherapy arm with the appearance of major PI resistance mutations.28 This analysis showed that in three patients with subtype CRF02_AG virus the pre-therapy EC50 FC was 0.57, 0.59 and 0.87 relative to the subtype B reference strain. In two subtype B-infected individuals, FCs of 1.46 and 1.49 were reported pre-therapy. The Phenosense assay result would suggest that subtype CRF02_AG viruses in MONARK are more susceptible than subtype B, in direct contradiction to our data using matched Gag and protease sequences (Figure 2 and Gupta et al.17).</p><p>However, the reduced PI susceptibility present in the variants from viral failure patients at baseline does not fully explain the subsequent treatment failure experienced. Six of the eight failure patients did initially achieve virological suppression <400 copies/mL on lopinavir/ritonavir monotherapy at week 24, before experiencing virological failure after >40 weeks of the trial. We hypothesize that PI monotherapy is potent enough to suppress viral replication with high adherence initially, but that reduced PI susceptibility lowers the tolerance or 'buffer zone' for subsequent suboptimal adherence. It is possible that patients with viruses demonstrating reduced PI susceptibility at baseline may be better suited to standard combined ART, which is likely to be more forgiving of reduced adherence, as three active agents are present. Alternatively, identification of these patients before treatment initiation would enable interventions to increase adherence and reduce this risk of failure. A third possibility is to closely monitor patients on PI monotherapy and intensify if low-level viraemia or viral rebound occurs.8,29</p><p>Limitations of our study include the relatively small sample size, the inclusion of more subtype CRF02_AG viruses in comparison with B and the possibility of viral recombination through our PCR and cloning strategy. Finally, our assay system did not incorporate the native gp160 envelope.20</p><p>In conclusion, we report an association between reduced PI susceptibility at baseline in the absence of known resistance mutations and subsequent virological failure on lopinavir/ritonavir monotherapy in patients harbouring subtype CRF02_AG/G viruses. This is an important finding as it indicates that it may be possible to predict treatment outcome on lopinavir/ritonavir monotherapy from baseline PI susceptibility. We hypothesize that reduced baseline PI susceptibility renders patients more vulnerable to virological rebound when their adherence is sub-optimal. This study suggests that lopinavir/ritonavir monotherapy should not be used in antiretroviral-naive patients infected with CRF02_AG. In maintenance therapy, studies are still needed to evaluate the impact of HIV-1 subtypes on virological response to PI/ritonavir monotherapy, as this regimen may be considered as a treatment option for individual patients in whom real-time viral load monitoring is available.</p><!><p>This study was funded by Wellcome Trust and a grant from AbbVie Laboratories. R. K. G. is funded by a Wellcome Trust Fellowship (WT093722MA). K. A. S. was funded by Public Health England (formerly the HPA). I. C. C. was involved in interpretation of data and review of the manuscript, but final content was decided by the principal investigator.</p><!><p>I. C. C. is an employee of AbbVie Laboratories and holds stock or options in AbbVie. The remaining authors have none to declare.</p><!><p>Table S1 and Figure S1 are available as Supplementary data at JAC Online (http://jac.oxfordjournals.org/).</p>

PubMed Open Access

Bleomycin Can Cleave an Oncogenic Noncoding RNA

Noncoding RNAs are pervasive in cells and contribute to diseases such as cancer. A question in biomedical research is whether noncoding RNAs are targets of medicines. Bleomycin is a natural product that cleaves DNA; however, it is known to cleave RNA in vitro. Herein, an in-depth analysis of the RNA cleavage preferences of bleomycin A5 is presented. Bleomycin A5 prefers to cleave RNAs with stretches of AU base pairs. Based on these preferences and bioinformatic analysis, the microRNA-10b hairpin precursor was identified as a potential substrate for bleomycin A5. Both in vitro and cellular experiments demonstrated cleavage. Importantly, chemical cleavage by bleomycin A5 in the microRNA-10b hairpin precursors occurred near the Drosha and Dicer enzymatic processing sites and led to destruction of the microRNA. Evidently, oncogenic noncoding RNAs can be considered targets of cancer medicines and might elicit their pharmacological effects by targeting noncoding RNA.

bleomycin_can_cleave_an_oncogenic_noncoding_rna

<p>The Encyclopedia of DNA Elements (ENCODE) project and functional studies have shown that 80 % of the genome is transcribed into RNA, but only 1–2 % is translated into protein.[1] Thus, RNA plays major roles in both healthy cells and disease biology that go beyond protein encoding. For example, long noncoding RNAs are involved in gene silencing and activation.[2] Small RNAs such as micro (mi) RNA play critical roles in decreasing the amount of protein translated from a given messenger (m) RNA.[3] In fact, in many cancers, miRNAs have been shown to cause disease through upregulation and subsequent silencing of pro-apoptotic proteins.[4] Thus, RNA is considered an important drug target.</p><p>The main manner in which RNA is drugged is through the use of oligonucleotides that target mRNAs.[5] These compounds can have suboptimal features, such as limited tissue and cellular permeability due to their large molecular weights, and often elicit off-target effects. Small molecules could be a preferred modality to target RNA; although bacterial RNAs such as the ribosome and riboswitches are tried and true targets of small molecules,[6] human RNAs are generally not considered druggable. One challenge is to develop small molecules that target human RNAs. Previous work in this area has included screening and sequence-based rational design enabled by Inforna.[7] One salient question remains: do known drugs actually target RNA and contribute to their pharmacological responses? In support of a positive answer to this question, the aminoglycoside antibiotic streptomycin was found to target a microRNA (miR) -21 precursor, albeit streptomycin is an antibacterial, not an anticancer agent.[8]</p><p>The antitumor natural product bleomycin has been known for decades to cleave DNA (Scheme 1).[9] There is an encyclopedia of information describing structure–activity relationships (SARs) for both the cleaved DNA and the bleomycin cleaver.[10] Interestingly, RNA has also been shown to be cleaved by bleomycin in vitro,[11] with a focus on cleavage of tRNA. More detailed analysis of the preferred RNAs that are cleaved could elucidate currently unknown cellular targets. However, RNA has not been shown to be cleaved by bleomycin in cells.</p><p>To assess the RNAs that are cleaved by bleomycin A5, we initially tested RNAs containing stretches of base pairs that were designed to be displayed in a unimolecular hairpin structure for cleavage. These RNAs had stems that contained the canonical AU and GC base pairs in each of their 5′ and 3′ orientations, for example, 5′AU/3′UA and 5′UA/3′AU, in a unimolecular GNRA hairpin, where N indicates any nucleotide and R indicates a purine nucleotide. The initial series also included the 5′GU/3′UG wobble pair found in RNA (see Figures S1–S3 for RNA secondary structures).[12] Interestingly, bleomycin A5 cleaved the RNAs containing AU pairs most efficiently, with 5′-AUAU-3′ sequences being cleaved more than 5′-AAUU-3′ sequences (Figure 1 A). Bleomycin A5 did not significantly cleave RNAs containing only GU and GC pairs (Figures S1–S3).</p><p>To further analyze the sequence selectivity of bleomycin A5 in RNA, we designed and tested six additional RNAs. These RNAs contained stretches of AU pairs of different lengths, separated by GC pairs with the GNRA hairpin loop (Figure 1 B). The RNAs with the longest stretches of AU pairs (AU-1 and AU-2) were efficiently cleaved by bleomycin A5; whereas, RNAs with only one or two AU pairs inserted between GC pairs were cleaved less efficiently (AU-3 and AU-4; Figures 1 C and S4). Interestingly, two of the RNA hairpins contain only purine bases on the 5′-side of the hairpin loop and only pyrimidine bases on the 3′-side of the hairpin loop. In both of these examples, bleomycin A5 cleaved only the purine-rich region of the RNA. Quantification of the percent cleavage of each RNA revealed that RNAs rich in AU base pairs and an RNA with a 5′-GAGA-3′ sequence (AU-6) were the most highly cleaved by bleomycin A5 (Figure 1 C).</p><p>A nearest-neighbor analysis of the cleaved sites of each RNA was completed, starting at the 5′-end of the RNA, in three base steps. This analysis revealed a strong preference for purine-rich sequences, especially those rich in AU base pairs (Figure 1 D). In the RNAs used in this study, three base sequences containing only pyrimidines were not cleaved by bleomycin A5. Three base steps containing three purines were cleaved 95 % of the time in the RNAs analyzed here, indicating that bleomycin A5 cleaves purine-rich sequences in RNA.</p><p>In studies of DNA cleavage, bleomycin prefers to cleave 5′-GC or 5′-GT sequences.[10] NMR spectroscopy and SAR studies of bleomycin and a DNA substrate have shown that this sequence selectivity in DNA is largely due to hydrogen bonding interactions between bleomycin's metal binding pyrimidine core and the N3- and C2-amino groups of guanine.[10, 13] Bleomycin binds the minor groove of DNA and is capable of making two hydrogen bonding interactions with guanine but only one with adenine; thus, bleomycin prefers to cleave 5′GC/T sequences (Figure 2 A).[13] Although bleomycin interacts with the minor groove of DNA, the minor groove of RNA is narrower; therefore, bleomycin is more likely to interact with the major groove of RNA. Interaction with the major groove of RNA would allow hydrogen bonding to occur with the C6 amino group of adenine that is displayed in the major groove (Figure 2 B). Previous in vitro studies have demonstrated that alteration of the position of the amino group in DNA changes the cleavage pattern and cleavage intensity of bleomycin.[14] DNA substitutions of guanine to inosine and adenine to 2aminopurine altered the cleavage of DNA in vitro.[14] A 2-aminopurine substitution to an AU-paired RNA would move the amino group to the minor groove of the RNA (Figure 2 B). Indeed, 2-aminopurine substitutions to an AU-paired RNA altered the cleavage of bleomycin A5, indicating that hydrogen bonding interactions with the amino group could play a role in cleavage specificity (Figure 2 C).</p><p>To determine whether bleomycin A5 would cleave naturally occurring, noncoding RNA containing AU-rich sequences, we searched all human miRNAs using bioinformatics to identify those with the potential to be good substrates for cleavage.[15] Among all human miRNAs, 13 miRNA precursors have a stretch of six consecutive AU base pairs in their secondary structures (Table S1). Of these 13 miRNAs, seven have been implicated in disease (Table S1). We identified the miR-10b primary hairpin precursor (pri-miR-10b) that contains a potentially reactive AU-rich region: 5′-AUAUAU/3′UAUAUA. The miR-10b is oncogenic and is overexpressed in many cancers, contributing to invasion and metastasis.[16] The putative target site for bleomycin A5 in pri-miR-10b is composed of the six consecutive AU pairs adjacent to the Drosha cleavage site (Figure 3 A). A small molecule that inhibits the biogenesis of miR-10b by binding to the Drosha cleavage site has previously been identified;[17] however, small molecule cleavage of miR-10b in particular, and miRNAs precursors in general, has not been studied.</p><p>We thus analyzed in vitro cleavage of pri-miR-10b. Indeed, bleomycin A5 cleaved the target at the predicted site and also had a secondary site of cleavage near the Dicer site of 5′GUG/3′CAC. The secondary cleavage was not surprising, as the target contains two purines, and these nucleotides were cleaved at the short hairpin RNAs that were initially studied for cleavage. Other minor cleavage sites included other purine-rich sequences, such as 5′-GAA-3′ sequences. Further analysis of the cleavage sites revealed that bleomycin A5 also cleaved the 5′-AUAUAU-3′ sequence at the 3′-end of the RNA (Figure S5). To confirm the sequence-specific cleavage of pri-miR-10b by bleomycin A5, the AU pairs were mutated to GC pairs, and the 5′GUG/3′CAC sequence was mutated to a 5′GCG/3′CGC sequence. At a concentration of 10 μm, bleomycin A5 cleaved the mutated RNA 30 % less than pri-miR-10b; cleavage was completely ablated at lower concentrations (Figure S6).</p><p>With the favorable activity and selectivity data observed in vitro, we sought to assess the effect of the compound on miR-10b and its precursors in cells. Because the cleavage sites for bleomycin A5 are near the Drosha and Dicer enzyme cleavage sites in the miRNA hairpin precursor, it is possible that small molecule cleavage of the miRNA could allow it to enter the miRNA pathway, or it could cause destruction of the miRNA. Therefore, HeLa (human cervical cancer) cells overexpressing pri-miR-10b through plasmid expression were treated with bleomycin A5. Treatment with as little as 100 nm bleomycin A5 resulted in a decrease in the levels of pri-miR-10b, as determined by RT-qPCR (Figure 4 A). Cleavage of pri-miR-10b resulted in a decrease in the levels of the mature microRNA as well (Figure 4 A). Interestingly, bleomycin A5 reduced the levels of mature miRNA at concentrations as low as 10 nm. Although bleomycin A5 cleaved the RNA near the Drosha cleavage site in vitro, chemical cleavage of pri-miR-10b in cells resulted in destruction of mature miR-10b. Importantly, in this transfected system, bleomycin A5 did not affect the levels of plasmid DNA (Figure S7). In an additional cellular experiment, the effect of bleomycin A5 on the triple-negative breast cancer cell line MDA-MB-231 was assessed. This cell line was used because it endogenously expresses higher levels of miR-10b than healthy, noncancerous cells.[16a] In MDA-MB-231 cells, bleomycin A5 at concentrations of 10 and 1 μm reduced the amount of mature miR-10b, showing that bleomycin A5 can also cleave endogenous pri-miR-10b (Figure 4 B).</p><p>In summary, analysis of the preferred RNA cleavage site for bleomycin A5 revealed that bleomycin A5 cleaved 5′-AUAU-3′ sequences and other purine-rich sequences with high efficiency. The identification of a biologically relevant RNA with a stretch of six AU base pairs, pri-miR-10b, showed that bleomycin cleaved the RNA at the predicted site in vitro. In cells, treatment with bleomycin A5 also resulted in cleavage of pri-miR-10b. This study shows that noncoding RNAs such as miRNAs can be targets of known drugs. Other targets of bleomycin outside of the canonical DNA targets could include RNA, which should now be considered a druggable target from phenotypic screens, for example. Furthermore, elucidation of the preferred cleavage site of bleomycin might help in designing selective RNA cleavers by attaching an RN- binding module to bleomycin A5 to target a specific RNA, thereby reprograming its cellular targets.[18] Importantly, bleomycin A5 might also serve as a useful reagent to map RNA AU base pairs in vitro and to map RNA secondary structure in cells.</p>

PubMed Author Manuscript

FUNCTIONALLY AND SPATIALLY DISTINCT MODES OF MUNC18-SYNTAXIN 1 INTERACTION*

Eukaryotic membrane trafficking is a conserved process under tight temporal and spatial regulation in which the fusion of membranes is driven by the formation of the ternary SNARE complex. Syntaxin 1a, a core component of the exocytic SNARE1 complex in neurones and neuroendocrine cells, is regulated directly by munc18-1, its cognate SM (Sec1p/Munc18) protein. SM proteins show remarkable structural conservation throughout evolution indicating a common binding mechanism and function. However, SM proteins possess disparate binding mechanisms and regulatory effects, with munc18-1, the major brain isoform, classed as atypical in both its binding specificity and mode. We now show that munc18-1 interacts with syntaxin 1a through two mechanistically distinct modes of binding, both in vitro and in living cells, in contrast to current models. Furthermore these functionally divergent interactions occur at distinct cellular locations. These findings provide a molecular explanation for the multiple, spatially distinct roles of munc18-1.

functionally_and_spatially_distinct_modes_of_munc18-syntaxin_1_interaction*

<!>Vectors and cell culture<!>Protein Biochemistry<!>Confocal laser scanning microscopy and image analysi<!>TCSPC-FLIM acquisition and analysis<!>RESULTS<!>DISCUSSION<!>

<p>In neuronal and neuroendocrine cells, exocytosis is mediated by the plasma membrane proteins (t-SNAREs) syntaxin and SNAP-25 (synaptosome-associated protein 25 kDa) and the vesicular protein synaptobrevin (v-SNARE)(​1,​2). The cytoplasmic regions of these three proteins interact to form a trimeric, four-helical complex, the generation of which drives fusion of the two opposing bilayers(3). This process is regulated by a conserved set of accessory proteins which operate throughout the trafficking pathway. The Sec1p/Munc18 (SM) protein family represents one such set of modulators, with SM protein mutations characterised by a severe disruption of general secretion or neurotransmitter release(4-​7). The mammalian SM protein munc18-1 was originally isolated as a syntaxin 1 binding protein which binds to the monomeric form of syntaxin 1, rendering the t-SNARE unable to form the SDS-resistant ternary SNARE complex(8,​9). Syntaxin 1 can act as a molecular switch, adopting two structurally distinct forms(10). In the open form, the SNARE helix does not interact with the N-terminal three helical regulatory domain (termed Habc) and has been shown to not interact with munc18-1(10). In contrast the closed form of syntaxin 1, in which the N-terminal Habc domain interacts with the SNARE helix, exhibits a high affinity for munc18-1. Association of syntaxin 1 with its SNARE partners to form the ternary SNARE complex, which prevents syntaxin adopting the closed conformation, has also been shown to preclude munc18-1 binding(11). However, a recent finding by Zilly et al using lysed cellular membrane sheets provided evidence that munc18-1 may interact with syntaxin 1 when in the binary SNARE complex (a heterodimer of syntaxin and SNAP-25)(​12).</p><p>The interaction of munc18-1 with its cognate syntaxin is in sharp contrast to the specificity of its yeast homologue Sec1p, which binds its cognate syntaxin, Sso1p, in the ternary SNARE complex and not in the monomeric state(13). This binding specificity has, however, since been questioned(14). Another yeast SM protein, Vps45p, binds to its cognate Golgi syntaxin, Tlg2p, both in the monomeric state and in the ternary SNARE complex(15,16). This interaction can occur through an N-terminal region of Tlg2p, with a similar interaction observed for its mammalian Golgi homologue, syntaxin 16(17). A similar ability to bind syntaxin in both monomeric and complexed forms has been observed for the SM protein Sly1p and its cognate syntaxin Sed5p(18,19) This apparent discrepancy in binding mode suggests differences in binding sites and recognition motifs of SM proteins and their syntaxin partners.</p><p>To date, the crystal structures of two SM proteins, munc18-1 and Sly1p (from S. cerevisiae), in complex with their cognate SNAREs, have been solved(20,21). Despite exhibiting high structural homology, munc18-1 and Sly1p appear to interact with their cognate syntaxins, syntaxin 1a and Sed5p, respectively, through independent sites. Munc18-1 binds to the closed conformation of syntaxin 1a with the N-terminal Habc domain folded back on to the helix involved in SNARE complex formation(10,22), whereas Sly1p binds the N-terminus of Sed5p through an independent binding site. The disparate modes of interaction between SM proteins and their cognate syntaxins, observed in different species and for different membrane compartments, are at odds with the conserved role for SM proteins in membrane fusion indicated by genetic experiments(2,23). Importantly, effects exerted by munc18-1 on different stages in the syntaxin lifecycle, from trafficking(24), through vesicle docking(25) to the final fusion event itself(26), cannot be explained using the current model of the munc18-1 - syntaxin 1a interaction. In this study we analysed in detail the syntaxin 1 munc18-1 interaction using mutations shown to lock syntaxin 1 in the open conformation(10). Here we show that munc18-1 binds to an evolutionarily conserved motif at the N-terminus of syntaxin 1 in addition to the closed form binding mode. Using quantitative colocalisation and fluorescence lifetime imaging microscopy (FLIM) we show that these two binding modes of syntaxin1 and munc18-1 are spatially segregated in living cells. Furthermore while the closed form binding mode prevents progress through the SNARE assembly pathway, syntaxin 1 in the open form can progress to the ternary SNARE complex with munc18-1 remaining associated. This is the first demonstration that the major neuronal and neuroendocrine SNARE, syntaxin 1, and its cognate SM protein, munc18-1, can interact through dual modes of binding.</p><!><p>Plasmids encoding glutathione-S-transferase (GST) fusion proteins with syntaxin 1a (a.a. 1-261, cytoplasmic domain), SNAP-25 (a.a. 1-206) and synaptobrevin (a.a. 1-96) were described previously(27,28). A plasmid encoding a poly histidine tagged munc18-1 (a.a. 1-594) was as previously described(29). Generation of N-terminal truncations of syntaxin 1a was performed by PCR and subsequent ligation into HindIII/KpnI and HindIII/XbaI sites of pmCerulean-C1 and PGEX-KG respectively or into pTargeT (Promega). Distance between the fusion tag and syntaxin 1a was maintained in all constructs. The [L165A, E166A] mutation was generated by site directed mutagenesis using a QuikChange II XL kit (Stratagene). An EYFP-munc18-1 fusion was generated using similar standard techniques. Neuroblastoma 2a (N2a) cells were grown in DMEM supplemented with 10% foetal bovine serum, 10 mM L-glutamine, 50 units penicillin, 50 μg/ml streptomycin and maintained at 37°C in 5% (v/v) CO2, 95% (v/v) air. All transfections were performed using ExGen 500 (Fermentas).</p><!><p>Recombinant GST fusion proteins were expressed and purified as previously described(30). For in vitro binding reactions, 2 μg GST-syntaxin 1a or SNAP-25 was immobilised on glutathione Sepharose beads (GE Healthcare) and incubated in a total volume of 100 μl of 20 mM HEPES, pH 7.4, 100 mM NaCl, 1 mM EDTA, 0.1% TX-100 (Buffer A) with 3 μg of protein to be tested for binding for 1 hour at 4°C. Beads were washed by low speed centrifugation and bound protein eluted in SDS-containing sample buffer followed by SDS PAGE and Coomassie staining. For binding reactions involving munc18-1, GST fused syntaxin 1a was incubated with either freshly prepared detergent rat brain extract, as previously described(30), or with freshly prepared detergent bacterial extract containing expressed recombinant His6-munc18-1. Circular dichroism experiments were performed as previously described(31). For the measurement of syntaxin 1a affinity for munc18-1, 0.2 pmol of GST fused syntaxin 1a was immobilised on glutathione Sepharose beads and incubated with increasing concentrations of munc18-1 in detergent bacterial extract in a reaction volume of 0.5 ml at 21°C. The beads were washed three times in buffer A and bound protein analysed by Western immunoblotting using a monoclonal anti-munc18-1 antibody (BD Biosciences) and West Dura enhanced chemiluminescence kit (Pierce). The chemiluminescent signal was imaged using a cooled 16-bit CCD and quantified using ImageJ (http://rsb.info.nih.gov/ij/). The concentration of munc18-1 in the detergent bacterial extract was measured by purification of munc18-1 and in-gel quantification using Sypro Red (Invitrogen). The calculated intensity volumes were fitted with a variable slope dose response relationship using Prism (GraphPad). For binding assays using a purified syntaxin 1a-munc18-1 complex, bacterial detergent extracts containing expressed GST-Syx1-261 (or variant) and His6-munc18-1 were mixed. The resulting protein complex was purified sequentially on glutathione Sepharose resin and IMAC resin (Bio-Rad). Purity was assessed by Sypro Red in-gel staining to confirm an equimolar stoichiometric ratio(30).</p><!><p>All imaging experiments were performed using a Zeiss LSM 510 Axiovert confocal laser scanning microscope, equipped with a pulsed excitation source (MIRA 900 Ti:Sapphire femtosecond pulsed laser, with a coupled VERDI 10W pump laser (Coherent). Data were acquired using a 1024 × 1024 pixel image size, using a Zeiss Plan NeoFLUAR 1.4 NA 63x oil immersion lens, or a Zeiss C-Apochromat 1.2 NA 63x water corrected immersion objective lens. All imaging was performed using living cells, maintained at 37°C in 5% (v/v) CO2, 95 % (v/v) air. Image data, acquired at Nyquist sampling rates, were deconvolved using Huygens software (Scientific Volume Imaging) and the resulting 3-D models were analysed using ImageJ software (http://rsb.info.nih.gov/ij/). Residual maps were generated by calculating the residual of each voxel from the linear regression fit to the intensities of each channel within each voxel. The resulting residuals are displayed on a colour scale from -1 to 1 (with zero residual coloured cyan) with brightness corresponding to the combined intensity of the two channels. Cell peripheries were determined using transmitted light imaging combined with CLSM data.</p><!><p>Time-correlated single photon counting (TCSPC) measurements were made under 800-820 nm TPE (two-photon excitation), which efficiently excited cerulean, without any detectable direct excitation or emission from EYFP, using a non-descanned detector (R3809U-50) multichannel plate-photomultiplier tube (MCP-PMT), or a fast photomultiplier tube (H7422; both Hamamatsu Photonics UK) coupled directly to the rear port of the Axiovert microscope. TCSPC recordings were acquired for between 10 s and 60 s, mean photon counts were between 105 - 106 counts per second. Images were recorded at 256 × 256 pixels from a 1024 × 1024 image scan with 256 time bins over a 12 ns period(32). Off-line FLIM data analysis used pixel-based fitting software (SPCImage, Becker & Hickl). The fluorescence was assumed to follow a multi-exponential decay. In addition, an adaptive offset-correction was performed. A constant offset takes into consideration the time-independent baseline due to dark noise of the detector and the background caused by room light, calculated from the average number of photons per channel preceding the rising part of the fluorescence trace. To fit the parameters of the multi-exponential decay to the fluorescence decay trace measured by the system, a convolution with the instrumental response function was carried out. The optimisation of the fit parameters was performed by using the Levenberg-Marquardt algorithm, minimising the weighted chi-square quantity. This approach can be used to separate the interacting from the non-interacting donor fraction in our FRET systems. The long lifetime component τ2 was determined by control assays with cerulean alone, or expressed with (non-interacting) EYFP as described above. This value was subsequently used as a fixed τ2 lifetime for all other experiments. As controls for non-specific FRET, or FRET between GFPs that may form dimers spontaneously when over-expressed in cells, we determined the fluorescence lifetimes of cerulean-Syx1-288 alone, cerulean alone, or cerulean-Syx1-288 co-transfected with EYFP (not shown). No FRET was detected in any of these experiments. Likewise, experiments using ceruleanA206K-fused Syx1-288 revealed no self-dimerisation between fluorescent proteins</p><!><p>We analysed the conservation of amino acids in syntaxin 1 homologues from evolutionarily distant species along with Sed5p from S. cerevisiae (Fig 1A). There is a high degree of similarity between the syntaxin homologues throughout the sequence corresponding to the SNARE helix and head domain of syntaxin. Of note is the additional spike in similarity in the first 15 amino acids. Analysis of the sequence contained in this region demonstrated a high degree of conservation of amino acids with a lysine-aspartate-arginine (KDR) motif absolutely conserved in both Sed5p and the syntaxin homologues. A structural alignment of munc18-1 and Sly1p demonstrates the highly conserved topology of these two proteins despite an overall sequence identity of 18% (Fig. 1B). Highlighted on the bound Sed5p peptide (Fig. 1B) are the residues absolutely conserved in the sequence alignment in Fig. 1A. The conservation of these residues suggested that the amino-terminal of syntaxin 1a may be capable of interacting with munc18-1 in a similar manner and that an "open" mutation of syntaxin 1a(10), could still support an interaction with munc18-1.</p><p>We therefore determined the ability of syntaxin 1a (Syx1-261) and an open mutant of syntaxin [L165A, E166A](10) (Syx 1-261 [Open]) to bind to munc18-1 in rat brain detergent extracts. Both forms of syntaxin readily bound munc18-1 (Fig. 1C). As the crystal structure of the syntaxin 1a - munc18-1 complex (notably excluding the first 26 amino acids of syntaxin) indicates that only closed syntaxin can bind to munc18-1(21), these findings suggest that an additional mechanism and site of interaction exists. To determine whether the conserved amino acids found at the syntaxin N-terminus are involved in a direct interaction with munc18-1, we generated a series of truncations of syntaxin in both the wild type and open mutant forms. These purified proteins were incubated with a bacterial detergent extract containing recombinant His6-munc18-1 (Fig. 1D). Syx1-261 and Syx 1-261 [Open] both readily bound recombinant munc18-1 from bacterial detergent extracts replicating the observation using detergent brain extracts. The N-terminal truncations of syntaxin 1a had no affect on their ability to bind munc18-1, but importantly, the combination of N-terminal truncation and the open mutations severely diminished munc18-1 interaction. As truncation of the first six amino acids produced a decrease in binding of munc18-1 to open mutants of syntaxin 1a (Fig. 1D), similar to the longer deletions, we concentrated on this short truncation in further experiments. Circular dichroism analysis confirmed that the mutations and truncations incorporated in to syntaxin 1a resulted in no major changes to the secondary structure of the proteins (Fig. 1E). Because binding reactions, as used in Fig. 1D are strongly dependent on the relative concentrations of the interacting partners used, the affinity of munc18-1 for syntaxin, and the mutant forms, was measured. Syx1-261, Syx1-261 [open] and Syx7-261 (Δ6) all had a similar affinity for munc18-1. Only when the open mutation was combined with truncation of the N-terminus (Syx7-261 [open] (Δ6)) was there a significant reduction in the affinity of syntaxin 1a for munc18-1 (Fig. 1F).</p><p>These in vitro data suggest that syntaxin 1a can interact with munc18-1 through two distinct binding mechanisms. It is unclear, however, whether these binding modes are employed in a cellular context. To address this question, fluorescent fusions of munc18-1 and syntaxin 1a were visualised in living Neuroblastoma 2A (N2A) cells. A quantitative approach was used to investigate colocalisation of syntaxin 1a and munc18-1. In addition to the merger of the two emission channels, to display areas of coincidence, the intensities of both channels within each voxel was displayed as a frequency histogram and fit by linear regression. From this fit, residuals for each voxel were calculated and displayed as a residual map to highlight areas of covariance (expected if 2 proteins interact with a defined stoichiometry) within an image. Syntaxin 1a and munc18-1 exhibited a high degree of coincidence and of intensity covariance on the plasma membrane and in intracellular membranes (Fig. 2A). Similarly, Syx7-288 (Δ6) and Syx1-288 [open] demonstrated a high degree of colocalisation with munc18-1. Importantly, however, munc18-1 no longer colocalised with syntaxin when both the N-terminal truncation and the open syntaxin mutations were combined (Syx 7-288 (Δ6) [Open]). Munc18-1 is a soluble protein with no localisation signals, and as such its membrane association relies on its interaction with syntaxin 1a(33); when co-expressed with Syx7-288 [open] (Δ6), munc18-1 adopted a cytoplasmic localisation. As a negative control for random colocalisation, unfused munc18-1 and EYFP were co-expressed with cerulean-syntaxin 1a. Pearson's correlation coefficient analysis demonstrated no significant difference between cerulean-Syx1-288, -Syx1-288 [open] and -Syx7-288 (Δ6) covariance with EYFP-munc18-1 (Fig. 2B). However, the combined N-terminal deletion and open mutation of syntaxin 1a (Syx7-288 [open] (Δ6)) resulted in a large decrease in covariance to a level not significantly different from the negative control.</p><p>This quantitative analysis of colocalisation between mutant proteins in living cells provided evidence for a role of the N-terminus of syntaxin 1a in binding to munc18-1. However, colocalisation data are limited by the resolution of the microscope (maximally 200 nm) and do not directly indicate interaction. In order to increase our understanding of each mode of syntaxin 1a binding to munc18-1, we employed fluorescence lifetime imaging microscopy (FLIM). FLIM quantifies the fluorescence lifetime of a fluorophore, the duration of which is sensitive to the microenvironment it inhabits(32). Thus, Förster resonance energy transfer (FRET), to an interacting acceptor molecule, shortens dramatically the donor fluorescence lifetime: this effect can be quantified directly in each pixel of an image. Furthermore, the low light levels required for FLIM allowed us to select cells expressing very low levels of exogenous proteins. Using this technique, the fraction of non-interacting and interacting donor fluorescence lifetimes in each pixel can be resolved (Fig. 3A-D)(​32).</p><p>This approach revealed that fewer Syx1-288 and munc18-1 molecules interact on the plasma membrane compared to intracellular locations (Mann-Whitney, p<0.007, n = 18), and confirmed that Syx1-288 [Open] and munc18-1 interact in cells (Fig. 3F). Surprisingly, FRET (i.e. donor fluorescence lifetime and amplitude ratios) between cerulean-Syx1-288 [Open] and EYFP-munc18-1 was substantially decreased in intracellular membranes compared to Syx1-288 and munc18-1 (p<0.001, n=8; Fig. 3F), with cell surface interactions maintained in evenly distributed puncta. In contrast, cerulean-Syx7-288 (Δ6) showed no decrease in interaction in intracellular membranes, but a large reduction in FRET on the cell surface (p<0.001, n=8). These effects were additive; cerulean-Syx7-288 [Open] (Δ6) exhibited decreased FRET both intracellular and on the cell surface (Fig. 3F). Thus, the two binding modes are spatially distinct, with syntaxin 1a N-terminal deletion affecting plasma membrane interactions, and abolition of syntaxin closed conformation binding affecting predominantly intracellular complexes. Why does deletion of the syntaxin N-terminal amino acids reduce interaction with munc18-1 at the plasma membrane? We have shown that N-terminal binding to munc18-1 is utilised when syntaxin is opened. This being the case, open syntaxin might be able to bind both munc18-1 and the other SNARE(s) simultaneously at the cell surface; loss of N-terminal interaction would thus dissociate munc18-1 in the presence of SNAP-25.</p><p>To explore this hypothesis, we used highly purified complexes of syntaxin 1a bound to munc18-1. It was thus possible to probe the effect of N-terminal truncation and the open mutation on progression of syntaxin through the SNARE assembly pathway (Fig. 4A). Incubation of Syx1-261/munc18-1 with immobilised SNAP-25 demonstrated that in this form, syntaxin is unable to engage into either the binary SNARE complex or the ternary SNARE complex (Fig. 4A left panel). The N-terminal truncation of syntaxin did not allow interaction of the Syx7-261 (Δ6)/munc18-1 complex, with no SNAP-25 binding detected (Fig. 4A centre panel). However, incorporation of the open mutation into syntaxin allowed Syx1-261[open]/munc18-1 to interact with SNAP-25 forming both the binary SNARE complex (t-SNARE heterodimer) and the ternary SNARE complex (Fig. 4A right panel). As a control munc18-1 alone was incubated with immobilised SNAP-25 (Fig. 4B). As previously shown(11), munc18-1 was unable to interact directly with SNAP-25. Importantly, munc18-1 remained bound to syntaxin 1a throughout the SNARE assembly pathway. This SNARE complex-munc18-1 interaction cannot include closed form syntaxin, and we conclude therefore that the N-terminal mode of interaction is essential at the cell surface for this reason.</p><!><p>Our data demonstrate that munc18-1 can interact with syntaxin 1a through two discrete modes of binding which are spatially distinct within the cell. The closed form of syntaxin-munc18-1 binding is a structurally constrained state with properties comparable to those previously described for this complex(8,21). This mode of interaction is of high affinity and occurs primarily, but not exclusively, on intracellular membranes (Figs. 1 and 3). We speculate that this conformation of syntaxin 1a is clamped by munc18-1 in order to allow the passage of syntaxin through the endoplasmic reticulum and Golgi apparatus, preventing ectopic intracellular SNARE complex formation. In addition to this form of binding we now show that syntaxin 1a and munc18-1 can interact through an evolutionarily conserved motif at the N-terminus of syntaxin. It is of note that an early clone of syntaxin 1a(22) lacks the conserved N-terminal amino acids identified in Fig. 1 and has since been widely used to examine syntaxin 1a protein-protein interactions. In contrast to a previous report(10) using rat brain extracts (as in Fig. 1D), munc18-1 readily binds to a mutation which prevents syntaxin 1a adopting a closed conformation, exhibiting no reduction in affinity for this interaction (Fig. 1). We are unable to explain this discrepancy, although it should be noted that the mutants used in the NMR studies had a 26-amino-acid N-terminal truncation, as this gave better spectra(10). Importantly the full-length form of this mutant syntaxin rescues the C. elegans unc13 phenotype(34), interpreted as circumventing the requirement for Unc13 to dissociate Unc18 and open syntaxin. However, we now show that this mutation permits syntaxin to progress through the SNARE assembly pathway, all the while still associated with munc18-1 (Fig. 4).</p><p>Little or no information is available to describe the intracellular locations of SM-syntaxin interactions. A recent report found that Vps45p, an SM homologue in S. cerevisiae, can interact in vitro with its cognate syntaxin through an undefined second mode of binding in addition to its well characterised N-terminal interaction(35). Additionally, the mammalian SM protein munc18-3 binds its partner syntaxin 4 principally through an N-terminal interaction in vitro(36). However, the authors of this report were unable to exclude that a second mode of binding exists. It is therefore credible that a dual mode of interaction with cognate syntaxins is a conserved feature of the SM protein family.</p><p>Why have SM proteins evolved multiple functions and binding modes, and how do they control vesicle fusion? We propose that syntaxin 1a is transported to the plasma membrane in a closed conformation, with the Habc regulatory head domain interacting with the SNARE helix, bound by munc18-1 (Fig. 4C). This is in agreement with a previously suggested chaperone role for munc18-1(24) and would most closely resemble the binding mode observed in the crystal structure of the syntaxin1a/munc18-1 complex(21), although may also include the additional interaction of munc18-1 with the N-terminus of syntaxin 1a. This mode of binding would preclude any aberrant interaction with intracellular or trafficking plasma membrane SNAREs until delivery to the plasma membrane. At the plasma membrane the syntaxin 1a would change conformation from a closed to an open state under the regulation of an as yet undefined plasma membrane factor(30,34,37). In this form munc18-1 binding would occur through the N-terminus of syntaxin 1a through a similar interaction as observed in the Sed5p/Sly1p crystal structure(20). This conformational change of syntaxin 1a would permit the subsequent interaction of syntaxin with SNAP-25 to form the binary t-SNARE heterodimer, whilst retaining the interaction with munc18 through the N-terminal mode of binding. This complex may form a scaffold for vesicle docking, a proposed function for munc18-1(25,38) that cannot be explained adequately using current models. Subsequent triggering of exocytosis would then drive engagement of synaptobrevin with the t-SNARE heterodimer to form the four helical ternary SNARE complex formation driving membrane fusion. The association of munc18-1 with the ternary SNARE complex through the N-terminus of syntaxin 1a can explain the observed effect munc18-1 can exert on the final fusion step (26). It is reasonable to expect that the function of munc18-1 here parallels that of sec1p, which apparently stabilises the yeast exocytic SNARE complex, and is required for vesicle fusion(13). The interaction of munc18-1 with different conformational forms of syntaxin, at spatially distinct sites, can explain the multiple, and somewhat controversial roles munc18-1 has been proposed to play in the cell, and so help unify thinking on SM protein structure and function.</p><!><p>This work was funded by a Wellcome Trust Fellowship award to R.R.D.</p><p>soluble N-ethylmaleimide sensitive factor attachment receptor</p><p>Sec1p/Munc18</p><p>target SNARE</p><p>vesicle SNARE</p><p>synaptosome-associated protein 25 kDa</p><p>fluorescence lifetime imaging microscopy</p><p>glutathione-S-transferase</p><p>Neuroblastoma 2A</p><p>Syntaxin 1a can interact with munc18-1 through its N-terminus in vitro. A, Sequence alignment of syntaxin 1 homologues and Sed5p with similarity scored at each position (left panel). The highly conserved N-terminal region (a.a. 1- 28, right panel) is shown on a colour coded scale (yellow - identical, cyan-conserved, green - similar). B, Structural alignment of Sly1p (red) bound to an N-terminal peptide of Sed5p (green) (PDB: 1MQS(20)) and munc18-1 (pink) (PDB: 1EPU(39)). Highlighted on an enlarged view (right panel) are the absolutely conserved residues from panel a. C, Both Syx1-261 and Syx1-261 [open], immobilised on Sepharose beads, readily bind munc18-1. Native brain GSTs bound directly to the Sepharose resin. D, Truncation of the N-terminus of Syx1-261 [open] reduces its ability to bind munc18-1. N-terminal truncations of GST-Syx1-261 and GST-Syx1-261 [open] were immobilised on glutathione Sepharose beads and incubated with His6-munc18-1 containing bacterial lysate. Bound material was analysed by SDS PAGE and Coomassie staining. E, Circular dichroism spectra of Syx1-261, Syx1-261 (Δ6), Syx1-261 [open] and Syx1-261 [open] (Δ6). F, Measurement of the equilibrium dissociation constant for the syntaxin 1a-munc18-1 complex. GST-Syx1-261, or a mutant form, immobilised on Sepharose beads, was incubated in the presence of varying concentrations of munc18-1. Bound material was analysed by Western immunoblotting. Error bars represent S.E.M. (n=3).</p><p>Munc18-1 colocalisation with syntaxin 1a in live cells is dependent on either closed form or N-terminal binding. A, Wild-type or mutant mCer-Syx (green), and EYFP-munc18-1 (red) were expressed in N2a cells and imaged by confocal laser scanning microscopy. The merge image shows areas of coincidence in yellow hues. The 2-D histogram represents the intensity for each channel in each voxel with a colour scale representing frequency. The residual map displays weighted residuals from the line fit to the histogram, thus indicating fluorescence channel covariance. The hue is from -1 to 1 with cyan corresponding to a zero residual. EYFP and unfused munc18-1 were used as a control. Scale bar: 2 μm. B, Combined covariance analysis of cerulean-Syx and EYFP-munc18-1 in N2a cells. Mean Pearson's coefficients are shown (n>4).</p><p>Different binding modes of munc18-1 - syntaxin 1a are spatially distinct. A, The excited state fluorescence decay of cerulean-Syx1-288 in the absence of an energy acceptor followed a mono-exponential decay, as previously described for cerulean(40) (dark circles). When co-expressed with EYFP-munc18-1, the cerulean-Syx1-288 decay data no longer fit to a single exponential, but were well described by a bi-exponential decay function (light grey circles). B, Cerulean-Syx1-288 fluorescence, in the absence of EYFP-munc18-1, exhibited a plasma membrane and intracellular membrane distribution (left panel) The colour scale in the "FLIM map" represents the fluorescence lifetime and brightness represents intensity (centre panel). The fluorescence lifetime values were plotted as a frequency distribution histogram (right panel) with a single fluorescence lifetime of 2288 ± 40 ps (mean ± S.E.M., n=18). C, The intensity localisation of cerulean-Syx1-288, in the presence of EYFP-munc18-1, was unaltered (left panel), but the FLIM map reveals a quenched fluorescence lifetime (right panel). The colours in this map represent the weighted mean of the two lifetimes for each pixel, one identical to the non-FRET lifetime (2288 ps) and one significantly shorter (680 ± 34 ps mean ± S.E.M; p<0.0001, n = 12). These data are plotted as the weighted mean (significantly different to the non-FRET distribution) (D) and as separate fluorescence lifetimes (E), representing the non-FRET and the FRET component contained in each pixel. The amplitudes of these components (which combine to 100%) represent the relative proportion of each lifetime. F, These approaches were applied to each syntaxin 1a mutant, in addition calculating the lifetime amplitude ratio of FRET:non-FRET amplitudes in each pixel. All data treatments were identical. Pixels containing only non-FRET values appear in greyscale. Amplitude ratio scale: zero (no FRET component, greyscale), 0.1-0.66 (red), 0.66-1.33 (green), 1.33-2.5 (blue). For all panels (A-F): scale bars represent 2 μm and fluorescence lifetime color scales 1250 ps (red)-2250 ps (blue).</p><p>Munc18-1 can remain associated with syntaxin 1a through the SNARE assembly pathway. A, Munc18-1 in complex with Syx1-261 (left panel), Syx7-261 (Δ6) (centre panel) or Syx1-261 [open] (right panel) was incubated with immobilised SNAP-25 in the presence or absence of synaptobrevin (Syb). B, As a control munc18-1 alone was incubated with immobilised SNAP-25. C, Proposed model for munc18-1 interaction with syntaxin 1a. An undefined factor releases syntaxin from its closed state on the plasma membrane permitting binary and ternary SNARE complex formation. Munc18-1 can remain bound to the N-terminus of syntaxin 1a through this process. It was not possible to purify a Syx7-261[open](Δ6)/munc18-1 complex for use in this binding assay because of the decreased affinity of Syx7-261[open](Δ6) for munc18-1.</p>

PubMed Author Manuscript

CANCER METASTASIS DIRECTLY ERADICATED BY TARGETED THERAPY WITHA MODIFIED SALMONELLA TYPHYMURIUM

Cancer metastasis is the life-threatening aspect of cancer and is usually resistant to standard treatment. We report here a targeted therapy strategy for cancer metastasis using a modified strain of Salmonella typhimurium. The genetically modified strain of S. typhimurium is auxotrophic for the amino acids arginine and leucine. These mutations preclude growth in normal tissue but do not reduce bacterial virulence in tumor cells. The tumor-targeting strain of S. typhimurium, termed A1-R and expressing green fluorescent protein (GFP), was administered to both axillary lymph and popliteal lymph node metastasis of human pancreatic cancer and fibrosarcoma, respectively, as well as lung metastasis of the fibrosarcoma in nude mice. The bacteria were delivered via a lymphatic channel to target the lymph-node metastases and systemically via the tail vein to target the lung metastasis. The cancer cells expressed red fluorescent protein (RFP) in the cytoplasm and GFP in the nucleus linked to histone H2B, enabling color-coded real-time imaging of the bacteria targeting the metastatic tumors. After 7\xe2\x80\x9321 days of treatment, the metastases were eradicated without the need of chemotherapy or any other treatment. No adverse effects were observed. This new strategy demonstrates the clinical potential of targeting and curing cancer metastasis with engineered bacteria without the need of toxic chemotherapy.

cancer_metastasis_directly_eradicated_by_targeted_therapy_witha_modified_salmonella_typhymurium

INTRODUCTION<!>S. typhimurium A1-R therapy of cancer cells in vitro<!>S. typhimurium A1-R therapy of experimental lymph node metastasis<!>S. typhimurium A1-R therapy therapy of spontaneous lymph node metastasis<!>S. typhimurium A1-R therapy of experimental lung metastasis<!>Statistical analysis<!>S. typhimurium A1-R therapy of cancer cells in vitro<!>S. typhimurium A1-R therapy of experimental lymph-node metastasis<!>S. typhimurium A1-R therapy of spontaneous lymph node metastasis<!>S. typhimurium therapy for lung metastasis<!>Effect of S. typhimurium on body weight<!>DISCUSSION<!>Supplemental Movie 1. Trafficking of bacteria to axillary lymph node metastasis<!>Supplemental Movie 2. Trafficking of bacteria to popliteal lymph node metastasis<!>Targeting of tumor cells by S. typhimurium A1-R in vitro<!>Targeted therapy of experimental lymph node metastasis with S. typhimurium A1-R<!>Growth of axillary-lymph-node metastasis in untreated animals<!>Targeted therapy of spontaneous popliteal lymph node metastasis with S. typhimurium A1-R<!>Targeted therapy of experimental lung metastasis with S. typhimurium A1-R<!>

<p>It has been known for approximately 60 years that anaerobic bacteria can selectively grow in tumors [Coley, 1906; Malmgren and Flanigan, 1955; Moese and Moese, 1964; Gericke and Engelbart, 1964; Thiele et al., 1964; Kohwi et al., 1978; Brown and Giaccia, 1998; Fox et al., 1996; Lemmon et al, 1997; Sznol et al., 2000; Low et al., 1999; Clairmont et al., 2000; Yazawa et al., 2000; Yazawa et al., 2001; Kimura, et al., 1980]. The conditions that permit anaerobic bacterial growth (i.e., impaired circulation and extensive necrosis) are found in many tumors. Several approaches to developing tumor-therapeutic anaerobic bacteria have been described. Yazawa et al. [2000; 2001] showed that the anaerobic bacterium Bifidobacterium longum could selectively grow in the hypoxic regions of solid tumors. Dang et al. [2001] created a strain of Clostridium novyi depleted of its lethal toxin (C. novyi-NT). C. novyi spores germinated within the avascular regions of tumors in mice and destroyed surrounding viable tumor cells [Dang et al., 2001]. The main efficacy of these anaerobic bacteria was in combination with chemotherapy [Dang et al., 2001]. Eradication of tumors was also achieved in mice by combining C. novyi-NT with radiation [Bettegowda et al., 2003]. It was recently shown that treatment of mice bearing large, established tumors with C. novyi-NT plus a single dose of liposomal doxorubicin could lead to eradication of the tumors. The bacterial factor responsible for the enhanced drug release was identified as a protein termed liposomase [Cheong et al., 2006; Juliano, 2007].</p><p>The facultative anaerobe Salmonella typhimurium was first attenuated by purine and other auxotrophic mutations to be used for cancer therapy [Low et al., 1999; Hoiseth and Stocker, 1981; Pawelek et al., 1997]. These bacteria replicated in the tumor to >1,000-fold compared with normal tissues [Low et al., 1999]. Salmonella lipid A was also genetically modified by disrupting the msbB gene to reduce septic shock [Low et al., 1999]. Melanomas in mice treated with the Salmonella msbB mutant were 6% the size of tumors in untreated controls [Low et al., 1999]. However, these Salmonella variants did not cause tumor regression or eradication, only growth inhibition. S. typhimurium with attenuated lipid A has been evaluated in a Phase I clinical trial [Toso, et al., 2002].</p><p>We have previously developed [Zhao et al., 2005] a genetically modified bacterial strain of S. typhimurium, auxotrophic for leucine and arginine, which also expresses green fluorescent protein (GFP), termed S. typhimurium A1. When introduced i.v. or intratumorally, A1 invaded and replicated intracellularly in various cancer cells in vivo as well as in vitro. When A1 was injected intratumorally, the tumor completely regressed by day 20. There were no obvious adverse effects on the host when the bacteria were injected i.v. or intratumorally. The S. typhimurium A1 strain grew throughout the tumor, including viable malignant tissue, but could not sustain growth in normal tissue. This result is in marked contrast to the anaerobic bacteria evaluated previously for cancer therapy that were confined to necrotic areas of the tumor as discussed above. The ability to grow in viable tumor tissue may account, in part, for the unique antitumor efficacy of the A1 strain [Zhao et al., 2005].</p><p>The A1-R strain was reisolated from A1-targeted tumor tissue in vivo. The idea was to increase the tumor targeting capability of the bacteria. As a consequence of this selective step, the tumor-cell targeting of the reisolated A1 increased in vivo as well as in vitro. The reisolated A1 bacteria, termed A1-R, administered i.v., resulted in human breast cancer [Zhao et al., 2006] and prostate cancer [Zhao et al., 2007] regression, including tumor eradication in orthotopic nude-mouse models.</p><p>Cancer metastasis is the life threatening aspect of cancer. Importantly, metastasis is most often treatment-resistant. In the present study, we demonstrate that S. typhimurium A1-R can directly target and eradicate cancer metastases demonstrating the potential of bacterial therapy to treat otherwise incurable disease.</p><!><p>XPA1 and HT-1080 cells labeled with RFP in the cytoplasm and GFP in the nucleus were grown in 24-well tissue culture plates to a density of 104 cells per well. Bacteria were grown in LB and harvested at late-logarithmic phase, diluted in cell culture medium and added to the tumor cells (1 × 105 colony-forming units (CFU) per well). After 1 h incubation at 37°C, the cells were rinsed and cultured in medium containing gentamycin sulfate (20 μg/ml) to kill external but not internal bacteria. Interaction between bacteria and tumor cells was observed at the indicated time points. An Olympus BH 2-RFCA fluorescence microscope equipped with a mercury 100-W lamp power supply and GFP filter set (Chroma Technology, Brattleboro, VT) was used for fluorescence microscopy.</p><!><p>To obtain metastasis in the axillary lymph node, XPA1-RFP cells were injected into the inguinal lymph node (afferent lymph node to the axillary) in nude mice. Nudemice were anesthetized with a ketamine mixture (10 μLketamine HCL, 7.6 μL xylazine, 2.4 μL acepromazine maleate, and 10 μL H2O) via s.c. injection. A one-cm incision was made in the abdominal skin to expose the inguinal lymph node. The inguinal lymph node was exposed without injuring the lymphatic. The skin was fixed on a flat stand. A total of 10 μL medium containing 5 × 105 cancer cells was injected into the center of the inguinal lymph node. Just after injection, cancer cells were observed trafficking in the efferent lymph duct toward the axillary lymph node. Seven days later (day 0), mice were anesthetized, and the axillary lymph node was observed for metastasis. A two-cm incision was made at the center of the chest wall. The greater pectoral muscle was released from the sternum to expose the axillary lymph node. The connective tissue on the axillary lymph node was separated. The lymph node was imaged for metastasis. Lymph nodes which had less than a one-mm tumor were included in this study. Five mice were treated with A1-R, and 5 were used as controls for each cell line. A1-R bacteria (108 CFU) were injected in the inguinal lymph node. After injection, the axillary lymph node was observed repeatedly at different time points with the Olympus OV100 Small-Animal Imaging System (Olympus Corp., Tokyo, Japan). The size of metastasis (fluorescent area [mm2]) was measured at each imaging time point.</p><!><p>To obtain spontaneous lymph node metastasis, 5 × 106 HT-1080-GFP-RFP cells in 20 μl of Matrigel basement membrane matrix (BD Bioscience, San Jose, CA) were injected into the foot pad in nude mice. The presence of popliteal lymph node metastasis was determined by fluorescence imaging every week after tumor injection. Mice were given the ketamine mixture for anesthesia and laid out in the prone position. The entire limb was observed with the OV100 imaging system without any traumatic procedures. Once metastasis was confirmed in the popliteal region, bacteria therapy was started targeting the metastasis (day 0). A1-R bacteria (108 CFU) were injected subcutaneously in the foot pad. The size of metastasis and primary tumor, and body weight was measured every week. Six mice were treated with A1-R, and 6 were used as controls. Another two mice were used for imaging immediately and at day-7 after bacteria injection. The experiment was terminated when the control primary tumor invaded the popliteal region or the mouse died. The experimental data are expressed as the mean ± SE. Statistical analysis was done using the two-tailed Student's t-test.</p><!><p>Four week old, female nude mice were used. To obtain lung metastasis, HT-1080-GFP-RFP cells (1 × 106 in 100μl PBS) were injected into the tail vein of the nude mice (day 0). On day 4 and day 11, 5 × 107 CFU bacteria per mouse were injected into the tail vein. Five mice were treated with bacteria and five mice were used as untreated control. On day16, all animals were sacrificed and the lungs were imaged to determine the efficacy of bacteria therapy for lung metastasis. To observe the lung metastasis at lower magnification, an RFP filter was used (excitation 545 nm, emission 570–625 nm).</p><p>All animal studies were conducted in accordance with the principles and procedures outlined in the National Institutes of Health Guide for the Care and Use of Laboratory Animals under assurance no. A3873-1.</p><!><p>The total number of metastases on the surface of the lung was counted using the Olympus OV100 Small Animal Imaging System (Olympus Corporation, Tokyo, Japan). The pictures were taken in same setting (exposure time 1000 msec, 0.56×lens). The number of metastases in each group was expressed as mean ± SE. Statistical analysis was done using the two-tailed Student's t-test.</p><!><p>We first tested the ability of S. typhimurium A1-R to kill human pancreatic cancer and fibrosarcoma cells in vitro. A1-R expressing GFP was observed to invade and replicate intracellularly in the XPA1 human pancreatic cancer and HT1080 human fibrosarcoma cell lines expressing GFP in the nucleus and RFP in the cytoplasm. Intracellular bacterial infection led to cell fragmentation and rapid cell death (Fig. 1).</p><!><p>A new experimental model of lymph node metastasis was developed for this study. To obtain an experimental metastasis in the axillary lymph node, XPA1-RFP human pancreatic cancer cells were injected into the inguinal lymph node in the nude mice (Fig. 2A). Just after injection, cancer cells were imaged trafficking in the efferent lymph duct to the axillary lymph node [Hayashi et al., 2007]. Metastasis in the axillary lymph node was subsequently formed. A1-R bacteria were then injected into the inguinal lymph node to target the axillary lymph node metastasis. Just after bacterial injection, a large amount of bacteria were visualized around the axillary lymph node metastasis (Fig. 2B and Supplementary Movie-1). By day 7, all lymph node metastases had been eradicated in contrast to growing metastases in the control group (Figs. 2C and 3). There were very few bacteria in the lymph node by day 7, and no bacteria were detected after day 10. This route of administration was therefore able to deliver sufficient bacteria to eradicate the lymph node metastasis after which the bacteria became undetectable.</p><p>The average tumor size (fluorescent area) in the axillary lymph nodes on day 0 was 0.4 ± 0.19 mm2 in the treatment group and 0.46 ± 0.08 in the untreated group, respectively. On day 7, it was 0 mm2 in the treatment group and 0.98 ± 0.17 in the untreated group.</p><!><p>We then tested bacterial therapy strategy for spontaneous lymph node metastasis from a tumor growing in the footpad. At first, only A1-R bacteria were injected in the foot pad in nude mice in order to determine any adverse effects. No infection, skin necrosis, or body weight loss or fatality was detected (data not shown). Then HT1080-GFP-RFP human fibrosarcoma cells were injected into the footpad of additional nude mice. Before treatment, the presence of popliteal lymph node metastasis was determined by weekly imaging. Once the metastasis was detected, A1-R bacteria were injected subcutaneously in the footpad.</p><p>Bacteria are small particles and when injected subcutaneously, the lymph system immediately collects them from the site of injection. The lymph system is well known as a drainage route for bacterial infection. One mouse was used to observe the injected bacteria trafficking in the lymphatic channel. The popliteal region was exposed just after bacteria injection and a large amount of GFP bacteria targeting the popliteal lymph node metastasis was observed by fluorescence imaging (Fig. 4A, Supplementary Movie-2). Dual-color labeling of the cancer cells distinguished them from the GFP bacteria. After treatment, the popliteal lymph node was observed every week by fluorescence imaging. One mouse was used to image the bacteria by exposing the popliteal lymph node on day 7. All lymph node metastases shrank and 5 out of 6 were eradicated within 7 to 21 days after treatment in contrast to growing metastases in the control group (Fig. 4B and C).</p><!><p>To obtain lung metastasis, dual-colored HT-1080 cells were injected into tail vein of the nude mice (day 0). On day 4 and day 11, bacteria were injected into the tail vein. On day 16, all animals were sacrificed and the lungs were imaged to determine the effect of bacteria therapy on lung metastasis. To observe the lung metastasis at lower magnification, an RFP filter was used (excitation 545 nm, emission 570–625 nm). In the bacterial-treatment group, only a few cancer cells were observed in contrast to multiple metastases in the control (untreated) group (Fig. 5A and B). The number of metastases on the surface of the lung was significantly lower in the treatment group than in the control group (P<0.005) (Fig. 5C).</p><!><p>There were no significance differences between the treated and untreated groups in body weight and primary tumor size (Table 1).</p><!><p>The strategy employed in our studies for bacterial therapy of cancer has significant advantages over that using obligate anaerobic bacteria such as Clostridium. Anaerobic bacteria only grow in the necrotic regions of tumors and therefore require additional cytotoxic chemotherapy in order to kill viable tumor tissue and effect eradication [Dang et al., 2001]. The facultative anaerobic S. typhimurium, in contrast, grows in viable as well as necrotic regions of tumors and therefore can eradicate tumors without additional chemotherapy. Thus, our method of bacterial treatment can target spontaneous metastasis and eradicate them without toxic chemotherapy, a major advance over other approaches to bacterial therapy which require combination with chemotherapy in order to effect eradication [Cheong et al., 2006; Juliano, 2007]. For lymph node metastasis, we used a lymphatic targeting approach that was highly effective. The lung metastases in our model were eradicated by systemic injection. However, no adverse effects were observed.</p><p>The lymphatic system is a critical route for both bacteria drainage and cancer spread. In humans, subcutaneous injection of tumor-targeting bacteria around the primary tumor should be an effective way to target lymphatic metastasis. Alternatively, the tumor-targeting bacteria can be injected directly in the lymph duct similar to lymphangiography. The bacteria could then reach regional lymph nodes and potentially eradicate the metastasis. The adverse effects should be much less than systemic administration. Intralymphatic administration of bacteria in humans should require only low doses of bacteria.</p><!><p>GFP-labeled S. typhimurium A1-R are seen trafficking in a lymphatic duct to the axillary lymph node metastasis of the XPA1-RFP pancreatic cancer cell line. The bacteria were injected in the inguinal lymph node. Real time images were obtained with the Olympus OV100 Small Animal Imaging System. The lymph duct was exposed with a skin-flap.</p><!><p>GFP-labeled S. typhimurium A1-R are seen trafficking in a lymphatic duct to the popliteal lymph node metastasis of the HT1080-GFP-RFP human fibrosarcoma cell line. The bacteria were injected in the foot pad. Real time images were obtained with the Olympus OV100 Small Animal Imaging System. The lymph duct was exposed with a skin-flap.</p><!><p>XPA1 human pancreatic cancer cells labeled with RFP in the cytoplasm and GFP in the nucleus [Yamamoto et al., 2004] were grown in 24-well tissue culture plates. Bacteria were added to the tumor cells (1 × 105 CFU per well). After 1 h of incubation at 37°C, the cells were rinsed and cultured in medium containing gentamycin sulfate (20 μg/ml) to kill external but not internal bacteria. Interaction between bacteria and tumor cells was observed at the indicated time points under fluorescence microscopy. (A to D) GFP-expressing S. typhimurium A1-R was able to invade and replicate intracellularly in the dual-color XPA1 cell line in vitro. The cytopathic effects of A1-R on XPA1 cells after infection were visible using dual-color fluorescence. Intracellular bacterial infection leading to eventual cell fragmentation and cell death was observed. Bars, 20 μm.</p><!><p>(A) Scheme for experimental lymph node metastasis model. Cancer cells were injected into the inguinal lymph node. After injection, the cells trafficked to the axillary lymph node and formed metastases. Small skin incisions were made around both the inguinal lymph node and axillary lymph node for purposes of imaging. (B) Seven days after XPA1-RFP human pancreas cancer cell injection in the inguinal lymph node, an experimental metastasis was observed in axillary lymph node. After A1-R was injected in the inguinal lymph node, bacteria were imaged around the tumor in the axillary lymph node. (C) Seven days after bacteria injection, the metastasis has been eradicated.</p><!><p>(A) Seven days after XPA1-RFP human-pancreas-cancer cell injection in the inguinal lymph node (day 0), metastasis was observed in the axillary lymph node. (B) On day 7, the metastasis was growing in contrast to the treatment group where the metastases were eradicated.</p><!><p>(A1) Dual color HT-1080 human fibrosarcoma cells were injected into the foot pad in nude mice. The popliteal lymph node metastasis was imaged every week after cancer-cell injection. Once metastasis was visualized in the popliteal region (arrow), bacterial therapy was started to target the metastasis. (A2) Bacteria were injected subcutaneously in the foot pad. The popliteal lymph node was exposed to image bacteria trafficking. GFP bacteria were observed in lymphatic channels connecting the popliteal lymph node. (A3) Higher magnification of lymphatic channel. GFP bacteria and dual color cancer cells are readily distinguished. (B1–B6) Representative weekly images of the popliteal lymph node metastasis after bacteria injection. (B1–B3) 0, 7 and 14 days after bacteria injection. The lymph node metastasis has been eradicated by 14 days. (B4–B6) 0, 7 and 14 days for control (no bacteria treatment). The lymph node metastasis continues to grow. (C1–C2) All lymph node metastases have regressed and 5 out of 6 are eradicated within 7 to 21 days after treatment in contrast to growing tumors in the control group.</p><!><p>To obtain lung metastasis, dual-colored HT-1080 cells were injected into the tail vein of nude mice (day 0). On day 4 and day 11, bacteria were injected into the tail vein. On day 16, all animals were sacrificed and the lungs were imaged to determine the effect of bacteria therapy on lung metastasis. To observe the lung metastasis at lower magnification, an RFP filter was used (excitation 545 nm, emission 570–625 nm). In the bacterial treatment group, only a few cells were observed (A) in contrast to multiple metastases in the control (non-treatment) group (B). The number of metastases on the surface of the lung was significantly lower in the treatment group than in the control group (P<0.005) (C).</p><!><p>Effect of bacterial therapy on body weight</p><p>S. typhimurium A1-R was used to treat nude mice with metastatic cancer as described in the Materials and Methods. In the course of bacterial treatment of the spontaneous lymph-node metastasis, there were no significant differences between the treated and untreated groups in body weight or primary tumor size in the foot pad.</p>

PubMed Author Manuscript

Glucose biosensor based on open-source wireless microfluidic potentiostat

Wireless potentiostats capable of cyclic voltammetry and amperometry that connect to the Internet are emerging as key attributes of future point-of-care devices. This work presents an \xe2\x80\x9cintegrated microfluidic electrochemical detector\xe2\x80\x9d (iMED) three-electrode multi-potentiostat designed around operational amplifiers connected to a powerful WiFi-based microcontroller as a promising alternative to more expensive and complex strategies reported in the literature. The iMED is integrated with a microfluidic system developed to be controlled by the same microcontroller. The iMED is programmed wirelessly over a standard WiFi network and all electrochemical data is uploaded to an open-source cloud-based server. A wired desktop computer is not necessary for operation or program uploading. This method of integrated microfluidic automation is simple, uses common and inexpensive materials, and is compatible with commercial sample injectors. An integrated biosensor platform contains four screen-printed carbon arrays inside 4 separate microfluidic detection chambers with Pt counter and pseudo Ag/AgCl reference electrodes in situ. The iMED is benchmarked with K3[Fe(CN)6] against a commercial potentiostat and then as a glucose biosensor using glucose-oxidising films of [Os(2,2\xe2\x80\xb2-bipyridine)2(polyvinylimidazole)10Cl] prepared on screen-printed electrodes with multi walled carbon nanotubes, poly(ethylene glycol) diglycidyl ether and flavin adenine dinucleotide-dependent glucose dehydrogenase. Potential application of this cost-effective wireless potentiostat approach to modern bioelectronics and point-of-care diagnosis is demonstrated by production of glucose oxidation currents, under pseudo-physiological conditions, using mediating films with lower redox potentials.

glucose_biosensor_based_on_open-source_wireless_microfluidic_potentiostat

Introduction<!>Materials<!>Hardware design<!>Power supply<!>Digital design<!>Analogue design<!>Microfluidic design<!>Electromechanical servos<!>Electrochemical measurements<!>Resistor polarization<!>Cyclic voltammetry<!>Amperometry<!>Microcontrolled syringe pump and programmed control of flow rates<!>Amperometric glucose biosensor<!>Conclusions

<p>Some of the most important challenges in point-of-care biosensor research are miniaturization, wireless data transmission and fully functional integrated devices [1,2]. Electrochemical biosensors explore measurements of an electrical signal produced on an electrochemical transducer applicable to addressing these challenges. A potentiostat is the electronic hardware required to control and run most electro-analytical experiments. The literature reports on many cost-effective open-source potentiostat platforms [3–10], however none integrate both microfluidics and wireless technology.</p><p>Designing systems with microfluidics rather than conventional laboratory analysis procedures can lead to miniaturization, increased throughput and reduced reagent consumption [11]. However, most existing fluid flow techniques rely on manual or semi-automated hardware. To combat this, programmed control over microfluidics is one target towards highly integrated, point-of-care clinical diagnostic devices for home and hospital use [12–17]. To this end, potentiostatic microfluidic biosensors must include some sort of standalone programmable controller for data acquisition, precise flow rate control and direction.</p><p>Automation in microfluidics has been applied to the extraction of nucleic acids [18], rotating disc based analysis of infectious diseases [19], flow cytometry [20], magnetic stirring [21], genetic stability with multiplexed PCR [22], chemiluminescence detection [23] and protein immunoblotting [24]. This research has brought about the search for a new generation of improved flow control and automation that satisfy requirements such as cost-effectiveness, all-in-one instrumentation and other complexity [25].</p><p>Recent efforts to automate liquid flow and direction include a report [26] on a bluetooth enabled microfluidic liquid handling system that utilised a microcontroller, solenoid valves and pneumatic pressure. The system capability was demonstrated by application to a bead-based HIV1 p24 sandwich immunoassay on a multi-layer poly-dimethylsiloxane (PDMS) chip. The Rusling group [27] demonstrated an automated electrochemiluminescence based system that combined a microcontroller, commercial micropumps and a charge coupled device camera for detection of proteins related to prostate cancer. Tan et al. [28] presented a platform for controlled fluid flow, where driving force is supplied by compressed nitrogen, via desktop computer for a multiplexed photonic crystal enhanced fluorescence detection of protein biomarkers interleukin 3 and Tumor Necrosis Factor α. Innovation in microfluidic directional control has seen introduction of valves that are torque-actuated [29,30], magnetic [29,31,32], solenoid [26,33–37] and pressure driven [14,38]. We integrate into the microfluidic platform for the first time, to our knowledge, independently controlled, electrically actuated servo-valves. The servo-valves are inexpensive relative to off-the-shelf commercially available automated microfluidic valve control options [13].</p><p>Here we integrate automated microfluidic flow and direction with the iMED, a wireless multi-potentiostat platform using inexpensive mass-produced components: the ESP-32 microcontroller, operational amplifiers (op-amps), high torque electromechanical servo-valves, and commercial syringe pump. Microcontroller incorporation allows for a generic and consistent route towards miniaturization, integration, automation and wireless connectivity of electrochemical protocols, identified as goals within the literature [39,40].</p><p>The iMED implements custom protocols, sequencing events and electrochemical analysis in a rapid and simple manner using the high-level programming language C/C++ with the open-source Arduino programming software. The ESP-32 microcontroller is approved by the FCC and CE regulatory requirements necessary for manufacture and sale in the United States and European Economic Area which could increase enthusiasm for the introduction of this technology to the market [13].</p><p>The entire iMED platform is designed to operate without a desktop computer. To this end, the iMED uses over the air updates (OTA) mechanism which allows the potentiostat to update itself based on data received while the normal firmware is running (for example, over WiFi). The device also sends all electrochemical data to an open-source cloud-based server for data post-processing [41]. This data is then visualised and accessible via any Internet connected device, such as a smart phone, tablet, laptop or wearable. The assembled wireless potentiostat requires only rudimentary instruction to use and here we demonstration the use of the iMED using an electrochemical enzyme-amplified biosensor to achieve detection of glucose concentrations in solution. Glucose sensing is critical to diagnosis, monitoring and treatment of the currently incurable diabetes disease [42].</p><p>Advancements in the application of enzymatic electrodes towards biosensors and bio-powered devices have been made in recent decades. Enzymes are an attractive route towards biocatalysis due to their inherent substrate specificity and ability to operate under mild conditions [43–45]. One such application of this is with immobilisation of glucose oxidising enzymes at an electrode surface. While glucose oxidase (GOx) is the most reported enzyme for this application, Flavin adenine dinucleotide-dependent glucose dehydrogenase (FADGDH) is more suitable than GOx due to its lack of reactivity with molecular oxygen. Previous work reported by us has demonstrated the improved suitability of FADGDH over GOx for such applications [46,47]. The combination of redox polymer films with enzymes on electrode surfaces to mediate electron transfer between active site and electrode surface has been widely demonstrated. Osmium based redox polymers consisting of an osmium centre attached to polymers such as poly(N-vinylimidazole) have been reported [48–54]. The suitability of osmium based redox centres is due to their relative stability, tunable redox potential and ability to operate at low potentials [55,56]. More recently, the addition of nano-supports such as multi-walled carbon nanotubes (MWCNTs) to the biofilm has resulted in increased current response as well as improved stability of the biofilms [57–61]. The use of crosslinkers such as glutaraldehyde or poly(ethylene glycol) diglycidyl ether (PEGDGE) to increase biofilm stability has been investigated previously [62,63]. Here we report on biofilms containing the redox polymer [Os(2,2′-bipyridine)2(poly-vinylimidazole)10Cl]+ (Os(bpy)PVI) with the glucose oxidising enzyme FADGDH and MWCNTs crosslinked using PEGDGE immobilised on screen printed carbon electrodes for the detection of glucose using the iMED potentiostatic microfluidic platform.</p><!><p>PDMS, Sylgard 184 was purchased from Sigma and used with a ratio of 1:10 PDMS per master mold. Sodium phosphate dibasic, sodium phosphate monobasic monohydrate, sodium chloride, potassium chloride, was purchased from Sigma. All solutions were prepared using deionized water. Rheodyne 7725i injector was purchased from Sigma and used as received. Screen-printed carbon arrays consisting of eight 850 micrometer diameter sensors were from Kanichi Research (UK). 4-Port Switching Valves were from Upchurch Scientific. The component cost of the potentiostat is below 30 USD. In this calculation, the equipment costs (syringe pump and microfluidic components) were not included. The full list of electronic parts are available in Section 4.4 of the Supporting Information.</p><p>The flavin dependent glucose dehydrogenase is from Aspergillus sp. (FADGDH 1.1.99.10, Sekisui, Cambridge, USA; product GLDE-70–1192). The MWCNTs (product 659258; Sigma-Aldrich) were pretreated under reflux in concentrated nitric acid for 6 h and isolated by filtration. Polyethylene glycol diglycidyl ether (PEGDGE) was purchased from Sigma-Aldrich (average Mn 526). All aqueous solutions unless otherwise stated were prepared in Milli-Q water (1818 MΩ cm). The [Os(2,2′-bipyridine)2(polyvinylimidazole)10Cl]+ (Os(bpy)PVI) redox polymer were synthesised according to literature procedures [64,65]. Characterization of (Os(bpy)PVI) is available in Section 2 of the Supporting Information.</p><p>The artificial plasma contained uric acid (68.5 mg l−1), ascorbic acid (9.5 mg l−1), fructose (36 mg l−1), lactose (5.5 mg l−1), urea (267 mg l−1), cysteine (18 mg l−1), sodium chloride (6.75 g l−1), sodium bicarbonate (2.138 g l−1), calcium sulfate (23.8 mg l−1), magnesium sulfate (104.5 mg l−1) and bovine serum albumin (7 g l−1).</p><!><p>The electronic design that forms the basis of the iMED is described in two parts: the digital and the analogue designs. In addition the power supply and microfluidic designs are discussed. Firstly the power supply is described because it is continually referenced in all subsequent subsections. Secondly the digital design includes the ESP-32 microcontroller for waveform generation and data acquisition, dual-core processing and WiFi upload to a cloud-based server and an analog-to-digital converter (ADC) because its output is digital. Thirdly, the analogue design includes an electronic circuit based around op-amps, and a digital-to-analog converter (DAC) because its output is analogue. Finally microfluidic design which discusses electromechanical servo-valves, syringe pumping and electrochemical detection chambers. Also present in this section is a program that was developed in MATLAB that post-processes data uploaded to a cloud based server that plots amperometry data. Detailed circuit schematics are available as Supporting Information Section 4. All electronic components are mounted on a solderless breadboard, commonly used for prototyping due to their speed and ease when adding or changing components [33,66–68], printed circuit boards can also be used and provide a more permanent solution [69], even though more difficult to rearrange components in the proof of concept phase.</p><!><p>The iMED is designed to be a completely integrated instrument, only requiring electrode connectors and a power supply. The iMED does not require a desktop PC or laptop to operate. A power supply is necessary to supply the correct voltage and current to each active component in the circuit. A reconfigured L305P-01 desktop ATX power supply (Dell) is used to reduce the overall cost of the design. The power supply can deliver 3.3, 5, −12 and 12 V. The 5 V rail supplies the ESP-32 via the 5 V in pin. The ADC/DAC, not introduced yet but are external to the ESP-32, require additional power supply in order to guarantee a stable voltage reference. In both cases this power is provided by the stable 5 V rail of the ATX power supply.</p><p>As for the current reading part of the design, the ADS1115 ADC operating in single ended mode (measuring the voltage between the analog input channel and analog ground) requires a bias potential of −5 V in order to offset the current reading so the maximum values of current, positive and negative, are in the range 0–5 V. This bias potential is achieved using the LM79L05 negative regulator with a 0.1 μF output capacitor that provides a stable −5 V from the −12 V rail of the ATX power supply.</p><p>The circuit uses LM324 op-amps that require a symmetrical power supply. The LM79L08 and LM78L08 negative and positive regulators are used for bipolar (−8 and 8 V) supply to the op-amps.</p><!><p>The ESP-32 is a Xtensa dual-core 32-bit microprocessor, 160 MHz, 600 DMIPS, WiFi: 802.11 b/g/n, Bluetooth Low Energy (BLE), that is used as the central processing unit for the potentiostat.</p><p>We use the WiFi capability instead of the BLE because of relatively higher bandwidth, larger coverage area, and increased cost-effectiveness due to WiFi network prevalence in buildings and cities compared to BLE [70,40]. To our knowledge, the literature reports on one previously published open-source wireless potentiostat using BLE [10], which confines wireless data transfer to a mobile phone. WiFi allows connection to any internet connected device and facilitates OTA, making it a more versatile device.</p><p>The ESP-32 is wirelessly programmed via WiFi using OTA updates, previously published potentiostats, both wired [3–9] and wireless [10], required a wired connection to the microcontroller to upload necessary computer programs, diminishing point-of-care applications. Once the ESP-32 is wirelessly programmed (details as Supporting Information, Section 5.5), a smart phone is used to connect to the ESP-32, which is then used to type WiFi credentials and set potential to be applied when in amperometry mode (detailed smart phone connectivity is available as Supporting Information, Section 5.4). In order to improve performance and versatility, both processor cores are programmed with ESP-IDF FreeRTOS (real-time operating system) embedded into the Arduino IDE core. This real-time operating system allows the ESP-32 to run tasks in both CPU cores. This dual core technology is used with one core dedicated to the DAC and the other to the ADC. It is worth noting that if only one core is used, accurate scan rates > 100 mV s−1 cannot be obtained because the time taken to read the ADC input exceeds the time needed to apply the potential via the DAC.</p><p>The op-amp in a summing amplifier configuration is used to apply a potential window output between negative and positive values controlled by the DAC input. The design of the summing amplifier is a compromise between the potential resolution and potential window width. The ESP-32 has a reported built in two channel 8-bit digital-to-analog converter (DAC) [71]. Instead, for digital-to-analog conversion, important when applying a programmable potential via the potentiostat, we use an external MCP4725 DAC ([72]). The key attributes of the MCP4725 DAC are; 12-bit resolution, fast settling time (0.6 μs) which is the time required for the output to reach and remain within a given error band following the ESP-32 input stimulus, single-supply operation: 2.7–5.5 V, and I2C communication. The external MCP4725 12-bit DAC is used to provide a resolution an order of magnitude over the ESP-32 internal lower resolution 8-bit DAC, and therefore greater precision within the 2 V potential window. The MCP4725 DAC provides 212 = 4096 quantization levels with a sensitivity of 5 V/4096 = 1.2 mV in the voltage output range. The DAC is powered by 5 V to create a 5 V potential for the summing amplifier circuit configuration. This 5 V potential overcomes the 3.3 V ESP-32 built in logic that would otherwise limit the potential window applied to the op-amp circuit for the potentiostat. The DAC requires a 5 V reference voltage, which must be stable, that is supplied via the reconfigured ATX power supply. For cost-effective comparisons of various DACs used within the literature, see Supporting Information, Section 4.2.1.</p><!><p>Electrical signals between the three electrode electrochemical cell and the ESP-32 is based on a circuit implemented with the op-amp in a transimpedance amplifier (TIA) configuration. TIA configuration is used to convert the signal current from working electrode (WE) into a voltage. This voltage is read by the external 16-bit ADC [66], that provides greater quantization levels over the internal 12-bit ADC on the ESP-32 and many custom built potentiostats in the literature [5,73,74], it provides equal quantization to one report by Cruz et al. [6], however not as powerful as one report by Dryden et al. [8] compromise being cost. The 16-bit ADC uses a signed integer [75]; therefore 215 = 32, 768 quantization levels with a sensitivity of 5 V/32768 = 0.15 mV in the voltage input range. The TIA circuit is duplicated allowing four channels whose timing is synchronized by the programmed ESP-32. The wireless potentiostat part of the iMED is 8 × 11 × 1.85 cm in dimensions, which is significantly smaller than the commercial CHI 1030 (36 × 24 × 12 cm) used to benchmark the iMED. This represents a significant percentage difference in volume cm3 (193%) in support of the iMED.</p><!><p>The ESP-32 controlled microfluidic platform configuration consists of one syringe pump (Aladdin AL-1000), three assembled servo-valves and four detection chambers (Fig. 1). Each electromechanical servo-valve is connected to the microcontroller directly using general-purpose input/output (GPIO) pins and powered by the 5 V rail of the ATX power supply. The ESP-32 is a versatile, low cost, standalone WiFI based microcontrolled platform with a small footprint (55 × 28 × 12 mm). This size makes it suitable for controlling many different types of microfluidic platforms. Alternatives to the ESP-32 include, Arduino, Teensy, Raspberry Pi and BeagleBone microcontrollers/processors that have been used in analytical chemistry platforms [33,34,36,67,68,76–81], although most do not have any kind of wireless connectivity, and the latter two are not cost-effective. For cost-effective comparisons of various microcontrollers used within the literature, see Supporting Information, Section 1.2. The syringe pump uses serial communication. Therefore, the syringe pump is connected to the ESP-32 via a MAX232 integrated circuit that converts signals from the syringe pump RS-232 serial port to signals suitable for use in TTL-compatible ESP-32 (Section 1.5 of the Supporting Information). The ESP-32 sends serial commands to the syringe pump at a default baud rate of 19200 bps, although the syringe pump can be programmed to different rates if required. The main function of the microfluidic part of the C/C++ program is to actuate the servo-valves at defined times and initiate the syringe pump at specified flow rates.</p><!><p>Servos are secured to valves using medical grade epoxy glue, each servo-valve can rotate 180° and has two flow path options (Fig. 2). Each actuator is controlled by the C/C++ program uploaded to the iMED. In order to change direction of fluid, a square wave of 5 V amplitude is applied to the actuator using the program-controlled ESP-32 (Section 1.3 of the Supporting Information). The first actuated valve in the sequence directs flow to either detection chambers 1 and 2 or 3 and 4, while the second and third actuated valve controls flow into 1–4 detection streams (Fig. 1). Each servo-valve has a reported low swept-volume, can withstand pressures up-to 500 psi (34 bar), is chemically inert and biocompatible [82]. Actuation is reported to occur over a short period of 150 ms through 60° at an applied potential of 6 V, or 190 ms through 60° at an applied potential of 4.8 V [83]. Actuation speed is therefore sufficiently fast enough for microfluidic direction change.</p><!><p>Placement for the working electrode (WE), counter electrode (CE) and reference electrode (RE) is inside a PDMS slab (created using a machined mold) with a 1.5 mm wide rectangular channel, creating four 63 μl electrochemical detection chambers previously reported [84]. The WE is a Kanichi array with a format of 8 carbon screen printed electrodes (dia. 850 μm). Kanichi electrode arrays have been used previously in biosensor research [84–86]. Each of the four Kanichi arrays used are short circuited to provide one WE for each channel of the potentiostat. All samples are injected manually into the microfluidic system using a 100 μl loop and Rheodyne 7725i sample injector, except Section 3.1.2 where a traditional beaker electrochemical cell is used.</p><p>Enzyme electrode assembly is achieved by depositing appropriate volumes from 5 mg ml−1 redox polymer aqueous solution, 10 mg ml−1 FADGDH aqueous solution, 46 mg ml−1 aqueous dispersion of acid-treated MWCNTs and 15 mg ml−1 PEGDGE aqueous solution on the surface of the Kanichi working electrode and allowing the deposition to dry for 24 h.</p><!><p>Ohms law states there is a linear relationship between voltage and current. Resistors are ohmic materials that are useful in the analysis of electrical circuits. Ohm's law is satisfied when the voltage vs current plot for a resistor of known value is a straight line through the origin. This known response is tested with various resistors and is plotted using data acquired from a commercial CHI 1030 (CH Instruments) potentiostat and the iMED: R2 = 0.999 for 5.1 kΩ, 7.1 kΩ, and 10 kΩ, Fig. 3. These results compare well to other reports in the literature [5,73,74], and suggest a strong positive linear relationship between the iMED and the CHI 1030 in the microamp scale throughout the 2 V ± 200 μA operation range.</p><!><p>Once the resistor polarization curve verified the iMED circuit was working as expected, the device was tested to validate its functionality. It was observed that the device effectively performed full range CV measurements on three-electrode set-ups in the voltage range between 1 and −1 V. The data obtained is stored in the internal memory of the microcontroller. Using both CPU cores, the C/C++ program is designed in such a way that voltage and current data is stored locally and then exported via WiFI upload to a ThingSpeak (MathWorks) web server to be stored as a comma separated values text file.</p><p>For validation, cyclic voltammetry (CV) studies of platinum working electrode with a diameter of 1.6 mm were performed in potassium chloride (0.1 M) containing 19.35 mM potassium ferricyanide using the CHI 1030 and the iMED. Potassium ferricyanide in this concentration range is consistently reported in the literature to validate potentiostats of this type [6,5,10].</p><p>The four channels of the iMED were also compared. Each channel is subjected to the same parameters with the scan rate (100 mV s−1) and the detection limits (−0.2 ≤ Vwe−re ≤ 0.8) for both experiments. The resulting plots were compared, and it was found that both CV curves exhibit well-defined oxidation and reduction peaks of the redox moieties with identical magnitude of current response (1% variation in magnitude of current response) Fig. 4. Redox potentials are also compared to the CHI 1030 and show a 1.25% variance in redox potentials, 0.219 V and 0.222 V, iMED and CHI 1030 respectively, Fig. 5.</p><!><p>Reproducibility within a single electrochemical detection chamber was tested using K3[Fe(CN)6] by running amperometry (reduction current) and injecting various concentrations of K3[Fe(CN)6] (2.5–15mM) in triplicate. The percentage relative standard deviation of the measurements for the 4 concentrations is, 3.66, 2.21, 0.89, 1.39 (% RSD), respectively. The equation of the line and correlation coefficient is y = 1.26162x + −1.9174 R2 = 0.98 and the standard deviation is indicated by error bars, (Fig. 6), confirming good peak-to-peak reproducibility of the potentiostat operating in amperometry mode. This data is in strong agreement with data from a similar experiment used to validate an open-source potentiostat [74]. The Supporting Information, Section 5.9 provides the source code for performing MATLAB graphical visualization of amperometric data to be viewed on an internet connected smartphone.</p><!><p>Automation of syringe pumps eliminates manual operational pump control and can achieve complete governance over flow rates and times in programmed biosensor experiments. While computer software is available to control syringe pumps at a cost [87,88], desktop computer connections diminish point of care applications. We report here, for the first time to our knowledge, microcontrol of a common Aladdin AL-1000 syringe pump using the ESP-32 microcontroller. To this end, the syringe pump is connected to the microcontroller via a MAX232 chip (see Section 2.6). The MAX232 is necessary to convert TIA-232 (RS-232) syringe pump serial port to signals suitable for use with the TTL-compatible digital logic circuits on the ESP-32, see Section 2.6. To test the hardware connection a terminal emulator program (RealTerm, on Microsoft Windows) is setup with RS-232 protocol (Section 1.5 of the Supporting Information) and used to communicate with the pump in hexadecimal commands via a PC. The command 'VER' is sent to which the pump responds with its model and firmware version. The benefit of initially using a terminal emulator to gain response from the pump is to verify that the hardware is working correctly. Syringe pump syntax is investigated using this terminal emulator concept. Development of the pumping sequence (Section 1.6 of the Supporting Information) is then coded onto the microcontroller. The pump and microcontroller maintain stable communication throughout 100 repetitions of a test pump flushing program, using water as a test medium. This programmed microcontrolled pump is capable of a reported flow rate range between 0.01 μL min−1 and 28.3 mL min−1, and can be expanded to achieve simultaneous control of multiple pumps connected in a network if required [89]. Back flow within microfluidic systems can cause possible reagent contamination [90], and because servo-valves can withstand pressures of 500 psi, any resulting back pressure due to excessive force needed to drive the syringe could stall the pump. To avoid this back pressure the microcontroller is not programmed to simultaneously close servo-valves and cease pumping at 2000 μL min−1 flow rates. Instead, by applying a square wave at 5 V for a defined time period, servo-valves close 10 s after terminating pump flow. By alternating square waves in this defined time sequence, back pressure opposed to the desired flow path that would otherwise result in a pump stall is avoided. Also, the fluid source, pump, and fluid outlet are maintained at the same height to minimize back pressure [91]. Connection to the microfluidic detection chambers is achieved using conventional 0.2 mm i.d. PEEK tubing.</p><!><p>For detection of glucose in solution, glucose concentrations were loaded into the sample loop, and injected into the detection chamber. Glucose oxidising currents were recorded by the iMED (Fig. 7) and the peak area currents increased linearly with physiologically relevant concentrations of 1 to 10 mM glucose (Fig. 8). Reproducibility was tested for glucose concentrations by running the experiment on different working electrodes with different Kanichi arrays. Standard deviation of the measurements for the 10 concentrations, indicated by error bars, confirmed good array-to-array and electrode-to-electrode reproducibility (Fig. 8).</p><p>Slow scan cyclic voltammograms in artificial plasma in the presence of glucose for films of co-immobilised with Os(bpy)PVI, PEGDGE, MWCNTs and FADGDH on Kanichi array electrodes show sigmoidal shaped curves, providing steady-state glucose oxidation currents above potentials for the Os(II/III) redox transition, indicative of EC' electrocatalysis of glucose, see Supporting Information, Section 3.</p><p>The redox potential of [Os(2,2′-bipyridine)2(poly-vinylimidazole)10Cl] is ~0.22 V [56]. Because the iMED works in the −1 V to 1 V ± 200 μA range, the device is suitable for detection of glucose oxidation currents using this electron mediator.</p><!><p>In this work, we designed, fabricated, and demonstrated a new wireless multi-potentiostat biosensor (iMED) performing integrated microfluidic electrochemical glucose detection. The main technical limitation, the potential window of the iMED compared to commercial potentiostat operational window (−1 V to 1 V), is overcome through the use of a tailored redox polymer, Os(bpy)PVI, within the potential range of the iMED. To assess the iMED performance, the system was compared with a commercial potentiostat commonly used in electrochemical research using K3[Fe(CN)6], with strong agreement between data obtained. Electrochemical cyclic voltammetry and amperometry is delivered in a simple, inexpensive, ambient lab environment that interfaces with a wirelessly programmed microcontroller that uploads all data to an open-source cloud-based server. The interface is comprised of a syringe pump and servo actuated valves that automate the flushing, washing and detection of glucose in solution. This control in microfluidic technology is applicable beyond glucose detection [11]. The system is sealed with conventional PEEK tubing and leak free flangeless fittings. This open-source approach could be implemented and adapted into modern bioelectrochemical laboratories in a variety of microfluidic experiments in a cost-effective, powerful, and wireless manner. Future works could focus on other biosensors based on similar enzyme-based designs that would utilise the multiple working electrodes, for example, multiplexed biomarker biosensors.</p>

PubMed Author Manuscript

First small molecular inhibitors of T. brucei dolicholphosphate mannose synthase (DPMS), a validated drug target in African sleeping sickness

Drug-like molecules with activity against Trypanosoma brucei are urgently required as potential therapeutics for the treatment of African sleeping sickness. Starting from known inhibitors of other glycosyltransferases, we have developed the first small molecular inhibitors of dolicholphosphate mannose synthase (DPMS), a mannosyltransferase critically involved in glycoconjugate biosynthesis in T. brucei. We show that these DPMS inhibitors prevent the biosynthesis of glycosylphosphatidylinositol (GPI) anchors, and possess trypanocidal activity against live trypanosomes.

first_small_molecular_inhibitors_of_t._brucei_dolicholphosphate_mannose_synthase_(dpms),_a_validated

<p>African sleeping sickness, also known as Human African Trypanosomiasis (HAT), is an infectious disease caused by the protozoan parasite Trypanosoma brucei (T. brucei). The parasite is transmitted by the bite of an infected tsetse fly, multiplies within the bloodstream of the mammalian host, and eventually invades the central nervous system. If untreated, HAT is invariably lethal. Despite becoming close to eradication by the late 1960s,1 the disease has made a dramatic re-emergence within the last half century, with prevalence in parts of Africa now as high as they were in the 1920s. In 2002 alone, around 48,000 deaths were reported.2 The WHO presently records around 17,500 new cases per year,3 with a cumulative rate of 50,000–70,000 cases3 and potentially over 60 million people at risk.2</p><p>The situation is further exacerbated by the growing resistance to established drug treatments for HAT, which have long been considered unsatisfactory.2 Serious side effects kill between 4% and 10% of patients who receive the arsenical melarsoprol,4 introduced in 1949, and the drug fails to cure between 10% and 30% of patients.5 The only relatively modern anti-HAT drug, the ornithine decarboxylase inhibitor eflornithine, is expensive, difficult to administer, and only effective against T. brucei gambiense.6 Thus, HAT has been described as one of the most neglected diseases of mankind,7 and novel therapeutic approaches to HAT are urgently needed.2</p><p>The T. brucei parasite is covered in a dense cell-surface coat of ~5 million variant surface glycoprotein (VSG) dimers8,9 which acts as a physical diffusion barrier for components of the innate immune system.10 The VSG coat undergoes constant antigenic variation,11,12 as the parasite switches between ~1000 immunologically distinct VSG genes,13 thus staying one step ahead of the host immune system. All VSG variants are linked to the trypanosomal plasma membrane via glycosylphosphatidylinositol (GPI) anchors (Fig. 1). Genetic14-16 and chemical17,18 studies show that GPI anchor biosynthesis is essential for viability of the bloodstream form of T. brucei, thus validating it as a drug target against HAT.19 Small molecular inhibitors of enzymes involved in GPI anchor biosynthesis therefore hold great promise as novel anti-trypanosomal agents.</p><p>GPI anchor biosynthesis in T. brucei involves three mannosyltransferases (MTs) on the luminal face of the endoplasmic reticulum, which catalyze the assembly of the trimannoside core structure (Fig. 1).20,21 All three of these MTs use dolicholphosphate mannose (Dol-P-Man) as their donor substrate, whose biosynthesis from dolicholphosphate (Dol-P) and the sugar-nucleotide GDP-mannose (GDP-Man) is catalyzed by another mannosyltransferase, dolicholphosphate mannose synthase (DPMS). All T. brucei VSG variants also contain at least one N-glycan, which requires another four Dol-P-Man-dependant mannosyltransferases for the formation of its lipid-linked oligosaccharide precursor.22-24 Consequently, T. brucei is doubly dependant upon Dol-P-Man for the synthesis of mature, N-glycosylated and GPI-anchored VSGs, and this double dependency makes DPMS an excellent target for inhibition of VSG biosynthesis. Recently, T. brucei DPMS has also been validated genetically as a drug target.25</p><p>Despite its promise as a therapeutic target, no inhibitors for T. brucei DPMS have been reported to date. The rational design of such inhibitors is complicated by the absence of a crystal structure for T. brucei DPMS at present. In search of a suitable lead structure for the development of DPMS inhibitors, we noticed striking structural similarities among small molecular inhibitors for other glycosyltransferases26-28 and sugar-nucleotide-dependent glycoprocessing enzymes.29-31 Several such inhibitors contain a rhodanine (2-thioxothiazolidin-4-one) scaffold, and derivatives of rhodanine-3-acetic acid 1 (Scheme 1) have been reported as inhibitors of the E. coli glycosyltransferase MurG26,27 and the C. albicans protein mannosyltransferase 1 (PMT1).28 It has been suggested that the thiazolidinone ring can act as a mimic of the pyrophosphate group,26,29-31 and that this mimicry may explain the inhibitory activity of thiazolidinone derivatives towards sugar-nucleotide-dependent enzymes. As DPMS is dependent on the sugar-nucleotide donor GDP-mannose, we reasoned that the thiazolidinone scaffold may also represent a good starting point for the development of DPMS inhibitors.</p><p>Herein, we describe the successful application of this strategy. We have prepared a small library of 5-benzylidene rhodanine-3-acetic acid analogs of the general structure 2, and report herein their inhibitory activity against T. brucei DPMS and GPI anchor biosynthesis as well as their trypanocidal activity against live trypanosomes.</p><p>The target rhodanine-3-acetic acid derivatives 2a–j (Scheme 1,Table 1) were prepared by Knoevenagel condensation of rhodanine-3-acetic acid 1 and substituted benzaldehydes 3a–j. To simplify the preparation and isolation of the target compounds, we explored different solvents and catalytic bases for this reaction, including DMF/sodium acetate, toluene/piperidine and ethanol/piperidine.28,32 In our hands, the ethanol/piperidine system was the most practical one, with short reaction times and straightforward product isolation. Under these conditions, all 5-benzylidene rhodanine-3-acetic acid derivatives precipitated from the ethanolic solution upon cooling to room temperature, and could be collected by simple filtration.33 Thus, all target compounds (Table 1) were obtained as yellow or yellow-orange solids in generally good yields. Remarkably, this procedure was also applicable to benzaldehydes containing a free phenolic hydroxyl group (e.g., 2f, 2g). This was particularly important as all attempts to prepare these analogs by debenzylation of the corresponding benzyloxy derivatives (e.g., 2b, 2c) had failed.</p><p>Fortuitously, our synthetic protocol provided benzylidene thiazolidinone 2b in crystalline form which allowed the unambiguous determination of its three-dimensional structure (Fig. 2).34 This is of particular interest, as in previous reports of structurally related thiazolidinones as glycosyltransferase inhibitors there had been ambiguity about the configuration of the exocyclic double bond, despite its obvious importance for structure–activity relationships.26,27 In our structure, the rhodanine acetic acid molecule 2b is essentially planar, with only the carboxy group out of plane and coordinating to one molecule of ethanol. The exocyclic double bond exists in the Z configuration, which for arylidene rhodanines has been reported as the thermodynamically stable configuration.35,36</p><p>In an initial biological screen, all target compounds were tested for inhibition of recombinant T. brucei DPMS in E. coli membranes (Table 1)37, At 1 mM, several thiazolidinone derivatives significantly inhibited T. brucei DPMS. A large benzyloxy substituent in position R2 and/or R3 appears to be advantageous for DPMS inhibition (2b–d), while a small polar substituent is less well tolerated in these positions (2f and 2g), as are rigid R3 substituents (e.g., nitrile 2h, acetylene 2j). Intriguingly, however, a polar substituent is beneficial for inhibitory activity when placed at the 2-position (R1), and the 2-hydroxy regioisomer 2e is among the most potent DPMS inhibitors in this series.</p><p>Next, compounds 2a–j were tested for their trypanocidal activity against cultured bloodstream form T. brucei. 38 Importantly, the in vitro activity of several thiazolidinones against DPMS did translate into in vivo trypanocidal activity against live trypanosomes. In particular, the 3-benzyloxy-substituted analog 2c and the 2-hydroxy regioisomer 2e showed trypanocidal activity with ED50 values in the medium micromolar range. However, only moderate activity was observed for some of the other potent DPMS inhibitors, notably the benzyloxy-substituted analogs 2b and 2d. The limited cellular activity of these DPMS inhibitors may be due to various factors, including limited cell penetration, which are currently being investigated.</p><p>In order to assess the parasite/host selectivity of these thiazolidinones, analogs 2a–f and 2j were tested at two concentrations (0.1 and 1.0 mM final) for their cytotoxicity against HeLa cells.38 Pleasingly, none of these thiazolidinones showed any cytotoxic effect, with the exception of 2b and 2j (20% inhibition at 1.0 mM, compared to DMSO-only control), suggesting low mammalian cytotoxicity (ED50 > 1.0 mM) and a promising margin of selectivity for T. brucei.</p><p>In order to gain further insight into the mode of action of these novel DPMS inhibitors, we investigated the effect of the thiazolidinones on GPI anchor biosynthesis in a T. brucei cell-free-system (Fig. 3).39 This assay monitors the DPMS-catalyzed formation of Dol-P-Man (lane 1) as well as the downstream formation of mannosylated GPI intermediates (lane 2). As expected, the potent DPMS inhibitor 2d abolished the formation of Dol-P-Man almost completely, and significantly reduced the formation of downstream GPI intermediates. A similar effect was observed for the analog 2f (Fig. 3).</p><p>Interestingly, thiazolidinones 2b and 2e did not affect Dol-P-Man production in the cell-free system, but did potently inhibit the formation of mannosylated GPI intermediates. Thus, the strong inhibitory effect of 2b and 2e on GPI anchor biosynthesis in the cell-free system may be due to inhibition of one or more GPI biosynthetic enzymes early in the pathway, for example, the GlcNAc transferase, de-N-acetylase or the first mannosyltransferase (MT1). Importantly, all of these enzymes represent therapeutic targets in their own right, and their inhibition may contribute to the trypanocidal activity of 2b and 2e.</p><p>In summary, we have identified the first small molecular inhibitors of trypanosomal DPMS. We demonstrate that DPMS inhibitors compromise trypanosomal GPI anchor biosynthesis in a cell-free system, and that inhibition of DPMS is a promising strategy for the development of anti-trypanosomal agents. Work on the optimization of these initial lead compounds with regard to their DPMS inhibitory potency as well as cellular activity is ongoing. Thiazolidinones 2b, 2d and 2e in particular are promising candidates for further development because of their respective activities against trypanosomal DPMS and GPI anchor biosynthesis.</p>

PubMed Author Manuscript

New approaches for extraction and determination of betaine from Beta vulgaris samples by hydrophilic interaction liquid chromatography-tandem mass spectrometry

Betaine is one of most studied biologically active compounds, due its role in the main biological processes. Although it may be found in several plants and roots, such as the Beta vulgaris family, present in typical diets, just a few analytical methods have been developed for its extraction from roots. A new, quick and effective procedure for the isolation and determination of betaine from two different varieties of B. vulgaris (red and gold) is presented. For betaine extraction, an accelerated solvent extraction (ASE) was coupled with solid-phase extraction. For betaine determination, a separation method based on hydrophilic interaction chromatography coupled with tandem mass spectrometry was optimized for a sensible detection of betaine by means of experimental design. Recoveries were about 93%, with RSD <5%, for both the matrices, without evidence of interfering species. The total content of betaine in extracts of various parts of plants (juice, peel, root) have been determined, obtaining concentrations in the range 3000–4000 mg/L for the juice and in the range 2–5 mg/g for the pulp and for the peel. The B. vulgaris gold species exhibited a higher concentration of betaine, compared to the red variety. Additionally, a micro extraction by packed sorbent technique and a modified quick, easy, cheap, rugged and safe (QuEChERS) procedure, were also tested and compared. Despite the lower recoveries of the latter, with respect to the ASE/SPE procedure (75–89%, RSD <1.5%), the ease of the method, which can be applied without the SPE purification procedure, can represent a positive improvement. Graphical abstractDetermination of betaine from Beta vulgaris samples.

new_approaches_for_extraction_and_determination_of_betaine_from_beta_vulgaris_samples_by_hydrophilic

<!>Introduction<!>Chemicals<!>Samples<!>Instrumentation<!>Optimization of MS parameters through experimental design<!>Betaine extraction<!>Solid-phase extraction<!>Accelerated solvent extraction<!>Micro extraction by packed sorbents<!>QuEChERS<!><!>Optimization of HILIC-MS/MS conditions through DOE<!><!>Optimization of HILIC-MS/MS conditions through DOE<!>Validation of HILIC-MS/MS technique<!>Optimization of the extraction protocol<!>Extraction of betaine from liquid matrices<!><!>Analysis of betaine from juice<!><!>Extraction of betaine from solid matrices<!>Evaluation of betaine content in B. vulgaris<!><!>Evaluation of betaine content in B. vulgaris<!>QuEChERS procedure<!>Conclusions

<p>Determination of betaine from Beta vulgaris samples.</p><p>Acid dissociation of betaine</p><!><p>Betaine is an oxidative derivative of choline. In the human body, it could be introduced by the diet, directly or from dietary choline [3].</p><p>Among the different, well documented roles of betaine, the following ones are the most relevant: (i) organic osmolyte, (ii) methyl donor for the remethylation of homocysteine, (iii) protector against alcohol-induced liver injury and (iv) biologically important in cancer development [4, 5].</p><p>Betaine is a promising agent that attenuates homocysteine rise after meals. Therefore, a diet rich in betaine (or choline) might benefit cardiovascular health through its homocysteine-lowering effects. However, some studies show how high betaine intakes could increase serum lipid concentrations, which of course increases the risk of cardiovascular disease.</p><p>As an organic osmolyte, betaine could be found in several plants and roots, which accumulate this highly soluble compound in response to water stresses. Among them, roots from the Beta vulgaris varieties (Chenopodiaceae family), which are present in typical food diets, are rich in this compound. One of the most well-known and consumed beetroots is the red B. vulgaris. Another diffuse variety is the golden (or gold) one, which has recently been investigated also as source of dyes, due to the high demand of yellow dyes [6].</p><p>Betaine extraction from food samples is mainly performed by liquid-liquid (LL) and solid-liquid extractions (SLE), depending on the nature of the sample. Methanol, dichloromethane and chloroform, sometimes partially mixed with water, were used as solvents for cold extraction [7, 8], while methanolic KOH was used together with a boiling Goldfisch apparatus [9]. High-performance liquid chromatography (HPLC) coupled with tandem mass spectrometry (MS/MS) detection [7], UV (using pre- or post-derivatization) [10] and light scattering detection (LSD) [8] were used for the analytical determination. Ion chromatography in the non-suppressed conductivity mode was also employed for analysis [11]. These methods, which were applied for the determination of betaine in feed additives, fruits, seeds or plasma samples, are characterized by limits of detection (LODs) which depend mainly upon the detection mode and the extraction technique used. For example, LODs ranging from 2 to 5 mg/kg were obtained using SLE extraction coupled with HPLC-MS/MS and non-suppressed ion chromatography (IC), respectively. Higher (150 mg/kg) limits were achieved by HPLC-LSD methods [8].</p><p>Beside the fair LODs, the above-mentioned extraction techniques are all non-selective for betaine and consequently co-extracted interfering species, affect the sensitivity of the technique. Accordingly, an enhancement in the selectivity of the extraction procedure is desirable in order to improve the detection limits of the entire method.</p><p>In light of the above-mentioned considerations, the aim of the present work was the optimization of a quick, easy and selective method for the determination of betaine in food samples. Recently, accelerated solvent extraction (ASE) has been proposed for the extraction of target polar compounds from water-rich food samples [12, 13]. For the first time, ASE was here coupled with solid-phase extraction (SPE) or micro extraction by packed sorbent (MEPS) for the extraction of betaine from B. vulgaris samples. Betaine was quantified by hydrophilic interaction chromatography coupled with tandem mass spectrometry (HILIC-MS/MS): the optimization of response, as a function of detector parameters, by experimental design allowed us to obtain improved LODs. Moreover, for the first time, a quick, easy, cheap, rugged and safe (QuEChERS) procedure was successfully optimized through the evaluation of different dispersive-SPE (d-SPE) sorbents and applied for a quick and easy isolation of betaine. Finally, the ASE/SPE method, previously optimized, was applied to the red and gold varieties, determining the distribution of betaine in each portion of the beetroots (peel, pulp and juice) and between the species considered.</p><!><p>All reagents used were of analytical grade. Betaine was purchased from Sigma-Aldrich (Chemie, Steinheim, Germany), as well as ammonium formate, sodium acetate, sodium hydrogen phosphate, acetonitrile, methanol and hydrochloride acid, which were used for eluent preparation and for extraction procedures. For the QuEChERS procedure, magnesium sulphate and sodium chloride were from UCT (Bristol, USA); PSA sorbent from J.T. Baker and graphitized black carbon from Supelco (Bellefonte, USA). High-purity water (18.2 MΩ cm resistivity at 25 °C), produced by a Milli-Q system (Millipore, El Passo, TX, USA), was used.</p><!><p>The two different varieties of beetroot studied, B. vulgaris red and golden, were acquired from local markets. Each sample was peeled and the vegetable was ground in order to obtain a homogenized and soft pulp. Finally, the pulp was filtered in order to separate it from the juice. The samples thus prepared were stored in a freezer at −10 °C until analysis.</p><!><p>An Agilent 1100 liquid chromatograph (Agilent, Waldbronn, Germany), equipped with a diode-array detector and an Agilent 6410 Triple Quad mass spectrometer (Agilent, Waldbronn, Germany), was employed for analysis. Based on the high polarity of betaine, a hydrophilic interaction chromatography column (HILIC) was used for its analysis. A HILIC Kinetex column (100 mm × 4.6 mm, 2.6 μm) (Phenomenex, Torrance, CA, USA) and a mobile phase (flow rate 0.6 mL/min) composed of acetonitrile (A, 75% v/v) and 10 mM ammonium formate buffer (B, 25% v/v), acidified to pH 3 with formic acid, in isocratic elution mode were used. Injection volume was 5 μL. Data were collected with the use of Agilent Mass Hunter, software version B.04.01.</p><p>Electrospray ionization (ESI) was applied in the positive ion mode. Nitrogen was used in the ion source and the collision cell. Full-scan mass spectra were recorded within the mass range of m/z 50–500 Da. Additionally, multiple reaction monitoring (MRM) mode was applied for the quantitative analysis. Three main operation parameters of MS/MS detector, namely collision energy (CoE), fragmentor voltage (FV) and temperature of the source (Temp), were optimized by means of a Central Composite Design (see 'Optimization of MS parameters through experimental design' section).</p><!><p>Design of experiments (DOE) is a useful approach for the optimization of the performance of a system with known variables. In this work, a central composite design (CCD) was used to estimate constant, linear terms, interactions between the different variables and quadratic terms [14], as indicated by the following model, in which in case of more than two variables the upper-level interactions are not taken into account):</p><p>\documentclass[12pt]{minimal} sepackage{amsmath} sepackage{wasysym} sepackage{amsfonts} sepackage{amssymb} sepackage{amsbsy} sepackage{mathrsfs} sepackage{upgreek} \setlength{\oddsidemargin}{-69pt} egin{document}$$ Y={b}_0+{b}_1{X}_1+{b}_2{X}_2+{b}_3{X}_3+\left({b}_{12}{X}_1{X}_2+{b}_{13}{X}_1{X}_3+{b}_{23}{X}_2{X}_3 ight)+{b}_{11}{X}_1^2+{b}_{22}{X}_2^2+{b}_{33}{X}_3^2 $$\end{document}Y=b0+b1X1+b2X2+b3X3+b12X1X2+b13X1X3+b23X2X3+b11X12+b22X22+b33X32</p><p>In this study, the response Y, was the betaine peak area while X 1, X 2 and X 3 were CoE, FV and Temp, respectively. b i are the coefficient of the linear term, b ij are the coefficients of the interactions and b ii are the coefficient of quadratic terms. Parameters levels and statistical data treatment are detailed in the 'Optimization of HILIC-MS/MS conditions through DOE' section.</p><!><p>For the extraction and the purification of betaine from beetroot samples, different approaches were evaluated and compared. For liquid samples, such as the beetroot juice, SPE or micro extraction by packed sorbents (MEPS) were used, since they ensure extraction and clean-up in the same step. For the solid portions of B. vulgaris (peel and pulp), ASE was used for the extraction of betaine from the matrices, while SPE and MEPS procedures were used to clean-up betaine from the co-extracted species.</p><!><p>For the solid-phase extraction (SPE) procedure, the performance of three sorbents (3 mL, 500 mg) was compared. In detail, a strong cation exchanger resin, Bakerbond SCX aromatic sulfonic acid (Agilent), a mixed mode hydrophilic/lipophilic polymer based resin, OASIS HLB (Waters) and a silica gel sorbent, Discovery SPE Pure Silica (Supelco), were used.</p><p>Each cartridge was conditioned with different solutions, depending on the type of adsorbent: (i) SCX was conditioned with 3 mL of water, followed by 3 mL of water acidified to pH 2 with HCl; (ii) HLB with 3 mL of methanol followed by 3 mL of water, while (iii) pure silica with 5 mL of methanol.</p><p>For each cartridge, aliquots of 4-mL solutions containing 50 μg/L of betaine were loaded with a flow rate of 1–2 mL/min flow rate. The eluents tested in the recovery step were chosen according to the expected analyte–sorbent interactions. Before elution, each cartridge was washed with 2 mL of the proper solvent (H2O acidified to pH 0.5 for SCX cartridge, H2O for polymeric sorbent and methanol for pure silica resin) to remove unretained compounds. To check the retention of the analyte, the following fractions were collected during each step of the SPE protocol and then analysed by HPLC-MS/MS: (i) the solution after loading, (ii) the washing solution and (iii) the eluate.</p><!><p>Dionex ASE 200 (Sunnyvale, CA) accelerated solvent extractor, equipped with 33-mL stainless steel cells and 60-mL collection vials was used. Extraction was performed mixing roast sand with the beetroot peel or pulp samples; two different extraction solvents were used (methanol and water-methanol, 50:50% solution). 1 g of sample was mixed with 3 g of sand and 3 cycles of 5 min with 4 mL each of extraction solvent were performed. Based on the only other paper present in literature (in which however betaine was extracted from algae) the temperature applied was 50 °C [15]. After extraction, the extract was diluted 1:5000 with methanol and purified with the optimized SPE procedure, using pure silica cartridges (see below).</p><!><p>The evaluation of MEPS extraction/purification technique was performed using the sorbent which turned out to be the best one in terms of recovery during the previous SPE extraction of betaine. MEPS sorbent was a pure silica adsorbent placed in a cartridge, located in the needle and it was obtained from SGE Analytical (Melbourne, Australia).</p><p>The main steps of the MEPS extraction are as follows: (i) conditioning of the cartridge with 200 μL of methanol, (ii) loading of 50-μL sample in the syringe, (iii) removal of the needle and the cartridge, (iv) drying of the cartridge with air, (vi) loading of the eluent solution (50 μL) and (vii) removal, again, of needle and cartridge, pushing the solution in a vial for analysis.</p><!><p>To propose a quick procedure to extract and isolate betaine that does not necessarily require two different purification approaches, a QuECheRS method was developed. In detail: 1 g of B. vulgaris peel was placed into a 50-mL centrifuge tube; 20 mL of water-methanol solution (1:1) were added, together with 4 g of MgSO4 and 1 g of NaCl. The tube was shaken for 1 min and next centrifuged at 6.000 rpm for 5 min (extraction step). Subsequently, 5 mL of the supernatant were transferred to a vial containing 150 mg PSA and 50 mg graphitized carbon black (GCB); the vial was shaken again for 1 min (clean-up step). The tube was centrifuged for 10 min, 10,000 rpm, to completely separate the dispersive phases. Finally, 1 mL of the supernatant was transferred to the vial for the chromatographic analysis.</p><p>For all the techniques evaluated, recoveries are expressed as the average of three independent extractions; in parallel, a blank was processed. Recovery for each protocol was assessed using HILIC-MS/MS technique.</p><!><p>Extraction protocol optimized and applied for betaine determination in beetroot pulp, peel and juice</p><!><p>Due to the polarity of betaine, the HILIC mode was selected for the analysis of the analyte. For the separation, silica gel was used as the most commonly used stationary phase in HILIC mode. Betaine was expected to be retained on the stationary phase due to polar interactions, e.g. hydrogen bonding and electrostatic interactions. In order to improve peak symmetry and to increase sensitivity, a buffer was introduced in the mobile phase. The final mobile phase composition consisted of 75% v/v of acetonitrile and 25% v/v of 10 mM ammonium formate buffer (pH 3). Acetonitrile-rich mobile phase, as known, are beneficial for MS sensitivity. Despite the slight peak width broadening observed with acid eluents, acid conditions were preferred to enhance MS sensitivity in the positive ion mode. Under the above-mentioned elution condition, the analysis of betaine can be performed within 6 min.</p><p>A full-scan mass spectrum of betaine was recorded obtaining the two parent ions at m/z 118.1 [M + H]+ and m/z 235.1 [2 M + H]+, corresponding to the molecular mass of betaine (117 Da). Additionally, one signal was observed for product ion spectra at m/z = 58.1 Da. On this basis, the transition 118.1 → 58.1 Da was used to quantify betaine in the MRM mode. Since it is well established that the MRM sensitivity depends drastically on the tuning of instrument parameters [16], an optimization of collision energy (CoE), fragmentor voltage (FV) and temperature of the source (Temp) was performed by means of a CCD. The response variable was the betaine MRM peak area. It is worth to be mentioned it is well known that the expected concentrations of betaine in beetroots samples are not in trace levels. However, the effort spent in optimizing the MS/MS parameters by the experimental design technique is justified by possibility to apply this chromatographic method also to other applications, quantifying matrices with trace amounts of betaine (which are several, as listed in the work of Slow et al. [17]).</p><p>To investigate the effect of each of the above-mentioned factors and to subsequently optimize the response, the (−1) and (+1) levels were the following: CoE 2 and 40 eV; FV 105 and 165 V and Temp 235 and 345 °C. At the central level (30 eV, 85 V and 55 °C), three replicates were performed.</p><!><p>Graphical representation of the coefficients obtained for the optimization of MS/MS detector parameters. 1, 2 and 3 (corresponding to b 1, b 2 and b 3) refer to linear terms of CoE, FV and Temp, respectively, while 4, 5 and 6 (corresponding to b 11, b 22, b 33) refer to quadratic ones</p><p>Graphical representation of peak area as a function of collision energy and temperature of the Source at FV = 105 V</p><!><p>Following the assumption that FV has a negative correlation with MRM peak area (higher values means lower sensitivity) and that using too low FV values fragmentation is not possible, the response graph was obtained keeping constant FV to its lower coded value −1 (corresponding to 105 V). Lines in Fig. 3 represent the iso-curves, which contain all the experimental conditions providing the same response. As shown, experimental conditions that lead to the highest peak area can be easily derived. The following conditions CoE = 40 eV, FV = 105 V and Temp = 345 °C were considered optimal and experimentally tested. The results obtained showed an increase of about 186% of betaine peak area if compared to initial conditions, thus confirming the hypothesis derived by the DOE.</p><!><p>Validation of the HILIC-MS/MS technique was performed investigating the main performance parameters of analytical method validation suggested by European Union (EUR-FA Guide, Annex I) and IUPAC guidelines [18], such as linearity, limits of detection, limits of quantification and instrument reproducibility.</p><p>The calibration curve for betaine was linear over the concentration range 5–500 μg/L. Standard curve used in this study was determined using the following linear regression: y = 239.02x − 19.45 (R 2 = 0.99992).</p><p>The linearity of the HILIC-MS/MS method was verified over two orders of magnitude with root mean square error (RMSE) for calibration equal to 65.80 for HILIC (related to chromatographic peak areas of 104/105 order of magnitude, respectively). Limit of detection (LOD) and limit of quantification (LOQ) were evaluated as LOD = 3 × SDxy/b and LOQ = 10 × SDxy/b (where SDxy is the standard deviation of the response and b is the slope of the calibration curve) [19]. There were found to be 0.95 and 2.87 μg/L, respectively.</p><p>The detection limits obtained for the HPLC-MS/MS in HILIC mode are lower (about 10 times) if compared with the ones obtained in literature [7, 20] using the same detector, mainly due to the CCD optimization of the main detector parameters. It should be remarked that very few approaches for determination of betaine by detectors different from MS are available. Shin et al. [8] propose betaine determination by evaporative light scattering detection at concentration levels higher than 1 mg/L, three orders of magnitude higher than the approach here proposed.</p><p>Inter- and intra-day reproducibility of the instrument (expressed as relative standard deviation, % RSD) were evaluated both for retention times and peak areas. A standard solution of betaine (5 μg/L) was injected repeatedly in the same day and for 2 weeks: the intra-day RSD obtained was 1.4% for retention times and 2.1% for the peak area, while the inter-day values were 2.8 and 4.1%, respectively, supporting the robustness of the chromatographic technique.</p><!><p>In order to extract betaine from beetroot samples and to isolate the molecule from interfering species, an extraction procedure was optimized comparing the performance of different approaches. In previous studies showing the determination of betaine in numerous food and vegetable matrices, solid samples were simply mixed with water, homogenized and centrifuged to recover the aqueous supernatant. The aqueous supernatant was then extracted with dichloromethane which removes hydrophobic compounds without removing betaine. Also, in case of liquid samples, they were shaken with an equal volume of dichloromethane, centrifuged and the resulting aqueous layer directly employed for analysis [17, 21]. However, these procedures are not selective towards betaine, and high hydrophilic and polar compounds are dissolved in the aqueous phase which is injected and analysed. As well, described by Wruss et al., beetroots and their derivative products (as for examples juices) have high concentration of sugars (glucose, fructose and sucrose, from 1.5 to 73.5 g/L for 15 mL of homogenized sample) and betalains that serve as colour pigments (mean concentration of 1.1 g/L). All the above-mentioned compounds are possible interfering species that should be removed and, therefore, the optimization of purification method to isolate betaine is discussed in the next paragraphs.</p><p>As previously mentioned, betaine could be present in both liquid (juice) and solid (peel and pulp) matrices. For the latter, extraction and clean-up steps are necessary, while for liquid samples, both the steps could be done within the same technique.</p><!><p>An SPE procedure was optimized comparing the performance of three different sorbents, see Table 1. As shown, all the tested sorbents interact with the analyte; retention was quantitative for SCX and pure silica cartridges, while a decreased but still significant retention of betaine was observed for the HLB sorbent (75%). The analysis of the washing solutions, before the elution step, showed no evidence of betaine. Since we observed that when using pure silica cartridge, betaine has to be dissolved in methanol; otherwise, interactions with sorbent are weak and low recoveries are obtained, this solvent was used during the loading step. This feature must be taken into account also for the selection of the extraction solvent for the ASE procedure (see 'Extraction of betaine from solid matrices' section).</p><p>For the elution of the betaine, different solvents (compatible with the HILIC-MS/MS analysis) were tested, depending on the physico-chemical characteristics of the sorbent and on the hypothesised interactions. Aromatic sulfonic acid groups present in the SCX interact with the positively charged amino group of betaine by electrostatic interactions. Consequently, eluents that can give rise to competitive interaction were evaluated. 1 M HCl and three different buffers, formate, acetate and phosphate, at different pH (2, 4, 8, respectively) at increasing concentrations (20, 50, 100 mM) were tested. At the tested conditions, elution recoveries ranged from 5 to 10%.</p><p>The elution from the polymeric OASIS HLB cartridge was evaluated with 4 mL of acetonitrile and a solution 50:50 v/v acetonitrile/water or water, which were not effective in the recovery of betaine.</p><p>The expected interaction between the silica sorbent and betaine is based on polarity and adsorption on hydrogen bonding. For the elution of betaine from silica, 2 aliquots of 2 mL of water were used for recovery, finally reaching a quantitative release of the analyte.</p><!><p>Recovery yields obtained for each tested substrate. SPE conditions: Sample: 50 μg/L betaine in 4 mL of water solution; elution volume: 4 mL. MEPS conditions: 10 μg/L betaine in 50 μL water solutions; elution volume: 50 μL. For activation and recovery procedures, see text</p><p>Elution solvents: a1 M HCl, bWater, cWater, dWater</p><!><p>In order to evaluate the efficiency of the SPE technique on real matrices, the procedure was used for the determination of betaine in beetroot juice. The study has been performed on both the red and golden beetroot juices. Recoveries were 92.7% (red) and 93.3% (golden), with high accuracy and repeatability (RSD <1.2%). Lower recovery values, compared to standard solution samples, could be ascribed to competition effects that occurred between the matrix of the beetroot juice and the silica sorbent. The same extraction was performed with optimized MEPS procedure, which confirmed slightly lower recoveries (93%) than SPE.</p><!><p>Analysis of the eluted fraction after SPE extraction of betaine from Beta vulgaris golden juice. Betaine full-scan spectrum and product ion spectrum (a and b, respectively), and chromatogram overlay of the betaine signal obtained both in MRM and full-scan mode (c). Instrumental conditions are detailed in the 'Instrumentation' section</p><!><p>Since it was not possible to evaluate the extraction efficiency on a blank solid sample, tests were performed comparing betaine concentration in pulp and peel of red and golden beetroots with the concentration of spiked samples.</p><p>An ASE procedure was used, comparing the performance of two different extraction solvents: a methanol and a solution of water-methanol 50:50% v/v. Even if both solvents lead to extraction yields of about 95%, methanol was selected since it is fully compatible with subsequent SPE step using pure silica cartridge (as previously optimized and described in the 'Extraction of betaine from liquid matrices' section). For both pulp and peel matrices, high recoveries (from 94.4% of red beetroot's peel to 96.5% of red beetroots pulp, RSD <1.05%) were achieved, confirming a high reproducibility and accuracy of the procedure. To the best of our knowledge, this is the first time that an ASE procedure is proposed for the extraction of betaine from beetroots.</p><p>Details of the optimized extraction procedure are summarized in Table 1, see 'Optimized extraction conditions' section, together with the ones optimized for liquid matrices. The overall ASE/SPE procedure takes advantages of both the high-pressure conditions of ASE, which in turn provide faster extraction times than traditional extraction apparatus [9] (12 min against 3 h), and of the higher safety conditions of SPE purification which is based on water and methanol, rather than on more toxic organic solvents (e.g. dichloromethane [10]).</p><!><p>Once optimized, the procedure for the extraction of betaine from both the solid and the liquid parts of the beetroots, a quantitative determination of this compound was performed in real samples. In literature, many studies determine the content of betaine in foods, especially in sugar beet and molasses, but this is the first time in which betaine content is specifically quantified for each portion of B. vulgaris red and gold.</p><!><p>Comparison of the betaine concentration level in all the three portions (juice, peel and pulp) of Beta vulgaris red and golden. Extraction procedure is summarized in Table 2 and 'Experimental section'</p><!><p>Data have been compared with determinations already performed on sugar beets and on different species of the same family of B. vulgaris (Chenopodiaceae). It is interesting to highlight that B. vulgaris species have higher concentrations of betaine if compared with other entities that belong to the same family. As an example, betaine concentration is typically about 0.2–0.3% (2–3 mg/g) also in sugar beets [23], while in fresh spinach, which belongs to the same family, betaine is present at concentration 5 to 10 folds lower [10].</p><!><p>The extraction procedures optimized in the previous paragraphs allow us to extract betaine from all portions of B. vulgaris vegetables. However, for solid samples (as peel and pulp), this extraction is accomplished by two different, consecutives techniques: ASE for extraction and SPE for isolation of betaine. In order to obtain betaine extraction by unique approach, a QuEChERS procedure for the extraction of betaine from B. vulgaris samples was here evaluated for the first time.</p><p>The QuEChERS technique, developed for the extraction of pesticides from food [24], with extensive applications event to other analytes and matrices [25, 26] is an easy and fast procedure, since it is based on liquid extraction of target analytes, assisted by hand shaking and followed by a dispersive-SPE (d-SPE) clean-up step, using selective adsorbents.</p><p>A modified QuEChERS procedure was here tested on the peel of golden beetroot: this portion was chosen since, according to the previous results, it has the highest concentration of betaine. In spite of acetonitrile, used in the traditional procedure, methanol was selected as the extraction solvent, due to the high affinity towards betaine observed in the extraction techniques previously commented. Magnesium sulphate and sodium chloride salts were added in order to enhance the salting-out effect, promoting a better extraction of betaine. For the d-SPE step, first tests were performed using a combination of PSA and graphitized carbon black (GCB) resins. PSA is necessary to remove co-extracted sugars and weak organic acids, while GCB is necessary to remove dyes and aromatic interferences. Results obtained after HILIC-MS/MS analysis showed that a recovery of 75.3 ± 0.8% was obtained. Lower recoveries than the ASE/SPE procedure could be explained by betaine interaction with the GCB sorbent. In order to confirm the hypothesis, a betaine solution was prepared at the same concentration present in Golden beetroot peel and was put in contact with 50 mg of GCB, shaken for 1 min, centrifuged and analysed, obtaining an adsorption of betaine of 25.2%.</p><p>It should be mentioned that the replacement of GCB with a C18 resin was not effective in removing co-extracted dyes and additional purification would be necessary, losing the intrinsic advantages of QuEChERS technique.</p><p>In conclusion, the method detection limits (MDLs) obtained for betaine extraction from B. vulgaris peel by ASE/SPE and QuEChERS were found to be 4.1 (93%) and 11.86 (75%) μg/kg, respectively. Despite the lower recoveries of QuEChERS if compared to the ASE/SPE procedure, and slightly higher MDLs, the ease of the QuEChERS technique, which can be applied without the SPE purification procedure, can represent a positive improvement. Moreover, if compared to ASE/SPE, QuEChERS procedure is also time-saving (as described in the 'QuEChERS' section, since only 1 min of shaking is needed for the extraction and the d-SPE steps) and far cheaper in terms of laboratory equipment required (plastic tubes and a centrifuge system) than the ASE system.</p><!><p>This work presents the first report on the study and optimization of the extraction and determination of betaine from two different B. vulgaris varieties, red and gold. The innovative coupling of the optimized ASE and SPE techniques allows to successfully isolate betaine from complex matrices such as liquid and solid portions of beetroot. The solvents used for both the procedures are fully compatible with the following steps of the analytical method, avoiding reconstruction and evaporation steps which could reduce the recoveries of the overall method. Central composite design was applied for the optimization of a HILIC-MS/MS technique with better sensitivity than previously published works. Excellent extraction performances were achieved in terms of method sensitivity and robustness with recoveries in real samples as high as 93%.</p><p>Moreover, for the first time, a QuEChERS procedure was successfully tested: although recoveries achieved were slightly lower than ASE/SPE procedure (recoveries about 75%), the high reproducibility and ease, this approach can justify its use for a quantification of betaine in complexes samples, such as B. vulgaris.</p><p>Finally, a quantification of betaine in red and golden B. vulgaris varieties and in each portion of them was performed, obtaining high concentrations, especially in golden B. vulgaris, supporting the possible administering of these vegetables and their juice in diets of patients that exhibit deficiency of this compound.</p>

PubMed Open Access

Sub-nanomolar Detection of Prostate Specific Membrane Antigen in Synthetic Urine by Synergistic, Dual Ligand Phage

The sensitive detection of cancer biomarkers in urine could revolutionize cancer diagnosis and treatment. Such detectors must be inexpensive, easy to interpret, and sensitive. This report describes a bioaffinity matrix of viruses integrated into PEDOT films for electrochemical sensing of prostate specific membrane antigen (PSMA), a prostate cancer biomarker. High sensitivity to PSMA resulted from synergistic action by two different ligands to PSMA on the same phage particle. One ligand was genetically encoded, and the secondary recognition ligand was chemically synthesized to wrap around the phage. The dual ligands result in a bidentate binder with high copy, dense ligand display for enhanced PSMA detection through a chelate-based, avidity effect. Biosensing with virus-PEDOT films provides a 100 pM limit of detection for PSMA in synthetic urine without requiring enzymatic or other amplification.

sub-nanomolar_detection_of_prostate_specific_membrane_antigen_in_synthetic_urine_by_synergistic,_dua

INTRODUCTION<!>Genetically-Encoded, Phage-Displayed Ligands Targeting PSMA<!>Cycloaddition to Generate the Secondary Recognition Ligand<!>Phage Wrapping to Maximize Ligand Density<!>Primary and Secondary Recognition Ligands on Phage for Bidentate Binding<!>Biosensing with Virus-PEDOT Films<!>EIS to Quantify PSMA Binding<!>Calculating the Hill Coefficient<!>PSMA Detection in Synthetic Urine<!>CONCLUSION<!>

<p>More effective biosensors could address a critical need for detecting cancer-associated biomarkers. An estimated 29,000 men in the US will succumb to prostate cancer in 2013.1 Unfortunately, the lack of validated clinical diagnostic markers complicates efforts to develop tests for early prostate cancer detection. For example, a recent report concludes that the Prostate Specific Antigen (PSA) test used for prostate cancer diagnostics is more harmful than beneficial.2 Despite this caveat, PSA remains an important biomarker for detecting recurrent prostate cancer. However, early detection of the disease could enable more effective treatment and prognosis.3 Thus, issues addressable by bioanalytical chemistry include the development of more sensitive measurements of protein concentration and then applying such measurements to identify and validate more effective biomarkers.</p><p>Unlike PSA, Prostate Specific Membrane Antigen (PSMA) concentrations in biological fluids appear to offer a more useful metric for prostate cancer diagnosis and prognosis.4 For example, elevated PSMA levels have been observed in prostate cancer patients' urine.5 The PSMA concentration increases from 0.25 nM to approximately 3.5 nM in prostate cancer patients' biological fluids including urine.6 PSMA, a 750-residue, 90 kD glycoprotein, is overexpressed on the surface of tumor cells as a non-covalent homodimer in >94.3 and >57.7% of primary and metastatic prostate cancers respectively.7,8 Elevated PSMA levels also correlate with the aggressiveness of tumor growth.9 Thus, PSMA offers an important biomarker for the development of biosensor-based diagnostic devices. This report describes the development of a biosensor capable of detecting clinically relevant concentrations of PSMA (<0.25 nM) in synthetic urine.</p><p>In 2003, Petrenko and Vodyanoy demonstrated the use of whole virus particles as a bioaffinity matrix for biosensors.10,11 In an improved generation of biosensors, T7 virus particles with a peptide antigen from the West Nile virus on their surfaces have been incorporated into conducting polymers by Cosnier and coworkers to allow detection of antibodies to the West Nile virus.12 This strategy can offer higher density ligands for biomarker binding, as T7 phage have a high density of peptides displayed on their surface. Improving biosensor sensitivity through increasing the density of ligands on the phage surface inspired in part the approach reported here.</p><p>M13 bacteriophage, or more commonly "phage," serve as receptors for biosensors reported by our laboratories. Viruses that infect only bacteria, the M13 bacteriophage have a readily customized protein coat, which can be tailored to bind to cancer biomarkers.13 The M13 viruses have ssDNA encapsulated by approximately 2700 copies of the major coat protein (P8) and five copies each of the four minor coat proteins. Manipulating the encapsulated DNA can provide peptides and proteins fused to the phage coat proteins, which are displayed on the phage surface.13 Combinatorial engineering of such polypeptides allows molecular evolution to obtain displayed ligands with specific binding affinities and specificities.14,15</p><p>For direct electrical detection of biomarkers, M13 bacteriophage have been incorporated into films of an electronically conductive polymer, PEDOT (poly-3,4-ethylenedioxythiophene).16–20 Synthesis of the biosensor film is accomplished by electropolymerizing EDOT on the surface of a planar gold electrode from a solution that contains virus particles. During biosensor measurements, the electrochemical impedance of the virus-PEDOT film increases upon exposure to the biomarker, providing a quantifiable readout for analyte binding.21</p><p>Modifications to the biosensing films could further improve the device's limit of detection (LOD) for translational relevance. In a previous report, our labs described phage-incorporated into PEDOT nanowires, which resulted in biosensors with a >66 nM LOD for PSMA in synthetic urine.22 Conventional phage display results in a low density of genetically encoded ligands displayed on the surface of the phage. Here, we focus on increasing the density of such ligands, as a strategy for more sensitive measurements with higher signal-to-noise ratios. The concept of "phage wrapping" to improve ligand density builds upon our previous reports of wrapping the negatively charged phage surface with positively charged polymers to prevent non-specific binding to the phage.23,24 The approach takes advantage of the presence of negatively charged residues, one Glu and two Asp, on the N-terminus of each P8. Since each phage includes 2700 copies of P8, such carboxylate-bearing residues result in a high negative charge on the outer surface of the virus particle.25 As reported here, additional ligands wrapped onto the phage surface due to this electrostatic interaction, lead to enhanced affinity and selectivity for PSMA.</p><!><p>The two forms of PSMA, monomeric and dimeric, offer different targets for ligand binding; the dimeric form is overexpressed by prostate cancer cells, and the monomeric form offers a closely matched negative control for non-specificity, as a protein only found in healthy prostate cells.9 The relative binding affinities of two previously reported phage-displayed ligands,22 phage-1 and phage-2, (sequences and nomenclature in Table 1) for the PSMA isoforms were first examined by ELISA (Figure 1). Phage-2 binds with higher affinity to the PSMA dimer, than phage-1. Neither phage-displayed ligand binds with significant affinity to the PSMA monomer. Thus, the peptide ligands selectively bind the dimeric form of PSMA. The specificity of phage-displayed ligands for the dimeric PSMA is critical for potential clinical applications. Additional negative controls include phage-displayed peptides targeting the blocking agent (bovine serum albumin, BSA) and Stop-4 phage targeting PSMA; the latter phage includes an analogous phagemid packaged into phage without ligands displayed on their surfaces. As expected, the negative controls failed to show any significant binding.</p><p>The two PSMA ligands, 1 and 2, provide a starting point for the development of biosensors. For translational relevance, the LOD of the resultant biosensor must be <0.25 nM, and a high-signal-to-noise ratio is essential for definitive diagnosis. In theory, the affinity of the ligands for their target analyte should correlate with their usefulness in biosensor applications. For example, a higher affinity ligand could enhance sensitivity for the analyte. In a previous report, we described using phage-displayed homolog shotgun scanning to improve phage-displayed ligand affinity for PSMA by >100 fold.22 However, this approach requires extensive mutagenesis and selections.</p><p>Nature applies evolution-guided affinity maturation, but also relies on another approach for more rapid affinity maturation. The immune system, for example, applies the principle of avidity to boost the apparent affinity of a weaker initial lead. During the initial immune response, the IgM protein presents receptors in a decavalent format, allowing weak initial binders to attain higher apparent affinity through proximity- and chelate-based avidity.26 The large, phage surface with repeatitive structural motifs appears well-suited to this approach, and, indeed avidity effects are often present during phage-based selections and screens.27 This concept could provide a generalized method for expedient improvement of ligand affinity and biosensor sensitivity.</p><!><p>To exploit this avidity effect, phage wrapping was used to boost ligand density, subsequent affinity, and the resultant sensitivity and the signal-to-noise of phage-based biosensors. Each "wrapper" consists of two parts linked together by the CuI-catalyzed azide-alkyne cycloaddition ("click") reaction (Scheme 1).28 The first component, an oligolysine (K14) peptide provides affinity to the phage surface. For the click reaction, an alkyne (4-pentynoic acid) was coupled to the N-terminus of the K14 peptide. The second component of the wrapper is the peptide ligand to PSMA. In previous studies, the PSMA-binding peptides, 1 and 2, exhibited limited solubility in water. The peptides were therefore synthesized as fusions to the solubilizing peptide sequence K3 on their N-termini. For the click reaction, the N-termini of the peptide ligands were coupled to an azide (4-azidobutanoic acid).</p><p>The two parts of each wrapper were chemically synthesized using solid phase peptide synthesis and purified by reverse-phase HPLC before the cycloaddition reaction. Here, click chemistry offers a convergent synthesis, and the reaction takes place at room temperature and in aqueous solution.28 The resultant secondary recognition ligands thus provide an oligolysine half to wrap the phage (termed KCS for "lysine, chemically synthesized"), and a second component, the PSMA ligand (1 or 2), to bind to the analyte (Figure 2). The products formed by the click reaction (KCS-1 or KCS-2) were characterized by MALDI-TOF MS (Figure S1 in the Supporting Information), and purified by reverse-phase HPLC to an estimated 90% purity.</p><!><p>Wrapping the phage with chemically synthesized PSMA ligands described above clearly enhances binding affinity to PSMA (Figure 3). The phage-displayed ligand (phage-2) was wrapped with the secondary recognition ligands (KCS-1, KCS-2 or a mixture of the two) to generate a phage surface displaying two PSMA ligands. The wrapped phage were then assayed for binding to the PSMA dimer. Negative controls, which resulted in no detectable binding, included the PSMA ligands targeting BSA and Stop-4 targeting PSMA. Phage-2 wrapped with KCS-2 exhibits ≈50 times higher affinity for PSMA than phage-2 without the ligand wrapper (Figure 3A). Additional optimization examined the concentration of the wrapper (Figure S2 in Supporting Information). As a result, the wrapper concentration can maximize the ligand density on the phage surface. The effectiveness of wrapping for improved binding affinity is dramatically demonstrated by comparing the binding affinities of the helper phage (KO7) versus KO7 wrapped with KCS-2 (Figure 3B and Figure S3 in the Supporting Information). Lacking a displayed ligand, KO7 phage displays no significant binding to PSMA, but KO7 phage wrapped with KCS-2 binds with significant affinity to PSMA. Taken together, the results demonstrate that the wrapping strategy achieves much higher affinity for the target and the wrapped ligands remain functional.</p><!><p>The arrangement and density of the primary, genetically encoded ligand and the secondary, chemically synthesized ligand determines the affinity of the wrapped phage for PSMA. For example, wrapping phage-2 with KCS-1 results in phage with an apparent 4-fold higher affinity for PSMA than phage-2 wrapped with KCS-2 (Figure 3C). Peptide 1, however, has a much lower apparent affinity for PSMA than peptide 2, as shown in Figure 1. Thus, the increased binding affinity of phage-2 wrapped with KCS-1, suggests that the two ligands target different sites on the surface of PSMA. Furthermore, a 1:1 mixture of KCS-1 and KCS-2 wrapped on the surface of phage-2, offers intermediate affinity between neat KCS-1 and neat KCS-2. The results demonstrate that phage-2 wrapped with KCS-1 results in improved affinity due to a bidentate binding interaction. Conversely, phage-2 wrapped with KCS-2 fails to access this bidentate binding mode. Thus, the two ligands displayed on phage, for phage-2 wrapped with KCS-1, can result in a chelate-based avidity effect, which enhances binding affinity beyond the gains achieved purely by maximization of ligand density.</p><p>In practice, chelate-based avidity effects can be challenging to design, as geometry and sterics must be satisfied to allow both ligands to reach an optimal interaction with the receptor. In fragment-based drug discovery efforts, for example, development of linkers with appropriate configuration is a non-trivial problem.29 Phage wrapping provides a more expedient solution to this problem. The second ligand, presented by a non-covalently bound wrapper, can equilibrate on the phage surface until finding a satisfactory geometry to allow simultaneous binding for synergistic effect.</p><!><p>Can these wrapped virus particles be exploited to create virus-electrode biosensors with a higher sensitivity for PSMA? To explore this question, films of PEDOT were prepared on gold electrodes by electropolymerization in the presence of phage-2 (Scheme 2). The PEDOT formed in solution during electropolymerization, associate with the negatively charged perchlorate ions from the electrolyte solution as it is deposited onto the gold electrode.30 Polymerization of EDOT in the presence of negatively charged phage particles leads to incorporation of virus particles into the polymeric film as counter-ion dopants due to electrostatic interactions.21 The cyclic voltammogram acquired during electrodeposition of the virus-PEDOT bioaffinity matrix indicates that the maximum current increases with every deposition cycle, consistent with the expected increase in the surface area of the film during growth (Figure 4A). SEM imaging confirms incorporation of phage into the bioaffinity matrix; we observe both filament-like and less-extended features having dimensions consistent with phage integrated as rope-like bundles into the polymer at various angles to the film (Figure 4B, 4C). The KCS-1 wrapper was then applied in vitro simply by exposing the resultant phage-2 film for a short time to an aqueous solution of the wrapper. For the biosensing measurements, only the highest affinity ligand combination of phage-2 wrapped with KCS-1 was used, and studied in comparison to phage-2 films.</p><!><p>As we have observed in our prior work,21 the electrochemical impedance of the virus-PEDOT film increases as PSMA selectively binds to the phage-displayed peptide ligands (Figure 4D). The real component of the impedance, R, in particular, increases upon PSMA binding. In previous work, we demonstrated that the increase in R, ΔR, normalized by the initial resistance, Ro, (ΔR/Ro) can be correlated with the concentration of a target molecule.21 Here, impedance data were acquired in PBF (phosphate buffered fluoride)-Tween buffer, spanning a frequency range from 0.1 Hz to 1 MHz in an electrochemical cell with a Pt counter electrode, and virus-PEDOT film electroplated on a planar gold working electrode (Figure 4E). The films incorporating phage-2 wrapped with KCS-1 provide higher sensitivity for PSMA detection than films incorporating phage-2 (Figure 4F). For example, the relative impedance change, ΔR/Ro, at each concentration of PSMA is three-fold higher. The noise present in this measurement (estimated as the standard deviation for five impedance measurements) is unchanged resulting in a much higher signal-to-noise ratio for phage-2 wrapped with KCS-1 relative to unwrapped phage-2. A series of negative controls validate the data obtained. The PEDOT films incorporating Stop-4 phage, PEDOT films lacking PSMA binding ligands, and PEDOT films incubated with KCS-1 result in no significant binding to PSMA, as expected. The specificity of PSMA binding was investigated by using an alternative target, transferrin receptor (TfR), which has 54% sequence similarity to PSMA. No significant binding affinity to TfR was observed, as expected. This experiment illustrates the negligible change in impedance caused by the wrapper due to the small size of the wrapper.</p><!><p>The biosensing data acquired for phage-2, and phage-2 wrapped with KCS-1 targeting PSMA, follows a Langmurian adsorption model. The data acquired was fit to the following Hill equation: Y=Ymax*[L]n[Kd]n*[L]n where Y is ΔR/Ro, L is the ligand concentration, and n is the Hill coefficient.31 Consequently, the dissociation constant, Kd and n were determined for phage-2 (Kd = 54 nM, n = 1.3, LOD = 6 nM), and phage-2 wrapped with KCS-1 (Kd = 33 nM, n = 1.5, LOD = 3.1 nM). Here, the two LODs, defined as 3x over background signal, were calculated from line fits to the data shown in Figure 4F. The response obtained for phage-2 wrapped with KCS-1 displays a much higher signal-to-noise ratio, compared to films incorporating only phage-2. Such sensitivity can play a crucial role in low concentration analyte detection. The n-values obtained are >1, indicating the presence of multiple binding sites and a cooperative binding effect. Phage-2 can access cooperative binding due to the avidity effect of multi-copy phage-displayed ligands. Wrapping phage with additional ligands leads to a further increase in cooperativity for phage-2 wrapped with KCS-1. Again, the synergism of the two ligands leads to higher PSMA binding affinity for phage-2 wrapped with KCS-1.</p><!><p>To further demonstrate the usefulness of the approach for potential clinical applications, biosensing data was next acquired in synthetic urine. This complex solution includes water, nitric acid, urea, sodium sulfate, potassium chloride, sodium dihydrogen phosphate, sodium chloride, ammonium chloride and 10 other components; the resultant solution has a high salt concentration (a calculated osmolality of 516.2 mOsm/kg and a pH of 5.8).32,33 The solution provides a good model for the clinical challenge of identifying cancer biomarkers found in urine samples. Substituting synthetic urine for PBF, impedance measurements with virus-PEDOT films were acquired as before (Figure 5A). Having already established phage-2 wrapped with KCS-1 as the most effective ligand combination for detecting PSMA in terms of sensitivity, specificity and signal-to-noise ratio, our experiments focused on this ligand combination for synthetic urine-based biosensing. In the presence of PSMA, the impedance scans for virus-PEDOT films of phage-2 wrapped with KCS-1 follow similar trends in both PBF and synthetic urine. However, the lower frequency ranges differ dramatically for the negative controls. For example, the negative control with Stop-4 phage in PBF displayed a much higher ΔR/Ro; this response at low frequencies was suppressed in synthetic urine. Thus, the measurements in synthetic urine resulted in higher specificity for the PSMA-ligand interaction.</p><p>Turning next to the measurement of PSMA concentration, the synthetic urine solution also appeared to enhance measurement sensitivity. The calibration curves in PBF and synthetic urine were overlaid for comparison (Figure 5B, green and purple, respectively). At high analyte concentrations, measurements in PBF and synthetic urine are superimposable. Both conditions reach saturation at high PSMA concentrations, hence the overlap in device response. At the lowest PSMA concentrations, the higher ΔR/Ro response in synthetic urine can be attributed to higher sensitivity and selectivity for PSMA under the high-salt conditions. Phage-2 films wrapped with KCS-1 targeting PSMA in synthetic urine, yields a 100 pM experimentally observed LOD. Applying the calculation to determine LOD described above yields a 10 pM LOD for the detection of PSMA in synthetic urine. Furthermore, this sensitivity requires no signal or enzymatic amplification. As before no significant change in impedance was observed for the negative controls.</p><p>The high salt concentration in synthetic urine appears to prevent non-specific charge-charge interactions. Binding of the biomarker to the virus-PEDOT film generates a positive ΔR/Ro, whereas a negligible ΔR/Ro for the negative controls indicates a non-significant extent of non-specific binding by the analyte. The resultant specificity increase in synthetic urine boosts the apparent sensitivity of the device by decreasing background binding. This effect also enhances the sensitivity by unmasking a higher concentration of ligands for analyte detection, which would otherwise be occluded through non-specific binding. Thus, a dramatic improvement in specificity and sensitivity can be obtained through the decreased non-specific interactions. The results suggest a general strategy for improving biosensor performance through focusing on decreased non-specific binding.</p><!><p>The 100 pM experimentally measured LOD obtained for phage-2 wrapped with KCS-1 is a significant improvement over our previous effort, which had a >66 nM LOD22 for PSMA. This detection sensitivity satisfies a key requirement for clinical applications. The chelate-based avidity effect and high specificity obtained due to the bidentate binding mode distinguishes this work from our previous efforts, and also accounts for the greatly improved signal-to-noise ratio and LOD obtained without enzymatic amplification. Furthermore, inclusion of synthetic urine as an analyte solution expands the applicability of the technique. However, the reported biosensing films have not been optimized. Further improvements could be made to achieve an increased response, without amplification, for lower PSMA concentrations, as demonstrated by earlier experiments with similar biosensing films.34</p><p>In conclusion, wrapping phage with additional ligands can increase the target affinity due to a chelate-based synergistic effect. Phage-2 wrapped with KCS-1 binds to PSMA with a 33 nM Kd and a 3.1 nM LOD in PBF, and a 100 pM experimentally measured LOD in synthetic urine. In the future, the biosensors will be optimized for improved PSMA detection.</p><!><p> ASSOCIATED CONTENT </p><p>Additional materials and methods are available in the supporting information. This information is available free of charge via the internet at http://pubs.acs.org.</p>

PubMed Author Manuscript

Electroreduction of carbon dioxide into selective hydrocarbon at low overpotential using isomorphic atomic substitution in copper oxide

The conversion of carbon dioxide into selective hydrocarbons is vital for green energy generation. Due to the chemical instability and lower activity, environmentally stable transition metal oxides (e.g. CuO) are unpopular for CO 2 electroreduction catalysis. Here, we demonstrate substitution of Cu with isomorphic atom i.e. Ni in the CuO and utilize it for improving the hydrocarbon selectivity by 4 times as compared to pristine CuO. Hydrocarbon formation is achieved at the lowest possible applied potential (-0.2 V RHE). This gives the overpotential of about 0.37 V for methane and 0.28 V for ethylene, the lowest ever reported. Employing the ionic interaction between Ni and Cu, this catalyst suppresses the hydrogen evolution reaction (HER) to improve the hydrocarbon selectivity prominently. It is observed that current normalized by the BET surface area gives 15 to 20 times enhancement in the case of Ni substituted CuO compared to undoped CuO. The in situ experiments indicate that Ni doped CuO prefers CO pathways compared to formate resulting into high hydrocarbon selectivity. The experimental observation is further supported by DFT studies, which reveals that the CO molecule is stabilized on Cu 0.9375 Ni 0.0625 O surface rather than the CHO intermediate, in comparison to the pristine CuO surface.

electroreduction_of_carbon_dioxide_into_selective_hydrocarbon_at_low_overpotential_using_isomorphic_

INTRODUCTION<!>Synthesis<!>Electrochemical studies and product analysis<!>In situ FTIR under gaseous environment<!>Computational Methodology<!>Characterizations<!>Electrocatalytic CO 2 reduction

<p>Conversion of CO 2 in to value added products is beneficial for both environment and energy. [1][2] Alongside thermocatalytic 3 and photochemical [4][5] , electrochemical conversion of CO 2 has attracted much interest due to its efficient production of a variety of gaseous and liquid products under ambient temperature and pressure. 2,[6][7] Among these numerous products, C 1 -and C 2 -hydrocarbons (CH 4 and C 2 H 4 ) are exceptionally attractive but remains elusive due to their high energy densities. In particular, ethylene is highly valuable as it is a key material for manufacturing of polyethylene. 8 There are many challenges associated with the electrochemical reduction of CO 2 such as large overpotential and unavoidable competition with hydrogen evolution reactions (HER) in aqueous environment. 9 Considerable effort has been devoted to develop new catalyst, which can promote CO 2 electroreduction activity with high efficiency and selectivity while suppressing the competing HER. Several heterogeneous catalysts have been developed to improve the electrocatalytic activity by modifying the bulk surface to nanostructures focusing on Cu, [10][11][12][13] Ni, 14 Au, 15 Ag, 16 Sn, 17 In, 18 Pd 19 . Out of these, Cu is unique as being the only material to convert CO 2 into hydrocarbons with high faradaic efficiency [10][11][12][13] but it has high overpotential, wide product distribution and undesired hydrogen evolution reaction. To overcome these challenges,alloying Cu with other metals [20][21][22][23][24] and several nano-structured metal oxides [25][26][27][28] and derived-metal oxides [29][30][31] and supported-metal oxides [32][33][34] have been employed as precursors and/or electrocatalysts for enhanced electrochemical reduction of CO 2 in to desired products.The reasons behind the catalytic activities of these oxides are optimum binding energy of reaction intermediates [25][26][27][28] , multiple valences metal, [29][30][31] and synergetic effect. [32][33][34] It is well-known that replacing a small fraction of the isomorphic cations in a host metal oxide with a different cations (doping) can change the electrochemical performance of the metal oxide catalyst due to its synergetic effect. 35,36 Adopting the similar substitutional chemistry, we developed a new electrocatalyst by substitution of isomorphic Nickel (Ni 2+ cation; as dopant) on copper (II) oxide (CuO as host) using solution combustion method. Nickel and Copper in their metallic state have highly similar over potentials for CO 2 electroreduction therefore, it is expected that synergy between the two in the ionic state can tune the electrocatalytic activity. By exploiting the efficient ionic interaction between Cu and Ni, we effectively enhance both CO 2 electroreduction and the hydrocarbon selectivity. This is a new strategy and has not been explored previously.</p><!><p>Solution combustion method has been widely known for preparing the doped metal oxides and simple metal oxides because of its rapidness and large yield of the final product with high crystallinity. In a typical preparation of Ni-doped CuO (Cu 1-x Ni x O; x= 0.05 and 0.1), 2.57 g (x=0.05) or 2.72 g (x=0.1) of oxalyldihydrazide (ODH) was dissolved in hot deionized water. This is followed by the addition of 5.0 g of Cu(NO 3 ) 2 •3H 2 O (99.9% purity) and 0.32 g (x=0.05) or 0.67 g (x=0.1) of Ni(NO 3 ) 2 •6H 2 O (99.9% purity) to make a clear solution. For CuO, 2.44 g of ODH and 5.0 g of Cu(NO 3 ) 2 •3H 2 O was used. For NiO, 2.03 g ODH and 5.0 g of Ni(NO 3 ) 2 •6H 2 O was used. Further, the solution was stirred using glass rod and evaporated by heating at 70°C until we get a homogeneous solution. The as-prepared clear solution was placed in a muffle furnace preheated at 400°C for 20 min. Then, the dish was cooled and powder was collected in crucible, which was kept for calcination at 500 °C for 10 h (Figure 1a). For NiO-deposited-CuO Powder XRD was carried out using Bruker D8 Discover diffractometer (Germany) in the range of 20−80° to the crystal phases. The Rietveld refinement was done using the FullProf-fp2k program varying 17 parameters simultaneously such as overall scale factor and background parameter. BET was carried out in Micrometrics 3Flex. Samples were degassed under vacuum at 150 o C before the analysis. High resolution scanning electron microscopy was performed by using JEOL (JSM7600F) with energy dispersive spectroscopy (Oxford, Model INCA Energy 250 EDS). The high resolution transmission electron micrographs were acquired using an FEI Titan Themis (s) transmission electron microscope. The XPS was performed with Axis Ultra DLD (Kratos Analytical Limited). The spectra were obtained by using the mono-chromatized AlKα source (1486.6 eV) run at 15 kV and 10 mA. For the individual peak regions, pass energy of 20 eV was used. Survey spectrum was measured at 160 eV pass energy. Analysis of the peaks was performed with the Casa XPS software, using a weighted sum of Lorentzian and Gaussian components curves after Shirley background subtraction. The binding energies were referenced to the internal C (1s) (284.9 eV) standard.</p><!><p>Electrochemical studies were performed using the conventional three-electrode system at CHI660E electrochemical workstation. All solutions were prepared using double-distilled water.</p><p>Working electrode was made by mixing 100 mg of prepared catalyst and 200 µL of 5% Nafion solution as binder. After that 300 µL of isopropanol was added to it to make thin slurry followed by deposition on a glassy carbon working electrode (GCE). Geometric area of the electrode was 0.071 cm 2 . Platinum wire was used as the counter electrode, and Ag/AgCl electrode was used as the reference electrode and 0.5M NaHCO 3 was used as supporting electrolyte. The estimated uncompensated resistance in the cell is 134.1 ohm. An air-tight electrochemical cell was used for the gaseous product formation and analysis. For quantification of gaseous products, a gas-tight syringe (Hamilton, 1000µL) was used to transfer the evolved gases into a gas chromatograph (GC; CIC Baroda). Haysep-A column was used for separation and a Flame Ionization Detector (FID) for detection of CO and hydrocarbons. Thermal Conductivity Detector (TCD) for H 2 detection. A standard gas mixture was used for calibrating the GC. For liquid products, Bruker-500MHz Nuclear Magnetic Resonance (NMR) spectrometer is used.</p><p>The faradaic efficiency (FE %) is calculated based on the equation</p><p>where, n is the number of electron transferred in the faradaic process; F is Faraday constant, 96485 C/mol; N is the amount of the generated product in this process; Q is the total charge passed through the whole reaction.</p><!><p>DRIFTS experiments were performed with CuO and Cu 0.9 Ni 0.1 O (mixture of KBr) under different gaseous composition and at different temperatures using a Thermo Scientific NICOLET-iS50 FTIR spectroscope. The DRIFTS cell (Pike Technologies, USA) with an environmental chamber was used to record the IR spectra. The cell was first purged with ultrahigh purity nitrogen (UHP N 2 ). All the experiments were carried out a fixed flow rate of 20 ml min -1 , which is controlled by mass flow controller. The catalysts were first heated from room temperature to 200 o C and then cooled to room temperature in UHP N 2 . FTIR spectra were then recorded at different temperatures in the presence of different gases. A spectrum of powdered KBr recorded in UHP N 2 at room temperature and 100 o C which served as the background at room temperature and higher temperatures.</p><!><p>Density functional theory (DFT) was used for theoretical calculations, implemented in the Vienna ab initio simulation (VASP) package. 37 Dudarev's approach was utilized for treating the localized nature of electrons in the d orbitals of transition metal oxides. 38 The Electron-ion interactions were described using all-electron projector augmented wave pseudopotentials, and Perdew-Burke-Ernzerhof (PBE) generalized gradient approximation (GGA) 39 was used to approximate electronic exchange and correlations. The Brillouin zone was sampled using a 10x12x8 and 5x4x1 Monkhorst-Pack k-grid for the bulk and slab calculations, respectively. All structures were relaxed using a conjugate gradient scheme until the energies and each component of the forces converged to 10 -5 eV and 0.01 eV Å -1 , respectively. The (111) surface of CuO was generated from their optimized bulk structures using Virtual Nano Lab version 2016.2. The structure of Cu 0.9375 Ni 0.0625 O random alloy was generated from CuO (111) surface using the Alloy Theoretic Automated Toolkit (ATAT) code. 40 The structure had 4 layers of atoms with the bottom three layers frozen, and allowing only the topmost layer to relax, in order to mimic the bulk behaviour. For both Cu and Ni atoms, the values of U (Hubbard parameter) and J (on-site Coulomb interaction) were taken to be 7.5 and 1, respectively. S1. Considerable decrease in the lattice parameter of CuO after the Ni doping is observed. This agrees with the smaller size of Ni 2+ ion than Cu 2+ ion. As a result of decrease in the lattice parameter, there is a decrease in the total lattice volume confirming the bulk doping of Ni 2+ in CuO host. Lattice shrinking will also have contribution from lattice strain and defects also which will prevent to validate Vegard's law. Surface areas of all the samples are measured by BET method and the surface are values are 0.58 m²/g, 0.74 m²/g and 1. 2(i,j). In both cases, the doublet of Cu(2p) showed two peaks with the binding energies at 933.3 and 953.4 eV are due to Cu (2p 3/2 ) and Cu (2p 1/2 ) of CuO, respectively. 41,42 Relatively higher binding energy peaks at 941.3 and 943.6 eV are shake up satellite peaks for Cu (2p 3/2 ) and peak at 961.8 eV is the shakeup satellite peak for Cu (2p 1/2 ). Likewise, the binding energies of Ni(2p), after deconvolution, is found as the doublet of Ni (2p 3/2 ) at 852.9, 854.5 eV and of Ni (2p 1/2 ) at 870.8, 872.7 eV. Strong shake-up satellite peaks at 860.4 and 878.6 eV, respectively also appear as expected. 41,42 The broad peaks at 854.5 and 872.7 eV along with shoulder peaks at 852.9 and 870.8 are the unique features of NiO (related to Ni 2+ state) and characteristic of bulk nickel oxide. 41,42 All these peaks collectively confirms the presence of free NiO on the surface. and NiO/CuO) drop-casted on glassy carbon electrode is studied in NaHCO 3 (0.5 M). Scan rate is kept at 40 mVs -1 for all the studies. This study is carried out to understand the basic electrochemical properties of the materials. In order to avoid the oxygen evolution and hydrogen evolution reactions, we restricted our working potential window from -0.8 V to 0.4 V vs. Ag/AgCl. In Figure 3(a), CV of CuO showed a diffused redox peak with enhanced anodic (i pa ) and cathodic (i pc ) current compared to glassy carbon electrode. Oxidation and reduction peak of Cu 2+ /Cu 0 appears with the equilibrium potential (E 1/2 ) of 144 ± 2 mV with a peak-to-peak separation (∆E p ) of 197 mV, measured from the last CV cycle. A characteristic pre-wave is also observed at -0.07 V due to the oxidation of Cu 0 to Cu 1+ (Cu 2 O). 32 The surface excess values (Г), i.e., number of moles of active sites available on the electrode surface, is 0.45 nmol cm -2 (Figure S2(a)), using the equation: Г= Q/nFA. 43 Here, Q is the charge obtained for the redox peak, n is the number of electrons involved, F is the faradaic constant and A is the area of the electrode.</p><!><p>Likewise, the Cu 0.95 Ni 0.05 O showed similar redox features with E 1/2 = 137 ± 2 mV and ∆E p =184 mV values (Figure 3b) and the calculated Г value is 0.66 nmol cm -2 (Figure S2(b)). In comparison, CV of Cu 0.9 Ni 0.1 O showed prominent redox peak at 0.051 and -0.192 volts. The enhancement in the anodic and cathodic current is due to Cu 2+ /Cu 0 redox reaction coupled with substituted Ni 2+ /Ni 0 (Figure 3c). Corresponding E 1/2 and ∆E p values are 124 ± 2 mV and 147 mV with surface excess value (Г) of 1.17 nmol cm -2 (Figure S2(c)). It is noteworthy that the current values of i pa and i pc and surface excess Г value of Cu 0.9 Ni 0.1 O is higher than Cu 0.95 Ni 0.05 O, with slight shifts (~13 ± 2 mV) in the peak potentials. This indicates more number of available active sites in the former due to higher Ni substitution. Comparatively, undoped CuO shows broad and less intense peaks with peak voltages shifted by ~20 mV. Since XRD has shown the presence of separated NiO phase (Figure 1b), CV of NiO alone and NiO deposited CuO (NiO/CuO) are studied in the similar potential range as control experiments. No redox feature is observed on NiO (Figure 3d), confirming that NiO phase is not a redox catalyst in the studied potential range.</p><p>In case of NiO/CuO (Figure S2(d)), CV looks similar to CuO (Figure 3c) demonstrating no effect of NiO addition. CV for Ni metal supported CuO (Ni/CuO) is also performed as a control experiment (Figure S2(e)). This is carried out to rule out the presence of any undetected metallic nickel which may be present in the substituted compounds. In the CV, the redox features and the overall response is quite close to CuO with diminished peak intensity. These experiments clearly indicate the redox electrochemistry is only due to the interaction between substituted Ni 2+ species and Cu 2+ and not from the separated Ni/NiO or NiO/CuO phase.</p><p>Further, the redox behavior and electron transfer properties of Cu 0.9 Ni 0.1 O and Cu 0.95 Ni 0.05 O, and CuO is studied by varying the scan rates from 10 to 150 mVs -1 (Figure S3) and potential ranges between -0.8V≤ 0 ≤ 0.8V (Figure S4). While increasing scan rates (v), the respective i pa and i pc currents at corresponding redox peaks, are increased significantly with slight shift in E p and changes in ∆E p for all three electrocatalysts (Figure S3(a-c)). Presumably, the potential shifts may be due to the sluggish electron transfer kinetics due to semiconducting nature of the materials. A linear plot of peak current (i p ) versus square root of the scan rate (v 1/2 ) is drawn for Once the potential window increased to -0.6 to +0.6 V, the current values of i pa and i pc increased for both Cu 0.9 Ni 0.1 O and Cu 0.95 Ni 0.05 O. Wider potential window allows the redox reaction to occur more profoundly causing the current to increase. This increase in current is more significant in for Cu 0.9 Ni 0.1 O. At more positive side (0 to +0.8 V), bubbling can be seen in both electrodes due to oxygen evolution reaction (OER). In negative direction from 0 to -0.8 V, featureless CV response with considerable reduction current at -0.8 V is seen. Once the range is extended to -0.9V and -1V, a bubbling can be observed due to hydrogen evolution reaction (HER) in both the electrodes. Similar explanation holds true for CuO also however; the current densities are significantly lower that the Ni doped systems (Figure S4(c)). Following from the results, the potential range is fixed from -0.8 to +0.4V vs. Ag/AgCl, in order to avoid OER/HER electrochemical reactions.</p><!><p>To examine the catalytic activity of all the prepared electrodes for CO 2 electroreduction, CV experiments were carried out in the N 2 (blank) and CO 2 saturated (purging for 30 mins) 0.5 M NaHCO 3 electrolyte at 20 mV s -1 . All the CVs are carried out in the potential range of 0.0 to -0.8 V vs. Ag/AgCl. Comparative CV responses of CuO (Figure 4a), Cu 0.95 Ni 0.05 O (Figure 4b) Cu 0.9 Ni 0.1 O (Figure 4c), and are shown in CO 2 and N 2 saturated solutions. Blank GCE response is also given with all the catalysts for comparison. In presence of CO 2 , a broad cathodic response due to the electroreduction of CO 2 starts at an onset potential of about -0.54 V on CuO (Figure 4a). In Cu 0.95 Ni 0.05 O, the cathodic current in CO 2 saturated electrolytes is 0.18 mA which is about 9 times higher than the N 2 saturated electrolyte (Figure 4b). Similarly, in Cu 0.9 Ni 0.1 O, cathodic response with an onset voltage of -0.35 V is attained (Figure 4c). Onset voltage is about 20 mV lower than undoped CuO demonstrating the catalytic effect of Ni 2+ in CuO. Cathodic current at -0.8 V is 2.9 mA in CO 2 saturated solution. This is about 14 times higher than N 2 saturated solution confirming the high activity of Cu 0.9 Ni 0.1 O for CO 2 electroreduction. Current density after normalizing the surface area gives about 20 times higher activity in the case of Cu 0.9 Ni 0. S2). Cleary, doped compounds are remarkably higher active than CuO, Ni/CuO and NiO/CuO further highlighting the importance of the ionic interaction for superior activity.</p><p>Steady state current responses for all the three catalysts in CO 2 saturated electrolyte at -0.8V vs. Ag/AgCl (-0.2 V RHE), is shown in Figure 4(d). Initially the current decreased gradually up to 400 s and attained steady state with a current value of 5.6 mA and 4.8 mA for Cu 0.9 Ni 0.1 O and Cu 0.95 Ni 0.05 O, respectively. This indicates that both the catalysts are stable and Cu 0.9 Ni 0.1 O has higher electroreduction activity for CO 2 . In contrast, the CuO initially showed a characteristic decrease followed by increase in current depicting an oxidation behavior. Steady state is not attained even after 1000 seconds due to the instability of the CuO. In addition, CV is carried out after this chronoamperometric experiment for Cu 0.9 Ni 0.1 O and CuO (Figure S6). It is observed that the redox behavior of Cu 0.9 Ni 0.1 O is preserved whereas it is diminished in CuO. This highlights the Ni 2+ substitution in CuO on one-hand increases the activity and on the otherhand imparts the stability. In order to analyze the products formed after CO 2 reduction, the evolved gases were collected by an air-lock syringe and the products were subjected to gas chromatography. The major gaseous products on all there electrocatalysts are CO, CH 4 , C 2 H 4 and H 2 . Since most of the copper based electrocatalysts have been pronounced for the formation of CH 4 and C 2 H 4 with H 2 as the side product, it is interesting to note that Ni substitution in CuO suppresses the hydrogen production significantly. This is also depicted in terms of percent Faradaic efficiencies (FE %) as shown in Figure 4(e,f). FE % of CO is same (1.0 %), for all the three catalysts as the copper based catalysts do not favor CO formation. 32 Comparatively, the formation of CH 4 and C 2 H 4 predominantly occurs on Cu 0.9 Ni 0.1 O with FE% up to 29.4% and 12.8%, respectively (Figure 4e).The same was 20.2 % and 8.7 % for Cu 0.95 Ni 0.05 O. On CuO, the FE% for CH 4 and C 2 H 4 is 3 and 4 times lower than Cu 0.9 Ni 0.1 O. Inferior hydrocarbon FE% with CuO is due to the undesired HER and possibly a different mechanism. FE for hydrogen on Cu 0.95 Ni 0.05 O and Cu 0.9 Ni 0.1 O is 25.4%and 28.7%, much lower than CuO (35.9%) as in Figure 4f. Thus, the presence of Ni in CuO host enhances the formation of hydrocarbons while suppressing the hydrogen production.</p><p>Comparative electrocatalytic performance of all three catalysts with literature is available in Table S3 (supporting information). From the table it is also clear that hydrocarbon efficiency of the doped catalysts is comparable or better at much lower applied potential. Total FE % is 58 to 68% for both the Ni substituted catalysts as only the gaseous products are analyzed. Very importantly, both Cu 0.95 Ni 0.05 O and Cu 0.9 Ni 0.1 O are able to generate hydrocarbons at -0.2 V (RHE) which is so far the lowest. In terms of overpotential, this is about 370 mV for methane and 280 mV for ethylene. To the best of our knowledge, this is the lowest overpotential ever reported. From Table S3, the lowest potential is -1.1 V (RHE) for methane and ethylene production. This is higher by 0.9 V than Cu 0.95 Ni 0.05 O and Cu 0.9 Ni 0.1 O, a very significant gain in voltage.</p><p>The stability of Cu 0.9 Ni 0.1 O catalyst is studied by depositing it onto FTO electrode and subjected to post-XRD and XPS after chronoamperometry in CO 2 saturated solution for 1000 sec (Figure S7 & S8). It is observed that apart from FTO peaks, Cu 0.9 Ni 0.1 O peaks are retained before and after CO 2 electrocatalysis confirming the stability of the bulk structure of Cu 0.9 Ni 0.1 O catalyst (Figure S7). The XPS also confirms the stability of the Cu 0.9 Ni 0.1 O catalyst surface as shown in Figure S8. XPS of as prepared catalysts is also given for comparison. It is worth noticing that there is no change in the XPS spectra before and after the chronoamperometry. Both copper and nickel show same oxidations state before and after the experiments proving the stability of Cu 0.9 Ni 0.1 O catalysts. This affirms our claim made in the chronoamperometry experiment (Figure 4d).</p><p>It is speculated that improvement in hydrocarbon selectivity along with the HER suppression is due to modification in the mechanism which is investigated by in situ FTIR and DFT studies. Mechanistic investigation for the CO 2 electroreduction pathways over CuO and Cu 0.9 Ni 0.1 O is executed using DRIFT in situ FTIR experiment (Figure 5). The schematic representation of DRIFT cell is displayed in Figure 5a. As the copper based materials are known for converting CO 2 to hydrocarbons both thermally and electrochemically, it is assumed that thermal reduction of CO 2 with gaseous hydrogen is similar to electrochemical reduction of CO 2 .</p><p>Firstly, the interaction of H 2 over CuO and Cu 0.9 Ni 0.1 O is investigated in the range of 2400 to 3200 cm -1 (Figure 5b). IR spectra of both as prepared sample show no indication of any band in the flow of nitrogen at room temperature. On passing H 2 , the DRIFT spectra of Cu 0.9 Ni 0.1 O, showed weak bands in the range of 2800 -3000 cm -1 at 50 o C. This band is due to the formation of formate on the surface by utilizing the atmospheric CO 2 . 44 The similar band on CuO is considerably higher than Cu 0.9 Ni 0.1 O under the same conditions. Same band at 100 o C show increase in the intensity on both the catalyst but intensity is still higher on CuO. At 200 o C, intensity of formate diminishes significantly over Cu 0.9 Ni 0.1 O but not much change is observed in the case of CuO. Temperature higher than 300 o C causes complete disappearance of formate band on both the catalysts. These experimental observations suggest that formate bands are more stable and prominent on CuO surface than Cu 0.9 Ni 0.1 O. Once Ni is substituted in CuO matrix, formate band are less stable and less prominent. Similar observation can also be drawn when the mixture of H 2 and CO 2 (1:1) is reacted on the catalyst's surface (Figure 5c). In this case also, over all intensity of formate band is less prominent on Cu 0.9 Ni 0.1 O compared to CuO. Intensity differences in the formate can be prominently observed at 100, 200 and 300 o C.</p><p>In the light of above observations, one can confirm that Ni substitution in CuO affects the catalyst's ability to stabilize formate (Figure 5d). This demonstrates the mechanistic differences between the two catalysts in which CuO shows the dominant formate pathway compared to Cu 0.9 Ni 0.1 O. Dominance of formate pathway with the generation of H 2 can be correlated well in the light of quantum mechanical calculations carried out by Cheng et. al. 45 This study reveals two pathways in copper bases materials; one is dominatingly forming formate and other forming CO.</p><p>The pathway leading to CO eventually leads to hydrocarbon formation while more formate leads to more hydrogen. This correlates well with the FTIR studies establishing that that Ni substitution in CuO shows less contribution from formate pathway resulting in less hydrogen formation and due to this more hydrocarbons. CuO on the other hand shows more contribution from formate pathway resulting into fewer hydrocarbons and more hydrogen generation. Further, Density functional theory (DFT) was used for theoretical calculations (Figure 6), implemented in the CO 2 electroreduction pathways on CuO and Cu 0.9 Ni 0.1 O model catalysts (Figure S9-11). The optimized lattice parameters of bulk monoclinic CuO (space group -C2/c) were found to be: a = 4.37, b = 3.86, and c = 3.92 Å (Figure S9,10), in excellent agreement with previous reports. [46][47] The (111) facet of CuO was chosen for calculations, as it has been found to be the most stable facet of CuO with lowest surface energy. 46 It contained 128 (64 Cu and 64 O) atoms. The Ni-doped structure contained 60 Cu and 4 Ni atoms, so the stoichiometry of alloy turned out to be Cu 0.9375 Ni 0.0625 O (Figure S11). Only one composition of Ni-doping was taken, as it is enough to observe the qualitative effect of change in the catalytic activity upon alloying.</p><p>The carbon dioxide electro-reduction involves 8 intermediates (8-electron transfer process), but the two most important intermediates which determine the overall activity on any catalyst surface are CO* and CHO*. 48 The symmetrically inequivalent adsorption sites on CuO(111) surface were found using the pymatgen code 49 , which came out to be 20 in number and are shown in Figure S12(a). Single-point calculations were performed on these adsorption sites to find the minimum energy configurations for CO 2 *, CO* and CHO* intermediates. For CO 2 , the minimum energy adsorption site lies in the region above Cu and O atoms, while those for CO and CHO lie above Cu and O atoms, respectively. However, for the case of Cu 0.9375 Ni 0.0625 O surface, the number of such sites increased to 134. Therefore, to save computational time, we performed single-point calculations only for the (12 symmetrically inequivalent) sites near Ni atom on the surface, shown in Figure S12(b). The minimum energy adsorption sites for CO 2 * and CHO* were similar as that of pristine CuO (111) surface, but CO was found to be most stable on Ni rather than Cu atom.</p><p>The intermediates were then completely relaxed on the most stable configuration sites (Figure 6(a)). From their energies, the intermediate adsorption energies were calculated using the formula:</p><p>Here * represents the catalyst surface.</p>

ChemRxiv

Pathway for Unfolding of Ubiquitin in Sodium Dodecyl Sulfate, Studied by Capillary Electrophoresis

This paper characterizes the complexes formed by a small protein, ubiquitin (UBI), and a negatively charged surfactant, sodium dodecyl sulfate (SDS), using capillary electrophoresis (CE), circular dichroism (CD), and amide hydrogendeuterium exchange (HDX; as monitored by mass spectroscopy, MS). Capillary electrophoresis of complexes of UBI and SDS, at apparent equilibrium, at concentrations of SDS ranging from sub-micellar and sub-denaturing to micellar and denaturing, revealed multiple complexes of UBI and SDS of the general composition UBI-SDSn. Examination of electrophoretic mobilities of complexes of UBI and SDS as a function of the concentration of SDS provided a new way to characterize the interaction of this protein with SDS and established key characteristics of this system: e.g., the reversibility of the formation of the complexes, their approximate chemical compositions, and the pathway of SDS binding to UBI. The work identified, in addition to SDS-saturated UBI, at least six groups of complexes of UBI with SDS, within which four groups were populated with complexes of distinct stoichiometries: UBI-SDS\xe2\x88\xbc11, UBI-SDS\xe2\x88\xbc25, UBI-SDS\xe2\x88\xbc33, and UBI-SDS\xe2\x88\xbc42. CD spectroscopy and amide HDX of the UBI-SDSn complexes suggested that many of the UBI-SDSn complexes (n > 11) have greater \xce\xb1-helical content than native UBI. Capillary electrophoresis provides a level of detail about interactions of proteins and SDS that has not previously been accessible, and CE is an analytical and biophysical method for studies of interactions of proteins and surfactants that is both convenient and practical. This study sheds light on the formation of the enigmatic protein-SDS complexes formed during SDS polyacrylamide gel electrophoresis and brings a new tool to the study of proteins and detergents.

pathway_for_unfolding_of_ubiquitin_in_sodium_dodecyl_sulfate,_studied_by_capillary_electrophoresis

Introduction<!>Proteins and Lipids/Surfactants in Biochemistry<!>Proteins and SDS<!>Capillary Electrophoresis (CE)<!>Selection of the Tris-Glycine Buffer, Its pH, and Working Temperature<!>Selection of Ubiquitin (UBI)<!>Choice of SDS<!>Quantitative Secondary Structure of UBI-SDSn Complexes by Circular Dichroism (CD)<!>Structure of UBI-SDSn Complexes by Amide Hydrogen-Deuterium Exchange (HDX) in SDS, Monitored by Electrospray Ionization Mass Spectroscopy (ESI-MS)<!>Charge Ladders: A Method for Evaluating the Stoichiometry of UBI-SDSnComplexes<!>Analysis of the Thermodynamics of Formation of UBI-SDSn by CE<!>Capillary Electrophoresis Detects Multiple Complexes of UBI-SDSn with Distinct Electrophoretic Mobilities after Equilibrium Dialysis against Concentrations of SDS Ranging from 0 mM to Greater than the Critical Micelle Concentration (cmc)<!>Increase in Mobility of UBI-SDSn in G1* Proceeds through Multiple Overlapping Peaks until the First Discrete Complex (G2) Forms at [SDS] = 0.8-1.4 mM<!>A Second Discrete Group of Intermediates (G4) Dominates at [SDS] = 2.0-3.0 mM<!>A Third Discrete Group of Complexes (G5) Dominates at [SDS] = 3.8 mM<!>Structure of UBI-SDSn Complexes on the Path from N to D by Circular Dichroism<!>Structure of UBI-SDSn Complexes on the Path from N to G4, Examined by Hydrogen-Deuterium Exchange (HDX)<!>Formation of UBI-SDSn Complexes Is Reversible According to Capillary Electrophoresis, Circular Dichroism, and Amide Hydrogen-Deuterium Exchange<!>Determination of the Stoichiometry of UBI-SDSn Complexes in G1* and G2 by Comparing Electropherograms of UBI-SDSn with UBI Charge Ladders<!>Determination of the Number of SDS Molecules Involved in the Transitions from G2 to G4 and in the Transition from G4 to G5 by Equilibrium Analysis<!>Determination of the Stoichiometry of G4 and G5 by Extrapolation of Mobilities of 4-Sulfophenylisothiocyanate Charge Ladders in the Range of Mobility Defining G4, G5, and D<!>Conclusions

<p>We have studied the association of a small protein, ubiquitin (UBI), with a negatively charged surfactant, sodium dodecyl sulfate (SDS), in aqueous buffers (e.g., tris-glycine, tris-borate, borate, phosphate). We chose SDS because it is a surfactant that is widely used in protein biochemistry [in SDS polyacrylamide gel electrophoresis (PAGE) and in studies of folding and unfolding]. Our analysis of mixtures of UBI and SDS by capillary electrophoresis (CE) revealed six groups of equilibrium complexes of composition UBI-SDSn. Analysis of these complexes by CE allowed us to (i) establish that they were formed reversibly, (ii) estimate (in some, exactly; in others, approximately) the number of equivalents of SDS bound in each complex, and (iii) propose a pathway for unfolding of UBI in presence of SDS. We also characterized elements of the structure of some of these complexes using circular dichroism (CD) and amide hydrogen-deuterium exchange (HDX). Structural analysis of these complexes of UBI and SDS - UBI-SDSn (where n increases approximately from 1 to 42 as the concentration of SDS increases from 0.05 to 10 mM) - by CD and HDX indicates that in some of these complexes the protein retained substantial native-like secondary structure (i.e., β-sheets and α-helices). Other complexes (e.g., those formed at denaturing levels of SDS as low as 3.0 mM) exhibited a surprisingly large amount of α-helical structure while retaining very little tertiary structure; this change in secondary structure suggests that the denaturation of ubiquitin by SDS involves substantial conversion of β-sheet to α-helical secondary structure.</p><!><p>Complexes of proteins with lipids (and surfactants) are ubiquitous in biochemistry and are widely used in biotechnology. Examples range from structures of phospholipids bilayers with integral membrane proteins,1 to the surfactant-protein micelles present during the estimation of the molecular weight of proteins using anionic detergents (e.g., SDS-PAGE).2 Unfortunately, the technical difficulties in studying interactions of proteins and surfactants, the apparent low specificity of the binding of surfactant to proteins, and the stoichiometric and structural heterogeneity of protein-surfactant complexes have made this area of structural biology and biochemistry particularly trying to study.3 A better understanding of protein-surfactant interactions will certainly benefit from, and probably require, the development of new analytical tools and techniques (e.g., new methods of separation and analysis) in biochemistry and biotechnology.</p><p>Understanding protein-surfactant interactions might also contribute to understanding the hydrophobic effect,4,5 because protein denaturation is believed to expose hydrophobic amino acid residues that are generally buried in the hydrophobic core of the native protein. The interaction between a protein and a charged surfactant such as SDS is complex and involves both hydrophobic and electrostatic interactions.6 Hydrophobic (and other noncovalent) interactions are at the core of molecular recognition in biology. Surfactants, both individually and in aggregates, can modify the function, conformation, and activity of proteins.</p><!><p>Understanding the interactions of SDS with proteins is also necessary to resolve the long-standing uncertainty surrounding the mechanism of SDS-PAGE and, perhaps, to suggest new analytical techniques based on protein-surfactant interactions. This legacy technique, which is based upon the binding of SDS to thermally and reductively denatured proteins, is arguably still one of the most widely used analytical tools in protein biochemistry.7 Its ubiquity not withstanding, the binding of SDS to proteins and the denaturation of proteins by SDS are not well understood.</p><p>During denaturing SDS-PAGE, proteins migrate as complexes with SDS (sometimes called, on the basis of little information, "SDS micelle-protein complexes"),8 roughly according to the molecular weight of the native protein; astonishingly, their mobility is largely insensitive both to amino acid sequence and to secondary, tertiary, or quaternary structure in the native conformation. The stoichiometry of SDS bound to proteins under denaturing conditions is relatively uniform: on average, 1.4 g of SDS associates with 1 gram of protein at saturation at concentrations of SDS greater than the critical micellar concentration ([SDS] > cmc).9,10 This consistent stoichiometry is the first rule of thumb for rationalizing behavior in SDS-PAGE; that is, on average, one molecule of SDS associates with two amino acid residues.11 The basis for this stoichiometry, and its apparent independence of structure of the proteins, remains a mystery.</p><p>Spectroscopic techniques (e.g., UV/visible, fluorescence, and CD spectroscopy) have been used to identify and study protein-SDS complexes that form under sub-denaturing SDS concentrations.12-17 Isothermal titration calorimetry (ITC) has also provided valuable information about the protein-SDS complexes that form under sub-denaturing concentrations of SDS.18 These techniques have approximated the stoichiometry of protein-SDS complexes and the thermodynamics of SDS binding;19-23 these methods do not directly observe distinguishable protein-SDS complexes.</p><p>We can obtain information about the stoichiometry of SDS complexes with proteins more directly by measuring the change in the electrophoretic mobility of a protein (using capillary electrophoresis) that occurs from the binding of a negatively charged SDS molecule to it. Capillary electrophoresis provides reproducible, quantitative information of a precision and resolution that cannot be achieved in gel electrophoresis.</p><!><p>The methods we applied here are, in principle, close to those used in affinity capillary electrophoresis24 and in surfactant capillary electrophoresis (SurfCE), a related technique that screens the conditions for the association of proteins with surfactants under non equilibrium conditions.25</p><p>CE separates molecular species according to their electrophoretic mobility (μ).26 To an (often quoted) approximation (eq 1), the electrophoretic mobility of a species is directly related to its net charge and inversely related to its hydrodynamic drag. In eq 1, Z equals the net charge of the protein or molecule, Cp and α are constants, and M is the molecular weight of the migrating species (here, protein or protein-SDS complexes).(1)μ=CpZMα</p><p>We used CE to separate complexes (UBI-SDSn) having differences in their composition (e.g., different values of n). The migration and separation of different UBI-SDSn complexes depend on changes in electrophoretic mobility that result from the binding of SDS molecules (SDS is a negatively charged molecule, ZSDS ≈ -1) in a UBI-SDSn complex.</p><p>The sequential binding of SDS to a folded protein (which remains folded upon binding) is fundamentally similar to the removal of positive charges on a protein (e.g., charge ladders),27,28 at least in terms of charge and mass variations. We therefore use charge ladders as a tool with which to calibrate the number of molecules of SDS associated with UBI in UBI-SDSn.</p><!><p>We used a buffer that is typically used in SDS-PAGE (tris-glycine buffer; 25 mM tris, 192 mM glycine, pH 8.4). All experiments were carried at room temperature (22 < T < 24 °C). The results presented here, however, are not dependent upon the buffer; similar results were obtained using four different buffers (e.g., tris-borate, pH 8.1; 2-amino-2-methyl-1,3-propanediol-glycine, pH 8.7; phosphate, pH 8.4; borate, pH 8.3; see Supporting Information, Figure S1).</p><!><p>Our primary motivation for using ubiquitin is that this protein exhibited a rich variety of complexes with SDS (as demonstrated in a survey using SurfCE).25 Our choice was also motivated by the extensive development of ubiquitin as a model system for studies of protein folding29 and NMR.30 Ubiquitin - as its name implies - is ubiquitous among all eukaryotic cells; its major role is to label other proteins for degradation by the ubiquitin/proteasome system (UPS) via an intracellular ATP-dependent polyubiquitination process.31</p><p>Ubiquitin (with indistinguishable structure, at least in bovine and human origins) is a polypeptide comprised of 76 amino acids; the native structure comprises one α-helix (three and one-half turns), one short 310 helix, a mixed β-sheet (comprising five β-strands), and seven reverse turns.32 Eighteen amino acid residues are involved in the α-helices (23%), 26 in the β-strands (∼34%), and 31 in coils and loops (41%) (PDB 1UBQ). Ubiquitin has no known metal binding sites, no cysteine residues, and a molecular weight of 8565 Da. Ubiquitin contains seven lysine (pKa ≈ 11.1), four arginine (pKa ≈ 12.5), six glutamic acid (pKa ≈ 4.5), five aspartic acid (pKa ≈ 4.5), one histidine (pKa ≈ 6.8), and one tyrosine residues (pKa ≈ 9.8) and is not acetylated at the methionine N-terminus (pKa ≈ 9.2). The number of positively charged residues (from the sequence) is thus 13 (α-NH3+ N-terminus included). The net charge (Z0) is estimated from the amino acid sequence to be -0.9 (Protein Calculator, http://www.scripps.edu). The measured net charge of UBI at pH 8.4 (determined by protein charge ladders and capillary electrophoresis) is Z0 = -0.2 (see Figure 4a, with ΔZ = -0.9).</p><!><p>SDS is the surfactant most used in studies of protein denaturation. The critical micelle concentration (cmc) of SDS in the tris-glycine buffer used here is 3.4 mM, as determined by CE (see the Supporting Information, Figures S2, S3).</p><p>The negative charge of SDS allows us to monitor its binding to UBI easily by CE: each bound SDS molecule will increase the negative charge of the resulting complex by ΔZ (which we assume to be indistinguishable from the charge increment in charge ladders). SDS is also transparent at the UV wavelength of the CE detector (214 nm); there is thus essentially no background in CE due to SDS, and the protein - in all of its conformations - can be studied without the requirement for an attached chromophore. This capability to observe the protein spectroscopically without interference from other species in the system is an enormous experimental convenience.</p><!><p>We used CD to study the secondary structure of complexes of UBI with SDS. After their formation by equilibrium dialysis, we examined UBI-SDSn complexes in the range 205-260 nm (at a total protein concentration of ∼50 μM). The CD spectrum is a sensitive measure of the secondary structure of proteins (λ < 260 nm).33 We quantified the secondary structure of UBI-SDSn by using three independent programs for deconvolution (e.g., CDSSTR, SELCON3, and CONTINLL).33 These programs reconstruct the CD spectrum of a protein (here a UBI-SDS complex) with a linear combination of individual k's (where k represents contributions from elements of secondary structure). Fractions of α-helices, β-sheets, turns, and random coils are the outputs from these three programs (fk, in %; see Figure 2b).</p><!><p>We also examined the structure of ubiquitin at various concentrations of SDS by measuring the rate of amide HDX using mass spectrometry.34 This rate is a sensitive method for detecting minor changes in the secondary and tertiary structured of proteins that arise from changes in hydrogen bonding or hydrophobic interactions (i.e., during thermal or chemical unfolding, ligand binding, or as a result of amino acid substitution).35</p><!><p>A protein charge ladder is a set of protein derivatives obtained by the conversion of charged residues (typically lysine ε-NH3+ and N-terminus α-NH3+) into electrically neutral (α- and ε-NHCOCH3) or negatively charged (α- and ε-NHCO-R-X-) residues (R is an organic spacer). Charge ladders make it possible to measure experimentally the change in mobility Δμ associated with the chemical removal of positive charge on UBI (either by acetylation or by acylation with reagents with negatively charged sulfonate groups) and to compare them to the values of Δμ obtained on the addition of a negative charge (by association with SDS). The mobility μn of the rung n of a protein charge ladder (here a UBI charge ladder) can be expressed as a function of the net charge of the unacylated protein (Z0), its molecular weight (MUBI), the number of acylated groups (n), the change in charge (ΔZ) resulting from acylation, the added mass (MX), and Cψ,n a correction factor accounting for nonlinearities in surface potentials associated with each acylation (eq 2; more details in the Supporting Information).(2)μn=CpCψ,nZ0+nΔZ(MUBI+nMX)α</p><p>The value of ΔZ observed on adding one unit of charge is less than one unit, due to so-called charge regulation (which is an adjustment of the extent of protonation of ionizable residues on the protein as a response to a change in electrostatic potential).36,37 One plausible origin of the difference between the theoretical value of ΔZ = -1.0, expected on changing the charge of a protein by one unit of charge (e.g., by converting Lys-ε-NH3+ into Lys-ε-NHCOCH3), and the inferred value of ΔZ = -0.9 (as established using charge ladders of BCAII)37 is a small change in local pH, or a shift in the pKa, of some ionizable residues (the two are equivalent analytically). The effect of charge regulation would be suppressed (e.g., ΔZ ≈ -1) if pKa values were significantly different from the working pH (at least, different by more than three pH units), according to the "charge-regulation model" proposed by Menon and Zydney36 and applied to BCAII.37 The distribution of values of pKa in a protein is such that this condition cannot, in practice, be met. Rather than working with an absolute relationship between ΔZ and Δμ, in this work we use relationships calibrated empirically using charge ladders of UBI.</p><p>The values of mobility (μn) of each rung of the charge ladder cannot be used directly (e.g., superposed) to determine the stoichiometry n for a species with composition UBI-SDSn which has an equivalent electrophoretic mobility: while we expect the association of SDS with UBI, and the neutralization of Lys-ε-NH3+ and N-terminal-NH3+ via acetylation, to increase the net negative charge by an indistinguishable amount, the molecular weight (and hydrodynamic drag) of two equally charged species (here UBI-SDSn and the nth-rung of a UBI charge ladder) are significantly different (dodecyl sulfate, DS-, has a molecular weight of MDS- = 265.4 Da; acetylation of lysine by acetic anhydride increases the mass by MX = 42.0 Da; chemical modification of lysine by 4-sulfophenylisothiocyanate increases the mass by MX = 214.2 Da).38 For expressing μUBI-SDSn (e.g., the mobility of a UBI-SDSn complex; eq 3) as a function of μn, we replaced Z0 + nΔZ in eq 3 by its expression in eq 2, assuming that the binding of a SDS molecule is undistinguishable electrostatically from acylation of Lys-ε-NH3+). These manipulations resulted in eq 4.(3)μUBI−SDSn=CpCψ,nZ0+nΔZ(MUBI+nMDS−)(4)μUBI−SDSn=μn(MUBI+nMXMUBI+nMDS−)αEquation 4 produces predicted values of μUBI-SDSn and describes complexes in which ubiquitin bears both additional negative charges and additional mass. We assume that the nonlinear dependence of the surface potential, Cψ,n, is indistinguishable for equally charged species in UBI-SDSn and UBI charge ladders. We also assume, in the absence of other information, that α = 2/3.39 Charge regulation is empirically included in this analysis, as we use experimental values of the mobilities of the rungs of UBI charge ladder where ΔZ ≈ 0.9 is assumed to be the same for both UBI charge ladders and UBI-SDSn.</p><p>For increasing numbers of associated SDS molecules (or for higher rungs of the charge ladder), however, ubiquitin becomes highly negatively charged, which might, in principle, affect the folding of UBI. We neglect this effect in light of the work by Makhatadze demonstrating that the neutralization of all arginines and lysines on the surface of UBI left the protein folded.40</p><!><p>We quantified the stoichiometry n in UBI-SDSn complexes by assuming the association of UBI and SDS to be at equilibrium, and by considering eqs 5 and 6.(5)UBI+nSDS⇄UBI−SDSnKn=[UBI−SDSn][UBI][SDS]n(6)log([UBI−SDSn][UBI])=nlog([SDS])+log(Kn)</p><p>The linear least-squares fit regression of log([UBI-SDSn]/[UBI]) as a function of log([SDS]) yields both the stoichiometry n and the binding constant Kn for the transition; [SDS] is exactly known if the protein is equilibrated with SDS by dialysis. The ratio [UBI-SDSn]/[UBI] can be determined by measuring the relative concentrations of UBI and of UBI-SDSn at a given value of [SDS] using the UV/visible detector in CE.</p><p>In order to treat the formation of UBI-SDSn complexes as reactions that have reached equilibrium, we first proved the reversibility of their formation. For this proof, we dialyzed native UBI (N) and fully denatured UBI (D) for 170 h against buffers containing SDS in the range between 0 and 10 mM (the time required for UBI denatured in 10 mM SDS - that is, D - to refold into a species having a mobility indistinguishable from that of the native protein under these dialysis conditions is 170 h). By generating a particular UBI-SDSn complex, at a specific concentration of SDS, in two different ways (i.e., starting from N or starting from D) and then characterizing these proteins (with CE, CD, and HDX), we established that the formation of UBI-SDSn was reversible (see Figure 3 for CE data, and Supporting Information for CD and HDX data, Figure S4).</p><!><p>Analysis of UBI dialyzed against buffered SDS by CE allowed us to observe distinct complexes of UBI and SDS, with electrophoretic mobilities (μ) that are intermediate between the value of mobility for native UBI in the absence of SDS (UBI in 0.0 mM SDS, denoted N; μ = 0.3 cm2 kV-1 min-1) and fully denatured41 UBI (UBI in 10 mM SDS, denoted D; μ = 22.6 cm2 kV-1 min-1) (Figure 1).42 We classified these intermediate complexes subjectively (based on the appearance and shape of peaks) into six groups of complexes which we denote G1*, G2, G3*, G4, G5, and G6* in addition to N (native UBI) and D (SDS-saturated UBI).</p><p>The observation of a number of stable (over the time required for a CE experiment) complexes of UBI with SDS establishes a multistep pathway for the unfolding of UBI in SDS. The appearance and disappearance of discrete peaks (e.g., N, G2, G4,G5, and D) as the concentration of SDS increases demonstrates the formation of UBI-SDSn species with distinct composition.</p><!><p>The association of a small number of SDS molecules with UBI takes place at concentrations of SDS as low as 0.4 mM. We interpret the conversion of the peak at μ ≈ 0cm2 kV-1 cm-1, N, to broad peaks spanning 1-14 cm2 kV-1 cm-1 to represent the formation of a range of different species of UBI-SDSn. These various species, G1*, all convert to a well-defined, discrete intermediate, G2, at [SDS] = 1.0 mM. G2 is a stable species at concentrations of SDS from 0.8 to 1.4 mM, beyond which the association of a discrete number of SDS molecules converts it into G3* and G4.</p><!><p>In this region, a second discrete peak appears at μ ≈ 20 cm2 kV-1 cm-1 (e.g., G4). The transition from G2 to G4 proceeds via G3* (Figure 1c). This transition also involves an additional complex observed as a shoulder on the G2 peak at 1.7 mM SDS (μ ≈ 16 cm2 kV-1 min-1; this shoulder is not observed at any other SDS concentration) (Figure 1c). The presence of this shoulder indicates that there is at least one additional species on the path from G2 to G4 (differing, we estimate later using charge ladders, by Δn ≈ 1). The mobilities of both major peaks in G2 and G4 become higher with increasing concentrations of SDS (Figure 1b,d); this behavior suggests the presence of species in which the number of associated SDS increases with the concentration of SDS but does not result in a transition to another distinct structure (e.g., another major unfolding or restructuring event).</p><!><p>The electropherograms in Figure 1d show a discrete transition from G4 to G5:G (μ ≈ 20.6 cm2 kV-1 min-1) converts to G5, a group of complexes with higher mobility: μ ≈ 21.5 cm2 kV-1 min-1. The transitions generating D from G5 (occurring in the group G6*) are less distinct (although still analyzable) than the transitions from G2 to G4 or G4 to G5. These peaks also undergo a distinct transition in shape: from triangular and sloping to the left (G4 and G5 are "right-handed" at 2.0 < [SDS] < 3.8 mM) to a peak sloping to the right (D is "left-handed" at [SDS] > 7.0 mM; we discuss these asymmetries in peak shape in terms of electrodispersion in the Supporting Information, Figure S7). Since the micellar phase begins to appear significantly at [SDS] > 3.4mM, those transitions - observed here above the cmc - suggest that SDS continues to associate with UBI-SDSn, while the literature generally states that, beyond the cmc, no further unfolding occurs and excess surfactant simply leads to further formation of micelles.11</p><!><p>As the concentration of SDS increases, the average ellipticity generally increases for UBI-SDS complexes (Figure 2a). As estimated from the deconvolution of CD spectra, the UBI-SDSn complexes in groups G1* and G2 ([SDS] < 1.4 mM) have secondary structure that is indistinguishable from native UBI (Figure 2b). For concentrations of SDS higher than 1.5 mM, we observe an approximately 2-fold increase in α-helical structure and a 2-fold decrease in β-strand structure (we infer a maximal content of α-helical secondary structure at [SDS] = 2.6 mM).</p><p>The overall structural picture that emerges from CD spectroscopy is that, as the concentration of SDS increases, UBI loses β-strands (and possibly loop structures) in favor of α-helices, with the maximum amount of α-helical structure observed at 2.6 mM SDS (G4). This increase is consistent with previous investigations into the structural effects of SDS on binding to proteins.8</p><p>While a small (∼1.2-fold) decrease in ellipticity is observed in the conversion of G4 to G5 (e.g., 2.6 < [SDS] < 3.5 mM), no significant variation in secondary structure was detected for the transition from G5 to D; this observation is in agreement with previous literature reporting that no further unfolding is observed above the cmc.11 This absence of variation in structure in the micellar regime could be due to a lack of sensitivity of CD to this transition, or perhaps because the secondary structure does not change once SDS micelles form.</p><!><p>Mass spectrometric results revealed that the binding of SDS to UBI decreases the rate of amide HDX as the concentration of SDS increased up to 1.4 mM (Figure 2c). The number of unexchanged or "protected" hydrogens therefore increased along the pathway from N to G2 (the increase in protected hydrogens is approximately linear with respect to [SDS], see Figure S4d). For example, under completely native conditions (e.g., [SDS] = 0 mM), the ubiquitin polypeptide exchanges ∼50 amide hydrogen very quickly (i.e., t < 10 min) and retains ∼10 protected hydrogens at t = 30 min. These protected hydrogens remained unexchanged with solvent throughout the course of the experiment (60 min). At higher concentrations of SDS (1.4 mM), the UBI polypeptide undergoes slower amide hydrogen exchange than at [SDS] = 0 mM: ∼21 hydrogens are protected for [SDS] = 1.4 mM. We speculate that the increase in protection on going from [SDS] = 0 to 1.4 mM is due to electrostatic interactions between the negative charge on SDS and OH-.43</p><p>HDX has been previously used in combination with CD spectroscopy to demonstrate that proteins (in the form of molten globules or partially folded states)44 can possess substantial secondary structure (as measured by CD spectroscopy) while also lacking tertiary structure (as indicated by a lack of protection from H/D exchange). Molten globule states are generally observed at partially denaturing conditions (e.g., mild denaturants or low pH) and are characterized by having minimal tertiary structure (e.g., fast HDX rates) and substantial, native-like secondary structure (e.g., a compact, partially folded state).45 The presence of secondary structure at 3.0 mM in conjunction with the rapid HDX kinetics makes it is tempting to hypothesize that UBI is populating a molten globule state at 3.0 mM SDS. The CD spectrum of UBI at 3.0 mM SDS, however, is different than the spectrum of native, SDS-free UBI; UBI could thus be considered a molten globule, but not a molten globule with native-like secondary structure.45</p><!><p>We investigated the reversibility of formation of UBI-SDSn with CE (Figure 3), CD spectroscopy (Figure S4a), and HDX measured by electrospray ionization MS (Figure S4b-d). In each experiment, the native (N) or denatured (D) UBI was dialyzed against sub-denaturing concentrations of SDS, so that the UBI-SDS complex was formed from native UBI (i.e., formed by unfolding UBI) or from denatured UBI (i.e., formed by refolding UBI).46 Figure 3 shows the similarity of the electropherograms for UBI-SDSn mixtures at various concentrations of SDS. To facilitate a comparison of the mixtures of UBI-SDSn formed from native and denatured UBI, the electropherograms from both the unfolding and refolding experiments are shown at each concentration of SDS, mirrored across the x-axis. The results demonstrate that the UBI-SDSn complexes produced from either refolding or unfolding have similar electrophoretic mobilities at each concentration of SDS. While they are not identical, the similarities in the mirrored electropherograms indicate that the association and dissociation of SDS and UBI are reversible under these conditions (0-10 mM SDS; 170 h of dialysis).</p><!><p>A lysine-ε-acetyl and N-terminal-α-acetyl protein charge ladder of UBI, prepared by allowing UBI to react with acetic anhydride, provides a useful "charge ruler" for correlating the electrophoretic mobility to the net negative charge on UBI (Figure 4a). We used the acetyl charge ladder to estimate the charge (and thus the composition) of UBI-SDSn complexes having less than eight added charges. To extend the range of negative charges beyond eight, we also prepared a charge ladder derivatized with 4-sulfophenylisothiocyanate (Figure 4b). Modification of reactive amino groups with this compound results in a change in charge of ∼ -2 (ignoring the effect of charge regulation) per acylation, making it possible to determine the stoichiometry of UBI-SDSn complexes associated with up to 14 SDS.</p><p>The first ruler (Figure 4c, Ruler 1) correlates the experimental values of mobility, μn, of rungs of UBI charge ladders with the number of acylations, n, for each rung of the various charge ladders. This ruler includes a total of 12 points (i.e., from 0-8, 10, 12, 14).47 The second ruler (Figure 4d, Ruler 2) represents the values of mobility μn that we estimated after using eq 4 to take into account the change in mass associated with the binding of n SDS molecules. This ruler can be used to determine the number of SDS molecules associated with UBI-SDSn complexes within G1* and G2 (assuming Cp and α are the same for all the protein-derived species to which we apply this equation; Figure 4e).</p><p>The second ruler shows that the conversion of N to G2 (via G1*) involves the stepwise binding of ∼11 SDS molecules to N with no significant changes of secondary structure (according to CD). The estimation of 11 SDS molecules associated with UBI in G2 is also in good agreement with previous work, stating that several proteins (i.e., lysozyme, ovalbumin)10 associate with SDS with the stoichimetry of 0.4 g of SDS per gram of protein at sub-denaturing concentrations of SDS (e.g., [SDS] < cmc).48 One hypothesis to rationalize the results is that the cationic residues nucleate the condensation of SDS in the transition from N to G2, yielding UBI-SDS∼11 (e.g., UBI contains 13 cationic sites, e.g. 7 lysines, 4 arginines, 1 histidine, and 1 unacetylated N-terminus; the extent of protonation of the single histidine and of the N-terminus at pH 8.4 is lower than that of arginine and lysine, and Lys27 is involved in a salt bridge with Asp52;49 these details might explain why fewer than 13 SDS molecules bind in this first transition).</p><p>If the electrostatic binding of SDS to cationic residues is the primary interaction between proteins and SDS at very low concentrations (i.e., the G2 state of UBI), then such electrostatic interactions might help to rationalize the retention of substantial native structure in UBI upon the binding of these first 11 SDS. This interaction would, in effect, neutralize the charge contributed by each lysine group; we know from the work of Makhatadze that the neutralization of lysine and arginine residues in UBI does not change its native structure.40</p><!><p>The equilibrium analysis aims at determining the stoichiometry of Gi (e.g., i = 4 for G4 and i = 5 for G5) by determining the number of SDS molecules involved in the transitions from Gi to Gj, e.g., Δni→j; the stoichiometry of G2 (UBI-SDS∼11) is already known from the charge ladder analysis. For this purpose, we first determined the relative concentrations of UBI-SDS complexes within each group (e.g., [G2], [G4], and [G5]) by integrating the peaks for each group in Figure 1.50 Second, we calculated log([Gj]/[Gi]) (that is, log([G4]/[G2]) for G2→G4, and log([G5]/[G4]) for G4→G5) for each concentration of SDS and plotted these values as a function of log([SDS]). Third, we generated a linear least-squares fit of log([Gj]/[Gi])= Δni→j log([SDS]) + log(Ki→j); this fit made it possible to estimate the slope, Δni→j, and the y-intercept, log(Ki→j). These numbers are, respectively, the number of SDS molecules that bind Gi to form Gj and the logarithm of the equilibrium constant for the transition from Gi to Gj (Figure 5).</p><!><p>Data for μUBI-SDSn(○, obtained using eq 4) and D (□, n = 42)51 were plotted versus n and fitted with a log-normal algorithm. The fitted curve was used to correlate the mobilities of complexes of UBI and SDS within each group to their stoichiometry n (Figure 6a). Interestingly, the stoichiometry of G4 (respectively G5) obtained from this independent analysis based on charge ladders agrees very well with the values from equilibrium analysis (Figure 6b).</p><p>The conversion of G2 to G4 (via G3*) involves the binding of ∼14 additional SDS to G2 (G4 is UBI-SDS∼25). The binding of these 14 SDS molecules to UBI-SDS∼11 causes a major change in the secondary structure (e.g., a 2-fold increase in ellipticity at 220 nm) and in the tertiary structure (e.g., faster rates of HDX) of UBI. We therefore conclude that the binding of a total of ∼25 SDS molecules to UBI is sufficient to "denature" the UBI polypeptide per se (at 25 °C, pH 8.4) into G4, perhaps a non-native molten globule state. The conversion from G4 to G5 occurs at concentration of SDS close to the cmc and was shown to involve the binding of ∼8 SDS to G4 according to the equilibrium analysis (respectively ∼9 SDS according to the charge ladder analysis).</p><p>Above the cmc (3.4 mM), the thermodynamic activity of SDS present in monomeric form is constant. Nonetheless, as [SDS] increases beyond the cmc, more molecules of SDS bind to UBI as we see a subsequent transition from G5 to D (via G6*) involving ∼9 molecules of SDS, based on the use of charge ladders to calibrate the stoichiometry of complexes within G5 and G6*. These last SDS molecules appear to bind to an unfolded (but still structured), non-native polypeptide, and the binding is accompanied by no significant conformational changes (as measured by CD).</p><!><p>Capillary electrophoresis, coupled with equilibrium dialysis and charge ladders, was used to characterize a multistep process for the denaturation of ubiquitin by SDS (Figure 7). The data show the formation of six distinguishable groups of complexes in the presence of sub-denaturing to denaturing concentrations of SDS, four of which have approximately defined compositions (e.g., G2, G4, G5, and D). Three of these UBI-SDSn complexes (G2, G4, and G5) have stoichiometries that involve smaller numbers of SDS molecules than required to saturate the protein (e.g., ∼42 SDS, as estimated from the "1.4:1 rule of thumb" applied to UBI).</p><p>Some of the complexes (UBI-SDS1, UBI-SDS2, ..., UBI-SDS10, UBI-SDS11) have native-like structure; others (UBI-SDS∼25) appear to have a more loosely organized structure (perhaps a molten-globule); still others (UBI-SDSn>25) appear to be completely non-native but still have more α-helical structure than the SDS-free, native protein. None of the complexes appeared to be "unstructured" (e.g., none lacked secondary structure). More importantly, the pathway described in Figure 7 is an oversimplication of what CE is able to detect ("unresolved" groups-e.g., G1*, G3*, G6* - and the shoulder on the G2 peak at 1.7 mM of SDS are not included in this unfolding pathway; we do not know with precision what those complexes are).</p><p>While we initially set out to answer the question of why most proteins show the same ratio of binding (about one SDS bound for every two amino acids at [SDS] > 10 mM), we found that the equilibrium pathway of SDS binding can be complicated (for example here with UBI), and we know from SurfCE25 that it can be significantly different among proteins at sub-denaturing levels of SDS. There must be, therefore, at least two regimes of binding: a sub-micellar regime where proteins show protein-specific pathways of interaction with SDS, and a micellar regime where, to the precision revealed in SDS-PAGE, proteins bind SDS independently of their primary, secondary, and tertiary structures.</p><p>The total number of solvent-accessible positive charges may determine (at least in part) the pathway of binding of SDS to proteins and the composition of intermediates. This type of electrostatic correlation could explain the observation that proteins with a higher content of β-sheets are more resistant to denaturation by SDS than proteins with a higher content of α-helices.52 Several studies that report the thermodynamic β-sheet and α-helical propensities for all 20 amino acids have shown that (i) positively charged residues stabilize α-helical structures (to unfolding by urea) more so than negatively charged and most uncharged residues53,54 and (ii) uncharged and bulky hydrophobic residues (W, Y, F, and I) stabilize β-sheet structures more than positively charged residues.55</p><p>Capillary electrophoresis has the potential to help to answer questions concerning the binding of proteins and SDS (and other charged surfactants, lipids, or other molecules) while providing a level of detail not accessible from other techniques such as 1D or 2D SDS-PAGE, circular dichroism, ITC, or fluorescence binding assays.</p><p>The study reported here suggests answers (or perhaps, more accurately, hypotheses) describing the interaction of UBI, a representative small protein, with SDS. It also demonstrates that CE is a powerful method to monitor the interaction of proteins with small charged molecules. We believe, but cannot yet prove, that the accessibility of charged residues on proteins, combined with secondary structure (α-helical vs β-sheet content) and conformational flexibility, strongly influences the sequence of intermediates formed on binding of SDS to proteins.</p>

PubMed Author Manuscript

Denoising DNA Encoded Library Screens with Sparse Learning

DNA-encoded libraries (DELs) are large, pooled collections of compounds in which every library member is attached to a stretch of DNA encoding its complete synthetic history. DEL-based hit discovery involves affinity selection of the library against a protein of interest, whereby compounds retained by the target are subsequently identified by next-generation sequencing of the corresponding DNA tags. When analyzing the resulting data, one typically assumes that sequencing output (i.e. read counts) is proportional to the binding affinity of a given compound, thus enabling hit prioritization and elucidation of any underlying structure-activity relationships (SAR). This assumption, though, tends to be severely confounded by a number of factors, including variable reaction yields, presence of incomplete products masquerading as their intended counterparts, and sequencing noise. In practice, these confounders are often ignored, potentially contributing to low hit validation rates, and universally leading to loss of valuable information. To address this issue, we have developed a method for comprehensively denoising DEL selection outputs. Our method, dubbed "deldenoiser", is based on sparse learning and leverages inputs that are commonly available within a DEL generation and screening workflow. Using simulated and publicly available DEL affinity selection data, we show that "deldenoiser" is not only able to recover and rank true binders much more robustly than read count-based approaches, but also that it yields scores which accurately capture the underlying SAR. The proposed method can, thus, be of significant utility in hit prioritization following DEL screens.

denoising_dna_encoded_library_screens_with_sparse_learning

Introduction<!>Results and Discussion<!>Notation<!>Tag imbalance<!>Truncated compounds<!>Benchmarking on simulated data<!>Benchmarking on experimental data<!>Discovering SARs<!>Extensions<!>Limitations<!>Software implementation<!>Conclusion<!>Author Information

<p>Since the seminal paper by Clark et al. 1 that saw the concept of DNA encoded libraries (DELs) reduced to practice, the technology has been gaining popularity as a novel hit discovery [2][3][4] and, more recently, target prioritization tool. 5 DELs represent large combinatorial libraries of small molecules that are typically generated using a split-and-pool methodology, with 3 to 4 cycles of chemistry providing routine access to millions or even billions of unique compounds. Unlike traditional chemical libraries, though, each compound in a DEL is attached to a sequence of DNA -the tag or barcode -that stores information on its complete synthetic history, and all DEL members are kept in a mixture. Once a DEL is prepared, affinity selection experiments utilizing immobilized protein targets can be used to capture high-affinity DEL binders from these mixtures, whose chem-ical identity can subsequently be determined by amplification and sequencing of the DNA tags. [6][7][8] We illustrate this process in Fig. 1. The scale and efficiency at which resolution of retained compound identities' can be made has been greatly improved with the introduction of next-generation sequencing 9 and associated statistical analysis 10 into DEL affinity selection workflows.</p><p>Figure 1: a. DNA encoded library consists of a mixture of molecules and a record of its design that provides one-to-one mapping between chemical identity and DNA tag sequence. b. Synthesis is performed in consecutive split-andpool cycles, in each of which tags get extended and building blocks are attached to the ligand. c. The target protein binds compounds with high affinity, which are subsequently sequenced to reveal their identity.</p><p>In "DNA recorded" DEL preparation, 2,11 which is the most commonly used approach, the tag sequence is incrementally built up by ligating short oligonucleotides to the nascent tag in each cycle of chemistry. A unique oligonucleotide sequence thus identifies each building block that went into the library. While, in theory, this DNA encoding scheme should enable one to unambiguously resolve the chemical identity of any DEL member by simply sequencing the corresponding tag, in practice, the correspondence between a given sequence and the chemical composition of a compound is not one-to-one. 12 Specifically, no chemical reaction leads to complete conversion of reactants to the desired product, instead yielding a mixture of the starting reactants, side products, and the expected product. Since tags are extended irrespective of the true chemical composition of compounds they are attached to, it follows that one tag sequence will be associated with more than one specific compound; the converse, products of failed ligation, are removed with HPLC purification. 13 While this issue can be mitigated by optimizing reaction conditions, and profiling building blocks on mock scaffolds so that only high-yielding reactions (e.g. conversion greater than 75-85%) are used in actual DEL generation, it is commonly accepted that every DEL will contain some proportion of truncates -compounds with one or more building blocks missing, when compared to the intended full-cycle products; and that these truncated products are indistinguishable from their full-cycle counterparts on basis of the DNA tag. 12 While presence of truncates in DEL mixtures is an inevitable artefact of the methodology used to prepare these libraries, understanding how they affect the results of DEL affinity screens is of significant relevance. Namely, identifying hits from DEL screens typically leverages the assumption that number of reads mapped to one tag is proportional to the binding affinity of a DEL member associated with that tag. 14,15 This assumption theoretically allows the investigator not only to identify the most potent binders in the library, but also derive structure-activity relationships (SARs) from affinity selection data, when suitable patterns emerge. As such, it has served as basis for interpreting the results of most selection experiments reported in the literature. [16][17][18][19][20][21][22][23] Yet this assumption is, implicitly, subject to a number of constraints, as can be gleaned from previously reported results of computational simulations of DEL affinity selections. 24 Notably, for the count-affinity relationship to robustly hold: (i ) all DEL members must be represented in equimolar amounts at the start of a screen or, alternatively, the starting concentration of each DEL member must be known; (ii ) each tag must unambiguously resolve to a single molecular entity; (iii ) there must be sufficient sequencing coverage and low level of sequencing noise present; (iv ) experimental conditions (e.g. concentration of the target protein, number of affinity selection cycles, number of wash steps per cycle etc.) must be carefully matched to the desired affinity of any binders one wishes to recover.</p><p>In practice, the above listed requirements will seldom be satisfied. Reactions used to generate individual DEL members have unequal yields, and it is not tractable to analytically determine them in complex mixtures. Likewise, as described above, presence of truncates that have tags identical to full-cycle products violates the assumption that compounds are uniquely tagged. Importantly, these truncates can have high binding affinities themselves. Furthermore, amplification and sequencing can introduce additional noise to the final output, constituting another important confounder.</p><p>It is appropriate to note that DEL affinity screens, even without any advanced postprocessing of read counts, are able to recover high-affinity binders. [16][17][18][19][20][21][22][23] However, most of the time, leveraging raw data would lead to erroneous ranking of hits, and yield spurious, lowfidelity SARs. 25 In turn, this could potentially lead to undue experimentation in the follow up to a DEL screen.</p><p>This fact has motivated others to propose specific methods for processing affinity selection outputs. Satz 25 demonstrated that truncated products constitute a major confounder when analyzing binding assays based on raw read counts, and proposed a data aggregation approach to more robustly identify true patterns in selection outputs. Kuai et al. 26 performed a large-scale replicate selection experiment, and demonstrated how DEL selection outputs -of an identical library -tend to be intrinsically noisy, especially at low read counts. The authors further suggested that random noise in these experiments can reliably be modelled using a Poisson distribution, and proposed a specific normalization approach to transform raw counts into an enrichment metric with associated confidence intervals. Similarly, Faver et al. 27 modelled selection data using a binomial distribution, and developed a normalized zscore enrichment metric, demonstrating its utility in quantitative comparison of results from parallel selection experiments. More recently, Gerry et al. 28 described an analysis framework that also takes into account non-uniform abundance of individual DEL members in screened libraries. Leveraging pre-and post-selection read counts modelled by a superposition of multiple Poisson distributions, the authors developed a normalized fold-change score they subsequently used to rank binders.</p><p>Each of the described approaches attempts to correct for a single, or a few of the factors contributing to noise associated with DEL affinity screens. For example, accounting for random noise in sequencing outputs to normalize read counts still does not correct for representation imbalance in a DEL, nor does it account for the fact some of the counts attributed to full-cycle products may be inflated due to binding of identically tagged truncates. Likewise, aggregating read counts over cycles does help identify genuine enrichment in a noisy selection output, but only in a largely qualitative manner.</p><p>Here, we propose a method for processing DEL affinity selection outputs, which can be used to obtain high-fidelity binding affinity estimates for DEL members. Unlike previously reported approaches, our method seeks to account for all the major sources of noise in selection data simultaneously: truncated products bearing tags equivalent to those of their fullcycle counterparts, representation imbalance, and sequencing noise. We base our approach on a previously demonstrated, formally derivable relation between read counts and binding association constants, 24 which we extend leveraging two key assumptions. These two assumptions are both simple and well founded in DEL practice, as will later be discussed. One, under what we term the "null-block model", we assume the majority of DEL members that are not fullcycle products can be treated as simple truncates, i.e. full-cycle product analogues missing one or more building blocks at respective diversity points. Two, we assume most DEL members will exhibit negligible affinity for the target of interest, with only a minor proportion binding with high association constants. With these two assumptions, and using data commonly available within a DEL generation and screening workflow, we developed a sparse learning method that, as demonstrated on simulated and publicly available data, can robustly recover true affinity rankings of DEL binders from inherently noisy selection data. Moreover, since our method also estimates truncates' binding affinity, it enables comprehensive evaluation of SAR. We also describe the implementation details of the method and make it freely accessible to the scientific community as a Python package and command line tool.</p><!><p>First, we introduce the basic notation used throughout this section. Then we show how a naive Bayes model can reduce sequence bias. Next, we proceed to our main result: first, describing our model, then showing how much noise suppression can be achieved on simulated and real experimental data. We finish this section with extensions to our model and details of its software implementation.</p><!><p>Each synthesized compound L r,q is uniquely identified by its DNA tag r = (r 1 , r 2 , . . . r C ), which records the sequence of reactions in which the molecules have taken part, and the list of attached building blocks q = (q 1 , q 2 , . . . q C ). Here, C is the number of synthesis cycles, r c is the index of the chemical reaction in cycle c, and q c is the index of the building block which is actually attached in that cycle. Ideally, q c = r c , indicating a successful synthesis step (which happens with probability Y c,rc ), but if a truncation happens (with probability 1 − Y c,rc ), we denote it with q c = 0, marking the molecules that are missing the corresponding building block. The amount of each compound relative to the ideal amount of the corresponding full-cycle product, L r,r , can be written as a product over reaction yields,</p><p>We denote the set of full-cycle building block combinations q, for which all q c = r c , with F, and the set of truncated building block combinations q, where there is at least one cycle for which q c = 0, with T . The binding assay depletes compounds whose building block combination q is unsuitable for interacting with the target protein. The fraction of molecules that survive C sel number of selection cycles, (each of which consists of equilibrating the library with anchored proteins, washing away free ligands and dissociating the bound complexes), can be written as</p><p>where [P] is the concentration of the target protein and K q is the association constant of the ligand L r,q and the protein in the reaction L r,q :P L r,q + P. (For derivation, see Supporting Information 1.2.) We call S q the survival rate of building block combination q.</p><p>Sequencing the tags of the surviving molecules produces reads that can be mapped to the set of pre-defined tag sequences. After dropping low-confidence and noisy reads, we can summarize the data in a form of read counts N r , each associated with a particular r tag. Taking into account the PCR amplification factor A, we can write the expected read counts as</p><p>where expectation value is denoted by . . . , N tot is the total number of cleaned reads, and k is a protocol-and apparatus-specific normalization constant that mathematically converts DNA concentration to number of sequencing reads.</p><!><p>Presence of DNA tags attached to compounds in a DEL is generally assumed not to govern the outcome of affinity selection assays. However, tags can be unevenly represented in sequencing outputs 28 even in absence of selection pressure. In other words even the [L r,r ] ideal concentrations can be different for different r tags, which, if disregarded, can adversely affect downstream analysis. Fortunately, such an imbalance would be apparent from results of sequencing performed before selection, and such data can be used to correct this bias.</p><p>We were able to investigate the extent of tag imbalance in pre-selection read counts thanks to data published by Gerry et al., 28 which includes the full set of read counts from a 8 × 114 × 119 DEL. In agreement with their findings, we identified two cycle-2 sequences and three cycle-3 sequences associated with higher than 2-fold tag imbalance, (see Supplementary Information 2.1). While the authors have attributed the imbalance to differential tag ligation efficiency, it is pertinent to note inhibition of the ligase (e.g. by leftover building blocks or catalysts), or PCR amplification bias can lead to similar artefacts. The latter can typically be avoided by including a degenerate region in the closing segment of a DNA tag, 13 which ultimately enables PCR deduplication. The tagging strategy employed by Gerry et al. seemingly does not make use of this approach. Irrespective of the source, however, this implies the need to accurately model pre-selection read counts. By assuming that the same experimental protocol is used for obtaining pre-selection results, we can write the expected read counts as</p><p>where A pre and k pre are the amplification factor and the normalization constant, respectively, specific to the pre-selection sequencing experi-ment. By dividing Eq. 3 by Eq. 4, we can write the post-selection read counts as</p><p>(5) Directly estimating the N pre r /N pre tot bias factor from the read counts N pre r is possible only if sequencing depth is at least 10, i.e. for libraries where complexity is lower than the total number of reads by at least one order of magnitude. Since many reported DELs have a numerical size between 10 7 and 10 9 , sequencing to the required coverage would be highly impractical, and resource-intensive, even on contemporary NGS platforms. To enable tag imbalance bias correction for larger libraries, we use a naive Bayes model, where we assume that tags associated with one cycle contribute bias factors independent of tags of other cycles. With this assumption, we write the expectation value of N pre r (denoted by λ pre r ) as a product of (much fewer) b c,rc bias parameters</p><p>The maximum likelihood estimates of the bias factors (see Supporting Information 1.1),</p><p>are robust even if the total number of reads is lower than library complexity. Since the complexity of a single cycle typically does not exceed 3 × 10 3 , a mere total of one million reads is enough to estimate the b factors with at most 5% uncertainty. Substituting estimates of b from Eq. 7 to Eq. 6 yields the estimates λpre r . Using the data of Gerry et al., 28</p><!><p>Ideally, all Y c,rc yields are 100%, the concentration of all truncates [L r,q =r ] are zero, and each expected read count N r is affected by only one compound, L r,r . In reality, most Y c,rc < 100%, which, combined with the possibility that some truncates q can exhibit significant binding to the target protein, i.e. S q > 0, allows truncates to survive the selection process and masquerade as full-cycle products. Here, we construct a statistical model capable of estimating S q up to an unknown global constant from the individual reaction yields Y c,rc and pre-and postselection read counts, N pre r , N r . It is worth noting that Y c,rc are typically unknown, as separa-tion and analytical quantification of individual DEL members is largely impractical. However, as we demonstrate later, yields of individual reactions on mock scaffolds, routinely obtained during building block validation, are a suitable surrogate for actual yields.</p><p>Combining Eqs. 1, 2 and 5 leads to the following linear relationship between N r and the survival probabilities S q . N r = q∈F ∪T X r,q F q , where (8)</p><p>y(r c , q c ),</p><p>where we separated the different factors so that X contains the known variables, and F the unknowns. The F q values, by virtue of being proportional to the survival probabilities S q , indicate how well each building block combination q withstands the selection step; we call F q the "fitness" of q. We assume independent, Poissondistributed sequencing noise,</p><p>which, together with Eq. 8, completes the main definition of our model. One may note that our model can be described as a generalized linear model with Poisson noise and identity link function over non-negative coefficients. This model is under-determined: there are fewer data points N r (r ∈ F) than unknown coefficients F q (q ∈ F ∪ T ). This comes at no surprise, because there is no read count which would directly inform us about truncated compounds. To enable robust inference, despite this difficulty, we turn to sparse machine learning techniques, 29 which work well under the assumption that many coefficients are exactly zero. Such an assumption is naturally fitting for DEL screens, where many of the full-cycle products are expected to not bind to the target protein, i.e. their association constant K q</p><p>[P] −1 , leading to S q ≈ 0 and F q ≈ 0.</p><p>From the long list of published sparse regular-ization strategies, 30 we chose the one used by LASSO 31 because of its mathematical compatibility with the Poisson distribution. Namely, we assume a common exponential prior density for the fitness coefficients,</p><p>As we show below, this choice allows us to average over the fitness of full-cycle products and compute the marginal posterior of the fitness of truncates in closed form. The new parameter, α, controls how strongly the model favors sparse solutions. Practical considerations (see Supporting Information 1.3) suggest that the optimal value is one tenth of sequencing depth. To simplify our exposition, we stick to this choice, but we also give the more general version of all following mathematical formulas, as well as their derivations, in Supporting Information 1.4 and 1.6.</p><p>According to Bayes theorem, the posterior of F , P (F | N ), is equal to the product of the likelihood P (N | F ) and the prior P (F ), up to a normalization constant.</p><p>where λ r = N r is a function of F , defined in Eq. 8. Due to the specific structure of y c (r c , q c ) (see Eq. 1), only one full-cycle product affects each N r read count, the one where q = r, i.e. X r,q = 0 only if q ∈ {r} ∪ T . This allows us to write the expected read count as</p><p>where B r = q∈T X r,q F q denotes the contribution of truncates. Our main goal here is to estimate B r background for all r tags. After that, we will be able to estimate F r directly from N r and X r,r .</p><p>Due to the compatibility of the e −λr and the e −αFq factors, we can average over the full-cycle fitness coefficients, F (F ) = {F q } q∈F , and obtain a closed-form expression for the posterior of truncate fitness coefficients, F (T ) = {F q } q∈T ,</p><p>where Γ r = Γ N r + 1, (1 + 1/X r,r )B r , and</p><p>dz is the upper incomplete gamma function, which is efficiently implemented in numerical software libraries.</p><p>The expression of the marginal posterior P (F (T ) | N ) in Eq. 12 can be numerically maximized with coordinate descent 32 that minimizes the cost function, f (F (T ) ) = − log P (F (T ) | N ). We terminate the optimization once all B r background contributions change less than 0.1 between consecutive iterations. We found that it requires fewer than 100 iteration cycles to converge. Any optimum found is guaranteed to be the global optimum because the cost function f is convex everywhere (see Supplementary Information 1.5 for proof).</p><p>Once the fitness of truncates F q ∈ F (T ) are estimated and the background Br is computed, we can estimate each full-cycle fitness, F r ∈ F (F ) ,</p><p>This takes the prior of F into account, which is apparent from the +1 term in the denominator of the first term, which would be missing from the maximum likelihood result. This completes our quest to estimate F . From the estimates F , we can compute the most likely breakdown of post-selection read counts. This splits each observed read count N r into different N r,q contributions (N r = q N r,q ) each counting how many reads come specifically from ligand L r,q . Under the assumption of Poisson sequencing noise, the posterior of read count breakdown {N r,q } q∈F ∪T is multinomial.</p><p>We use its conditional expectation value to estimate it:</p><p>The estimates Nr,q=r can be regarded as a denoised version of the input data N r , from which the effects of truncates have been statistically subtracted. Although it is tempting to consider N r,q the final output of our model, fitness F is a more reliable metric for distinguishing compounds by affinity, as we show next.</p><!><p>We investigated the accuracy and robustness of our model on data simulated under a large variety of realistic conditions. First, we describe the settings of the simulation, then we summarize our findings.</p><p>Libraries from 1 million to 1 billion compounds were simulated, the default size being 100 million. The number of synthesis cycles were chosen to be 2, 3, 4, 5 and 6, with default being 3. Reaction yields Y c,rc were sampled from a beta distribution with mean from the range [40%, 80%] (default: 70%) and fixed standard deviation of 10 percentage points. The association constants K q were chosen relative to the inverse protein concentration [P] −1 . Their logarithms, log 10 K q , were drawn from a halfnormal distribution (with σ = 1.5, to match realistic spread 25 ) whose minimum was chosen to result in a pre-defined "density", i.e. the fraction of K q values being larger than [P ] −1 . We simulated libraries with densities between 10 −6 and 10 −2 , default being 10 −4 . Fig. 3 shows the distribution of simulated yields and association constants with the default choice of parameters. Sequencing depth, i.e. the average number of reads per DNA tag, were chosen to be between 10 −4 and 10, with default being 0.1. We also investigated how robust our model is against measurement uncertainty of the yields, considering the fact that we expect yields from mock scaffolds to be used in actual computation. To do this, we fed noisy yield values to the model, where noise standard deviation was ranging from 0 to 25 percentage points, default being 15. Finally, we tested how much accuracy we lose if we neglect to correct tag imbalance by not providing pre-sequencing data to the model, forcing it to assume the absence of any bias. The logarithm of tag imbalances log 10 λ pre r were sampled from a normal distribution centered at 0, the standard deviation of which we increased gradually from 0 to log 10 (300%), the default being log 10 (200%). We set the number of selection cycles to C sel = 2. After computing X r,q and F q , from the simulated yields and association constants, we drew each read count N r,q (broken down by q) from a Poisson distribution with mean X r,q F q . The simulated "observed" read counts were computed by the sum N r = q N r,q .</p><p>We conducted seven sequences of numerical experiments, one for each parameter (library size, cycles, mean yield, density, sequencing depth, yield noise, tag imbalance), where we change the value of the selected parameter while keeping all other parameters at default value. To evaluate how well the model performed on the simulated data, we visualize the estimated number of full-cycle reads Nr,r and the estimated fitness values Fq vs the true read counts and association constants. We include the plots here for the simulation where each parameter was set to its default value (100 million compounds, 3 cycles, 70% mean yield, 10 −4 density, 0.1 sequencing depth, 15% yield noise, tag imbalance of 1.0), and provide plots from all runs in Supporting Information 2.2. Square roots of observed N r and estimated N r,r are plotted against the true N r,r and log 10 (K[P]) in Fig. 4, showing that Nr,r is confounded by less noise than raw N r . Fig. 5 shows the fitness of full- Figure 5: Estimated and true fitness F q (scattered points and dashed lines, respectively) for full-cycle products and truncates as functions of log 10 (K q [P]). Association between K q and Fq is strong for both groups. Low fitness values are overestimated. For high fitness values, truncates are accurately recovered, but fullcycle products retain some of the original noise. (We plot the transformed fitness values, i.e. (F q ) 1/C sel , because their true curve is symmetric around its center point at log 10 (K[P]) = 0, and independent of C sel .)</p><p>The main purpose of the DEL screen is to distinguish strongly-binding ligands (which we define with the condition K q > [P] −1 ) from weakly-binding ones. We compute how well different metrics perform in this binary classification problem. We compare the classification performance, i.e. false discovery rate (FDR) and false negative rate (FNR), of three metrics: observed read count N r , estimated fullcycle read counts Nr,r , and estimated fitness coefficients Fr . We define FDR and FNR with</p><p>where false positives (FP) refer to compounds that are above a chosen threshold of the metric but bind weakly, false negatives (FN) are compounds that are below the threshold but bind strongly, and true positives (TP) are the ones where the metric correctly predicts strong binding. One may note that FDR = 1 − Precision and FNR = 1 − Recall, using the usual definitions of Precision and Recall. 33 By plotting corresponding FDR and FNR values for a sequence of different thresholds, we visualize the detection error trade-off 34 (DET) curves of the three metrics. This is shown on Fig. 6 for the simulation with default parameters. At the point, where FDR = 5%, the false negative rates of the three metrics are 99%, 92% and 54%, for N r , Nr,r and Fr , respectively. This means, if one had access to only the raw read counts but needed to keep false discovery rate below 5%, then the acceptance threshold would have to be set so high that 99% of strong binders would be missed. By using the estimated fitness coefficients to assess binding, only 54% of the strong binders would be missed at the same FDR level, a 50-fold increase in Recall.</p><p>To determine under what circumstances a particular metric performs well, we selected two indicators: First, the FNR at the point where FDR = 5%, which indicates how reliably a metric is able to distinguish strong binders from weak binders. Second, the Spearmancorrelation between the metric and K q among strong binders, which measures how accurately can the metric rank compounds by their binding strength. We plot these metrics for six of the seven sequences of numerical experiments in Fig. 7, and for tag imbalance in Fig. 8.</p><p>For the observed counts N r , FNR stays above 90%, whereas Fr can achieve significantly lower FNR under all tested library sizes, cycles, mean yields, density values highlights the usefulness of the null-block model as a post-processing step.</p><p>Although FNR and Spearman rank correlation reflect different facets of model accuracy, their trends are mirror images of each other: when FNR is low, Spearman correlation is high for Nr,r and Fr across all seven parameters.</p><p>Metric Nr,r fails to improve Spearman correlation beyond the level already achievable with N r , and improves FNR only marginally. Because computing Fr is possible from Nr,r only if we know the yields with some degree of accuracy, this highlights that knowing the yields is important for being able to predict stronglybinding compounds. This is also highlighted by the 6th sub-figure column in Fig. 7, which shows that FNR is increasing and Spearman correlation is decreasing quickly as the measurement uncertainty of the yields grows beyond 0.1.</p><p>The most difficult conditions, where all three metrics have a difficult time estimating strong binders, are low mean yield (≤ 0.5), low sequencing depth (≤ 2), high density of strong binders (≥ 3%), and high yield noise (≥ 0.15). It is encouraging to see that the performance of Nr,r and Fr are constant with increasing library size, suggesting that libraries much larger than the ones we simulated can be accurately analyzed with our model. As the number of synthesis cycles is increasing from 3 to 6, the null-block model loses some accuracy, but this is less of a concern, because the number of cycles is limited to 4 in most DELs.</p><p>Sensitivity to tag imbalance shows a similar trend as sensitivity to yield noise, but here we Figure 7: False negative rate (at the point where false discovery rate is 5%) for three different metrics (N r , Nr,r , Fr ) and Spearman's rank correlation between the metric and the true association constant K r among strong binders, plotted as functions of six simulation parameters: library size, number of synthesis cycles, mean of reaction yields, sequencing depth, density (fraction of strong binders), and noise on measured yields. Vertical dashed lines mark the default values of each parameter, which was used in all other sequences of simulations where other parameters were changed. distinguish between two scenarios. First, we do not use pre-selection sequencing data, and implicitly assume that there is no tag imbalance. Second, we fit the naive Bayes model to pre-sequencing read count data and correct tag imbalance. Fig. 8 shows that correcting tag imbalance greatly improves FNR and Spearman correlation for the metric Fr , but does nothing for Nr,r . This is expected, since the true read count breakdown N r,r (which Nr,r estimates) is subject to the same tag imbalance bias as the observed read counts N r .</p><!><p>We show that high fitness values, estimated by the null-block model, are good indicators of strong binding on real experimental data. Once again, we make use of the supplementary data published by Gerry et al. 28 We take advantage of the fact that index 23 denotes a "null" reaction in cycle 3, as designed by the authors. All read counts N (r 1 ,r 2 ,r 3 ) where r 3 = 23 actually measure the fitness of one of the truncates q = (r 1 , r 2 , 0), where r 1 and r 2 where σ is the standard deviation of log 10 λ pre r , which is normally distributed, e.g. tag imbalance of 0.5 means that σ = std(log 10 λ pre r ) = log 10 (1 + 0.5) = 0.17.) are non-zero cycle-1 and cycle-2 indexes, respectively. First, we run our model on the pre-and post-sequencing data, and obtain estimates of all truncate fitness values F q , for q = (r 1 , r 2 , 0), without informing our model about the existence of a "null" reaction (in fact, using Y 3,23 = 0.99, which implies that r 3 = 23 is a proper reaction). Then we compute the observed fold change, N r /N pre r for all r of the from (r 1 , r 2 , 23). In Fig. 9, we compare the two Figure 9:</p><p>Left: Comparison of estimated fitness Fq and observed fold change N r /N pre r for cycle-3 truncates, i.e. q = (r 1 , r 2 , 0), and r = (r 1 , r 2 , 23), because reaction 23 in cycle 3 is a "null" reaction. The Pearson correlation of the two from the experiment "hrp_exp_r2" is 0.834. Right: Person correlation from all six experiments. (Scatter plots of all six experiments are included in Supporting Information 2.3.) metrics, and compute their Pearson correlation. Their correlation is statistically significant in all six experiments, and especially strong in the two "hrp_exp" data sets. This shows that estimated fitness of truncated compounds, which is central for denoising, is consistent with what one can obtain with a more direct measurement of the truncates.</p><p>We further evaluated the concordance of results obtained using our method to those discussed in Gerry et al. 28 For horseradish peroxidase (HRP), the authors observed a strong dependency between enrichment and the electrophilic character of the N-capping group (building block 1, BB1), further highlighting three sulfonyl chloride-derived Michael acceptors, ranked according to their electrophilicity, which furnished compounds of highest affinity for HRP. As can be seen in Fig. 10, results obtained using "deldenoiser" consistently recapitulate the aforementioned dependency, which is broadly observable across most scaffolds and the second diversity point, suggesting the latter two play only a minor role in governing binding. Similarly, for carbonic anhydrase IX (CA9), "deldenoiser" successfully recovered aryl sulfonamide-based building blocks at the second diversity point (building block 2, BB2) as privileged substructures, particularly in trans stereoisomers of azetidine derivatives (e.g. scaffold ID 8); the activity of these products was experimentally validated off-DNA by Gerry et al. Finally, the preference for para-sulfonamide building blocks over their meta-positional isomers is also captured (Fig. 10). It is pertinent to note these trends do not become readily appreciable when using a simple count-based enrichment metric (see Figure 10 in Supporting Information), primarily due to numerous outliers; however, an aggregate visualization does render them obvious.</p><!><p>Structure-activity relationships (SARs) manifest themselves as elevated read counts along one or more reaction indexes. E.g. if the combination of reaction 47 in cycle-1 and reaction 38 in cycle-2 already produces a structure that binds strongly to the target, irrespective of the cycle-3 reaction, then most r = (47, 38, r 3 ) tags will have high read counts. Unfortunately, such a data set is similar to what a high-affinity truncate q = (47, 38, 0) would produce. Still, we can clean the read count data from the effect of potential truncates and faithfully retain the SAR signature, if the design of the DEL includes "null" reactions, i.e. synthesis steps where nothing is done to the compounds (see details in "Extensions" subsection). To continue our example, let us imagine that r 3 = 101 is a "null" reaction. The, usually low, read count N (47,38,101) prevents the fitness of the affecting truncate, F (47,38,0) , from being overestimated. Accurate estimation of the truncate fitness, even if it is much smaller than all full-cycle fitness values, enables accurate estimation of the entire SAR series.</p><p>We demonstrate this capability of our model by first simulating DEL results with default parameter settings, (see "Benchmarking on simu-Figure 10: Results from the Gerry et al. 28 affinity screens, reanalyzed using "deldenoiser". Top: Ratio of mean fitness values (across replicates) for protein-loaded vs. beads-only affinity selections against horseradish peroxidase (HRP), grouped according to scaffold, and colored by the presence of a specific building block, as summarized in the legend. Inset bar chart provides an aggregate representation of the data, grouped by the identity of BB1, with height of the bars reflecting the mean fitness ratio across all species with the corresponding building block. Bottom: Corresponding results for carbonic anhydrase IX. Inset bar chart reflects grouping on the identity of BB2. lated data" subsection), and adding ten artificial SAR series to each data set. We randomly selected ten 1-index SAR series, where values of all except one r c reaction indexes are fixed, and drew their association constants from the K q ≥ [P] −1 tail of their distribution. We chose all K q [P] of the confounding truncates to be a fixed value from 0.01, 0.03, 0.1, 0.3, 1, 3, 10, 30, and 100. We used the direct estimation formulas (see Supporting Information 1.10) to fit our model. To benchmark our model, we compute the same metrics as before, but this time, we focus on the compounds that are part of the SAR series: i.e. Spearman correlations and false negative rates are evaluated only on the set of these compounds. Fig. 11 shows the false negative rate at the threshold where the false discovery rate is 5% (which we compute among all full-cycle compounds), and Spearman correlation of the three metrics (N r , Nr,r and Fr ) among SAR compounds, plotted as a function of log 10 (K q [P]) of the truncate q that directly confounds the SAR series. For Fr , false negative rate hovers around 0.5, independent of truncate binding strength, while Spearman correlation slowly decreases from 0.75 to 0.6 as log 10 (K[P]) increases from -2 to +2, a significant improvement over what is achievable with the two other metrics N r and Nr,r . This suggests that our model, with the help of read counts of "null" reactions, can accurately separate the effects of truncates, no matter their affinity, and recover</p><!><p>The null-block model, as presented above, can be extended in several ways. Here we give a summary of the different avenues, and details can be found in the Supporting Information.</p><p>First, if the DEL is designed to include "null" reactions, as has commonly been reported, 17,20,21,35 then the null-block model can be adapted to take this information into account and estimate truncate fitness not only from the background contributions B r , but also from the read counts corresponding to these "null" reactions. Setting Y c,rc = 0 for the "null" reaction indexed by r c in cycle c results in X r ,r = 0 for all r where r c = r c . This forces the model to try to explain the read count N r using only truncates, providing a boost to the accuracy of estimating truncate fitness values, which translates to more accurate estimates of the full-cycle products. Such an input data also enables direct fitting of our model, as we explain in Supporting Information 1.10.</p><p>Second, although we found that the naive Bayes model can account for the majority of the effects resulting in overdispersion of the read counts, one may wish to model the remaining dispersion (which we found to be about 3.37 for data from Gerry et al.). This can be efficiently done by replacing the Poisson distribution in Eq. 9 with a dispersed Poisson distribution, Third, so far we were dealing with the problem of estimating the fitness F q , and we did not discuss what is needed to estimate the survival chance S q and association constant K q . The added difficulty stems from not knowing the factor k/k pre . To calibrate this value, we need additional information. Sequencing on Illumina machines is performed with an additional DNA spike-in to be used as positive control, and which helps maintain diversity in libraries originating from DELs. 15 Sequences from the virus PhiX are added to the prepared DNA library in controlled amount. Taking note of the added amounts and the number of reads mapping to PhiX genome provide sufficient information to estimate the ratio k/k pre . Alternatively, one could use a compound with known binding affinity, resynthesized on-DNA, and spiked into the library at a reasonable concentration. Combined with the PCR amplification rates A and A pre , which can be estimated from the experimental protocol, one can compute the proportionality constant between F q and S q , enabling direct computation of Ŝq from the estimated fitness Fq . Then, the protein concentration [P] and the number of selection cycles C sel can be used to estimate K q . Formulas can be found in Supporting Information 1.7. Note that depending on the details of the experimental protocol, equations more involved than Eq. 2 may be needed to establish the relation between S q and K q , but this is beyond the scope of this article.</p><p>Fourth, our model is equally applicable for pooled DEL screens. After separating the data from a pooled screen by libraries, our algorithm can be run on each part separately to obtain fitness estimates. Added difficulty is created by the fact that concentrations of the pooled libraries may be uncertain. To overcoming this, pre-selection sequencing data is crucial. Previously we used pre-selection sequencing data to correct tag imbalance due to unequal sequence generation, but the exact same method corrects the effect of unequal concentrations between libraries. More details can be found in Supporting Information 1.8. Furthermore, analyzing pooled libraries is directly parallelizable, saving wall clock time.</p><p>Finally, our formalism allows developing more complex truncation models. Whenever the exact same side product get produced alongside different full-cycle products, correlated noise is present in the observed data, which opens the possibility of deconvolving the side product from the full-cycle products. Only the formula for computing the X matrix needs to be changed to incorporate reaction branches, and the rest of the machinery will function without change.</p><!><p>An inherent difficulty in developing a method for analysis of DEL affinity selections is the lack of experimental data that can serve as a robust ground truth dataset in method validation. Ideally, one would desire having association constants for several hundred compounds on-DNA, along with the sequencing data produced after the affinity selection of the parent library. In this scenario, one would not need to account for factors contributing to differences in binding affinity when hits are resynthesized off-DNA, facilitating a straightforward comparison. In reality, however, published data typically includes only biochemical or biophysical affinity measurements for off-DNA compounds, with the number of data points seldom exceeding 10-20, and with no or incomplete sequencing data released. In validating the described method, we therefore resorted to heavily leveraging simulated data. While we made every effort to ensure this data was generated by incorporating reasonable assumptions pertaining to the affinity selection and sequencing steps, these involve complex thermodynamic processes, and are subject to numerous potential experimental artifacts. This renders any simulated data only a modest surrogate for real selection outputs. Correspondingly, prospective validation of our method is highly warranted, and will be instrumental in confirming its practical utility. None the less, the results we obtained by analyzing data from Gerry et al. -which, to date, presents the most comprehensive publicly available DEL selection dataset -provide encouraging evidence to this end. It should be noted, though, that even this dataset contains only 8 IC50 measurements for library members synthesized off-DNA, precluding any statistically robust assessment. Further advances in algorithmic developments supporting DEL screens, as well as systematic cross-method benchmarks, would benefit greatly from broader availability of additional datasets akin to that released by Gerry et al.</p><p>Beyond real-world validation of the method we described, its underpinnings features a key limiting assumption that is worth further underlining. The null-block model assumes all library members are either full-cycle (intended) products or their truncated counterparts. Yet, chemistry involved in many library designs is clearly conducive to the generation of highly diverse arrays of side products. These are challenging to account for on several grounds: (i ) their presence does not typically result in correlated noise; (ii ) they render library complexity an almost arbitrary parameter; (iii ) no estimates on yields of individual side products can be obtained. Although a more complex model can be envisioned to address these challenges, we opted against pursuing it. For one, most practitioners will choose to include only highyielding building blocks into the final library, and will optimize designs and reaction conditions to minimize the generation of alternate products. Secondly, the bulk of DEL literature cites truncates as key contributors to spurious SAR, with significant utility coming from determining their putative affinity. Finally, a more complex model would ultimately be of questionable utility, given the inherent difficulties in fitting it, and validating its performance. Prospective adopters should, none the less, be aware of this limitation, as libraries that are suspected to contain complex side product mixtures in non-negligible quantities may not be suited for analysis by this approach.</p><!><p>The null-block model can be fitted with our software implementation "deldenoiser". We developed a python package and command line tool under the same name, and made them available under GNU General Public License v3.0 at https://github.com/totient-bio/ deldenoiser.</p><p>The input data about reaction yields Y c,rc , pre-and post-selection read counts N pre r , N r , are processed and output files are created that contain the estimated read count breakdown Nr,r and fitness coefficients Fq . The logical flow is shown in Fig. 12.</p><p>Our implementation, based on python's numpy framework, 36 takes 12 minutes to run for a library of 100 million compounds, using 8 CPUs and 1.5 GB of memory. Analyzing bigger libraries require more computational resources. Fig. 13 shows how running time and peak memory usage increases with increasing library size, reaching 5.5 hours and almost 2.6 GB for a library of 1 billion compounds.</p><!><p>We developed and benchmarked a statistical method capable of reducing the noise affecting sequencing results of DNA encoded libraries due to truncated compounds. Numerical experiments conducted with simulated data showed that one can select and rank strongly binding compounds more reliably with the metric produced by our model, compared to using the raw</p><!><p>M.K. identified the problem, set the scope of the work, and advised on numerical experiments. P.K. designed, implemented the benchmarked the statistical model. P.K. and M.K. co-wrote the manuscript and Supporting Information.</p>

ChemRxiv

On the evolving open peer review culture for chemical information science

Compared to the traditional anonymous peer review process, open post-publication peer review provides additional opportunities -and challenges- for reviewers to judge scientific studies. In this editorial, we comment on the open peer review culture and provide some guidance for reviewers of manuscripts submitted to the Chemical Information Science channel of F1000Research.

on_the_evolving_open_peer_review_culture_for_chemical_information_science

Introduction<!>Pre-review and review<!>Open peer review specifics<!>Guidelines<!>Conclusions

<p>The Chemical Information Science (CIS) channel of F1000Research has been introduced as a publication platform 1 that covers all aspects of chemical information science 2 and positions the full spectrum of chemoinformatics approaches 3 within this broader context. The CIS channel specifically aims to attract high-quality manuscripts. Therefore, submissions to the CIS channel undergo a two-layer expert review, as described below. This editorial is intended to provide specific guidance for reviewers of studies published in the CIS channel.</p><!><p>The review of papers submitted to the CIS channel takes place in two stages, an initial pre-review by members of the channel Editorial Board, followed by open peer review. Once a submission has been processed by F1000Research editorial staff and passed on to the guest editors, members of the channel Editorial Board evaluate a manuscript on the basis of its scientific potential to advance the field. This initial assessment (pre-review) is not meant to result in formal reviews, but a collection of expert opinions. The conclusions of the channel Editorial Board are then forwarded to the authors. If a positive pre-review consensus is reached or if views of the channel Editorial Board on a submission remain controversial, the paper is published in the CIS channel and reaches the stage of open peer review. If a negative pre-review consensus is reached by the channel Editorial Board, the manuscript is not published in the CIS channel (but the authors have still the opportunity to publish their work in F1000Research).</p><p>Upon publication of a paper in the CIS channel the authors are asked to make reviewer suggestions; members of the channel Editorial Board may suggest additional reviewers. Authors must agree with the final reviewer line-up before F1000Research editorial staff initiates the post-publication review. The review, approval, and indexing process of CIS channel publications follows standard F1000Research procedures.</p><!><p>The open post-publication peer review presents referees with different opportunities and challenges compared to the conventional anonymous peer review process. The general philosophy of open peer review is that the reviewer identity will be disclosed and the review directly presented to the scientific community including the authors (without editorial interference). In addition, authors and readers have the opportunity to comment on reviews. In the following, we provide some specific comments and guidelines for reviewers of CIS channel publications.</p><!><p>(1) The primary function of a review is to evaluate whether a given study is scientifically sound, understandably presented, and reproducible. Frequent lack of reproducibility is a major issue concerning chemoinformatics publications in many journals 4. Reviewers of CIS channel publications must determine whether data and methods used in a given study are accessible to the scientific community and that sufficient details are provided to reproduce reported calculations and re-implement a method (provided an implementation of the method is not made available as part of the study). Answering these key questions should directly lead the reviewer to conclude that a study should be "approved", "approved with reservations", or "not approved". The reviews can be brief as long as they clearly address the key questions.</p><p>(2) Because these questions are among those already considered during the pre-review, members of the channel Editorial Board are encouraged to convert/extend pre-review comments into a post-publication review. This will inevitably reduce the time required for a CIS channel publication to reach approval status (or a status requiring revisions).</p><p>(3) Open peer reviews not only provide feedback for authors, they also help to position a paper within the CIS channel and spark the interest of the scientific community. As such, these reviews and subsequent on-line discussions become an essential part of a publication. An open review process can also dramatically reduce the time between submission, publication, and indexing of a paper, thus supporting its dissemination. Short review times are highly desired and particularly important for Data or Method articles, which often report tools made freely available to the scientific community. In addition, short review times are an additional attraction for authors to submit their work to the CIS channel.</p><p>(4) Of course, reviewers are at liberty -and encouraged- to provide detailed reviews, which might also suggest more or less extensive revisions. This particularly applies to Research Articles or Reviews. It is also appropriate to further extend reviews of a paper after approval status is reached. This can be accomplished, for example, by adding comments to initial reviews. We expect that diligent authors will take reviewer comments seriously and submit revisions and/or responses. If authors disagree with review conclusions or requests, they can comment on them and articulate their viewpoints. Authors are specifically encouraged to publish appropriate revisions in a second version of their manuscript. If they do or do not take reviews seriously will be clearly visible to the scientific community; another bonus of an open review culture.</p><p>(5) Answering the key questions if a study is scientifically sound, clearly presented, and reproducible in a timely manner becomes especially important for off-the-beaten path contributions, which are explicitly encouraged by the CIS channel. For example, such papers might introduce novel, provocative, and/or controversial concepts that are far from being established, report negative results, or principal shortcomings of current methods. In such cases, views of authors, reviewers, and readers might often differ. Regardless, conceptually novel or controversial investigations that are viewed differently must still be scientifically sound. Even in the presence of different opinions, a careful assessment of the key questions is an essential task for reviewers of such CIS channel publications.</p><!><p>We hope that our comments will help to foster a culture of open peer review, for the CIS channel and beyond. As discussed, open peer reviews are not written for editors but directly address authors and the scientific community. As such, they become a part of a publication and are thought to make important contributions to the further scientific development of our field. Open peer reviews must evaluate the key questions whether a publication is scientifically sound, understandably presented, and reproducible and may go well beyond answering these questions. Short review times are important when a paper is presented to the scientific community. Timely reviews, be they positive or negative, indicate that studies are taken seriously, make publications more interesting to readers, and help to disseminate them. Open peer review catalyzes scientific perception. The scientific community has the last word.</p>

PubMed Open Access

The Hydrogen Catalyst Cobaloxime \xe2\x80\x93 a Multifrequency EPR & DFT Study of Cobaloxime\xe2\x80\x99s Electronic Structure

Solar fuels research aims to mimic photosynthesis and devise integrated systems that can capture, convert, and store solar energy in the form of high-energy molecular bonds. Molecular hydrogen is generally considered an ideal solar fuel as its combustion is essentially pollution-free. Cobaloximes rank among the most promising earth-abundant catalysts for the reduction of protons to molecular hydrogen. We have used multifrequency EPR spectroscopy at X-band, Q-band, and D-band combined with DFT calculations to reveal electronic structure and establish correlations between structure, surroundings and catalytic activity of these complexes. To assess the strength and nature of ligand cobalt interactions, the BF2-capped cobaloxime, Co(dmgBF2)2, was studied in a variety of different solvents with a range of polarities and stoichiometric amounts of potential ligands to the cobalt ion. This allows the differentiation of labile and strongly coordinating axial ligands for the Co(II) complex. Labile, or weakly coordinating, ligands like methanol result in larger g-tensor anisotropy than strongly coordinating ligands like pyridine. Additionally, a coordination number effect is seen for the strongly coordinating ligands with both singly-ligated LCo(dmgBF2)2 and doubly-ligated L2Co(dmgBF2)2. The presence of two strongly coordinating axial ligands leads to the smallest g-tensor anisotropy. The relevance of the strength of the axial ligand(s) to the catalytic efficiency of Co(dmgBF2)2 is discussed. Finally, the influence of molecular oxygen and formation of Co(III) superoxide radicals LCo(dmgBF2)2O2\xe2\x80\xa2 is studied. The experimental results are compared with a comprehensive set of DFT calculations on Co(dmgBF2)2 model systems with various axial ligands. Comparison with experimental values for the \xe2\x80\x9ckey\xe2\x80\x9d magnetic parameters like g-tensor and 59Co hyperfine coupling tensor allows the determination of the conformation of the axially ligated Co(dmgBF2)2 complexes. The data presented here are vital for understanding the influence of solvent and ligand coordination on the catalytic efficiency of cobaloximes.

the_hydrogen_catalyst_cobaloxime_\xe2\x80\x93_a_multifrequency_epr_&_dft_study_of_cobaloxime\xe2\x80

Introduction<!>Sample preparation<!>EPR spectroscopy<!>Density Functional Theory calculations<!>Experimental Results - EPR under anaerobic conditions<!>Correlation Dependence of g- and A-tensors<!>Experimental Results - EPR under aerobic conditions<!>DFT Results<!>Cobaloxime with axial ligands (except molecular oxygen)<!>Cobaloxime with oxygen and PPh3 as axial ligands (molecular oxygen adduct)<!>Conclusion

<p>The production of carbon-neutral and sustainable fuel sources by solar energy conversion is considered to be a vital component of our future energy landscape.1,2 Biology provides several examples for processes necessary to transform abundant solar energy to energy stored in the form of chemical bonds in small, energy-dense molecules, such as molecular hydrogen. Molecular hydrogen is generally considered as an ideal energy carrier, since its combustion is essentially pollution-free. A major goal in the field of solar H2 production is to generate molecular systems which ideally have both a high turnover number (stability of the molecular device) and high turnover frequencies (efficiency of catalyst) using earth-abundant first-row transition metals rather than unsustainable precious metals. A variety of different catalysts designed towards this goal have been structurally and mechanistically characterized for the reduction of protons to molecular hydrogen (reviewed in 2,3). However, both the efficiency and stability of these devices still need significant improvement. The development of an efficient system for hydrogen generation driven by sunlight has even been termed a "Holy Grail" of solar-driven water splitting.3 Among currently known complexes, cobaloxime compounds (cobalt macrocycles with di-glyoxime ligands) have been emerged as productive electrocatalysts for the reaction 2H+ + 2e−→ H2 with low overpotential in the presence of a proton source in organic solvents. Cobaloximes are not only among the best synthetic transition metal complexes known for H2 production, they are also relatively easy to synthesize, oxygen tolerant, are amenable for coupling to natural and artificial photosynthetic systems,3,4 and rely only on earth-abundant materials. The pseudo-macrocyclic cobaloxime compounds were originally developed and investigated as vitamin-B12 analogs,5,6 and have been extensively studied as a model system for vitamin B12.7–9 About twenty years later, Connolly & Espenson found that difluoroboryl-cobaloxime Co(dmgBF2)2 catalyzes proton reduction in acidic media.10 The urgent need for earth-abundant materials capable of solar energy conversion has prompted renewed interest in the catalytic properties of cobaloxime compounds. Both electrocatalytical systems for hydrogen evolution which utitilize cobaloxime as well as systems using photosensitizers were reported.3,11–18 Recently, the first photocatalytic supramolecular assemblies based on pyridyl coordination of a variety of photosensitizers to the Co(II) center of the cobaloxime macrocycle have been described.4,19–22 However, a significant problem for many cobaloxime compounds is their low stability at low pHs and rapid decomposition, which limits their general utility for hydrogen catalysis, while the (BF2-capped) difluoroboryl-cobaloximes –like Co(dmgBF2)2 (Figure 1) – have been found to be more resistant toward acidic hydrolysis than H-capped cobaloxime compounds.3,10,23 In addition, the reduced nucleophilicity of its hydride derivative limits undesired hydrogenation reactions. Furthermore, for the BF2-capped cobaloxime, the ground state is more easily reduced with its Co(II)/Co(I) potential about 0.5 V more positive than the H-capped cobaloxime.3 Another benefit is that the presence of a conjugated bridging ligand facilitates the transfer of electrons to the cobalt ion. The BF2-capped cobaloxime Co(dmgBF2)2 studied here is quite resistant to the oxidation by molecular oxygen and the ensuing degradation of the cobalt macrocycle. These factors make it a better model system for studying the interaction of the cobalt complex with molecular oxygen, as compared to H-capped cobaloxime.24</p><p>The catalytic efficiency is determined to a large extent by the electronic structure of the cobaloxime. The electronic structure in turn is strongly influenced by the macrocycle's surroundings and in particular by the ligand(s) directly bound to the central metal ion. Also, other secondary interactions, such as hydrogen bonding to the fluorine atoms and the dielectric properties of the medium cannot be neglected. An excellent illustration of the importance of the surroundings is the redox potential. The cobaloxime Co(II)/(I) reduction potential is sensitive to its coordination environment and shifts significantly on pyridyl coordination to the Co(II) center.3,4,21 Thus, the knowledge of the electronic properties is essential for an understanding of the catalytic properties of the complex.</p><p>So far, only cursory experimental data regarding changes in the electronic structure as a result of axial coordination are available. The resting state of cobaloxime's catalytic cycle is 3d7 low spin Co(II).25 Since the highest occupied molecular orbital (HOMO) of the Co(II) complex, which carries a single unpaired electron, has mainly dz2 character, the effect of axial coordination by fifth and sixth ligands to the cobalt ion is crucial for the function of cobaloxime, and particularly in catalytic processes involving axial ligand exchange. Electron paramagnetic resonance (EPR) spectroscopy is an excellent tool to characterize any changes in electronic structure which result from changes to the surrounding environment or properties of the orbital carrying the unpaired electron. Changes in the electronic properties can then be correlated with different catalytic efficiencies of cobaloxime.</p><p>Several EPR studies on cobaloxime compounds exist but there are number of inconsistencies and gaps in the published data, e.g. refs 24–35. First, almost all studies focus on H-capped cobaloxime and to the best of our knowledge, there has been only one previous cursory EPR study of Co(dmgBF2)2.25 The second drawback is that most previous cobaloxime EPR studies (including the one by Bakac and co-workers on Co(dmgBF2)2) performed EPR spectroscopy only at the conventional X-band frequency (9–10 GHz). In most cases, the X-band spectra are very congested, with g-tensor anisotropy and 59Co hyperfine interaction having the same order of magnitude, which makes it difficult to determine the principal components of the g-tensor, leaving one of the most important parameters far from precisely determined. In fact, in many studies both g-tensor and the 59Co A-tensor were reported as axial, whereas it is shown in our study that in nearly all cases both tensors are essentially rhombic. Furthermore, the previous studies were either lacking quantum mechanical calculations completely or the treatment was restricted to Hückel theory.</p><p>Here we use EPR spectroscopy at X-band (9 GHz), Q-band (34 GHz), and D-band (130 GHz) microwave frequencies to distinguish clearly between field-dependent and field-independent parameters. The multifrequency approach allows us to determine g-tensor anisotropy and hyperfine splitting due to the central metal (59Co) and coordinating solvents molecules/ligands (L) that have 14N, 15N, or 31P magnetic nuclei. A variety of different solvents from different groups are studied, including (i) polar protic solvents, (ii) polar aprotic solvents, (iii) non-polar solvents, and with and without potential ligands to the central cobalt ion. The experimental investigation were accompanied by an extensive Density Functional Theory (DFT) study on the cobaloxime, taken into account a variety of different possible conformations of the cobaloxime macrocycle and ligand/solvent molecules.</p><p>The catalytically relevant ligands studied both in experiment and calculation include the "small and simple" ligand pyridine as well as two rather bulky organic molecules, N-cyclohexyl-N′-4-pyridyl-1,7-dipyrrolidinylperylene-3,4:9,10-tetracarboxylic acid bisimide (PDI) and triphenylphosphane (PPh3) (Figure S1). The PDI scaffold is widely used in many light-harvesting applications due to its high extinction coefficient, broad absorbance and redox tunability, exceptional structural integrity, and thus presents an attractive alternative to other photosensitizers which contain precious metals. PDI derivatives are most often employed as acceptor fragments in donor-bridge-acceptor constructs.36 A supramolecular cobaloxime assembly was recently reported with a pyridyl-functionalized PDI molecule coordinated to cobaloxime, acting as a photosensitizer.21 Triphenylphosphine (PPh3) is the second bulky ligand studied. The central phosphorous atom acts as the ligating atom of PPh3 where the 31P nuclear spin of I = ½ allows the hyperfine interaction of the phosphorus atom with the unpaired electron to be monitored. Interaction of cobaloximes with molecular oxygen were studied on the example of a O2: Co:PPh3 complex which efficiently formed a stable, but reversible, paramagnetic adduct. The oxygenation is found to dramatically change the EPR spectrum of cobaloxime, resulting in a pronounced reduction of g-tensor anisotropy and decrease in both 31P and 59Co hyperfine interaction, which is consistent with a pronounced shift of electron spin density to molecular oxygen. We anticipate that this comprehensive study of the changes in the electronic properties in response to a variety of environmental factors can be correlated with demonstrated catalytic efficiencies of cobaloxime and eventually provide a predictive tool for the design of new cobaloxime catalyst structures.</p><!><p>All commercial reagents including organic solvents were of ACS grade and purchased from Sigma-Aldrich unless otherwise noted. 1H NMR (500 MHz) was performed on a Bruker DMX 500 spectrometer and referenced to TMS or residual solvent peak as an internal standard. ESI-MS was conducted on a ThermoFisher LCQ Fleet from dilute acetonitrile or methanol solutions. UV-Vis absorbance measurements were performed on a Shimadzu UV-1601 spectrophotometer. The synthesis of the cobaloxime Co(dmgBF2)2•H2O was carried out as previously described and matched all characterization methods.23 Solvents were methanol, glycerol/H2O, acetone, toluene, dichloromethane (CH2Cl2), and dimethylformamide (DMF). The synthesis of N-cyclohexyl-N′-4-pyridyl-1,7-dipyrrolidinylperylene-3,4:9,10-tetracarboxylic acid bisimide (PDI) was described previously.21 15N-labeled pyridine was obtained from Cambridge Isotope Laboratories (Andover, MA). Co(dmgBF2)2•H2O was dissolved in the respective solvent (or solvent mixture, possibly containing potential ligand molecules) to yield a 1–3 mM solution. The solutions were thoroughly purged with N2 before filling the quartz tubes used for the EPR measurements under nitrogen atmosphere in a dry box. The samples were then frozen quickly in liquid nitrogen. The samples of the oxygen adduct LCo(dmgBF2)2O2, were prepared outside the dry box under aerobic conditions and frozen quickly in liquid nitrogen.</p><!><p>CW X-band (9 GHz) EPR experiments were carried out with a Bruker ELEXSYS E580 EPR spectrometer (Bruker Biospin, Rheinstetten, Germany), equipped with a TE102 rectangular EPR resonator (Bruker ER 4102st) and a helium gas-flow cryostat (Air Product, Allentown, PA). The temperature was controlled by a Lakeshore cryogenic temperature controller (Westerville, OH).</p><p>Pulse X-band experiments were performed on the same spectrometer, using a Flexline dielectric ring resonator (Bruker ER 4118X-MD5 or Bruker EN 4118X-MD4-W1) and a helium gas-flow cryostat (CF935, Oxford Instruments, UK). The temperature was controlled by an ITC (Oxford Instruments, UK).</p><p>CW Q-band (34 GHz) EPR experiments were carried out with the same EPR spectrometer, equipped with a Q-band bridge (Bruker ER 051 QG) a cylindrical EPR resonator (Bruker ER 5106 QT-W) and a helium gas-flow cryostat (CF935, Oxford Instruments). The temperature was controlled by an ITC (Oxford Instruments, UK). The microwave (MW) frequency was monitored by a frequency counter (5352B, Hewlett Packard).</p><p>High frequency EPR measurements were performed on a home-built D-band (130 GHz) spectrometer equipped with a single mode TE011 cylindrical cavity.37 EPR spectra of the samples were recorded in pulse mode in order to remove the microwave phase distortion due to fast-passage effects. The absorption line shape of the EPR spectra was recorded by monitoring the electron spin echo (ESE) intensity from a two microwave pulse sequence as a function of magnetic field. The duration of a π/2 microwave pulse was 40–60 ns, and typical separation times between microwave pulses were 150–300 ns. All D-band spectra have been pseudo-modulated to facilitate comparison with the respective X- and Q-band spectra.38</p><p>Data processing was done using Xepr (Bruker BioSpin, Rheinstetten) and Matlab™ 6.5 (MathWorks) environment. The magnetic parameters were obtained from theoretical simulation of the EPR and ENDOR spectra. The simulations were performed using the EasySpin software package (version 3.1.7).39 Quadrupole interactions were not included since within the first order of perturbation theory they do not contribute to the EPR spectra. Several simulations were repeated with the program SIMFONIA Version 1.25 (Bruker BioSpin, Rheinstetten, 1996), using second order perturbation theory, and delivered virtually identical parameters. The accuracy in determination of the electronic g-tensor for the set of multifrequency EPR spectra is estimated to be ±0.001.</p><!><p>Starting structures for all geometry optimizations presented in this paper were obtained from the coordinates of the crystal structure of Co(dmgBF2)2 with acetonitrile as axial ligands, provided in the Supplementary Information. This structure is very similar to the one described by Bakac et al. for Co(dmgBF2)2 with methanol as axial ligands. The geometry optimizations were carried out using density functional theory (DFT) with the B3LYP functional40–42 using the 6–31G* basis set, as implemented in PQSMol.43 All structures were first optimized in the gas phase and the resulting structures were re-optimized with solvent effects included by employing the continuum solvation model COSMO with a dielectric constant set to 80.44 Geometry optimizations were also performed with lower dielectric constants and led to nearly identical structures. In addition, a selected set of geometry optimizations were performed using larger basis sets with the program package ORCA.45 These geometry optimizations again led to nearly identical structures and subsequently g-values. After confirming by the absence of imaginary frequencies that the stationary points obtained were minima, the spectroscopic parameters were obtained via single point DFT calculations, performed with the program package ORCA.45 For all calculations of magnetic parameters, the program ORCA with the B3LYP functional was used in combination with the TZVP triple-ζ basis set of Ahlrichs and co-workers on all atoms except cobalt.46,47 For the transition metal 59Co, the Wachters basis set, developed for third row transition metals, was used.48,49 The Wachters basis set has proven quite successful for transition-metal compounds and is well suited for the description of structures, energies, and other properties, see e.g. 50,51 These single point calculations also employed COSMO to account for the dielectric screening of surrounding molecules. The principal g-values were calculated employing the coupled-perturbed Kohn-Sham equations,52 in conjunction with a parameterized one electron spin-orbit operator. The magnetic dipole and the isotropic Fermi contact contributions to the hyperfine coupling were calculated for all atoms, while for 59Co second order spin-orbit hyperfine contributions were also calculated.</p><p>In an additional set of calculations, the effect of the displacement of the central cobalt ion from the plane was systematically investigated. The complex contained a single pyridine molecule as the axial ligand in the "boat towards pyridine" conformation. A restricted geometry optimization was performed with the cobalt ion being pushed/pulled systematically in and out of the plane of the macrocycle using a torsion constraint (Co-N-N-N, Figure 1). The planar nature of the macrocycle portion of the complex was maintained by constraining the nitrogen atoms to remain in plane via constraining the torsion angle of the four nitrogen atoms to the initial, unconstrained value of −0.0158°. The spectroscopic parameters were obtained via single point DFT calculations as described above.</p><!><p>Figure 2 shows selected EPR spectra of the cobaloxime Co(dmgBF2)2 in several different solvents or solvent mixtures under anaerobic conditions, recorded at X-band (9–10 GHz), Q-band (34 GHz), and D-band (130 GHz) microwave frequencies. The multifrequency approach allows us to clearly distinguish between magnetic field-dependent parameters (electronic g-tensor) and magnetic field-independent parameters (hyperfine coupling tensors, A-tensors).</p><p>We will first discuss the D-band spectra since their analysis is clearest. At D-band, the g-tensor anisotropy of Co(dmgBF2)2 is by far larger than the other magnetic interactions, i.e. any hyperfine interaction, and the line width. All our frozen solution D-band EPR spectra show a substantial g-tensor anisotropy (gx–gz > 0.18) and are typical for a low-spin d7 (S = ½) electron configuration of 59Co. This is in complete agreement with previous investigations of several cobaloximes and related compounds.25,30,33,53 Hence, the (pseudomodulated) D-band spectra allow a straightforward determination of the three principal values of the electronic g-tensor, which was found in all cases to be rhombic. In contrast, many previous studies of cobaloxime at lower magnetic fields approximated the g-tensor to be axial. The values obtained from analysis of the D-band spectra were used as constraints in the simulation of X-band and Q-band data. In the following, we will refer to the principal g-values as gx (for gmax), gy (for gmid), and gz (for gmin). In all cases, the smallest principal g-value gz is found to be close to the value of a free electron (ge). This result is expected for a complex where the unpaired electron resides preferentially on the dz2 orbital (see below for a more detailed discussion). The largest g-tensor anisotropy (gx–gz > 0.34) is found in the case of a 1:1 ratio of PDI:Co (see Table 1). The presence of PDI in the toluene solution could result in a mixture of doubly-ligated, singly-ligated, and non-ligated cobaloxime complexes, depending on the free enthalpy of the respective complexes. Since Co(dmgBF2)2 is minimally soluble in neat toluene, at least one PDI molecule must be ligated to a cobaloxime in solution. The spectrum shows no indication of the presence of several paramagnetic species, demonstrating that only cobaloxime with one axial ligand is present. This conclusion is clearly verified by analysis of the ligand superhyperfine structure (14N) at lower magnetic fields (see below). The presence of triphenylphosphine (PPh3) in a solution of CH2Cl2:toluene (1:1) results in a slightly lower gx-value than observed for PDI in toluene. This value is similar to that in methanol and several other neat solvents without strong axial ligand, which have gx -values around 2.26–2.28 (see Table 1 & Figure S2). Interestingly, the gy-value is much smaller than in these other solvents, a property which has been also detected for H-capped cobaloxime.29,33 The unusual ratio of gx/gy has been explained as a PPh3 -induced distortion of the macrocycle.29 The possible PPh3 -induced distortion of the macrocycle is discussed below in more detail. Smaller g-tensor anisotropy is found when pyridine is present in slight stoichiometric excess as compared to the cobaloxime. The smallest g-tensor anisotropy (gx–gz ≈ 0.18) is found in the case of a large pyridine ligand excess with respect to the cobaloxime. The spectrum shows no sign of the presence of multiple species, indicating that two pyridine molecules act as axial ligands. In all D-band spectra, the hyperfine structure of 59Co (I = 7/2) is resolved only at gz orientation (high-field edge) resulting in a splitting into 2I+1= 8 lines. In case of PPh3 as axial ligand, the 59Co hyperfine structure is not resolved even at the high-field edge. This is be due to the substantial 31P hyperfine interaction in addition to the 59Co hyperfine interaction (see Table 1 & discussion below).</p><p>At Q-band frequency the g-tensor anisotropy is resolved in most cases, while at X-band frequency the g-tensor anisotropy is generally unresolved, since the cobalt hyperfine interaction is of comparable magnitude. As in the D-band spectra, the cobalt hyperfine structure is resolved only for gz, while for the Ax- and Ay- components of the 59Co A-tensor only an upper limit can in most cases be estimated by line broadening. The accuracy of determination of Ax and Ay is thus lower than for Az. To determine Ax and Ay more precisely, we performed pulsed ENDOR and pulsed ELDOR-detected NMR experiments at X-band frequency on cobaloxime in methanol. Unfortunately, we were not able to detect signals of 59Co in ELDOR-detected NMR experiments. Only weak and broad ENDOR signals were detected, which we attribute to 59Co (Figure S4). The strongly anisotropic 59Co A-tensor and the non-negligible quadrupolar coupling are most probably responsible for this.</p><p>The chemical nature and number of axial ligands can in many cases not be determined by inspection of the EPR spectra. However, if the coordinating atom of the axial ligand molecule possesses a magnetic nucleus (I ≠ 0), the hyperfine interaction with the unpaired electron of the cobalt ion can deliver the necessary information. For pyridine or PDI with their nitrogen (I(14N) = 1) atom as ligand, a superhyperfine structure can be observed, at least in the high-field part, of the X-band and Q-band EPR spectra. For the samples with a stoichiometric ratio of PDI to Co(dmgBF2)2, each resolved cobalt hyperfine line is split in three lines with equal intensity (1:1:1) and demonstrates the presence of one nitrogen axially coordinating the Co(II) ion. The same superhyperfine structure is observed for 1:1 pyridine:Co, indicating that also here only one nitrogen is axially coordinating the cobalt ion. In contrast, in the case of a large excess of pyridine, the superhyperfine structure observed consists of five lines with an intensity ratio of 1:2:3:2:1. This pattern is indicative for two magnetically equivalent nitrogen atoms that are interacting with the same paramagnetic cobalt ion. A smaller excess of pyridine (<10:1) results in a mixture of doubly- and singly-coordinated complexes. Similar results have been obtained for H-capped cobaloxime.28,35 To exclude the possibility that the strongly interacting nitrogen atoms(s) responsible for the superhyperfine structure are from the equatorial dimethylglyoxime ligands, we performed these measurements also with 15N-labeled (I = ½) pyridine instead (Figure S3). The superhyperfine structure changed with respect to the 14N-labeled pyridine, exhibiting a pattern of 1:1 for singly-ligated cobaloxime and 1:2:1 for doubly-ligated cobaloxime. This proves unequivocally that the resolved superhyperfine structure is due to the axial ligand(s). We did not observe in any case resolved hyperfine structure from the equatorial (macrocycle) nitrogen atoms, indicating that their hyperfine couplings are rather small, at least for Az (assuming that the A-tensor is close to collinear to the g-tensor). This is in agreement with previous studies of H-capped cobaloximes, where also hyperfine coupling constants of the equatorial nitrogen atoms could not be determined from the EPR spectra.29,32 The quantitative appearance of cobaloxime complexes singly-ligated by PDI or pyridine in samples with a 1:1 stoichiometric ratio demonstrates the far higher affinity of the nitrogen-containing pyridine bases and derivatives to the central Co(II) ion as compared to the oxygen-containing polar (e.g. acetone) or polar and protic (e.g. methanol) solvents, which are readily replaced as axial ligands.</p><p>Somewhat different results were obtained for PPh3 as axial ligand. In the case of the bulky ligand PPh3, the same EPR spectrum was found for different stoichiometric ratios of cobalt and PPh3, as long as PPh3: Co(dmgBF2)2 >1 (Figure 2). This is a clear sign of the stability of the five-coordinated complex, if compared to pyridine as the axial ligand, where a tenfold stoichiometric excess results mostly in doubly-ligated cobaloxime complexes. Our observation is in agreement with previous results obtained in a study of H-capped cobaloxime,26,29 where no doubly-ligated cobaloxime was observed when PPh3 or related compounds were present. All these results indicate that the binding of a second PPh3 molecule is not as thermodynamically favorable as it is for pyridine, and that the singly-ligated PPh3: Co(dmgBF2)2 complex is rather stable. A possible reason for this behavior is that the binding of the first bulky ligand molecule results in a displacement of the cobalt ion from the equatorial plane of the macrocycle, possibly including a deformation of the macrocycle. This would make the binding of a second bulky ligand thermodynamically unfavorable and also lower the accessibility of the cobalt ion for catalytic reactions. Since the Co(II)-coordinating part of PDI is identical to pyridine, we consider the explanation that PDI directly causes a much more pronounced cobalt displacement or macrocycle distortion unlikely. A more detailed discussion is given below. In a study of H-capped cobaloxime, Baumgarten et al. observed a larger g-anisotropy when using a toluene:THF mixture instead of methanol.33 For our BF2-capped cobaloxime the increase of g-tensor anisotropy is even larger in neat toluene with PDI. It should be mentioned, that the experimental data do show a large increase of gx as compared to pyridine, but the 59Co hyperfine interaction stays similar in magnitude. The effect of cobalt displacement from the plane is discussed in more detail in the DFT section, where we performed a systematic study of the dependence of magnetic parameters on the Co(II) out-of-plane distance.</p><!><p>For simulation of the EPR spectra it has been assumed that the g-tensor principal axes and the principal axes of the 59Co A-tensor are collinear. The ligand superhyperfine principal axes (14N, 31P) were assumed to be collinear to the g-tensor principal axes as well. A rotation of the A-tensor with respect to the g-tensor did not lead to an improved fit. The assumption that the g-tensor principal axes and the 59Co A-tensor principal axes are close to collinear is also justified on basis of the theoretical DFT calculations (see below). A similar observation concerning the collinearity of g-tensor and 59Co A-tensor has also been made for H-capped cobaloxime.27,33</p><p>Figure 3a shows a plot of the gy-values vs. gx values. Ignoring the large, bulky ligands PDI and PPh3, there exists a linear relation between gy and gx. A linear interpolation to smaller gx -values intersects the point where both gx and gy are near free electron g-value, ge (gx, gy ~ ge). This linear dependence can be justified on the basis of the simplified Stone formula as a first approximation for the prediction of g-tensor anisotropy due to spin-orbit coupling: (Eq. 1)Δg∼(ρ∗ζ)/ΔE where Δg is the deviation from ge; ζ is the spin-orbit coupling constant for the respective atom; ρ is the unpaired spin density in the particular atomic orbital containing the unpaired electron; ΔE is the energy difference between the orbital containing the unpaired electron and lower or higher laying orbitals.54 According to Stone's theory of g-values, negative deviation of g-values from the free electron value is due to spin-orbit coupling with empty higher laying orbitals while spin-orbit coupling with lower laying occupied orbitals leads to positive deviations. The difference between Δgx and Δgy is related to the symmetry properties of the particular atomic orbitals. Total deviation of the g-values is a sum of the deviations induced by each atom in the molecule. As Co(II) in 3d7 configuration has the biggest spin-orbit coupling constant, 515 cm•1, as compared to other atoms in the molecule,55 the analysis can be simplified by taking into account only Co orbitals. Thus, if the electronic structure of the cobaloxime does not change substantially upon change of the axial ligand, i.e. the molecular orbitals are in the first order approximation the same combination of the atomic orbitals, the plot of gy vs. gx should demonstrate linear dependence. As it was shown previously,53,56,57 the complete absence of axial ligands results in very large g-tensor anisotropy and very large 59Co hyperfine coupling constants, much larger than observed in our study. We thus conclude that we always have at least a weak axial ligation of the Co(II) ion. According to Eq. 1 the same linear behavior should hold for the plot of Az vs. gx or gy, as Az is proportional to the spin density of the unpaired electron within the first order of approximation. Indeed, the plot of Az vs. gx (Figure 3b) shows also linear dependence. This is important, since the slopes of the lines on Figure 3 are sensitive characteristics of the electronic structure of the compound and might be a reference point for testing the validity of DFT calculations along with g- and A-tensors.</p><p>A large g-tensor anisotropy for Co(dmgBF2)2 is observed for "weakly" coordinating ligands like alcohols, acetone or DMF where an oxygen atom is involved in coordination of the cobalt ion. Ligands termed "weakly coordinating or labile" can be easy replaced by small amounts of another type of ligands referred to as "strongly coordinating". Two general types of "strong" Co(II) coordination can be distinguished. In the six-coordinated L2Co(dmgBF2)2, complex two strong axial ligands like pyridine are bound to the central cobalt ion. In the five-coordinated LCo(dmgBF2)2 only one strong axial ligand is present. If in the case of one strong axial ligand like pyridine a second, weak ligand like methanol or acetone binds at the open axial coordination site can not be verified on basis of our experimental results.</p><p>In both cases (singly- or doubly-ligated by pyridine) the g-tensor anisotropy is significantly smaller than for the weakly coordinating solvents like methanol. The doubly-ligated pyridine:cobaloxime complex exhibits even smaller g-tensor anisotropy than the singly-ligated cobaloxime. Since cobaloxime in a solvent of "weak" ligand molecules has most probably two axial ligands, the g-tensor anisotropy is unlikely to be predominantly caused by displacement of the Co(II) ion from the equatorial plane. More likely, the interaction strength with ligand molecules leads to reduced g-tensor anisotropy. The relative small hyperfine interaction with the pyridine nitrogen atoms contradicts the assumption of a substantial shift of unpaired spin density from Co(II) to the ligand molecules, which is supported by the DFT calculations (see below). Concomitantly to the reduced g-tensor anisotropy a reduction of the 59Co Az-value is found, while Ax and Ay both increase in magnitude. The main effect of axial ligand field could be a modification of ΔE term in Eq. 1 and/or a reduction of spin density in the dz2 orbital, which seems unlikely. A direct determination of relative orbital energies is complicated, since multiple parameters influence the relation between magnetic properties and orbital energies.57</p><p>So far, the two bulky ligands PDI and PPh3, which do not fit into the rather simple model described above, have not been discussed. Unfortunately, due to the hydrophobic nature of there ligands, their complexes with cobaloxime can exist only in non-polar solvents such as toluene or CH2Cl2/toluene mixture. On the other hand the hydrophilic nature of cobaloxime does not allow existence of pyridine:cobaloxime complexes in strongly non-polar solvents, thus prevents from the direct comparison of magnetic resonance parameters with bulky hydrophobic ligands in the same solvents. These two bulky ligands in combination with the non-polar solvents also exhibited different thermodynamic properties, showing a lower affinity for a second molecule to ligate the Co(II) ion. This matter will be further addressed in the DFT section.</p><!><p>It is well known that many Co(II) complexes including cobaloximes form adducts with molecular oxygen.24,25,58–60 The BF2-capped cobaloxime Co(dmgBF2)2 studied here is expected to be quite resistant against the oxidation by molecular oxygen and subsequent degradation of the cobalt macrocycle.3 This makes it the ideal cobaloxime for studying the interaction of the complex with molecular oxygen, as compared to previous studies of the H-capped cobaloxime. A certain resistance against oxidation and degradation by molecular oxygen is also desirable for any catalyst involved in hydrogen production.</p><p>Bakac and co-workers already observed the reversible binding of molecular oxygen by Co(dmgBF2)2 in a variety of different solvents, including water, methanol, acetonitrile, and acetone.25 Schrauzer & Lee studied H-capped cobaloxime, using different bases as axial ligands.24 The EPR spectra changed dramatically upon oxygenation, indicating a strong change of the electronic structure of the complex. "Unspecific" oxidation of the macrocycle could be excluded, since the process was reversible and could be repeated multiple times, if water was carefully removed from the solution.24 Lubitz and co-workers studied oxygen adducts of H-capped cobaloxime in a zeolite matrix.61 However, the X-band spectra presented previously exhibited rather low g-tensor resolution and did not allow a unique determination via magnetic parameters. Here we used the multifrequency approach described above to study the oxygenation of Co(dmgBF2)2. We choose a solvent mixture of toluene and CH2Cl2, which is known to solubilize Co(dmgBF2)2 well without exerting strong interactions on the cobalt center, and also generates a good glass at cryogenic temperatures. These solvents have been previously used to study the oxygenation of cobalt complexes.58–60 A small amount of PPh3 was added to act as axial ligand for multiple reasons. In general, cobalt complexes without strong axial ligand are poor oxygen binders, while the presence of one strong axial ligand allows binding of molecular oxygen even at ambient pressures.58 As mentioned above and observed previously,26,29 the presence of PPh3 in slight excess in the solution results exclusively in a singly-ligated cobaloxime complex, displacing weakly interacting solvent molecules from one binding site and leaving the other coordination site open for another small ligand like molecular oxygen. For a 2:1 ratio of PPh3:cobaloxime investigated here, we observed quantitative binding of molecular oxygen as shown by the almost complete disappearance of the EPR signal typical for anaerobic cobaloxime samples and appearance of an intense signal from the superoxide. This species was efficiently generated and rather stable in the presence of PPh3, which is similar to the findings of Schrauzer & Lee for H-capped cobaloxime,24 while the use of other axial ligands resulted in much lower yields of this paramagnetic species. Subsequent bubbling of the solution containing oxygenated cobaloxime with nitrogen gas largely restored the "original" signal as shown in Figure 2. These findings exclude the possibility that the cobaloxime has undergone unspecific oxidation of the macrocycle, which would instead show irreversible spectral changes. Prolonged incubation of the solution for days under aerobic conditions and ambient temperature lead to a steady decrease of the EPR signal, indicating slow degradation of the complex and built-up of possible diamagnetic di-nuclear Co-O2-Co complexes.24,58 The presence of 31P with a nuclear spin of I = ½ as the ligating atom of PPh3 also provides the advantage to monitor changes in the spin density distribution via 31P hyperfine interaction with the unpaired electron(s) of the cobaloxime, while a possible shift of spin density to or from an axial ligand coordinating via an oxygen atom, i.e. acetone, methanol, or THF - can not easily be observed.</p><p>The EPR spectra of oxygen-exposed Co(dmgBF2)2 recorded at the three different microwave frequencies are depicted in Figure 4. They are well resolved and could be simulated with the same set of parameters, indicating the presence of only one paramagnetic species. The D-band EPR spectrum (Figure 4a) is rather narrow (<0.15 T) as compared to the spectra of the Co(II) cobaloxime under anaerobic conditions (Figure 2). It exhibits the typical shape for a S = ½ system with a rhombic g-tensor. In the following, we will keep the nomenclature introduced above to name the principal g-values as gx (for gmax), gy (for gmid), and gz (for gmin). Note, that different nomenclatures have been used previously.60–62 In contrast to the spectra of the Co(II) cobaloxime recorded under anaerobic conditions, resolved hyperfine structure is not visible. The absence of the prominent cobalt hyperfine structure at the high-field part of the spectrum is a further indication of the significant electronic changes experienced by the cobalt ion upon binding of oxygen. The principal values of the g-tensor can be directly determined from the spectrum, as 2.0390, 2.0014, and 1.9879. The g-values determined by Bakac et al. for acetone ligated cobaloxime and by Schrauzer & Lee do not agree well with our findings.24,25 This could be due to the difference in one of the axial ligand, however, the poorly resolved X-band EPR spectra reported previously might have resulted in an erroneous determination of the g-tensor. The pronounced reduction of the g-tensor anisotropy is typical for Co(II) low spin complexes singly-ligated by molecular oxygen.58–60 At Q-band frequencies (Figure 4b), the spectrum is less than 40 mT wide, and the three principal g-values are still resolved, confirming the absence of any large hyperfine interaction. However, a clear broadening of the gy component can be observed, which can be attributed to unresolved hyperfine interaction.</p><p>At X-band frequency, the whole spectrum is about 20 mT wide. Its appearance resembles many X-band spectra of oxygenated forms of related cobalt complexes and cobaloximes, see e.g.60,63,64 A well-resolved hyperfine structure is visible around gy, although the hyperfine structure around gx is not resolved. This is in contrast to vitamin B12 complexes where the hyperfine structure at low field is typically well resolved.61,64,65 The number of lines could only be reproduced in the spectral simulation by the assumption of significant hyperfine interaction of one phosphorus nucleus and one cobalt nucleus, both being very similar in magnitude, around 35 MHz. The presence of substantial 31P hyperfine interaction was confirmed by additional experiments. The use of PDI instead of PPh3 as axial ligand resulted in a decrease of hyperfine lines from 9 to 8 (see Figure S7). The absence of a visible hyperfine structure at gx allows no reliable estimation of Ax neither for 59Co nor 31P. The EPR signals arising from gz severely overlap with the hyperfine structure of gy. The Ax and Az values used for the simulation thus bear a rather large error. The much lower magnitude of the 31P hyperfine interaction as compared to the deoxygenated cobaloxime is in line with the observed dramatically reduced 59Co A-values due to the lower spin density on the coordinated cobalt ion. In this type of complex, the 59Co A-tensor and 31P A-tensor are expected not be collinear anymore with the g-tensor.60 Lifting the assumption of collinear axes systems and using similar angles as in studies of related cobalt oxygen adducts60,63,65 lead only to minor changes in the spectral simulations. Schrauzer & Lee estimated for the H-capped cobaloxime a reduction of unpaired spin density at the cobalt by at least a factor of 5 on the basis of the 59Co isotropic hyperfine coupling. Even larger reductions have been estimated for H-capped cobaloxime 61 and related compounds.63,65</p><p>Cobalt complexes can form several different complexes with molecular oxygen, see e.g. 58. One complex which may be formed by cobaloximes dissolved in organic solvents in the presence of molecular oxygen are dinuclear μ-peroxides, as shown previously for H-capped cobaloxime.24 This type of dinuclear complex is diamagnetic and thus not detectable by EPR spectroscopy. A paramagnetic μ-superoxo radical cation L-Co(dmg2H2)-O2-Co(dmg2H2)-L+ with a characteristic 15-line signal in liquid solution was also observed. This signal is due to the interaction of two magnetically equivalent cobalt nuclei with the unpaired electron symmetrically delocalized over the Co-O2-Co moiety.24 We did not observe the dinuclear μ-superoxo radical under our experimental conditions. Mononuclear superoxide complexes can be described as L-Co(dmg2H2)-O2•and characterized by a hyperfine splitting multiplicity of eight, indicative of the presence of only one cobalt ion in the complex. This complex was obviously observed in our study. The frozen solution spectra reported here show no indication of S > ½. As Co(II) has one unpaired electron in the non-oxygenated complex, and molecular oxygen is a ground state triplet (S = 1) due to two unpaired electrons, the resultant spin is S = ½. The coordination environment in Co(II) Schiff base compounds resembles that of cobaloxime, and a qualitative interpretation of the EPR spectra of these compounds binding molecular oxygen has been given by Hoffman et al.63 In that work, the authors concluded that the cobalt ion is essentially oxidized to Co(III) and the molecular oxygen is reduced to the superoxide ion, O2− and estimated the transfer of spin density from cobalt to oxygen to be roughly 90%. Schrauzer & Lee determined the 59Co isotropic hyperfine coupling for the H-capped cobaloxime and estimated a reduction of unpaired spin density at the cobalt by a factor of 5–10. Since we observe similar hyperfine coupling for the difluoroboryl cobaloxime, we may follow their line of reasoning. Additional experimental evidence that related cobalt complexes form under presence of molecular oxygen Co(III) superoxide complexes, provides further support of an assignment of our complex as being formally LCo(III)O2• superoxide.58,59,61,65–67</p><p>We thus conclude that there is almost complete transfer of the cobaloxime's unpaired electron spin density to the oxygen. The g-tensor anisotropy is then mostly determined by the spin density on the oxygen atoms, and the principal component of the g-tensor with the largest deviation from ge will be in direction of the inter-oxygen axis. The cobalt hyperfine coupling instead is determined by the 3d orbital of the cobalt ion, and this presents a pronounced difference to the non-oxygenated complex. From our experimental data no direct conclusion is possible regarding the presence of the three different conformers (chair, boat-PPh3, boat-O2). A comparison with the results from the DFT calculation let us favor the boat-PPh3 conformation (see discussion below).</p><!><p>The calculated g-values, 59Co A-values, and superfine coupling constants are presented in Table 2. To the best of our knowledge, these are the first published EPR parameter calculations for singly- and doubly-ligated Co(II) cobaloxime complexes. Previous semi-empirical calculations focused on the effect of the cobalt displacement on the energetics of the compound29 and four-coordinate Co(II) complexes without axial ligands were reported.68 Recent DFT studies on cobaloxime did not report any magnetic parameters.69,70 The structure of the cobaloxime complex allows multiple conformations of each cobaloxime-ligand(s) complex. As shown in Figure 5, using the common cyclohexane notation, the conformations can be labeled "chair" and "boat" with the possibility of further specifying the boat as towards or away from the attached ligand. If only one axial ligand molecule was included in the calculation, the expressions "boat towards ligand" or "boat away" are used. The "boat" and "chair" conformations refer to the orientations of the BF2 caps with respect to the macrocycle plane. Unlike the four-coordinate Co(II) complexes investigated by Zbiri,68 our calculated g-values do not change significantly when other hybrid functionals are used. Before discussing in detail the results of the calculations, some critical issues regarding the accuracy and reliability of the calculations are addressed briefly.</p><p>One problem of DFT calculations on large open shell transition metal complexes is the high accuracy required to distinguish between conformations which are quite close in energy as compared to the overall energy of the complex. This is partially the result of inherent approximations made in DFT calculations.71,72 At least part of the discrepancy between actual and calculated energies is due to insufficient inclusion of the interactions with solvent. Increasing the number of solvent molecules in the calculation and thus forming an extended solvent shell around the complex may lead to an improved prediction of conformational energies and also to improved magnetic parameters as compared to pure gas phase calculations.73–75 However, the explicit inclusion of a large number of solvent molecules is computationally demanding, thus such time consuming calculations were not performed. Instead only the direct ligands to the central cobalt atom were explicitly included and the residual surrounding solvent molecules were implicitly included using the dielectric screening model COSMO.44 This neglect is considered to be most problematic for those solvent molecules where significant direct interactions with the cobaloxime are expected. For example, it has been shown that hydrogen bonding is in many cases essential to geometries and electronic structures.76–78 However, since in our study the experimental spectra of cobaloxime in methanol:pyridine and acetone:pyridine mixtures were almost identical (see Table 1 for the magnetic parameters), possible hydrogen bonding to the cobaloxime by solvent molecules is not considered critical for the electronic structure of the cobalt complex and thus no calculations with solvent molecules ligated to the fluorine atoms were performed.</p><p>If conformers exist with comparable energies (same order of magnitude as kT), the complexes in frozen solution studied by EPR spectroscopy may also consist of a mixture of conformers trapped during the freezing process. In addition, our calculations determine the energetic minimum but do not include entropic contributions. Hence, instead of choosing only the lowest energy conformation, we present the magnetic parameters calculated for the various conformations. It is expected that the comparison with the experimentally determined magnetic parameters allows us to more reliably exclude conformers than a discrimination based solely on energy. To validate the quality of the geometry optimization, we repeated selected calculations with different functionals, basis sets, and dielectric constants. Use of a sufficiently large dielectric constant for the COSMO screening model was found to be needed to obtain reasonable structures, i.e. to avoid structures with strong distortion of the macrocycle. Otherwise, all calculations resulted in nearly identical geometries and energetic orderings within the conformations, which suggests that the obtained structures are reliable. Problems with precisely calculating the magnetic parameters for our specific transition metal complexes are discussed below in relation to the available experimental data.</p><!><p>As shown in Table 2, the position of the cobalt with respect to the plane of the cobaloxime macrocycle varies greatly depending on number and nature of the axial ligand(s). Most notably, the di-water and di-methanol complexes result in the cobalt remaining nearly planar while the nitrogen and phosphorus containing ligands pull the cobalt out of plane. Additionally, singly-ligated complexes pull the cobalt out of the plane to a greater degree than their doubly-ligated counterparts and the boat conformations show greater displacements than the chair. To evaluate systematically the effect of this cobalt out-of-plane movement on the electronic g-tensor and the cobalt hyperfine interaction independent of the chemical nature of the ligand, we investigated a model system composed of just the cobaloxime with one pyridine molecule in the "boat towards pyridine" conformation. The energetic minimum for this system has the Co(II) ion located 0.32787 Å below the plane of the macrocycle pulled towards the pyridine. In these calculations, the cobalt atom was moved in and out of the macrocycle nitrogen plane fixed there while the rest of the molecule was allowed to relax. Further details of the calculation are given in the Experimental Procedures section. The effects on the g-values and 59Co hyperfine coupling constants Ax, Ay, and Az are presented in Figure 6 and listed in Table S1. In the case without forced displacement of the Co(II) ion, the principal g-values, 59Co A-values, and 14N A-values resemble those obtained experimentally. The displacement of the cobalt ion leads to a systematic change in both the principal g-values and the 59Co A-values while the 14N A-values (not shown) do not appreciably vary from the non-displaced case. Notably, the 59Co A-values exhibit an almost linear relationship with cobalt out-of-plane distance. Ax, Ay, and Az values all decrease as the cobalt moves further away from the plane. Moving the cobalt into the plane, beyond the energetic minimum position, has a weak influence on the g-values. Despite the correlation shown in the test calculation, the chemical nature of the ligand and the chair/boat conformation seems to obscure this effect on the values of the principal g-values and 59Co A-values. Thus it is not possible to directly predict the cobalt position from the g-values or the 59Co A-values. For example, consider the singly-ligated water case (Table 2). The two boat conformations differ in the cobalt distance by 0.13 Å, and the 59Co Az-values decrease as expected. The chair conformation has an even larger distortion and a concurrent larger drop in the 59Co Az-value. However, the g-values do not increase for the chair conformation as might have been expected from the model. In general, the 59Co Az-values are most sensitive to the cobalt position while the g-values are a poor indicator of cobalt position. This is in agreement with prior work on d1 five-coordinate systems that showed only a weak dependence of the g-values on geometry.79 In terms of directly predicting the experimental g-tensor, the calculations correctly reproduce the trend of decreasing g-values (gx, gy) as the complex goes from singly-ligated by pyridine to doubly-ligated. However, the calculations underestimate the gx value for all ligands except the PPh3 oxygen adduct. This is not a problem specific to our study. The systematic underestimation of g-values of transition metal complexes by DFT is common,72 nevertheless, they can be used to indicate a trend within the series of calculations.</p><p>Turning to the 59Co A-values, qualitative rather than quantitative agreement is present. The three contributions to the (transition metal) hyperfine interaction (Fermi contact, spin-dipolar, and spin-orbit coupling) are known to be difficult to calculate with accuracy.80,81 In agreement with experiment, the Az-values of Co(dmgBF2)2 in methanol and water are greater than the corresponding values for pyridine.</p><p>The 14N A-values of the equatorial (dimethylglyoxime) nitrogen atoms were also calculated. They are found to be small, below 8 MHz in all (energetically reasonable) cases and in majority of cases well below that (see Table S2). The hyperfine coupling constants of the equatorial nitrogen atoms were not resolved in the EPR spectra. HYSCORE data of Co(dmgBF2)2 in methanol (see Figure S6) allows us to estimate these 14N hyperfine coupling constants as below 4 MHz which is in a good agreement with the principle values obtained by the DFT calculation. The 14N and 31P A-values of the axial ligands which could be determined from the EPR spectra were nicely reproduced by the DFT calculations, see Tables 1 & 2. The general better agreement of superhyperfine interactions than hyperfine interactions of the central metal ion is typical for DFT calculations.82</p><p>The binding energies for the six-coordinate complexes (two axial ligands) predicted by the calculations are in general agreement with experiments (pyridine > methanol; PPh3 > toluene, CH2Cl2). In some of these cases, the singly-ligated conformation appears to be more energetically favorable than the doubly-ligated one. As noted previously, some ligands like PPh3 seem to result in a stabilization of the singly-ligated Co(II) complex.29 Prior authors attributed this observation to a distortion of the cobaloxime structure where the Co(II) is pulled out of the plane towards the ligand and the equatorial glyoxime ligands are folded away from the axial ligand. By comparing binding energies, it is clear that while water and methanol are more stable as doubly-ligated complexes, this is not true for other complexes. The pyridine complex is nearly isoenergetic as a 2:1 and a 1:1 complex while the calculations predict that the PPh3 and PDI complexes prefer the mono-substituted complex. These results are in good agreement with the experimental results. For PDI and PPh3 we observed only singly-ligated complexes, while already a moderate excess of pyridine leads to a significant presence of doubly-ligated complexes. However, the optimized structures of complexes singly-ligated by pyridine or PDI are very similar and yield similar magnetic parameters, while the latter is not found experimentally. From calculations it is unclear whether a solvent effect is present (methanol or acetone for pyridine, and toluene or CH2Cl2:toluene for PDI and PPh3). It seems likely, that for cobaloxime singly-ligated by pyridine in the methanol solution, a methanol molecule acts as a second but "weak" ligand.</p><p>Computationally, the binding of a single PDI molecule to the cobalt induces out-of-plane displacements almost identical to that of the binding of pyridine. Experimentally, the single PDI molecule binding results in the largest gx-value. However, in the DFT calculation this large g-tensor anisotropy is not reproduced. In fact, not surprisingly, the calculated magnetic parameters of cobaloxime singly-ligated by one PDI molecule resemble relatively closely those of cobaloxime singly-ligated by pyridine (Table 2). This might be explained by our tentative assumption that the quite non-polar solvents (toluene or CH2Cl2:toluene mixture) have an additional effect on the geometry and electronic structure of the cobaloxime complex (see above). For PDI in neat toluene, the second axial ligand may either be absent or hardly coordinating, in contrast to methanol or acetone as solvents. This might be especially prominent since both PDI and PPh3 are the most bulky ligands studied here. A highly simplified picture would be a kind of "folding" of the macrocycle. However, computationally, optimizing the singly-ligated PDI complex using a low dielectric constant but not explicitly solvent molecules does not appreciably change the geometry or the g-values.</p><p>Experimentally, the single PPh3 ligand results in the second largest gx component determined, but in a rather small gy value. While the geometry of the singly-ligated PPh3 complex after energetic minimization is quite distorted; the calculations still under-estimate gx but match gy quite well.</p><p>In addition to influencing the location of the cobalt atom, the ligands also seem to influence the stability of the boat/chair conformational preference of the cobaloxime itself. In particular, water and methanol are most stable in the chair conformation shown in Figure 5 while pyridine is most stable in the boat conformation (Table 2). The "chair" structure obtained for two methanol ligand molecules is in good agreement with the crystal structure determined by Bakac et al.25 Our calculated structure has a cobalt-methanol oxygen distance of 2.246 Å and the crystal structure reports a distance of 2.264(4) Å. Within this work the crystal structure of Co(dmgBF2)2 with two acetonitrile molecules as ligands was determined (see Supplementary Information). It exhibits a similar distance with 2.25 Å and shows also "chair" conformation. Both the calculated and the two crystal structures have in-plane cobalt atoms. While no di-pyridine crystal structure exists, support for the boat structure is evident both in the energetics and in the g-and A-values.. The calculations also predict that PDI would exhibit the boat conformation. For the PPh3:Co (1:1) complex, the energetics suggest that the "boat towards" conformers is less likely, but all three conformers have so similar magnetic parameters that we can not distinguish them with reference to the experimental data.</p><p>The DFT calculations provide not only the magnitude of the magnetic interactions, but also the orientation of the g-tensor principal axes and those of the A-tensors (59Co, 14N, 31P) within the molecular frame. Figure 1 presents the molecular axes system of cobaloxime, with the axes labeled accordingly. For all cases without molecular oxygen bound to the Co(II) ion we observed that the lowest g-value, gz, is perpendicular to the equatorial plane, i.e. pointing along the dz2 orbital. The middle g-value, gy, is always found to be collinear to the boron-boron axis and the highest g-value, gx, is perpendicular to the two other. The 59Co A-tensor principal axes are also to a good approximation collinear to the molecular axes system. The deviation of the electronic g-tensor and 59Co A-tensor from the molecular axes system was usually less than 5°. Only the binding of PPh3 which caused a slight distortion of the macrocycle, resulted in deviations up to 10°. All ligands (14N, 31P) A-tensors were collinear to the molecular axes system with deviations of less than 5°.</p><!><p>Finally, DFT calculations on the O2:Co:PPh3 complex were done for the three conformers: chair, boat towards PPh3, and boat towards O2. All three conformers are rather close in energy (Table 2). Geometry optimizations were performed several times with different starting structures which differed regarding the orientation of the oxygen molecule with respect to the cobaloxime. All calculations resulted in the O-O-Co angle remaining near 120°. For the conformers chair and boat towards PPh3 the minimum is very shallow (less than 2 kcal/mol as the O2 is rotated over the macrocycle). In contrast, the boat towards O2 structure results in structures with unbound oxygen if the initial O-O-Co-N dihedral angle deviates from the one found in the energetic minimum. For all three, the magnetic parameters exhibit no pronounced differences and show reasonable agreement with the experiment, regarding the electronic g-tensor, the 59Co A-tensor and the 31P A-tensor. However, the g-tensor of the "boat towards PPh3" conformer shows the best agreement with the experiment and we discuss this boat conformer in the following. The analysis of the other conformers is similar. In the "boat towards PPh3" conformer, the distance between cobalt and the bound phosphorous atom is 2.57 Å, and hence hardly changed as compared to the corresponding 2.63 Å found in the de-oxygenated complex. The distance between the two oxygen atoms is 1.35 Å. The calculated distance between the two oxygen atoms of unbound molecular oxygen is 1.22 Å, while the calculated distance in O2− is 1.35 Å (both distances from DFT calculations with 6–31G* basis set). The virtually identical distance between the two oxygen atoms in the oxygen adduct and the O2− anion is a strong hint that a substantial shift of the electron density from cobalt towards the oxygen has occurred and thus supports the formal description of the oxygen adduct as a LCo(III)O2− superoxide. The Co-O-O bond angle is 119°, which is in the typical range found for cobalt superoxide complexes.58 Also the spin density (shown in Figure 7) exhibits the expected shift from the Co(II) ion to the O2 molecule. This calculation thus clearly refutes for the cobaloxime the spin-pairing model proposed by Tovrog et al. 83,84 and supports the view of the oxygen-adduct as a LCo(III)O2• superoxide. The principal axes of the g-tensor and the A-tensors (59Co and 31P) deviate with respect to each other and the molecular axes system (see Table S3 in the Supporting Information). This is typical for O2:Co:L complexes.60,62,63,65</p><!><p>In conclusion, we have presented a comprehensive study of the molecular H2 catalyst Co(dmgBF2)2 by multifrequency EPR spectroscopy at X-band (9 GHz), Q-band (34 GHz), and D-band (130 GHz), combined with DFT calculations. Cobaloxime was investigated in a number of solvents with different polarity, proticity and stochiometric amounts of potential ligands to the cobalt ion. This approach allows a clear distinction between labile and strongly coordinating types of axial ligands to the Co(II) macrocycle, as well as two types of strong Co(II) coordination: singly-ligated LCo(dmgBF2)2 and doubly-ligated L2Co(dmgBF2)2. Weakly coordinating ligands like methanol result in larger g-tensor anisotropy than strongly coordinating ligands like pyridine. The presence of two strongly coordinating axial ligands leads to the smallest g-tensor anisotropy. Strongly binding N-donating solvents displace the weaker O-donating solvents as axial ligands to the Co(II) ion in the cobaloxime. Therefore, O-donating solvent environments can more readily facilitate proton binding at the cobalt center to initiate H-H bond formation, rather than forming a tight and rather stable complex with solvent molecules and hindering proton binding. The influence of molecular oxygen and formation of Co(III) superoxide radicals LCo(dmgBF2)2O2• was also studied. The proportional decrease in both ligand and 59Co hyperfine coupling constants for the reversible LCo(dmgBF2)2 + O2 ⇆ LCo(dmgBF2)2O2• transformation is consistent with a substantial shift of electron spin density to oxygen. Finally, the results obtained experimentally are compared to a comprehensive set of DFT calculations on Co(dmgBF2)2 model systems with various axial ligands, and on oxygen adducts in the presence of PPh3. Comparison with experimental values for the "key" magnetic parameters like g-tensor and 59Co A-tensor allows the determination of the identity of the ligands and conformation of the axially ligated Co(dmgBF2)2 complexes.</p><p>In this study, we chose this BF2-capped cobaloxime in particular as it stands out among earth-abundant H2 catalysts as extremely efficient and quite stable under relevant conditions for catalysis. Our observations provide important insight on electro- and photocatalytic efficiencies, and explain for example the results of Fihri et al. on supramolecular cobaloxime photocatalysis, who found a nearly six-fold enhancement in H2 turnovers in acetone solution as compared to acetonitrile.4</p><p>Future work will be directed toward understanding the catalytical activity of supramolecular assemblies of cobaloxime as catalyst and photosensitizers, including conventional artificial photosensitizers as well as natural photosentizers (i.e. photosynethetic proteins).85 We anticipate that the combination of multifrequency EPR structural analysis and high-level DFT calculations will provide a predictive role in the development of new photocatalysts.</p>

PubMed Author Manuscript

Phase I dose escalation study of KOS-1584, a novel epothilone, in\npatients with advanced solid tumors

Purpose First-in-man study of KOS-1584, a second generation epothilone. Methods Patients with advanced solid malignancies received KOS-1584 every 3 weeks until disease progression. Using a modified Fibonacci dose escalation scheme, one patient was enrolled at each dose level until the first instance of grade 2 toxicity. Thereafter, a standard 3 + 3 design was utilized. Results Sixty-six patients in 14 cohorts were dosed from 0.8 to 48 mg/m2. Diarrhea, arthralgias, and encephalopathy were dose-limiting toxicities (DLTs) at doses \xe2\x89\xa536 mg/m2. At the recommended phase II dose (RP2D), the most common adverse effects were peripheral neuropathy (low grade), fatigue, arthralgias/myalgias, and diarrhea (31, 6%). The incidence of neutropenia was low. The overall clearance, volume of distribution, and half-life of KOS-1584 were 11 \xc2\xb1 6.17 L/h/m2, 327 \xc2\xb1 161 L/m2, and 21.9 \xc2\xb1 8.75 h, respectively. The half-life for the seco-metabolite (KOS-1891) was 29.6 \xc2\xb1 13.8 h. KOS-1584 exhibited linear pharmacokinetics. A dose-dependent increase in microtubulin bundle formation was observed at doses \xe2\x89\xa527 mg/m2. Two patients achieved partial responses and 24 patients had stable disease (SD). Conclusions The RP2D of KOS-1584 is 36 mg/m2. The lack of severe neurologic toxicity, diarrhea, neutropenia, or hypersensitivity reactions; favorable pharmacokinetic profile; and early evidence of activity support further evaluation.

phase_i_dose_escalation_study_of_kos-1584,_a_novel_epothilone,_in\npatients_with_advanced_solid_tumo

Introduction<!>Eligibility criteria<!>Dosage, dose escalation, and drug administration<!>Safety and efficacy assessments<!>Pharmacokinetics<!>Tubulin polymerization in PBMCs<!>Statistical analysis<!>General<!>Safety and tolerability<!>Pharmacokinetics<!>Pharmacodynamics<!>Anti-tumor activity<!>Discussion

<p>Microtubules play a pivotal role in cell division and are validated targets for cytotoxic agents. Vinca alkaloids and taxanes interfere with microtubule function and induce apoptotic cell death [1–3]. Despite efficacy in a variety of malignancies, these agents have limitations of multi-drug resistance (MDR), neutropenia, neurotoxicity, and hypersensitivity reactions [3–6]. The development of second generation agents has aimed to overcome these problems.</p><p>Epothilones, derived from the myxobacterium Sorangium cellulosum, are a novel class of microtubule interactive agents. Importantly, they do not exhibit cross-resistance to taxanes (due to poor susceptibility to p-glycoprotein (P-gp)-mediated drug efflux and high affinity to various β tubulin isoforms) and have a more favorable side effect profile [4, 7, 8]. Epothilones in recent development include ixabepilone, patupilone, sagopilone, and KOS-862 [9–14].</p><p>Figure 1 depicts the chemical structures of epothilone derivatives in recent development. Small chemical variations result in important differences in terms of susceptibility to P-gp, need for cremophor-based formulation, pharmacokinetics, toxicity profile, and in vivo anti-tumor activity [3]. The epoxide group at carbon 12–13 seen with ixabepilone and patupilone is highly reactive and contributes to epothilone toxicity. Desoxyepothilone B (epothilone D, KOS-862) was designed with the epoxide reduced to a double bond in an effort to give this compound less toxicity. KOS-862 was shown in vivo to have greater activity compared with ixabepilone and patupilone [11, 15]. However, phase II trials of KOS-862 have shown significant neurotoxicity, limiting further development [16–18].</p><p>KOS-1584 [(E)-9,10-didehydroepothilone D] is a second generation epothilone specifically designed to have a longer elimination half-life, larger volume of distribution, less toxicity, more potency, and higher solubility (negating cremophor) compared with KOS-862 [3, 18–20]. Modifications included a reduced double bond at carbon 9–10 and flattening the 16-member ring [3]. In P-gp-overexpressing cell lines, KOS-1584 was more potent than paclitaxel. In paclitaxel-resistant xenografts, anti-tumor activity has been demonstrated. KOS-1584 is highly bound to plasma proteins (97%). The primary seco-metabolite (KOS-1891) was inactive in cytotoxicity assays up to 10 μM [18].</p><p>We performed this first-in-human, phase I dose escalation study of KOS-1584 in patients with advanced cancers. Primary objectives included determination of safety, DLTs, maximum tolerated dose (MTD), and recommended phase II dose (RP2D). Secondary objectives included evaluation of pharmacokinetics, pharmacodynamics, and anti-tumor activity.</p><!><p>Eligible patients had advanced or metastatic solid tumors measurable by Response Evaluation Criteria in Solid Tumors (RECIST, Version 1.0) [21]; Eastern Cooperative Oncology Group performance status ≤1; age ≥18 years; life expectancy ≥3 months; and adequate bone marrow (ANC ≥1.5 × 109/L, hemoglobin ≥8.5 g/dL, platelets ≥75 × 109/L), renal (Cr <1.5 ULN) and hepatic (AST ≤2.5 ULN, total bilirubin ≤1.5 ULN) function.</p><p>Exclusion criteria included uncontrolled or hemorrhagic diarrhea; active peptic ulcer disease; grade ≥2 neurological symptoms; hypersensitivity reaction to hydroxypropyl-β-cyclodextrin, ethanol, or propylene glycol; intracranial metastasis; pregnancy/lactation; clinically significant cardiac disease; dementia or altered mental status; and other conditions interfering with study participation.</p><p>All patients gave written informed consent. Approval was obtained from the institutional review boards at both institutions.</p><!><p>The starting dose was 0.8 mg/m2 (one-sixth the MTD in dogs) [18] given intravenously over 3 h, every 3 weeks. Dose escalation followed a modified Fibonacci scheme [22]. Prophylactic anti-emetics and premedications were not routinely administered; however, anti-emetics were allowed at the investigator's discretion after documented nausea during a previous infusion. DLT was defined as any first cycle grade 4 neutropenia for ≥7 consecutive days or febrile neutropenia; grade 4 thrombocytopenia or bleeding episode requiring platelet transfusion; grade ≥3 nausea and/or vomiting despite maximal medical intervention; all other grade ≥3 non-hematological toxicity; or delay ≥4 weeks from the time of scheduled retreatment, if due to delayed recovery of drug-related toxicity. One patient per cohort was enrolled until the first instance of grade ≥2 drug-related toxicity (except nausea, vomiting, fatigue, anorexia, or alopecia) or upon observation of cumulative toxicity (meeting the definition of DLT in cycle 2 and above). Thereafter, a standard 3 + 3 design was utilized.</p><p>Toxicity was graded according to the National Cancer Institute Common Terminology Criteria for Adverse Events, v3.0 [23]. Dose reduction by one level was allowed for patients who experienced DLT. The MTD was defined as the highest dose at which no more than one of six patients in the cohort developed DLT. Once the MTD was defined, an additional 10 patients were enrolled to further define the RP2D and to characterize toxicity, pharmacokinetics, and pharmacodynamic relationships.</p><p>KOS-1584 was supplied by Kosan Biosciences, Inc (Hayward, CA) and diluted with KOS-1584 diluent (133 mg/mL hydroxypropyl-β-cyclodextrin in water for injection) and the appropriate amount of drug was withdrawn and mixed with saline (0.9% w/v sodium chloride) to achieve a final concentration of 0.05 or 0.1 mg/mL.</p><!><p>A medical history and physical history with detailed neurological examination, vital signs, performance status assessment, and laboratory determinations including pregnancy test, prothrombin time, activated partial thromboplastin time, and urinalysis were obtained at baseline and at the beginning of each cycle. Complete blood count and serum chemistries were obtained weekly. Three serial 12-lead electrocardiograms (ECGs) were obtained prior to and 30 min after treatment during cycles 1 and 2. The QT interval was determined using automated readings and corrected for heart rate according to Bazette's formula (QTcB) [24]. Fecal occult blood tests (FOBT) were performed at the pretreatment evaluation and on day 3 of the first two cycles. RECIST response evaluation was performed every two cycles.</p><!><p>Blood specimens were collected preinfusion, just prior to the end of infusion, 5 min and 0.5, 1, 2, 3, 5, 8, 10, 24, 48, and 72 h after the end of infusion during cycles 1 and 2. Blood was collected into EDTA-containing tubes, placed on ice, and plasma was separated from whole blood and stored at −70°C until analysis.</p><p>At the MTD level, urine specimens were collected at baseline and during the intervals of 0–5 and 5–24 h following the start of the infusion of the first two cycles and stored at 2–8°C. The total volume was recorded, and three 30-mL aliquots from each interval were frozen at −70°C until analysis. Completeness of the 24-h collection was determined by a 24-h urinary creatinine. A proprietary LC/MS/MS method developed by Kosan Biosciences, Inc was used to identify urinary KOS-1584 and metabolites over time.</p><!><p>Blood specimens were collected into heparin-CPT tubes prior to treatment, at the end of infusion, and 1, 3, and 24 h following the end of the first two infusions. Peripheral blood mononuclear cells (PBMCs) were isolated by centrifugation, and red blood cells were removed by brief hypotonic lysis. PBMC slides were prepared in a cytospin (Thermo Shandon, Inc., Pittsburgh, PA) and sent to Kosan Biosciences, Inc. for immunohistochemistry and quantitation of cells exhibiting tubulin bundle formation using a previously described method [25].</p><!><p>Tumor response and safety data were reported by descriptive statistics. Plasma concentration data were analyzed by non-compartmental methods [26] using Kinetica™, Version 4.4.1 (Thermo-Fisher Scientific Corporation, Philadelphia). The ratio of KOS-1891 AUC0–∞ to KOS-1584 AUC0–∞ was used as a measure of the extent of metabolism of KOS-1584 to KOS-1891. KOS-1584 and its metabolites were profiled in the urine. A sigmoidal maximum effect pharmacodynamic model was fit to describe the dose–response relationship between plasma concentration of KOS-1584 and percent microtubule bundle formation. The maximum effect (Emax), dose at which 50% of the maximum is produced (ED50), and steepness factor (γ) were estimated using WinNonlin Version 1.5 (Pharsight, Cary, North Carolina).</p><!><p>Between November 2004 and August 2007, sixty-six patients were treated across 14 cohorts, at doses ranging from 0.8 to 48 mg/m2. A total of 49 cycles (median 2, range 1–8) were given to the 16 patients at 36 mg/m2 (MTD) cohort. Patients received a median of four prior chemotherapy regimens, including taxanes (62%), platinums (81%), or both (49%). Thirty-one patients (47%) had baseline grade 1 sensory neuropathy. Patient demographics and characteristics are listed in Table 1.</p><!><p>Common treatment-related adverse events (AEs) are listed in Table 2. Six DLTs were observed. The third patient treated at 36 mg/m2 experienced a DLT of arthralgia, after which three additional patients were treated at this dose level without DLT. Two patients were then treated at 48 mg/m2; one developed a DLT of arthralgia and the other (a patient with extensive stage small cell lung cancer) developed a DLT of encephalopathy in the setting of dehydration and hyponatremia. Two of six patients subsequently treated at 42 mg/m2 developed DLT of diarrhea. Therefore, the MTD was defined as 36 mg/m2, and the cohort was expanded with an additional ten patients treated, one of whom developed a grade 3 diarrhea. At dose level 6.5 mg/m2, one patient had G3 dyspnea, occurring in cycle 2 and was not considered DLT. Another patient died of streptococcus pneumonia and sepsis, and although this was subsequently felt not to be related to drug, the cohort was expanded to 6 patients to further explore toxicity at this dose level.</p><p>Most AEs, including sensory neuropathy, were generally low grade and transient. Sensory neuropathy occurred more frequently at doses ≥20 mg/m2 with onset during the first two cycles. Patients who developed either new or worse grade neuropathy generally had symptom resolution or return to baseline shortly after drug discontinuation. Similarly, diarrhea and arthralgia had early onset, were not cumulative, and were managed effectively with anti-diarrheal and anti-inflammatory agents, respectively. Five patients had positive FOBT or hematochezia. There were no cases of bowel perforation, necrosis, or significant GI bleeding. Additional grade 3 toxicities not listed in Table 2 occurred at doses generally higher than RP2D and included dehydration (42 mg/m2), hypocalcemia (42 mg/m2), hypoalbuminemia (27 mg/m2), tremor (42 mg/m2), hypotension (42 mg/m2), encephalopathy (48 mg/m2), and painful respiration (36 mg/m2). One episode of grade 3 small bowel obstruction was observed during the second week of cycle 1 in a patient with metastatic non-small cell lung cancer treated at the 42 mg/m2 dose level, after having been treated with loperamide for grade 3 diarrhea in the first week. Comparisons of baseline or pre- and post-infusion ECGs did not show any effects of KOS-1584 on QTcB interval.</p><p>Five patients discontinued therapy due to adverse effects from KOS-1584: one patient each at 27 mg/m2 (grade 1 nausea, vomiting, and neuropathy), 36 mg/m2 (grade 2 neuropathic pain), and 48 mg/m2 (encephalopathy); and two patients at 42 mg/m2 (grade 2 neuropathy; fatigue and grade 3 neuropathy). Two patients died within 4 weeks of receiving study drug: one at 6.5 mg/m2 due to streptococcal pneumonia with septicemia occurring on day 3 of the first cycle, and one at 20 mg/m2 due to complications from a malignant pleural effusion.</p><!><p>Concentration–time profiles of KOS-1584 were available for all 66 patients treated. KOS-1584 concentrations reached a maximum at the end of the infusion and declined biexponentially. Plasma profiles for the seco-metabolite, KOS-1891, demonstrated much lower concentrations but were similar in shape for cohorts dosed at 6.5 mg/m2 and higher.</p><p>Table 3 summarizes the PK parameters of KOS-1584 and KOS-1891 for cycle 1. The mean overall clearance, volume of distribution, and half-life of KOS-1584 were 11.0 ± 6.17 L/h/m2, 327 ± 161 L/m2, and 21.9 ± 8.75 h, respectively. The half-life for KOS-1891 was 29.6 ± 13.8 h, and the KOS-1891/KOS-1584 AUC ratio was 9.43 ± 4.86%. There appears to be no dose dependency in clearance values, consistent with linear kinetics.</p><p>Urinary excretion of unchanged KOS-1584 represented <10% of drug-related materials. In addition to KOS-1891, three glucuronide metabolites and seven oxidative metabolites were excreted in urine, suggesting metabolism of KOS-1584 by multiple pathways.</p><!><p>The degree of microtubulin bundle formation (MBF) was assessed as a surrogate marker of binding of KOS-1584 to tubulin in PBMCs. For all cohorts, the percentage of PBMCs with microtubule bundles increased to a maximum at the end of drug infusion and then decreased by 24 h after drug infusion (Fig. 2a). A dose-dependent increase in MBF was observed with a plateau of maximal effect between 50 and 60% at the 3–4 highest dose levels (27–48 mg/m2). The pattern of MBF was similar across cycles 1 and 2.</p><p>Figure 2b describes the relationship between the percentage of PBMCs with MBF and KOS-1584 concentration in plasma. A sigmoidal maximum effect pharmacodynamic model (R2 = 0.943) fits this relationship with an EC50 = 24.1 ng/mL, Emax = 52.7%, slope = 1.10, and E0 = 3.69%.</p><!><p>Two patients experienced objective partial responses (PR). The first patient, treated at the 6.5 mg/m2 dose level, had advanced ovarian cancer and was heavily pretreated with 6 prior regimens including paclitaxel and docetaxel. She had PR (−31%) of target retroperitoneal lymph nodes; however, a confirmatory scan was not performed. The second patient, treated at the 36 mg/m2 dose level, had metastatic pancreatic cancer that was previously treated with four chemotherapy regimens including paclitaxel. He had confirmed PR, determined by CR of his target retroperitoneal lymphadenopathy, and SD of non-target, subcentimeter pulmonary nodules and hilar lymphadenopathy. Stable disease (SD) was observed in 24 patients, 13 of whom had durable SD of ≥3 months. Among these 24 patients with SD, the median number of prior regimens was 4.5 and the best response to their most recent previous therapy included progressive disease (PD) in 15, SD in 5, PR in 1, and not evaluable/data not available in 3 patients.</p><!><p>Epothilones are microtubule-active agents with potential to overcome the issues of MDR, myelosuppression, neurotoxicity, and hypersensitivity reactions associated with the currently available agents. Ixabepilone, currently the only FDA-approved epothilone, has DLTs of neutropenia, neuropathy, and abdominal pain/nausea [9, 10, 27–29]. It is formulated with a cremophor base; therefore, similar to taxanes, hypersensitivity has been observed and premedication is required. Other epothilones currently under investigation include patupilone, which has DLTs of diarrhea, peripheral neuropathy, and fatigue [11, 12, 30, 31], and sagopilone, with DLTs of peripheral neuropathy, infection, hyponatremia, diarrhea, and central ataxia [11, 13, 14]. KOS-1584, specifically designed to have more favorable pharmacokinetics and less toxicity compared with first-generation epothilone D derivatives, has demonstrated activity in paclitaxel-resistant xenografts and in cell lines with MDR due to overexpression of P-gp.</p><p>In this first-in-man, phase I dose escalation study, we evaluated the safety, tolerability, and biologic activity of KOS-1584. The MTD and RP2D are 36 mg/m2 each. We treated patients on 14 cohorts and saw a 45-fold difference between the starting dose and RP2D, illustrating the shortcomings of dose translation from animal to human studies as previously documented by others [32–34]. Observed DLTs included diarrhea, arthralgias, and encephalopathy. In contrast to a paralleled conducted study of KOS-1584, given on two weekly dosing schedules where diarrhea was increasingly more severe after successive infusions despite maximal supportive care [35], diarrhea was managed effectively with loperamide or diphenoxylate/atropine in our trial. Arthralgias also were effectively treated and subsequently prevented, with anti-inflammatory agents or corticosteroids.</p><p>It is unclear at this point whether the DLT of encephalopathy was related to drug penetration into the blood–brain barrier (BBB). Evaluation of penetration of KOS-1584 across the intact BBB has not been evaluated in humans. However, in tissue distribution studies performed in mice, KOS-1584 did penetrate the BBB [18]. Continued vigilance for central nervous system toxicity will be required in future studies of this drug.</p><p>Peripheral neuropathy was common, although grade ≥3 peripheral neuropathy occurred only at the two highest dose levels (above the MTD). This observation was very encouraging given that most of our patients were previously treated with taxanes and/or platinums, and many had grade 1 neuropathy at study entry. Generally, new or worsening neuropathy occurred during the first two cycles, persisted at the same severity, and resolved upon drug discontinuation. At the RP2D, the frequency of all-grade and grade ≥3 peripheral neuropathy was 69 and 0%, respectively. Peripheral neuropathy is also a common adverse effect of ixabepilone [36] and was a DLT for sagopilone [13] and patupilone [12, 31]. KOS-1584 was not associated with any hypersensitivity reactions and had a very low incidence of myelosuppression, even among this group of heavily pretreated patients. KOS-1584 exhibited linear pharmacokinetics, with no accumulation of KOS-1584 or its secometabolite. As KOS-1584 is a potent inhibitor of CYP3A4/5 in vitro, potential interactions with CYP3A4/5 substrates should be evaluated in future trials [18]. There were no discernable differences in the pharmacokinetics of KOS-1584 in patients who experienced grade ≥3 drug-related toxicities compared to those who did not.</p><p>Similar to ixabepilone, a dose-dependent increase in PBMC MBF was observed. At the RP2D, the mean percentages of MBF were 50 and 26% at 1 h and 24 h after infusion, respectively. Ixabepilone at the RP2D of 40 mg/m2 (using this same assay) demonstrated mean MBF of 63% and 16–23% (n = 27) at 1 h and 24 h after infusion, respectively [10, 25]. An association between the maximal percentage of PBMC MBF and severity of neutropenia has been previously reported with ixabepilone [10, 37]. In contrast, we did not observe any correlation between the maximal MBF of KOS-1584 and any grade neutropenia or grade ≥3 toxicity.</p><p>The disease control rate (CR + PR + SD) was 39%. MBF was not an accurate predictor of clinical benefit among evaluable patients; however, the effects of KOS-1584 on this target were not measured in pre- and post-exposure tumor samples.</p><p>The RP2D of KOS-1584 is 36 mg/m2 administered intravenously every 3 weeks. The lack of prohibitive severe neurologic toxicity, neutropenia, or hypersensitivity reactions; favorable pharmacokinetic profile; and early evidence of activity support further evaluation.</p>

PubMed Author Manuscript

Lead Optimization of Dehydroemetine for Repositioned Use in Malaria

Drug repositioning offers an effective alternative to de novo drug design to tackle the urgent need for novel antimalarial treatments. The antiamoebic compound emetine dihydrochloride has been identified as a potent in vitro inhibitor of the multidrug-resistant strain K1 of Plasmodium falciparum (50% inhibitory concentration [IC50], 47 nM ± 2.1 nM [mean ± standard deviation]). Dehydroemetine, a synthetic analogue of emetine dihydrochloride, has been reported to have less-cardiotoxic effects than emetine.

lead_optimization_of_dehydroemetine_for_repositioned_use_in_malaria

TEXT<!><!>TEXT<!>Molecular modeling of ligand interactions with P. falciparum 80S ribosome.<!><!>Molecular modeling of ligand interactions with P. falciparum 80S ribosome.<!><!>Molecular modeling of ligand interactions with P. falciparum 80S ribosome.<!><!>Experimental determination of IC50s to test activity against the K1 strain of P. falciparum.<!><!>Time course analysis for determination of speed of action of (−)-R,S-dehydroemetine and (−)-S,S-dehydroisoemetine.<!><!>Stage-specific profiling of (−)-R,S-dehydroemetine and (−)-S,S-dehydroisoemetine using synchronized cultures of Plasmodium falciparum.<!><!>MTT assay for cell cytotoxicity against HepG2 cells.<!><!>MTT assay for cell cytotoxicity against HepG2 cells.<!>Determination of cross-resistance through 3H-hypoxanthine incorporation assay.<!><!>Determination of cross-resistance through 3H-hypoxanthine incorporation assay.<!>In vitro IC50s against P. falciparum male and female activated gametes.<!><!>hERG channel inhibition assay.<!>Data analysis of hERG channel inhibition assay results.<!><!>Staining with rhodamine 123 for measurement of mitochondrial membrane potential.<!><!>Staining with rhodamine 123 for measurement of mitochondrial membrane potential.<!><!>Validation of the CalcuSyn assay for drug interaction analysis for malaria.<!><!>CalcuSyn-based drug interaction analysis of the atovaquone-proguanil combination.<!><!>CalcuSyn-based drug interaction analysis for the (−)-R,S-dehydroemetine–atovaquone combination.<!><!>CalcuSyn-based drug interaction analysis for the (−)-R,S-dehydroemetine–proguanil combination.<!><!>CalcuSyn-based drug interaction analysis for the (−)-R,S-dehydroemetine–proguanil combination.<!>DISCUSSION<!>Culture of Plasmodium falciparum.<!>Synthesis of (−)-2,3-dehydroemetine.<!><!>Synthesis of (−)-2,3-dehydroemetine.<!>Molecular modeling.<!>Drug preparation.<!>Experimental determination of IC50 to test activity against the K1 strain of P. falciparum.<!>SYBR green staining of erythrocytic-stage P. falciparum parasites for flow cytometry.<!>Time course analysis through IC50 speed assay using unsynchronized cultures of Plasmodium falciparum.<!>Stage-specific profiling of (−)-R,S-dehydroemetine and (−)-S,S-dehydroisoemetine using synchronized cultures of Plasmodium falciparum.<!>MTT assay for cell cytotoxicity against HepG2 cells.<!>Derivation of dose-response curves and IC50 values.<!>Determination of cross-resistance through IC50 determination in multidrug-resistant P. falciparum strains.<!>Determination of transmission-blocking potential through in vitro inhibition of gamete activation.<!>hERG channel inhibition (IC50 determination) assay protocol.<!>Staining with rhodamine 123 for mitochondrial membrane potential disruption.<!>Drug interaction analysis for (−)-R,S-dehydroemetine.<!><!>Ethics statement.

<p>Malaria presents a huge burden on the economic development of countries of endemicity (1). In 2017, WHO reported 219 million malaria cases globally, with an estimated 435,000 deaths, occurring mostly among African children. The 20th century witnessed the development of a range of antimalarials, including quinine alternatives like mepacrine, chloroquine, and primaquine, antifolates like sulfadoxine and pyrimethamine, and artemisinin (2–7). However, the emergence of resistance against all known classes of antimalarial drugs, including artemisinin chemotherapy, has warranted research into the development of new drugs with novel targets against the parasite (8, 9). Drug discovery is hindered by long development timelines, high attrition rates, and soaring research and development costs (10). Antimalarial chemotherapy has predominantly relied on compounds based on natural products (11). Emetine dihydrochloride hydrate, a natural product alkaloid derived from Psychotria ipecacuanha, was superseded by the introduction as an antiamoebic drug of a safer drug, metronidazole, by the 1970s (12). Repositioning screens carried out at the University of Salford, United Kingdom, identified emetine dihydrochloride as having potent, nanomolar in vitro antimalarial efficacy in the multidrug-resistant (MDR) Plasmodium falciparum parasite strain K1. The comparatively significant difference in in vitro antiprotozoan efficacies (50% inhibitory concentration [IC50] of 47 ± 2.1 nM [mean ± standard deviation] in P. falciparum compared to IC50 of 26.8 ± 1.27 μM in Entamoeba histolytica) dictate that the safety profile for its repositioned use as an antimalarial could be different (13, 14). The pleiotropic natural product drug has been reported to have antiviral and anticancer properties, including recent reports of interrupting viral replication and cell entry for Zika and Ebola viruses (15). The 40S ribosomal subunit of the eukaryotic 80S ribosome was recently reported to be the site of action of emetine and is available as a cryo-electron microscopy (cryo-EM) structure with PDB code 3J7A (16). The definition of the target binding site for emetine enables a chimeric approach to refine, using rational design, a drug discovered through repositioning.</p><p>The (R) configuration at C-1′ and the presence of secondary nitrogen at position 2′ are important for emetine's biological activity (Fig. 1). Indeed, the (S) configuration at C-1′ (isoemetine) or the substitution of the secondary amine results in loss of activity. Even if the asymmetry at carbons 2 and 3 is lost with unsaturation at position 2-3 (2,3-dehydroemetine), the biological activity is retained (17). In 1980, a study conducted on the cross-resistance of emetine-resistant mutants of Chinese hamster ovary cells to related compounds found that the distance between the two aromatic rings and the angle between the nucleophilic element, such as nitrogen, and the rings were essential for biological activity (18).</p><!><p>Structure of emetine hydrochloride.</p><!><p>Emetine in its crude form has been in use since 1658 for the treatment of dysentery and was first brought to Europe from Brazil by Piso (19). Powdered ipecacuanha introduced in Mauritius in 1858 reduced the annual death rate from severe dysentery from 10 to 18% to 2% (20). Generalized muscle weakness, vomiting, and cardiotoxicity were the side effects with prolonged use of emetine (21, 22). A less emetic and safer synthetic analogue of emetine which could be given as a resinate, 2,3-dehydroemetine, was introduced in 1959 (23). At high doses, the electrocardiographic changes were similar but were less marked and of shorter duration. The observed changes in heart conduction, contractility, automaticity, and electrocardiogram (ECG) abnormalities caused by emetine and 2,3-dehydroemetine could be due to their effect on membrane permeability to Na+, K+, and Ca++ ions. Dehydroemetine has been reported to have less-cardiotoxic effects than emetine for the treatment of amoebic liver abscesses (24–26).</p><p>Previous studies have found that dehydroemetine was eliminated faster from the body and more rapidly from the heart than from the liver, while the reverse was found to be true for emetine (27). Based on the anecdotal evidence, two diastereomers of dehydroemetine, (−)-R,S-dehydroemetine and (−)-S,S-dehydroisoemetine, were synthesized. Molecular modeling tools (28, 29) were used to predict the activities of the diastereomers and their potencies against the multidrug-resistant K1 strain of P. falciparum.</p><p>The parasites are transmitted from their mammalian hosts to mosquito vectors through mature Plasmodium gametocytes, and hence, the reinfection cycle could be broken with the use of transmission-blocking antimalarials (30). The cross-resistance and gametocidal activities of emetine dihydrochloride and its synthesized analogues were tested by GSK in a bid to determine the potential of these compounds as transmission-blocking drugs.</p><p>A number of studies have reported that dehydroemetine and emetine are potent inhibitors of protein synthesis (17, 31). It has also been reported that emetine affects the myocardium in a dose-dependent manner (32). Drugs inhibiting the cardiac potassium ion channel encoded by the human ether-a-go-go-related gene (hERG) can prolong the QT interval and cause a dangerous cardiac arrhythmia, Torsades de pointes, which has hampered a number of drug discovery and development projects (33). In this study, we tested the activity of emetine dihydrochloride and its two synthetic analogues against the hERG potassium channel. The three compounds were also tested for any effect on mitochondrial membrane potential (MMP), as unpublished data from previous studies conducted by our group predicted atovaquone-like activity affecting MMP by the parent compound, emetine.</p><p>The preference for combinatorial regimes over monotherapy for the treatment of malaria has affected the drug discovery pipeline in a crucial way (34). Newer drug candidates have been tested for synergistic activities with existing antimalarial treatments for dose reduction to improve therapeutic and safety profiles. Furthermore, combinatorial regimes expand the effective life of the antimalaria drugs by delaying the emergence of resistance (35, 36).</p><p>Chou and Talalay developed a method based on the argument that the issue of synergy is more physiochemical rather than statistical in nature and employed the law of mass action to derive a median-effect equation where additivity could be defined using the resulting combination index (CI = 1), with antagonism and synergism defined as >1 and <1, respectively (37). CalcuSyn software based on the complex algorithms for median-effect analysis allows automation and eliminates subjectivity during data analysis (37). We have previously demonstrated the use of CalcuSyn as a reliable method to define antimalarial drug interactivity for combinatorial regimes (38).</p><p>Hence, to reduce dose-dependent side effects, combinatorial partner drugs showing synergistic activity were sought. (−)-R,S-Dehydroemetine was taken forward for drug interaction studies as it was found to be highly potent against the K1 strain of P. falciparum. Atovaquone and proguanil were first evaluated to test the efficacy of the method, as the two drugs are known to exhibit synergism (39). The methodology was then applied to evaluate combinatorial partner drugs for the potent antimalarial candidate (−)-R,S-dehydroemetine.</p><!><p>In order to explore the structural basis of the relative inhibitory activities of (−)-R,S-dehydroemetine and (−)-S,S-dehydroisoemetine, we predicted and compared their molecular interactions with the P. falciparum 80S ribosome using computational docking. As the receptor structure, we used the recently solved cryo-EM structure of the 80S ribosome of P. falciparum, determined to a resolution of 3.2 Å (16). The electron density corresponding to the bound emetine is located in the E site of the ribosomal small subunit, i.e., the P. falciparum 40S ribosome (Pf40S); this E site, a binding site that pactamycin also recognizes in the bacterial 30S subunit, is at the interface between 18S rRNA helices 23, 24, and 45 and the C terminus of protein uS11.</p><p>On closer examination of the published cryo-EM structure of the complex, the ligand structure originally modeled into the density erroneously corresponds to the nonnatural enantiomer of emetine, (1S,2R,3S,11bR)-emetine. This modeled structure was not the ligand employed in the associated cryo-EM experiments (16). Therefore, we docked emetine, i.e., the (1R,2S,3R,11bS) structure, into the identified Pf40S binding site using MOE-Dock (2016; Chemical Computing Group, Inc., Montreal, QC, Canada). The docked emetine geometry maps well into the electron density envelope (shown at a contour level of 0.1542 electrons/Å̂3 [3.50 root mean square deviation {RMSD}] in Fig. 2a); the bound pose broadly follows the twisted U shape conformation of the observed electron density, with improved alignment in comparison to the previously published nonnatural emetine enantiomer geometry (Fig. 2b).</p><!><p>(a) Overlay of the docked pose of emetine (green) with its enantiomer present in the cryo-EM structure (maroon); observed electron density envelope is also shown (wireframe surface with contour level 0.1542 electrons/Å̂3 [3.50 RMSD]). (b) Interactions of docked emetine (green) with Pf40S residues and comparison with previously modeled interactions of its enantiomer (blue) (16).</p><!><p>In this U shape, an intermolecular T-shaped π-stacking interaction is observed between the two cyclic systems of emetine, i.e., benzo[a]quinolizine rings A/B/C and isoquinoline rings D/E (Fig. 1 and 2a). The docked emetine also forges a number of comparable interactions with the Pf40S subunit as its modeled enantiomer (Fig. 2b), with a key π-stacking interaction between the A/B/C rings of emetine and the purine ring of G973 of h23 (16). The locations of the emetine secondary and tertiary amines are broadly similar (within 1 to 2 Å), allowing hydrogen bonding interactions with a backbone oxygen atom of U2061 (h45) and the 2′-hydroxyl group of U1068 (h24), respectively (Fig. 2b). The tertiary amine also forms a salt bridge interaction with the carboxylate side chain of C-terminal residue Leu151 of uS11 (Fig. 3b). This interaction was not highlighted in the cryo-EM study and is a consequence of modeling the natural emetine geometry into the E site.</p><!><p>(a) Overlay of docked poses of (−)-R,S-dehydroemetine (red) and (−)-S,S-dehydroisoemetine (cyan) with emetine (blue). (b) Interactions of (−)-R,S-dehydroemetine (red) and (−)-S,S-dehydroisoemetine (cyan) with the Pf40S binding site. Distances (dotted lines) in Å.</p><!><p>Subsequently, the two diastereomers of emetine, (−)-R,S-dehydroemetine and (−)-S,S-dehydroisoemetine, were docked in turn into the emetine binding region of the Pf40S subunit. We found that the preferred docked pose of the (−)-R,S-dehydroemetine adopts the familiar U-shaped conformation, superimposing rather closely onto the bound pose of emetine (Fig. 3a). The docking scores are correspondingly similar for emetine, with a London dG value (a scoring function in molecular docking that estimates the binding free energy of the ligand from a given pose) of −7.2 kcal/mol, and (−)-R,S-dehydroemetine, which has a dG score of −7.3 kcal/mol. As observed for emetine, (−)-R,S-dehydroemetine forms the π−π stacking interaction with the G973 pyrimidine ring and polar interactions with U2061, U1068, and Leu151 (Fig. 3b).</p><p>However, (−)-S,S-dehydroisoemetine docks into the binding site with a lower dG score of −6.5 kcal/mol and does not overlay in conformation with emetine or (−)-R,S-dehydroemetine so readily (Fig. 3a). The particular stereochemical configuration of (−)-S,S-dehydroisoemetine appears to result in its secondary amine being more distant from the E-site residues of Pf40S. Consequently, this amine N…O U2061 distance extends by 0.8 Å in proceeding from the R to the S isomer (Fig. 3b). The tertiary amine interaction with U1068 is maintained, however, as is the interaction with the terminal carboxylate of Leu151 (Fig. 3b). The π stacking interaction with G973 is also present, but at a slightly larger distance between planes, increased by ∼0.7 Å. Figure 4a and b show the predicted binding site residues for (−)-R,S-dehydroemetine and (−)-S,S-dehydroisoemetine.</p><!><p>(a) Docking through MOE showing the binding site residues for (−)-R,S-dehydroemetine molecule. (b) Docking through MOE showing the binding site residues for (−)-S,S-dehydroisoemetine molecule.</p><!><p>Based on the anecdotal evidence, two synthetic analogues of emetine dihydrochloride, (−)-R,S-dehydroemetine and (−)-S,S-dehydroisoemetine, were synthesized. Experiments to test drug efficacy were set up, and (−)-R,S-dehydroemetine was tested in 2-fold serial dilutions from 12.5 nM to 200 nM. A dose-response experiment was set up on synchronized ring stage cultures of the K1 strain of P. falciparum to be read at 72 h. The IC50 of (−)-R,S-dehydroemetine was observed to be 71.03 ± 6.1 nM (Fig. 5a). (−)-S,S-dehydroisoemetine was tested in 2-fold serial dilutions from 0.625 μM to 10 μM. The IC50 was observed to be 2.07 ± 0.26 μM (Fig. 5b). The results shown in Fig. 5 are derived from representative experiments performed thrice, with each concentration of (−)-R,S-dehydroemetine and (−)-S,S-dehydroisoemetine tested in triplicates.</p><!><p>(a) The effective dose of (−)-R,S-dehydroemetine (dose range tested in 2-fold serial dilutions from 12.5 nM to 200 nM) on P. falciparum K1 infection after an incubation period of 72 h using SYBR green-based plate reader assay. (b) The effective dose of (−)-S,S-dehydroisoemetine (dose range tested in 2-fold serial dilutions from 0.625 μM to 10 μM) on P. falciparum K1 infection after an incubation period of 72 h using SYBR green-based plate reader assay. The experiments were performed thrice with each concentration of (−)-R,S-dehydroemetine and (−)-S,S-dehydroisoemetine (tested in triplicates). Data were analyzed using GraphPad prism. Error bars show standard deviations.</p><!><p>Accurate determination of the parasite killing rate in response to treatment is crucial in the development of drugs against P. falciparum, as chemotherapy remains the primary element in the control of malaria. The speed of action of compounds on the viability of parasites is difficult to measure through traditional techniques (40). It is important to identify new drugs with rapid parasite-killing kinetics early in the drug development process. Besides rapid relief of symptoms, a fast-acting drug also helps to curtail the mutations causing the development of new mechanisms of drug resistance. In 2012, Tres Cantos developed a labor-intensive low-throughput assay taking up to 28 days to determine in vitro the parasite reduction ratio (PRR) and the presence or absence of a lag phase in response to a drug (41). The method used in this study to differentiate between fast- and slow-acting compounds gives initial results in 4 to 7 days (42).</p><p>The IC50 speed assay was performed for (−)-R,S-dehydroemetine and (−)-S,S-dehydroisoemetine (with concentrations of 25 nM, 50 nM, 100 nM, 200 nM, and 400 nM and 0.63 μM, 1.25 μM, 2.5 μM, 5 μM, and 10 μM, respectively) using unsynchronized cultures of P. falciparum. The ratios of 24-h IC50s to 72-h IC50s for (−)-R,S-dehydroemetine and (−)-S,S-dehydroisoemetine could not be determined, as the IC50s could not be reached within 24 h. Both compounds only achieved 50% inhibition at the previously defined IC50s after 48 h of exposure, indicating that the isomers have delayed action against the multidrug-resistant K1 strain of P. falciparum. Figure 6 shows the results from the IC50 speed assays for (−)-R,S-dehydroemetine and (−)-S,S-dehydroisoemetine [error bars in the figure represent the standard errors from experiments performed twice with each concentration of (−)-R,S-dehydroemetine and (−)-S,S-dehydroisoemetine, tested in triplicates].</p><!><p>Time course analysis through IC50 speed assay using unsynchronized cultures of Plasmodium falciparum. The graphs show that IC50s could not be reached within 24 h. (−)-R,S-dehydroemetine and (−)-S,S-dehydroisoemetine reached the IC50 by 48 h. Error bars represent the standard errors of the results from experiments performed twice with each concentration of (−)-R,S-dehydroemetine and (−)-S,S-dehydroisoemetine (tested in triplicates). (a) (−)-R,S-dehydroemetine at concentrations of 25, 50, 100, 200, and 400 nM at 24 h, 48 h, and 72 h. (b) (−)-S,S-dehydroisoemetine at concentrations of 0.63, 1.25, 2.5, 5, and 10 μM at 24 h, 48 h, and 72 h.</p><!><p>The two isomers of 2,3-dehydroemetine were tested on synchronous cultures [(−)-R,S-dehydroemetine was tested in 2-fold serial dilutions from 132.81 nM to 8,500 nM, and (−)-S,S-dehydroisoemetine was tested in 2-fold serial dilutions from 3.13 μM to 200 μM] to determine the stage specificities of the compounds by measuring the concentration-dependent growth of schizonts and rings following incubation with the two compounds.</p><p>It was observed that (−)-R,S-dehydroemetine and (−)-S,S-dehydroisoemetine affect both the ring and trophozoite/schizont stages of the parasite. The isomers were found to be more active in the late trophozoite/schizont stage, which is consistent with the proposed protein synthesis target of the parent compound. They displayed comparatively less activity against the rings even at high concentrations. At the highest concentration tested, the growth rates for the rings and trophozoites/schizonts 24 h after the (−)-R,S-dehydroemetine postexposure wash were 50.21% and 19.75%, respectively. At the highest concentration tested, the growth rates for the rings and trophozoites/schizonts 24 h after the (−)-S,S-dehydroisoemetine postexposure wash were 70.98% and 24.67%, respectively. The lower potency against ring stage parasites may also explain in part the lag phase observed during the speed assay, where the IC50 was not reached after 24 h of exposure against unsynchronized cultures.</p><p>Figure 7 shows the results for stage-specificity assays of (−)-R,S-dehydroemetine and (−)-S,S-dehydroisoemetine, respectively [error bars in the figure represent the standard errors from experiments performed twice with each concentration of (−)-R,S-dehydroemetine and (−)-S,S-dehydroisoemetine, tested in triplicates].</p><!><p>Stage-specific profiling using synchronized cultures of Plasmodium falciparum. The decrease in growth of trophozoites/schizonts was observed to be more marked than for rings. The drug effects are expressed as percentages of growth of rings relative to growth of trophozoites/schizonts 24 h after the postexposure wash. Error bars represent the standard errors of the results from experiments performed twice with each concentration of (−)-R,S-dehydroemetine and (−)-S,S-dehydroisoemetine (tested in triplicates). (a) Cultures were exposed to a serial dilution of (−)-R,S-dehydroemetine (from 132.81 nM to 8,500 nM) for 24 h. (b) Cultures were exposed to a serial dilution of (−)-S,S-dehydroisoemetine (from 3.13 μM to 200 μM) for 24 h.</p><!><p>MTT [3-(4,5-dimethyl-2-thiazolyl)-2,5-diphenyl-2H-tetrazolium bromide] assays for cell cytotoxicity were performed for (−)-R,S-dehydroemetine and (−)-S,S-dehydroisoemetine. Emetine and cisplatin were used as control drugs (Fig. 8). The plates were read at 48 h (43).</p><!><p>Forty-eight-hour MTT assays. Cells were seeded at 5,000 cells per well. Cell viability was determined using a standard MTT assay. Data were analyzed using GraphPad prism. Error bars represent the standard deviations of the results from experiments performed thrice with each concentration of emetine, cisplatin, (−)-R,S-dehydroemetine, and (−)-S,S-dehydroisoemetine (tested in triplicate). Forty-eight-hour MTT assays for emetine, tested in 2-fold serial dilutions from 31.25 nM to 2,000 nM (a), cisplatin, tested in 2-fold serial dilutions from 0.78 μM to 25 μM (b), (−)-R,S-dehydroemetine, tested in 2-fold serial dilutions from 11.72 nM to 750 nM (c), and (−)-S,S-dehydroisoemetine, tested in 2-fold serial dilutions from 0.312 μM to 10 μM (d).</p><!><p>Emetine was tested in 2-fold serial dilutions from 31.25 nM to 2,000 nM, with the 50% lethal dose (LD50) observed to be 199.2 ± 9.3 nM. (−)-R,S-dehydroemetine was tested in 2-fold serial dilutions from 11.72 nM to 750 nM, with the LD50 observed to be 168.07 ± 8.65 nM. (−)-S,S-dehydroisoemetine was tested in 2-fold serial dilutions from 0.312 μM to 10 μM, with the LD50 observed to be 1.429 ± 0.18 μM. Cisplatin was tested in 2-fold serial dilutions from 0.78 μM to 25 μM, with the LD50 observed to be 10.16 ± 2.6 μM. The selectivity index, calculated as LD50 HepG2 cells/IC50 parasites (44), was 2.37 for (−)-R,S-dehydroemetine and 0.69 for (−)-S,S-dehydroisoemetine.</p><!><p>The cross-resistance assay used relies on the parasites' incorporation of labeled 3H-hypoxanthine, which is proportional to P. falciparum growth. In vitro cross-resistance of the compounds is measured as the ratio between the IC50 for the tested P. falciparum strain and the IC50 for strain 3D7A. Every replicate from a multidrug-resistant (MDR) strain involved a simultaneous determination using a 3D7A replicate to avoid any artifact linked to experimental conditions. Table 1 shows the IC50s of emetine dihydrochloride, (−)-R,S-dehydroemetine, and (−)-S,S-dehydroisoemetine in the sensitive P. falciparum strain (3D7A) and two resistant P. falciparum strains (Dd2 and W2). Using strain 3D7A as a reference, the ratios of in vitro cross-resistance of emetine dihydrochloride were found to be 1.15 for strain Dd2 and 0.77 for strain W2. The ratios for (−)-R,S-dehydroemetine in the two resistant strains, Dd2 and W2, were found to be 1.21 and 1.15, respectively.</p><!><p>Analysis of results of 3H-hypoxanthine incorporation assay to determine cross-resistance</p><p>In vitro IC50s of emetine dihydrochloride, (−)-R,S-dehydroemetine, and (−)-S,S-dehydroisoemetine in sensitive P. falciparum strain 3D7A and resistant P. falciparum strains Dd2 and W2, as well as ratios of in vitro cross-resistance of emetine dihydrochloride and (−)-R,S-dehydroemetine in both resistant strains (Dd2 and W2), using strain 3D7A as reference. ND, not determined.</p><!><p>The results showed that the inhibitory potencies observed for emetine dihydrochloride and (−)-R,S-dehydroemetine in both multidrug-resistant strains (Dd2 and W2) are similar to their inhibitory potencies for the sensitive strain 3D7A. These results suggest that there is no cross-resistance with any of the MDR strains tested.</p><!><p>A bioassay was performed to assess the malaria transmission-blocking potential of compounds on P. falciparum strain NF54 by estimating their ability to prevent male mature gametocytes from progressing to male microgametes or/and to inhibit female gamete activation, as indicators of gametocyte functionality. The NF54 strain was selected for its increased ability to produce gametocytes under in vitro conditions (45). The activation of male gametocytes into mature microgametes is evaluated by the process of exflagellation (extrusion of rapidly waving flagellum-like microgametes from the infected erythrocyte). The activation of female gametocytes is evaluated based on the specific expression of the P. falciparum s25 (Pfs25) protein at the surface of the female activated gametes (46). Male P. falciparum gametocytes exflagellate when activated, causing movement of the surrounding red blood cells (RBCs) in the medium. By detecting these changes in cells' positions, we were able to detect activated male gametes. Female P. falciparum gametocytes round up when activated, and the Pfs25 protein becomes widely distributed in the membrane of the gamete. Using a monoclonal antibody against this protein, we were able to specifically detect activated female gametes. A dual gamete formation assay was performed, and the results are shown in Table 2.</p><!><p>In vitro IC50s of (−)-R,S-dehydroemetine and (−)-S,S-dehydroisoemetine against male and female gametocytes in P. falciparum strain NF54</p><p>In vitro IC50s of (−)-R,S-dehydroemetine and (−)-S,S-dehydroisoemetine in P. falciparum dual gamete formation assay.</p><!><p>Dose-related cardiovascular side effects observed following treatment of amoebiasis with emetine included ECG changes such as T-wave inversion, prolongation of QT interval, and widening of the QRS complex and PR interval. Hypotension, tachycardia, and precordial pain were also observed (25). Stoppage of treatment resulted in complete recovery of cardiovascular functions. Cardiac microscopic examination revealed a separation of muscle fibers and destruction of myocardial fibers but an absence of inflammatory cells, leading to the interpretation that the myocarditis is toxic rather than inflammatory in origin. In another study on 32 patients, pain was noted at the injection site, along with ECG abnormalities, myalgia, muscle weakness, and increased levels of serum creatinine phosphatase (47). In a study on guinea pigs by Schwartz and Herrero (27), it was observed that dehydroemetine was excreted faster than emetine. It was also postulated that reduced cardiotoxicity of dehydroemetine could be due to decreased tissue affinity to the heart in comparison to that of emetine (25).</p><p>In a resting cardiac cell, the concentration of K+ ions is high intracellularly, which creates a chemical gradient for K+ ions to diffuse out of the cells. A subunit of the rapid delayed rectifier potassium ion channel is involved in the cardiac repolarization (48). It is encoded by the hERG gene (human ether-a-go-go-related gene). Since emetine is known to affect the movements of Na+, K+, and Ca2+ ions, 2,3-dehydroemetine is also thought to affect ion permeability. hERG channel inhibition assays were therefore carried out to determine potential influences on the cardiotoxicity previously observed with emetine therapy.</p><!><p>For each replicate, the hERG response was calculated using the following equation: % hERG response = post-compound-application current (nA)/pre-compound-application current (nA) × 100. The % hERG response was plotted against the concentration of the test compound, and where concentration-dependent inhibition was observed, the data were fitted to the following equation and an IC50 value calculated:y=ymax −ymin 1+(IC50x)s+ymin where y is the hERG response, ymax is the mean of the 100% vehicle control response, ymin is the mean of the 0% vehicle control response, x is the concentration, IC50 is the concentration required to inhibit current by 50%, and s is the Hill slope.</p><p>(−)-R,S-dehydroemetine has an IC50 of 19.3 μM for the hERG channel, whereas (−)-S,S-dehydroisoemetine has an IC50 of 2.99 μM (Table 3). The IC50 of quinidine, which was used as a positive control, is 1.99 μM. The selectivity index was calculated as IC50 hERG/IC50 parasites. Thus, it was found that, with a selectivity index (SI) of over 271, (−)-R,S-dehydroemetine is not an hERG channel inhibitor, but (−)-S,S-dehydroisoemetine is a potent inhibitor (selectivity index = 1.48).</p><!><p>hERG channel inhibition assay results for (−)-R,S-dehydroemetine and (−)-S,S-dehydroisoemetinea</p><p>This experiment was outsourced to Cyprotex, UK. Each value is the mean of triplicate values.</p><!><p>Changes in mitochondrial membrane potential were measured using rhodamine 123 (excitation wavelength, 511 nm; emission wavelength, 534 nm), a membrane-permeating cationic fluorescent dye (49), which accumulates by electrostatic attraction in the mitochondria because of its negative transmembrane potential. A change in the dye's concentration in the mitochondria is caused by a depolarization event and can be visualized as a shift in the fluorescence intensity of rhodamine 123 (50). Draq5 (excitation wavelength, 647 nm; emission wavelength, >665 nm) is a far-red fluorescent DNA dye which is cell permeating and was used to distinguish the parasites in flow cytometry in the allophycocyanin (APC)-Cy7-A channel. SYBR green was not preferred for this experiment as it emits in the same channel as rhodamine 123 and would result in the overlap of emission signals. Fluorescence microscopy was used to visualize the localization of rhodamine 123 within the cytoplasm of the parasites (Fig. 9).</p><!><p>Staining of parasites with rhodamine 123 and Draq5. Merge with bright-field image shows localization of the dye within the parasite. P. falciparum K1 strain trophozoites stained with 10 μM Draq5 (a) and 200 nM rhodamine 123 (b), Draq5 and rhodamine merge (c), and Draq5, rhodamine 123, and brightfield merge (d). Visualized under fluorescence microscope at ×100 magnification.</p><!><p>A decrease in fluorescence intensity measured in the fluorescein isothiocyanate (FITC)-A channel indicates a loss of mitochondrial membrane potential. Atovaquone, a known mitochondrial inhibitor, was used as a control. Shifts in fluorescence intensity were observed after treatment with all three compounds at IC50s and 10× IC50s. Emetine and (−)-R,S-dehydroemetine showed shifts in the fluorescence intensity of rhodamine 123 in a direction similar to that of atovaquone, indicating a possible mitochondrial effect (Fig. 10). Atovaquone produced 41.95% and 43.17% changes in mean fluorescence intensity at the IC50 and 10× IC50, respectively, emetine produced changes of 31.12% to 35.13% at the IC50 and 10× IC50, respectively, in a direction similar to that of atovaquone, and (−)-R,S-dehydroemetine produced changes of 26.76% to 32.98% at the IC50 and 10× IC50, respectively, in a direction similar to that of atovaquone (Fig. 10a).</p><!><p>Disruption of mitochondrial membrane potential. Changes in mitochondrial membrane potential observed on treatment with IC50s of atovaquone, emetine, and (−)-R,S-dehydroemetine. (a) Graphical representation of changes in fluorescence intensity on application of drugs at IC50s. (b) Table showing changes in fluorescence intensity (FITC-A mean) on application of drugs at IC50s and 10× IC50s. FITC-A mean value of infected blood without drugs was considered to be the base value (0), and FITC-A mean values of the compounds are represented as percentages of deviation from the base value.</p><!><p>The CalcuSyn method has been validated in our laboratory as a useful tool for antimalarial drug interaction analysis (38). To establish the robustness of the method, the known synergistic drug combination atovaquone and proguanil was used. Triplicate samples were analyzed after 72 h using the SYBR green-based fluorescence plate reader method. In accordance with published literature, CalcuSyn predicted strong synergism between the two drugs.</p><p>Following validation of the CalcuSyn software using the atovaquone-proguanil combination, the interactions of (−)-R,S-dehydroemetine–atovaquone and (−)-R,S-dehydroemetine–proguanil were studied. The doses used for each compound were based on the known 50% effective dose (ED50) values, which served as the midpoints for 2-fold, constant-ratio dose series as shown in Table 4.</p><!><p>Dose series used for the combination of existing antimalarials with (−)-R,S-dehydroemetine</p><p>The ratios for combination with (−)-R,S-dehydroemetine were 1:25 for atovaquone and 140:1 for proguanil.</p><!><p>The CalcuSyn-based analysis of the drug interactivity between atovaquone and proguanil was carried out using a constant-ratio combination of 1:7,000. The output included a dose-effect curve and a median-effect plot in addition to the combination index (CI) and an isobologram plot, to figuratively depict the compounds' potency and conformity to the mass action law (Fig. 11). Specifically, CI values of 0.20, 0.34, and 0.57 at the ED50, ED75, and ED90 levels (Table 5), respectively, were obtained, inferring strong synergism at the ED50 and synergism at the ED75 and ED90. Good correlation coefficients of the median-effect plot were reported for atovaquone (r = 0.99), proguanil (r = 0.89), and their combination (r = 0.93), inferring good conformity to the mass action law.</p><!><p>CalcuSyn-based drug interactivity analysis for atovaquone and proguanil. CalcuSyn-based median effect plot (a), isobologram (b), and dose-effect curve analyzed through GraphPad Prism (c) for drug interactivity between atovaquone and proguanil. The combination of atovaquone and proguanil (1:7,000) was found to be strongly synergistic at the IC50 (combination index [CI] = 0.21) and IC75 (CI = 0.34) and synergistic at the IC90 (CI = 0.57). fa, fraction affected; fu, fraction unaffected.</p><p>CalcuSyn-based drug interaction analysisa</p><p>The combination ratios and CalcuSyn determined the combination index (CI) values for the assays performed. The r value represents the linear correlation coefficient for the median effect plot and indicates conformity to the mass action law.</p><!><p>The CalcuSyn-based analysis of the drug interactivity between (−)-R,S-dehydroemetine and atovaquone was done using a constant-ratio combination of 25:1 (Fig. 12). Specifically, CI values of 0.88, 0.88, and 0.89 at the ED50, ED75, and ED90 levels, respectively, were obtained, inferring slight synergism at all measurement points (Table 5). Good correlation coefficients of the median-effect plot were reported for atovaquone (r = 0.94), (−)-R,S-dehydroemetine (r = 0.94), and the combination (r = 0.96).</p><!><p>CalcuSyn-based drug interactivity analysis for (−)-R,S-dehydroemetine–atovaquone combination. Dose-effect curve analyzed through GraphPad Prism for drug interactivity between (−)-R,S-dehydroemetine and atovaquone. The combination of atovaquone and (−)-R,S-dehydroemetine (1:25) was found to display slight synergy at the IC50 (CI = 0.88), IC75 (CI = 0.88), and IC90 (CI = 0.89).</p><!><p>The CalcuSyn-based analysis of the drug interactivity between (−)-R,S-dehydroemetine and proguanil was done using a constant-ratio combination of 140:1 (Fig. 13). Specifically, CI values of 0.67, 1.04, and 1.62 at the ED50, ED75, and ED90 levels, respectively, were obtained, implying synergism at the ED50, a nearly additive effect at the ED75, and antagonism at the ED90 (Table 5). Good correlation coefficients of the median-effect plot were obtained for proguanil (r = 0.90), (−)-R,S-dehydroemetine (r = 0.86), and the combination (r = 0.95), suggesting good conformity to the mass action law.</p><!><p>CalcuSyn drug interactivity analysis for (−)-R,S-dehydroemetine–proguanil combination. Dose-effect curve analyzed through GraphPad Prism for drug interactivity between proguanil and (−)-R,S-dehydroemetine. The combination of proguanil and (−)-R,S-dehydroemetine (140:1) was found to be synergistic at the IC50 (CI = 0.67), nearly additive at the IC75 (CI = 1.04), and antagonistic at the IC90 (CI = 1.62).</p><!><p>After 72 h of incubation, the (−)-R,S-dehydroemetine–atovaquone interaction was found to show slight synergism at all inhibitory levels analyzed. For the (−)-R,S-dehydroemetine–proguanil combination, the ED50 level of inhibition was classified as synergism, the ED75 as nearly additive, and the ED90 as antagonism.</p><!><p>We report here the nanomolar antimalarial efficacy of the synthetic emetine analogue (−)-R,S-dehydroemetine in the multidrug-resistant K1 strain of Plasmodium falciparum (IC50 of 71.03 ± 6.1 nM). The clinical use of emetine dihydrochloride as an antiamoebic drug was superseded by the better-tolerated synthetic analogue 2,3-dehydroemetine in the 1970s. The isomer (−)-S,S-dehydroisoemetine was found to be less potent, with an IC50 of 2.07 ± 0.26 μM. Molecular modeling suggests that the greater potency of (−)-R,S-dehydroemetine is linked to its ability to mimic the interactions of emetine with the E site of the Pf40S subunit more fully than the (−)-S,S-dehydroisoemetine isomer does. The improvements in the dose-dependent cardiotoxicity previously reported for 2,3-dehydroemetine in comparison to that of emetine dihydrochloride was linked to decreased affinity to cardiac myocytes and increased clearance from the body (27). We propose that the very significant differential in vitro activities of emetine dihydrochloride in Entamoeba histolytica (IC50 of 26.8 ± 1.27 μM) and Plasmodium falciparum (IC50 of 47 ± 2.1 nM) (13, 14) would mean that the dose-related toxicity profile for its repositioned use will be different.</p><p>Analysis done by GSK showed that in asexual blood stages, the compounds exhibited no cross-resistance issues. Mosquito vector dynamics, the number of people with peripheral gametocytemia in the population, and the infectiousness of circulating gametocytes to mosquitoes determine the transmission of malaria. Interrupting transmission is an important aspect of preventing malaria in areas of endemicity, and there has been renewed interest in compounds preventing the formation of gametocytes. (−)-R,S-dehydroemetine was found to be gametocidal. It displayed activity against both male and female gametes. Thus, it also has the potential to block the transmission of malaria.</p><p>The selectivity indices calculated using MTT assays in HepG2 cell lines were similar for emetine and its analogues. It is important to note that the low selectivity indices were expected, given the documented anticancer properties of emetine (51). This could explain the low LD50s obtained on HepG2 lines, as these are rapidly multiplying cancer cell lines. The selective advantage of (−)-R,S-dehydroemetine for cardiotoxic effects is linked to its faster elimination from the body, and hence, the appropriate toxicity investigation to define this would be the use of in vivo animal models. It is imperative to progress this work in this direction. It was also clearly established that the cardiac toxicity for (−)-R, S-dehydroemetine was not effected through the hERG channel. Our observations suggest that emetine and its two synthetic analogues produce changes in the mitochondrial membrane potential at their IC50s, indicating a possible multimodal mechanism of action.</p><p>The failure of monotherapy has been unambiguously demonstrated in malaria, and hence, there is forceful insistence by WHO on the use of artemisinin-based combination therapy (ACT) as a policy standard. In addition to more potent therapeutic efficacies, other benefits of combination regimes include decreased toxicity, favorable synergistic interactions, and most significantly, the potential to impede or delay the onset of resistance. The drug interaction analysis presented here emphasizes a route to further dose reduction to minimize toxicity. (−)-R,S-dehydroemetine was found to exhibit synergistic activity with proguanil and display slight synergism with atovaquone.</p><p>Emetine and its analogues constitute a pleiotropic group of natural product-derived compounds that could potentially enhance the depleted antimalarial armamentarium should a catastrophic gap in the drug market occur as a result of the spread of resistance to frontline antimalarials. The work presented here provides strong justification for further optimization, particularly for use in a hospital setting in cerebral malaria, where appropriate monitoring could obviate progression of the reversible cardiotoxicity previously reported. Their hepatic concentration ability could indeed have favorable consequences for the treatment of Plasmodium vivax infections. The history of antimalarial chemotherapy has been largely reliant on natural product-derived leads. We provide here a significant body of evidence to progress and further optimize a previously overlooked, potent, and affordable natural product compound.</p><!><p>RPMI 1640 medium containing 25 mM HEPES and 0.3 g/liter l-glutamine (Gibco; Life Technologies, UK) supplemented with 2.5 g sterile filtered AlbuMax (Sigma, UK), 2.5 ml hypoxanthine (Sigma, UK), 2.5 ml 40% glucose (anhydrous dextrose; Fisher Scientific, UK), and 0.5 ml gentamicin (Sigma, UK) was used for the culture of erythrocytic-stage strain K1 P. falciparum parasites (gifted by John Hyde, University of Manchester, United Kingdom; original source, Thai-K1 clone) under a 5% CO2, 5% O2, and 90% N2 gas mixture (BOC Limited, UK) at 37°C. All routine culture methods were consistent with those described in reference 52.</p><p>O positive human blood (purchased from NHS Blood Bank, Manchester, UK) was used routinely to maintain the parasites. Continuous cultures were maintained at 5% hematocrit. Since synchronous parasite development is observed in natural hosts and this synchrony is lost quite rapidly in in vitro cultures, sorbitol was used to keep the parasites in tight synchrony. In brief, the parasitized blood pellet was resuspended 1:10 in 5% sorbitol (prepared in distilled water and filtered using a 0.22-μm-porosity Millipore filter), incubated at room temperature for 5 min, and then centrifuged at 3,000 rpm for 5 min. The supernatant was removed, and complete medium was used to wash the pellet 3 times before setting up a new culture.</p><!><p>(−)-2,3-Dehydroemetine was synthesized according to the methods in the established literature (16, 53). The synthesis was outsourced to Chiroblock GMBH, Germany. Patent literature previously published by chemists at Hoffmann-La Roche did not include modern analytical data, such as mass spectrometry or nuclear magnetic resonance (NMR) (53). The methodology adopted for the synthesis of (−)-R,S-dehydroemetine (compound 7) used in this study followed the published patent, with minor modifications, as depicted in Fig. 14.</p><!><p>Synthesis of 2, 3-dehydroemetine. Conditions were as follows. (a) (i) Iodomethane, ethanol, room temperature (RT), 24 h, (ii) compound 2, KOAc, reflux, 3 h, 72%; (b) malononitrile, ammonium acetate/acetic acid, toluene, reflux, 2.5 h, 69%; (c) (i) 20% HCl, reflux, 5 h, (ii) methanolic HCl, RT, 18 h, 45%; (d) (+)-dibenzoyl tartrate, methanol, two recrystallizations, 25%; (e) (i) 3 M HCl, reflux, 90 min, (ii) 3,4-dimethoxyphenethylamine, xylenes, reflux, 18 h, (iii) POCl3, benzene, reflux, 1 h, 28%; (f) NaBH4, methanol, RT, 1 h, 81% of a 1:1 diastereomeric mixture. Compound 7, (S)-2-((R)-6,7-dimethoxy-1,2,3,4-tetrahydro-isoquinolin-1-ylmethyl)-3-ethyl-9,10-dimethoxy-1,6,7,11b-tetrahydro-4H-pyrido[2,1-a]isoquinoline * dihydrobromide. Compound 8, (S)-2-((S)-6,7-dimethoxy-1,2,3,4-tetrahydro-isoquinolin-1-ylmethyl)-3-ethyl-9,10-dimethoxy-1,6,7,11b-tetrahydro-4H-pyrido[2,1-a]isoquinoline * dihydrobromide.</p><!><p>The two compounds synthesized were (−)-R,S-dehydroemetine, compound 7, and (−)-S,S-dehydroisoemetine, compound 8. Briefly, the Mannich base, compound 1, was quaternized and reacted with the imine, compound 2, giving the piperidinone, compound 3, as an approximately 5:1 mixture of diastereomers that were not separated. Knoevenagel condensation with malononitrile led to unsaturated dinitrile, compound 4, in a 69% yield. Hydrolysis and decarboxylation, with concomitant alekene isomerization, gave compound 5 after methylation. Resolution with (+)-dibenzoyl-d-tartaric acid gave the homochiral compound, compound (−)-5. Hydrolysis and amidation with homoveratrylamine, followed by Bischler-Napieralski cyclization, led to a dihydroisoquinoline derivative, compound 6, also known as 2-dehydro-O-methyl psychotine. The cyclic imine was reduced with sodium borohydride in methanol to give a 1:1 mixture of the desired 2,3-dehydroemetine, compound 7, and 2-dehydroisoemetine dihydrobromide, compound 8, separated by fractional crystallization. The purity of (−)-R,S-dehydroemetine obtained through this method was 85%, with the balance consisting of the diastereomer, compound 8.</p><!><p>In preparation for predicting the bound poses of emetine, (−)-R,S-dehydroemetine, and (−)-S,S-dehydroisoemetine, the cryo-EM structure of the P. falciparum 80S ribosome bound to emetine dihydrochloride was obtained from the Protein Data Bank (PDB code 3J7A) (16). We note that this cryo-EM structure features the enantiomer of emetine rather than emetine itself; therefore, an initial three-dimensional (3-D) structure of emetine was obtained from the OMEGA-generated conformer in the PubChem database (54). Subsequently, emetine derivatives (−)-R,S-dehydroemetine and (−)-S,S-dehydroisoemetine were constructed using the software package MOE (55). Prior to in silico docking, crystallographic water molecules and a magnesium ion were removed (the latter was within 13 Å of the active site and was found to unduly influence docking solutions). Hydrogens were assigned consistent with physiological pH. Docking of ligands to the Pf40S subunit was performed using MOE-Dock (2016; Chemical Computing Group, Inc., Montreal, QC, Canada). The binding site region was defined using the cryo-EM ligand location. Ligand placement used the Triangle Matcher protocol. Internal flexibility was permitted for ligands but not ribosome. Poses were scored via the London dG scoring function. While this docking protocol was able to reproduce well the cryo-EM pose of the emetine stereoisomer bound to the Pf40S subunit (not shown), please see the comment in Results regarding this ligand's structure.</p><!><p>Emetine dihydrochloride was obtained from Sigma-Aldrich, UK. The dehydroemetine analogues were synthesized as described in the established literature (16; scheme 1 in reference 53). Derivatives were validated using NMR, high-performance liquid chromatography (HPLC), and Fourier transform infrared spectroscopy (FTIR). The synthesis process was outsourced to Chiroblock GMBH, Germany. Drug stock solutions were prepared in dimethyl sulfoxide (DMSO) in accordance with the manufacturer's instructions, and the primary stock concentration of 5 mM was aliquoted and stored at –20°C until further use. Serial dilutions of the working solution using complete medium were made for the experiments.</p><!><p>Refined dose ranges were selected for emetine dihydrochloride (12.5 nM to 200 nM), (−)-R,S-dehydroemetine (12.5 nM to 200 nM), and (−)-S,S-dehydroisoemetine (0.625 μM to 10 μM) to permit accurate calculation of their IC50s against the K1 strain of P. falciparum. Ring stage parasites were diluted to 0.5 to 1% parasitemia, 2.5% hematocrit (in a 96-well-plate format, 200-μl final well volume) and treated for 72 h. Dose-response parasitemia was determined using the SYBR green-based flow cytometer or plate reader method previously optimized in the laboratory (14).</p><!><p>Following drug efficacy experiments, 150-μl amounts from the control and drug-treated wells on a 96-well plate were transferred to microcentrifuge tubes, and samples were washed once with phosphate-buffered saline (PBS). The supernatant was removed, and samples were incubated in the dark at room temperature for 20 min following the addition of 1 ml of 5× SYBR green solution (prepared by adding 5 μl of 10,000× SYBR green to 10 ml PBS). After staining, the samples were centrifuged for 1 min at 14,000 rpm and the supernatant was discarded. Samples were resuspended in 250 μl of 0.37% formaldehyde fixation solution prepared by diluting 36.5% formaldehyde (Sigma, UK) with PBS to the specified final concentration. The samples were placed in the refrigerator and incubated at 4°C for 15 min. Subsequently, PBS was used to wash the samples 3 times, and finally, the samples were suspended in 1 ml PBS. Parasitemia was determined by measuring SYBR green fluorescence using the FITC channel of the BD FACSVerse flow cytometer system (blue laser; excitation laser line, 488 nm, with excitation maximum of 494 nm and emission maximum of 520 nM) and cell size (forward scatter [FSC-A]). Fifty thousand events were recorded, in three replicates for each. Fluorescence events in drug-treated samples were compared with those in their infected and uninfected blood counterparts and gated accordingly to obtain the percentages of parasitemia.</p><!><p>The IC50 speed assay was used to determine the speeds of action of the emetine analogues. Unsynchronized cultures of K1 strain P. falciparum were used, and parasites were grown in the presence of (−)-R,S-dehydroemetine and (−)-S,S-dehydroisoemetine for three incubation periods of 24 h, 48 h, and 72 h. The assays were analyzed by determining SYBR green fluorescence as described above.</p><!><p>Parasite cultures with ≥80% trophozoites and ≥80% rings were obtained by synchronization with 5% sorbitol (42). The cultures were synchronized twice, at 0 h and 31 h, to obtain young rings which were up to 3 h old. To obtain early schizont stages, the cultures were synchronized twice, with the second synchronization being 6 to 8 h after the first. Each synchronous stage was incubated for 24 h at 37°C on a 96-well microtiter plate with a 2-fold serial dilution of the drugs ranging from 1.6- to 100-fold IC50 of each drug. The plates were washed 4 times after the incubation in order to dilute the drug concentration by >1,000-fold. The plates were incubated for another 24 h at 37°C, following which SYBR green staining was used to read the plates as described above.</p><!><p>HepG2 cells (University of Salford stocks, purchased from American Type Culture Collection [ATCC], USA), were grown in medium consisting of RPMI 1640, 2 mM l-glutamine, HEPES, 10% fetal bovine serum, 1% nonessential amino acids, and 1% penicillin-streptomycin. Cells were calculated on a hemocytometer to make a new cell-medium solution to a concentration of 5,000 cells/100 μl, and 100 μl of this new cell solution was added into each well of the 96-well plate. After 24 h of incubation, 100-μl amounts of the drug prepared in the medium were added to the wells. This was repeated in triplicate, and the plate was incubated for a further 48 h (56). After the incubation period, 50 μl of MTT solution was added to each well and incubated for 3 h. The liquid in each well was aspirated carefully, 200 μl of DMSO was added to each well, and the results were read on the Ascent plate reader.</p><!><p>The results for the infected-blood controls were set at 100% for normalization, and percentages of parasitemia for drug-treated samples were calculated relative to the results for the infected controls. For IC50 values, GraphPad Prism 5.0 was used to further process the data. IC50 values were calculated using nonlinear regression (GraphPad Prism 5.0) by using log-transformed drug concentrations plotted against the dose response. The log(inhibitor) versus normalized response-variable slope option was used for IC50 calculation. Emetine dihydrochloride was used as a control drug to validate the method.</p><!><p>The three strains used for the cross-resistance assay were 3D7A, Dd2, and W2, based on their resistance profiles; the assay was carried out by GSK, Tres Cantos. Strain 3D7A is chloroquine sensitive, strain Dd2 is resistant to chloroquine, mefloquine, and pyrimethamine (57), and strain W2 is resistant to chloroquine, quinine, pyrimethamine, cycloguanil, and sulfadoxine (from the Malaria Research and Reference Reagent Resource Center [MR4]) (58). A culture of red blood cells (RBCs) parasitized by the corresponding strain (0.5% parasitemia, 2% hematocrit) in RPMI 1640, 5% AlbuMax, and 5 μM hypoxanthine was exposed to 9 dilutions (3-fold serial dilutions) of the compound starting at 5 μM. One hundred microliters of culture volume was plated in 96-well flat-bottom microtiter plates with 0.5 μl drug (200× stock in DMSO). Plates were incubated for 24 h at 37°C, 5% CO2, 5% O2, 90% N2. Next, 3H-hypoxanthine was added and plates were incubated for another 24-h period. After that, parasites were harvested on a glass fiber filter using a TomTec cell harvester 96. Filters were dried, and melt-on scintillator sheets were used to determine the incorporation of 3H-hypoxanthine. Radioactivity was measured using a MicroBeta counter. Data were normalized by incorporating the results for the positive control (parasitized red blood cells without drug). IC50 values were determined using the Grafit 7 program (41).</p><!><p>Asexual cultures of P. falciparum strain NF54 parasites were used to seed gametocyte cultures at 0.5% parasitemia, 4% hematocrit in a 50-ml total volume under 3% O2, 5% CO2, 92% N2 gas. Culture medium (RPMI, 25 mM HEPES, 50 mg/liter hypoxanthine, 2 g/liter NaHCO3 without l-glutamine plus 5% human serum and 5% AlbuMax) was replaced daily for 14 days. At day 14, the concentration of nonpurified cultures was adjusted to plate 700,000 total cells per well in each 384-well plate. The test drugs were then added in 2-fold serial dilutions starting at 10 μM (10 μM to 0.00976 μM) and incubated for 48 h at 37°C (3% O2, 5% CO2, 92% N2). DMSO was used as the negative control, and thiostrepton as the positive control. Activation was performed with ookinete medium (same RPMI base used for culture but supplemented with 50 μM xanthurenic acid) supplemented with anti-Pfs25-Cy3 antibody at a final concentration of 1/2,000 (from 1 mg/ml stock). Plates were analyzed to detect exflagellation centers. "Triggered" cultures were then incubated (protected from light) at 26°C for 24 h (in a thermoregulated incubator). Then, plates were analyzed to detect female activated gametes. Activation of male gametes was detected based on light changes provoked by flagellar movements that caused movement of surrounding cells. A 10-frame video was taken and then analyzed to determine these changes in cell position based on changes in pixels. Then, the script determined where exflagellation centers were located, also based on the size and intensity of light changes. Activation of female gametes was based on detection of fluorescent Cy3-anti Pfs25 antibody (as the primary parameter), followed by a selection of events according to their size and roundness and the intensity of the fluorescence. Both measurements were performed using an automated inverted Ti-E Nikon microscope and JOBS software. Analysis of images and videos was performed with the ICY program. The IC50 values were determined using Microsoft Excel and GraphPad.</p><!><p>The experiment to test the potential of emetine, (−)-R,S-dehydroemetine, and (−)-S,S-dehydroisoemetine to inhibit the hERG channel was outsourced to Cyprotex, UK. One-hundred-microliter amounts of a 20 mM concentration of each of the three compounds in DMSO was provided to Cyprotex. Compound dilutions were prepared by diluting a DMSO solution (10 mM default) of the test compound into DMSO using a 5-fold dilution scheme, followed by dilution into extracellular buffer such that the final concentrations tested were typically 0.008, 0.04, 0.2, 1, 5, and 25 μM (final DMSO concentration, 0.25%). Chinese Hamster ovary cells expressing the hERG potassium channel were dispensed into 384-well planar arrays, and hERG tail currents were measured by whole-cell voltage clamping. Amphotericin B was used as a perforating agent that was circulated underneath the PatchPlate to gain electrical access to the cell. The pre-compound-application hERG current was measured. Emetine, (−)-R,S-dehydroemetine, and (−)-S,S-dehydroisoemetine in ranges of concentrations were then added to the cells, and a second recording of the hERG current was made. The test compound was left in contact with the cells for 300 s before currents were recorded. Quinidine, an established hERG inhibitor, was included as a positive control, and a vehicle control (0.25% DMSO) as the negative control. All buffers, cell suspensions, and drug compound solutions were at room temperature. The percentage of change in hERG current was measured and used to calculate an IC50 value.</p><p>Each concentration was tested in 4 replicate wells on the PatchPlate (maximum of 24 data points). Filters were applied to ensure that only acceptable cells were used to assess hERG inhibition. The cell must maintain a seal resistance of greater than 50 MΩ and a pre-compound-application current of at least 0.1 nA and ensure cell stability between pre- and post-compound-application measurements.</p><!><p>Rhodamine 123 and Draq5 were used to observe the effects of emetine and its two synthetic analogues on mitochondrial membrane potential. Rhodamine 123 is a mitochondrial-specific dye which emits in the FITC channel in flow cytometry, whereas Draq5 stains the DNA and emits in the APC-Cy7-A channel. These two dyes were chosen to perform the experiment as there is very minimal overlap in their emission signals. A synchronized culture of P. falciparum strain K1 at the trophozoite stage was incubated with 1 ml of 200 nM rhodamine 123 for 1 h, followed by a further incubation with 100 μl of 10 μM Draq5 for 20 min. Smears prepared after the incubation were immediately viewed under the fluorescence microscope at ×100 magnification to visualize the localization of the dyes inside the parasite.</p><p>To test the effects of emetine and its analogue on mitochondria, synchronized cultures of P. falciparum strain K1 at the trophozoite stage were incubated at 2.5% hematocrit (in a 96-well-plate format with a final well volume of 200 μl) for 2 h with IC50s and 10× IC50s of atovaquone, emetine hydrochloride, and (−)-R,S-dehydroemetine. Compounds were then washed away by centrifuging the cultures, and a pellet was prepared for each drug concentration. Pellets were then incubated with 1 ml of 200 nM rhodamine 123 for 1 h, followed by a further incubation with 100 μl of 5 μM Draq5 for 20 min. After a wash with PBS, the experiment was read with a flow cytometer on the FITC channel. Loss of mitochondrial membrane potential was indicated by a decrease in fluorescence intensity.</p><!><p>Primary stock solutions of (−)-R,S-dehydroemetine were prepared as described above. For the experimental setup, the primary stock solutions were further diluted with complete medium to give final test concentrations. A dose range of 0.125 to 8 times the ED50 was made by 2-fold serial dilutions for atovaquone, proguanil, and (−)-R,S-dehydroemetine. For atovaquone, the doses ranged from 0.25 nM to 16 nM for combination with proguanil and from 0.5 nM to 32 nM for combination with (−)-R,S-dehydroemetine. For proguanil, the doses ranged from 1.75 μM to 112 μM. For (−)-R,S-dehydroemetine, the doses ranged from 12.5 nM to 800 nM. At each level, the compounds were coadministered; for example, the ED50 of atovaquone was combined with the ED50 of (−)-R,S-dehydroemetine, 2 times the ED50 of atovaquone was combined with 2 times the ED50 of (−)-R,S-dehydroemetine, and so forth. Parasites were treated at ring stage and incubated for 72 h in a 96-well-plate format. The SYBR green-based plate reader method was used to determine drug susceptibility. The data were analyzed for the median effect using CalcuSyn software (Biosoft) by converting triplicate data to an averaged percentage. The r values are also reported for all sets of data. The r value represents the linear correlation coefficient for the median-effect plot and indicates conformity to the mass action law. The CalcuSyn software generates the CI over a range of fraction-affected (fa) levels at different growth inhibition percentages. The interpretation of CI was done in accordance with the classification presented in Table 6 (37).</p><!><p>Classification of synergism or antagonism using CI values generated by the Chou-Talalay method (CalcuSyn manual, Biosoft [38])</p><!><p>For routine malaria culture, anonymized whole-blood packs deemed unfit/outdated for clinical use were purchased from the NHS Blood Bank at Plymouth Grove, Manchester, UK. For experiments carried out in GlaxoSmithKline, Diseases of the Developing World Medicines Development Campus, Tres Cantos, Spain, the human biological samples were sourced ethically and their research use was in accord with the terms of the informed consents under an IRB/EC-approved protocol.</p>

PubMed Open Access

Photolysis of dimethoxynitrobenzyl-“caged” acids yields fluorescent products

Carboxylic acids conjugated with 4,5-dimethoxy-2-nitrobenzyl photoremovable protecting group are well known and widely used for biological studies. In this paper, we study the photolysis of likewise "caged" acetic, caprylic and arachidonic acids. Unexpectedly, we observed huge growth of fluorescence emission at ~430 nm during photolysis. Following further UV irradiation, a product with fluorescence at longer wavelength was formed (470 nm excitation / ~500-600 nm emission). While it may be used to monitor the "uncaging", these fluorescent products may interfere with widespread dyes such as fluorescein in biomedical experiments. This effect might be negligible if the photolysis products dissolve in the medium. On the other hand, we observed that arachidonic and caprylic acids derivatives self-organize in emulsion droplets in water environment due to long lipophilic chains. Illumination of droplets by UV rapidly induces orange fluorescence excited by 488 nm light. This fluorescence turn-on was fast (~0.1 s) and apparently caused by the accumulation of water-insoluble fluorescent residuals inside droplets. These self-organized lipophilic structures with fluorescence turn-on capability may be of interest for biomedical and other application. We have identified and hypothesized some compounds which may be responsible for the observed fluorescense.

photolysis_of_dimethoxynitrobenzyl-“caged”_acids_yields_fluorescent_products

<!>Results and Discussions<!>Conclusion<!>Methods

<p>Light is used in different aspects of our life, from natural photosynthesis and to artificial photonic networks. All these processes rely on compounds that absorb light and transform its energy to the desired effect, including electromotive force [1][2][3] , chemical reactions 4 and luminescence 5,6 . Application of photoremovable protecting groups to control the activity of biological molecules has been widely used and extensively studied [7][8][9][10] . It allows one to use light to "uncage" the molecule and rapidly induce the desired effect. For instance, lipo-or amphiphilic molecules conjugated with nitrobenzyl-based moieties have been used for preparation of photosensitive liposomes 11,12 . In such self-organized structures, light may be used to modify the properties of particles, for instance, destabilize the surface inducing cargo release. On the other hand, fatty acids are important messengers in biology, especially arachidonic acid. "Caged" derivatives of arachidonic acid were reported previously 13 , including that with water-soluble protective group 14 .</p><p>The ability of liquid hydrophobic substances to make emulsions in water is of great importance for life and technology. Emulsion contains dispersed phase consisting of small droplets or -in special cases -microdroplets, micelles or liposomes. Thus, water-insoluble molecules may be transferred and used in a water environment; a perfect example is dairy fat in milk and, more generally, fat digestion in a body. The dispersed phase may be used for delivery of chemicals, for instance, with diagnostic and therapeutics purposes. This possibility has attracted a lot of attention in recent years [15][16][17][18] . Stimuli-responsive particles like micelles or liposomes are especially promising since they allow the spatiotemporal control of their action 19 .</p><p>In the present work, we describe arachidonic and caprylic acids conjugated with 4,5-dimethoxy-2-n itrobenzyl-based protecting group (so called "dimethoxynitrobenzyl-caged" or "DMNB-caged"). We show that these compounds form emulsions in water (with small fraction of DMSO), in contrast to acetic acid modified in the same way. Apparently, photolysis products accumulate inside these self-organized lipophilic droplets. Surprisingly, we observed that the droplets become highly fluorescent after ~0.1 s illumination by UV LED.</p><p>The fluorescence of DMNB-"caged" compounds residuals after photolysis was observed in previous studies 20,21 , but it was attributed to the release of fluorescent target compound and used as a measure of "uncaging" degree. Our study shows that this approach could be a source of systematical errors. The majority of studies focus on the released ("uncaged") molecule, while the properties of cleaved protective group are usually out of scope, with notable exception of 22 . Little is known of the primary photoproduct 2-nitroso-4,5-dimethoxybenzaldehyde and its subsequent derivatives. To our knowledge, all known properties are that it is potentially toxic 23 and is considered non-fluorescent 13 , although quenches the fluorescence of coumarin 343 24 . In the paper, we confirm and study the fluorescence of the photoproducts, both in droplets and in a solution. It may be used to detect the "uncaging", but also may interfere with widespread dyes such as fluorescein isothiocyanate (FITC) in biomedical experiments. On the other hand, the possibility to switch on fluorescence might be useful in imaging applications 25 .</p><!><p>Figure 1 shows compounds I-III which we synthesized as described in supplementary materials and used in experiments. All compounds were dissolved in DMSO to the concentration of ~10 mM (stock solution).</p><p>Photolysis mechanism of compounds conjugated with nitrobenzyl-based protecting groups has been described 26,27 . Although it may depend on the solvent and pH, generally two products are formed: the "uncaged" compound and the "byproduct", or protecting group residual. The naive scheme for carboxylic acids is shown in Fig. 2, but, according to our results, it is only a part of the whole picture. Aromatic nitroso compounds could undergo photochemical transformations leading wide range of products 28,29 .</p><p>We first performed the "uncaging" experiment for 100 µM of Ia in DMSO. Figure 3A shows changes in absorption spectra during UV illumination as described in methods (365 nm, ~1 W). These changes almost stopped after 4-5 min., indicating the complete photolysis of the sample. We also measured fluorescence emission spectra upon 355 nm excitation wavelength and observed huge growth in fluorescence intensity following the UV illumination (Fig. 3B). The intensity at maximum (430 nm) has grown 150-fold after 4 min. of illumination. Interestingly, the growth was faster than linear, as shown in the inset, which means that formation of fluorescent product involves further reactions, e.g. dimerization (typical process for aromatic nitroso compounds like IV 30 ).</p><p>Further illumination caused the decrease of fluorescence at 430 nm. Absorption spectra did not change significantly during this process. The decrease was accompanied by the appearance of orange fluorescence excited by 470 nm light. Figure 3C shows the corresponding spectra and kinetics of the maximum value at 550 nm.</p><p>The fluorescence during photolysis of fatty acids II and III in DMSO follow the same pattern as Ia, although have different time course and maximal values (Fig. 4). Interestingly, the initial fluorescence growth is the same for all three compounds.</p><p>The kinetics of further decay of short-wavelength fluorescence and appearance of long-wavelength one depends on the acid used. It seems that free acid influences the ratio of different products, which the protective group forms after the photolysis. However, the normalized emission spectra of the ultimate product coincide exactly (Fig. S3). It shows that this fluorescent compound is likely the same for all three DMNB-based compounds. The compound thus should consist of several protective group residuals, either the same or different.</p><p>We can conclude that compound IV under our experimental conditions transforms into highly fluorescent derivative (355 excitation/430 emission). This derivative is subsequently converted by light to a form with longer-wavelength emission. Interestingly, the transition between two forms depends on concentration: in a S2 shows the appearance of fluorescent product in a cuvette. To avoid possible artifacts and contamination, we reproduced this effect in several independent experiments, using both quartz and disposable plastic cuvettes, as well as a solvent from different sources.</p><p>Further, we tried to identify the fluorescent products. The in situ NMR study of reaction mixtures was performed simultaneously with fluorescence measurements using DMSO-d 6 as a solvent and the only identified product was free acid (Fig. S1). The residual of protective group seems to form tens of different products with too little amount to identify by NMR. This was confirmed by GC-MS studies of reaction mixture after irradiation of compound II. However, we were able to identify 2-hydroxy-4,5-dimethoxybenzaldehyde and its o-caprilyc derivative (Fig. S5). Despite of these molecules has no data on fluorescence, their close analogs such as vanillin and 4-hydroxy-3,5-dimethoxybenzaldehyde has emission at 380-420 nm in polar solvents 31 , so we may hypothesize that these substances are responsible for shortwave emission. Presence of the acid residue in molecule with M r = 308 explains the differences in graphs in Fig. 4. Indeed, the "uncaged" acid participates in chemical transformations of the protective group. Longer wavelength fluorescence may appear due to formation of dimeric products from 4,5-dimethoxy-2-nitrosobenzaldehyde under irradiation. Indeed, nitrozobenzenes are known to undergo photochemical dimerization with further transformations 29 to form benzo[c]cinnoline derivatives, which has fluorescence maxima around 500 nm 32 (Fig. S5). 3-Hydroxy-4,5-dimethoxybenzaldehyde also could be engaged in photochemical oxidative dimerisation similar to vanillin 33 . In GC-MS spectra, we observed substances with molecular weights of 356 and 359 which may correspond to such dimeric products. In previous studies 20,21 , the increase of fluorescence intensity was explained by the release of fluorescent target compound and was used as a measure of "uncaging" degree. Our study shows that this approach could be a source of systematical errors.</p><p>To test whether this effect is unique for the protective group with methoxy substituents, we synthesized compound Ib, which bears nitrobenzyl group. The photolysis was performed by LEDs with 340 nm peak wavelength. Interestingly, we observed similar growth of short-wavelength fluorescence with emission spectrum shifted to the left (Fig. S4). Thus, the short-wavelength fluorescent product retains some properties of the protective group, including the absorption region. It supports the idea that this fluorescence is a discharge of energy absorbed by a residual of protective group, which no longer can be spent on the dissociation. In this case the residual is 2-nitrosobenzaldehyde, which is known to be fluorescent 34 , or its derivatives.</p><p>Kinetics of fluorescence intensity at 430 nm is shown in Fig. 4 (dashed line). In contrast to DMNB-based compounds, there is no acceleration of growth for Ib. The formation of long-wavelength fluorescent product is almost absent. We can conclude that the reactions which yield that product are much more efficient and fast in case of DMNB protective group. www.nature.com/scientificreports www.nature.com/scientificreports/ Usage in biological systems implies water environment, and we conducted separate experiments in phosphate buffered saline (pH 7.3). Stock solutions of I-III were gently mixed with ten-fold volume of PBS. For II and III we immediately observed characteristic milky appearance of mixture indicating the formation of emulsion. In contrast, acetic acid Ia dissolves well in the medium, which shows that self-organization into droplets occurs due to long lipophilic chains of fatty acids.</p><p>Light microscopy of emulsions showed round droplets with sizes in a range of 1-10 µm. We have not tried to obtain more narrow size distribution, although we expect that it can be easily done with more extensive mixing or using ultrasound. The droplets showed no fluorescence, but started to glow with orange color after 1-second pulse of 365 nm UV light without any background fluorescence.</p><p>To study this phenomenon and relate it with our photolysis experiments in DMSO, we illuminated the droplets by 0.1 s UV flashes every 0.5 s and recorded the fluorescence intensity. Figure 5A-D shows images of droplets of III before illumination, after 4, 11 and 50 flashes. As particles were moving, we used the TrackMate FIJI plugin 35 to measure the kinetics of fluorescence intensity for each particle. Figure 5E shows mean fluorescence value for all detected particles, Fig. 5F -the total number of detected particles over time and Fig. 5G shows kinetics of fluorescence for three single particles.</p><p>One can see that fluorescence started to grow immediately after the first flash, which indicates that all reaction stages are much faster in this case. We observed similar effect for compound II. The incredibly fast turn-on and the absence of background lead to the conclusion that photodecomposition product IV stays inside droplets after the photolysis, which is consistent with its water-insoluble structure. Its local concentration is much higher than that used in our photolysis experiments: it might be that of stock solution or even higher as the solvent (DMSO) may go outside the droplet. The fluorescence in droplets was stable for hours; after drying for 3 days, fluorescent fractal-like structure was formed, which is shown in Fig. S6.</p><!><p>In conclusion, we have synthesized two "caged" fatty acids by conjugation with dimetoxynitrobenzol protecting group. The compounds showed efficient decomposition under UV light (365 nm). We also for the first time report fluorescent properties of the protecting group residuals after photolysis. Besides the fluorescence excited by ~355 nm, which is also present for usual nitrobenzol protecting group, DMNB version produces some compound whose fluorescence is significant under 488 nm excitation. This unknown compound would interfere with widespread dyes such as FITC and Green Fluorescent Protein in biomedical experiments, which should be kept in mind when using dimethoxynitrobenzyl protecting group.</p><p>Next, we showed that our fatty acids derivatives self-organize in emulsion droplets in water environment due to long lipophilic chains. Finally, the illumination of droplets by UV light induces fast fluorescence turn-on due to the accumulation of protecting group residuals inside. We have not identified the specific fluorescent compound, leaving it for the future research. However, such self-organized particles (droplets) may be of interest for biomedical and other applications 25 . For instance, a system where such particles are "turned on" by passing the laser beam may be used to trace hydrodynamic flows. More generally, transport of such particles from the particular place marked by a pulse of light can be studied in detail, including intracellular pathways.</p><!><p>Synthesis of compounds I-III is described in supplementary materials.</p><p>The photolysis ("uncaging") experiments were carried out in fluorimeter quartz cuvette with 10 × 10 mm light path. Compounds were dissolved in 3 mL of DMSO to the final concentration of 100 µM. We prepared a setup consisting of two UV LEDs with 365 nm central wavelength and 480 mW optical power. Another LEDs with peak wavelength of 340 nm were used in experiments with compound Ib. A microcontroller was used to set the illumination time, which was in a range from 0.1 s to 5 min in different experiments. After each session of illumination we measured UV-VIS absorption spectrum of the sample using Shimadzu UV-1900 spectrophotometer and fluorescence excitation/emission spectra with Shimadzu RF-6000 fluorimeter. Although we performed one measurement for each time point to avoid delays, all the described effects were reproduced in several independent experimental series.</p><p>Microscopic studies were conducted with Carl Zeiss AxioVert A1 with 450-490 excitation and >515 emission filters. All experiments were recorded using AxioCam 503 monochromatic high-sensitive camera with 3x analog gain and exposure time of 0.1 s. Illumination of samples was performed in place of observation with the same UV LED used in cuvette photolysis experiments. TrackMate FIJI plugin 35 was used to measure the kinetics of fluorescence intensity for particles. We used LoG detector with the following parameters: estimated diameter 20 pixels, threshold 0.05.</p><p>NMR studies were performed on Brucker AV-500 spectrometer. DMSO-d6 was used as a solvent.</p>

Scientific Reports - Nature

Optimizing optogenetic constructs for control over signaling and cell behaviors

Optogenetic tools have recently been developed that enable dynamic control over the activities of select signaling proteins. They provide the unique ability to rapidly turn signaling events on or off with subcellular control in living cells and organisms. This capability is leading to new insights into how the spatial and temporal coordination of signaling events governs dynamic cell behaviors such as migration and neurite outgrowth. These tools can also be used to dissect a protein\xe2\x80\x99s signaling functions at different organelles. Here we review the properties of photoreceptors from diverse organisms that have been leveraged to control signaling in mammalian cells. We emphasize recent engineering approaches that have been used to create optogenetic constructs with optimized spectral, kinetic, and signaling properties for controlling cell behaviors.

optimizing_optogenetic_constructs_for_control_over_signaling_and_cell_behaviors

Introduction<!>Naturally occurring light activated signaling proteins<!>Light-activated adenylyl cyclases<!>G protein coupled opsins<!>Achieving optical control of a desired signaling output<!>Light induced heterodimerization \xe2\x80\x93 controlling a protein\xe2\x80\x99s proximity to its effectors<!>Light dependent unmasking of proteins and peptides<!>Optically controlled association and dissociation of protein clusters<!>Optimization of optogenetic constructs<!>Tuning the masking characteristics of LOV domains<!>Engineering improved light-induced heterodimerization schemes<!>Spectral tuning<!>Conclusions

<p>Photoreceptor proteins mediate light-dependent responses in organisms ranging from bacteria and algae to plants and mammals. Two features of this diverse class of proteins have made it possible to engineer novel genetically encoded constructs to control a wide range of biochemical processes in intact cells using light. First, photoreceptors are often found to retain their light-dependent signaling capabilities when expressed in different organisms and cell types. Second, light-responsive domains often act in a modular fashion, which has enabled their integration into engineered constructs that retain light sensitivity and incorporate novel signaling outputs.</p><p>Recent reviews have focused on optogenetic tools based on photoreceptors such as opsins, phytochromes, cryptochromes, and light-oxgen-voltage sensing (LOV) domains used to control signaling activities with light and the biological insights that these tools can provide 1-3. Here, we focus on how optical modulators of cell signaling have been designed by selecting light sensitive proteins and recruiting their unique properties such as activation spectra, light-sensitivity, kinetics and signaling characteristics to control cellular responses.</p><!><p>Some signaling functions can be controlled using naturally occurring light-sensitive versions of enzymes or transmembrane receptors. In this section we focus on light activated adenylyl cyclases and G protein coupled opsins. Both offer examples of light-sensitive proteins that can be used in a variety of organisms and cell types to achieve subcellular control over signaling events. They also illustrate how light-sensing and signaling domains can be combined in new ways to achieve desired spectral, kinetic and signaling properties.</p><!><p>Photoactivated adenylyl cyclases from the flagellate Euglena gracilis (euPAC), and the soil bacterium Beggiatoa sp. (bPAC) have been shown to be capable of generating cAMP in mammalian cells and model organisms in response to blue light 4-9. bPAC is advantageous because of its smaller size (350 amino acids), lower dark activity and larger light-induced increase in activity 7. Its cyclase activity increases >100-fold in response to blue light and decays within 20s upon remove of the optical stimulus 7. bPAC encodes an N-terminal BLUF (blue light receptor using flavin adenine dinucleotide) domain, and a C-terminal adenylyl cyclase and functions as a homodimer. Understanding the mechanism of photoactivation remains an active field of study 10, 11. An LOV regulated adenylyl cyclase was recently discovered from the cyanobacterium Microcoleus chthonoplastes PCC 7420 (mPAC) 12. Compared to bPAC, mPAC exhibited higher constitutive activity as well as higher light induced activity 12.</p><p>Subcellular compartmentalization and temporal control of second messengers such as cAMP are thought to be central to how this ubiquitous second messenger achieves diverse cellular responses that are specific to different signaling inputs. For example, some G protein coupled receptors (GPCRs) are now thought to be capable of generating cAMP both at the plasma membrane and at endosomes 13. Light activated adenylyl cyclases are enabling new insights into this regulation. By fusing bPac with domains that targeted it to either the plasma membrane or endosomes, cAMP production was optically stimluated at these sites, showing distinct patterns of changes in gene expression 14.</p><!><p>G protein coupled receptors (GPCRs) regulate diverse cellular behaviors and physiological responses. GPCRs are therefore the most important target for therapeutic drugs. Methods that allow optical control over these pathways can be valuable both for probing the mechanistic basis of these pathways and identifying their roles in regulating cell physiology. Optical control can be especially useful if it allows spatially restricted regions of a single cell to be activated. This can help create asymmetric signaling activity and direct polarized cell behavior such as migration, asymmetric cell division and neurite outgrowth. Since light can be switched on and off almost instantaneously, temporal control over signaling is also potentially possible with appropriate optical triggers.</p><p>G protein coupled opsins possess unique properties that make them excellent candidates for acting as such optical triggers 2, 15, 16. They are spectrally distinct and coupled to different G protein subunits with contrasting effects on second messengers. They can be used in the native form without the introduction of a light sensing domain. Their wavelength selectivity allows intracellular signaling activity to be imaged using fluorescent proteins of spectrally distinct excitability. Opsins sense the extracellular signal first, so it is possible to measure the activity of all signaling molecules downstream. Responses to opsin activation are mediated by endogenous molecules thus maintaining the molecular integrity of the cell. Importantly, color opsin deactivation and recovery occur rapidly in contrast to rhodopsin 17 and since they diffuse relatively slowly along the plasma membrane activated areas are restricted and temporal control is acute.</p><p>Since chimeric forms of different GPCRs in which the extracellular and intracellular loops have been swapped are functional, it is possible to obtain opsins with residues and domains swapped to create novel combinations of spectral sensitivity and second messenger effects. Distinct opsins of this nature can be used to control multiple signaling pathways with different wavelengths of light. For example, jellyfish opsin normally optimally senses green light but a chimera has been developed in which the extracellular loops have been substituted with human blue opsin so that cAMP increase is induced on sensing blue light 15. A Gs coupled blue sensing opsin can be used with fluorescent proteins spanning a larger part of the visual spectrum.</p><p>Opsins require 11-cis-retinal as a cofactor for light activation. Rhodopsin appears to be capable of functioning in cell lines and in the whole brain without the need for exogenous 11-cis-retinal 18, 19. In contrast, cone opsins require 11-cis-retinal addition to function in cell lines 15 but are able to show activity in the brain with endogenous retinal19. When using opsins to regulate signaling and cell behavior they are repeatedly exposed to pulses of light. Rhodopsin rapidly loses the capability to respond to subsequent light pulses 20, and this process is found to be slower in the case of color opsins 15. A valuable property of some opsins -- bistability 21 -- can be recruited to overcome this deficit. For example, OPN3 is a bistable opsin that is capable of being activated and deactivated with two distinctly different wavelengths of light without progressive inactivation 22. This opsin can be activated repeatedly without the requirement of endogenous or exogenous 11-cis retinal which can be valuable for extended periods of optical control. Alternatively, the light activated photoisomerase RGR converts all-trans-retinal to 11-cis-retinal and might be used to replenish 11-cis-retinal levels in experiments requiring prolonged opsin activation 23.</p><!><p>For most signaling proteins, naturally occurring light-sensitive versions are unavailable. However, optical control over several such proteins has now been achieved using genetically encoded constructs that combine a light-sensitive input domain with a signaling output domain from the protein of interest. Light-induced conformational changes in the input domain can be relayed through various mechanisms to achieve control over the signaling domain as outlined below. An emerging theme has been the versatility of different light-sensing domains. A given type of photosensory module, whether it be a cryptochrome, phytochrome, or LOV domains can often be utilized in multiple different approaches to control signaling.</p><!><p>Light-induced heterodimerization has been leveraged to control protein activity through two main approaches. In the first approach, each partner in the dimerization pair is fused to a protein of interest. Optically induced dimerization brings them together to facilitate their interaction. In the second approach, one partner of the dimerization module is fused to a subcellular targeting sequence. The second partner is fused to the protein of interest, so that upon optical activation, it is recruited to the targeted subcellular region. Light induced recruitment of various signaling protein domains to the plasma membrane has been used to control cellular responses such as those regulated by small G proteins 24, 25, lipid kinases and phosphatases 26, and heterotrimeric G protein subunits 27. Light induced recruitment to other subcellular locations such as the nucleus, cytoskeleton or organelle membranes has also been used to control a variety of processes 28-30.</p><p>Several light-inducible dimerization pairs have been applied in optogenetics, but two naturally occurring dimerization pairs, both from Arabidopsis, have been the most widely used. The first is the red light induced interaction between the phytochrome phyB and its binding partners PIF3 31 or PIF6 24. The second is the blue light induced interaction between the cryptochrome CRY2 and its binding partner CIB1 32. Each scheme has unique properties that can be advantageous depending on the application.</p><p>Like other plant phytochromes, phyB contains a photosensory core module (PCM) that binds a phycocyanobilin (PCB) cofactor. An advantage of the phyB/PIF scheme is that it is activated by non-toxic red light, and can be rapidly reversed using infrared light 24. A disadvantage is that mammalian cells do not contain endogenous PCB, so the cofactor must be added. To overcome this constraint, cells can be genetically engineered to synthesize their own PCB 33. Bacterial phytochromes use a biliverdin chromophore that is present in mammalian cells 34. This can form the basis for an engineered inducible dimerization tool that is sensitive to red light and does not require exogenous cofactor.</p><p>In contrast, CRY2 binds to the flavin cofactor FAD which is ubiquitous in mammalian cells. Its activation by blue light is convenient for live cell imaging because it allows imaging of longer wavelength fluorescent proteins to be performed without causing photoactivation of CRY2. Both full length CRY2 and a truncated version consisting of only the photolyase homology region (CRY2PHR) have been used, and the optimal choice was reported to be context dependent 35.</p><p>It is also important to consider the design of the signaling domain and how it is fused to the light-sensing domain. For example, optical control over the actin-remodeling protein cofilin used a partially impaired mutant version in order to minimize background activity in the dark prior to light induced recruitment to the cytoskeleton 36. Likewise, light-induced activation of Wiskott-Aldrich Syndrome Protein (WASP) involved heterodimerization with a GDP bound version of Cdc42 that would normally have minimal capability to activate WASP 31. Although the GDP bound states of small G proteins such as Cdc42 are often considered "off states", there was sufficient activity that when combined with a large increase in effective concentration it could activate WASP.</p><!><p>Optical control over protein activity can also be achieved by masking its interaction surfaces in a light-dependent manner, as exemplified by work using a light-oxygen-voltage sensing domain from Avena sativa (oat) phototropin 1 (AsLOV2). Like other LOV domains, AsLOV2 contains a flavin-based blue light sensing chromophore. In the dark state, the LOV domain interacts with a C-terminal helix termed the Jα helix. Light exposure causes unwinding of the Jα helix, and this change has been used to achieve unmasking of a protein fused to the C-terminus. The approach has been used to obtain optical control over the interaction between a constitutively active Rac1 mutant and its effectors 37. A similar principle has been used to optically control access of peptides that activate or inhibit specific proteins inside the cell 38, 39. It has also been used for optical control over the concentration of peptide ligands on the outer surface of the plasma membrane 40.</p><p>Optogenetic constructs based on light induced unmasking generally require more careful engineering than those based on light induced heterodimers, because of the need for effective masking in the dark state. However, this approach offers the advantage of only requiring a single genetically encoded construct. Also, it can be used to control proteins that may not be effectively regulated by subcellular targeting alone. In addition to the unmasking scheme described above, a diverse array of optogenetic tools have been engineered using LOV domains, as described in the sections below. Together, this collection of tools shows the remarkable versatility of light-sensitive domains for controlling protein functions.</p><!><p>Optical control of protein clustering has been used for both light induced activation 41 and inactivation. Oligomerization induces activation of proteins such as the small G protein Ras, and several types of receptor tyrosine kinases (RTKs). Blue light activation of CRY2 not only induced binding to CIB1, but can also generate CRY2 oligomers in the absence of CIB1. This property has been exploited to activate Ras or various RTKs 41-43. A modified version called CRY2olig was identified in a yeast two hybrid screen that exhibits dramatically enhanced oligomerization 44. An alternative approach for inducible clustering, also demonstrated to provide optical activation of RTK signaling, used a bacterial LOV domain capable of forming light induced homodimers 45.</p><p>Whereas clustering leads to increased activity of certain proteins, in other cases it can lead to an inhibition of activity by sequestering the protein away from its effectors. This property has been leveraged for light-induced inhibition by cluster formation using a CRY/CIBN based scheme that incorporated a multivalent protein to enhance light-induced clustering 41. Conversely, an alternative approach used a mutant of the fluorescent protein Dronpa that oligomerizes in the dark and dissociates under cyan light to enable light-induced protein activation 46.</p><!><p>Optimizing optogenetic constructs for control of protein activity and cellular responses involves several considerations such as spectral characteristics, light-sensitivity, binding affinities, and kinetic properties. Variations in these properties can often be found across different members of families of naturally occurring light sensitive proteins. Additionally, mutagenesis can often uncover variants not found in nature that have properties better suited for engineering optogenetic tools. Information from structures and computational modeling has helped identify residues of interest for directed mutagenesis. High throughput methods involving imaging 47 or phage display 48 technologies have allowed optimized properties to be selected from large numbers of photoreceptor mutants. Below we describe some of the advances that these approaches have enabled toward generating optimized optogenetic tools.</p><!><p>Most of the photoreceptors used for engineering optogenetic tools respond very rapidly upon absorption of a photon, and this step does not typically limit kinetic control of cell signaling. In contrast, the rate at which a light-activated optogenetic construct reverts back to its "inactive" dark state upon removal of the light source varies widely among photoreceptors. Mutagenesis studies of the LOV domains from a variety of proteins have demonstrated the ability to drastically alter the off-rate kinetics by controlling the stability of a conserved cysteinyl-flavin adduct associated with photoactivation. For example, mutations in AsLOV2 designed to alter side chain interactions with the flavin cofactor and the surrounding water molecules were identified that alter photocycle times from 2s to over 2000s 49. For the LOV protein YtvA from Bacillus subtilis, mutagenesis identified a variant that reverts to the dark state 85 times faster than wild-type following photoactivation. Mutations of the slow-cycling fungal LOV photoreceptor Vivid (VVD) altered the off-rate kinetics by over for order of magnitude 50.</p><p>Ideal optogenetic tools have a large photoswitching dynamic range with minimal signaling in the dark state and a significant increase in signaling upon optical activation. Guided by an analytical model of photoswitching, mutational stabilization of the dark state interaction between the Jα helix and core domain of AsLOV2 helped increase the dynamic range of an LOV based optogenetic tool from 5-fold to 70-fold 51.</p><!><p>Two recently engineered light-induced dimerization schemes, iLIDs and Magnets 48, 52, may prove advantageous over the commonly used CRY2/CIBN and PhyB/PIF schemes for certain applications. Both schemes leverage LOV domains as optical switches, but they do so in distinct ways as described below (Fig. 1). These are not the first LOV based dimerization schemes 53, 54, but they are the most extensively optimized and represent the cutting edge for optical control over heterodimer formation. They offer small size and different ranges of affinities and kinetics that will allow them to be tailored to control of different protein activities.</p><p>The iLID scheme is based on the known interaction of the bacterial SsrA peptide with its binding partner SspB. Optical control over the interaction was achieved by embedding SsrA in the C-terminal helix of asLOV2 38. A multifaceted optimization procedure using computation, phage display and high-throughput binding assays improved the light-induced change in binding affinity from twofold to over 50-fold 48. Two iLID pairs were developed: iLID nano switches from 4.7μM to 130 nM affinity upon blue light exposure, whereas iLID micro switches from 800 nM to 4μM. As noted by the authors, the availability of inducible dimerizers with different ranges of affinities will be useful because the concentration threshold required for activity varies for different signaling proteins.</p><p>The Magnets heterodimerization pair was generated through mutagenesis of Vivid (VVD), a fungal photoreceptor derived from Neurospora crassa that is almost entirely composed of an LOV domain and can switch from a monomer to a homodimer in response to blue light 55-57. Mutations were introduced at the N terminal cap of VVD to alter the electrostatic properties of the homodimer interface and generate two VVD mutants, pMag and nMag, that selectively form heterodimers when exposed to blue light 52. Additional mutations in the Per-Arnt-Sim (PAS) core of VVD resulted in modified pMag/nMag interactions with dissociation rates in the dark ranging from 25 sec to 4.7 h 52. The ability of the fastest dissociating pMag/nMag pair, as well as both iLID pairs, to dissociate in less than a minute will enable better dynamic control over protein activity within a cell compared to the CRY2/CIBN scheme. Alternatively, the slowest dissociating pMag/nMag pair will be useful for controlling slow processes where minimizing repeated exposure to blue light could be important to avoid phototoxicity.</p><!><p>The magnitude of signaling output produced by an optogenetic construct depends on the intensity and wavelength of light used for photoactivation. Tuning the activation spectrum can generate new capabilities (Fig. 2). For example, spectral tuning can enable independent optical control of two different processes in the same cell using different wavelengths of light. Alternatively, it can allow live cell imaging of fluorescence proteins to be performed orthogonally to photoactivation. It can also enable the signaling response to be activated by far-red or near infrared wavelengths of light that are advantageous for in vivo applications due to their deeper tissue penetration.</p><p>The activation spectrum is determined largely by the specific chromophore and the neighboring residues within the protein. For example, 11-cis-retinal in solution has peak absorption at 380nm but by interacting with different opsins exhibits peaks varying from 360 to 560 nm (17). This natural variation in spectral characteristics can be leveraged for spectrally selective control over two different opsins. For example, mouse blue SWS1 opsin and human red opsin have been shown to both control GIRK currents with wavelength specificity 19. It will be highly desirable to extend this approach to control the activation of two different families of G proteins in the same cell.</p><p>A further red shifted opsin would also be useful for dual control in cells coexpressing CRY2 or LOV based constructs. Screening of naturally occurring opsins could uncover such a red shifted opsin. Alternatively, a mutagenesis approach will be aided by the extensive characterization of the residues involved in spectral tuning of opsins 17, 58-61. Spectral tuning can be altered by substituting the specific retinal bound to the opsin with an alternative form. For example, when 11-cis-retinal in a goldfish opsin is replaced with 11-cis-3,4-dehydroretinal, the λmax is red shifted more than 25nm 61. Synthetic retinal analogues can be used for this purpose. Such analogues have been shown to modify spectral and kinetic characteristics of microbial opsins 62. Regardless of the method used to red shift opsins, receptors activated at lower energy infrared wavelengths are likely to suffer from increased background noise due to thermal activation 63.</p><p>The flavin cofactors used by BLUF, LOV, and CRY modules are far less amenable to spectral tuning than the retinal and tetrapyrrole cofactors for opsins and phytochromes 10. For example, mutagenesis of bPAC would be unlikely to achieve a red or infrared sensitive adenenyly cyclase. However, a near infrared activated cyclase was engineered by an alternative approach using a bacteriophytochrome. In the native protein, the light sensing domain is linked to activation of a histidine kinase. By replacing the histidine kinase with a structurally homologous adenylyl cyclase, infrared light activated cyclase activity was achieved 34. A similar method was utilized to engineer a red light activated phosphodiesterase 64.</p><!><p>The ability to recruit the unique properties of a variety of photosensitive proteins has helped design optical tools to modulate signaling activity in a single cell with spatial and temporal precision. The challenges ahead are in overcoming some of the limitations of existent approaches. This includes the ability to quantitatively image the molecular and cellular responses to stimulation by optogenetic tools which at present is constrained by spectral overlap. A potential avenue towards surmounting this is to utilize the properties of additional light sensitive proteins and by developing more innovative strategies for optical control. This can be achieved by identifying novel light sensitive proteins with useful properties such as a collection of algal phytochromes that sense light optimally at wavelengths that span the entire visual spectrum 65. It can also be done by finding appropriate applications for existing protein systems in which conformational changes and protein-protein interactions have been shown to be light driven (e.g., 66-68).</p><p>Among the most important challenges is to apply these optogenetic tools to whole animals to address long standing basic questions with regard to development, differentiation and tissue physiology. The ability to shine light on a single cell and achieve subcellular control can be translated to individual cells within a population and can be useful for studying cell-cell communication. For example, activating ERK in single cells lead to propagation of ERK activity pulses in neighboring cells 69. Optical control can be used to direct cell migration during morphogenesis or extend neurites in predetermined directions. The long term goal should be to apply these tools towards light driven therapeutic intervention.</p>

PubMed Author Manuscript

The Transmission Interfaces Contribute Asymmetrically to the Assembly and Activity of Human P-glycoprotein*

Background: The P-gp drug-binding domain is linked to the two nucleotide-binding domains (NBDs) by NBD1 and NBD2 transmission interfaces.Results: Only mutations at the NBD2 transmission interface blocked P-gp assembly. Tariquidar repaired all 25 mutants.Conclusion: The mechanism of P-gp folding requires precise hydrophobic interactions at the NBD2 interface.Significance: We identified a linchpin for assembly and repair of ABC proteins.

the_transmission_interfaces_contribute_asymmetrically_to_the_assembly_and_activity_of_human_p-glycop

Introduction<!>Construction of Mutants<!>Expression and Maturation of ICL Mutants<!>Purification of P-gp and Measurement of ATPase Activity<!>Disulfide Cross-linking Analysis<!>Maturation of P-gp is Highly Sensitive to Mutations at the NBD2-TMD Interface (ICL2/3) but Insensitive to Mutations at the NBD1-TMD (ICL1/4) Interface<!><!>Maturation of P-gp is Highly Sensitive to Mutations at the NBD2-TMD Interface (ICL2/3) but Insensitive to Mutations at the NBD1-TMD (ICL1/4) Interface<!><!>Maturation of P-gp is Highly Sensitive to Mutations at the NBD2-TMD Interface (ICL2/3) but Insensitive to Mutations at the NBD1-TMD (ICL1/4) Interface<!>Rescue of Processing Mutants<!><!>The IH4-NBD1Contact Site Is Less Hydrophobic than the Equivalent IH2-NBD2 Site<!><!>The IH4-NBD1Contact Site Is Less Hydrophobic than the Equivalent IH2-NBD2 Site<!>Clamping of IH1 to IH4 Inhibits Activity<!><!>Clamping of IH1 to IH4 Inhibits Activity<!>The R262A(IH2)/R905R(IH4) Mutant Retains Robust ATPase Activity<!><!>Discussion<!><!>Discussion<!>

<p>The ATP-binding cassette (ABC)2 family is the largest class of transmembrane (TM) proteins (1). They use ATP hydrolysis to translocate a wide variety of substrates including metabolic products, lipids, peptides, sterols, ions, and drugs across extra- and intracellular membranes.</p><p>Human ABC proteins have been the subjects of intense research efforts since most are clinically important. Cystic fibrosis (CF), gout, intrahepatic cholestasis types 2 and 3 (liver bile diseases), Tangier disease (HDL deficiency), Dubin-Johnson syndrome (jaundice), hyperinsulimic hypoglycemia of infancy, pseudoxanthoma elasticum, multidrug resistance, secretory diarrheas, anemia, diabetes, and atherosclerosis are examples of potentially lethal or debilitating conditions caused by genetic mutations or altered activity of one of the 48 human ABC proteins (2).</p><p>The most common genetic defect is expression of an ABC processing mutant that is defective in folding and trafficking. The classic example is the ΔF508-CFTR (ABCC7) mutant in CF (3). Other examples includes ABCG2 (gout), ABCB4/ABCB11 (progressive familial intrahepatic cholestasis), ABCA1 (Tangier disease), ABCC2 (Dubin-Johnson syndrome), ABCC6 (pseudoxanthoma elasticum), and ABCC8 (hyperinsulimic hypoglycemia of infancy) (4). A high priority is to understand how processing mutations impact synthesis of ABC proteins and use this knowledge to develop therapies to repair the defects.</p><p>Human ABC proteins appear to be particularly sensitive to point mutations because they are multi-domain membrane proteins (core structure of two nucleotide-binding domains (NBDs) and two transmembrane domains (TMDs)) that require formation of specific domain-domain contacts to adopt a native structure (5). ABC proteins differ from other multi-domain proteins because much of the folding occurs post-translationally (6–8).</p><p>The human P-glycoprotein (P-gp) drug pump has been a very useful model system for studying repair of ABC protein defective in processing/folding because processing mutations throughout the molecule can be repaired by expression in the presence of drug substrates (9) or by introduction of arginine suppressor mutations into TM segments (10, 11). P-gp is a classic ABC protein as it is a single polypeptide of 1280 amino acids that forms a structure containing two NBDs and two TMDs. Its physiological role is block entry or export toxic compounds out of the body.</p><p>The interfaces between the domains play critical roles in the P-gp drug efflux mechanism. Drug substrates bind within a cavity located at the interface between the TMDs (11–15). Two ATP-binding sites are located at the interface between the NBDs. ATP hydrolysis occurs by an alternating site mechanism (16–20).</p><p>Coupling of ATP hydrolysis in the NBDs to drug efflux from the drug-binding sites in the TMDs is mediated by ball-and-socket joints (21) at the NBD-TMD transmission interfaces. Ball-and-socket joints are intracellular loops (ICLs) connecting TM segments that contain a central intracellular helix (IH) that is in close contact to an NBD. Contacts between TMD1-NBD1 or TMD2-NBD2 are mediated by the first ICL of each TMD (ICL1 or ICL3). The second ICL of each TMD mediates TMD1-NBD2 (ICL2) or TMD2-NBD1 (ICL4) contacts.</p><p>Contacts between the homologous halves (TMD1-NBD2 or TMD2-NBD1) appear to be particularly important for mammalian drug pumps. For example, the BCRP/ABCG2 drug pump lacks contacts equivalent to P-gp ICL1/ICL3 (22). In addition, we found that mutations to IH1 (ICL1) had little effect on activity or maturation while mutations to IH2 (ICL2) suggested the presence of a hydrophobic interface that was highly sensitive to changes (23). The presence of such an interface was unexpected since modeling studies suggested that charged residues would play a dominant role (24).</p><p>Hypersensitivity to mutation at the IH2/NBD2 transmission interface compared with the IH1/NBD1 interface might explain the differences in the maturation of NBD2 deletion mutants of P-gp and CFTR (25). Deletion of NBD2 inhibits folding and trafficking of P-gp but not CFTR although both are structurally similar ABC proteins (26). It is possible that ICL2/ICL3 interactions with NBD2 are critical for P-gp maturation and the protein is destabilized when NBD2 is deleted. Understanding P-gp domain-domain interactions will also be of importance when extrapolated to folding defects of other mutant ABC proteins associated with disease and in the design of methods to counteract their effects. It has been shown that in processing mutations of P-gp or CFTR trap the proteins as partially folded intermediates with incomplete NBD-TMD interactions (7).</p><p>In this study, we tested whether the NBD2 transmission interface was particularly critical for P-gp assembly and activity by comparing the effects of mutations at the NBD1-TMD and NBD2-TMD interfaces. We report that the interfaces are asymmetric since P-gp maturation was only sensitive to mutations at the NBD2 interface. The results suggest that the NBD2-TMD interface is a linchpin for assembly and repair of ABC proteins.</p><!><p>Mutations were introduced into the wild-type or Cys-less P-gp cDNAs containing the A52-epitope or 10-histidine tags (27) by site-directed mutagenesis as described by Kunkel (28). In most cases, residues in ICL1 (residues 146–177), ICL2 (residues 250–277), ICL3 (residues 787–818) or ICL4 (residues 893–920) were replaced with alanines to test for effects on maturation. Exceptions were that alanines were replaced with leucine, glycines were replaced with valine, and leucines or valines were replaced with serine as performed previously (23, 29, 30). Mutants were constructed to contain an A52 epitope tag at their C-terminal ends for use in whole cell immunoblot assays (29). The presence of the epitope tag distinguished the mutant proteins from any endogenous P-gp. P-gp contains three N-linked glycosylation sites can be used to monitor maturation of human P-gp from an immature 150 kDa protein to a mature 170 kDa protein. P-gp cDNA was modified to contain a 10-histidine tag at the COOH-terminal end to facilitate purification of the expressed protein by nickel-chelate chromatography (31).</p><!><p>The ICL mutant cDNAs were transiently expressed in HEK 293 cells by a calcium phosphate precipitation approach as described previously (7). Briefly, 10 μl of 2.5 m CaCl2 was added to 90 μl H2O containing 2 μg of DNA followed by addition of 100 μl of BES solution (50 mm N,N-bis(2-hydroxyethyl)-2-aminoethanesulfonic acid, 280 mm NaCl and 1.5 mm Na2HPO4, pH 6.96). After 10 min at room temperature, 4 ml of HEK 293 cells (about 100,000 cells/ml) in Dulbecco's modified Eagle's medium (DMEM) with high glucose (supplemented with nonessential amino acids, 4 mm l-glutamine, 10 IU/ml penicillin, 10 μg/ml streptomycin, and 10% (v/v) bovine calf serum) was added and 1.5 ml of the mixtures were added to duplicate well of 6-well culture plates. After 5 h at 37 °C, the medium was replaced with fresh medium with or without 0.5 μm tariquidar. About 16 h later, the cells were harvested, washed with PBS, and cell pellets suspended in 150 μl of 2× SDS sample buffer (125 mm Tris-HCl, pH 6.8, 4% (w/v) SDS, 4% (v/v) 2-mercaptoethanol containing 25 mm EDTA. Samples were applied to 6.5% SDS-PAGE gels (minigels, 1.5 mm spacers, 15 wells). The gels were electroblotted onto a sheet of nitrocellulose and P-gp proteins detected using A52 monoclonal antibody, horseradish peroxidase conjugated anti-mouse secondary antibody, and enhanced chemiluminescence. The signals were imaged and levels of mature (170 kDa) P-gp relative to total P-gp (mature 170 kDa plus immature 150 kDa) determined using ChemidocTM XRS+ with Image LabTM software (Bio-Rad Lab. Inc., Mississauga, Ontario). An equivalent amount of the sample was loaded onto 10% (v/v) SDS-PAGE gels and subjected to immunoblot analysis with a monoclonal antibody against glyceraldehyde-3-phosphate dehydrogenase (GADPH) (internal control).</p><!><p>HEK 293 cells were plated onto fifty plates (10-cm diameter) and transfected with the cDNA of the histidine-tagged P-gp mutant when about 50% confluent. After 16 h at 37 °C, the medium was replaced with fresh medium containing 5 μm cyclosporin A. Cyclosporin A is a substrate of P-gp and acts as a potent pharmacological chaperone in promoting maturation and yield of P-gp (9). Cyclosporin A rather than tariquidar was used to rescue the mutants for purification because it is considerably less expensive when used in scaled up experiments. The cells were harvested after another 24 h at 37 °C and washed three times with phosphate-buffered saline (PBS, pH 7.4). The cells were then suspended in PBS and solubilized at 4 °C by addition of one volume (0.75 ml) of PBS containing 2% (w/v) n-dodecyl-β-d-maltoside (Anatrace Inc., Maumee, OH). After 10 min at 4 °C, insoluble material was removed by centrifugation at 16 000 × g for 15 min at 4 °C. DNA in the supernatant was removed by passage through a DNA miniprep microfuge column (Bio Basic Canada Inc., Markham, ON). The flow-through material was then applied onto a nickel spin column (Ni-NTA, Qiagen, Mississauga, ON) that had been pre-equilibrated with buffer A containing 50 mm NaPO4, pH 8.0, 500 mm NaCl, 50 mm imidazole and 20% (v/v) glycerol and 0.1% (w/v) n-dodecyl-β-d-maltoside. The column was then washed twice with 0.6 ml buffer B containing 10 mm Tris-HCl, pH 7.5, 500 mm NaCl, 80 mm imidazole, pH 7.0, 20% (v/v) glycerol and 0.1% (w/v) n-dodecyl-β-d-maltoside and twice with 0.6 ml buffer C (buffer B containing 50 mm imidazole). Histidine-tagged P-gp was then eluted with 0.2 ml of buffer B but containing 300 mm imidazole. Recovery of P-gp was monitored by immunoblot analysis with rabbit anti-P-gp polyclonal antibody (32). A sample of the isolated histidine-tagged P-gp (about 100 ng) was mixed with an equal volume of 10 mg/ml sheep brain phosphatidylethanolamine (Type II-S, Sigma) that had been washed and suspended in TBS. ATPase activity (33) was measured in the presence of 0.4 mm verapamil.</p><!><p>Cys-less P-gp or the ICL1/ICL4 double cysteine mutants I160C/F904C or F163C/R905C were transiently expressed in HEK 293 cells at reduced temperature (30 °C) to promote maturation. The P-gps were isolated by nickel-chelate chromatography. The isolated P-gps were incubated 20 °C for 10 min in the presence or absence of 0.5 mm copper phenanthroline (oxidant to promote disulfide bond formation). EDTA was then added to a final concentration of 2 mm. Immunoblot analysis and assay of ATPase activity were performed as described above.</p><!><p>The NBDs of P-gp are linked to the TMDs by four ICLs (Fig. 1, A and B). The IH located in the middle of the ICLs interacts with the NBDs to form ball-and-socket joints (21). The NBD1 transmission interface is linked to TMD1 and TMD2 by ICL1 and ICL4, respectively. The NBD2 transmission interface is linked to TMD1 and TMD2 by ICL2 and ICL3, respectively.</p><!><p>Models of human P-gp. A, secondary structure of human P-gp showing the four ball-and-socket joints of the NBD-TMD transmission interfaces. The intracellular loops (ICLs) containing intracellular helices (IHs) that interact with the NBDs are shown in color. The cylinders represent TM segments, and the branched lines in the loops connecting TM segments 1 and 2 represent glycosylated sites. B, predicted structure of human P-gp in an open conformation was based on the crystal structure of mouse P-gp (56). The intracellular loops interacting with the NBDs are colored. The model was viewed using the PyMol system (57).</p><!><p>The transmission interfaces might play critical roles to promote folding of P-gp into a native structure. P-gp is initially synthesized in the endoplasmic reticulum to yield a protease-sensitive loosely folded protein with incomplete packing of the TM segments and incomplete domain-domain interactions (34–36). P-gp then matures into a compact protease-resistant native conformation that leaves the endoplasmic reticulum for addition of complex carbohydrate in the Golgi and trafficking to the plasma membrane (36). Removal of the NBDs inhibits maturation of P-gp. P-gp is different from CFTR because deletion of NBD2 only inhibits P-gp maturation (26). This indicated that the NBD2-TMD interface might play a particularly important role in the mechanism of P-gp folding. Truncation mutants lacking NBD2 or both NBDs are trapped in the endoplasmic reticulum in protease-sensitive loosely folded conformations (37).</p><p>A mutational approach was used to test if residues in the ICLs at the NBD2 transmission interface played more important roles in maturation of P-gp compared with residues in the ICLs at the NBD1 interface. Maturation of P-gp can readily be monitored in whole cell assays as the protein contains three N-glycosylation sites in the extracellular loop connecting TM segments 1 and 2 (Fig. 1A). The protein is initially synthesized as a 150 kDa core-glycosylated protein. If the protein correctly folds into a compact structure it can exit the endoplasmic reticulum for modification of the carbohydrate in the Golgi to yield a 170 kDa mature protein.</p><p>Accordingly, 84 mutants were constructed that contained point mutations to residues in IH4/ICL4 and residues in ICL1, ICL2, and ICL3 flanking IH1, IH2, and IH3, respectively (segments of about 30 residues). The 36 IH1, IH2, and IH3 mutants constructed in previous studies (23, 38) were included for comparison.</p><p>Mutations were made to amino acids flanking each IH as it had been reported that some of these residues contributed to NBD-TMD interactions in the crystal structure of P-gp from Caenorhabditis elegans (21). In general, residues were replaced with alanine as it has a small side chain. Exceptions were that alanines were replaced with leucine, glycines were replaced with valine, and leucines or valines were replaced with serine as performed previously (23, 29, 30).</p><p>Mutants were transiently expressed in HEK 293 cells for about 16 h and whole cell SDS extracts were subjected to immunoblot analysis to determine the steady state levels of mature and immature forms of P-gp. Examples of the effects of the mutations compared with wild-type P-gp are shown in Fig. 2A. Wild-type P-gp showed efficient maturation as about 80% of the protein was present as the 170 kDa mature form of the protein. The remainder of the protein was the 150 kDa immature P-gp. We previously showed that the 150 kDa protein was core-glycosylated as it was sensitive to endoglycosidases H and F (39). The 170 kDa P-gp was sensitive only to endoglycosidase F.</p><!><p>Maturation of P-gp is highly sensitive to point mutations at the NBD2 transmission interface (ICL2/ICL3) but not point mutations at the NBD1 transmission interface (ICL1/ICL4). HEK 293 cells were transiently transfected with A-52-tagged P-gps containing point mutations in the ICLs. A, representative immunoblot of SDS extracts of whole cells transfected with wild-type P-gp (WT), control vector (Cont), or P-gp mutants R789A, V907S, Y790A, or V908S showing no maturation (R789A), partial maturation (V907S, Y790A) or maturation similar to wild-type P-gp (V908S). The positions of mature (170 kDa) and immature (150 kDa) forms of P-gp are indicated. The amount of mature 170 kDa P-gp relative to total (mature 170 kDa plus immature 150 kDa protein) was determined for wild-type P-gp (WT) or mutants with point mutations in ICL2 (B), ICL4 (C), ICL3 (D), or ICL1 (E) was determined (Percent Mature). Each value is the mean ± S.D. Residues in the predicted IH segments are indicated.</p><!><p>Mutant R789A (ICL3) represents an example of a mutation that blocked maturation as only the 150 kDa immature form of P-gp was detected (Fig. 2A). Mutants V907S (IH4) and Y790A (ICL3) represent examples of mutations that partially inhibited maturation. The V907S mutant yielded about equivalent levels of mature and immature P-gp. Mutation of Tyr-790 to alanine substantially reduced maturation as about 85% of the product was the 150 kDa immature protein. Mutant V908S (IH4) represents an example of a mutation that did not detectably reduce P-gp maturation. In general, the mutations did not markedly reduce the steady-state levels of P-gp expression 16 h after transfection. There was less than a 2-fold change in total P-gp compared with wild-type P-gp. For example, the total P-gp yields in mutants R789A, V907S, Y790A, and V908S (Fig. 2A) were about 60, 75, 75, and 98% of wild-type P-gp, respectively.</p><p>The effects of ICL mutations are shown in Fig. 2 (panels B–E). Maturation of P-gp was quite insensitive to mutations at the NBD1 transmission interface. None of the 32 residues in ICL1 when mutated affected maturation of P-gp (Fig. 2E; >80% mature P-gp). Similarly, the mutations to residues in ICL4 (except for V907S) also did not affect maturation of P-gp. Mutant V907S yielded a slightly lower amount (about 55%) of mature P-gp.</p><p>By contrast, the NBD2 transmission interface was highly sensitive to point mutations (Fig. 2, B and D). Fourteen of the twenty-eight mutations in ICL2 (A250L, G251V, V253S, A254L, E256A, L258S, I261S, V264S, F267A, G268V, G269V, L274S, R276A, Y277A) inhibited maturation of P-gp (<15% mature P-gp) while four other ICL2 mutations (A2650L, R262A, T263A, and I265S) partially reduced maturation of P-gp (about 55–60% mature P-gp). The results suggest that the NBD1 and NBD2 transmission interfaces make asymmetric contributions to maturation of P-gp. Maturation of P-gp was highly sensitive to mutations that could disrupt formation of the ICL2/ICL3 tetrahelix bindle at the NBD2 transmission interface.</p><!><p>Drug substrates and modulators can act as pharmacological chaperones to rescue P-gp processing mutants (9, 40). There were 25 ICL mutations at the NBD2 transmission interface that yielded immature 150 kDa P-gp as the major product (Fig. 2, panels B–E). To test if maturation of the processing mutants could be restored, they were expressed in the presence of tariquidar. Tariquidar was selected because it is the most potent pharmacological chaperone for rescue of P-gp processing mutants (41). For example, tariquidar but not cyclosporine A could repair the F804D mutation at the ICL3-NBD1 interface (41). Rescue of a representative misprocessed mutant by tariquidar is shown on Fig. 3A. No detectable mature P-gp was observed when mutant A250L (ICL2) was expressed in the absence of drug substrate. Expression in the presence of tariquidar however, promoted maturation of the A250L mutant to yield mature 170 kDa P-gp as the major product. Accordingly, all 25 of the ICL processing mutants were expressed in HEK 293 cells in the absence or presence of tariquidar. A sample of whole cell SDS extracts was subjected to SDS-PAGE and immunoblot analysis and the amount of mature 170 kDa P-gp was quantified. It was found that all of the misprocessed mutants could be efficiently rescued with tariquidar to yield mature P-gp as the major product (Fig. 3, B and C; >90% mature P-gp). The results show that the defects caused by the ICL3 (Fig. 3B) or ICL4 (Fig. 3C) point mutations at the NBD2 transmission interface could be overcome by binding of tariquidar to the TMDs. We previously showed that a P-gp truncation mutant lacking the NBDs could be efficiently rescued with tariquidar (41).</p><!><p>Rescue of P-gp Processing Mutants with Tariquidar. HEK 293 cells were transiently transfected with A52-tagged wild-type P-gp or ICL processing mutants that yielded immature P-gp as the major product (Fig. 2) and expressed with (+) or without (−) 0.5 μm tariquidar (Tar). A, representative immunoblot of SDS extracts of whole cells transfected with wild-type P-gp (WT) or mutant A250L. The positions of mature (170 kDa) and immature (150 kDa) forms of P-gp are indicated. The amount of mature P-gp relative to total for wild-type P-gp or mutants with processing mutations in ICL2 (B) or ICL3 (C) was determined (Percent Mature). Each value is the mean ± S.D. (n = 3 different transfections).</p><!><p>Mutational analysis of the ICLs (Fig. 2) suggested that the NBD1 and NBD2 transmission interfaces were asymmetric in their contribution to P-gp folding. To test if these interfaces also contributed asymmetrically to activity, we examined whether the hydrophobic residues at the TMD2(IH4)-NBD1 site were critical for activity because the TMD1(IH2)-NBD2 transmission interface appeared to have a hydrophobic contact site (23). The IH2-NBD2 joint was hydrophobic because replacement of Phe-1086 or Tyr-1087 aromatic residues at the NBD2 socket (Fig. 1) with small or hydrophilic residues inhibited P-gp maturation and activity. For example, the F1086A mutation abolished activity. Activity of the F1086A could be restored if the opposing Ala-266 residue in IH2 was replaced with an aromatic residue (23).</p><p>Residues equivalent to Phe-1086 and Tyr-1087 at the IH4-NBD1 site are Leu-443 and Tyr-444. To test if the IH4-NBD1 contact point was also hydrophobic we first compared the effects of changes to Leu-443 and Phe-1086 on maturation and activity. Accordingly, A52-tagged mutants L443X (X = A, S, F, or R) were constructed. These mutants and A52-tagged mutants F1086X (X = A, L, W, or R) (23) were transiently expressed in HEK 293 cells and whole cell SDS extracts subjected to immunoblot analysis. The amount of mature P-gp was then quantified. All of the L443X mutations had relatively minor effects on maturation of P-gp (Fig. 4A). The amount of mature P-gp in mutants L443A, L443S, and L443F P-gps were similar to wild-type while mutant L443R yielded about 35% mature P-gp. Mutants F1086A, F1086L, and F1086W yielded mature 170 kDa protein as the major product, whereas F1086R yielded little mature P-gp (<5%) (Fig. 4A).</p><!><p>Point mutations at the NBD2-TMD contact point have a greater impact on P-gp maturation and activity than those at the NBD1-TMD site. A, HEK 293 cells were transiently transfected with A52-tagged wild-type P-gp (WT) or mutants with changes to residues in NBD1 that is adjacent to IH4 (Leu-443, Tyr-444) or to homologous residues in NBD2 (Phe-1086, Tyr-1087). Whole cell SDS extracts were subjected to immunoblot analysis and the amount of mature P-gp relative to total was determined (Percent Mature). Each value is the mean ± S.D. (n = 3). B, histidine-tagged mutants were expressed in the presence of cyclosporine A to promote maturation. P-gps were isolated, mixed with lipid and ATPase activity determined in the presence of verapamil. The results are derived from three different transfections + S.D. (n = 3 different transfections).</p><!><p>Histidine-tagged versions of the L443X (X = A, S, F, or R) and F1086X (X = A, L, W, or R) mutants were expressed in HEK 293 cells in the presence of cyclosporine A to promote maturation of the mutants. Expression in the presence of cyclosporine A promoted maturation of all the mutants to wild-type levels (data not shown). The mutants were isolated and ATPase activity measured in the presence of verapamil. Verapamil was used because it is a substrate (42) that highly activates P-gp ATPase activity (over 10-fold) (43). There is a good correlation between drug transport and activation of ATPase activity (44).</p><p>It was found that P-gp was again less sensitive to changes at the Leu-443 position. While the F1086A and F1086R mutations blocked verapamil stimulated ATPase activity, the L443A mutant resembled wild-type activity, and the L443R mutant retained about 35% activity (Fig. 4B).</p><p>It was possible that the asymmetric effects of changes to positions 443 and 1086 were related to the fact that different amino acids were found at the two locations in the wild-type protein (Leu-443 at NBD1 and Phe-1086 in NBD2). Therefore, we next examined the effects of changes to the residues at the Tyr-444(NBD1) and Tyr-1087(NBD2) positions (Fig. 1).</p><p>A52-tagged mutants Y444X were constructed so that tyrosine was replaced with a small (Ala), hydrophobic (Phe, Leu) or charged (Arg, Glu) residue. For comparison, the previously constructed Y1087A, Y1087F, and Y1087L mutants were included (38) and additional mutants Y1087R and Y1087E were also constructed. The mutants were transiently expressed in HEK 293 cells and whole cell SDS extracts subjected to immunoblot analysis. The amount of mature P-gp was quantified. All of the Y1087X mutations except for Y1087F blocked P-gp maturation (Fig. 4A). By contrast, the amount of mature P-gp in Y444F and Y1087F was similar to that of wild-type P-gp (Fig. 4A). All the Y444X mutants except Y444E differed from their Tyr-1087 counterparts as they yielded detectable levels of mature 170 kDa P-gp. The Y444A and Y444L mutants yielded about 25% mature P-gp while the Y444R mutant yielded about 40% mature P-gp (Fig. 4A).</p><p>We then tested whether the Y443X and Y1087X mutants had activity. Histidine-tagged mutants were expressed in HEK 293 cells in the presence of cyclosporine A to promote maturation of the mutants. The mutant P-gps were isolated, mixed with lipid and assayed for ATPase activity in the presence of saturating concentrations of verapamil. Fig. 4B shows that Y1087F retained about 30% of the activity of wild-type enzyme. The activity of wild-type P-gp was 2.2 + 0.2 μmol Pi/min/mg protein. The activity of Y1087 was almost completely inhibited when it was mutated to A, L, R, or E. In contrast, the Y444X mutations had less severe effects. With the exception of Y444E, the Tyr-444 mutations had less severe effects on P-gp activity. Both the Y444F and Y444R mutants showed over 75% of wild-type activity. Mutants Y444A and Y444L showed about 25% activity.</p><p>Results suggest that the IH4-NBD1 contact site is less sensitive to mutations and less hydrophobic than the equivalent IH2-NBD2 site. The difference is particularly evident when the aromatic residues were replaced with the positively charged arginine residue. Replacement of Phe-1086 or Tyr-1087 with arginines blocks P-gp maturation and activity. The L443R and Y444R mutations only showed modest reductions (Fig. 4). The results show that the NBD1 and NBD2 transmission interfaces contribute asymmetrically to activity.</p><!><p>Clamping the IH segments in close proximity by cysteine cross-linking can activate or inhibit P-gp ATPase activity. For example, cross-linking of cysteines in ICL1 and ICL3 to hold IH1 and IH3 in close proximity activated P-gp ATPase activity over 10-fold in the absence of drug substrates (45). By contrast, cross-linking of cysteines in IH2 and IH3 together (NBD2 transmission interface) did not activate basal ATPase activity and blocked drug-stimulated ATPase activity (38).</p><p>Since IH1 and IH4 at the NBD1 transmission interface were less sensitive to point mutations compared with IH2 and IH3 at the NBD2 transmission interface (Fig. 2), it was possible that P-gp activity might not be inhibited if IH1 were cross-linked to IH4. To test this, the first step was to identify pairs of cysteines in IH1 and IH4 that could be cross-linked. Pairs of cysteines were introduced into the N-terminal segments of IH1 (Glu-159 to Asp-164) and IH4 (Phe-904, Arg-905) of Cys-less P-gp. Cysteines were placed in the N-terminal regions of IH1 and IH4 as this was the strategy employed to identify disulfide cross-linking between IH2 (A259C) and IH3 (W803C). The 12 double cysteine histidine-tagged mutants were expressed in HEK 293 cells at low temperature (30 °C) to promote maturation. The mutants were isolated by nickel-chelate chromatography and subjected to cross-linking and assayed for activity (data not shown).</p><p>Mutants I160C/F904C and F163/R905C (Fig. 5, A and B) were selected to test the effects of IH1/IH4 cross-linking as the mutants yielded mature P-gp that showed efficient cross-linking in the presence of oxidant (copper phenanthroline) (Fig. 5C). Cross-linking can readily be detected because cross-linking between different domains causes P-gp to migrate slower on SDS-PAGE gels (7). The cross-linked products disappeared when samples were treated with dithiothreitol prior to immunoblot analysis (Fig. 5C). No cross-linked product was observed when the I160C, F904C, F163C, or R905C mutants were treated with oxidant (data not shown).</p><!><p>IH1/IH4 cross-linking inhibits P-gp ATPase activity. A, model of human P-gp showing parts of the various ICLs. The boxed inset is expanded in B to show the positions of residues I160C(IH1), F904C(IH4), F163C(IH1), and R905C(IH4). The distances (Å) between the α carbons of I160(IH1)/F904(IH4) and F163(IH1)/R905(IH4) are indicated. C, histidine-tagged Cys-less P-gp or mutants containing pairs of cysteines introduced into IH1 and IH4 (I160C(IH1)/F904C(IH4), F163C(IH1)/R905C(IH4)) (in Cys-less background) were transiently expressed in HEK 293 cells and isolated by nickel-chelate chromatography. Samples of the isolated P-gps were treated without (−) or with (+) copper phenanthroline (CuP). The reactions were stopped by addition of EDTA. Samples from the mutant P-gps were subjected to immunoblot analysis before (−) or after (+) treatment with dithiothreitol (DTT). The positions of cross-linked (X-link), mature (170 kDa), and immature (150 kDa) P-gps are indicated. D, samples were also assayed for ATPase activity in the presence of verapamil. Each value is the mean ± S.D. (n = 3).</p><!><p>To test for the effects of cross-linking on activity, isolated Cys-less P-gp and mutants I160C/F904C and F163/R905C were treated without or with copper phenanthroline for 10 min a 20 °C. The reaction was stopped by addition of EDTA. The samples were then assayed for verapamil-stimulated ATPase activity. Before treatment with oxidant, both mutants exhibited robust drug-stimulated ATPase activity. Mutants I160C/F904C and F163C/R905C exhibited over 70% of the activity of the Cys-less parent (Fig. 5D). Cross-linking of mutants I160C/F904C and F163C/R905C with oxidant however, inhibited more than 90% of the activity. Treatment with oxidant had little effect on the Cys-less parent. Activity of the cross-linked mutants was restored when the disulfide bond was reduced with dithiothreitol (Fig. 5D).</p><p>The results show that IH1/IH4 and IH2/IH3 cross-linking had similar effects as they both severely inhibited the activity of P-gp. The results are in agreement with modeling studies that predict that the IH1/IH4 and IH2/IH3 segments undergo significant conformational changes during the reaction cycle that would alter the relative positions of residues at the IH1/IH4 and IH2/IH3 interfaces (24, 46). Although the transmission interfaces contribute asymmetrically to folding and activity, the cross-linking studies show that movement between the helices at both interfaces are important for activity.</p><!><p>Pajeva et al. (24) predicted that homologous arginines at positions 262 (IH2) and 905 (IH4) were critical for coupling ATP hydrolysis to drug efflux. We found however, that the R905C mutation caused only a modest reduction in drug-stimulated ATPase activity (Fig. 5D).</p><p>A study of Q loop mutations Q475A(NBD1) and Q1118A(NBD2) suggested that the P-gp transport mechanism shows redundancy (47). It was found that the single Q475A or Q1118A mutants retained transport activity and 35–50% of wild-type ATPase activity but the double Q475A/Q1118A mutant was inactive. These results with the Q-loop mutations suggest that ATP hydrolysis at one site might be sufficient for transport (47). A similar mechanism might operate in other ABC proteins (48, 49).</p><p>We constructed histidine-tagged R262A/R905A and T263A/T906A mutants (in wild-type background) to test if mutations in both IH2 and IH4 were needed to inactivate P-gp ATPase activity. It was found that both mutants retained about 70% of wild-type verapamil stimulated ATPase activity (Fig. 6). A Q475A/Q1118A control P-gp however, showed little detectable ATPase activity. The results show that residues Arg-262 or Arg-905 were not essential for coupling of drug binding to activation of ATPase activity.</p><!><p>Double mutations introduced into equivalent positions in IH2 and IH4 caused modest reductions in ATPase activity. Histidine-tagged wild-type P-gp (WT) or mutants containing point mutations to equivalent positions in IH2 and IH4 (R262A(IH2)/R905A(IH4), T263A(IH2)/T906A(IH4)), or NBD1 and NBD2 (Q475A(NBD1)/Q1118A(NBD2)) were expressed in HEK 293 cells, isolated by nickel-chelate chromatography and assayed for ATPase activity in the presence of verapamil. The results are derived from the mean of three different transfections + S.D.</p><!><p>The major novel mechanistic findings in this study were that the transmission interfaces make asymmetric contributions to folding and activity of P-gp. The results show that the NBD2 transmission interface is particularly important for folding of P-gp and helps to explain why P-gp maturation is sensitive to deletion of NBD2 (26). The NBD2 transmission interface appeared to require formation of a very precise tetrahelix bundle structure between ICL2 and ICL3 because P-gp maturation was highly sensitive to changes in these segments (Fig. 7). About half the ICL2 point mutations and a third of the ICL3 point mutations blocked maturation (Fig. 2, B and D).</p><!><p>P-gp maturation is highly sensitive to point mutations at the second transmission interface but relatively insensitive to changes at the first transmission interface. Predicted structure of human P-gp in an open conformation (based on the crystal structure of mouse P-gp) (56). Mutation of residues in ICL2 and ICL3 inhibited maturation of P-gp. The side chains of these residues are shown in blue and red, respectively. None of the mutations in ICLs 1 and 4 inhibited maturation of P-gp and their side chains are shown in green.</p><!><p>By contrast, maturation of P-gp was highly insensitive to changes at the NBD1 transmission as none of the ICL1 or ICL4 point mutations severely reduced maturation (Fig. 7). Apparently, P-gp only required a hydrophilic flexible NBD1 transmission interface as maturation was relatively insensitive to structural perturbations of the ICL1/ICL4 tetrahelix bundle or IH1/4 segments.</p><p>The NBD2 contact site required more precise and hydrophobic interactions for P-gp maturation and activity compared with NBD1. The F1086R or Y1087R mutations at the NBD2 socket blocked maturation and activity while P-gp showed substantial activity when the comparable mutations (L443R, Y444R) were made to the NBD1 socket.</p><p>Indeed, a very conservative change of Y1087F in NBD2 reduced P-gp activity by about 70% (Fig. 4B). Removal of the hydroxyl side chain may have detrimental effects on activity because the equivalent residue in the crystal structure of C. elegans P-gp (Tyr-1129) shows hydrogen bond interactions with Asp-846 in IH3 (equivalent to human Asp-805) and Arg-286 in IH2 (equivalent to human Arg-262) (21). In a modeling study of human P-gp, it was also predicted that the hydroxyl group of Tyr-1087 would form a hydrogen bond with Asp-805 (24). All other replacements to Tyr-1087 (Ala, Leu, Arg, and Glu) reduced P-gp maturation and activity to undetectable levels.</p><p>The C. elegans P-gp structure showed that the equivalent residues at the NBD1 transmission interface made the same contacts (21). The tyrosine at position 468 in NBD1 (equivalent to human Tyr-444) formed hydrogen bonds with Asp-188 in IH1 (equivalent to human Asp-164) and Arg-946 (equivalent to human Arg-905). In the human P-gp modeling study, it was also predicted that the hydroxyl group of Tyr-444 would form a hydrogen bond with Asp-164 (24). The presence of a hydroxyl group at position 444 was not essential however, as the Y444F mutant yielded wild-type levels of maturation and activity (Fig. 4).</p><p>Differences between equivalent IH4-NBD1 and IH2-NBD2 contact sites were also observed in a cross-linking study (7). Both mutants L443(NBD1)/S909C(IH4) and A266C(IH2)/F1086C(NBD2) could be cross-linked with copper phenanthroline but only the L443C/S909C mutant was active. The activity of this mutant was lost upon cross-linking and restored with dithiothreitol.</p><p>Residues Asp-164 in IH1 and Asp-805 in IH3, predicted to form hydrogen bonds with Tyr-444 and Tyr-1087 respectively, have also been postulated to interact with the adenine ring of ATP (50). To test if residues Asp-164 or Asp-805 were essential for activity, Kapoor et al. (51) tested the effects of introducing D164C or D805C mutations into human Cys-less P-gp. They found that both mutations reduced trafficking of P-gp to the cell surface when they were expressed in HeLa cells. The D164C mutation reduced cell surface expression to about 40% of the Cys-less parent while the presence of both the D164C and D805C mutations reduced cell surface expression to about 20% of the parent. Immunoblot analysis of whole cell extracts expressing D164C/D805C showed that the major product was immature P-gp. The observation that P-gp mutants R262A/R905A, T263A/T906A (this study), and D164C/D805C (51) retain substantial activity suggests that there are multiple NBD-TMD contacts as reported by Jin et al. (21).</p><p>The D164C/D805C mutant could be efficiently rescued when expressed in the presence of cyclosporine A to yield mature P-gp as the major product. The rescued mutant showed transport levels of various substrates (rhodamine 123, NBD-cyclosporine, daunomycin, calcein-AM) similar to the Cys-less parent. Their results showed that Asp-164 or Asp-805 were not critical for drug transport. In agreement with cysteine mutagenesis results, we found that the D164A mutation did not reduce verapamil-stimulated ATPase activity (23).</p><p>A difference was that the D164A mutation did not reduce maturation of P-gp (23). The likely explanation is that the D164C mutation was introduced into a Cys-less P-gp background while the D164A mutation was introduced into a wild-type background. Mutation of P-gp's 7 endogenous cysteines reduces its maturation efficiency to make the protein more sensitive to processing mutations. For example, the F1086C mutation inhibits maturation of Cys-less P-gp but not wild-type P-gp (23).</p><p>Our results suggest that the NBD2 transmission interface is particularly sensitive to mutations relative to the NBD1 transmission interface. Studies on mutants of P-gp's structurally similar sister proteins, ABCB4 and CFTR, also suggest that the NBD2 transmission interface plays a key role in activity and protein assembly.</p><p>Human ABCB4 is a phosphatidylcholine transporter that shows 78% amino acid identity to P-gp (52). An ABCB4 Q1174E mutant was predicted to be the cause of progressive familial intraheaptic cholestasis type 3 in a patient (53). Analysis of the mutant suggested that the mutation disrupted the NBD2 transmission interface to inhibit substrate transport and substrate-induced ATPase activity (53).</p><p>CFTR is a chloride channel predicted to be structurally similar to P-gp (25). The NBD2 transmission interface was also found to be particularly important for CFTR assembly as processing mutations in other domains that cause cystic fibrosis (such as ΔF508 in NBD1, G91R in TMD1, L1093P in TMD2) were found to impair the conformational stability of NBD2 (54). Defects in folding of NBD2 can also be relayed through the second transmission interface to impair folding of the rest of the protein. For example the CFTR N1303K (54) and P-gp L1260A (55) NBD2 mutations inhibit maturation of the proteins.</p><p>NBD2 is more important for P-gp maturation than CFTR. A ΔNBD2 P-gp truncation mutant did not mature while ΔNBD2 CFTR showed robust maturation (26). The results of the present study suggest that NBD2 interactions are critical for stabilizing the ICL2/ICL3 tetrahelix bundle. The sensitivity of the P-gp second transmission interface to point mutations makes it an attractive target to develop compounds to inhibit activity or delivery of the protein to the cell surface to increase the effectiveness of chemotherapy or enhance drug delivery.</p><p>All of the ICL2/ICL3 processing mutations could be repaired with tariquidar. Binding of tariquidar to the TMDs (41) might stabilize the ICL2/ICL3 interactions in these mutants and promote proper interactions between various domains. The implication of these results is that binding of compounds to the TMDs of processing mutants in other ABC proteins could be a potential method for overcoming their folding defects.</p><!><p>This work was supported by a grant from the Canadian Institutes of Health Research (to D. M. C.). The authors declare that they have no conflicts of interest with the contents of this article.</p><p>ATP-binding cassette</p><p>P-glycoprotein</p><p>nucleotide-binding domain</p><p>human embryonic kidney</p><p>transmembrane</p><p>transmembrane domain</p><p>intracellular loop</p><p>intracellular helix.</p>

PubMed Open Access

Japonica Array NEO with increased genome-wide coverage and abundant disease risk SNPs

AbstractEthnic-specific SNP arrays are becoming more important to increase the power of genome-wide association studies in diverse population. In the Tohoku Medical Megabank Project, we have been developing a series of Japonica Arrays (JPA) for genotyping participants based on reference panels constructed from whole-genome sequence data of the Japanese population. Here, we designed a novel version of the SNP array for the Japanese population, called Japonica Array NEO (JPA NEO), comprising a total of 666,883 markers. Among them, 654,246 tag SNPs of autosomes and X chromosome were selected from an expanded reference panel of 3,552 Japanese, 3.5KJPNv2, using pairwise r2 of linkage disequilibrium measures. Additionally, 28,298 markers were included for the evaluation of previously identified disease risk markers from the literature and databases, and those present in the Japanese population were extracted using the reference panel. Through genotyping 286 Japanese samples, we found that the imputation quality r2 and INFO score in the minor allele frequency bin >2.5–5% were >0.9 and >0.8, respectively, and >12 million markers were imputed with an INFO score >0.8. From these results, JPA NEO is a promising tool for genotyping the Japanese population with genome-wide coverage, contributing to the development of genetic risk scores.

japonica_array_neo_with_increased_genome-wide_coverage_and_abundant_disease_risk_snps

<!>Tag SNP selection for JPA NEO<!>Selection of disease-related markers for JPA NEO<!>Development of JPA NEO<!>DNA samples<!>Genotyping with JPAs<!>QC analysis and genotype imputation<!>Tag SNP selection for improved genome-wide coverage<!><!>Tag SNP selection for improved genome-wide coverage<!><!>Tag SNP selection for improved genome-wide coverage<!>Selection of disease-related markers based on published evidence<!>JPA NEO has genome-wide coverage and contains disease risk SNPs<!><!>JPA NEO has genome-wide coverage and contains disease risk SNPs<!><!>JPA NEO has genome-wide coverage and contains disease risk SNPs<!>High imputation performance of JPA NEO<!><!>High imputation performance of JPA NEO<!><!>High imputation performance of JPA NEO<!><!>High imputation performance of JPA NEO<!><!>Large-scale genotyping by JPAs in the TMM project<!><!>Large-scale genotyping by JPAs in the TMM project<!><!>Large-scale genotyping by JPAs in the TMM project<!>Discussion<!>Author Contributions<!>Supplementary Data<!>

<p>The Tohoku Medical Megabank (TMM) Project was launched as part of reconstruction efforts following the Great East Japan Earthquake on March 11, 2011, and aims to establish a next-generation medical system for precision medicine and personalized healthcare (1). To accomplish the purpose, we have been conducting prospective genome cohort studies in connection with the establishment of an integrated biobank. Between 2013 and 2017, the Tohoku Medical Megabank Organization (ToMMo) and the Iwate Tohoku Medical Megabank Organization recruited 157,602 participants and conducted a baseline assessment, including the collection of biospecimens in Miyagi and Iwate Prefectures. The study population comprised two cohorts: the TMM Community-Based Cohort Study (TMM CommCohort Study) cohort, consisting of 84,073 adults (2), and the TMM Birth and Three-Generation Cohort Study (TMM BirThree Cohort Study) cohort, consisting of 73,529 pregnant women and their family members (3).</p><p>We have performed genome/omics analyses within the TMM project and established an integrated biobank that includes biospecimens, health and clinical information and genome/omics data to develop a research infrastructure for genomic medicine (4). Taking advantage of the two abovementioned cohorts, we planned a strategy for genomic analysis as follows: development of a whole-genome reference panel using the TMM CommCohort, large-scale genotyping and genotype imputation of both cohorts, and collection of accurate haplotype information from the TMM BirThree Cohort. Based on this strategy, we first established an allele frequency panel called 1KJPN, which includes the whole-genome sequencing (WGS) data of 1,070 participants (5). The reference panel was sequentially expanded to the latest version, 3.5KJPNv2, which consists of 3,342 and 210 samples from the participants of the TMM project and other cohorts in western Japan, respectively (6). Based on the updated reference panel, we have developed and refined custom single-nucleotide polymorphism (SNP) arrays for genotyping all 157,602 participants, as described below.</p><p>The International HapMap and the 1000 Genomes Project have shown that human genomes comprise regions with an extended linkage disequilibrium (LD) and of limited haplotype diversity, depending on the population (7, 8), and that SNPs within regions could be inferred from genotypes of a smaller number of SNPs. Carlson et al. showed that the selection algorithms of the set of SNPs for genotyping (referred to as tag SNPs) based on the r2, which are widely used for pairwise LD measures using reference genome sequences from different populations (9). Using tag SNPs, untyped sites can be complemented by genotype imputation using a reference genome to increase the number of SNPs that can be used for further association studies (10, 11). In large-scale multiethnic studies, four kinds of ethnic-specific SNP arrays were first designed for European, East Asian, African American and Latino populations with simulations of genotype imputation (12, 13). Similarly, biobanks and/or cohort projects developed ethnic-specific SNP arrays, such as the UK biobank Axiom Array (14), the Axiom-NL Array based on GoNL reference data in Netherlands (15) and the Axiom Array for Finnish of the FinnGen project. In East Asia, ethnic-specific custom arrays were also developed by the Taiwan Biobank as the TWB Array, based on the Axiom Genome-Wide CHB 1 Array (16); by the Korean biobank as Axiom KoreanChip, based on 2,576 WGS data (17) and by the Axiom China Kadoorie Biobank Array (18).</p><p>Most of these large-scale projects adopted the Axiom system because of the flexibility of the manufacturing array, the highly automated assay process and the robust sample tracking with a 96-array layout. Concurrent with the trend towards developing ethnic arrays, we also selected the Axiom system to design a Japanese-specific SNP array [the Japonica Array (JPA)]. We designed the first version of the Japonica Array (JPAv1) in 2014 (19). JPAv1 contains tag SNPs selected by means of a statistical measure called 'mutual information' with minor allele frequency (MAF) ≥0.5% to cover rare variants from a reference panel comprising 1,070 Japanese genomes, and a number of characterized SNPs from the genome-wide association studies (GWAS) catalog plus some other databases. In 2017, we updated JPAv1 and developed the second version (JPAv2) by increasing nontag SNPs, such as human leukocyte antigen (HLA), killer cell immunoglobulin-like receptor (KIR) regions and Y chromosome, and by replacing markers that were not working in JPAv1. Genotyping of TMM participants was conducted primarily with JPAv2 until 2018.</p><p>To enhance direct genotyping of previously identified disease risk variants and to obtain maximum genomic coverage with the expanded whole-genome reference panel including nearly 4,000 Japanese individuals (6), we aimed to design a novel and substantially revised version of the JPA, which we call Japonica Array NEO (JPA NEO). In this paper, we describe how we have improved upon JPAv2 to create JPA NEO. Of the various improvements, one salient point is that we have changed the selection algorithm for tag SNPs from using the mutual information criteria to the global standard of using r2 of the LD measure, aiming to improve imputation accuracy and to standardize the data for use in meta-analyses conducted anywhere in the world. We also report the progress of genotyping TMM participants by using all three versions of the JPA.</p><!><p>A target set was constructed using founders in the repository of our new genome reference panel consisting of genomes from 3,820 Japanese participants (6). For X chromosome, only females of above panel (2,066 Japanese individuals) were used. Tag SNPs were selected by the standard greedy pairwise algorithm based on the pairwise r2 of LD statistics (9, 12, 13). Briefly, starting with a set of target sites with an MAF higher than a specific threshold, one site with the maximum number of others exceeding the r2 threshold was selected. Then, this maximally informative site and all other associated sites were grouped as a bin of tag SNPs and removed from the target set. These steps were iterated until the total number of tag SNPs matched that of JPAv2. When multiple tag SNPs were selected in the same step, we prioritized them according to the following three criteria: (1) the maximum score of annotation by ANNOVAR (20) (exonic or splicing = 6, ncRNA = 5, 5′-UTR or 3′-UTR = 4, intronic = 3, upstream or downstream = 2, intergenic = 1 and no annotation = 0); (2) not A-T or G-C of alternative-reference alleles and (3) yielding the maximum variance of base-pair positions.</p><!><p>Disease-related markers were selected primarily from published lists of disease-related genes and GWAS results of Japanese populations, with expert advice. In addition, markers in the NHGRI GWAS catalog (21) and the UK Biobank Array (14) were also selected. From the latter, we extracted markers present in the Japanese population by referring to the 3.5KJPNv2 panel.</p><!><p>The list of tag SNPs and disease-related markers was combined with those of Y chromosome and mitochondrial markers. Based on the combined list, the array was produced using the Axiom myDesign service (Themo Fisher Scientific, Inc.). Multiple probes were designed for markers that were not included in the Axiom™ validated probe sets. Then, control markers were added, and the total number of markers was adjusted to the maximum number for the Axiom 96-array layout. The full marker list and detailed list of disease-related SNPs are available at the jMorp website (https://jmorp.megabank.tohoku.ac.jp/downloads/#jpa).</p><!><p>Isolation and quality control (QC) of genomic DNA from blood and saliva samples in the TMM biobank were performed as described previously (22). Genomic DNA samples isolated from the blood of TMM participants, but not those included in the reference panel, were used to evaluate JPA NEO. The study was approved by the Research Ethics Committee of ToMMo, Tohoku University.</p><!><p>A genotype assay was performed according to the manufacturer's protocol (i.e. Axiom™ 2.0 Genotyping Assay User Manual for 8-plate workflow). Briefly, target DNA was enzymatically amplified and fragmentated, and after confirmations of concentration and fragment length by NanoDrop (Thermo Fisher Scientific) and TapeStation System (Agilent Technologies), hybridization, ligation and scanning were processed by a semiautomated machine, GeneTitan™ Multi-Channel Instrument (Thermo Fisher Scientific). These processes were conducted using liquid-dispensable robots (Nimbus™, Hamilton; Biomek FXP, Beckman Coulter) and managed by a laboratory information management system (LIMS, LabVantage Solutions). For the QC, the dish quality control (DQC), sample QC call rate and plate pass rate were analysed using control markers for the Axiom platform (around 19,000) according to the Axiom™ Genotyping Solution Data Analysis Guide using Axiom™ Power Tools (APT, version 1.16.1). Genotyping data satisfying the criteria were used for the following QC analyses.</p><!><p>Genotyping data were further analysed for SNP QC, sample QC and plate QC using all markers per plate, in accordance with the abovementioned analysis guide. After filtering out variant sites with low call rates, low MAF, or showing substantial deviation from Hardy–Weinberg equilibrium, SHAPEIT2 (23) and IMPUTE2 (24) were used to conduct prephasing and genotype imputation, respectively. The imputation accuracy was evaluated using the squared correlation, r2, with leave-one-out SNP masking methods (12, 13, 25). Briefly, genotype imputation was performed by masking an input SNP and the imputed SNP was compared with the masked one to obtain r2, after which the average r2 in each MAF bin was calculated. Another metric, the information measure (INFO score) given by IMPUTE2, was used to analyse the imputation quality for each marker, where the value 0–1 indicated the uncertainty about the imputed genotype (11).</p><!><p>In JPA NEO, our updated version of the JPA, we used the maximum number on a single array of the Axiom 96-array layout, and the total of nearly 670,000 markers was divided into about 650,000 tag SNPs and tens of thousands of disease-related markers. The selection process of JPA NEO is essentially the same as previous versions of the JPA. However, we have selected these markers by using the latest version of our genome reference panel, which contains the genomes of 3,552 Japanese individuals (3.5KJPNv2) (6), which is about three times greater than that used for the previous versions (JPAv1 and JPAv2). Of note, while the previous two versions of the JPA used mutual information for tag SNP selection (19), in JPA NEO we decided to change the method for selecting tag SNPs to one based on the standard protocol using pairwise r2 (9) (Table I). This has the advantage of allowing us to harmonize our data with those of other studies. We believe that it is of great importance to perform meta-analyses with other large-scale GWAS utilizing the same concept. A comparison of the design of JPA NEO with those of JPAv1 and JPAv2 is summarized in Table I.</p><!><p>Overview of JPA design</p><p>r 2 of LD measures (9);</p><p>r 2 ≥ 0.8</p><p>Including samples outside from the Tohoku region.</p><p>Markers present in the Japanese population extracted by 3.5KJPNv2.</p><!><p>To optimize the selection of tag SNPs, we first selected tag SNPs from chromosome 10 of the 3.5KJPNv2 reference panel by using greedy pairwise algorithm (9) with different combinations of thresholds of MAF; i.e. ≥0.005, ≥0.01 or ≥0.05 and pairwise r2 of LD measures; r2 ≥ 0.5 or ≥0.8. Two metrics were used to evaluate tag SNP performance: (1) genomic coverage, which is the proportion of untyped sites with at least one tag SNP with r2 greater than a given threshold and (2) the number of variants obtained by genotype imputation above the threshold of a given INFO score, which is an index of imputation accuracy. When tag SNPs were selected by pairwise r2 ≥ 0.8 and MAF ≥ 0.01, the genomic coverage with r2 ≥ 0.8 and the number of imputed variants from the 2KJPN reference panel (2,049 Japanese genomes) with INFO > 0.9 were better or comparable to those of JPAv2 and Infinium Omni2.5-8 (Fig. 1). Based on these results, we decided to select tag SNPs with pairwise r2 of LD measures ≥0.8 and MAF ≥ 0.01 from the target set of autosomes and the X chromosome. For the design of JPA NEO, a substantial number, >1,000 of sex-chromosome SNPs on two pseudoautosomal regions were newly selected, whereas only about 10 SNPs on these regions were available in JPAv1 and JPAv2.</p><!><p>Evaluation of tag SNPs performance selected by different conditions. Tag SNPs were selected by different thresholds of pairwise r2 (0.5 or 0.8) and MAF (0.005, 0.01 or 0.05) from the target set of chromosome 10. The markers on JPAv2 or Infinium Omni2.5-8 (Omni 2.5) in the same chromosome were used for the control. (A) Genomic coverage was analysed with r2 threshold of 0.2, 0.5 or 0.8. (B) The number of imputed variants by the 2KJPN haplotype reference panel with an INFO score threshold of 0.5 or 0.9.</p><!><p>We also selected Y chromosomal markers for the Y haplogroup classification of the International Society of Genetic Genealogy (26) and from those in JPAv1 and JPAv2, which were selected using preexisting Axiom arrays for Asian populations. Mitochondrial markers were extracted mainly from 3.5KJPNv2 by removing those with MAF < 0.5% as well as those with multiple alleles. Most markers corresponding to the HLA and KIR regions were taken over from those adopted for JPAv1 and JPAv2.</p><!><p>For the selection of disease-related markers, we picked ∼9,000 markers present in the Japanese population, mainly from among published lists of disease genes and GWAS-identified risk variants. The former includes known and candidate functional variants on gene lists from the American College of Medical Genetics and Genomics (27) and 1,866 pharmacogenomics markers in 38 genes, 18 of which were obtained from drug guidelines published by the Clinical Pharmacogenetics Implementation Consortium as of April 2020 (28). The latter includes published risk variants for various complex diseases identified by GWAS of the Japanese population and a meta-analysis of East Asian populations. Representative examples are shown in a Supplementary Table S1, which includes 99 markers (96 genes) of type 2 diabetes (29), 100 markers (94 genes) of lipid metabolism, 45 markers (35 genes) of obesity, as well as 12 markers (7 genes) and 33 markers (24 genes) of late-onset Alzheimer's disease identified by GWAS of the Japanese population and meta-analyses of European populations, respectively.</p><p>Moreover, ∼13,000 and 12,000 markers were selected from the NHGRI GWAS catalog (21) and UK Biobank Array (14), respectively. We used reference panel 3.5KJPNv2 to extract the markers present in the Japanese population. The novel Axiom SNP array specific to the Japanese population was developed as JPA NEO.</p><!><p>The developed JPA NEO contains a total of 666,883 markers; the number of markers in each category is shown in Table II in comparison with JPAv1/JPAv2. In JPAv1/JPAv2, tag SNPs from autosomes and the X chromosome account for ∼98% (>650,000 SNPs). In contrast, nearly 8,500 SNPs from the Y chromosome (779 markers), mitochondria (409 markers) and HLA and KIR regions (6,757 and 532 markers, respectively) were also included to realize genome-wide coverage and genotyping of specific functional variants.</p><!><p>Number of markers for each category of JPAv1, JPAv2 and JPA NEO</p><p>Including markers present in multiple categories.</p><p>Extracted markers present in Japanese population by 3.5KJPNv2.</p><!><p>Although there is some overlap with the above SNPs, a total of 28,298 disease-related SNPs in 12 disease categories and pharmacogenomics are included as well (Table III). These SNPs include risk alleles for complex diseases, including dementia, depression and autism spectrum disorder among psycho-neurologic diseases (5,556 markers), type 2 diabetes and hyperlipidaemia among metabolic diseases (2,948 markers) and asthma and atopic dermatitis among immunological diseases (6,426 markers). In addition, variants related to physical traits (height, blood protein levels, etc.), expression quantitative trait locus, and so on are categorized as 'others'.</p><!><p>Summary of disease-related markers in JPA NEO</p><p>Some markers are included in multiple categories.</p><p>Physical traits, expression quantitative trait locus, etc.</p><p>Without overlap.</p><!><p>Of note, 3,472 markers (0.52%) in JPA NEO were MAF < 1% as confirmed by 3.5KJPNv2 (Supplementary Table S2). This is due to the adoption of some disease-related markers regardless of their MAF in 3.5KJPNv2. We have compiled the full list of disease-related markers with keywords and disease categories as a Supplementary Table S3, which can be downloaded from the jMorp website (30).</p><!><p>We modified the tag SNPs for JPA NEO from the previous versions with the aim of improving the imputation coverage of the microarray. To verify this point, we analysed the performance of JPA NEO in comparison with that of JPAv2. To this end, the same 286 samples, which were not included in the 3.5KJPNv2 reference panel, were genotyped using both JPA NEO and JPAv2. We found that the median call rates of JPAv2 and JPA NEO for all markers per sample were >99.6% and 99.8%, respectively (Supplementary Table S4), indicating that the call rate of JPA NEO is slightly better than that of the JPAv2.</p><p>More than 99% of markers were polymorphic in both JPAv2 and JPA NEO, as we intended (Table IV). Some microarrays are designed to cover a wide range of ethnicities, which is in contrast to the aim and scope of our Japanese-specific arrays. We hypothesized that the former type of microarrays may have lower performance compared with ethnic-specific ones. To address this point, we compared the performance of JPAv2 and the Infinium Asian Screening Array (ASA), which covers a wide range of Asian populations, including Japan, by using the genomes of 191 Japanese in the TMM cohorts. We found that >17% of markers were monomorphic in the ASA array, while >99% worked as polymorphic markers in JPAv2 (Table IV) with a median call rate of >99% for both arrays (Supplementary Table S4). This observation supports our contention that ethnic-specific microarrays are critical for analysing each ethnic population.</p><!><p>Numbers of polymorphic markers according to small-scale genotyping of Japanese individuals</p><p>Used the same DNA sample set for each comparative analysis.</p><p>Used the markers within the recommended probe set list created during the QC analysis.</p><p>Removed tri-allelic markers and markers for missing and overlapping positions.</p><!><p>When we closely inspected the MAF distributions of JPA NEO in comparison with those of JPAv2, we noticed that JPAv2 showed low numbers of MAF markers (15–25%) compared with JPA NEO (Fig. 2). We envisage that this may be due to the method for selecting tag SNPs. However, our new selection method has significantly improved the marker distribution in this region.</p><!><p>MAF distributions from small-scale genotyping. The MAF distributions of (A) JPA NEO and (B) JPAv2 were obtained by genotyping 286 individuals and analysing 659,754 and 643,417 markers, respectively. The number of markers present in each MAF bin (0.01 interval) is shown.</p><!><p>We performed genotype imputation of autosomes by using the haplotype reference panel of 3.5KJPNv2 and evaluated the imputation accuracy according to two metrics, imputation quality r2 and INFO score. The mean r2 and INFO score were >0.9 and 0.8, respectively, in MAF bin >2.5–5% of two arrays (Fig. 3), indicating reliable imputation accuracy for both JPAv2 and JPA NEO. However, importantly, we also noticed that there was a significant decrease in mean r2 in the region over MAF 20% in JPAv2. Whereas the precise reason for this decrease remains to be clarified, the decrease has been abrogated in JPA NEO.</p><!><p>Imputation accuracy of JPA NEO compared with that of JPAv2. Imputation accuracy was measured by (A) the coefficient of determination, r2, and (B) the INFO score. Genotyping was performed for 286 individuals using both arrays, and genotype imputation was performed using the 3.5KJPNv2 haplotype reference panel. The mean values in each MAF bin are shown.</p><!><p>As shown in Table V, slightly but clearly more imputed markers with INFO > 0.8 were obtained from genotyping data by JPA NEO than JPAv2, especially those with MAF < 1% (1.08-fold). We found that a total of >12 million markers were imputed by the small-scale analyses of the two arrays. These results indicate that while both JPA NEO and JPAv2 provide sufficient power for genotyping the Japanese population and following genotype imputation, JPA NEO shows better imputation performance without any bias throughout MAF bins. Thus, we conclude that JPA NEO is the most reliable imputation array ever developed for the Japanese population.</p><!><p>Number of imputed markers (INFO score > 0.8)</p><p>Used the same DNA sample set for genotyping.</p><!><p>To establish a solid research infrastructure for genomic medicine in Japan, the TMM project aimed to generate as much genotype data as possible from its 150,000 participants. To this end, we have been genotyping TMM cohort participants using the JPA since 2014. To complete such as large-scale genotyping efficiently, we established an elaborate three-group system from sample selection to genotyping, which connects to the data qualification.</p><p>We prepared our own special workflow for the ToMMo analysis, which ensures efficient and reliable sample processing and supports high-throughput measurement (Fig. 4). The first step is preparing the target sample lists containing the thousands of participants corresponding to a specific purpose, such as the TMM CommCohort participants with respiratory function data. The selection of participants and availability of DNA samples or biospecimens are supported by LIMS at the TMM biobank (22). This step is conducted by Center for Genome Platform Projects. The second step is extracting and dispensing the DNA into 96-well plates. To divide samples into individual plates in a well-ordered and formulated manner, the correspondence between sample identifier (ID) and well position is manifested by creating the plate map before dispensing the DNA samples. This step is conducted by Group of Biobank. The final step is transporting the DNA plates and plate maps to the genotyping facility attached to the TMM Biobank, which is operated using LIMS by Group of Microarray-based Genotyping Analysis. For security control, different sample IDs were used for sample collection, storage and analysis (31).</p><!><p>Workflow for large-scale genotyping in the TMM project. Based on the plate maps created from the target sample lists, the DNA plates were prepared and transported to the genotyping facility.</p><!><p>Capitalizing on this workflow, in May 2020, we obtained JPA data of ∼130,000 participants who met the criteria for QC analysis using control markers. The dataset comprises ∼2,000 JPAv1, 101,000 JPAv2 and 27,000 JPA NEO data (Table VI). We have already analysed >63,000 samples from the TMM CommCohort by using JPAv2, whereas the TMM BirThree Cohort samples were analysed by either JPAv2 or JPA NEO. Considering further association analyses, we are in the process of designing a rigid protocol that would allow each family unit to be analysed by the same JPA.</p><!><p>Progress of genotyping TMM samples by JPA as of May 2020</p><p>Genotyping conducted by Toshiba Inc.</p><p>Including samples analysed by multiple JPAs.</p><p>Including one failed plate below the criteria of mean QC call rate of passed samples.</p><!><p>We have been using DNA samples obtained primarily from peripheral or cord blood. When samples from these sources were not available, mostly those from the children of TMM BirThree Cohort participants, DNA from saliva samples was used and analysed separately with the one from blood. In our operation, the QC pass rate has been >99% for blood samples using both JPAv2 and JPA NEO. In contrast, that of saliva samples as ∼90% using JPAv2, likely due to the presence of lower-quality samples. We believe that with this accomplishment, JPA NEO now has enough control data of the resident population to be an important and useful array for the entire Japanese population.</p><!><p>The TMM project is one of the first large-scale prospective genome cohort studies in Japan and aims to realize precision medicine and personalized healthcare. To construct genome research infrastructure, we had to consider a cost-effective and high-throughput strategy for the acquisition of genomic data of >150,000 participants. Based on previous studies on genomic variants in diverse populations (32), we recognized that commercial arrays for global or even Asian populations were not sufficient for our purpose. Therefore, we decided to develop a custom ethnic-specific SNP array, the JPA, to maximize the acquisition of polymorphic markers in the Japanese population and provide genomic coverage with reliable genotype imputation accuracy while reducing cost.</p><p>In the TMM project, the whole-genome reference panel was expanded from 1KJPN to 2KJPN and 3.5KJPN. The latest version, 3.5KJPNv2, was constructed not only with an increased number of single-nucleotide variants but also added those from the X chromosome and mitochondria (6). JPA NEO was designed by reselecting the tag SNPs of autosomes and the X chromosome from this panel. The haplotype reference panel for genotype imputation was also updated from 2KJPN to 3.5KJPNv2. This update to the imputation panel yielded an increase of >5 million imputed variants in the preliminary analysis of 335 samples using JPAv2 data compared with those obtained by genotype imputation with 2KJPN in the same sample analysis (data not shown). Thus, 3.5KJPNv2 is more effective than previous reference panels in providing genome-wide coverage in terms of both tag SNP selection and genotype imputation.</p><p>The genotype imputation performance of JPA NEO was evaluated in comparison with that of JPAv2 by performing a small-scale analysis. In the genotyping data obtained by both arrays, monomorphic markers were scarcely observed and the large number of variants were imputed with a high imputation quality r2 and INFO score. Of note, JPA NEO showed better statistics compared with JPAv2 but without any bias, suggesting that JPA NEO is the best-ever SNP array developed for the Japanese population. The compatibility of markers in JPAv2 and JPA NEO is ∼40% (data not shown). Therefore, it seems important to develop a method for utilizing the genotyping data obtained by different JPA array platforms, which we plan to provide as a user guideline when the full data of all 150,000 TMM participants are released.</p><p>JPA NEO incorporates nearly 30,000 disease-related variants previously reported in the literature and stored in databases, to allow for the evaluation of known functional risk alleles in the Japanese population. Because some SNPs with MAF < 1% were included, their SNP cluster plots and the concordance with genotypes obtained by WGS analysis must be carefully assessed. However, qualified disease risk variants can be used for association studies along with phenotype data.</p><p>The JPAs have been used to perform large-scale genotyping of TMM samples. Whereas we have not experienced any issues with plate QC assessments conducted so far, we are planning to carefully implement batch-based as well as statistical genetic QC analyses to assess whether a plate effect is caused by sample selection bias. Indeed, in the UK Biobank, a sample picking algorithm has been used for genotyping experiments to prevent clustering of participants in the same plate by time or date of collection, collection centre, geography or participant phenotypes (33). In contrast, we did not intentionally randomize sample picking; we selected samples according to the aim of our analysis. For example, TMM CommCohort samples with respiratory function data for GWAS were selected and analysed using the same plates. Therefore, each plate should include samples collected from the same periods, regions and families.</p><p>Among the ∼130,000 genotyping data of TMM participants that we have processed so far, samples satisfying the criteria of sample DQC and QC call rate are quite high, especially when using blood samples (>99.8%). However, the pass rate of saliva samples was slightly worse (>89.8%) than that of blood samples in JPAv2. The use of saliva has been reported yield a low rate in other large-scale genotyping projects, for instance, 93.8% in the Genetic Epidemiology Research on Adult Health and Aging cohort (34). This may be due to the lower quality of saliva-derived samples, which is sometimes observed by electrophoresis as DNA degradation; this is likely due to problems during sample collection by participants, such as when they mix the sample with Oragene preservative solution. We are sharing the direct and imputed genotyping data with the research community upon completion of the QC analyses and genotype imputation. More than 54,000 JPA data have already been released as of June 2020, with associated data such as biochemical examinations and questionnaires. The full genotype data of the TMM project is expected to be released soon.</p><p>Data obtained by the genotype imputation array have been successfully utilized for GWAS. Summary statistics of large-scale GWAS are precious for the development of genetic risk scores, such as the polygenic risk score (PRS) (35). PRSs will be used to identify groups of individuals for therapeutic intervention, initiation and interpretation of disease screens and life planning (36). So far, the number and scale of GWAS in the European population greatly exceed those in non-European populations (37). However, the application of PRSs based on European cohorts to other populations is limited due to biases originating from the genomic diversity among populations, for instance, the difference in LD structure around causal variants. Further studies are required to develop PRSs for the Japanese population, and to evaluate their clinical utility used together with conventional clinical risk scores as pointed out recently (38, 39).</p><p>We believe that our future efforts should be focussed on acquiring genotype data from all participants of the TMM cohorts as well as implementing a GWAS to develop and evaluate genetic risk scores, including PRSs, optimized for the Japanese population. Genomic data obtained by the TMM project will serve as an excellent control for the GWAS executed using other biobanks/cohorts in Japan, and it will also be exploited for GWAS of associated phenotypes and omics data from the TMM project. We also believe that the JPA should continue to be updated. For the next version, we are planning to design a medical checkup array with a minimal set of tag SNPs that nevertheless contains abundant risk SNPs. These efforts will also contribute to further identifying genetic determinants of diseases in those of East Asian ancestry (32).</p><p>In conclusion, we designed a new version of the JPA, JPA NEO, to improve both genome-wide coverage and genotyping of disease risk variants. Disease risk variants were selected from the literature and filtered by our reference panel to extract those expected to be present in the Japanese population. Experimental verification using the developed JPA NEO showed greater imputation performance without any bias through a wide range of MAF and with increased imputed variants compared with the previous version. Large-scale genotyping of TMM samples using JPA NEO is now underway. JPA NEO will provide highly accurate, efficient and cost-effective genotyping for the Japanese population. Combining the JPA data of TMM participants with those of other Japanese biobanks/cohorts will be helpful for better understanding the genetic risks of complex diseases, leading to its application for disease risk prediction and prevention and consequently personalized healthcare.</p><!><p>K.Ku., C.G., S.M., A.U., S.T., I.N.M., A.O., A.N. and Y.A. performed the computational analyses. M.S.-Y., K.Ku., M.K., S.I., A.O. and H.Ku. prepared the samples and conducted the genotyping experiments. M.S.-Y., M.Y. and K.Ki. wrote the manuscript with assistance from the other authors. M.S.-Y., I.D., J.Y., H.Ka., N.M., S.K., N.F., G.T., M.Y. and K.Ki. conceived and supervised the project. All authors read and approved the final manuscript.</p><!><p>Supplementary Data are available at JB Online.</p><!><p>Click here for additional data file.</p>

PubMed Open Access

Excitation Energy Transfer and Exchange‐Mediated Quartet State Formation in Porphyrin‐Trityl Systems

AbstractPhotogenerated multi‐spin systems hold great promise for a range of technological applications in various fields, including molecular spintronics and artificial photosynthesis. However, the further development of these applications, via targeted design of materials with specific magnetic properties, currently still suffers from a lack of understanding of the factors influencing the underlying excited state dynamics and mechanisms on a molecular level. In particular, systematic studies, making use of different techniques to obtain complementary information, are largely missing. This work investigates the photophysics and magnetic properties of a series of three covalently‐linked porphyrin‐trityl compounds, bridged by a phenyl spacer. By combining the results from femtosecond transient absorption and electron paramagnetic resonance spectroscopies, we determine the efficiencies of the competing excited state reaction pathways and characterise the magnetic properties of the individual spin states, formed by the interaction between the chromophore triplet and the stable radical. The differences observed for the three investigated compounds are rationalised in the context of available theoretical models and the implications of the results of this study for the design of a molecular system with an improved intersystem crossing efficiency are discussed.

excitation_energy_transfer_and_exchange‐mediated_quartet_state_formation_in_porphyrin‐trityl_systems

<!>Introduction<!><!>Introduction<!><!>Results and Discussion<!><!>Results and Discussion<!><!>Results and Discussion<!><!>Results and Discussion<!><!>Results and Discussion<!><!>Results and Discussion<!>Conclusions<!>Supporting Information<!>Conflict of interest<!>

<p>O. Nolden, N. Fleck, E. R. Lorenzo, M. R. Wasielewski, O. Schiemann, P. Gilch, S. Richert, Chem. Eur. J. 2021, 27, 2683.</p><!><p>Due to their versatility, covalently‐linked chromophore‐radical systems have found a wide range of applications in the fields of information technology, artificial photosynthesis, as well as spin catalysis.[ 1 , 2 , 3 , 4 ] In all of these applications, the covalent linkage of a stable radical to the chromophore is used as a means of altering—and thereby controlling—the excited state dynamics of the chromophore, and many of them rely on the ability of the stable radical to act as an efficient triplet sensitiser by enhancing the intersystem crossing rate constant.</p><p>An increased triplet yield can for instance serve to improve the efficiency of processes such as triplet‐triplet annihilation photon‐upconversion,[ 5 , 6 , 7 ] while other applications in organic solar cells or OLEDs make use of the photoluminescence (doublet emission) of a particular class of these π‐radical systems.[ 3 , 8 , 9 , 10 ]</p><p>In the area of molecular spintronics, photogenerated organic multi‐spin systems have proven invaluable for exploring the fundamental requirements for spin‐information transfer and storage on a molecular level.[ 11 , 12 , 13 , 14 ] By photoexcitation, well‐defined initial spin states with strong spin polarisation can be generated, which may then be characterised by advanced magnetic resonance techniques with regard to their magnetic properties and spin‐spin interactions. Through systematic variations of the molecular design and a careful study of the resulting magnetic properties, it will ultimately be possible to establish design protocols for new magnetic materials exhibiting the desired properties for an efficient transfer and storage of spin information.</p><p>Motivated by the large number and potential impact of these applications, a multitude of recent studies have focused on increasing our understanding of the excited state dynamics and deactivation mechanisms in chromophore‐radical systems, with the long‐term goals of achieving photocontrol of the magnetic properties of materials via spin coupling and exploring the influence of molecular topology on the interaction efficiency.[ 1 , 2 , 3 , 15 , 16 , 17 , 18 , 19 , 20 , 21 , 22 , 23 ]</p><p>However, despite this effort, the mechanistic behaviour of organic multi‐spin systems is up to now only poorly understood and truly systematic studies that could help to elucidate the details of the excited state dynamics, the mechanisms of excited multiplet generation and spin‐information transfer, as well as the role of the individual interacting building blocks (i.e. chromophore triplet, bridge, stable radical) remain scarce.</p><p>Here, we investigate the spectroscopic properties of a series of covalently‐linked chromophore‐radical systems using femtosecond transient UV‐vis absorption (fs‐TA) and electron paramagnetic resonance (EPR) techniques. Porphyrins were chosen as the chromophores since they are highly photostable, [24] exhibit characteristic EPR signatures, [25] and their central metal can easily be modified, enabling a variation of the excited state energetics while avoiding major changes to the molecular structure. A tetrathiaryl trityl radical, derivatives of which are commonly employed as spin labels,[ 26 , 27 , 28 , 29 ] was attached to the porphyrin via a phenyl linker and is used here as the stable radical moiety. Compared to more commonly employed stable radicals, such as nitroxides, trityl radicals are characterised by a very narrow EPR line and slow spin‐lattice relaxation, [30] which might be particularly interesting for spin information storage applications. In addition, the chosen molecular design and synthetic approach allows a systematic modification of the individual building blocks of this photogenerated three‐spin system. [31] The three investigated structures only differ with respect to the central metal bound by the tetraphenylporphyrin (TPP) chromophore as shown in Figure 1 and are henceforth referred to as H2TPP‐trityl, ZnTPP‐trityl, and MgTPP‐trityl.</p><!><p>Chemical structures of the investigated chromophore‐radical systems and their building blocks.</p><!><p>Making use of the time‐resolution of fs‐TA spectroscopy, we explore the mechanism underlying excited multiplet generation in the porphyrin‐trityl systems and determine the efficiencies of the various competing excited state reactions. Complementary information is obtained from transient continuous wave EPR experiments, allowing us to identify and characterise the different spin states and spin–spin interactions. It is found that the interaction between the porphyrin triplet and the radical falls within the strong coupling regime, resulting in the formation of (doublet and) quartet states at cryogenic temperatures. Optical spectroscopy reveals that ultrafast energy transfer (∼10 ps) from the porphyrin to the trityl radical largely dominates the excited state dynamics at room temperature and in frozen solution, implying a very low quartet formation efficiency. However, despite this low yield, transient EPR enabled the determination of the characteristic magnetic parameters and spin polarisation patterns of the formed quartet states and provided further information on the mechanistic details of quartet formation, as well as on the internal dynamics of the spin system. Finally, the differences observed for the three dyads containing different porphyrin central metals are discussed, together with the implications of the results of this study for the design of a molecular system with an improved intersystem crossing efficiency.</p><p>In all transient spectroscopic experiments, the porphyrins were excited at a wavelength of 550 nm, corresponding to an absorption maximum within the porphyrin Q‐band region (S1 state). After photoexcitation of the porphyrin chromophore, a number of different excited state reactions can occur, as schematically shown in Figure 2, mediated by the close proximity of the third spin of the stable radical. The possible processes are (i) excited state electron transfer (ET), (ii) enhanced intersystem crossing (EISC), (iii) excitation energy transfer (EET) and (iv) enhanced internal conversion (EIC) back to the ground state.</p><!><p>Schematic overview of the photophysical processes expected to occur in covalently‐linked chromophore‐radical systems assuming a non‐negligible exchange interaction J TR. Abbreviations: EIC enhanced internal conversion; ET electron transfer; EET excitation energy transfer; EISC enhanced intersystem crossing.</p><!><p>Synthesis and characterisation: H2TPP‐trityl and ZnTPP‐trityl were synthesised according to procedures reported previously, [31] while the synthesis and characterisation of MgTPP‐trityl is outlined in the Supporting Information. An NMR characterisation of the compounds and precursor molecules (cf. Figure S1) revealed that MgTPP‐trityl is prone to oxidation during the purification process. The batch of MgTPP‐trityl thus contained some diamagnetic impurity (∼30 %), where the radical centre had been transformed into the corresponding triphenylmethanol moiety. Although no such observation was made for ZnTPP‐trityl and H2TPP‐trityl, the presence of very small amounts (∼0–3 %) of a diamagnetic impurity in the samples cannot be completely excluded due to the limited sensitivity of 1H NMR, especially with respect to the radical content of the sample. The presence of such an impurity would be observed as a triplet contribution in the spectroscopic studies, but is by no means expected to influence the photophysical behaviour of the chromophore‐radical compounds.</p><p>UV‐vis absorption spectroscopy: The room temperature UV‐vis spectra of the compounds and their porphyrin and trityl building blocks are shown in Figures S4 and S5. As is well‐known from the literature, [24] the spectra of the three tetraphenylporphyrins are qualitatively similar, showing an intense absorption band around 420 nm (porphyrin Soret band, S2 state) and significantly less intense absorption peaks in the wavelength region of the porphyrin Q‐bands from roughly 500 to 620 nm. The individual peak positions, however, differ for the different porphyrins and are also influenced by the solvent polarity. While the spectra of ZnTPP and MgTPP (approximate D4h symmetry) show one prominent absorption peak in the Q‐band region, this Q‐band peak is shifted to higher wavelengths and split into two major peaks in H2TPP, due to symmetry breaking induced by the presence of the two protons.</p><p>The spectrum of the trityl radical is characterised by a very prominent absorption band in the visible at 460 nm, a strong UV absorption, and a very broad but less intense absorption band extending over the whole visible range into the NIR (∼900 nm). [32] The absorption spectra of the porphyrin‐trityl compounds are similar to the sum of the UV‐vis spectra of their respective building blocks as can be seen from Figures S4 and S5 in the Supporting Information.</p><p>Femtosecond transient absorption: In order to get an idea about the efficiencies associated with the individual competing excited state reactions and their temperature dependence, femtosecond transient absorption measurements of the porphyrin‐trityl compounds as well as their building blocks were performed in toluene at room temperature and in isotropic frozen 2‐methyltetrahydrofuran solution at 85 K. Details on the sample preparation, experimental setup [33] and data treatment can be found in the Supporting Information. Since the observations for the investigated porphyrin‐trityl compounds were qualitatively very similar, only the results obtained for ZnTPP‐trityl and its building blocks are presented in the main text, while the corresponding data for the free base analogue are shown in the Supporting Information (Figure S6).</p><p>Figure 3 shows the room temperature fs‐TA data of the ZnTPP chromophore and trityl radical moieties. The photophysics of ZnTPP is well‐known and has already been studied extensively.[ 34 , 35 ] In agreement with these previous results, we observe a strong ground‐state bleach around 420 nm, accompanied by a considerably weaker excited state absorption extending over the entire visible range, with a maximum close to the onset of the ground state bleach at 460 nm. These typical features of the ZnTPP excited singlet state decay with a time constant of about 2.6 ns to form the porphyrin triplet state in high yield.[ 36 , 37 , 38 ] The latter then lives for about 1 μs in solution at room temperature. [36]</p><!><p>Femtosecond transient absorption data for ZnTPP (top) and the trityl radical (bottom) recorded in toluene solution at room temperature after photoexcitation at 550 nm. The central panel shows a contour plot of the data, where the red and blue colour coding represents positive and negative signals, respectively. The vertical coloured lines in the contour plot indicate the positions corresponding to the kinetic traces shown in the left panel, while the dotted horizontal lines indicate the time delays associated with the spectra shown on the right. The spectrum overlaid in blue corresponds to the spectrum of the ZnTPP excited triplet state, obtained from a global kinetic analysis of the data.</p><!><p>A global kinetic analysis of the data, as presented in detail in the Supporting Information, revealed that the evolution of the ZnTPP spectra with time can be described satisfactorily with two time constants of 11 ps and 2.6 ns. In addition, we detect an offset which is in line with the ∼μs lifetime of the porphyrin triplet state. The first time constant is likely to be associated with relaxation processes within the S1 state, while the second time constant can be assigned to the decay of the excited singlet state and simultaneous rise of the T1 state. As shown in Figure 3, the signatures of the porphyrin excited singlet and triplet states are very similar and may therefore be difficult to distinguish.</p><p>The fs‐TA spectra obtained for the trityl radical are shown in Figure 3 (bottom). They are characterised by a ground state bleach centred at about 460 nm, and a single, broad, and rather featureless excited state absorption extending from roughly 480 to ∼ 800 nm, with a maximum around 540 nm. The excited state absorption features do not change markedly over the course of the excited state lifetime. These features and the ground state bleach decay completely with a time constant of 120 ps, as determined by a global kinetic analysis of the data (cf. SI).</p><p>In the proximity of the trityl radical, the excited state dynamics of ZnTPP are considerably altered as evident from Figure 4, showing the spectra of ZnTPP‐ trityl recorded at room temperature (top) and in frozen solution (bottom). Shortly after photoexcitation of the porphyrin moiety, the typical features of the ZnTPP singlet excited state are observed and found to decay quickly, with a time constant of ∼10 ps. The decay of the S1 state is accompanied by the simultaneous rise of a new absorption band in the range from 480 to 800 nm and the formation of a new ground state bleach centred at 460 nm. A global kinetic analysis of the room temperature data (cf. SI) revealed that these newly formed excited state features subsequently decay with a time constant of roughly 120 ps. After their complete decay, a small percentage (∼ 4–5 %) of the porphyrin ground state bleach remains, together with some excited state absorption, attributed to the porphyrin triplet state.</p><!><p>Femtosecond transient absorption data for ZnTPP‐trityl recorded in toluene solution at room temperature (top) and frozen 2‐methyltetrahydrofuran solution at 85 K (bottom) after photoexcitation at 550 nm. For a comparison of the spectral shape, a representative spectrum of the excited trityl radical is superimposed onto the ZnTPP‐trityl data (red line). The wavelength ranges dominated by artefacts from either excitation light scattering or optical saturation were cut out from the spectra recorded at 85 K. In the room temperature data, the ground state bleach extending to ΔA=0.22 has been truncated for a better comparability of the relative spectral signals.</p><!><p>In frozen solution, the overall reaction dynamics are qualitatively similar, as shown in Figure 4 (bottom). However, the spectral positions of the individual excited state features appear to be slightly redshifted, due to the different solvent environment (e.g. dielectric constant) and temperature, and the excited state absorption is found to decay with a slower time constant of 650 ps.</p><p>The room temperature fs‐TA experiments on ZnTPP‐trityl were also performed at a different excitation wavelength (i.e. 400 nm) and in 2‐methyltetrahydrofuran solution (cf. SI). No marked differences in the spectra were observed, suggesting that the excited state deactivation mechanism in these systems is largely independent of solvent polarity and excitation wavelength (porphyrin Soret vs. Q‐band excitation). Even in polar solvents, no spectral features characteristic of ZnTPP ion formation could be observed, [39] indicating that any contribution of excited state electron transfer to the excited state dynamics can likely be excluded.</p><p>When comparing the room temperature fs‐TA spectrum obtained for ZnTPP‐trityl after the initial decay of the porphyrin S1 state, e.g. at 25 ps, with that of the excited trityl radical (cf. Figure 4, top), a strong similarity is evident. The spectral features are nearly identical, which is a clear indication for excitation energy transfer (EET) from the porphyrin to the trityl radical taking place in ZnTPP‐trityl. A detailed analysis of the relative spectral amplitudes and time constants (cf. SI) demonstrated that this energy transfer (i) is very efficient with a quantum efficiency of Φ EET≥95 % and (ii) occurs equally fast at room temperature and in frozen solution, with a time constant of ∼10 ps (no thermal activation barrier).</p><p>For energy transfer from the chromophore to the radical to occur, the radical needs to have low‐lying electronic states that are accessible from the chromophore's excited singlet state. In principle, two different energy transfer mechanisms could be invoked to explain the experimental observations: Dexter‐type energy transfer requires orbital overlap as it is based on an electron exchange mechanism. The associated transfer rates show an exponential distance dependence and a strong dependence on the bridge structure.[ 40 , 41 ] On the other hand, Förster resonance energy transfer is based on a dipolar mechanism, requiring spectral overlap of the fluorescence spectrum of the energy donor (ZnTPP) and the absorption spectrum of the energy acceptor (trityl), but no orbital overlap. The transfer rate depends on the distance between donor and acceptor as r DA −6. [42]</p><p>In order to evaluate whether the experimentally observed rate constants could be in agreement with Förster theory, we performed calculations of the expected Förster energy transfer rate for ZnTPP‐trityl, as detailed in the Supporting Information. The employed formalism relies on the validity of the point‐dipole approximation and is therefore likely not to yield accurate results in the case of the investigated structures, since the centre‐to‐centre distance between chromophore and radical is not significantly larger than their individual molecular sizes. Nevertheless, a good estimate can be obtained. The calculations were carried out with a centre‐to‐centre distance of r DA=1.3 nm and for different values of κ 2, describing the relative orientation of the transition dipole moments within the porphyrin and trityl moieties with respect to the vector connecting them. The correct value for κ 2 cannot easily be predicted, but given the covalent linkage between donor and acceptor and the known orientation of the transition dipole moments within the porphyrin and trityl moieties, it might be reasonable to assume that κ 2 is larger than one and maybe even close to its maximum value of four. Provided this assumption holds true, energy transfer time constants between 10 and 40 ps should be feasible (cf. SI).</p><p>The experimentally observed EET time constant of ∼10 ps thus seems to be consistent with a Förster‐type mechanism, although a contribution from Dexter‐type energy transfer cannot be excluded in view of the rather short distance between donor and acceptor and the covalent linkage between the two reaction partners, resulting in non‐negligible exchange coupling. [31]</p><p>The high efficiency of excitation energy transfer in these covalently‐linked porphyrin‐trityl systems naturally imposes severe limitations on the efficiency of the enhanced intersystem crossing (EISC) process and therefore the efficiency of excited multiplet formation. Efficient EISC should result in rapid quenching of the porphyrin S1 state accompanied by a simultaneous rise of the porphyrin T1 state. Here, we find that excited state deactivation is dominated by EET, suggesting that the rate constant of EISC is at least an order of magnitude smaller than that found for EET. From the ratio of the intensities of the initial ground state bleach and the ground state bleach remaining after complete decay of the trityl excited state absorption, we can estimate that the quantum efficiency of EISC can at most reach a value of ∼5 % in these systems.</p><p>In addition, part of this remaining triplet absorption could also result from a contribution of molecules where either (i) the trityl radical has been deactivated (diamagnetic impurity) or (ii) the orbital overlap between the chromophore and trityl moieties is negligibly small (unfavourable molecular conformation). Such molecules would undergo "normal", i.e. spin‐orbit coupling induced, intersystem crossing (ISC) to the porphyrin triplet state.</p><p>Since different origins for the triplet state contribution to the fs‐TA spectra of ZnTPP‐trityl can be envisioned, only an upper limit for the EISC yield of Φ EISC≤5 % can be given as a result from the analysis of the femtosecond transient absorption data.</p><p>Electron paramagnetic resonance: In order to verify whether excited multiplet formation can be observed for the different porphyrin‐trityl compounds and to characterise the magnetic properties of the individual spin states formed by the interaction of the porphyrin triplet and the trityl radical, transient continuous wave (cw) EPR measurements were performed in isotropic frozen 2‐methyltetrahydrofuran solution at 80 K. Details on the sample preparation and experimental setup are given in the Supporting Information.</p><p>Figure 5 (left) shows a comparison of the transient cw EPR spectrum of the ZnTPP triplet state and the spectrum of the trityl radical, recorded in pulse mode. The spectrum of the trityl radical is very narrow, due to the lack of hyperfine coupling interactions to nearby magnetic nuclei, and exhibits an isotropic g value of 2.0028, as determined by a spectral simulation of the corresponding room temperature cw EPR spectrum (cf. Figure S13). Some slight structural broadening is observed in the wings of the spectrum due to 13C satellite‐transitions. [31]</p><!><p>Comparison of the transient cw EPR spectrum of the triplet state of ZnTPP with the field‐swept FID‐detected spectrum of the trityl radical (left) and transient cw EPR spectra of ZnTPP‐trityl at two different time delays after photoexcitation at 550 nm (right). All spectra were recorded at the X‐band (9.75 GHz) in frozen 2‐methyltetrahydrofuran at 80 K.</p><!><p>Compared to the spectrum of the trityl radical, the triplet state spectrum of ZnTPP is very broad. As it is typical for the triplet states of organic chromophores, [43] the spectral shape and width is dominated by the zero‐field splitting (ZFS) interaction, described by the following Hamiltonian:(1)ℋZFS=SDS=DSz2-13S2+ESx2-Sy2</p><p>and parametrised by the two ZFS parameters D and E. In the case of the investigated porphyrins, D is known to be positive as determined from magnetophotoselection experiments, [44] implying an oblate spin density distribution. Due to differences in the intersystem crossing rates to the individual triplet sublevels, triplet spectra are typically spin‐polarised, leading to the observation of absorptive (a) as well as emissive (e) transitions.</p><p>To determine the zero‐field splitting D values as well as the relative initial populations of the three triplet sublevels, numerical simulations of the triplet state spectra were carried out. In good agreement with previously published results,[ 45 , 46 ] D T values of 930 MHz, 1150 MHz and 890 MHz were obtained for ZnTPP, H2TPP and MgTPP, respectively. While the out‐of‐plane triplet sublevel (Z) is primarily populated in ZnTPP, resulting in an aaaeee spin polarisation pattern, the in‐plane levels (X, Y) are overpopulated in H2TPP and MgTPP as evident from the eeeaaa polarisation pattern. The experimental spectra, simulations and further simulation parameters for all three tetraphenylporphyrins are shown in Figure S14 in the Supporting Information.</p><p>Compared to the spectra of the two building blocks, significant differences are observed in the transient cw EPR spectra of ZnTPP‐trityl as shown in Figure 5 (right). The overall spectral width is considerably reduced compared to that of the ZnTPP triplet state and a strong absorptive feature is found to dominate the central part of the transient spectrum. In addition, the spectral shape is found to change with time: While an overall a/e spin polarisation could be detected shortly after laser excitation, the spectra detected after a few microseconds are found to be entirely in absorption and are largely dominated by the central absorptive feature. [47]</p><p>This central feature with broad, significantly less intense, wings, observed after a few microseconds, is characteristic for a quartet state with Boltzmann population. In addition, a similar time evolution of the spin polarisation (a/e polarisation turning into net absorptive features) has been observed before for a porphyrin‐nitroxide system,[ 1 , 16 ] and could be attributed to thermal equilibration of the populations within the quartet state. [48] We therefore tentatively assign the spectral features observed here for ZnTPP‐trityl to the quartet state.</p><p>From a detailed analysis of the characteristic magnetic parameters (i.e. zero‐field splitting parameters D Q, E Q and the g value g Q) and spin polarisation patterns of the quartet state, important information on the quartet state formation mechanism can be obtained. However, before we discuss the implications of the present experimental results, the proposed mechanism for quartet state formation,[ 3 , 18 , 19 , 49 ] graphically summarised in Figure 6, shall briefly be outlined here.</p><!><p>Processes and energetic states involved in quartet state formation. The individual energetic states are labelled with terms of the form j Ai, where the superscript j indicates the overall spin multiplicity, A the spin multiplicity of the chromophore, and the subscript i the energetic ordering (low to high) of the states with identical multiplicities.</p><!><p>According to this model, light absorption leads to π–π* singlet excitation of the porphyrin moiety. From this so‐called excited sing‐doublet state (2S1), [24] EISC occurs to yield the chromophore triplet state. The triplet state is split into trip‐doublet (2T) and trip‐quartet (4T) states by the exchange interaction (ΔE DQ=3J TR≪k B T). Typically, the lowest trip‐doublet state (2T1) is populated first by EISC, since, due to different exchange interactions of the two triplet electrons with the electron of the radical, the trip‐doublet state acquires some sing‐doublet character, implying that the relaxation from the excited sing‐doublet state to the trip‐doublet becomes partially allowed. [24] The following transition from the doublet to the quartet state is spin‐forbidden, but can be mediated by spin‐orbit coupling (SOC).[ 3 , 18 , 19 ] Efficient mixing of quartet‐doublet eigenstates by SOC is however only possible when involving a nearby state with different orbital angular momentum. In metal porphyrins, the lowest two triplet states (E u,x, E u,y; π–π* excitation) are nearly degenerate and are likely to fulfil this role. The energetic difference between the two states involved in the mixing is referred to as Δ and needs to be (significantly) smaller than the thermal energy.</p><p>The mixing of 2T1 and 4T2 by SO‐ISC is then followed by fast internal conversion from 4T2 to 4T1. Through SOC‐mediated mixing, the doublet states acquire some quartet character (and vice versa), which will depend on the energy difference between doublet and quartet states (3J TR) and will also be spin sublevel specific. This spin‐selectivity leads to spin polarisation of the observed quartet state (4T1). In the following, back and forth transitions between quartet and doublet states (i.e. thermal repopulation of the trip‐doublet through mixing of 4T1 with 2T2) will deplete any initially overpopulated sublevels, causing a thermal equilibration of the populations. The decay to the ground state then primarily occurs from the doublet states, since the transition is spin allowed.</p><p>By SO‐ISC the quartet ±12 and ±32 sublevels are always expected to be equally populated. Therefore, the +12↔+32 and -12↔-32 transitions will have equal and opposite polarisations, the so‐called multiplet polarisation. The oppositely polarised multiplet contributions cancel each other in the centre of the spectrum and thus do not contribute to the central quartet feature which results from so‐called net polarisation.</p><p>Since net and multiplet polarisation have different properties (spectral positions, orientation dependence, magnetic field dependence, time behaviour),[ 18 , 19 ] the time evolution of the polarisation and the ratio of net to multiplet polarisation in the quartet spectra can provide valuable information on the internal dynamics of the system, especially the relative magnitudes of Δ and 3J TR.</p><p>The disappearance of the a/e multiplet polarisation in less than a few μs in the spectrum of ZnTPP‐trityl (cf. Figure 5) suggests that the doublet‐quartet mixing (equilibration of the quartet populations) is rather fast on the EPR time scale, implying that the energy gaps (3J TR, Δ) are small compared to the thermal energy (k B T ≅ 55 cm−1 at 80 K and 205 cm−1 at room temperature). Further, from a numerical simulation of the data, as shown in Figure 7, the relative ratios of the net to multiplet polarisation, the quartet zero field splitting parameters, D Q and E Q, as well as the g Q value of the quartet state can be obtained. The D Q value is of particular interest, since it is related to the dipolar interaction between the triplet and radical D TR as D Q=13 (D T+D TR). The obtained g Q value of 2.001 is well in line with the predictions for a quartet state from calculations based on the magnetic parameters of the chromophore triplet and stable radical. [50] This and the fact that the spectrum can be fit as a combination of net and multiplet polarisation, using the same magnetic parameters for both contributions, strongly supports our statement that the transient cw EPR spectra observed for ZnTPP‐trityl arise from a single species with quartet multiplicity. An overview of the experimental and calculated g and D values of all involved spin states is provided in Table S2. [51]</p><!><p>Transient cw EPR spectra of ZnTPP‐trityl, H2TPP‐trityl and MgTPP‐trityl (from left to right) at the X‐band (9.75 GHz) and ZnTPP‐trityl (right) at the Q‐band (34.0 GHz) together with the best numeric fit to the experimental data. All spectra were recorded in frozen 2‐methyltetrahydrofuran solution at 80 K at about 1 μs after photoexcitation at 550 nm. The quartet state spectrum was simulated as the sum of net and multiplet components (red dotted lines) with different weights. The simulation parameters are indicated in the figure.</p><!><p>Transient cw EPR spectra were also acquired for H2TPP‐trityl and MgTPP‐trityl as shown in Figure 7. While the spectral widths are found to be similar in all cases, significant differences in the initial spin polarisation patterns are observed. In particular, it is noted that multiplet polarisation is observed for the quartet states of ZnTPP‐trityl and MgTPP‐trityl, whereas only net polarisation could be detected on the EPR time scale in the case of H2TPP‐trityl. The central quartet peak is found to be positive for all investigated compounds, implying a positive sign of the exchange interaction J TR between chromophore triplet and radical. [19]</p><p>Although the formation of quartet states should go hand in hand with excited doublet state formation,[ 23 , 52 , 53 , 54 ] no doublet signatures could be detected experimentally. Possible reasons for this could either be (i) fast relaxation or (ii) spectral overlap with the quartet feature: Fast relaxation of the doublet excited state (on the EPR time scale) could be favoured since the transition to the ground state doublet is spin‐allowed. Fast relaxation also implies a broad spectral feature, which might be difficult to detect. In addition, since the g values of radical and triplet are very similar, also the g values of the formed quartet and doublet states will differ only little. Therefore, it is also plausible that the (weaker) signature of the doublet state is buried underneath the quartet spectrum and accounts for minor asymmetries observed in the experimental spectra. In addition, assuming the validity of the model for quartet formation outlined above, the doublet state is not expected to be spin polarised, implying a rather weak signal.</p><p>The opposite multiplet polarisation pattern observed for ZnTPP‐trityl and MgTPP‐trityl (a/e vs. e/a) might suggest a strong correlation between the quartet state polarisation and the polarisation of the triplet precursor (aaaeee vs. eeeaaa), i.e. a direct influence of the relative initial triplet state populations and the sign of D T, in agreement with previous observations. [55] However, an exception is known in the literature, where the polarisation pattern of the quartet state formed in a free‐base porphyrin‐verdazyl system was found to differ from that of the corresponding porphyrin triplet state, [56] implying that the mechanism is likely to be more complex and that other factors, such as the signs of D Q, D TR, and J TR or the proximity of vibrational states for mixing, also need to be taken into consideration.[ 19 , 22 ]</p><p>The qualitative similarity of the transient cw EPR spectra recorded for ZnTPP‐trityl at X‐ (9.75 GHz) and Q‐band (34.0 GHz) frequencies (cf. Figure 7) indicates that this system clearly falls within the strong coupling regime. Within this regime, the exchange interaction between triplet and radical, J TR, has to be larger than any other relevant magnetic interactions in the system, i.e. (i) all hyperfine couplings, the difference in the Zeeman frequencies between triplet and radical, and (iii) the zero‐field splitting in the triplet state. This situation is easily reached in porphyrin‐trityl systems, since the hyperfine couplings in the triplet state are no larger than about 4 MHz,[ 43 , 44 ] and the difference in Zeeman frequencies is proportional to the difference in g values, which is almost negligible here (cf. Table S2). The strong coupling limit is thus reached as soon as J TR exceeds the magnitude of the triplet D value (∼0.033 cm−1).</p><p>When having a closer look at the spectra acquired for ZnTPP‐trityl at the X‐ and Q‐bands, it further appears that the ratio of net vs. multiplet polarisation is slightly increased at the Q‐band (no negative intensities on the high‐field side due to cancellation of the signal with the contribution of the absorptive net polarisation). This is in line with the theoretical prediction that the relative contribution of the net polarisation to the spectra should increase linearly with the field strength and thus supports the validity of the proposed mechanism for quartet state formation (cf. Figure 6).</p><p>For all compounds it is observed that |D Q|∼|3E Q|, which should imply a comparatively large contribution of the net polarisation. [19] Nonetheless, compared to typical net:multiplet ratios expected based on the above‐mentioned mechanism, the contribution of the net polarisation to the quartet spectra of Figure 7 is surprisingly large. In particular, the complete absence of multiplet polarisation in the spectrum of H2TPP‐trityl is intriguing and cannot readily be explained since spin‐selective spin‐orbit induced intersystem crossing from the doublet to the quartet excited state should always lead to the observation of multiplet spin polarisation.[ 17 , 19 , 48 , 49 , 55 ]</p><p>This apparent inconsistency can be resolved by assuming that doublet‐quartet mixing is fast on the EPR time scale: By the time of detection, a significant part (or all) of the multiplet polarisation might already have decayed. However, to explain the experimental observations, these equilibration dynamics would need to be considerably faster in H2TPP‐trityl as compared to ZnTPP‐trityl or MgTPP‐trityl, implying a smaller Δ and/or J TR in H2TPP‐trityl (cf. Figure 6). Small differences in the molecular geometry or spin delocalisation of the triplet wavefunction could result in marked differences in J TR. Free base porphyrins are non‐symmetric and non‐planar. The porphyrin core is also likely to be more flexible, resulting in a larger distribution of J TR. Differences in geometry and/or triplet delocalisation are also suggested by the larger experimental D T value of H2TPP as compared to ZnTPP and MgTPP. On the other hand, no big differences in terms of delocalisation and J TR are expected between ZnTPP and MgTPP, in line with their near identical D T values.</p><p>Compared to the variations in J TR, even more significant differences could be imagined between H2TPP‐trityl and ZnTPP‐trityl/MgTPP‐trityl regarding the excited state energetics (and therefore Δ): In metal porphyrins, the splitting between Qx and Qy is typically about 30 cm−1,[ 24 , 25 ] implying that the degeneracy of the 3Eu states is only slightly lifted and that these two states could well be involved in quartet‐doublet mixing, since the energy gap is reasonably small compared to the thermal energy. In H2TPP, the D4h symmetry is broken, leading to a much larger splitting between Qx and Qy of the order of 2800 cm−1. [24] The mechanism thus needs to involve a different state in the case of the molecules containing H2TPP as the chromophore since any mixing processes involving the second triplet state would clearly be too slow.</p><!><p>In the present study we investigated the photophysics and magnetic properties of a series of three porphyrin‐trityl compounds, that only differ with respect to the porphyrin central metal. A detailed analysis of the femtosecond transient absorption data recorded at room temperature and in frozen solution enabled us to determine the efficiencies of the individual competing excited state reactions. It was found that excitation energy transfer largely dominates the excited state dynamics with a temperature‐independent time constant of ∼10 ps and an efficiency of ≥95 %, imposing severe limitations on the yield of enhanced intersystem crossing and therefore the efficiency of excited multiplet formation.</p><p>Nonetheless, the sensitivity of transient cw EPR was sufficient to detect the formation of quartet states for all three investigated compounds in frozen solution and numerical simulations of the data allowed the characterisation of the relevant magnetic parameters. No significant differences were observed between the spectra recorded at different microwave frequencies, as expected for a system that clearly falls within the strong coupling regime. Differences in the quartet state characteristics between the different compounds were discussed and could be rationalised in the context of previously invoked theoretical models and predictions,[ 3 , 18 , 19 , 22 ] strongly supporting the validity of the proposed mechanism for this class of compounds. However, in view of the potential applications of such triplet‐radical systems, the low EISC yield observed here is, of course, a substantial drawback. Future studies in this direction should thus focus on improving the design of the molecular system to increase the enhanced intersystem crossing efficiency. To this end, it will be necessary to find out how to suppress excitation energy transfer while keeping the exchange interaction between chromophore triplet and radical in a favourable range for EISC:</p><p>In the case of the investigated porphyrin‐trityl compounds, it is likely that the frozen solution samples are composed of molecules with different conformations (i.e. varying dihedral angles between the porphyrin plane and the phenyl spacer, and therefore different J TR). The EPR signal is thus expected to arise from only a small percentage of molecules in the sample that happen to adopt a conformation favouring EISC over EET. The efficiency of both mechanisms (EET and EISC) depends on the molecular conformation (i.e. orientation of transition dipoles and/or orbital overlap), but if the most favourable conformation for the two excited state deactivation mechanisms were different, one might have a chance to increase the efficiency of EISC by a slight modification of the present molecular design (e.g. by changing the bridge structure: topology, rigidity, extent of conjugation and/or length).</p><p>However, it is well possible that a change of the bridge structures is not sufficient to effectively suppress EET. In this case, the excited state energetics of radical and chromophore would need to be modified to eliminate the spectral overlap required for EET, which would most likely imply the choice of a different radical since the UV‐vis absorption of the trityl radical extends over the entire visible range. To safely exclude the possibility of energy transfer, the energies of both the S1 and T1 states of the chromophore would need to be smaller than the lowest excitation energy of the radical.</p><!><p>Synthesis and characterisation of the compounds, UV‐vis absorption spectra, approximation of the Förster rate, description of the experimental setups for fs‐TA and transient EPR, additional fs‐TA data and global kinetic analysis, additional transient EPR data and simulations, as well as DFT calculations of the spin densities.</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

Anion Dependent Self-Assembly of Polynuclear Cd-Ln Schiff Base Nanoclusters: NIR Luminescent Sensing of Nitro Explosives

Two types of polynuclear Cd-Ln complexes [CdLnL(NO3)Cl2(DMF)2] [Ln = La (1) and Nd (2)] and [Ln2CdL2(NO3)2(DMF)2](OH)2 [Ln = La (3) and Nd (4)] were constructed using a new Schiff base ligand which has a long backbone with two phenyl groups. The Schiff base ligands show a “twist” configuration in 1–4. The crystal structures show that the molecular dimensions of 3 and 4 are about 6 × 10 × 15 Å. The Cd-Nd complexes 2 and 4 exhibit the typical NIR luminescence of Nd3+. Interestingly, 4 shows the luminescent sensing of nitro explosives and exhibits a high sensitivity to 2-nitrophenol at the ppm level.

anion_dependent_self-assembly_of_polynuclear_cd-ln_schiff_base_nanoclusters:_nir_luminescent_sensing

Introduction<!><!>Materials and Methods<!>Preparation of the Schiff Base Ligand H2L<!>Preparation of [CdLaL(NO3)Cl2(DMF)2] (1)<!>[CdNdL(NO3)Cl2(DMF)2] (2)<!>Preparation of [La2CdL2(NO3)2(DMF)2](OH)2 (3)<!>Preparation of [Nd2CdL2(NO3)2(DMF)2](OH)2 (4)<!>Crystallography<!>Photophysical Studies<!>Synthesis and Structures<!><!>Synthesis and Structures<!><!>Synthesis and Structures<!><!>Synthesis and Structures<!>Photophysical Properties<!><!>Photophysical Properties<!>Luminescent Sensing of Explosives<!><!>Luminescent Sensing of Explosives<!><!>Luminescent Sensing of Explosives<!>Conclusions<!>Author Contributions<!>Conflict of Interest Statement<!><!>Supplementary Material<!>

<p>Currently, a great deal of attention is being paid to the lanthanide-based fluorescent chemosensors due to their unique optical properties (i.e., long lifetimes, line-like emission bands, and large Stokes' shifts; Jankolovits et al., 2011) and potential application in the detection of various analytes such as metal ions (Chen et al., 2009; Tang et al., 2013), anions (Qiu et al., 2009; Shi et al., 2015), and small molecules (Guo et al., 2011; Liu et al., 2013). Many visible luminescent Eu- and Tb-based frameworks such as Eu- and Tb-MOFs have been designed for this purpose (Chen et al., 2009; Qiu et al., 2009; Guo et al., 2011; Liu et al., 2013; Tang et al., 2013; Shi et al., 2015). In contrast, there are very few reports on the near-infrared (NIR) luminescent probes based on polynuclear lanthanide complexes, for example, Yb(III), Nd(III), and Er(III) complexes (Wu et al., 2018). In fact, NIR luminescent lanthanide complexes have been used as luminescent labels in the study of biological imaging and bioanalytical detection due to their low signal-to-noise ratios in living organisms (Hemmila and Webb, 1997; Stouwdam et al., 2003; Zheng et al., 2014).</p><p>It is known that light-absorbing Zn(II) and Cd(II) chromophores can be used as sensitizers for lanthanide emission in d-f complexes ("antenna effect") (Zheng et al., 2004; Zhu et al., 2006). We recently reported our studies focused on sensing with NIR luminescent Zn-Ln and Cd-Ln clusters formed by flexible salen-type Schiff base ligands, with long carbon-carbon (-CH2-CH2-) backbones (Jiang et al., 2018; Wang et al., 2018). The backbones of Schiff base ligands can efficiently affect the structures of d-f complexes. Thus, we present the synthesis and the structural characterization of two types of Cd-Ln complexes, with a specially designed Schiff base ligand 6,6′-((1Z,1′E)-(((ethane-1,2-diylbis(oxy))bis(2,1-phenylene))bis(azanylylidene))bis(methanylylidene))bis(2-methoxyphenol) (H2L), which has a long backbone with two phenyl groups (Scheme 1). These new complexes are [CdLnL(NO3)Cl2(DMF)2] [Ln = La (1) and Nd (2)] and [Ln2CdL2(NO3)2(DMF)2](OH)2 [Ln = La (3) and Nd (4)]. The length of H2L is approximately 20 Å, which helps to form large metal complexes. For example, molecules 3 and 4 are of nanoscale proportions, with the molecular dimensions ~6 × 10 × 15 Å. The Schiff base ligand H2L has four phenyl groups, which is advantageous to the formation of π···π electrostatic interactions with added explosives. Of particular note, 4 shows NIR luminescent sensing of nitro explosives, and exhibits high sensitivity to 2-nitrophenol (2-NP).</p><!><p>Schiff base ligand H2L.</p><!><p>Metal salts and solvents were purchased from Meryer and used directly without further purification. All reactions were performed in dry oxygen-free dinitrogen atmospheres using standard Schlenk techniques. Physical measurements: Powder XRD: D8ADVANCE; IR: Nicolet IS10 spectrometer. Melting points were obtained in sealed glass capillaries under dinitrogen and were uncorrected. Elemental analyses (C, H, N) were carried out on a EURO EA3000 elemental analysis. The thermogravimetric analyses were carried out on a TA Instruments Q600 under flowing N2 (200.0 mL/min) with a heating rate of 10.00°C/min from ambient temperature to 900°C. Field emission scanning electron microscopy (FESEM) images and EDX spectra were recorded on a Nova NanoSEM 200 scanning electron microscope.</p><!><p>2-[2-(2-aminophenoxy)ethoxy]phenylamine (2.80 mmol, 0.6840 g) in 20 mL EtOH was added drop by drop under reflux, to a solution of 2-hydroxy-3-methoxybenzaldehyde (5.60 mmol, 0.8520 g) in 10 mL EtOH. The yellow solution was then stirred for 3.5 h under reflux. The resulting yellow solid was filtered off, washed with 5 mL EtOH three times, and air dried. Yield (based on 2-[2-(2-Aminophenoxy)ethoxy]phenylamine): 1.3911 (97%). m.p. = 194.2°C. Elemental analysis: Found: C, 70.45; H, 5.62; N, 5.58%; Calc. for C30H28N2O6: C, 70.30; H, 5.51; N, 5.47%. IR (cm−1): 1,606 (m), 1,470 (w), 1,348 (w), 1,268 (w), 1,119 (m), 1,059 (m), 950 (w), 855 (m), 799 (w), 746 (s), 673 (s), 658 (s). 1H NMR (DMSO, 500 MHz): δ 13.94 (s, 2H), 8.93 (s, 2H), 7.41 (d, 2H), 7.24 (t, 4H), 7.05 (dd, 6H), 6.80 (t, 2H), 4.25 (s, 4H), 3.79 (d, 6H).</p><!><p>CdCl2 (0.2 mmol, 0.0367 g), La(NO3)3·6H2O (0.2 mmol, 0.0650 g) and H2L (0.2 mmol, 0.1024 g) were dissolved in 5 mL MeOH, 5 mL EtOH and 2 mL DMF at room temperature, respectively, and then mixed together. A solution of NEt3 in EtOH (0.35 mol/L, 1 mL) was added into the mixture. The yellow solution was stirred for 30 min under reflux and then filtered. The filtrate was transferred into a test tube, and then the test tube was placed in a jar with diethyl ether. The diethyl ether diffused slowly into the filtrate to create a pale yellow crystalline solid. The crystalline product was filtered off and air dried. Yield (based on La(NO3)3·6H2O): 0.0750 (36%). m.p. > 150.4°C (dec.) Elemental analysis: Found: C, 41.58; H, 3.91; N, 6.79 %. Calc. for LaCdCl2C36H40O11N5: C, 41.50; H, 3.84; N, 6.72 %. MS(ESI): 399.2294 (M+H)+, 513 ([H2L+H]+), 625 ([CdL+H]+), 648 ([LaL]+), 771 ([Cd2LCl]+), 1005 ([M-Cl−]+). IR (cm−1): 1,648 (m), 1,530 (m), 1,384 (w), 1,258 (w), 1,079 (m), 1,009 (m), 917 (m), 839 (m), 738 (s), 678 (s).</p><!><p>The pale-yellow crystalline product of this complex was obtained using Nd(NO3)3·6H2O (0.2 mmol, 0.0661 g) by a similar method described for 1. Yield (based on Nd(NO3)3·6H2O): 0.0711 (34%). m. p. > 133.8°C (dec.). Elemental analysis: Found: C, 41.50; H, 3.89; N, 6.82 %. Calc. for NdCdCl2C36H40O11N5: C, 41.30; H, 3.82; N, 6.69 %. IR (cm−1): 1,636 (m), 1,530 (m), 1,401 (w), 1,275 (w), 1,179 (w), 1,053 (m), 991 (w), 882 (m), 809 (m), 738 (m), 670 (s).</p><!><p>Cd(NO3)2·4H2O (0.2 mmol, 0.0617 g), La(NO3)3·6H2O (0.2 mmol, 0.0650 g) and H2L (0.2 mmol, 0.1,024 g) were dissolved in 5 mL MeOH, 5 mL EtOH and 5 mL DMF at room temperature, respectively, and then mixed together. A solution of NEt3 in EtOH (0.35 mol/L, 3 mL) was added into the mixture. The resulting yellow solution was processed in the same way described for 1 to obtain a pale yellow crystalline product of this complex. Yield (based on La(NO3)3·6H2O): 0.1225 (33%). m. p. > 245.8°C (dec.). Elemental analysis: Found: C, 45.35; H, 4.55; N, 6.83%. Calc. for La2CdC70H83O26N9: C, 45.26; H, 4.47; N, 6.79%. IR (cm−1): 1,633 (w), 1,540 (m), 1,377 (m), 1,232 (w), 1,062 (m), 968 (w), 849 (w), 758 (s), 666 (s).</p><!><p>The pale-yellow crystalline product of this complex was obtained using Nd(NO3)3·6H2O (0.2 mmol, 0.0661 g) using a similar method described for 3. Yield (based on Nd(NO3)3·6H2O): 0.1307 (35%). m. p. > 246.6 °C (dec.). Elemental analysis: Found: C, 45.10; H, 4.53; N, 6.80 % Calc. for Nd2CdC70H83O26N9: C, 44.99; H, 4.45; N, 6.75 %. IR (cm−1): 1,596 (m), 1,500 (m), 1,384 (m), 1,291 (w), 1,132 (m), 1,062 (m), 951 (w), 867 (s), 787 (m), 646 (s).</p><!><p>The diffraction experiments were carried out on a Smart APEX CCD diffractometer in the θ−2θ mode with monochromated Mo-Kα radiation (λ = 0.71073 Å). The structures were solved by direct methods (SHELX 97 program) (Sheldrick, 1997). All non-hydrogen atomic coordinates were refined anisotropically. Hydrogen atoms at their calculated positions were included in the structure factor calculation but were not refined. Selected bond lengths (Å) and angles (°) in the structures of 1–4 are shown in Tables S1–S4 (ESI). The CCDC reference numbers for the crystal structures are 1,865,266–1,865,269, respectively.</p><p>For 1: C36H40N5O11Cl2CdLa, monoclinic, space group P2(1)/n, a = 12.177(5), b = 19.331(7), c = 19.141(8) Å, α = 90°, β = 96.738(7)°, γ = 90°, V = 4,475(3) Å3, Z = 4, Dc = 1.545 g cm−3, μ(Mo-Kα) = 1.594 mm−1, F(000) = 2,072, T = 190 K. R1 = 0.1013, wR2 = 0.2029, GOF = 1.115.</p><p>For 2: C36H40N5O11Cl2CdNd, monoclinic, space group P2(1)/n, a = 11.800(5), b = 17.258(7), c = 19.859(8) Å, α = 90°, β = 93.755(8)°, γ = 90°, V = 4,035(3) Å3, Z = 4, Dc = 1.722 g cm−3, μ(Mo-Kα) = 1.995 mm−1, F(000) = 2084, T = 190 K. R1 = 0.0554, wR2 = 0.1765, GOF = 1.016.</p><p>For 3: C70H83N9O26CdLa2, orthorhombic, space group Pbcn, a = 32.423(10), b = 30.831(10), c = 15.887(5) Å, α = 90°, β = 90°, γ = 90°, V = 15,882(8) Å3, Z = 8, Dc = 1.547 g cm−3, μ(Mo-Kα) = 1.403 mm−1, F(000) = 7,432, T = 190 K. R1 = 0.0996, wR2 = 0.2557, GOF = 1.178.</p><p>For 4: C70H83N9O26CdNd2, orthorhombic, space group Pbcn, a = 32.458(11), b = 30.698(9), c = 15.924(6) Å, α = 90°, β = 90°, γ = 90°, V = 15,866(9) Å3, Z = 8, Dc = 1.558 g cm−3, μ(Mo-Kα) = 1.636 mm−1, F(000) = 7,480, T = 190 K. R1 = 0.0858, wR2 = 0.2489, GOF = 1.075.</p><!><p>The UV-visible absorption spectra were recorded at RT using an UV-3600 spectrophotometer. The solvent employed was of HPLC grade. Luminescence spectra in the visible and NIR regions were recorded on a FLS 980 fluorimeter. The light source for excitation and emission spectra was a 450 W xenon arc lamp with a continuous spectral distribution from 190 to 2,600 nm. A liquid nitrogen cooled Ge PIN diode detector was used to detect the NIR emissions from 800 nm to 1,700 nm. The temporal decay curves of the fluorescence signals were stored using the attached storage digital oscilloscope. The overall emission quantum yields (Φem) were obtained using an integrating sphere, according to Equation Φem = Nem/Nabs, where Nem and Nabs are the numbers of emitted and absorbed photons, respectively. The intrinsic quantum yields (ΦLn) of Ln3+ emission is calculated using ΦLn = τ/τ0, where τ and τ0 are the observed emission lifetime and the natural lifetime of Ln3+, respectively. Systematic errors were deducted through the standard instrument corrections. All measurements were carried out at room temperature. For the luminescent response experiment, the lanthanide NIR emissions of 4 were recorded when various concentrations of explosives were added into the solution of the complex with the initial concentration of 15 μM.</p><!><p>The synthesis of the new Schiff base ligand H2L was accomplished using preparations from literature (Lam et al., 1996), with a yield of 97% (Figure S1, Supporting Information) . The Cd-Ln complexes were synthesized from the reactions of H2L with CdCl2 and Ln(NO3)3·6H2O (Ln = La and Nd). The isomorphous 1 and 2 were obtained as pale-yellow crystalline solids. As shown in Figure 1, in 2 the Nd3+ and Cd2+ ions were bridged by two phenolic oxygen atoms of the Schiff base ligand with a separation of 3.872 Å. The coordination number of Nd3+ ion is ten, coordinated with eight O atoms from one L ligand, one NO3- anion and two DMF molecules and two N atoms from the L ligand. The Cd2+ ion is surrounded by four oxygen atoms from the L ligand and two Cl− anions. The Schiff base ligand coordinated with both metal ions through its two N and six O atoms. The bond lengths of Cd-O, Nd-N and Nd-O in 2 are 2.284–2.595, 2.758–2.783, and 2.424–2.740 Å, respectively.</p><!><p>A view of the crystal structure of 2. (Nd3+: blue; Cd2+: green; Cl: brown; N: purple; O: red; C: gray).</p><!><p>The nature of anions that existed in the reactions appears to have affected the self-assembly process of the clusters. Thus, the reactions of H2L with Cd(NO3)2·4H2O and Ln(NO3)3·6H2O (Ln = La and Nd) under similar experimental conditions produced isomorphous 3 and 4. The crystal structure of 4 is shown in Figure 2. Two Nd3+ and one Cd2+ ions are coordinated with two Schiff base ligands. The outer Nd3+ ion is bound by the O2N2O2 core of one L ligand in addition to four O atoms from two NO3- anions, resulting in a ten-coordinate geometry. While the center Nd3+ ion is eight-coordinates and bound by the O2O2 cavities of two L ligands. The Cd2+ ion is surrounded by four O atoms from the L ligand and two DMF molecules and two N atoms from the L ligand. The center Nd3+ ion is bridged with the outer Nd3+ and Cd2+ ions through four phenolic oxygen atoms of the Schiff base ligands. The Nd-Nd and Nd-Cd distances are 3.823 and 3.719 Å, respectively. In 4, the bond lengths of Cd-O, Nd-N, and Nd-O are 2.262–2.322, 2.699–2.950, and 2.218–2.888 Å, respectively.</p><!><p>A view of the crystal structure of 4. (Nd3+: blue; Cd2+: green; N: purple; O: red; C: gray).</p><!><p>The long Schiff base ligands show a "twist" configuration in 1–4, resulting in large molecular dimensions of the complexes. For example, the molecular sizes of 3 and 4 are about 6 × 10 × 15 Å. The panoramic scanning electron microscopy (SEM) image and energy dispersive X-ray spectroscopy (EDX) spectrum of 4 are shown in Figure 3. The molar ratio of Cd:Nd in 4 is confirmed to be 1:2 (Figure 3b), which is consonant with the crystal structure. The powder XRD patterns of the 1 and 4 show large background peaks, indicating that they are predominantly amorphous (Figure S2, Supporting Information). Thermogravimetric analyses show that 1–4 lose about 2% of the weight before 100°C (Figure S3, Supporting Information), due to the escape of uncoordinated solvent molecules such as MeOH and H2O.</p><!><p>Scanning electron microscopy image (a) and energy dispersive X-ray spectroscopy spectrum (b) of 4.</p><!><p>Melting point measurements indicate that 1–4 begin to discompose from 133 to 246°C (Experimental Section). Besides the molecular ion peak (m/z = 1,005), the mass spectrum of 1 shows fragments of the free ligand [H2L+H]+, [CdL+H]+, [LaL]+, [Cd2LCl]+ at 513, 625, 648, and 771, respectively (Figure S4, Supporting Information). This indicates that besides the product of 1, other species such as Cd-L, La-L, or Cd-Cd-L complexes may exist in the solution after the reaction. The products of 1–4 were collected form their solutions as crystalline solids.</p><!><p>The photophysical properties of 1–4 were studied in solution. The UV-vis absorption spectra of the free Schiff base ligand and 1–4 are shown in Figure 4. Compared to the absorption bands of the free ligand H2L, some of 1–4 are red-shifted. It is noticeable that, a broad absorption band at about 400 nm was found for 1–4, which may be from the ligand-to-metal charge transfer (LMCT) transition due to the existence of Cd(II) ions in the complexes (Blasse, 1994). For the Cd-La complexes 1 and 3, excitations of the ligand-centered absorption bands result in broad visible ligand-centered 1π-π* emission bands at 548 and 554 nm, respectively (Figure S5, Supporting Information), which are blue-shifted compared to that of the free ligand H2L (λmax = 602 nm). While, for the Cd-Nd complexes 2 and 4, besides the visible ligand-centered emission bands, they also show NIR luminescence of Nd3+ (4F3/2→4Ij/2 transitions, j = 9, 11, and 13) (Figures 5, 6). For the NIR luminescence, both 2 and 4 show broad excitation bands (i.e., λex = 327 and 386 nm for 4), indicating that the chromogenic Cd/L moieties can act as effective sensors for the luminescence of Nd3+ ions (Sabbatini et al., 1993; María et al., 2017). The excitation and emission wavelengths (λex and λem) as well as the absorption of excitation wavelengths (ε), luminescence lifetimes (τ) and overall luminescence quantum yields (Φem) of 2 and 4 in solution are listed in Table 1.</p><!><p>UV-Vis spectra of the free ligand H2L and 1–4 in CH3CN. (C = 10−6-10−5 M).</p><p>NIR luminescence spectra of 2 in CH3CN. (λem = 1,064 nm, λex = 388 nm).</p><p>NIR luminescence spectra of 2 in CH3CN. (λem = 1,060 nm, λex = 386 nm).</p><p>The excitation and emission wavelengths (λex and λem), the absorption of excitation wavelengths (ε), lifetimes (τ), and quantum yields (Φem) of 1–4 in solution.</p><p>The addition of 400 μM 2-NP.</p><!><p>2 and 4 show typical NIR emission bands of Nd3+ from 875 to 1,338 nm (Table 1). The luminescence lifetimes (τ) of 2 and 4 in CH3CN are 8.21 μs and 7.81 μs, respectively (Figure S6, Supporting Information). Therefore, the intrinsic quantum yields (ΦLn) of Nd3+ in 2 and 4 can be estimated at τ/τ0 = 3.28 and 3.12%, respectively, where τ0 = 250 μs [the natural lifetime of Nd3+ (Klink et al., 2000)]. As shown in Table 1, the overall NIR luminescence quantum yields (Φem) of 2 and 4 are 0.39 and 0.78%, respectively, indicating that 4 shows better luminescence properties than 2. This may be due to their different conformations and cooperative effects. For example, 4 has one more Schiff base ligand than 2 and can absorb and transfer more energy to the lanthanide ions. The efficiency (ηsens) of the energy transfer from ligand to Ln3+ can be calculated from ηsens = Φem/ΦLn (Bünzli and Piguet, 2005). Thus, the ηsens values in 2 and 4 are estimated to be 11.89 and 25.0%, respectively. For 1 and 3, the La3+ ion does not have f-f transition energy levels, and therefore cannot accept any energy from the sensitizer. As shown in Table 1, the ligand-centered emission quantum yields of 1 and 3 in visible range are 7.16 and 15.48%, which are higher than those of 2 and 4, respectively, due to no energy transfer to La3+ ion.</p><!><p>The NIR luminescent complex 4 has a larger surface area and more phenyl groups than 2, which is favorable to the formation of intermolecular interactions between 4 and guest molecules. Thus, the NIR luminescent response of 4 to nitro explosives such as 2-nitrophenol (2-NP), 2,4,6-trinitrotoluene (TNT), cyclotetramethylene tetranitramine (HMX), 1,4- dinitrobenzene (1,4-DNB), cyclotrimethylene trinitramine (RDX), 1,3-dinitrobenzene (1,3-DNB), 4-nitrobenzene acetophenone (4-NBAP), nitrobenzene (NB) and 4-nitrotoluene (4-NT) has been studied in CH3CN (Scheme 2). The intensity of the strongest emission peak of 4 at 1,060 nm was recorded with the addition of the explosives. Interestingly, the NIR luminescence intensities of 4 are gradually decreased with the addition of explosives increase (Figure S7, Supporting Information). It is noticeable that the addition of 2-NP leads to much more rapid luminescence quenching than other explosives (Figure 7). For example, the addition of 400 μM 2-NP solution makes the emission intensity at 1,060 nm decrease about 50%.</p><!><p>The structures of nitro explosives.</p><p>Decrease in the luminescence intensity of 4 (15 μM) in CH3CN upon the addition of different concentrations of 2-NP. Inset: linear relationship between the luminescence intensity and the concentration of 2-NP.</p><!><p>The Stern-Volmer (SV) equation, KSV = (I0/I- 1)/[A], can be used to calculate the luminescence enhancement or quenching constants of 4 to the explosives (Xiao et al., 2010). In this equation, I0 and I are the luminescence intensities before and after the addition of the explosive, respectively, and [A] is the molar concentration of the explosive. The KSV values of 4 to all explosives are shown in Figure 8 (Figure S7, Supporting Information). It was found that 4 shows the highest KSV value to 2-NP (3,020 M−1), indicating that 4 is most sensitive to this explosive. The KSV values to other explosives are from 225 to 2,240 M−1. The luminescence detection limits of 4 to the explosives can be calculated using the 3σ/Ksv equation, where σ is the standard deviation (Qi et al., 2017). The detection limit of 4 to 2-NP is found to be 14.70 μM, indicating that 4 shows high luminescence sensitivity to this explosive at the ppm level.</p><!><p>The luminescence enhancement or quenching constants (KSV) of 4 (15 μM) toward nitro explosives.</p><!><p>The perturbation of the added explosives, to the electronic structure of the ligand, may affect the ligand-to-lanthanide energy transfer process in 4. The luminescent quenching response of lanthanide-based sensors, toward nitroaromatic explosives, can be explained by photoinduced electron transfer (PET) and resonance energy transfer (RET) mechanisms (Li et al., 2013). In both mechanisms, the efficiency of ligand-to-lanthanide energy transfer is an important contributor to the luminescence intensity of the lanthanide complex (María et al., 2017). It was found that the intensities of ligand-centered fluorescence at about 559 nm of 4 are gradually increased with the addition of 2-NP (Figure S8, Supporting Information), indicating that more excitation energy of the Schiff base ligand may be consumed by visible emission. When the concentration of added 2-NP is 400 μM, the NIR emission lifetime and quantum yield of 4 is decreased to 6.42 μs and 0.41%, respectively (Table 1). Thus, the efficiency (ηsens) of the energy transfer is decreased to 15.95% from 25.0% (without the addition of explosives), demonstrating that the addition of 2-NP may efficiently affect the ligand-to-lanthanide energy transfer process and decreases the luminescence intensity of 4. The reason for the differences in explosive sensing properties of 4 is more difficult to understand since we do not know the precise nature of the interactions between the complex and the explosives that are introduced. A discussion of the precise nature of these kinds of interactions, as well as the difference between the luminescent response behavior of 2 and 4, is too speculative to be included in this paper. Our current studies are focused on attempts to isolate and characterize species which may interact with external explosives since this will provide useful information relating to explosive sensing.</p><!><p>In summary, two types of Cd-Ln complexes 1–4 have been successfully synthesized using a new Schiff base ligand (H2L), which has a long backbone with two phenyl groups. The length of H2L is about 20 Å, which is advantageous for the formation of large metal complexes. 3 and 4 are of nanoscale proportions and their molecular sizes are about 6 × 10 × 15 Å. The long Schiff base ligands show a "twist" configuration in all complexes. The chromogenic Cd/L moieties in 2 and 4 can act as efficient sensitizers to absorb and transfer energy to the Nd3+ centers, resulting in typical lanthanide luminescence. The Cd-Nd nanocluster 4 shows NIR luminescent sensing of nitro explosives. The luminescence quenching constant of 4 to 2-NP is 3,020 M−1, which is much larger than others (from 225 to 2,240 M−1). The detection limit of 4 to 2-NP is 14.70 μM, indicating that 4 has a high sensitivity for this explosive at the ppm level.</p><!><p>XY and SH design the Cd-Ln nanoclusters. HC, WJ, DJ, DS, BY, FW and LZ finish the experiment.</p><!><p>WJ was employed by company Guangzhou Sysmyk New Material Science & Technology Co., Ltd. The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.</p><!><p>Funding. The work was supported by the National Natural Science Foundation of China (No. 21771141 and 51025207).</p><!><p>The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fchem.2019.00139/full#supplementary-material</p><!><p>Experimental and characterization details, additional figures and tables, and CIF files (CCDC 1865266-1865269 for 1-4).</p><p>Click here for additional data file.</p>

PubMed Open Access

Structural Alterations of the FAS Gene in Cutaneous T-Cell Lymphoma (CTCL)

FAS (TNF receptor superfamily member 6, also known as CD95) plays a major role in T-cell apoptosis and is often dysregulated in CTCL. We searched for structural alterations of the FAS gene with the potential to affect its function. Although several heterozygous FAS promoter single nucleotide polymorphisms (SNPs) were detected, the only homozygous one was the -671 GG SNP present in 24/80 CTCL cases (30%). This SNP maps to an interferon response element activated by STAT-1. EMSA and supershift EMSA showed decreased CTCL nuclear protein/STAT-1 binding to oligonucleotides bearing this SNP. Luciferase reporters showed significantly less interferon-alfa responsive expression by FAS promoter constructs containing this SNP in multiple CTCL lines. Finally, FAS was upregulated by interferon-alfa in wildtype CTCL cells but not those bearing the -671 GG SNP. These findings indicate that many CTCL patients harbor the homozygous FAS promoter -671 GG SNP capable of blunting its response to interferon. This may have implications for CTCL pathogenesis, racial incidence and the response of patients to interferon-alfa therapy. In contrast, functionally significant mutations in FAS coding sequences were detected uncommonly. Among CTCL lines with the potential to serve as models of FAS regulation, FAS-high MyLa had both FAS alleles, FAS-low HH was FAS-hemizygous and FAS-negative SeAx was FAS-null.

structural_alterations_of_the_fas_gene_in_cutaneous_t-cell_lymphoma_(ctcl)

INTRODUCTION<!>Cells and Lesional Tissues<!>Flow Cytometry<!>Immunohistology<!>Cytogenetic and FISH Analysis<!>Sequencing of FAS Exons and Promoter<!>Electrophoresis Mobility Shift Assay (EMSA) and Supershift EMSA<!>Luciferase Reporter Constructs<!>Statistical Analysis<!>Molecular analysis of the FAS promoter region in CTCL demonstrates frequent germline single nucleotide polymorphisms (SNPs)<!>The -671 G SNP reduces STAT-1 binding to the FAS promoter<!>Luciferase reporter constructs demonstrate that the FAS promoter SNPs detected in CTCL can reduce FAS promoter function<!>The -671 G SNP reduces the response of the FAS promoter to interferon-alfa<!>The -671 GG genotype correlates with resistance to interferon-alfa induced upregulation of FAS protein<!>Molecular analysis of the FAS coding region in CTCL demonstrates significant mutations in only a small minority of cases<!>Cytogenetic and FISH analyses demonstrate chromosome 10 losses in CTCL lines with low or absent FAS expression<!>DISCUSSION<!>Conclusions<!>EMSA and supershift EMSA show reduced CTCL nuclear protein/STAT-1 binding to FAS promoter oligonucleotide probe bearing the -671 G SNP<!>FAS promoter SNPs can alter basal promoter activity<!>No simple correlation between basal FAS protein expression and -671 SNP genotype<!>Luciferase reporter constructs bearing the FAS promoter -671 G SNP exhibit reduced response to interferon-alfa<!>CTCL cell bearing the FAS promoter -671 GG SNP do not upregulate FAS protein in response to interferon-alfa<!>Karyotyping and FISH analysis of CTCL lines<!>

<p>Cutaneous T cell lymphoma (CTCL) is a neoplasm of well differentiated CD4+ memory T cells belonging to the skin associated lymphoid tissue (SALT) [1]. It includes mycosis fungoides (MF) and its erythrodemic and leukemic variant, the Sézary syndrome (SS). There are multiple lines of evidence supporting the hypothesis that early CTCL is primarily a lymphoaccumulative disorder rather than a lymphoproliferative disorder, i.e., that tumor cells persist and accumulate primarily due to defective apoptosis rather than enhanced proliferation. This evidence includes the indolent clinical behavior of early CTCL, its resistance to therapy that targets rapidly proliferating cells, its relatively low proliferative rate as assessed by mitotic index or Mib-1/Ki-67 expression, and its low apoptotic rate as assessed by terminal dUTP nick-end labeling (TUNEL) assay [2,3].</p><p>One of the major systems mediating apoptotic activity in T cells is the FAS pathway [4]. FAS dysregulation by CTCL tumor cells has been reported in a variable proportion of cases using a variety of immunohistological, flow cytometric and PCR techniques [3,5-13]. In prior studies of CTCL, we determined that there is a mechanistic connection among FAS transcript level, expression of FAS protein on the cell surface, and functional sensitivity to FAS-mediated apoptosis in vitro [3]. However, structural factors affecting FAS transcript level and integrity in CTCL are largely unexplored. This set the stage for the current study in which we analyzed the primary structure of the FAS gene (see Supplemental Figure 1) in order to search for alterations with the potential to influence FAS expression.</p><!><p>CTCL- derived (MyLa, HH, Hut-78, SeAx, SZ4, MJ) and other T-cell lines (Jurkat, JFL) were obtained from multiple sources and cultured as reported previously [3]. CTCL lesional skin and involved blood samples were obtained from our local cutaneous lymphoma tissue bank. Seven blood samples from SS were generously provided by Dr. Alain Rook (University of Pennsylvania). Specimens were collected with informed consent and Institutional Review Board approval.</p><!><p>Surface FAS expression by T-cell lines was determined by staining with FITC or PE conjugated anti-FAS monoclonal antibody DX2 (Becton Dickinson, San Jose, CA). Isotype-matched monoclonal antibodies of irrelevant specificity were used as negative controls as described previously [3]. For interferon experiments, 2 × 105 cells were treated with 100u/ml of interferon-α2b (Merck & Co., Inc., Whitehouse Station, NJ) for 48 hours before staining for surface detection.</p><!><p>We used a 3-stage murine monoclonal antibody/biotinylated goat anti-mouse IgG/avidin-HRP immunoperoxidase method applied to acetone-fixed frozen sections to assess FAS expression by CTCL and inflammatory skin disease controls. Two different anti-FAS monoclonal antibodies were used: clone APO-1-1 (Alexis, Farmingdale, NY) and clone DX2 (Dako, Carpinteria, CA) [3].</p><!><p>Cytogenetic analysis was performed as described previously [10,14]. Fluorescence in situ hybridization (FISH) analysis was performed according to the ACT Cytogenetics Laboratory Manual [14,15]. FAS probe was made using established procedures [16] by labeling the BAC Clone RP11-399O19 (chr10:90,718,801-90,775,625), which spans 56.8 kb including FAS promoter, all of the FAS exons/introns but not the nearby gene PTEN. The control probes for the 10 centromere and PTEN were purchased from Vysis (Des Plaines, IL). The 10 centromere probe hybridizes to alpha satellite sequences specific for the chromosome 10 centromere, while the PTEN probe contains sequences that span the PTEN gene at both the 5′ and 3′ ends.</p><!><p>As described previously [3], we used genomic DNA to amplify and sequence the FAS promoter and coding regions (Supplemental Figure 1). Nucleotide numbering of the polymorphisms in the FAS promoter was based on GenBank (accession # X87625) and related reports [17,18]. Coding region and promoter primers are shown in Supplemental Tables 1a, 1b and 1c, respectively. Some samples showing heterozygous FAS promoter polymorphisms by direct sequencing were further confirmed by DNA cloning and subsequent re-sequencing.</p><!><p>Nuclear extracts were prepared from MyLa cells (5 × 106/ml) using the NE-PER nuclear and cytoplasmic extraction reagents kit (Pierce, Rockford, IL). Double-stranded oligonucleotide FAS promoter probes (-663 to -683) were chemically synthesized with wildtype A or SNP G at the -671 site. Probes were labeled using the Biotin 3' end DNA labeling kit (Pierce, Rockford, IL). The sequence of this oligonucleotide, 5′-TGTCCATTCCAGA/GAACGTCTG-3′, contains a GAS binding site. EMSA was performed using the LightShift Chemiluminescent EMSA kit (Pierce, Rockford, IL). DNA-bound protein was identified by supershift EMSA with an anti-STAT-1 antibody (α-p91 and β-p84, Santa Cruz Biotechnology, Santa Cruz, CA).</p><!><p>Generation of DNA fragments with SNPs was performed using the QuikChange II site-directed mutagenesis kit (Stratagene, La Jolla, CA). The constructs and primers used to generate them are shown in Supplemental Table 2. Wildtype and SNP constructs were sequenced to confirm the appropriate promoter and transcription initiator sequences were present and then ligated into pGL3 luciferase basic vector. Plasmid carrying the β-galactosidase gene (2ug/sample) was co-transfected as an internal control. Cells were stimulated by PMA (100ng/ml) immediately after transfection. For interferon stimulation assays, 2 × 106 cells were treated with interferon-α2b (IFN) (100 U/ml) at the time of transfection.</p><!><p>Statistical analysis for the luciferase assays was performed using Student's t-test. A two-tailed p-value <0.05 was considered statistically significant and is represented as (*) in the figures. Analyses utilized SAS statistical software version 9.2 (SAS Institute, Inc. Cary, NC). Statistical analysis of FAS promoter genotypes among CTCL patients compared to controls was performed using the Chi-square method and two-tailed p values.</p><!><p>To search for mutations in regulatory regions of the FAS gene in CTCL, we used automated nucleotide sequencing of genomic PCR products to detect alterations of the 1781 bp FAS promoter immediately upstream of the start codon. We analyzed this region in 29 CTCL cases including 9 early MF (stages IB-IIA), 7 advanced MF and 10 SS (stages IIB-IVA), and 3 CTCL-derived cell lines: MyLa, HH and SZ4. PCR of FAS targets was negative in the SeAx CTCL line, consistent with absence of the FAS gene by FISH analysis (see below). As shown in Table 1, several single nucleotide polymorphisms (SNPs) at known germline SNP sites were detected among 24/29 CTCL cases (83%). Interestingly, each of these SNPs involved a transcription factor binding site including SP1 (-1378, RefSNP ID: rs2234767), AP1 (-1092, RefSNP ID: rs9658675), YY1 (-691, RefSNP ID: rs2234768), GAS (-671, RefSNP ID: rs1800682) and TEF (-436, RefSNP ID: rs9658676). All of these SNPs were heterozygous except for 8/29 CTCL cases (28%) that contained the homozygous -671 GG SNP. Other than these SNPs, no other mutations of the FAS promoter were identified by PCR. This indicates that somatic FAS promoter mutations are absent or rare in CTCL and do not accumulate with disease progression or in response to therapy. Consistent with our results using primarily lesional skin samples, a study of leukemic Sézary cells detected somatic mutations in the FAS promoter only very rarely [9].</p><p>In addition to searches of GenBank for FAS promoter polymorphisms, we also scanned the NCBI dsSNP database. This includes SNP maps from the Human Genome Project and SNP frequency data from the Allele Frequency Project of The SNP Consortium. Based on these searches, ten germline SNPs have been mapped within the FAS promoter, ranging from 0.5-50% in allelic frequency within the general population. In addition to the two commonest polymorphisms registered in GenBank (-1378, -671), the three other mutations we detected in CTCL are also reported germline SNPs (-1092, -691, -436), consistent with the interpretation that these are all germline polymorphisms rather than acquired somatic mutations in the CTCL tumor clone.</p><p>To assess further the true prevalence of the homozygous -671 GG SNP, we used genomic PCR and automated nucleotide sequencing to determine the genotype of the -671 site in another 51 CTCL cases. These included 10 early MF (stages IA-IIA), 10 advanced MF and 29 SS (stages IIB-IVA), and two CTCL lines (Hut-78 and MJ). To best match our initial data set of 29 cases (27 of which were Caucasian), all additional patients were also Caucasian. In aggregate, the -671 GG SNP was detected in 24/80 CTCLs (30%). The genotypic distribution among the 80 CTCL cases was 23 AA, 33 AG and 24 GG.</p><p>Some of our CTCL cases exhibited more than one FAS promoter SNP. All of these multiple SNPs were heterozygous. Four cases had the -1378A/-671G combination. One case each had the -1092A/-671G or -1378A/-671G/-436A combination. Interestingly, the FAS promoter in the Jurkat T cell line was heterozygous for C/T within the YY1 element at nt -691, as confirmed by cloning and sequencing (Supplemental Figure 2a). Although not previously reported in Jurkat, this is a known SNP found in approximately 20% of the general population.</p><!><p>We used electrophoretic mobility shift assays (EMSA) to determine the effect of the -671 G SNP on the binding of nuclear proteins extracted from the MyLa CTCL line. As shown in Figure 1, FAS promoter oligonucleotide probe containing the G SNP binds less nuclear protein than the wildtype A probe. Furthermore, anti-STAT-1 antibody supershifted the protein-probe complex, indicating that the bound nuclear protein is rich in STAT-1.</p><!><p>To determine if FAS promoter mutations can affect FAS transcription (and thereby FAS protein expression and apoptotic sensitivity), luciferase reporter constructs containing the homozygous equivalent of the FAS promoter SNPs detected among our CTCL samples were tested for transcriptional activity in CTCL and other T-cell tumor lines. Reporter activity was compared to the homozygous equivalent of the wild-type promoter sequence. As shown in Figure 2, the wild-type sequence generally showed the greatest activity except for constructs containing the -436A SNP which exhibited enhanced activity. This confirmed the ability of our assay to detect both increases and decreases in reporter activity. The -1378 solitary SNP had little impact on promoter function. Otherwise, constructs containing individual SNPs trended toward lower luciferase activity with statistically significant (p < 0.05) reductions relative to wild-type promoter in 5/5 CTCL and 2/2 Jurkat cell lines for constructs containing the -671 G SNP, and in 3/4 CTCL lines for the -1092 A SNP.</p><!><p>The T-cell lines used in this study have the following -671 FAS promoter genotype: AA (Jurkat, JFL), AG (MyLa, Hut-78, SZ4), GG (MJ), hemizygous A (HH) and FAS-null (SeAx). As shown in Figure 3, there is no simple correlation between these genotypes and the level of basal FAS expression. However, the -671 SNP lies within a "gamma activating site" (GAS) that is responsive to interferons that act through the JAK/STAT pathway to activate STAT transcription factors that bind to this site and regulate FAS promoter function. To determine the impact of the -671 G SNP on FAS promoter function, multiple CTCL lines were transfected with luciferase reporter constructs containing either -671 wildtype A or SNP G in the FAS promoter (Figure 4). Confirming our earlier experiments shown in Figure 2, baseline promoter function was significantly reduced in the -671 G constructs relative to the wildtype construct in all five CTCL lines (p < 0.04). Furthermore, relative to wildtype, the -671 G construct also showed a significantly reduced response to interferon-alfa in all five CTCL lines (p < 0.03).</p><!><p>To confirm that the reporter data were relevant to actual FAS protein expression, we determined the effect of interferon-alfa on cell surface FAS expression by neoplastic T cells bearing either the homozygous -671 GG SNP or the homozygous -671 AA wildtype genotype. As shown in Figure 5, CTCL line MJ (-671 GG) had no response to interferon while freshly isolated SS blood cells from patient LM (-671 AA) and the Jurkat line (-671 AA) showed significant FAS upregulation (p < 0.05). Even CTCL line HH (hemizygous for -671A) was able to respond significantly to interferon, underscoring the importance of even a single wildtype allele for interferon responsiveness.</p><!><p>A schematic diagram of the FAS gene is shown in Supplemental Figure 1. We used a combination genomic PCR, cDNA PCR and nucleotide sequencing to screen for mutations in all nine FAS exons in 20 cases of CTCL (9 early MF (stages IB-IIA), 7 advanced MF and 1 SS (stages IIB-IVA), and 3 CTCL cell lines: MyLa, HH, SZ4). Genomic DNA was sequenced in each case to maximize detection of exon mutations that might be under-represented in cDNA due to surveillance mechanisms such as nonsense modulated mRNA decay. The tolerability of amino acid substitutions was calculated using the SIFT score which represents the normalized probability that the amino acid change will be tolerated functionally [19]. Based on this analysis, we found functionally significant mutations in only 2/20 samples (Table 2 and Supplemental Figure 2b). These involved Exon 9 frameshifts that led to a truncated FAS protein lacking an intact death domain and an Exon 8 amino acid substitution predicted to interfere with FAS function. The few other amino acid substitutions detected were not predicted to be functionally significant (Table 2).</p><!><p>In order to better utilize existing CTCL lines as models for future studies of FAS regulatory pathways, we performed an in-depth characterization of FAS gene copy number in selected CTCL lines. Cytogenetic analysis indicated the presence of the chromosome 10 pair and FAS gene (10q24.1) in FAS-high MyLa, absence of one 10q region in FAS-low HH, and absence of the chromosome 10 pair in FAS-negative SeAx (see Figure 6). Fluorescent in situ hybridization (FISH) analysis was performed on two hundred interphase nuclei from each T-cell line using a probe specific for FAS together with control probes. There was complete absence of FAS genes in SeAx. Nevertheless, FISH analysis of SeAx detected signals for the chromosome 10 centromere and the PTEN gene (10q23) which is situated near the FAS gene locus. None of the SeAx nuclei showed signals for the FAS gene while positive control cells demonstrated two FAS gene signals in each cell. Therefore, PTEN signals are from disrupted chromosome 10 material retained in SeAx. SeAx cells were often tetraploid with four PTEN and four chromosome 12q control signals. HH demonstrated only one copy of FAS together with two control signals in 133 cells (66.5%), and two copies of FAS with four copies of the control gene in 67 tetraploid cells. These results indicate that one copy of FAS has been lost, and that tetraploidy is about 33%. MyLa demonstrated two copies of FAS and 2 control signals in 184 cells, while 10 cells had 4 FAS and 4 control signals, consistent with minor tetraploidy. Six cells had only a single control signal but still 2 copies of the FAS gene. These results are consistent with presence of two copies of the FAS gene overall.</p><p>Jurkat (which expresses FAS at moderate levels) demonstrated two copies of FAS in 194 cells, 4 of which had lost the control signal, as well as 6 tetraploid cells with four copies of both FAS and the 12q control. JFL (a FAS-low variant of Jurkat) demonstrated two copies of FAS together with two control signals in 155 cells (77.5%) and 4 copies of each in 45 tetraploid cells (22.5%). There were 3 tetraploid cells that lost either one copy of the control (1 cell) or one copy of FAS (2 cells). Therefore, there was generally no evidence of FAS loss in either Jurkat or JFL.</p><p>In aggregate, these findings indicate that existing CTCL lines can serve as models for roughly normal (MyLa, Jurkat, JFL), hemizygous (HH) and null (SeAx) FAS gene complements. In particular, the FAS-null SeAx line has the potential to serve as a substrate for functional studies of transfected FAS gene variants.</p><!><p>Because most promoter mutations result in reduced or lost function, we hypothesized that germline SNPs involving the FAS promoter in CTCL could potentially contribute to FAS dysregulation via their impact on FAS transcription, especially since all the SNPs we detected were within transcription factor binding sites. Although CTCL has not been studied previously, prior work has shown that mutations in AP1, GABP and GAS elements of the FAS promoter in other cell types can influence FAS promoter function and consequent levels of tumor cell FAS transcript and protein as well as circulating soluble FAS protein [4,20-27]. As shown in Figure 2, we used CTCL lines to provide proof of principle for this concept in CTCL by demonstrating decreased FAS promoter function in luciferase reporter constructs corresponding to homozygous versions of the actual transcription factor binding site SNPs we detected among our CTCL samples. When combined with the reduced STAT-1 binding (Figure 1), the blunted response of both FAS reporters and FAS protein expression to interferon-alfa (Figures 4 and 5) and a similarly diminished response to interferon-gamma reported previously [20-22,26,27], it is likely that FAS promoter SNPs like the homozygous -671 GG are relevant to the tumor response to endogenous and therapeutic interferons whose mechanism of action involves the regulation of FAS expression via STAT transcription factor binding. Furthermore, as shown in Figure 5, the presence of even one copy of the wildtype allele in the hemizygous HH CTCL line was sufficient to maintain sensitivity to upregulation of FAS by interferon-alfa.</p><p>MyLa and Hut-78 (both heterozygous for the -671 SNP) expressed high levels of FAS transcript and protein (Figure 3). This suggests that the functional impact of a single copy of the -671 G SNP found in several heterozygous CTCL cases is likely to be minimal. Furthermore, although the level of FAS protein expressed by MJ (which is homozygous for the -671 G SNP) was well below that of normal blood lymphoid cells (Figure 3), it was still greater than that of other T-cell lines containing one or both wildtype alleles (Jurkat, HH and SZ4). Therefore, the status of the -671 SNP is unlikely to control basal FAS expression which is probably the net result of a multiplicity of factors impacting promoter function. A case in point is the JFL line which is a very FAS-low variant of Jurkat (Figure 3) yet shares the same -671 AA genotype. This interpretation is consistent with our prior immunohistologic and flow cytometric studies which have shown that FAS protein expression is low in a larger proportion of CTCL cases and lines than the 30% that bear the homozygous -671 GG SNP [3]. It is clear from our own work [28] and others [9] that additional factors such as DNA methylation can affect basal FAS expression in CTCL. Therefore, the -671 GG SNP, which does not involve CpG islands subject to regulation by methylation, is likely more important for governing responses to interferons than governing basal FAS expression in general.</p><p>Because the -671 GG SNP appears to reduce the responsiveness of the FAS promoter to interferons, it may promote a TH1/TH2 imbalance favoring TH2 cells which are normally down-regulated by interferon produced by TH1 and other cells. CTCL has been postulated to arise from the TH2 T-cell subset, so this imbalance might provide fertile soil for the development of CTCL. The -671 GG SNP may also help explain racial differences in CTCL incidence. Relative to Caucasians, African Americans have an increased incidence of CTCL (~1.5-fold) and also an increased frequency of the -671 GG SNP (~ 4-fold) [29,30] (African American frequencies at: http://www.ncbi.nlm.nih.gov/projects/SNP/snp_ref.cgi?rs=1800682). Given these observations, it will be important to perform follow-up studies to determine the prevalence of the -671 GG SNP among both Caucasians and African-Americans with CTCL relative to controls, the impact of this SNP on the clinical response of CTCL patients to treatment with interferons, and the impact of this SNP on the response of normal T cells, melanomas and other tumors to interferons and other factors that act through STAT transcription factors. For example, among Caucasian CTCL patients in our current study, the 30% prevalence of the -671 GG SNP (24/80) was significantly higher (p < 0.03) than its 20% prevalence among Caucasian controls aggregated from multiple sources (217/1097) [31] (and NCBI-SNP database: http://www.ncbi.nlm.nih.gov/projects/SNP/snp_ss.cgi?subsnp_id=24084001,id=66857087,andid=13453099).</p><p>As shown in Table 2, functionally significant mutations (presumably somatic rather than germline) affecting the FAS coding region were relatively uncommon (2/20 CTCLs). The low prevalence of these FAS exon mutations is consistent with a few prior studies and suggests that overall they are relatively uncommon in CTCL [32,33]. This indicates that while FAS coding region mutations causing abnormal protein structure might account for reduced FAS protein detection and/or resistance to apoptosis in a small minority of CTCL cases, they cannot explain the phenomenon generally. Nevertheless, treatments aimed at restoration of silenced but otherwise normal FAS could prove problematic in this subset of CTCL cases.</p><p>Cytogenetic and FISH analysis showed gross FAS gene abnormalities in 2/3 CTCL lines. Presumably, these major genetic defects were somatic alterations restricted to the donors' CTCL tumor cells. Nevertheless, there was an interesting correlation between these genetic abnormalities and the level of FAS cell surface protein expression. In contrast to the presence of the chromosome 10 pair and its FAS alleles (10q24.1) in FAS-high MyLa, there was absence of one 10q region in FAS-low HH, and complete absence of FAS genes in FAS-negative SeAx (see Figure 3 for FAS levels of these lines). These data provide one potential explanation for the low and absent FAS protein expression characteristic of HH and SeAx, respectively [3]. They also define key features of these cell lines that will make them useful models for future studies of the FAS pathway. Their potential clinical relevance is underscored by recent studies that have used comparative genomic hybridization arrays to document recurrent loss of heterozygosity of the chromosome 10q24 region containing the FAS gene in SS tumor cells [34,35]. Of particular relevance to our current findings, loss of heterozygosity could convert a tumor with a heterozygous -671 AG FAS promoter genotype into one with a hemizygous -671 G genotype that would acquire interferon-alfa resistance due to loss of its only wildtype allele.</p><!><p>Our study shows that many CTCL patients harbor the homozygous FAS promoter -671 GG SNP capable of blunting the promoter's response to interferon mediated by STAT1. In contrast, functionally significant mutations in FAS coding sequences were detected uncommonly. Among CTCL lines with the potential to serve as models of FAS regulation, FAS-high MyLa had both FAS alleles, FAS-low HH was FAS-hemizygous and FAS-negative SeAx was FAS-null.</p><!><p>Lanes 3 and 4 show that labeled wildtype FAS promoter oligonucleotide probe A forms complexes better with CTCL (MyLa) nuclear protein extract (NE) than does labeled oligonucleotide probe G bearing the -671 G SNP. Lanes 5 and 6 show that anti-STAT-1 antibody (ab) supershifts these same complexes, consistent with the presence of STAT-1 transcription factor. Lanes 1 and 2 are controls containing only labeled probes that are much smaller and migrate off bottom of gel. Lanes 7 and 8 are also controls that show 200-fold excess of unlabeled probes (200X) can block binding of labeled probes A and G to nuclear protein extracts.</p><!><p>Luciferase reporter constructs show that most individual SNPs and SNP combinations resulted in decreased promoter function. Y axis shows luciferase levels. Asterisks (*) mark results that were statistically significant relative to wildtype.</p><!><p>Flow cytometric analysis of neoplastic T cell lines with different -671 SNP genotypes: AA (Jurkat, JFL), AG (MyLa, Hut-78, SZ4), GG (MJ), hemizygous A (HH) and FAS null (SeAx) as well as pooled mononuclear blood cells from three normal donors. Y axis shows cell surface FAS protein levels.</p><!><p>Asterisks (*) show that both baseline and interferon-stimulated activity of the -671 G SNP reporter were significantly less than that of the wildtype (-671 A) reporter. Y axis shows luciferase levels.</p><!><p>Flow cytometric analysis of neoplastic T cells with the homozygous -671 GG SNP (MJ CTCL line) compared to -671 AA wildtype (LM Sezary cells and Jurkat cell line) and -671 A hemizygous CTCL line HH. Presence of the -671 A wildtype allele but not the -671 GG genotype correlated with interferon responsiveness. Y axis shows cell surface FAS protein levels. Asterisks (*) mark results that were statistically significant relative to no-treatment controls.</p><!><p>FAS targeted FISH analysis of multiple T-cell lines, and karyotypic analysis of MyLa, HH and SeAx CTCL lines. A shows FISH results of a FAS-positive control cell line CLG 2144, both FAS (red) and Chromosome12q control (green) signals are present. B and C show the FISH results of SeAx cell line. Although signals are detectable for PTEN (red) (B) and control Chromosome 12q (green) (B and C), there is no FAS signal (red) (C). D, E, F and G show FISH results for MyLa, HH, Jurkat and JFL, respectively. H, I and J are karyotyping results for MyLa, HH and SeAx, respectively. As indicated by the red arrows, Chromosome 10 is complete in MyLa; one of the long arms of Chromosome 10 is deleted in HH; and the intact chromosome 10 pair containing FAS is absent in SeAx.</p><!><p>Genotypic Frequencies of FAS Promoter SNPs in CTCL</p><p>TFBS: transcription factor binding site.</p><p>TFBS were searched using search tools: TESS (Transcription Element Search System: http://www.cbil.upenn.edu/cgi-bin/tess/tess) and TRANSFAC® Public database (http://www.generegulation.com).</p><p>Exon mutations found in 3/20 CTCL cases.</p><p>The mutation positions were designated from FAS mRNA sequence (GenBank access# NM_000043).</p><p>The SIFT score is the normalized probability that the amino acid change is tolerated. Not tolerated (<0.05); borderline tolerated (0.05-0.1); tolerated (>0.1).19</p>

PubMed Author Manuscript

Binding of YC-1 or BAY 41-2272 to Soluble Guanylyl Cyclase Induces a Geminate Phase in CO Photolysis

Soluble guanylyl/guanylate cyclase (sGC), a heme-containing heterodimeric protein of ~150 kDa, is the primary receptor for nitric oxide, an endogenous molecule of immense physiological importance to animals. Recent studies have identified compounds such as YC-1 and BAY 41-2272 that stimulate sGC independently of NO binding, properties of importance for the treatment of endothelial dysfunction and other diseases linked to malfunctioning NO signaling pathways. We have developed a novel expression system for sGC from Manduca sexta (the tobacco hornworm) that retains the N-terminal two-thirds of both subunits, including heme, but is missing the catalytic domain. Here, we show that binding of compounds YC-1 or BAY 41-2272 to the truncated protein leads to a change in the heme pocket such that photolyzed CO cannot readily escape from the protein matrix. Geminate recombination of the trapped CO molecules with heme takes place with a measured rate of 6 \xc3\x97 107 s\xe2\x88\x921. These findings provide strong support for an allosteric regulatory model in which YC-1 and related compounds can alter the sGC heme pocket conformation to retain diatomic ligands and thus activate the enzyme alone or in synergy with either NO or CO.

binding_of_yc-1_or_bay_41-2272_to_soluble_guanylyl_cyclase_induces_a_geminate_phase_in_co_photolysis

<!>Supporting Information Available<!>

<p>Nitric oxide (NO) regulates numerous physiological processes in animals, including blood pressure, platelet aggregation, memory formation and tissue development.1 Soluble guanylyl/guanylate cyclase (sGC), a heme-containing heterodimeric protein of ~150 kDa, is the primary NO receptor. The protein is composed of two subunits, α and β, that are evolutionarily related and have similar domain structures. The N-terminus of each subunit begins with a predicted H-NOX domain followed by a PAS domain, a coiled-coil region and a cyclase domain. A single catalytic site is formed at the interface of the two C-terminal catalytic domains and catalyzes the conversion of GTP to 3′,5′-cyclic GMP (cGMP) and pyrophosphate (PPi). Ferrous (FeII) heme resides in the N-terminal H-NOX domain of the β subunit and is coordinated through His 105. NO binding to the distal side of the sGC heme leads to release of the proximal histidine, formation of a five-coordinate nitrosyl complex and an ~200 fold stimulation of cyclase activity. CO binding also stimulates cyclase activity, but by only ~5-fold in the absence of allosteric activators. However, on binding allosteric activators such as synthetic compound YC-1 (3-(5′- hydroxymethyl-2′-furyl)-1-benzylindazole) or its derivative BAY 41-2272, CO stimulates sGC catalysis to nearly the same extent as NO, but without cleavage of the proximal histidine bond.2 Molecules that stimulate sGC have long been sought for the treatment of cardiovascular and related diseases. Organic nitrates and other NO releasing compounds are commonly used for this purpose but suffer from the nonspecific side reactions of NO and a tendency for tolerance to develop with prolonged usage. Compounds that activate sGC in a NO-independent manner, such as YC-1 or BAY 41-2272, may therefore provide a substantial therapeutic advantage.3</p><p>Very little is known about how YC-1 functions or where it binds in the protein. To address this question, we developed a novel expression construct that contains the N-terminal two thirds of α1/β1 sGC from the tobacco hornworm (Manduca sexta) and demonstrated that this protein retains the YC-1 binding site despite the absence of a cyclase domain.4 YC-1 binding leads to increased affinity for CO and decreased release rates for both CO and NO, suggesting that YC-1 binding somehow traps CO and NO in the heme binding pocket.4 Here, we demonstrate, through direct observation of a geminate recombination phase for CO photolysis, that binding of YC-1 or BAY 41-2272 alters the heme distal pocket.</p><p>We examined the transient kinetics of photolyzed CO using a truncated Manduca sGC containing residues α1 1-471 plus an N-terminal His-tag, and β1 1-400 with a ferrous heme, together referred to as msGC-NT. In a typical experiment, 3 μM msGC-NT was present in a 50 mM potassium phosphate buffer (pH 7.4) containing 100 mM KCl and 5% glycerol. The solution was purged with CO gas for 20 minutes. Binding of CO to msGC-NT shifted the Soret absorption band maximum from 433 nm to 425 nm.4 Photolysis of CO resulted in initial loss of A425 amplitude, initial appearance of A433, and complete recovery of both absorbances on the milliseconds time scale with the same kinetics (Figure 1, top panel). The rate of slow recovery was dependent on the CO concentration, indicating that rebinding was from CO free in solution. The bimolecular rate constant of 41.5 ± 4.0 mM−1 s−1 was obtained for this slow process (Figure S1 in Supporting Information), a value similar to that previously determined from rapid mixing in a stopped-flow spectrophotometer (28 mM−1 s−1).4</p><p>Addition of YC-1 to the protein led to a 2 nm blue shift in the A425 Soret maximum (Figure S2) and a loss of photolysis amplitude in the milliseconds time scale (Figure 1, bottom panel). Possible explanations for the loss of signal amplitude include a change in photolytic quantum yield or the appearance of a geminate phase such that some of the photolyzed CO molecule was unable to escape from the heme pocket before recapture by the heme. To investigate these possibilities, we undertook laser photolysis on the nanosecond time scale, which is typical for CO geminate recombination in heme proteins (Figure 2). In the absence of an allosteric activator, a single slow phase is observed for recovery of the photolyzed species. In the presence of YC-1 (50 µM) or BAY 41-2272 (5 µM), a new, faster phase appears (Figure 2, top panel) with a characteristic lifetime of ~20 ns that behaves as a single exponential, suggesting a single site mechanism (Figure 2, bottom panel). The extent of photolysis (i.e. the yield of free Fe(II)) is similar in the presence or absence of YC-1; however, the faster phase (1) comprises ~50% of the total amplitude. Both YC-1 and BAY 41-2272 gave rise to a fast phase with similar rate constants (Table 1). Moreover, for both compounds, the measured fast phase rate constants are independent of CO concentration (Table 1), indicating that this fast decay is due to geminate recombination. We also examined full-length Manduca sGC and found that CO stimulates as well as NO in the presence of YC-1 or BAY 41-2272, but not in their absence (Figure S3), behaving much like its mammalian counterparts.2</p><p>Taken together, these data strongly indicate that YC-1 or BAY 41-2272 induce a change in sGC conformation that blocks escape of CO from the distal heme pocket and facilitates catalysis. Electrostatic effects, such as those that govern dioxygen binding to myoglobin, are unlikely to influence NO and CO escape due to the apolar nature of the Fe-NO and Fe-CO complexes.5 However, a change in protein conformation can have a substantial effect on NO and CO escape as demonstrated by the nitrophorins, proteins that make use of a change in protein conformation for transporting NO.6</p><p>It remains unknown where YC-1 and related compounds bind to the protein, and whether their binding modes are similar to that of an endogenous compound. Several binding sites have been proposed, including in the catalytic domain,7 the α1 subunit,8 or within the heme pocket.9 Our present and previous results4 confirm that activator binding occurs somewhere in the N-terminal two-thirds of the protein; however, our results do not rule out a second binding site in the catalytic domain. A recent study indicates that both YC-1 and BAY 41-2272 induce a change in the EPR and IR spectra of five-coordinate nitrosyl heme for the full-length rat protein but not for a truncated version of the protein containing only the β1 heme domain, suggesting binding does not occur in the heme domain.10 We have proposed that the YC-1 binding site may be in the α1 H-NOX domain, which our modeling studies indicate may have a similar fold to that of the β1 H-NOX domain but without heme.4 Support for this model also comes from photoaffinity labeling using a modified BAY compound containing an azido group (BAY-9491), which was found covalently linked to α1 Cys 238,8 a residue near in sequence to the putative α1 H-NOX domain.4</p><p>Binding of both YC-12,9,11 and nucleotide12 to full-length sGC affects the sGC heme pocket, making their individual contributions to function difficult to resolve. For example, binding of GTP and YC-1 to full-length bovine sGC leads to three CO association rates, a slower bimolecular phase (90 mM−1s−1), a faster bimolecular phase (97 µM−1s−1) and a geminate recombination phase (not characterized), while in the absence of YC-1, only the slowest phase is seen.9 In contrast, msGC-NT displays only a slow bimolecular phase in the absence of YC-1 (41.5 mM−1 s−1), and both the slow phase and a second geminate recombination phase in the presence of YC-1 that is ~1000 times faster than the bovine bimolecular fast phase. Importantly, msGC-NT, which does not contain the cyclase domain, does not respond to cGMP, GTP or ATP, indicating that the YC-1 binding site is distinct from that for nucleotide.4 We conclude from this that both YC-1 and nucleotides induce a change in the heme pocket of the fulllength protein, but do so independently of one another, resulting in multiple ligand binding effects.11,12 By isolating the YC-1 binding site in msGC-NT through removal of the cyclase domain, we provide a powerful new tool for investigating NO-independent stimulation of sGC and the discovery of improved compounds for the treatment of cardiovascular disease.</p><!><p>Experimental procedures and Figure S1–Figure S3. This material is available free of charge via the Internet at http://pubs.acs.org.</p><!><p>Effect of YC-1 on photolysis of CO from msGC-NT. Top panel: Photolysis by a 386 nm nitrogen dye laser pulse leads to bleaching of the A425 band, appearance of the A433 band, and a complete recovery exhibiting a single phase. Monitoring was at 424 and 435 nm, respectively. Bottom panel: Addition of YC-1 leads to the loss of signal amplitude (A422) on the milliseconds time scale.</p><p>Nanosecond laser photolysis of CO from msGC-NT. Top panel: Photolysis by a 532 nm Nd:YAG laser pulse in the presence of YC-1 or BAY 41-2272 yields fast and slow phases (red and green lines, monitored at 422 nm); in the absence of activator, only a single slow phase is evident (black line, monitored at 424 nm). Bottom panel: Single exponential fitting of the faster phase for the YC-1 bound sample.</p><p>Rate constants for CO geminate rebinding to msGC-NT</p><p>Saturated CO is of ~1 mM; 10% CO is of ~0.1 mM.</p>

PubMed Author Manuscript

Photochemical and thermal intramolecular 1,3-dipolar cycloaddition reactions of new o-stilbene-methylene-3-sydnones and their synthesis

New trans- and cis-o-stilbene-methylene-sydnones 3a,b were synthesized by transforming the trans- and cis-o-aminomethylstilbene derivative, obtained by reduction of corresponding o-cyano derivatives, into glycine ester derivatives (43 and 31% yield) followed by hydrolysis (90 and 96% yield), nitrosation and ring closure with acetic acid anhydride (30 and 40% yield). The products were submitted to photochemical and thermal intramolecular [3 + 2] cycloadditions to afford diverse heteropolycyclic compounds. Photochemical reactions afforded cis-3-(4-methylphenyl)-3a,8-dihydro-3H-pyrazolo[5,1-a]isoindole (11, 12.5% yield) and trans-3-(4-methylphenyl)-3a,8-dihydro-3H-pyrazolo[5,1-a]isoindole (12, 5% yield). Thermal reactions afforded 3-(4-methylphenyl)-3,3a,8,8a-tetrahydroindeno[2,1-c]pyrazole (14, 50% yield) and 11-(4-methylphenyl)-9,10-diazatricyclo[7.2.1.02,7]dodeca-2,4,6,10-tetraene (15, 22% yield).

photochemical_and_thermal_intramolecular_1,3-dipolar_cycloaddition_reactions_of_new_o-stilbene-methy

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

<p>Sydnones belong to the group of five-membered heterocyclic compounds referred to as being "mesoionic" and have been widely studied since their discovery [1–5]. They can be represented as hybrids of a number of mesomeric ionic structures (Figure 1).</p><!><p>Resonance structures of the sydnone ring.</p><!><p>One of the most characteristic reactions of sydnones is the intermolecular 1,3-dipolar cycloaddition. In the presence of acetylenic or ethylenic dipolarophiles, sydnones undergo cycloaddition reactions, which can be induced thermally [4,6–7] or photochemically [8–17], giving different pyrazole and/or pyrazoline derivatives, depending on the applied dipolarophile (Scheme 1). Namely, sydnones are masked 1,3-dipoles that by photolysis give nitrile imine intermediates, or in thermal reactions react as cyclic azomethine imines.</p><!><p>Thermal and photochemical intermolecular [3 + 2] cycloadditions.</p><!><p>Intramolecular 1,3-dipolar cycloadditions of sydnone derivatives have not been as thoroughly investigated, and so far only a few examples are known [18–20]. Photochemically induced intramolecular 1,3-dipolar cycloadditions have been studied on 3,4-disubstituted sydnone derivatives [18–19] (Figure 2, A and B), wherein indolopyrazole and pyrazolobenzoxazine structures are formed (Figure 2, C and D).</p><!><p>Illustration of intramolecular [3 + 2] cycloadditions.</p><!><p>Heimgartner and coworkers also carried out the thermally induced reaction of 3-(o-allylphenyl)-4-phenylsydnone (A) and obtained the cycloadduct E (Figure 2) with the oxycarbonyl group remaining in the structure [18].</p><p>We have been studying photochemical reactions of conjugated heterostilbene derivatives in which the sydnone moiety is part of a heterostilbene [17] (1, Figure 3) or is directly attached at the ortho position to the stilbene 2 [21–23]. Upon photolysis of compound 1, where the sydnone moiety is part of a heterostilbene system, cis–trans isomerization was the main process, and no intramolecular cycloadducts were found owing to the unfavourable conformation of the formed intermediate in the trans configuration. The existence of the nitrile imine intermediate as a result of competitive photolysis of the sydnone moiety was confirmed on irradiation of 1 in the presence of acrolein and isolation of the pyrazoline derivative (F, Figure 3) [17]. In the case of stilbenylsydnones 2, where the sydnone moiety is directly connected to the ortho position of the stilbene, the cyclization of the formed nitrile imine intermediate leading to benzodiazepine ring closure (G, Figure 3) was the main intramolecular process [22]. In a continuation of our interest in the synthesis of heteropolycyclic compounds we extended our research to new stilbene-sydnone derivatives 3 (Figure 3). In such a system, where two chromophores, stilbene and sydnone, are divided by a methylene bridge, an intramolecular 1,3-dipolar cycloaddition and the formation of diverse polycyclic compounds could be expected. Herein we describe, for the first time, the synthesis of cis- and trans-3-(stilbenylmethyl)sydnones and their photochemical and thermal intramolecular transformations to heteropolycyclic structures.</p><!><p>Styryl-sydnone 1 and stilbenyl sydnone 2 and their photoproducts F and G, respectively; target molecules 3 in this work.</p><!><p>In the investigation of 3-{2-[2-(4-tolyl)ethenyl]phenyl}methylsydnones, 3a (trans) and 3b (cis), were prepared by a sequence of reactions (Scheme 2) starting from o-cyanotoluene (see Supporting Information File 1 for full experimental data). Bromination of o-cyanotoluene afforded 2-(bromomethyl)benzonitrile (4) [24], which was transformed to triphenylphosphonium salt 5 [25] followed by Wittig reaction with 4-methylbenzaldehyde to 2-(4-methylstyryl)benzonitrile (6a,b) [26]. The product was obtained as a mixture of 6a (trans isomer, 40%) and 6b (cis isomer, 60%).</p><!><p>Synthesis of the target molecules 3a and 3b.</p><!><p>The isomers were separated by column chromatography and further treated separately to achieve the final products in cis and trans configurations. Reduction of 6a (trans) or 6b (cis) with LiAlH4 in anhydrous ether afforded amino derivative 7a (trans, 94%) or 7b (cis, 93%). In the 1H NMR spectra new signals from the methylene protons appeared at 3.88 ppm (7a) and 3.62 ppm (7b) confirming the reduction. By further nucleophilic substitution, from 7a (trans) or 7b (cis) and ethyl bromoacetate, the ester 8a (trans) or 8b (cis) was prepared. On purification by column chromatography, the byproducts, obtained by disubstitution reaction of the amino compound, were separated, and the pure 8a (43%) or 8b (31%) was isolated. The obtained esters showed the presence of the carbonyl group at ~1740 cm−1 in the IR spectra and the carbonyl carbon at ~172 ppm in the 13C NMR spectra. The esters 8a or 8b were hydrolysed to the amino acid 9a (trans, 90%) or 9b (cis, 96%). The obtained amino acids were transformed to N-nitroso glycine 10a or 10b and, without isolation or further purification, were submitted to dehydration with acetic acid anhydride to give sydnone 3a (trans) or 3b (cis). After column chromatography the pure 3a (30%) or 3b (42%) was isolated. The best indication that the sydnone structures were formed was given by the singlet at ~6 ppm in the 1H NMR spectrum, characteristic for the proton H-4 in the sydnone ring, as well as those in the 13C NMR spectrum, namely the CH and CO sydnone carbons at ~94 and ~169 ppm, respectively.</p><p>The irradiation experiments with the trans isomer (3a) or cis isomer (3b) were performed in ~10−3 M benzene solution in a Rayonet reactor at 300 nm under anaerobic conditions (purged with argon). The absorption maximum of trans isomer (3a) is at 300 nm (ε 37453) and of cis isomer (3b) at 291 nm (ε 16228), thus upon irradiation both isomers were excited. The irradiation of the isomers was performed until full conversion. The irradiation of either the trans or cis isomer, or of the mixture of isomers, resulted in the formation of two products in the same mutual ratio, along with large amounts of unidentified high-molecular-weight products. Separation by column chromatography in combination with thin layer chromatography gave dihydropyrazolo-isoindoles, 11 (12.5%) and 12 (5%) (Scheme 3).</p><!><p>Photolysis of cis- or trans-3.</p><!><p>The structure of the photoproducts was determined by spectroscopic methods. The molecular ions of 11 and 12, m/z 248 in the mass spectra, indicate that the structures have lost CO2 relative to the starting compound. In the 1H and 13C NMR spectra, the signals are found in an area which is characteristic for saturated cyclic compounds. The major product is assigned to compound 11 based on the following data: In 1H NMR spectrum the signals at 4.98 ppm and 4.43 ppm are doublets with coupling constants of 15.6 Hz and are assigned to geminal protons G-1 and G-2. The other two signals at 4.92 and 4.24 ppm are also doublets but with coupling constants of 10.8 Hz. Based on the interaction of the proton at 5.88 with the proton at 4.92 ppm in the NOESY spectrum, the doublet at 5.88 is assigned to proton H-5, the doublet at 4.92 ppm to proton B and the doublet at 4.24 to proton C. Interaction of proton H-5 with proton B is also visible in HMBC spectrum. The signal at 6.53 ppm was assigned to proton A based on COSY interaction with proton C. The rather large high-field shift of the aromatic proton H-5 can be explained by an anisotropic effect of the tolyl group and thus confirms the cis orientation of protons B and C.</p><p>The structure of the minor photoproduct 12, different from the structure 11 only in trans orientation of the protons B and C, was also evident from NMR spectra by using 2D NMR techniques. The doublets at 4.90 ppm and 4.37 ppm with a coupling constant of 15.6 Hz are assigned to geminal protons G. The singlet at 6.57 ppm is assigned to proton A. The chemical shift of proton A is the same as in structure 11. The other two signals at 3.97 ppm and 4.70 ppm appear as singlets, and they are assigned to protons C and B, respectively, based on weak interactions in the COSY and NOESY spectra. Nevertheless, the proton H-5 of the fused benzene ring in structure 12 is in the multiplet together with other aromatic protons, which is in accordance with the proposed structure.</p><p>The structure of the photoproducts was confirmed by an additional experiment on the crude reaction mixture with DDQ (Scheme 4) in which, as expected for the predicted structures, the aromatization reaction took place forming the pyrazolo-isoindole 13. Compound 13 arose also on silica gel during the purification of either 11 or 12.</p><!><p>Aromatization with DDQ.</p><!><p>The irradiation of 3a or 3b until full conversion, as previously mentioned, produced a mixture of 11 and 12 along with decomposition and high-molecular-weight products. On shorter irradiation time (10 min, in benzene or acetonitrile) 3a (trans isomer) afforded, according to 1H NMR, the photomixture of predominantly 3b (cis isomer) with only traces of starting 3a and tricyclic photoproducts 11 and 12. Under the same irradiation conditions, 3b (cis isomer) as starting compound gave a photomixture of cis isomer and the newly formed product 11 in 1:1 ratio with only traces of 3a (trans isomer), along with some unidentified side products. The experimental results show that the trans-(3a) and cis-(3b) isomerize and react with different efficiency, and that the isomerization, as in the case of stilbene itself [27], is shifted toward the cis isomer. It follows that the reaction is stereospecific and that photoproduct 11 is formed from the cis configuration of the stilbene moiety and the photoproduct 12 from the trans configuration, although the formation of 12 via epimerization of 11 could not be eliminated. It is also evident that there are several competitive processes, which are summarized in Scheme 5.</p><!><p>Possible mechanism for the formation of the photoproducts.</p><!><p>On irradiation of 3a (trans) or 3b (cis) parallel competitive processes are in operation, namely, trans–cis and cis–trans isomerization of the stilbene moiety, and photolysis of the sydnone ring resulting in the formation of the nitrile imine intermediate. The nitrile imine species is, in intramolecular dipolar [3 + 2] cycloaddition, trapped by the cis- or trans-double bond of the stilbene, giving cycloadducts 11 or 12, respectively.</p><p>We also performed the thermal intramolecular reactions with the starting compounds 3a and 3b. Theoretically the intramolecular 1,3-dipolar cycloaddition of the sydnone moiety, acting as a masked azomethine dipole, and the double bond of the stilbene moiety could proceed in different ways. The orientation of the sydnone ring toward the π bond of the stilbene in combination with the double bond configuration can give various formal [3 + 2] intramolecular cycloadducts.</p><p>On heating of 3a (trans) in toluene until full conversion (4 h) one product in 50% yield was isolated from the reaction mixture after column chromatography. From the molecular ion (m/z 248) of the product and its 13C NMR spectrum it was obvious that in the cycloaddition CO2 elimination took place. Fragmentation of the product and the presence of an ion at m/z 220 suggests a structure in which the expulsion of nitrogen is possible. The structure 14 (Scheme 6) was determined by additional NMR techniques, NOE and HMBC interactions, and by single crystal X-ray structure analysis (Figure 4) of the crystal formed in an NMR tube by slow evaporation of the solvent.</p><!><p>Thermal reaction of trans-3.</p><p>ORTEP of compound 14.</p><!><p>Also, when compound 3b (cis) was refluxed in xylene (9 h) or toluene (19 h) only one cycloadduct 15 was isolated (Scheme 7), besides decomposition products, in 22% yield. The structure of 15 was determined by spectroscopic methods.</p><!><p>Thermal reaction of cis-3.</p><!><p>In the 1H NMR spectrum two pairs of geminal protons were found at 4.51 and 4.21 ppm (A, B) and at 3.71 and 3.35 ppm (E, D). The doublet at 3.95 ppm, coupled with one geminal proton (D), was assigned to proton C. In the 13C NMR spectrum, one of the five quaternary carbons is shifted to 176 ppm, which corresponds to an sp2-carbon in the vicinity of nitrogen. In the NOESY spectrum the interaction of protons A and B with an aromatic proton (H-2) at 7.00 ppm is seen, as well as the interaction of proton C with protons E and D. Since the NOE interaction is seen between protons A and E we concluded that protons A and E must lie on the same side of the six-membered ring. In addition, the interaction of proton C with tolyl (H-10) and H-5 protons was seen.</p><p>In order to explain the diverse structures (14 and 15) and their formation mechanism, we analysed the possible ways of intramolecular [3 + 2] cycloaddition relating to the arrangement of the sydnone ring towards the cis and trans double bond (Figure 5 and Figure 6).</p><!><p>Proposed stereochemical pathway of sydnone ring (CH–N) and trans- and cis-stilbene (α–β).</p><p>Proposed stereochemical pathway of sydnone ring (N–CH) and trans- and cis-stilbene (α–β).</p><!><p>As presented in Figure 5, the sydnone ring could be oriented to the double bond in such a way that the bonds are formed at the C(Sy)–α(St) and N(Sy)–β(St) positions, or, as presented in Figure 6, at the N(Sy)–α(St) and C(Sy)–β(St) positions. The favoured arrangement of the sydnone ring toward the cis and trans double bond, leading to the products, is the pathway presented in Figure 5. The regiospecific and stereospecific formation of the products 14 and 15 could be explained by this approach of the sydnone ring (Scheme 8). The cycloadducts, cA from trans isomer and cB from cis isomer, lose CO2 under the reaction conditions to afford intermediates 14A and 14 B, respectively. Owing to the favourable conformation in the case of biradical 15A, the 1,3-H abstraction and formation of the C–N double bond in product 15 is possible. In the biradical 14A the intramolecular hydrogen abstraction is not favourable, but 1,2-alkyl shift takes place followed by formation of the N–N double bond in product 14.</p><!><p>Possible formation of thermal products 14 (from trans-3) and 15 (from cis-3).</p><!><p>Monitoring the reaction by thin layer chromatography revealed that the [3 + 2] cycloaddition is much faster in the case of the trans isomer (3a). After the 4 h reflux of the toluene solution of the trans isomer, the 1H NMR spectrum of the crude reaction mixture showed complete conversion, while the cis isomer (3b) under the same conditions showed complete conversion only after 19 h. This evidence led us to believe that the formation of the "C–α/Ν–β" adduct cA proceeds via an energetically favoured transition state due to a possible secondary π–π interaction of the tolyl and carbonyl groups.</p><!><p>In photochemical and thermal intramolecular reactions the investigated compounds 3a and 3b, in which the stilbene and sydnone ring are bridged by a methylene group, show the characteristic reaction for stilbene and sydnone moieties. The stilbene moiety photochemically isomerizes and the process of trans–cis isomerization is in competition with the photolysis of the sydnone ring. Photolysis of the sydnone moiety leads to a nitrile imine, followed by its intramolecular trapping by the cis or trans double bond of stilbene moiety, affording polycyclic compounds 11 and 12, respectively. The same starting compounds also react thermally: The sydnone moiety in 3a reacts as a masked azomethine dipole with trans configuration on the stilbene moiety by intramolecular [3 + 2] cycloaddition, giving polycyclic compound 14, while the sydnone moiety in the cis isomer 3b gives polycyclic compound 15. Stilbene-methylene-sydnones are useful substrates for photochemical and thermal intramolecular [3 + 2] cycloaddition reactions to heteropolycyclic compounds.</p><!><p>Experimental details and characterization data for all compounds.</p><p>1H NMR and APT spectra of 3a, 3b, 11–15, NOESY spectra of 11, 12, 14 and 15 and X-ray data for 14.</p>

PubMed Open Access

Leveraging heterogeneous data from GHS toxicity annotations, molecular and protein target descriptors and Tox21 assay readouts to predict and rationalise acute toxicity

Despite the increasing knowledge in both the chemical and biological domains the assimilation and exploration of heterogeneous datasets, encoding information about the chemical, bioactivity and phenotypic properties of compounds, remains a challenge due to requirement for overlap between chemicals assayed across the spaces. Here, we have constructed a novel dataset, larger than we have used in prior work, comprising 579 acute oral toxic compounds and 1427 non-toxic compounds derived from regulatory GHS information, along with their corresponding molecular and protein target descriptors and qHTS in vitro assay readouts from the Tox21 project. We found no clear association between the results of a FAFDrugs4 toxicophore screen and the acute oral toxicity classifications for our compound set; and a screen using a subset of the ToxAlerts toxicophores was also of limited utility, with only slight enrichment toward the toxic set (odds ratio of 1.48). We then investigated to what degree toxic and non-toxic compounds could be separated in each of the spaces, to compare their potential contribution to further analyses. Using an LDA projection, we found the largest degree of separation using chemical descriptors (Cohen’s d of 1.95) and the lowest degree of separation between toxicity classes using qHTS descriptors (Cohen’s d of 0.67). To compare the predictivity of the feature spaces for the toxicity endpoint, we next trained Random Forest (RF) acute oral toxicity classifiers on either molecular, protein target and qHTS descriptors. RFs trained on molecular and protein target descriptors were most predictive, with ROC AUC values of 0.80–0.92 and 0.70–0.85, respectively, across three test sets. RFs trained on both chemical and protein target descriptors combined exhibited similar predictive performance to the single-domain models (ROC AUC of 0.80–0.91). Model interpretability was improved by the inclusion of protein target descriptors, which allow the identification of specific targets (e.g. Retinal dehydrogenase) with literature links to toxic modes of action (e.g. oxidative stress). The dataset compiled in this study has been made available for future application.Electronic supplementary materialThe online version of this article (10.1186/s13321-019-0356-5) contains supplementary material, which is available to authorized users.

leveraging_heterogeneous_data_from_ghs_toxicity_annotations,_molecular_and_protein_target_descriptor

Introduction<!>Dataset collation<!>GHS category annotation<!>Structural preprocessing<!>Descriptor generation<!>Binary toxicity classes<!>Exploratory data analysis<!>Toxicophore screening<!>Class separation analyses<!>Predictive modelling<!><!>Exploratory data analysis<!><!>Toxicophore analyses<!><!>Toxicophore analyses<!><!>Class separation analyses<!>Predictive modelling<!><!>Predictive modelling<!><!>Predictive modelling<!>Conclusions<!>

<p>There are two competing pressures in contemporary chemical risk-assessment. On the one hand, there is increased demand for safety data, for example under the EU's REACH regulations firms are obliged to submit detailed risk and hazard notifications to the EU for any substance they introduce in significant quantity [1]. On the other hand, there is decreased regulatory and societal acceptance of large-scale traditional in vivo toxicity studies on animals; moreover, on a practical level, covering the vast regions of chemical space requiring toxicity data using the traditional in vivo toxicological techniques would be too time- and resource-intensive to be feasible [2]. The consequence of these pressures is an increased demand for novel methodologies to complement and in some cases replace in vivo studies, including in silico computational toxicology techniques.</p><p>Conventional in silico approaches for toxicity prediction include quantitative structure–toxicity relationship modelling, analogous to quantitative structure–activity relationship (QSAR) modelling, whereby machine learning technologies are applied to derive a regression or classification function that maps from chemical structures to their in vivo effect [3]. These approaches have been extensively applied in the domain of toxicity prediction, with some success in cytotoxicity [4, 5], hepatotoxicity [6] and off-target effect prediction [7].</p><p>An extension of this approach is the integration of high-throughput in vitro screening data and/or protein target annotations into predictive toxicity modelling. In comparison to toxicity prediction methods that only utilise toxicity structure/structural alerts data alone to provide a prediction [5, 8] these heterogenous approach operate under the hypothesis that chemical, protein target, and phenotypic data domains each contribute partially independent and therefore complementary information about the potential toxic effects of a compound in vivo, and that utilising them in combination may therefore improve the performance of predictive models. Sedykh et al. [9] showed that enhancing a QSAR-style toxicity model with in vitro qHTS data improved its performance. This technique was recently reviewed by Low et al. [10], who, while noting mixed success in its application thus far, expressed optimism that the approach would play a greater role in future of toxicology and drug discovery as the data and expertise required to implement such techniques become more available.</p><p>In a previous study [11], we extended this approach by augmenting the dataset of Sedykh et al. with protein target descriptors corresponding to the likelihood of a ligand-target interaction derived from a Bayesian prediction model, representing protein target affinity predictions, and investigated the performance of Random Forest classification models trained to predict binary toxicity classes using successive integration of data domains. We found that, for this data set, inclusion of heterogeneous data domains did indeed generally tend to improve model performance—with a models trained using all three descriptor domains outperforming other combinations of descriptors, with an average correct classification rate (CCR), defined as the mean of sensitivity (true positive rate) and specificity (true negative rate) and also called the balanced accuracy, of 0.82 compared to 0.80 for the next-best model. Models showed the most improvement when chemical descriptors were added to a model which previously lacked them. We also found that the improvement over chemistry-only models was a consequence of more accurate extrapolation of the models' applicability domain into wider chemical space. However, this study concerned only one small dataset (367 compounds).</p><p>Despite the significant increase in data available in the chemical and biological domains, the development of models using several heterogenous data domains still represents a significant challenge. This is due to the collation of a suitable set, given that compounds must simultaneously possess readouts with overlap across several data types, e.g. the structural, bioactivity, phenotypic readout, and toxicological domains (or a subset thereof). In this study, we hence made use of the wealth of toxicity data made available through the Globally Harmonized System of Classification and Labelling (GHS) in order to maximise the overlap of the ToxCast and Tox21 chemical library [12] with a toxicity classification for modelling.</p><p>The Globally Harmonized System of Classification and Labelling (GHS) [13] is an international framework for standardising chemical health and safety information. The GHS encompasses a broad spectrum of physical, health and environmental hazards; pertinently for the purposes of this study, this includes the collation of the outcomes of independent toxicity assessments into a set of categories corresponding to the severity of the exhibited toxicity. For three routes of exposure (dermal, inhalation and oral) five categories are defined, each corresponding to a quantitative median lethal dose (LD50) interval specified in mg/kg, ppmV or mg/l as appropriate, with the three most severe categories (1–3) necessitating a "toxic" label, category 4 necessitating a "harmful" label, and category 5 requiring no label. The European Chemicals Agency (ECHA), Japan's National Institute of Technology and Evaluation (JP NITE), New Zealand's Environmental Protection Authority (NZ EPA) and Safe Work Australia (SWA) provide public access via their websites to governmentally mandated or recommended acute toxicity classifications under the GHS. Further, ECHA publishes the industrial submissions it receives under the requirements of the EU legislation, which include declaring GHS classifications. This data takes the form of the number of notifications received for each GHS hazard category, along with the total number of notifications received. The common classification standards provided by the GHS system enable the collation of acute oral toxicity data from all of these resources with the confidence that the data they hold are mutually commensurate by design. Apart from the ability to look up information on individual compounds, this also represents a valuable means of annotating large compound sets with toxicity labels as performed in this work. The ECHA database has been previously identified as suitable for toxicological data analysis [14], and GHS categories have been used as a framework for defining toxicity thresholds for predictive modelling [15].</p><p>In the present study, we annotated 3336 of 8540 standardized chemical structures from the ToxCast and Tox21 chemical library [16] with toxicity classifications derived from regulatory GHS information. We compared the overlap of this compound set with each GHS data source, and the correlations between the sources, and investigated the degree to which GHS-derived toxicity classifications could be discriminated through substructure based screens. We then derived three sets of descriptors for our compounds: molecular descriptors from MOE [17]; in silico-derived protein-target descriptors using an in-house Random Forest ligand-target prediction algorithm [18]; and qHTS activity scores taken from the Tox21 assays disseminated via PubChem [19]. We sought to compare how the GHS toxicity classifications related to these three descriptor sets through analysing the nearest-neighbour distance distributions and linear discrimination analysis projections using the chemical and protein-target descriptors, and the Tox21 qHTS assay data. Finally, we defined and applied three training-test set splits (one random, and two designed to be more challenging) to build and assess the performance of Random Forest classifiers on using these different descriptor sets, and analysed the effect of the inclusion of the different combinations of heterogeneous descriptors on model interpretability.</p><!><p>For this study, we required compound data for structures, targets, in vitro results, and a toxicity endpoint, which necessitated data collation from multiple sources. To this end, we were able to collate a dataset of 3055 compounds, each of which was annotated with: (1) a chemical structure in SMILES format, from which 2D molecular descriptors were calculated using MOE; (2) qHTS assay results from the Tox21 project published via PubChem; (3) protein target descriptors, representing probabilities of bioactivity against 109 human protein targets; and (4) regulator-derived GHS acute toxicity categorisation for oral, dermal and inhalation exposure routes.</p><p>Our starting point was the full ToxCast & Tox21 chemical library [12, 20], as made available for download on the website of the United States' Environmental Protection Agency [16]. From this, we discarded all compounds which (a) were labelled as "Mixture/Formulation", "Polymer" or "Macromolecule" in the "Substance_Type" field, or (b) did not possess a CAS registry number (required to lookup GHS categories in regulatory databases). SMILES strings were downloaded from PubChem for any remaining compounds which did not already possess them using the PubChem substance IDs provided or else the CAS registry numbers. This process yielded a set of 8540 compounds for which GHS toxicity annotations could be sought.</p><!><p>Authoritative GHS categorisations were derived from four regulatory classification databases: the harmonized classifications present in ECHA's Classification and Labelling Inventory [21], the NZ EPA's Classification and Information Database [22], the GHS classification results published by JP NITE [23], and SWA's Hazardous Chemicals Information [24]. (Note that while the classifications provided by the NZ EPA are not strictly GHS classifications, the two systems are standardized [25] such that the conversion of the acute toxicity categories to GHS categories is trivial). The authoritative classifications from ECHA, JP NITE and NZ EPA were accessed through the OECD's eChemPortal service [26] via the CAS registry numbers of compounds; classifications from ECHA and JP NITE were provided directly by eChemPortal, while classifications from NZ EPA were indirectly provided via a link to the compound's entry on the NZ EPA website. Classification data from SWA were not accessible via eChemPortal, but rather were downloaded directly from the SWA website as a flat file and once again matched to compounds via CAS registry numbers.</p><p>There are five acute toxicity categories defined by the GHS, ranging between category 1 (most severe) to category 5 (least severe). However, there is no acute toxicity GHS category directly representing general non-toxicity, since even the least-toxic category represents a closed LD50 interval [13]. We have therefore employed the concept of "implied nontoxicity" to ensure sufficient nontoxic compounds were included in the dataset: as GHS classifications are intended to provide a complete and comprehensive overview of a chemical's hazards, presence of a substance in a GHS database and absence of an acute toxicity category implies, according to our rationale, nontoxicity. Therefore, we have treated any compound which is present in an authoritative GHS classification database, but which is not categorised under acute toxicity for that administration route, as an implied nontoxic for that administration route. Overall, GHS acute toxicity classifications (including implied nontoxic classifications) from authoritative databases were found for 2770 compounds.</p><p>For the remaining 5770 compounds, a GHS categorisation was instead sought using the industrial notifications submitted to ECHA's Classification and Labelling inventory. Again, eChemPortal was used to connect a CAS registry number to an entry in the ECHA inventory. To minimize the impact of false positives when utilising industrial notification data, no categorisation was applied unless at least 10% of all notifications included that categorisation at either the same or a more severe level (This is the same threshold applied by ECHA to determine whether to advertise a notified hazard on their website and via their data contributions to PubChem). Using this method, 1591 of the total compounds could be associated with a notification-derived classification. Where both a notification-based classification and an authoritative classification were available for the same compound, we preferred the authoritative classification and discarded the notification-based classifications. Overall, 566 additional compound annotations were provided through notification-based classifications, bringing the total number of compounds with a GHS acute toxicity classification to 3336, or 39% of the 8540 compounds for which a classification was sought.</p><!><p>The last stage of dataset collation was compound filtering and standardization, and the removal of duplicates. The structures as represented by the SMILES strings were standardized using ChemAxon's Standardizer [27] (the protocol followed was: "remove fragment", "neutralize", "remove explicit hydrogens", "clean 2D", "mesomerize", "tautomerize"). Following standardization, the resultant structures were filtered to retain only small organic molecules, by discarding those with no carbon atoms, those containing elements of atomic number 21–32, 36–52 and > 53, and those with molecular weight over 100 Da, leaving 8328 structures.</p><p>Finally, a duplicate removal procedure was applied as follows: (1) duplicated structures were identified by converting standardized SMILES to InChIs [28]; (2) all compounds with a duplicated structure but no GHS annotation were discarded; (3) where only one compound in a set of duplicates was also an exact duplicate of its unstandardized structure, that compound alone was retained as the closest representation of the substance for which a GHS annotation was found, and the others discarded (this only occurred for two structures); and (4) any remaining sets of duplicates were discarded. Following this, 7732 substances remained, of which 3060 (40%) possessed GHS annotations.</p><!><p>For each remaining unique compound, molecular descriptors, protein target bioactivity probabilities, and qHTS-derived features were next derived as described in the following.</p><p>Firstly, structures were transformed to the major tautomer present at pH 7.4 using ChemAxon's Calculator [27]. Secondly, a set of 201 2D molecular descriptors were calculated using the Chemical Computing Group's Molecular Operating Environment (MOE) [17].</p><p>Next, to calculate protein target affinity probability profiles we made use of PIDGIN v2 [18], a collection of 3394 target prediction algorithms, trained on over 13 million bioactivity points, with actives (cutoff 10 μM) extracted from ChEMBL [29] and labelled inactives extracted from PubChem [19]. PIDGIN is itself a suite of predictive structure-bioactivity models, providing for each input compound a Platt-Scaled probability of affinity for each target. As with any predictive model its accuracy depends on (in this case) the particular input structure and target class. For that reason, when including its output in further predictive modelling, we chose to only use the output of reasonably reliable models whose applicability domains extended into the dataset at hand in order to minimise the error-carried-forward. To that end, the performance of individual PIDGIN models on the dataset was estimated by measuring the models' recall on the overlap of the input set and known activities in its training set. Using this estimate, a well-performing model was defined as one which achieved a recall (i.e. proportion of known actives assigned a probability of activity of over 50%) of at least 0.5 on the overlap of the query compounds and the model's training data. Further, we filtered well-performing models to retain only those with a training set having a mean nearest-neighbour Tanimoto similarity to the 7732 compound set used in this study of over 0.25, calculated using the circular fingerprints utilised by PIDGIN for prediction. These requirements can be considered stringent, as only 109 human target bioactivity models (3% of the total) were retained. This process afforded a set of annotations used as descriptors subsequently. To further ensure that the protein target bioactivity probabilities were as accurate as possible, known bioactivities (i.e. those included in the training data set extracted from ChEMBL) for the compound set on the selected targets were included as probabilities of 1 (i.e. certainty).</p><p>Lastly, we assembled qHTS data for our compound set from the Tox21 assay data made publicly available via PubChem. First, all PubChem assay data for all 192 assays listed with "Tox21" as their source were downloaded (accessed on 23 Aug 2018). Next, less relevant assays (i.e. counter-screening assays, autofluorescence assays, and those confirmatory assays for which a summary assay combining its results with counter screens was also available) were discarded, leaving 76 remaining assays (Additional file 1: Table S1). For these assays, the PubChem activity score was used to provide a single qHTS feature summarising the behaviour of a compound against an assay record as a continuous numerical descriptor. Such scores were available for nearly all compounds (7713 out of 7732), but compounds for which no Tox21 scores were available were excluded where necessary. The activity score ranges from 0 to 100, with inactive compounds having a score of 0, active compounds having a score from 40 to 100, and inconclusive compounds having a score in between. The score is provided by the depositor, and the exact way in which the score is calculated depends on the assay, but it is commonly assigned based on potency, efficacy, curve class or a combination of these. Where multiple scores were available for the same structure-assay pairing due to repeated measurements, the median score was used (36% of compounds had at least one repeated measurement). Missing values (11% of all data points) were assumed to be inactive, and assigned a score of 0.</p><p>The dataset collated in this study, alongside the code necessary for reproduction of the results obtained, is made publicly available via the Additional files included alongside this article (Additional file 2).</p><!><p>For the purposes of defining a binary acute toxicity classification for certain analyses, we considered acute oral toxicity only and took the GHS categories which require a "toxic" (skull and crossbones) pictogram—i.e. categories 1–3—as the toxic class, and those which require no pictogram—i.e. category 5 and implied nontoxicity—as the nontoxic class. Compounds in category 4 (requiring a "harmful" pictogram) were treated as marginal, and disregarded when performing binary analyses. Following this transformation, 2006 compounds were retained of which 579 were classed as toxic and 1427 as nontoxic. Of these, only three compounds lacked Tox21 assay outcomes.</p><!><p>The sources' contributions to the final data were compared by calculating the relative overlap of compounds between the sources, with relative overlap quantified as the intersection over the union. The degree to which the sources were commensurate with one another was considered by calculating the agreement of labels between the GHS acute toxicity categories of the compounds present in both data sources (for the purposes of this analysis, implied nontoxic compounds were allocated to a hypothetical category 6).</p><p>The oral bioavailability of the compound set was quantified by considering the fraction of compounds which passed Lipinski's rule, which was calculated as part of the MOE [17] molecular descriptor set. The druglikeness of the compound set was quantified using DataWarrior's [30] fragment-based druglikeness score, in which positive values indicate more and negative values less druglike structures.</p><p>Visualisation of the coverage of the dataset's chemical space with GHS annotations was performed using DataWarrior's Self-Organising Map function [30], using SkelSpheres fingerprints, a Gaussian neighbourhood function, and 100 neurons per axis (i.e. 10,000 neurons in total).</p><!><p>The structures in the compound set were analysed for the presence of toxicophores using FAFDrugs4 [31] and ToxAlerts [32] in order to investigate the overlap of GHS toxicity classifications with established screens. FAFDrugs4 is an online server designed for filtering compound libraries prior to in silico screening experiments and related modelling studies, which through the "filter undesirable substructure moieties" filtration option provides for the identification of structures containing substructures involved in toxicity which were collated through a manual survey of the literature. For each compound screened by the FAFDrugs4 server, in addition to an inventory of undesirable substructures present, a qualitative screening outcome has been generated (one out of "rejected", "intermediate" or "accepted"), which depends on the quantity and severity of toxicophores identified.</p><p>ToxAlerts is another online platform, available via the Online Chemical Modelling Environment [33], providing structural alerts for the virtual screening of chemical libraries to flag compounds containing toxicity-related substructures. In contrast to FAFDrugs4, ToxAlerts is an open platform allowing user-contributed alerts; however, all alerts must be accompanied by an endpoint and a reference and must be moderated before approval. ToxAlerts therefore contains a wide range of alerts and sources, including sets of alerts derived from literature and from industry. While ToxAlerts contains alerts are all associated with an "endpoint", certain of these endpoints are generic in nature and do not relate to any defined toxic outcome (e.g. the endpoint "extended functional groups" represents structural features for use in chemical space analysis or as features in further modelling rather than a screening set [34]), and hence the full battery of alerts contained ToxAlerts is not itself suitable for application as a toxicity screen. Rather, a relevant subset of the alerts should be employed. In this study, we used the alerts annotated with the "reactive, unstable, toxic" endpoint, which includes structural alerts defined by ChemDiv, LifeChemicals and Enamine used for their internal compound selection purposes, which was also previously explored in the publication introducing ToxAlerts [32].</p><!><p>We next analysed the various descriptor spaces to determine how informative each was with respect to the GHS toxicity labels. Nearest-neighbour distance analyses were performed as follows: for each compound in our set with a binarized toxicity, the distance to its nearest-neighbour (the most similar compound) in its own class (intra-class) and to its nearest-neighbour in the other class (inter-class) was calculated and the two distributions compared in each space. For the molecular descriptor and the protein target probability spaces, Euclidean distance in MOE and protein target descriptor space was employed to define the nearest-neighbour; for the Tox21 assay scores, due to the sparsity of the data matrix (i.e. a large proportion of zero values), Cosine distance was employed instead. Before calculating distances, the molecular descriptors were centred and scaled (this was unnecessary for the protein target and Tox21 assay descriptor sets, which are each already internally on the same scale). Differences in distribution were tested for significance using a paired t test (implemented by SciPy [35]), and effect size measured using Cohen's d (i.e. the difference between two means divided by their pooled standard deviation).</p><p>The classes were next analysed to determine which predicted protein targets were more associated with toxic compounds. We used SciPy [35] to perform a paired t-test and a Kolmogorov–Smirnov test to determine which target probability descriptors showed a significant difference in distribution between the two classes. We considered a distribution to show a significant difference where the larger of the two p values obtained was beneath the Bonferroni-corrected threshold equivalent to α = 0.05. Cohen's d was used to quantify the effect size.</p><p>We then performed linear discriminant analyses (LDA) as implemented in scikit-learn [36] on the three descriptor spaces to compare their relevance towards the endpoint, by assessing the degree to which a simple linear model might be able to separate the classes in those spaces. An LDA performs feature reduction (to at most one dimension fewer than the number of classes, i.e. to a single dimension in the binary class case) by deriving the linear transformation of the feature space which achieves a maximum separation of the classes; the degree to which the classes separate in the LDA is therefore indicative of the maximal extent to which the classes can be linearly separated in the feature space. The significance of the separation observed under the LDA projection was measured using an unpaired t-test, and the effect size using Cohen's d. We examined the weights assigned to the original protein target descriptors by the LDA projection to determine which of these features were found to be more important in linearly separating the classes, and hence which proteins might be involved in causing toxicity in man.</p><!><p>In order to explore the degree to which GHS-derived toxicity labels can be predicted using machine learning models, we defined a predictive modelling set comprising the compounds for which a binary acute oral toxicity class could be defined. Such predictive models are increasingly relevant to chemical hazard assessment, but the robustness of any model is a reliant on the quantity and quality of the data available [37]. For evaluation of predictive models, we defined three test sets. Firstly, we defined a random test set, comprising a simple class-stratified random 20% subsample of the modelling set. We then defined two test sets to approximate a more realistic scenario where a predictive model is applied to novel data: a rare scaffolds test set (representing novel chemistry) and a single source test set (indicating generalisability to novel datasets). The rare scaffolds test set comprised all those compounds having a Murcko scaffold [38] present in the modelling set no more than twice (26% of the compounds), the scaffolds having been calculated using RDKit [39]. The single source test set comprised a further class-stratified random 20% subsample of the modelling set, but was restricted to include only compounds found in exactly one authoritative GHS data source. In each case, all compounds not included in a test set were used as the corresponding training set.</p><p>We applied Random Forest classification models [40] in the predictive modelling analysis, as implemented in scikit-learn [36], because they (a) are non-linear models (complementing the linear transformations performed as part of the exploratory data analysis), (b) are relatively quick to train, even on large numbers of features, (c) require little in the way of feature selection, pre-processing or parameter tuning, and (d) are well-studied algorithms with previous well-performing application in toxicity prediction [11]. For each modelling run, a Random Forest classification model of 200 trees was trained on the relevant training set using scikit-learn default parameters: e.g. criterion (the metric for selecting a split) set to "gini", max_features (the number of features considered at each split) set to the square root of the number of features, max_depth (the maximum depth of each tree) set to "None" to permit unlimited depth. A fivefold cross-validation routine was employed within the training set to determine the optimum probability threshold (corresponding to the proportion of concurring trees in the ensemble) to be used as the decision boundary, maximising CCR. Each model was then applied to the relevant test set, and its performances assessed in terms of area under the ROC curve and average precision (summarised from the precision-recall curve), along with the sensitivity, specificity and CCR achieved at the probability threshold determined to be optimal from the cross-validation performed within the training set.</p><p>Finally, to demonstrate the robustness of the performance statistics computed for these models, and the degree to which they depend upon the random selection of test compounds, the training-test splits were repeated an additional 20 times for those splitting methods where randomness played a role (i.e. the random method and the single source method). The modelling workflow above was undertaken the performance recorded for each of these repeated splits.</p><!><p>Overlap heatmap for common compounds across the GHS data sources. Overlap here is quantified as intersection over union. Absolute numbers of compounds present in each source are given in the x axis. With the exception of the ECHA and SWA data sources (SWA uses ECHA as one of its own sources of classification), the heatmap indicates that other data sources contribute significant quantities of unique data, illustrating the benefit of collating toxicity data from multiple commensurate sources</p><p>Agreement heatmap between GHS acute toxicity categorisations of common compounds in the GHS data sources. For the purpose of this analysis, implied nontoxic compounds were treated as belonging to a hypothetical category 6. The heatmap indicates that common compounds' acute toxicity classifications tend to exhibit higher agreement when comparing classifications for the same route of exposure; however, agreement between different routes of exposure are generally substantially lower</p><p>Lipinski's rule failure rate (a) and DataWarrior fragment-based druglikeness score (b) for the structures in our compound set. Compounds with an available GHS-derived acute oral toxicity classification (including implied nontoxicity) more frequently pass the Lipinski filter, which may indicate higher bioavailability among those compounds. The distributions (median and inter-quartile range) of druglikeness among the classifications are very similar, though the tail length varies. Hence, we determined that the annotated compounds did not substantially differ with regard to their druglikeness</p><!><p>The distribution of GHS-derived toxicity classes within the chemical space of the dataset was visualized in through a Self-Organizing Map (Additional file 1: Fig. S1) in which compounds were coloured according to their GHS acute oral toxicity classification. This plot illustrates that GHS categories had been derived for diverse subset of the chemical space described by the wider chemical library, although classifications were not uniformly distributed. The visualisation also reveals a small degree of clustering among the classes.</p><!><p>The results of toxicophore screens on the compound set, using the FAFDrugs4 screen (a, b), and the "reactive, unstable, toxic" endpoint alerts from ToxCast (c, d). The results of the FAFDrugs4 test are unable to usefully screen for acute oral toxicity as encoded by GHS classes: in B, both rejection (red bar) and acceptance (green bar) is more common in nontoxic compounds. However, there is a weak relationship between these classes and the presence/absence of the "reactive, unstable, toxic" ToxAlerts toxicophores, as illustrated by the higher red bar for the toxic compounds in D</p><!><p>To complement FAFDrugs4 screen, we next used the ToxAlerts webserver to screen our compounds as outlined in the Methods section. We identified that ToxAlerts associated with the "reactive, unstable, toxic" endpoint were present in slightly larger proportions of compounds in the more severe acute oral toxicity categories (present in 79%, 67% and 61% of compounds in categories 1–3, 4 and 5, respectively) (Fig. 4c). When the compound set was divided into binary "toxic" and "nontoxic" classes (Fig. 4d), the enrichment of "reactive, unstable, toxic" toxicophores among the compounds annotated as toxic via GHS classifications was measured as having an Odds Ratio of 1.56 with a p value of 1.3 × 10−4 (representing a significant enrichment, albeit only at a small effect size). This means that while the agreement between FAFDrugs4 alerts and GHS toxicity labels is small, the "reactive, unstable, toxic" substructures from ToxAlerts show a greater prevalence among toxic compounds and therefore may be implicated in contributing towards these compounds' toxic behaviour.</p><!><p>Top 5 enriched "reactive, unstable, toxic" endpoint ToxAlerts substructures in the binarized toxic set (GHS acute oral toxicity category of 1–3) versus nontoxic set (acute oral toxicity category of 5 or implied nontoxic)</p><p>Only enrichments with a p value below the Bonferroni-corrected cut-off equivalent to α = 0.05 were considered. The remaining significant enrichments were ranked according to their odds-ratio, or effect size. These alerts generally represent reactive functionalities that might be anticipated to afford nonspecific toxicity</p><!><p>While the substructures in the ToxAlerts server labelled with the "reactive, unstable, toxic" endpoint were enriched in compounds classed as "toxic" using GHS acute oral toxicity categories, the size of the effect was small. Indeed, we found that the majority of our toxic and nontoxic compounds contained at least one relevant ToxAlerts alert. These findings corroborate previous findings in the literature concerning the limited utility of relying upon (non-quantitative) structural alerts for accurate toxicity assessment [44], not least due to the propensity for alerts to be present in large proportions of both toxic and nontoxic compounds [45]. Nonetheless, as the GHS is the internationally recognised standard for categorising and communicating chemical hazard generally and acute toxicity specifically, and structural alerts are a widely accepted technique in toxicity screening, we would have anticipated toxicophore analysis to be a more useful means of forecasting its acute oral toxicity categories.</p><!><p>Nearest-neighbour distance distributions for intra-class and inter-class pairs among compounds annotated with a binary toxicity class, and linear discriminant analysis projections on the same data. The molecular descriptor space is used in a, b, the protein target descriptor space in c, d, and the in vitro qHTS space in e, f. Larger differences in nearest-neighbour distributions and LDA projections are evident in the molecular and protein target descriptor spaces than in the in vitro qHTS space</p><p>Protein target descriptors exhibiting a significant difference in distribution between binarized toxic set versus nontoxic set</p><p>Enrichment effect size was quantified via Cohen's d, and the top 5 targets with the largest effect size indicating more association with the toxic set are shown. Only enrichments with a p value below the Bonferroni-corrected threshold equivalent to α = 0.05 were considered. The full table is given in Additional file 11: Table S3. The relevance of these targets to toxicity is explored in the text</p><p>Top 5 highest-weighted human target descriptors in the linear discriminant analysis projection used to discriminate between binarized toxic and nontoxic classes</p><p>Central to nerutotranmitter activity i.e.—acetylcholine transport</p><p>GPCR not the most toxic—could link it to pain [n/a]</p><p>The relevance of these targets to toxicity is explored in the text</p><!><p>We finally performed the nearest-neighbour and LDA analysis on Tox21 assay space, which shows a lower degree of separation on the qHTS descriptors (Fig. 5e, f). The trend of intraclass nearest-neighbours tending to be nearer together than interclass nearest-neighbours was still apparent, but the two distributions were largely overlapping and the effect size was very small (Cohen's d of 0.11, p value of 1.4 × 10−40). Despite LDA deriving a transformation which achieves the maximal linear separation, still the maxima of the two classes overlapped in the LDA dimension, and the effect size was much less than for the other descriptor spaces (Cohen's d of 0.67, p value of 7.5 × 10−40). These results suggested that a linear method of class discrimination would have limited applicability using the qHTS values as descriptors. We therefore conclude that the relationship between Tox21 assays and acute oral toxicity is non-trivial, and that the biological mechanisms of toxicity represented in the GHS-derived classifications are not fully captured by the endpoints measured by Tox21; this is partially consistent with the findings of Huang et al. [55], who reported that the performance of models using Tox21 assay data without chemical descriptors to model in vivo toxicity was highly end point dependent, with success only for a subset of endpoints studied.</p><!><p>We next utilized Random Forest classification models in order to evaluate how well different input feature spaces can be used to predict GHS classes computationally. Briefly (for details see methods section), the models were evaluated using three test sets: a random test set, a rare scaffold test set (containing only compounds having a Murcko scaffold present no more than twice, representing novel chemistry), and a single source test set (containing compounds present in only one authoritative GHS data source, indicating generalisability to novel datasets). For each test set, models were trained on all compounds not included in the test set. For each training/test set split, models were generated using each descriptor set in turn. In addition, a fourth class of models were generated using a descriptor set comprising a combination of chemical and protein target descriptors. This resulted in 12 models in total across three test sets and four descriptor sets.</p><!><p>Summary of Random Forest classifier performances across the three different test sets and the four different combinations of descriptors</p><p>Generally, the best performing models were those trained using either molecular descriptors alone or in combination with protein target descriptors. Classifiers found the random test set less challenging to predict than the two more challenging test sets</p><p>Performance of the four classes of Random Forest classifiers trained on the dataset, quantified by ROC AUC, average precision, and the sensitivity, selectivity and CCR achieved at the optimal prediction threshold, across the three training-test set splits. It can be observed that the best-performing class of models were those utilising molecular descriptors alone or in combination with protein target descriptors. The random test-training split afforded the best-performing models, while performance predicting the toxicity of the rare scaffolds and single source test sets was lower</p><!><p>The models trained on protein target probabilities achieved good performance overall across the test sets (centre-left, green bars in Fig. 6), with ROC AUCs between 0.85 and 0.70, average precisions between 0.71 and 0.51, and CCRs (at the optimum thresholds) between 0.77 and 0.65 across the test sets. Despite this, ROC AUC, average precision and CCR were each always lower than the equivalent value for the model built using molecular descriptors for every test set. This result corroborates the finding of our prior work, [11] in which we found an average CCR drop of 0.11 across 100 training-test splits comparing Random Forest models trained on molecular descriptors to those trained on protein target descriptors.</p><p>The predictive performances of the models trained on molecular descriptors and the models trained on protein-target descriptors were further analysed by considering only those compounds which were classified differently by the two models in any of the training-test set splits, using their optimum thresholds. A full list of these compounds is provided in Additional file 3, and summarised in Additional file 1: Table S4. It is observed that for such compounds, models trained using molecular descriptors are generally more likely to make the correct prediction. However, this was not the case for toxic compounds in the rare scaffolds test set (for which the protein-target descriptor-trained model made the correct prediction 61.2% of the time) and toxic compounds in the single source test set (for which the protein-target descriptor-trained model made the correct prediction 62.1% of the time). Where the two models disagreed on nontoxic compounds in these test sets, the molecular descriptor-trained models more frequently made the correct classification (78.3% of cases for the rare scaffolds set and 74.7% of cases for the single source set). For compounds within the random test set where the two classifiers disagreed, the molecular descriptor-trained model performed better on both toxic and nontoxic compounds (88.2% and 60.6% correct, respectively). However, these compounds were the only ones for which the protein-target descriptor-trained model made a larger proportion of correct predictions on nontoxic than toxic compounds—though it was out-performed by the very strong molecular descriptor-trained model (ROC AUC of 0.92) in each case.</p><p>In contrast to the aforementioned models, the models built using data derived from in vitro data from the Tox21 project exhibited comparatively poor predictivity across the test sets (centre-right, red bars in Fig. 6), with ROC AUCs varying from 0.61 to 0.57, average precision scores varying from 0.40 to 0.36, and CCRs at the optimum threshold varying from 0.57 to 0.55. These results indicate that the model performed only slightly better than a random classifier, which is particularly evident in the ROC plots in Additional file 1: Fig. S2. Though this is a disappointing result, it reflects the smaller separations observed between classes seen in earlier exploratory analyses (Fig. 5). Indeed, there has been mixed success in the literature when attempting to use high-throughput screening assays to predict in vivo toxicity. In order to link our work to existing studies, we consider the review of Thomas et al. [56] who found in their comprehensive review of chemicals and assays provided under ToxCast phase I (a closely related endeavour to the Tox21 project) that those assays have "limited applicability for predicting in vivo chemical hazards using standard statistical classification methods." In contrast, the study of Huang et al. [55] which reported variable success utilizing Tox21 assays alone to predict in vivo endpoints found that "combing structure and activity data resulted in better models than those built with structure or activity data alone" for most of the endpoints they studied. The difference in results may be surprising given the overlap between the compound set and the Tox21 assays, however this previous study differed from ours in that they made use of a Self-Organizing Map-based approach in contrast to our Random Forest, and different toxicity endpoints were considered. It is reasonable to conclude that the success of utilizing the Tox21 assay data depends on optimizing the analysis for that purpose, rather than comparing their performance as simply another class of descriptor in a study such as this. Moreover, the such qHTS assay data has applications beyond predictive modelling, including deriving putative mechanisms for adverse events [57].</p><p>We also generated a fourth class of models using both molecular and protein target descriptor sets. The predictive performance of this model class is also given in Table 4 and illustrated by the far right, purple bars in Fig. 6. Performance measured by ROC AUC, average precision and CCR for this class of model tended to be very similar to the molecular descriptors-only model class across the three test sets (differing by at most 0.03), though sensitivity and specificity showed more variation (differing by at most 0.19 and 0.18, respectively). The maximum discrepancy occurred for the rare scaffold set, for which the combination model exhibited a sensitivity 0.19 higher than the molecular descriptor only model, and a selectivity 0.18 lower. Overall, however, the CCR remained comparable.</p><p>Finally, to examine the dependence of these results on the randomness in the division of test and training instances, the training–testing routines were performed as above for 40 further models: 20 using further class-stratified random splits, and 20 using further single-source splits (in which a class-stratified random subsample of single-source compounds are selected for the test set). Due to the design of the rare scaffold set, it was not possible to repeat this split to generate multiple test sets. The mean and sample standard deviation (SD) of the performance metrics for multiple random-split models are given in Additional file 1: Table S5, and for the multiple single-source splits in Additional file 1: Table S6. These results indicate that the performance of the models analysed in detail elsewhere in this study are broadly representative of other random splits that might have been chosen. In particular, the same relative performance trends for the models trained both on varying descriptor sets and varying training sets are observed in the models analysed in detail and in the summary statistics. One key observation is that, on average, models trained using only molecular descriptors and models trained using both molecular and protein-target descriptors perform very similarly on average for both random splits [mean ROC AUC of 0.941 and 0.913 (SD of 0.013 in each case), respectively] and single source splits [mean ROC AUC of 0.851 and 0.852 (SD of 0.013 in each case again), respectively]. The average performances of models trained using the Tox21 assay descriptors remained low, with the best average performance seen in the random test sets producing a ROC AUC of 0.619 (SD of 0.034) and a CCR of 0.619 (SD of 0.029).</p><!><p>Highest-importance features in the two Random Forest classifiers with the highest ROC AUC scores, i.e. those generated using the random test-training set split using (a) molecular descriptors only and (b) both molecular and protein target descriptors</p><p>a_nN: Number of nitrogen atoms</p><p>Q_RPC-: Relative negative partial charge</p><p>a_ICM: Atom information content (mean)</p><p>h_pavgQ: Average total charge sum across protonation states at pH 7</p><p>GCUT_PEOE_0: First GCUT descriptor calculated from the eigenvalues of a modified graph distance adjacency matrix where the diagonal takes the values of the partial charges</p><p>Q_RPC-: Relative negative partial charge</p><p>a_nN: Number of nitrogen atoms</p><p>GCUT_SLOGP_0: First GCUT descriptor calculated using atomic contributions to logP instead of partial charge</p><p>bpol: Sum of the absolute value of the difference between atomic polarizabilities of all bonded atoms in the molecule</p><p>chi1v_C: Carbon valence connectivity index (order 1)</p><p>P18031: Tyrosine-protein phosphatase non-receptor type 1</p><p>P51449: Nuclear receptor ROR-gamma</p><p>P00352: Retinal dehydrogenase 1</p><p>P23219: Prostaglandin G/H synthase 1</p><p>P11473: Vitamin D3 receptor</p><p>The table illustrates the difference in interpretability between the two classes of descriptors, since molecular descriptors may be either be too broad to interpret or nontrivial to understand, while protein target descriptors provide a specific biological hypothesis which can be subsequently tested to validate a mechanism of action</p><!><p>While the MOE-generated molecular descriptors are evidently highly predictive for the data set employed in the present study, and therefore of great utility where the priority is simply the accuracy of the model, it's not immediately obvious what some of the more esoteric descriptors represent (e.g. connectivity indices), nor how they might relate to acute oral toxicity. This means that generating any further insight from these models is a challenge. In contrast, as discussed above, protein target descriptors can be more readily interpreted through literature validation. Two of the most of important protein targets were not previously identified through the exploration of descriptor spaces, namely Tyrosine-protein phosphatase non-receptor type 1, a kinase linked with cytoskeletal machinery and interferon pathways (GO:0060338) [58], and Prostaglandin G/H synthase 1 which has links to cellular stress events [59]. Three of the five were previously identified in the aforementioned nearest-neighbour and LDA analysis (Nuclear receptor ROR-gamma, Retinal dehydrogenase 1 and Vitamin D3 receptor), which illustrates that the RF algorithm is able to identify these target descriptors as an important feature to separated toxicity classes in addition to its predictive functionality. The targets identified in this work, via the feature importance and descriptor distribution analysis, share overlap with the cytotoxic target enrichment analysis (Fisher Test) presented in the complete list included in Mervin et al. [5], namely DNA dC → dU-editing enzyme APOBEC-3F, Glucagon-like peptide 1 receptor and Tyrosine-protein phosphatase non-receptor type 1, in addition to the histone deacetylases 5 and 6, as identified in both Mervin et al. and Liggi et al. [4, 5].</p><!><p>We have presented the collation of a novel dataset, not previously analysed, which comprised acute toxicity labels (derived from GHS data made available by regulatory authorities), chemical structures and qHTS results from Tox21. Molecular descriptors were derived from the chemical structures, compounds were annotated protein target descriptors using in silico target prediction, and the qHTS results were summarised into a descriptor set.</p><p>In our exploration of the descriptor dataset, we found those compounds with a GHS-derived acute oral toxicity labels were not substantially more or less druglike than the full ToxCast & Tox21 chemical library. We found that acute oral toxicity, as encoded by the GHS system, was not well aligned with the FAFDrugs4 toxicophore-based screen. In contrast, a subset of the toxicophores from the ToxAlerts server exhibited a modest relationship with GHS-encoded acute oral toxicity. We therefore conclude that toxicophore-based screens cannot alone discern the acute toxicity encoded within the GHS. We found that the acute oral toxicity classes derived from GHS data were partially linearly separable in chemical and protein-target space, as illustrated using nearest-neighbour distance distributions and linear discriminant analyses. Little separation was observed in the Tox21 descriptor space, in agreement with our previous studies.</p><p>Predictive models could be created by training Random Forest models on the dataset using molecular descriptors and protein target bioactivity probabilities as input features, with the model trained on the molecular descriptors outperforming that trained on the bioactivities (CCRs of 0.85–0.72 compared to 0.77–0.65). However, the qHTS data from the Tox21 assays could not be successfully employed in GHS class prediction, with the Random Forest model trained using these as features exhibiting a CCR little better than a random guess. We conclude from this that the endpoints captured by the Tox21 project may not be relevant to the acute in vivo toxicity encoded by the GHS classifications, or else that the relationship between the two may not be captured by the simplistic application of the Tox21 assay results as features in a supervised machine learning algorithm as expected from the separations observed in the three spaces.</p><p>A combined model trained on both chemical descriptors and protein target bioactivity descriptors had similar predictive performance to that trained on chemical descriptors only. This result was confirmed by measuring the average performance of 40 further models trained and tested on repeated randomised splits. We suggested that, given that the performance of the combined model was comparable with the chemistry-only model, such a combined model may be primarily useful due to the increased interpretability of the feature importance extractable from the model. The five most relevant protein target descriptors had links to toxicity in the literature. Further work on this dataset may focus on exploring the degree to which this interpretability can be harnessed toward interpretability and mode-of-action hypothesis generation on a compound-by compound basis. We have made the dataset publicly available to this end.</p><!><p>Additional file 1. Supplementary information.</p><p>Additional file 2. Research code and data.</p><p>Additional file 3. List of compounds for which molecular and protein target models disagree.</p><p>(class-balanced) correct classification rate, i.e. the mean of sensitivity and specificity</p><p>European Chemicals Agency</p><p>Japan's National Institution of Technology and Evaluation</p><p>linear discriminant analysis</p><p>New Zealand's Environmental Protection Authority</p><p>receiver operating characteristic curve</p><p>area under the receiver operating characteristic curve</p><p>Safe Work Australia</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

Retooling Asymmetric Conjugate Additions for Sterically Demanding Substrates with an Iterative Data-Driven Approach

The development of new catalytic enantioselective methods is routinely carried out using easily accessible and prototypical substrates. This approach to reaction development often yields asymmetric methods that perform poorly using substrates sterically or electronically dissimilar to those used during the reaction optimization campaign. Consequently, expanding the scope of previously optimized catalytic asymmetric reactions to include new substrates is decidedly non-trivial. Here we address this challenge through the development of a systematic workflow to broaden the applicability and reliability of asymmetric conjugate additions to substrates conventionally regarded as sterically and electronically challenging. The copper-2 catalyzed asymmetric conjugate addition of alkylzirconium nucleophiles to form tertiary centers, although successful for linear alkyl chains, fails for more sterically demanding linear α,βunsaturated ketones. Key to adapting this method to obtain high enantioselectivity was the discovery of new phosphoramidite ligands, designed using quantitative structure-selectivity relationships (QSSR). Iterative rounds of model construction and ligand synthesis were executed in parallel to evaluate the performance of twenty chiral ligands. The copper-catalyzed asymmetric addition is now more broadly applicable, even tolerating linear enones bearing tertbutyl -substituents. The presence of common functional groups is tolerated in both nucleophiles and electrophiles, giving up to 99% yield and 95% ee across twenty examples.

retooling_asymmetric_conjugate_additions_for_sterically_demanding_substrates_with_an_iterative_data-

INTRODUCTION<!>4<!>Initial results<!>Structure-selectivity relationships<!>Multivariate modelling<!>Multi-objective optimization<!>Scope<!>CONCLUSIONS<!>ASSOCIATED CONTENT Supporting Information.

<p>The development of catalytic asymmetric methods usually begins with the examination of a simple, readily available, benchmark substrate. While this approach is undeniably useful, it also often leads to a reaction protocol that is not widely applicable beyond the simple starting scaffold. Extending the scope of new reactions to include a variety of more complex substrates offers a wider range of potential applications. Reoptimization of reactions is often driven by empirical trial-and-error screening, a process that relies heavily on chance and intuition, making this a formidable challenge. There is a pressing need for a rational and operationally simple process to extend catalytic asymmetric methods to encompass electronically and/or sterically different starting materials.</p><p>The copper-catalyzed asymmetric conjugate addition (ACA) of organometallic species is a powerful tool to synthesize new C-C bonds from α,β-unsaturated carbonyl compounds [1][2][3][4][5][6] but remains underused in synthesis despite the intuitive appeal of this C-C bond disconnection. One likely reason for the underutilization of this method stems from the fact that methodology development has focused on simple substrates, whereas more highly-decorated reaction partners either do not display the desired reactivity or fail to reach suitable levels of enantioselectivity. 7,8 A lack of robustness in Cu-catalyzed ACAs is also well known and widely implicated in preventing the approach from enriching mainstream synthetic strategies and methods. 9,10 Substrate limitations in ACAs include the presence of large groups, aromatic moieties and electron-rich or -withdrawing functional groups. The origins of these limitations may be due to steric effects inhibiting C-C bond formation, or undesirable interactions of organometallic reagents with functional groups disrupting the catalytic cycle. Simple linear substrates are also more challenging than their cyclic counterparts, as the population of both s-cis and s-trans conformers of the substrate can lower the enantioselectivity. 8,11 There is a significant shortfall in the scope of products theoretically accessible through ACA methods and those that can be produced in practice. An incomplete understanding of how to close this gap limits the application of these methods in complex molecule settings. To make the ACA approach a "go-to methodology" 9 several advances are necessary, including the commercial availability of reagents, an operationally simple reaction set-up under convenient conditions, and the tolerance to a wide variety of substrates.</p><p>During our work on Cu-catalyzed asymmetric conjugate additions [12][13][14][15] of alkylzirconium species (generated from olefins) we found that asymmetric additions to linear enones bearing linear alkyl chains work well (>90% ee), 16 but additions to electronic or sterically deactivated enones gave only very poor results (<50% ee). This limitation is not unusual in ACA chemistry. 8,17 The use of such substrates is a longstanding challenge in asymmetric catalysis that motivated us to explore rational approaches to expand the scope of previously optimized catalytic asymmetric reactions.</p><!><p>Here we report that new phosphoramidite ligands, 18 developed with the aid of quantitative structure-selectivity relationships (QSSR), allow highly enantioselective Cu-catalysed ACAs of alkylzirconium species to linear enones bearing branched substituents or conjugated aromatic rings (Scheme 1). Selection of the best ligand from this series achieved high selectivity and reactivity with linear α,β-unsaturated ketones bearing -substituents as bulky as tert-butyl groups.</p><p>Scheme 1. Limitations in ACAs of alkylzirconium species to acyclic α,β-unsaturated ketones bearing branched or aromatic moieties and our approach tackling these limitations.</p><!><p>Benzylideneacetone 1a was previously found to be a challenging substrate for the Cu-catalyzed ACA 16 and was therefore chosen as a good candidate for examination. Previous conditions for related asymmetric additions were reoptimised and we subsequently found that the use of copper(I) triflate and a phosphoramidite ligand in the presence of TMSCl were critical to achieve 5 high reactivity. A combination of Et 2 O and DCM in a 1:1 mixture at 0 C also proved to be optimal for selectivity, similar to a previously published reaction. 13</p><!><p>We then examined structurally diverse phosphoramidite ligands (see Scheme S1 in ESI for more details) to explore the Phosphoramidite Ligand Space with the objective of finding a ligand "lead" to build upon. This preliminary screen uncovered an initially promising ligand L1, giving 90% yield and 71% ee. Structural diversification of the L1 scaffold provided the qualitative ligand-structure-enantioselectivity relationship shown in Scheme 2A. Several trends in ligand performance are apparent from these data: the aminoindane ring size is relatively unimportant (cf. L1-L2); the BINOL configuration dictates which is the major product enantiomer; the stereogenic center on the indane provides a matched-mismatched effect (cf. L3-L4); enantioselectivity can be tuned by variation of the R group, giving results from 67% to 92% ee (cf. L2, L4-L6). The variation in enantioselectivity as a function of relatively minor changes to the alkyl group was unexpected. Assuming Curtin-Hammett behavior, 19 the Gibbs energy difference between competing diastereomeric transition states (∆∆G ⱡ ) for ligand L4 with an isopropyl moiety is 3.8 kJ/mol at 0 C, whereas a simple replacement of isopropyl to isononyl (L6) more than doubles this value to 7.7 kJ/mol. Non-intuitive effects of ligand structure on enantioselectivity are common in asymmetric transition metal catalysis, 12 usually due to the complexity of interactions involved. As shown by the collected data, qualitative conclusions can be drawn from a structure-selectivity relationship but it offers limited design guidance apart from intuitively increasing the length of the alkyl chain, without any notion of shape or properties.</p><p>Ideally, one would prefer to make decisions based on a predicted ee value to start the next ligand synthesis.</p><!><p>Inspired by Sigman's development of predictive and mechanistic multivariate linear regression models for reaction development, 20 we recently reported the optimization of Cu-catalysed ACA to β-substituted cyclopentenones 12 and cyclohexenone 21 with the aid of QSSR. This approach allows one to correlate observed enantioselectivity values against molecular descriptors, quantitative parameters that capture structural and/or electronic differences between the ligands used. These descriptors may be derived from experimental or computed properties in the absence of detailed mechanistic knowledge, and indeed the models may be useful in formulating a mechanistic hypothesis. Computational mechanistic studies (e.g., using density functional theory) have previously aided the optimization of phosphoramidite ligands used in metalcatalyzed asymmetric transformations. 22,23 However, these approaches are significantly more expensive and require prior detailed knowledge of mechanism and competing stereodetermining transition structures. Impressive predictive accuracies of ~2 kJ/mol have been obtained using QSSR models, which should be viewed in a favorable light when compared with the bounds of chemical accuracy attainable by quantum chemical calculations of around ~4 kJ/mol. 24 Furthermore, statistical modelling can accelerate the design of new ligands by ranking their syntheses, which remains the principal bottleneck of the design process. 25 Other promising methods also exist. 26 Incomplete mechanistic understanding and the absence of quantitative guidelines led us towards the use of QSSR, where our strategy was to carry out statistical model construction and ligand synthesis in parallel. Iterative refinement of the model could be accomplished as new data was collected. We aimed to achieve high enantioinduction of 1a, and assumed higher reactivity could also be obtained in the process, as we have previously observed new ligands could increase both yield and enantioselectivity.</p><p>Our ligand design workflow started with the collection and curation of all the available data, regardless of the achieved selectivity (Scheme 2B). This was followed by the generation of steric and electronic descriptors for each ligand, optimized after a conformational search. Internal and external validation of the model was a critical step to get a statistically valid model. One could finally predict the enantioselectivity of ligands in-silico and discard unpromising structures. We only synthesized ligands that would provide useful information to the model or that would likely achieve high enantioselectivity. These synthesized ligands could then be fed to the model such that the QSSR model would gradually get stronger in an iterative way.</p><p>Guided by the qualitative structure-selectivity relationship (Scheme 2A), we restricted ligand modification to structural diversification of the aliphatic R-group only. We reasoned that this reduced search space for ligand optimization could be explored more efficiently, while still providing sufficient variation in selectivity values (as discussed above) from which to extract meaningful structure-selectivity trends. The BINOL backbone and indanyl group were not modified further and were retained in a matched configuration. Following these criteria, nine data points were initially used for model building out of sixteen ligands explored in the initial screening (see Scheme S2 in ESI).</p><p>Computable feature descriptors were generated to quantify the steric and electronic properties of the phosphoramidite ligands (see Table S2 We set boundaries for the exploration of Phosphoramidite Ligand Space based on synthetic accessibility. Ligand synthesis currently represents the bottleneck in our approach, and so we considered only those structures accessible from readily available commercial sources or fragments that could easily be synthesized within four well-established synthetic steps. Ligand synthesis and enantioselectivity prediction were carried out in parallel. Although there is no singular definition for the applicability domain (AD) of a statistical model, and the utility of this concept is contested, 29 we only envisaged potential in-silico ligands possessing aliphatic R groups. Therefore no heteroelements were added to the alkyl substituent even if the lipophilicity value could have been improved. As a rule of thumb, the quality of extrapolative predictions deteriorates further away from the area of feature space spanned by the training data. Inside this space interpolative predictions can be made confidently. 30 An in-silico library of 22 synthetically accessible ligands (see Scheme S3 in ESI) was developed to satisfy the above considerations. Molecular descriptors were computed for each of these ligands and submitted into the model, represented as gray dots in Figure 1. The predicted levels of enantioselectivity were used to plan the next phase of ligand synthesis. We selected evenly spaced values along the range of predicted selectivities (gray labels), focusing our efforts in the region above 6.0 kJ/mol (>85% ee). Inspired by Bayesian Optimization approaches, 31 for which data acquisition is a tradeoff between exploring regions of high uncertainty vs. exploring regions of lower uncertainty, but higher expected values, we set out to improve the predictive power of our model while also targeting higher enantioselectivities.</p><p>The enantioselectivity of L12 was predicted between 79-93% ee (95% confidence interval)</p><p>according to the initial model. Experimentally this was determined as 94% ee. This new data point could now be used to refine (i.e. re-train) the statistical model. By expanding the feature space spanned by the training data, predictions for new ligands can be made more confidently.</p><p>Accordingly, incorporating the newly generated data into model training led to almost identical statistical performance across the training set, but with narrower error intervals. The in-silico ligand library was then predicted again, guiding us next to synthesise L13, predicted to give between 90-98% ee and afforded 92% ee. Slightly narrower confidence intervals were again achieved by feeding the model with more information and similar model quality was achieved.</p><p>Synthesis and testing of L14 resulted in 92% ee, close to the predicted range of 94-99% ee. This approach is illustrative of how a targeted data-collection strategy can be used to iteratively refine an underlying statistical model and generate more confident predictions. For a (multivariate) linear regression, optimization of the output necessarily involves extrapolation to a previously unexplored region of feature space, so the above approach proves particularly useful. Unlike linear models, the optimal values of nonlinear parametric models (e.g. higher order polynomials, 32 support vector machines, 30 random forests 33 ) can lie within the bounds of existing feature space, such that extrapolative prediction may not be necessary to accomplish reaction optimization. Nevertheless, predictive performance can still be enhanced by additional datacollection in sparsely covered regions of chemical space.</p><p>We hypothesized that the correlation of enantioselectivity and lipophilicity might be due to catalyst solubility, whereby lipophilic R groups could help to either solubilize the active catalyst or disperse inactive aggregates. The concentration of active catalyst was varied by an order of magnitude to test this hypothesis. As shown in Figure 1, both reactivity and selectivity were unaffected by concentration, forcing us to abandon this assumption.</p><p>We decided to challenge Model I by preparing phosphoramidite ligands with unsymmetric and more branched alkyl groups, with the indane and BINOL moieties unchanged. Even though the predictions were acceptable and allowed for a slight improvement of enantioselectivity, we decided to build more predictive models with tighter confidence intervals through a more widely-distributed, uniform sample of data points.</p><p>L15, containing a β-cyclocitral derivative in the R group, behaved surprisingly well as it afforded 75% ee with ee values predicted between 75% and 88%. L16, L17 and L18 however behaved unexpectedly and the correlation started to break. L17 gave a striking difference between predicted and measured enantioselectivity, and shows how small structural changes can result in large "cliffs" in terms of enantioselectivity. Such cliff-edge effects are unpredictable by nature, and have similarities to the so-called "magic methyl effect" encountered in drug discovery. 34 As shown in Figure 2A, we observed a jump in selectivity and reactivity in moving from L4 (93%, 67% ee) to L17 (99%, 92% ee). Our model only focuses on enantioselectivity but our objective as always is to achieve good selectivity and reactivity with ACA. Thus L17 ligand afforded similar level of selectivity as previously achieved with ligand L6, but far better reactivity (99% vs 63% isolated yield). Reaction kinetics were also about an order of magnitude faster, with the reaction now typically complete in 30 min.</p><p>This substituent effect was not captured by changes in the lipophilicity descriptors. As shown in Figure 2A, the conformation of L17 is different to L2 such that it affects the ∆∆G ⱡ by +2.92 kJ/mol. The gauche conformation somehow causes a long-distance change in the active 13 catalyst that leads to better enantioselectivity. Superimposition of L4 and L2 proved identical whereas L17 and L18 both had similar gauche conformations that avoid destabilizing synpentane interactions, and which was consistent with the grouping of the observed enantioselectivities for these four ligands. We examined whether the inclusion of additional descriptors would allow us to capture the effect of methylation (exemplified by ligand L4, L2, L17 in Figure 2A). Conformations likely play an important role in enantioselectivity here as highlighted by the improvement obtained by comparing L2 (73% ee) to L6 (92% ee) (Scheme 2A) but no steric parameters initially showed promise (e.g Sterimol). Molecular descriptors might then fail to grasp the important features responsible for enantioinduction since flexible chains are often treated statically in a single conformation. For example, Sterimol steric parameters refer to a particular geometry and do not automatically take into account effects of a conformational ensemble. 20 In contrast to this, weighted Sterimol (wSterimol) 35 parameters report on the Boltzmann average along with minimum and maximum values across the ensemble. Upon examination, wSterimol parameters confirmed the anticipated impact of conformation on the output values and its error (on average ±6 kJ/mol, see Figure S1 in ESI), although no meaningful correlation was obtained using these descriptors.</p><p>Inspired by Doyle's use of electronic structure calculations to generate atomic and molecular descriptors, 33 we used the Spartan package 36 to generate parameters from which the highest occupied molecular orbital (HOMO) energy and dipole moment of the global minimum energy conformer of each ligands were found to correlate with reaction outcome (Figure 1). L19 was then predicted at 78% ee and actually afforded 75% ee, so we decided to continue with this new model. The descriptors of synthetically accessible ligands (represented as gray dots) were computed again and were fed to the newly generated model. L20 followed by L21 and L22 were thus predicted and then synthesized.</p><p>The final model, called model II, possesses a good fit (fourteen ligands, R 2 = 84%, RMSE = 0.91 kJ/mol). The external test set also showed satisfactory correlation (six ligands, R 2 = 86%, RMSE = 0.93 kJ/mol)) and LOOCV remains acceptable (R 2 = 75%, RMSE = 1.16 kJ/mol).</p><p>There are fourteen ligands in the training set for only two descriptors in the model equation and the ANOVA test confirmed the statistical significance of the descriptors (p < 0.05).</p><p>The ligand HOMO energy relates to the non-bonding phosphorus lone pair. Although the classification of molecular descriptors as either electronic or steric is not absolute, 37 a higher HOMO energy is indicative of a more electron-rich -donating ligand with a stronger metalligand bond. A positive coefficient in the regression model indicates that higher HOMO energies lead to higher levels of selectivity. On the other hand, the dipole moment describes the overall charge distribution in the ligand, which also captures the gross molecular shape (e.g. 2.26 D with L4 and 2.09 D with L17, which is an 8% relative difference arising due to changing the length of the alkyl chain). This parameter therefore indirectly reflects steric as well as electronic differences, and is sensitive to the length and branching of the N-alkyl substituent. The model coefficient is negative meaning that smaller dipole moments lead to higher levels of selectivity.</p><p>In total, an in-silico library of 24 synthetically accessible ligands was predicted using the final model. As none of the newly predicted selectivities were in excess of previously realized experimental values, ligand optimization was halted at this stage. We had reached the maximum in selectivity based on the structural diversification of aliphatic R groups.</p><!><p>We finally decided to rank all our synthesized ligands according to their yield and enantioselectivity. Plotting yield versus enantioselectivity, the equation in Figure 2B represents the normalized distance to the origin (0% yield, 0% ee). The simultaneous optimization of more than one objective function (e.g., yield and selectivity) produces sets of equally good, nondominated solutions rather than a singular value. The Pareto set, 26 depicted as the bold curve among the isodistances in the Figure 2B connects those ligands for which there are no other examples superior in both yield and selectivity. The analysis showed that L17 was the best ligand in our library, placed equal first with L14. The synthesis of L14 is more tedious due to the need to synthesize the corresponding ketone in three steps with mediocre yields. It was therefore decided to continue with L17 (derived from a commercially available ketone) as the best ligand in our library. It quickly proved to have an impact outside this work, giving higher levels of reactivity in other reactions such as in the desymmetrization of meso-bisphosphates. 38</p><!><p>The scope of the reaction was finally investigated with our new ligand L17. As well as varying the nucleophiles used we also probed the effects of putting substituents in various positions that were not tolerated in our previous system (Scheme 3). A phenyl ring at the R 2 position (2) gave the desired product with 72% yield and 92% ee. An isopropyl bearing electrophile (3) led to similar levels of selectivity. To our delight, even a tert-butyl group in 4, which is well known to be unsuitably reactive, gave satisfactory yield (71%) and 82% ee. The examination of two other branched and hindered electrophiles at the 4-position provided product with high ee (5 and 6).</p><p>ACA was also effective when R 2 phenyl rings were substituted with a nitro group (7, 91% ee), although an electron donating methoxy group gave poor results (8, 72% ee). Halogen substitution at different position (9, 10 & 11) afforded between 81% and 95% ee. Heteroaromatic rings (12 & 13) were also tolerated, however giving moderate selectivity.</p><p>Substitution on R 1 is well accepted by the catalyst, providing high ee and excellent reactivity in the case of branched aliphatic or aromatic substituents (14, 15 & 16). Even chalcone, to give 17, was tolerated although this was obtained with a lower selectivity (96% yield, 78% ee).</p><p>Different nucleophiles were examined. 18 was obtained with 99% yield and 92% ee (59% yield and 33% ee achieved in previous work). 16 Functionalized alkenes such as bromo-styrene afforded 90% yield of 19 with 93% ee. 6-Chlorohexene gave 20 in high yield (88%) and high ee (91%). Use of protected alcohol (21) provided similar results (62%, 93% ee), with somewhat lower yield due to competitive slow in-situ TBS deprotection.</p><p>Scheme 3. Optimised conditions and substrate scope of the ACA on α,β-unsaturated ketone bearing branched or aromatic moieties.</p><!><p>In conclusion, an iterative protocol has guided the development of a new ligand for transition metal catalyzed asymmetric reactions (L17). The addition onto linear α,β-unsaturated ketones possessing bulky or aromatic groups, chosen as a challenging case study, now proceeds satisfactorily even with bulky tert-butyl -substituents. Key to selectivity was the fine tuning of phosphoramidite ligands, designed with the aid of quantitative structure-selectivity relationships.</p><p>The QSSR approach allowed us to quickly discard unpromising potential ligand structures, which easily justifies the time spent generating models as ligand synthesis remains the bottleneck of the design process. A key lesson from this work is that one should aim for tighter confidence intervals and not just statistically significant models as this allows for a more useful ranking of the in-silico ligands. Selectivity optimization using multivariate linear regression is fundamentally and inescapably an exercise in extrapolative prediction: the targeted collection of new data in unexplored areas of chemical space should be prioritized. At the end, we improved our understanding to reach higher levels of enantioinduction and the method now achieves up to 99% yield and 95% ee on a broader range of substrates. We hope that this work will be used as an example on how to 'fix' an asymmetric reaction, but we also showcase how copper-catalysed ACA is becoming a more robust reaction potentially capable of enriching mainstream synthetic methodologies.</p><!><p>All experimental procedures, characterization data, preliminary results, modeling methods and details about the ligand design (PDF). Code in R and data in CSV files used to generate all the multivariate models (ZIP).</p>

ChemRxiv

Estimation of Migration-time and Mobility Distributions in Organelle Capillary Electrophoresis with Statistical-Overlap Theory

The separation of organelles by capillary electrophoresis (CE) produces large numbers of narrow peaks, which commonly are assumed to originate from single particles. In this paper, we show this is not always true. Here, we use established methods to partition simulated and real organelle CEs into regions of constant peak density and then use statistical-overlap theory to calculate the number of peaks (single particles) in each region. The only required measurements are the number of observed peaks (maxima) and peak standard deviation in the regions, and the durations of the regions. Theory is developed for the precision of the estimated peak number and the threshold saturation above which the calculation is not advisable due to fluctuation of peak numbers. Theory shows that the relative precision is good, when the saturation lies between 0.2 and 1.0, and is optimal when the saturation is slightly greater than 0.5. It also shows the threshold saturation depends on the peak standard deviation, divided by the region\xe2\x80\x99s duration. The accuracy and precision of peak numbers estimated in different regions of organelle CEs are verified by computer simulations having both constant and non-uniform peak densities. The estimates are accurate to 6%. The estimated peak numbers in different regions are used to calculate migration-time and electrophoretic-mobility distributions. These distributions are less biased by peak overlap than ones determined by counting maxima and provide more correct measures of the organelle properties. The procedure is applied to a mitochondrial CE, in which over 20% of peaks are hidden by peak overlap.

estimation_of_migration-time_and_mobility_distributions_in_organelle_capillary_electrophoresis_with_

Introduction<!>Assumptions<!>Review of basic equations<!>Protocol<!>Interpretation of calculated m\xcc\x84<!>Standard deviation of me distribution<!>Largest interpretable p value and threshold saturation<!>Threshold values<!>CE simulations<!>Calculation of me<!>Assessment of eqs 4a and 4b<!>Analysis of migration-time distributions<!>Application to mitochondrial CE<!>Threshold values of p\xcc\x84 vs m\xcc\x84 curve<!>Analysis of simulations with constant peak density<!>Accuracy of me<!>Precision of me<!>Analysis of simulations with different migration-time distributions<!>Accuracy of me,tot<!>Precision of me,tot<!>Estimation of migration-time distribution<!>Application to mitochondrial CE<!>Conclusions<!>

<p>Biological particles including microorganisms, viruses, intact cells, and organelles have negative electrophoretic mobilities that are complex functions of the particles' surface zeta potentials, morphologies, and electrical membrane potentials, as well as the analytical conditions and medium (e.g., electric field and ionic strength)1–6. The mobility of individual particles has been characterized using capillary electrophoresis (CE) and is highly heterogeneous even when particles appear to be identical7. When fluorescently labeled particles are analyzed by CE coupled to a laser-induced-fluorescence detector (LIF), the electropherogram consists of a collection of narrow peaks that migrate out in a migration-time window that is dependent on the separation conditions and, most importantly, on the mobility range of the particles in the biological sample. The resultant window can be interpreted as a migration-time distribution, which can be transformed into a mobility distribution.</p><p>Mobility distributions are important, because they show quantitative differences among different organelle types8 and the influence of organellar content on mobility6. For example, they reveal the differences among mitochondria from different muscle types of animals with different ages9. The distributions also are indicators of the gentleness or harshness of procedures for sample preparation10. For distributions to have diagnostic potential, they must be determined without bias. A source of bias addressed by this paper is the overlap of peaks in the CEs, from which the distributions are determined.</p><p>A "peak" is defined as a fluorescence intensity profile of a single organelle as it travels through the LIF detector, whereas an "observed peak" is a detected maximum that is comprised of one or more organelle peaks. The numbers of peaks and observed peaks differ because of peak overlap. If we could separate most peaks, we could approximate the migration-time distribution by partitioning a CE into bins and counting the peaks in each. In a recent paper, we used statistical-overlap theory (SOT) and an analogy to the Type-II error of hypothesis testing to predict conditions, for which peak overlap is small enough that the numbers of peaks and observed peaks in such bins are statistically indistinguishable11. Under these conditions, migration-time and mobility distributions can be estimated simply by counting observed peaks in the CE.</p><p>For any analysis the organelle concentration has a value, beyond which these conditions are not met because too many peaks overlap. In principal, the analysis can be improved or the sample can be diluted or injected in smaller amounts. However, it may be difficult to improve the analysis. Furthermore, sample dilution and subsequent analysis are not always possible, because organelles have a finite lifetime. Samples also may be limited in size or to a single injection, as are individual cells12,13. In such cases, we may be forced to interpret organelle CEs with many overlapping peaks. In this paper, we develop procedures for using SOT to estimate the actual numbers of peaks in the different bins from the numbers of observed peaks. With these procedures, the total number of peaks in the CE, and the migration-time and mobility distributions, can be estimated.</p><p>Because organelle CE entails the injection of only hundreds to thousands of discrete particles, statistical fluctuations among otherwise replicate injections affects the numbers of peaks and observed peaks. Because these numbers differ, the histograms determined from different CEs vary, causing an uncertainty in the migration-time and mobility distributions determined from any one CE. We must know the uncertainty to evaluate the distributions' reliability.</p><!><p>The objectives of this paper are to use SOT to estimate accurate and precise numbers of peaks, migration-time distributions, and mobility distributions from organelle CEs. Several approaches exist in SOT14–18, but in all cases a separation is a member of a large ensemble of separations, in which the number, migration times, and heights of peaks are governed by probability distributions. Here, the CE bins are chosen such that the peak density in each is nearly constant. However, the density can vary among different bins, and each bin has its own ensemble. The variation of peak numbers models the variation of the injected number of organelles; the variation of migration times models the variation of mobilities among different organelles. The number of observed peaks also varies among ensemble members.</p><!><p>Let m and p equal the numbers of peaks and observed peaks in an ensemble member (i.e., bin), with m̄ and p̄ equaling the means of these numbers in the ensemble. The means are related by the saturation14,17</p><p> (1a)α=4m¯σRs∗/X which is a metric of peak crowding. Peak numbers and migration times usually are modeled by Poisson statistics for empirical15,19,20 and theoretical14,16,21,22 reasons. For a Poisson distribution of peaks having constant density and width14</p><p>In eq 1a, σ is the peak standard deviation, X is the bin duration, and Rs∗ is the average minimum resolution that separates adjacent peaks. The value of Rs∗ is not arbitrary but depends on the ensemble's distribution of peak heights23 and amount of peak overlap24 in a manner that is non-intuitive to most practitioners. The peak-height distribution usually is modeled by an exponential function, which is consistent with the empirically measured or estimated distribution of peak heights in complex mixtures19,25–28 and also the prediction of theory28,29. Recently, this Rs∗ function was absorbed into the saturation by defining a new variable, the effective saturation αe30</p><p> (2a)αe=α/Rs∗=4m¯σ/X which allowed eq 1b to be fit by the empirical expression</p><p> (2b)p¯=m¯/(1+β1αe+β2αe2) over the range, 0 ≤ αe ≤ 25, with β1 = 0.775 ± 0.002 and β2 = 0.0750 ± 0.0009.</p><p>Eqs 1 and 2 express the same result. In this paper, we express all theory relative to eq 1, because SOT fundamentally depends on the saturation α. Thus, the effective saturation αe is not explicitly used, although some findings are reported relative to it. Practitioners may prefer eq 2 because it is independent of Rs∗ and avoids the numerical analysis needed with eq 1 (see Procedures section). Conversions between α and αe are reported in Part One of Supplementary Material to this paper.</p><p>For Poisson statistics, the variance of the peak number m, σm2, among ensemble members is m̄. The variance of the observed peak number p, σp2, is31</p><!><p>We partition CEs into bins of known duration X, as shown in Figure 1a, using a standard statistics formula and count the number of observed peaks (maxima) p in each bin. For each bin, we set p equal to p̄, eq 1b, and calculate m̄ from the known peak standard deviation σ and predicted Rs∗ value. Here, σ is assumed constant to model peak widths in CEs having a large post-column sheath flow32. Unlike previous calculations of m̄ determined by least-squares fittings to eq 1b of multiple p values from different separations19,27,33, the calculation of m̄ is exact, requiring new theory that restricts p to values less than eq 1b, avoids the "double-value" problem14, and evaluates the calculation's precision. This theory is discussed here but details are deferred to Supplementary Material for brevity's sake. These calculations produce an m̄ estimate and its standard deviation for each bin. The sum of the m̄'s is the total number of peaks in the CE, whereas the m̄'s of different bins determine the discrete SOT-estimated migration-time distribution. The dashed curve, dashed-line histogram, and solid-line histogram in Figure 1b are the actual migration-time distribution in Figure 1a, the SOT-estimated distribution, and the approximate distribution obtained by simply counting the numbers of observed peaks. The graph ordinate is the density, or the number of peaks (or observed peaks) per unit time. The first two distributions agree closely, but the last one is biased because of peak overlap, illustrating the need for SOT calculations. The error bars are the standard deviations of the m̄'s, which gauge the precision of the SOT-estimated distribution. The inset to Figure 1b is the SOT-estimated mobility distribution, calculated from the SOT-estimated migration-time distribution by mapping the temporal bin boundaries into mobilities. Because the relation between time and mobility is non-linear and because mobilities are signed, the orientation, bin height, and bin width differ from the SOT-estimated migration-time distribution.</p><p>Figure 1c is a graph of four model migration-time distributions used to simulate CEs and test the protocol. The distributions are graphed against reduced time ζ (which lies between 0 and 1) and scaled relative to ordinate f (ζ), such that the area under each is unity. The distributions differ markedly and are assigned names for easy reference (Gaussian, bimodal, asymmetric, and constant). On interpreting the simulations by SOT, we assume the distributions are unknown and subsequently assess their accuracy.</p><!><p>The implications of calculating m̄ from a single p value and eq 1b are discussed. Figure 2a is a graph of p̄ vs m̄ based on eq 1, when peak heights are exponentially random. This curve describes a relation between means. The p value of any bin rarely equals the mean p̄, however, because of statistical fluctuation. Figure 2b shows the consequences of equating them. The circle on the p̄ vs m̄ curve represents the means (m̄, p̄) of a particular ensemble. The bold curve in the figure's lower center is the probability distribution of the number of peaks (i.e., the m distribution). The bold curve to the right is the probability distribution of the number of observed peaks (i.e., the p distribution). Their means coincide with the circle. Both distributions are discrete but are shown as continuous functions for simplicity.</p><p>Consider the p value identified by the horizontal arrow, a, in Figure 2b. On its identification with eq 1b, it is mapped by the p̄ vs m̄ curve into the m̄ value associated with the vertical arrow, a′. However, this m̄ is not the ensemble mean. Rather, it is the mean of another m distribution, represented by the finely dashed curve. A similar argument can be made for the p and m̄ values connected by arrows b and b′; this m̄ is the mean of a third m distribution, represented by the coarsely dashed curve. However, the m̄'s so determined almost coincide with m values actually belonging to the ensemble, as represented by the squares in Figure 2b. This coincidence is the basis of our analysis: m̄'s determined by identifying eq 1b with single random p values form a distribution related to the m distribution. These determined m̄'s henceforth are called me to distinguish them from the ensemble mean, and they comprise the me distribution.</p><!><p>The me distribution is broadened by our analysis. This is shown in Figure 2c, in which circles represent two different (m̄, p̄) coordinates. The arrows have the same meaning as in Figure 2b and show that the me distribution broadens as ∂p̄/∂m̄ decreases (the derivative is written as a partial to show that σ/X is constant). This broadening decreases the precision of me, reducing its reliability. The standard deviation σme of me is derived in Part Two of the Supplementary Material and can be expressed as</p><p> (4a)σme(σ∣X)1/2=12(αRs∗[1−2αexp(−α)])1/2exp(α/2)[α−1−dlnRs∗/dα]α−1−1−dlnRs∗/dα or as the coefficient of variation (CV), equal to 100 σme/m̄</p><p>These equations can be evaluated at any α if Rs∗ is known. For exponential peak heights, Rs∗ can be predicted from theory24; alternatively, the numerical solution so predicted is well fit by the empirical equation</p><p> (4c)Rs∗=Rs∗(α)=0.725+β3α+β4α2+β5α3 over the range, 0 ≤ α ≤ 3.85, with β3 = −0.1910 ± 0.0001, β4 = 0.0039 ± 0.0001, and β5 = 0.00188 ± 0.00002. The coefficient 0.725 in eq 4c is the theoretical value of Rs∗ at α = 0 23. The functions σme and CV in eqs 4a and 4b equal or exceed their counterparts for the Poisson distribution, which are m¯ and 100/m¯. Monte Carlo simulations of the me distribution are reported in Part Three of the Supplementary Material.</p><p>For good precision, me should be calculated to the left of the maximum of the p̄ vs m̄ curve in Figure 2a (i.e., the low-m̄ side), where ∂p̄/∂m̄ is large. This action also avoids the "double-value problem"14, i.e., the correspondence of p̄ to two m̄'s except at the maximum. However, p̄ should not be near the maximum for another reason.</p><!><p>Figure 2d shows that p̄, m̄, and α have threshold values on the low- m̄ side of the p vs m̄ curve, beyond which large outliers of the p distribution can exceed the curve maximum. If this happens, eq 1 has no solution. We can derive a relation among the threshold value of p̄, the curve maximum, and the width of the p distribution that assures us that this rarely happens.</p><p>The maximum p̄ value, p̄max, is found by calculating ∂p̄/∂m̄ from eq 1 and setting it equal to zero</p><p>The expression ∂α/∂m̄ is given by the reciprocal of eq S-9 in Part Two of the Supplementary Material, whereas dlnRs∗/dα is evaluated from eq 4c. The solution to eq 5 is given in the Results and Discussion; for the moment, we observe that p̄max is a multiple of the reciprocal σ/X ratio, i.e., p̄max = δ X/σ, where δ is a constant.</p><p>For m̄ ≥ 10, the p distribution is described well by a Gaussian envelope having mean p̄ and standard deviation σ p 11,15. A Gaussian distribution has negligible density three to four standard deviations from its mean. Therefore, the threshold saturation αt, or the largest saturation at which statistical fluctuations of p cause very few (if any) problems with estimating me, is established by the equation</p><p> (6a)p¯t+γσp=p¯max where p̄t is the threshold value of p̄ and γ is a parameter between 3 and 4. Figure 2d shows the relation among p̄t, the threshold peak number m̄t, γ σp, and p̄max. Using eqs 1a, 1b, and 3, and the expression p̄max = δ X/σ, we can explicitly write eq 6a as</p><p> (6b)αt4Rs∗(αt)Xσexp(−αt)+γ2αtRs∗(αt)Xσexp(−αt)[1−2αtexp(−αt)]=δXσ with Rs∗(αt) equal to eq 4c, as expressed with α =αt. Eq 6b shows the threshold saturation αt is determined by parameter γ and the reciprocal σ/X ratio.</p><!><p>Eq 5 was solved by bisection to determine the scalar δ. For various σ/X, eq 6b was solved for αt by bisection, with γ = 3. The threshold m̄t value then was calculated from αt, eq 1a, and eq 4c. Values of the threshold effective saturation were calculated from αt as described in Part One of the Supplementary Material. CEs were simulated to verify α t.</p><!><p>CEs were simulated as previously described24 along a reduced time coordinate ζ = (t − to)/1D between 0 and 1, with t, to, and 1D equaling time, the time of the first peak's elution, and the CE duration, respectively. The Box Muller transform34 was used to mimic Poisson distributed peak numbers. In any CE, peaks were Gaussians with constant standard deviations and exponentially random heights. The number of observed peaks (maxima) was determined.</p><!><p>The peak number me was calculated from the number of observed peaks p in a bin, the σ/X ratio, and eq 1. The calculation was a dual bisection, in which m̄ = me determined α via eqs 1a and 4c, and then in which m̄ was adjusted so that eq 1b equaled p. (Alternatively, practitioners may wish to determine me from eq 2b, whose solution is</p><p>Evaluating eq 7 is a straightforward computation and simpler than performing the dual bisection, but the me so determined is slightly less accurate since it is based on a fit.)</p><!><p>Five hundred CEs of constant peak density (i.e., single-bin CEs) were simulated for different m̄ and for σ/X ratios ranging from 0.00005 to 0.0013. Peak numbers me were estimated as described above. To assess accuracy, the percentage error between the average me and m̄ was calculated. The standard deviation and CV of me were compared to eqs 4a and 4b. The minimum of eq 4b was determined with the golden search algorithm34.</p><!><p>Five hundred CEs were simulated with the four model migration-time distributions in Figure 1c. The ratio of the peak standard deviation to the CE duration, σ/1D, was 8 × 10−5 or 8 × 10−6 (these are realistic values in organelle CE, e.g., σ = 0.008 s and 1D = 100 – 1000 s). Equations for the distributions and their migration times are reported in Part Four of the Supplementary Material.</p><p>For each distribution and σ/1D ratio, the mean number of peaks in the entire CE, m̄tot, was incremented (e.g., m̄tot = 100, 200, …) until the saturation of the bin of greatest peak density exceeded the threshold saturation αt determined as described below. The CE was partitioned into N bins of equal duration X</p><p> (8)N=trunc[(2ptot)1/3]+1 where ptot is the total number of observed peaks in the entire CE and trunc means truncation. Eq 8 is one bin larger than the minimum bin number of a common statistics formula35 (with ptot equaling the number of data points for binning) and it determined the σ/X ratio as Nσ/1D (i.e., X = 1D/N). The peak number me in each bin was calculated as described above. The number of peaks in the entire CE, me,tot, was calculated by summing estimated peak numbers in all bins</p><p> (9)me,tot=∑i=1Nmei where m is the me estimate for the ith bin. The average, standard deviation, and CV of me,tot values were calculated.</p><p>To assess accuracy, the percentage error between the average value of me,tot and m̄tot was calculated. The standard deviation of me,tot was compared to its theoretical counterpart calculated by error propagation of eq 9</p><p> (10)σme,tot=[∑i=1Nσmei2]1/2 with σmei2 equaling the variance of the ith bin, as computed from eqs 4a and 4c. The saturation α in these equations was calculated by estimating the mean numbers of peaks and observed peaks in each bin from SOT equations for non-uniform migration-time distributions36, identifying these estimates with p̄ and m̄, and substituting them in eq 1b. This calculation provides the best estimate of the saturation, with which to associate eq 1b. The theoretical CV was calculated as the product of eq 10 and 100/m̄tot, and was compared to the CV's determined by simulation.</p><p>Average histograms of the four migration-time distributions were calculated for different m̄tot from me's determined from 500 simulations. They were compared to the actual normalized distributions by scaling the histogram areas to equal the average of me,tot, divided by m̄tot. The scaling makes it easy to compare distributions for different m̄tot and to assess if the distributions are underestimated or overestimated.</p><!><p>The total number of peaks in, and the migration-time and mobility distributions of, a mitochondrial CE37 were estimated. The CE was acquired at 200 Hz. Maxima produced by baseline noise were removed by clipping as described elsewhere11. In brief, the standard deviation of the baseline noise was estimated from three signal-free regions spanning 50 to 98.6 s, six multiples of it were added to a linear interpolation of the lower noise bound estimated at 145 points, and all signal below the sum was clipped. This action is appropriate for Gaussian noise, which is negligible three or more standard deviations from the mean. The peak standard deviations σ of 10 isolated observed peaks spanning the CE were determined by moments analysis. The migration-time distribution was constructed as described above and transformed into a mobility distribution by mapping the times t of its bin boundaries into mobilities μ</p><p> (11)μ=L/(Et)−μeo where L is the capillary length (0.4 m), E is the electric field strength (40 kV/m), and μeo is the electroosmotic mobility (5.2 × 10−8 m2/V-s).</p><!><p>By solving eq 5, we find the p̄ vs m̄ curve maximizes at α ≈ 1.601. (If Rs∗ were constant, it would maximize at α = 114.) On substituting this α and its associated Rs∗ value (see eq 4c) into eq 1, we discover that p̄max is 0.185 X/σ. Thus, the scalar δ in eq 6 is 0.185.</p><p>Figure 3a is a graph of the threshold saturation αt, threshold effective saturation αe,t, and logarithm of the threshold peak number m̄t vs the logarithm of the σ/X ratio, as determined from eq 6b. All thresholds increase with decreasing σ/X because σp/p̄ decreases, allowing p̄t to move closer to the curve maximum. For a Gaussian p distribution, the probability that p exceeds p̄max is [1−erf(3/2)]/2=0.00135. The circles in Figure 3a are αt values determined by extrapolating p in 50,000 simulations to the expected number of excesses, 67.5 = (0.00135)(50000). Least-squares fits to the curves in this figure are reported in Part Five of the Supplementary Material.</p><!><p>Our procedure is based on applying SOT to different bins of a partitioned CE. The peak density in each bin is roughly constant, but it can vary among different bins. We first assess the accuracy and precision of me values determined from a single bin of constant peak density.</p><!><p>Figure 3b is a graph of the percentage error between m̄ and the average me of 500 simulations vs saturation α for five σ/X ratios spanning a 26-fold range. (The σ/X ratios in subsequent determinations of migration-time distributions lie within this range.) The results are somewhat scattered but the trend is clear. The accuracy decreases with α, because eq 1 slightly underestimates peak overlap at high saturation. For α ≤ 1, the accuracy is −6% or better.</p><!><p>Figure 3c is a graph of the standard deviation of me, σme, multiplied by (σ/X)1/2 vs α. Figure 3d is a graph of the coefficient of variation, CV = 100σme/m̄, divided by (σ/X)1/2 vs α. The symbols are results of 500 simulations for the same σ/X ratios in Figure 3b; the curves are graphs of eqs 4a and 4b. These equations are based on a propagation of errors (see Part Two of the Supplementary Material) and overestimate σme and CV at large σme but they agree with simulation for α ≤ 0.8 or so. The bold curves are graphs of</p><p> (12)σme(σ/X)1/2=12αRs∗;CV(σ/X)1/2=200Rs∗α which are the limits of eqs 4a and 4b as α approaches zero. On substituting eq 1a into eq 12, we find the limits are σme=m¯ and CV=100/m¯, that is, the Poisson limits stated earlier. With increasing α, σme and CV exceed these limits, because the me distribution is broadened by the decreasing slope, ∂p̄/∂m̄, of the p̄ vs m̄ curve.</p><p>The graph of CV/(σ/X)1/2 has a shallow minimum at α ≈ 0.524, with a value of 320.5. The minimum occurs, because Poisson statistics reduces the CV as m̄ (or α) increases, whereas the diminishing slope, ∂p̄/∂m̄, increases the CV as m̄ (or α) increases. Ultimately the latter effect dominates. It is fortuitous that the minimum is shallow, with roughly the same precision over the range, 0.2 ≤ α ≤ 1.0. Regardless of the threshold saturation, we suggest that α not exceed unity if good precision is sought.</p><p>Figure 3d shows that the relative precision is best, when σ/X is small. For example, if m̄ is decreased by 10-fold and σ/X is increased by 10-fold, the saturation α doesn't change. Therefore, the right-hand sides of eqs 4a and 4b don't change, but the left-hand sides do. The estimate me is 10-fold smaller, but σme is smaller by only 10 and the CV is larger by 10.</p><!><p>Our estimates from CE simulations having the migration-time distributions in Figure 1c follow the same trends as do simulations of constant peak density. This is not surprising, since the estimates entail repetitive applications of eq 1 to contiguous bins of approximately constant peak density. However, the quantitative results vary with the distribution.</p><!><p>Figure 4 is a graph of the percentage error between the mean total number of peaks m̄tot and the average me,tot of 500 simulations vs the logarithm of m̄tot. Each me,tot was calculated from eq 9 for the four model distributions and the two σ/1D ratios, 8 × 10−5 and 8 × 10−6. The m̄tot range is greater for the smaller σ/1D ratio, because the saturation is proportional to the product of m̄ and σ. The number N of bins ranged from 6 to 45, with σ/X ratios between 1.0 × 10−4 and 1.3 × 10−3. The symbols at the curve ends have m̄tot values, at which the saturation in the bin of maximum peak density exceeded the threshold saturation. As in Figure 3b, the results are somewhat scattered but the trend is the same, with the average of me,tot underestimating m̄tot at high saturation and an accuracy of roughly −6% or better. Some m̄tot values greatly exceed typical particle numbers in organelle CE, but they show the potential of theory.</p><!><p>Figure 5a is a graph of σme,tot vs m̄tot for σ/1D = 8 ×10−6. The symbols are results of 500 simulations for the four model distributions. The curves were calculated from eq 10 and agree with simulation, indicating the total variance can be modeled by a discrete sum of variances for different bins. As before, σme,tot exceeds the Poisson limit, σme,tot=m¯tot, represented by the bold curve. For a given m̄tot, σme,tot differs for the different distributions because eq 10 and α vary with changing peak density. Figure 5b is the corresponding graph of the coefficient of variation, CV, vs m̄tot. As before, minima are found, but the m̄tot at which minimization occurs and the minimum itself differ for different distributions.</p><p>The CV is a metric of precision for the entire CE. We may be more interested in good precision in bins of high peak density, since bins of low peak density have little impact on the distribution. By weighting the CV by the fraction of peaks in different bins, we can target m̄tot values for optimal precision where most peaks are found. Unlike most separations, in favorable cases the CV can be adjusted in organelle CE by diluting or concentrating the sample. Accordingly, we define a weighted CV</p><p> (13)CVwt=∑i=1Nm¯iCVim¯tot≈∑i=1Nme,iCVime,tot where CVi is the CV of the ith bin. Figure 5c is a graph of CVwt vs m̄tot for the four distributions. The small discontinuities in the curves occur as the bin number changes. As before the m̄tot producing the minimum CVwt differs with the distribution, but now the minimum itself is the same (about 6). The unifying attribute is shown in Figure 5d, which is a graph of the saturation α in different bins vs the reduced time ζ. For nonuniform distributions, the α's of bins with large peak density are slightly greater than 0.5. This α range produces a small CV for a single bin, as shown in Figure 3d, and it is not surprising that a sum of CV's weighted by the fraction of peaks follows the same trend. Similar trends to those in Figure 5 are found for the σ/1D ratio, 8 × 10−5.</p><!><p>Figure 6 reports graphs of the four normalized distributions f (ζ) vs reduced time ζ, first shown in Figure 1c, for the two σ/1D ratios. The symbols and error bars are the means and standard deviations of me's determined from 500 simulations for three m̄tot values, scaled such that the distribution area is the average me,tot, divided by m̄tot. Each symbol corresponds to a different bin. For each distribution, one value of m̄tot corresponds to the minimum CVwt, whereas the other m̄tot values are smaller and larger. Although all distributions are estimated well, the precision is best at the minimum CVwt if we consider only the bins of large peak density. If possible, separation conditions should be adjusted such that these bins have saturations slightly greater than 0.5.</p><!><p>Figure 7a is a noise-clipped mitochondrial CE containing 1707 observed peaks and first discussed in ref 37. The inset expands the CE and shows overlapping maxima, suggesting that SOT estimates could be useful. The CE first was partitioned among uniform bins. For such bins, however, detail was lost around the maximum peak density near 380 s and slowly varying densities were binned to excess. Accordingly, eq 8 was applied to three regions spanning 0 to 332 s (ptot = 99; N = 6; X = 55.3 s), 332 to 770 s (ptot = 1355; N = 14; X = 31.3 s), and 770 s to 1848.6 s (ptot = 253; N = 8; X = 134.8 s). The average standard deviation σ calculated from 10 observed peaks was 11.1 ms, with a coefficient of variation equaling 8.4. These values are almost constant due to a large post-column sheath flow. It also is likely they are peaks of single mitochondria, because the coefficient of variation is small. The σ/1D ratio is 6.0 × 10−6; the σ/X ratios range from 8.2 × 10−5 to 2.0 × 10−4, for which the threshold saturation αt is 1.1 or larger (see Figure 3a).</p><p>The estimated peak number me in each bin was determined from eq 1 and the number of observed peaks p. The saturation α of each bin was approximated from eq 1a as19</p><p> (14)α≈−ln(p/me) and used to evaluate equations dependent on α. Eq 14 was used to estimate α instead of the detailed procedures outlined earlier, because p and me are the only metrics we have (i.e., we do not know p̄ and m̄). The saturation in the bin of greatest peak density is 0.52, which is less than αt and the upper saturation, α ≈ 1, for precise determinations of me.</p><p>Figure 7b reports histograms of migration-time distributions, as determined by estimating me (dashed line) and counting observed peaks (solid line). Because the bins have different widths, the ordinate was scaled such that the bin area equals the number of peaks or observed peaks. The error bars are the scaled standard deviations of me. The estimated total number of peaks me,tot is 2163̄ ± 56̄ (CV = 2.6). Over 20% of the peaks are hidden by peak overlap, mostly in the second region.</p><p>The numbers of peaks and observed peaks differ in the three bins of highest density, and they differ somewhat in the two bins to their right (Figure 7b). We evaluated the observed-peak numbers in these bins with our simple procedure based on the Type-II error analogy11. As noted earlier, this procedure allows us to decide if the numbers of peaks and observed peaks are statistically the same. For the three bins of highest density, the procedure is not satisfied, and for the two other bins it shows the numbers of observed peaks are questionable estimates. However, the observed-peak numbers in the other bins satisfy the procedure. Thus, our current work is consistent with our earlier work.</p><p>Figure 7c shows the mobility distributions calculated from the migration-time distributions and eq 11. As before, the bin areas equal the numbers of observed peaks or peaks. Because the relation between time and mobility is non-linear, and the mitochondrial and electroosmotic mobilities have opposite signs, the two distribution types appear differently. Most peaks overlap in the mobility range, −2 × 10−8 to −3 × 10−8 m2/V-s.</p><!><p>We have shown that SOT can be used to estimate both accurate and precise migration-time and mobility distributions in partitioned organelle CEs from a series of single observed-peak numbers, as long as the saturation does not exceed unity. Usually, organelle CEs are assumed to be free of peak overlap. Our work shows this is not always true. The mitochondrial CE in Figure 7a has significant peak overlap, requiring SOT calculations to reduce distribution bias. We recommend first screening organelle CEs for peak overlap by our simple procedure based on a Type-II error analogy11, and then making SOT calculations if necessary.</p><p>It may be tempting first to transform the migration times of observed peaks into mobilities and then to make the peak-overlap analysis in mobility space, with bins of equal width in mobility. This is unwise, because peak standard deviations that are constant in time (as in organelle CE with post-column sheath flow) are not constant in mobility. They not only vary from bin to bin but within a bin, complicating the analysis.</p><p>We recognize that the number of bins in the SOT-estimated migration-time and mobility distributions is not optimal, because the bin number is calculated from the total number of observed peaks (see eq 8). Ideally, the bin number should be calculated from the total number of peaks, but of course this is unknown prior to the SOT estimation. Fortunately, Figure 6 shows the error in the estimated distributions is small.</p><p>Unfortunately, our procedure cannot be extended with high precision to chromatography, except in special cases. Consider gas and liquid chromatography, wherein one might find (for example) 300 observed peaks in a 30-min separation. According to eq 8, these separations would partition among 9 bins of duration X = 3.33 min. For peak standard deviations σ of 2 – 4 s, σ/X would range from 0.01 to 0.02. Even the minimum CV of a bin would vary from 32 to 45, as evaluated from eq 4b. Much smaller peak widths would be required for a precise application.</p><!><p>a) Simulated organelle CE. Dashed lines represent bin boundaries spanned by interval X. b) Histograms of SOT-estimated (dashed line) and observed-peak (solid line) migration-time distributions of CE in a). Dashed curve is actual migration-time distribution. Inset is SOT-estimated mobility distribution. c) Graphs of four normalized model migration-time distributions f (ζ) vs reduced time ζ used to simulate organelle CEs.</p><p>Distribution and precision of me. a) Graph of p̄ vs m̄ for exponentially random peak heights. b) Mapping of p into m̄ by the p̄ vs m̄ curve. Circle represents means of a particular ensemble; bold curves are p and m distributions of the ensemble; dashed curves are m distributions of other ensembles; squares are members of the ensemble's m distribution. c) Graph of p̄ vs m̄, showing broadening of the me distribution from reduction of ∂p̄/∂m̄. Error bars represent standard deviation σp of p distributions. d) Relation among upper limit to p̄, curve maximum, and width of the p distribution.</p><p>Results for single-bin separations of constant peak density. Symbols in panels b) - d) are simulation results for σ/X equaling 0.00005 (), 0.0001 (○), 0.0005 (□), 0.0009 (◇), and 0.0013 (△). a) Graph of threshold saturation αt, threshold effective saturation αe,t, and logarithm of threshold peak number m̄t vs log (σ/X). Circles are αt values determined by simulation. b) Graph of percentage error between m̄ and average me vs saturation α. c) Graph of standard deviation of me, σme, multiplied by (σ/X)1/2 vs α. d) Graph of coefficient of variation, CV = 100σme/m̄, divided by (σ/X)1/2 vs α.</p><p>Accuracy of SOT estimates for migration-time distributions in Figure 1c. Graph of percentage error between m̄tot and average me,tot of simulations vs log (m̄tot). Line types are the same as in Figure 1c (e.g., solid for Gaussian migration-time distribution). Normal-weight curves are results for σ/1D = 8 × 10−5; bold curves, for σ/1D = 8 × 10−6. Ordinates of symbols are m̄tot values, at which the largest saturation exceeds αt</p><p>Precision of SOT estimates for migration-time distributions in Figure 1c and σ/1D = 8 × 10−6. a) Graph of standard deviation σme,tot vs m̄tot. Symbols are simulation results (Gaussian (○), bimodal (△), asymmetric (◇), constant (□)). b) As in a), but for the coefficient of variation CV. c) Graph of weighted coefficient of variation CVwt vs m̄tot. d) Graph of saturation α vs reduced time ζ, when m̄tot corresponds to the minimum CVwt. Values of m̄tot are 10750 (Gaussian), 19000 (bimodal), 16000 (asymmetric), and 23000 (constant).</p><p>Scaled SOT-estimated migration-time distributions for m̄tot values less than (circles), equal to (squares), and greater than (diamonds) m̄tot at the minimum CVwt. a) σ/1D = 8 × 10−6, with m̄tot equaling 750, 10750, and 21800 (Gaussian); 1450, 19000, and 34750 (bimodal); 1275, 16000, and 31300 (asymmetric); and 1575, 23000, and 51000 (constant). b) σ/1D = 8 × 10−5, with m̄tot equaling 160, 950, and 1675 (Gaussian); 270, 2000, and 2700 (bimodal); 250, 1650, and 2425 (asymmetric); and 325, 2500, and 4900 (constant).</p><p>Experimental organelle CE data. a) Mitochondrial CE. Arrows near time axis are demarcations among three regions. Inset shows peak overlap in region of maximum peak density. b) Histograms of migration-time distributions estimated by SOT (dashed line) and obtained by counting observed peaks (solid line). Main panel shows distributions in second region; inset shows distributions in first and third regions, with break between them. Error bars are SOT-estimated standard deviations. c) Mobility distributions determined from SOT-estimated and observed-peak migration-time distributions.</p>

PubMed Author Manuscript

Characterization of the Human Sigma-1 Receptor Chaperone Domain Structure and Binding Immunoglobulin Protein (BiP) Interactions*

Background: Sigma-1 receptor is a ligand-regulated membrane protein chaperone involved in BiP regulation and the ER stress response.Results: The chaperone domain of human sigma-1 receptor is mostly helical with short extended regions.Conclusion: Regions of the sigma-1 receptor chaperone domain implicated in ligand and cholesterol binding can be mapped to separate helices.Significance: A structural framework for delineating sigma-1 receptor BiP and ligand interactions is presented.

characterization_of_the_human_sigma-1_receptor_chaperone_domain_structure_and_binding_immunoglobulin

Introduction<!><!>Protein Sample Production<!>S1R(cd) Sample Preparation<!>NMR Spectroscopy<!>Water-soluble Paramagnetic Studies<!>BiP Interactions<!>Circular Dichroism<!>S1R(cd) Sample Production and Structural Overview<!><!>S1R(cd) Sample Production and Structural Overview<!><!>S1R(cd) Sample Production and Structural Overview<!><!>S1R(cd) Sample Production and Structural Overview<!>Backbone Dynamics of S1R(cd)<!><!>Water and Micelle Interactions of S1R(cd)<!><!>Water and Micelle Interactions of S1R(cd)<!>S1R(cd) Interactions with BiP<!><!>S1R(cd) Interactions with BiP<!>DISCUSSION<!><!>DISCUSSION<!>

<p>The sigma-1 receptor (S1R)3 is a ligand-regulated membrane protein chaperone involved in the ER stress response and interorganelle communication (1–3). S1R is localized to mitochondria-associated ER membranes (4, 5), which are sites for regulation of mitochondrial bioenergetics via ER calcium release (6). S1R is expressed primarily in cerebral cortex, hippocampus, and cerebellum Purkinje cells (7, 8), and has been proposed as a target for treatment of central nervous system diseases, including amnesia, pain, schizophrenia, clinical depression, Alzheimer disease, stroke, and addiction (9, 10). S1R activity is modulated by N,N-dimethyltryptamine (11), progesterone (12), and sphingosine (13). In addition, S1R is regulated by a large number of exogenous small molecules, including opiates, antipsychotics, antidepressants, antihistamines, phencyclidine-like compounds, β-adrenergic receptor ligands, and cocaine (reviewed in Ref. 10).</p><p>Activated S1R dissociates ankyrin from the inositol triphosphate receptor (3, 14), which results in calcium release at the mitochondria-associated ER membrane that is efficiently taken up by mitochondria to increase energy production. S1R appears also to have other roles in the stress response. ER calcium depletion or agonist binding dissociates S1R from the Hsp70 protein BiP, resulting in activation of protein chaperoning activity in both BiP and S1R (1). Chaperone activity and BiP binding (1) and the interaction of S1R with inositol triphosphate receptor (14) have been localized to the C-terminal domain of S1R, and truncation of the C terminus leads to dysfunction in mitochondrial stress response (15).</p><p>The chaperone domain is C-terminal to two putative transmembrane domains (residues 11–29 and 80–100) but contains a predicted membrane associated region (residues ∼176–204) containing two cholesterol recognition motifs (CRM) (Fig. 1) (16, 17). We have studied the S1R chaperone domain (residues 112–223; S1R(cd)) reconstituted into detergent micelles by solution NMR. S1R(cd) solubilized in dodecylphosphocholine (DPC) adopts a conformation competent to bind BiP in a calcium-dependent manner. Analysis of the structure and dynamics indicates that S1R(cd) contains five helices and at least two short extended regions. Three of the helices in the C-terminal membrane-associated domain correspond closely to regions previously implicated in cholesterol and drug interactions.</p><!><p>Full-length S1R topology schematic showing two predicted transmembrane helices (TM1 and TM2) and the membrane-associated domain. The human S1R(cd) construct studied here contains residues 112–223 that include a predicted disordered region (residues ∼134–167 (40)) and a predicted membrane-associated domain (residues ∼176–204 (47)). The N and C termini and the approximate positions of residues 112 and 223 (the C-terminal residue) are indicated. The membrane topology is based on Refs. 1 and 48. TM1, transmembrane domain 1; TM2, transmembrane domain 2; Cyt., cytosol.</p><!><p>An ACA-free gene construct (GeneArt) containing a (His)6-tag, a factor Xa cleavage site, and residues 112–223 of human S1R was subcloned into the pCOLD-I vector (Takara) and confirmed by sequencing. The N-terminal sequence preceding residues 112–223 of S1R was MNHKVHHHHHHIEGRHM. The S1R(cd) plasmid and a pMazF plasmid containing the gene for the RNA interferase MazF (Takara) were transformed into C43(DE3) cells. Transformed cells were grown to an OD of 0.8–0.9, cold shocked on ice, and incubated for 45 min at 15 °C. Cells were pelleted and washed with M9 salt solution, spun again, and resuspended into isotopically labeled medium with a 10-fold condensation. Cells were then incubated for a further 45–60 min at 15 °C before induction with 2 mm isopropyl 1-thio-β-d-galactopyranoside. Expression proceeded for 16 h at 15 °C. Both membranes and inclusion bodies were collected with a 40,000 rpm spin and incubated overnight in a solution containing 6 m guanidine, 200 mm NaCl, 1% Triton X-100, and 20 mm Tris, pH 8.0. S1R(cd) was separated by nickel affinity chromatography and dialyzed against water to remove guanidine. The precipitated protein was resolubilized in hexafluoro-2-propanol and purified by HPLC on a C3 reverse phase column over a gradient from buffer A (95% water, 5% acetonitrile, 0.1% trifluoroacetic acid) to buffer B (57% 2-propanol, 38% acetonitrile, 5% water, 0.1% trifluoroacetic acid). Fractions containing S1R(cd) were pooled and lyophilized. Yields of pure S1R(cd) from triple-labeled medium were between 8 and 10 mg per liter of labeled medium.</p><p>Polyhistidine-tagged human BiP (residues 26–654; "full length"), the isolated nucleotide binding domain (NBD; residues 26–410), and the isolated substrate binding domain (SBD; residues 418–638) of BiP (Herwig Schüler, Karolinska Institutet) were expressed in BL21(DE3) cells at 18 °C according to a previously published protocol (18). An ammonium sulfate precipitation was introduced as an initial purification step. The proteins were further purified by nickel chromatography, anion exchange chromatography using a Q-Sepharose column, and gel filtration using a Superdex200 column.</p><!><p>S1R(cd) was reconstituted with detergent micelles for NMR measurements from thin films of protein and DPC that were solubilized in hexafluoro-2-propanol and dried down under a nitrogen stream (19). The film was resolubilized in urea and reconstituted by dialysis against NMR buffer containing 20 mm HEPES, pH 6.5. Based on integration of peaks in 1H one-dimensional spectra calibrated against samples with known detergent concentrations, the final amount of DPC in S1R(cd) NMR samples was ∼30 mm. Where indicated, certain samples of S1R(cd) were run over gel filtration rather than dialyzed to reduce the DPC concentration to 5 mm. Final protein concentrations for all NMR samples were ∼250 μm.</p><!><p>All NMR experiments were collected at 37 °C at field strengths of 500, 600, 750, or 950 MHz (1H). The 750 and 950 MHz spectrometers were equipped with room temperature probes (home-built), and the 500 and 600 MHz spectrometers were equipped with cryogen-cooled probes (Bruker). NMR spectra were processed using NMRPipe (20) and analyzed using NMRView or CARA. Backbone 1HN, 15N, and 13Cα, and side chain 13Cβ assignments were obtained from triple resonance HNCA, CBCA(CO)NH, HNCACB, and HNCO spectra on 15N, 13C, and 2H-labeled S1R(cd). Assignments were confirmed and extended to sidechain protons with 15N-edited NOESY (90- or 180-ms mixing times at 950 or 600 MHz, respectively) and 15N-edited TOCSY spectra (40-ms mixing time; 600 MHz). 15N R1, R2, and 1H-15N heteronuclear NOE experiments were collected on a 0.27 mm 15N-labeled S1R(cd) sample at 600 MHz using the following relaxation delays: 4, 350, and 480 ms (T1) and 0, 40, and 60 ms (T2).</p><!><p>Mn2+EDDA2− was prepared as described (21). 1H,15N HSQCs were recorded before and after addition of 1 mm Mn2+EDDA2−. The accessibility to water was taken to be inversely proportional to the ratio of cross-peak intensity in the absence (I) and presence (Io) of Mn2+EDDA2−.</p><!><p>BiP titrations were carried out by preparing concentrated stocks of BiP constructs in 100 mm NaCl, 30 mm DPC, 5 mm 2-mercaptoethanol, and 20 mm HEPES, pH 6.5. Isotope-labeled samples of S1R(cd) in identical conditions were titrated with BiP, and 1H,15N SOFAST-HMQC spectra (22) were collected to monitor changes in cross-peak intensities and chemical shifts.</p><!><p>CD spectra were collected on a Jasco J-815 circular dichroism spectropolarimeter by collecting spectra from 190–250 nm with 10 accumulations (far UV) or 250–350 nm with 10 accumulations. Samples for CD contained 17.5 μm (far UV) and 240 μm (near UV) S1R(cd) in 5 mm DPC and 20 mm potassium phosphate at pH 6.5.</p><!><p>An S1R construct (S1R(cd)) containing the residues following the second putative transmembrane domain of S1R was produced with an N-terminal (His)6 tag using the single protein production approach (23) and purified to homogeneity (Fig. 2). Attempts to reconstitute S1R(cd) into aqueous solution in the absence of lipids or detergent resulted in protein aggregation. Therefore, S1R(cd) was reconstituted in the presence of DPC micelles. The resulting sample yielded homogenous NMR spectra permitting acquisition of high resolution data (Fig. 3).</p><!><p>SDS-PAGE stained with Coomassie Blue showing bacterial production of triple-labeled (2H, 13C, and 15N) S1R(cd) using the single protein production system. The S1R(cd) (theoretical mass of 14.7 kDa) band is indicated by an arrow. Lane 1, molecular weight standard; lane 2, whole cell lysate before induction; lane 3, whole cell lysate after induction in minimal medium for 16 h at 15 °C; lane 4, precipitated protein after elution from nickel column; and lane 5, HPLC-purified protein.</p><p>1H,15N HSQC spectrum (600 MHz, 1H) of S1R(cd) in 5 mm DPC at 37 °C. Backbone amide resonance assignments are indicated. The positions of the cross-peaks corresponding to residues Arg-114 and Gly-118, which are weak, are indicated with dashed line circles. The five cross-peak pairs expected from asparagine and glutamine H2N(ϵ2) groups of the S1R(cd) construct are indicated with horizontal lines. Inset, the four tryptophan HN(ϵ1) cross-peaks.</p><!><p>The overall secondary and tertiary structure of S1R(cd) was probed with CD. Far UV CD of S1R(cd) in 5 mm DPC exhibited minima at 208 nm and ∼220 nm (Fig. 4A). A lack of a well defined minimum at 222 nm suggested the presence also of a small amount of β-strand structure. Tertiary structure was assessed by CD at near UV wavelengths (Fig. 4B). A negative peak at 280 nm and a positive peak at ∼290 nm indicated the presence of tertiary structure. To test whether the tertiary structure could be disrupted at high temperature, a second spectrum was collected at 95 °C. At 95 °C, the 290-nm peak disappears and the 280-nm peak is decreased, indicating unfolding at high temperature.</p><!><p>Circular dichroism spectra of S1R(cd) in 5 mm DPC and 20 mm HEPES, pH 6.5, plotted as the mean residue molar ellipticity ([θ] MR). A, far UV spectrum of S1R(cd) (17.5 μm) at 37 °C. B, near UV spectra of S1R(cd) (240 μm) at 37 °C (solid line) and 95 °C (dashed line).</p><!><p>Conventional amide proton-based NMR experiments enabled backbone resonance assignment of 104 of the 107 nonproline residues of S1R(cd) (Fig. 3). Many of the amide proton NOESY strips for residues ∼140 to 160 contained cross-peaks at the water proton chemical shift indicating chemical exchange, and few medium range NOEs (i.e. dαN(i + 3)) (Fig. 5A). In contrast, NOE strips for residues ∼120–140 and ∼160–220 exhibited extensive short and medium range NOEs.</p><!><p>A, summary of NOE cross-peaks to backbone amide protons. Amide proton to amide proton NOEs (dNN(i+1)), and amide proton to α-proton NOEs (dαN(i+1), dαN(i+2), and dαN(i+3)), as well as cross-peaks at the position of the water proton (dN H2O), were measured in a three-dimensional 15N-edited NOESY with 180-ms mixing time at 600 MHz (1H) on an 15N-labeled S1R(cd) in 5 mm DPC. Amide proton to DPC methyl proton NOEs (dN DPC (CH3)) were obtained from a three-dimensional 15N-edited NOESY with 90 ms of mixing time at 950 MHz (1H) on an 15N- and 2H-labeled (∼70%) S1R(cd) in 30 mm DPC. The x axis (arbitrary units) is the cross-peak intensity normalized to that of the diagonal. No attempt was made to deconvolute overlapped cross-peaks. B, the chemical shift indices (CSI) of 1Hα, 13Cα, 13C′, and 13Cβ are shown in the top four panels. In panels five and six are shown the secondary structure and random coil index derived from chemical shift analysis using TALOS+ (49) and the method of Berjanskii et al. (50), respectively. For the TALOS+ plot, values above or below the line indicated that the calculated φ/ψ values correspond to helical (H) or extended conformation (E), respectively. Shown schematically at the top are the S1R(cd) residues determined to be helical (cylinders) or extended (arrow) (see main text for full description of secondary structure assignment).</p><!><p>Secondary chemical shift analysis was used to facilitate determination of S1R(cd) secondary structure (Fig. 5B). Chemical shift indexing of individual nuclei and TALOS+ analysis predicted helices at residues ∼121–137, ∼167–175, ∼180–189, and ∼193–219. Although chemical shifts predicted a continuous helix from residues ∼193–219, a large increase in the measured rate of exchange of the amide protons of residues 211–213 with water indicated disruption of the helical hydrogen bonding network and a break in this helix (see below). In addition, secondary chemical shifts predicted a φ/ψ angle for Ile-128 that corresponded to an extended conformation, suggesting that helix 1 may also be disrupted. However, there were no corresponding increases in the amide exchange rate or dynamics for this residue (see below).</p><p>Regions of extended structure could be determined with less confidence than helical regions, although an extended conformation is likely for residues ∼145–147 and ∼153–155. Chemical shift indices analysis and intense daN(i + 1) NOEs suggested the possibility of a third β-strand in residues ∼160–162, but the RCI indicated a high degree of flexibility in these residues.</p><!><p>Backbone amide dynamics of S1R(cd) were probed by measuring 15N transverse (R2) and longitudinal (R1) relaxation rates and 1H-15N heteronuclear NOEs (Fig. 6). Small 15N R2 values and negative or small heteronuclear NOEs were found throughout the region of residues 142–163, consistent with the low amount of structure in this region, although increased heteronuclear NOEs in residues 145–147 correlated with the chemical shift-based prediction that these residues are in a stable extended conformation. In addition, the relaxation properties of some of the interhelical regions at the C terminus exhibited decreased heteronuclear NOEs and 15N R2 relaxation rates, indicating increased flexibility, particularly in Ile-178 between helices 2 and 3, Val-190 between helices 3 and 4, and Arg-211 between helices 4 and 5.</p><p>R2/R1 ratios correlate with the effective rotational correlation time (24) and are plotted in Fig. 6D for S1R(cd). The average R2/R1 value for helical residues was 14.0 ± 2.2, which corresponds to a rotational correlation time (τc) of ∼11.6ns. This τc is nearly twice that expected for a 14.7-kDa protein at 37 °C (∼6.1 ns), suggesting that the protein is tightly associated with a detergent micelle. Although residues in helices 1, 2, 3, and 5 exhibited similar mean τc values (10.9–11.5 ns), residues in helix 4 had a mean τc value of 12.4 ns. Although effects from rotational anisotropy or contributions to the R2 from conformational exchange could not be ruled out, the higher τc value for helix 4 was consistent with this helix anchoring S1R(cd) to the detergent micelle (see below).</p><!><p>Relaxation properties of S1R(cd). A, 1H-15N heteronuclear NOEs (het-NOE); B, 15N transverse relaxation rates (15N R2); C, 15N longitudinal relaxation rates (15N R1); and D, ratio of 15N R2 and R1, as a function of residue number for S1R(cd) at 600 MHz (1H).</p><!><p>Based on observation of cross-peaks at the water proton resonance frequency in the NOESY, a large number of backbone amides within the region of residues 140–160 had apparent amide proton chemical exchange with water. Therefore, the rates of backbone amide exchange with water were measured using CLEANEX experiments (Fig. 7A) (25). In the region between helices 1 and 2, residues 137, 139–144, 149–151, and ∼155–161 had the largest exchange rates. Exchange rates in the predicted extended regions (residues 145–147 and 153–155) were low, suggesting the presence of β-sheet structure. Increases in exchange were also observed in the interhelical regions in the C-terminal half of S1R(cd) and helped to assign breaks in the helices.</p><!><p>Summary of water interactions with S1R(cd). A, backbone amide proton exchange of S1R(cd) with water measured using CLEANEX experiments (25). Water proton exchange is shown as the ratio of the peak volume in CLEANEX experiments using 10-ms (filled circles) or 30-ms (open squares) mixing times to the peak volume in a fast HSQC (600 MHz 1H). Measurement error has been estimated from the spectral noise. B, resonance broadening after addition of the water-soluble paramagnetic agent Mn2+EDDA2−. A decreased ratio of the peak intensity after (I) and before (Io) addition of Mn2+EDDA2− indicates increased water accessibility. Measurement error has been estimated from the spectral noise. Data for resonances that exhibit high water exchange rates as indicated by CLEANEX ratios of I/Io > 0.4 (dashed line in A) were excluded from the analysis due to potential artifacts in cross-peak intensities due to changes in relaxation properties.</p><!><p>Low amide proton exchange rates may be due to hydrogen bonding and/or inaccessibility to bulk water. Therefore, we measured line-broadening effects from the water-soluble paramagnetic agent Mn2+EDDA2−. Broadening of many resonances corresponding to the N-terminal half of S1R(cd) was observed, indicating few continuous regions of protection from water. In contrast, the two segments of residues ∼174–184 and ∼197–206 showed relatively high levels of protection (I/Io > ∼0.5), indicating interactions with the detergent micelle and/or other regions of the protein.</p><p>Regions of interactions of S1R(cd) with the detergent micelle were assessed from an 15N-edited NOESY (90 ms) recorded on an 15N-labeled and partially deuterated sample at high field (950 MHz) to resolve cross-peak overlaps and reduce spin diffusion (Fig. 5A). The largest clusters of residues with NOEs to detergent occurred in residues 183–189 in helix 3 and residues 197–204 in helix 4. Helix 4 was also strongly protected from the water-soluble paramagnetic agent. The region ∼198–206 forms an amphipathic helix, with residues Leu-199, Phe-200, Leu-203, and Tyr-206 forming the hydrophobic face (26).</p><!><p>Previous work by Hayashi et al. (1) showed that S1R residues 116–223 are sufficient for binding to the Hsp70 protein BiP. Therefore, the NMR spectrum of 15N-labeled S1R(cd) was monitored as a function of unlabeled BiP concentration (Fig. 8). The S1R(cd) cross-peak intensities decreased in a BiP concentration-dependent manner, suggesting that a BiP·S1R(cd) complex was formed. After several hours, a white precipitate formed, indicating that the complex was unstable in solution. By contrast, BiP alone in the titration buffer (containing 30 mm DPC) was more stable. It was speculated that binding of the S1R(cd)·micelle complex to BiP may increase the local concentration of DPC to destabilize BiP.</p><!><p>S1R(cd) interactions with BiP. A, backbone amide cross-peak intensities are plotted as a function of residue number after addition of 0.5 (black bars) and 1.0 (red bars) molar equivalents of full-length BiP (no calcium). The intensities are normalized to the cross-peak intensities in the absence of BiP. B, the normalized average backbone amide cross-peak intensity of S1R(cd) as a function of the molar ratio of BiP NBD (no calcium) or full-length BiP to S1R(cd) in the absence or presence of 2.5 mm calcium chloride. The intensity average is for only the helical regions. C, spectral overlays of 15N-labeled S1R(cd) with (red) or without (black) full-length BiP (1:1) in the presence of 2.5 mm calcium. D, spectral overlays of 15N-labeled S1R(cd) with (red) and without (black) addition of the BiP SBD (1:1) in the presence of 2.5 mm calcium.</p><!><p>The changes in 1H-15N cross-peak intensities upon addition of full-length BiP were plotted as a function of residue number at molar ratios of 0.5 and 1 ([BiP]/[S1R(cd)]) (Fig. 8A). Decreases in cross-peak intensities were higher for residues in the structured helical regions and smaller in the flexibly disordered regions at the N and C termini and residues ∼140–160, suggesting that these disordered residues retained some flexibility in the complex.</p><p>BiP chaperone (27) and ATPase (1) activity has been shown to be inhibited by calcium, and the interaction between BiP and S1R is proposed also to be modulated by calcium under physiological conditions, with the presence of calcium leading to increased association and inactivation of S1R and BiP (1). Therefore, a second titration of full-length BiP against S1R(cd) was carried out in the presence of 2.5 mm calcium chloride. A comparison of the cross-peak intensities for helical residues as a function of BiP indicates an increased association of S1R(cd) for full-length BiP in the presence of calcium (Fig. 8B). No chemical shift changes were observed upon addition of calcium to S1R(cd) alone, consistent with a calcium binding site on BiP (18).</p><p>To test whether the NBD of BiP was sufficient for S1R binding, the isolated domain was titrated against S1R(cd). Addition of the NBD resulted in similar changes in intensities as full-length BiP (Fig. 8B), indicating that the BiP NBD was sufficient for S1R interactions. The NBD domain was generally less stable than full-length BiP and more unstable in the presence of calcium, preventing evaluation of the effects of calcium on the NBD interaction with S1R(cd). In contrast, no changes in intensities or chemical shifts were observed after addition of the SBD (Fig. 8D).</p><!><p>Most membrane-associated proteins remain difficult to express at sufficient levels to enable inexpensive isotope labeling. We have found that the chaperone domain of human S1R could be efficiently expressed in Escherichia coli using the single protein production approach (23). In single protein production, the cells remain metabolically active and express the target protein, but cell growth is halted, which appears to greatly diminish expression toxicity. The approach permitted 10-fold condensation of the cultures when transferring the cells into labeled medium and resulted in yields of 8–10 mg of pure S1R(cd) per liter of labeled medium.</p><p>The S1R chaperone domain is predicted to contain a C-terminal membrane-associated region (residues ∼180–203). Consistent with this prediction, solubilization of S1R(cd) required the presence of lipid or detergent, suggesting that membrane interactions may be required also for proper folding of S1R(cd) in vivo.</p><p>For the NMR studies reported here, S1R(cd) was reconstituted from 8 m urea into DPC, which contains a phosphocholine headgroup and has been used extensively for studying folded integral membrane and membrane-associated proteins (28–31). The S1R(cd) samples permitted measurement of high quality NMR spectra, and near UV CD indicated the presence of temperature-sensitive tertiary structure. The finding that the S1R(cd) construct purified from E. coli and reconstituted with DPC micelles interacted with BiP in a calcium-dependent manner indicated that S1R(cd) adopted a native-like conformation in the NMR conditions, at least in respect to those elements required for the BiP interaction.</p><p>The N-terminal half of S1R(cd) contains a ∼17-residue helix, and at least two short β-strands. The region containing the extended regions, residues ∼140–160, is highly dynamic, based on secondary chemical shifts, relaxation measurements, and amide proton exchange rates.</p><p>The putative membrane associated C-terminal half of S1R(cd) is largely helical, with four helices separated by short, flexible, water-exposed linkers. By combining information from NOEs to detergent and resonance broadening from a water-soluble paramagnetic agent, a stretch of ∼9 amino acids (residues 198–206) within helix 4 was identified as a site for strong interactions with the detergent micelle. Although helix 4 is ∼18 residues long, it is not highly hydrophobic (containing two arginines and an aspartic acid), and residues 196, 197 and 207 have detectable rates of amide proton exchange with water. This suggests that it is not entirely embedded within the detergent micelle as could be expected for a transmembrane helix. Instead, residues 198–206 form an amphipathic helix, which likely interacts with the surface of the detergent micelle, and is therefore also a likely site for attachment to the ER membrane.</p><p>A cluster of NOEs to detergent were also observed in residues ∼183–189, suggesting that this region may also facilitate micelle or membrane binding. However, these residues are also largely water-accessible and are proposed to have long-range intramolecular interactions in the intact receptor (35). Thus, it appears unlikely that the chaperone domain contains a helix capable of spanning the ER membrane. The lack of a transmembrane helix in the chaperone domain is consistent with the finding that this domain can be extracted from cells with chaotropic salt washes (14).</p><p>The chaperone domain contains several regions implicated in cholesterol and drug interactions (Fig. 9). Based on a cholesterol recognition consensus sequence (32, 33) (L/V)X1–5YX1–5(R/K) two cholesterol recognition motifs have been identified within the structured region of S1R(cd) (17, 34, 35). The first motif (CRM1; residues 171–175; amino acid sequence, VEYGR) corresponds approximately to helix 2. The second motif (CRM2; residues 199–208; amino acid sequence, LFYTLRSYAR) contains two overlapping cholesterol recognition motifs (Leu-199/Tyr-201/Arg-204 and Leu-203/Tyr-206/Arg-208), and corresponds to the central region of helix 4. Residues in the vicinity of both CRM1 and CRM2 exhibited heightened protection from the water-soluble paramagnetic agent, although a larger number of NOEs to DPC were detected in the CRM2, suggesting a more intimate interaction with the detergent micelle at this site.</p><!><p>The helical (cylinders) and extended (arrow) residues of S1R(cd) determined from a combination of chemical shifts, amide-water proton exchange, 15N relaxation rates, and NOEs. Residues and regions previously implicated in cholesterol binding are indicated: cholesterol binding motifs (CRM1 and CRM2), cocaine binding (SBDLII; residues Asp-188 and Val-190 are indicated by ↓), and haloperidol binding (residues Asp-126 and Glu-172 are indicated by an asterisk). Residues 198–206, which are proposed to have the strongest interactions with the ER membrane, are indicated. The amino acid sequence and corresponding residue numbers are shown at the bottom.</p><!><p>Although the chaperone domain does not bind drugs in the absence of the N-terminal transmembrane domain (1, 35), the secondary structure determined here may reflect the structural propensities of the intact receptor. Studies using a chemically reactive affinity probe have provided information on the binding site of cocaine in guinea pig S1R, which is 98% similar to human S1R (35–37). Those studies defined a steroid binding-like domain (SBDLII) at residues 176–194 and located residues Asp-188 and Val-190 close to the cocaine interaction site. SBDLII together with a steroid binding-like domain (denoted SBDLI; residues 91–109) in the second putative transmembrane helix stabilizes cocaine binding (35). SBDLII corresponds closely to helix 3 (residues 180–189) and the adjacent residues connecting helix 3 to helices 2 and 4 (Fig. 9). Asp-188 and Val-190, which are proposed to be close to the cocaine binding site, are at the C-terminal end of helix 3. Both Asp-188 and Val-190 are solvent exposed in S1R(cd), and Val-190 is one of the most flexible residues within the C-terminal helical region of S1R(cd). It is unknown whether the flexibility observed here is preserved in full-length S1R, but such flexibility may facilitate binding-induced conformational changes necessary to mediate S1R downstream interactions.</p><p>Mutational studies of S1R have also implicated Asp-126 and Glu-172 in haloperidol binding (8). Asp-126 and Glu-172 are found in helices 1 and 2 of S1R(cd), respectively, suggesting that these helices may interact in the haloperidol bound conformation of the receptor.</p><p>BiP is an ER resident chaperone that regulates the unfolded protein response in addition to assisting protein folding (reviewed in Ref. 38). S1R is proposed to sequester BiP in the absence of ER stress. Upon depletion of ER calcium or S1R ligand binding the S1R·BiP complex dissociates leading to chaperone activity and downstream signaling of S1R, including inositol triphosphate receptor-mediated calcium release (1, 3). We have shown here that S1R(cd) under NMR conditions interacts with BiP in a calcium-dependent manner and that the NBD of BiP is sufficient for these interactions. The finding that S1R(cd) interacts with the NBD of BiP is consistent with the known regulatory interactions with other Hsp proteins. For example, the co-chaperone BAG1 binds to the NBD of Hsc70 (42, 43), and bacterial GrpE binds to the NBD of DnaK (44, 45).</p><p>The role of the region ∼140–160 remains unknown. This region did not appear to tightly associate with BiP in our studies, and substitution of acidic residues within this region has been shown to have no impact on haloperidol binding (8). Based on sequence analysis (39), several short β-strands are predicted for this region in residues 143–145, 151–153, and 159–164, which corresponds approximately to those regions observed to be extended here (residues 145–147, 153–155; secondary chemical shifts also indicate transient extended conformation in residues 160–162). However, sequence analysis also predicts a high degree of disorder in this region (40), which is confirmed by the NMR data. By contrast, residues 124–137 and 168–173 are predicted from sequence to be extended but found here to be helical. DPC has previously been shown to disrupt β-sheets in a concentration-dependent manner (41). However, the concentrations used here are as much as 10-fold lower than in that study, and the spectra of S1R(cd) in 5 and 30 mm DPC are essentially identical.</p><p>In summary, we have studied the chaperone domain of S1R by solution NMR and have characterized its secondary structure and dynamics. S1R(cd) is composed of five helices and at least two short extended regions in a dynamic region between helices 1 and 2. Three of the helices in the C-terminal membrane-associated region map to residues previously identified as important in cholesterol and cocaine binding. A fourth helix (helix 4) is implicated in membrane association. In addition, we have shown that the NBD domain of BiP is sufficient for interaction with the S1R chaperone domain. These results advance our understanding of S1R and are likely to be useful in refining models of the S1R drug binding sites (46). Future studies are needed to identify tertiary interactions in the S1R chaperone domain and to further delineate S1R/BiP and S1R ligand and cholesterol interactions.</p><!><p>This work was supported by Medical Research Council Grant K018590 and the University of Oxford Department of Biochemistry.</p><p>sigma-1 receptor</p><p>sigma-1 receptor chaperone domain</p><p>endoplasmic reticulum</p><p>binding immunoglobulin protein</p><p>dodecylphosphocholine</p><p>nucleotide binding domain</p><p>substrate binding domain</p><p>steroid binding-like domain</p><p>cholesterol recognition motif.</p>

PubMed Open Access

Non-adiabatic Matsubara Dynamics and Non-adiabatic Ring Polymer Molecular Dynamics

We present the non-adiabatic Matsubara dynamics, a general framework for computing the time-correlation function (TCF) of electronically non-adiabatic systems. This new formalism is derived based on the generalized Kubo-transformed time-correlation function, using the Wigner representation for both the nuclear degrees of freedom (DOF) and the electronic mapping variables. By dropping the non-Matsubara nuclear normal modes in the quantum Liouvillian and explicitly integrate these modes out of the TCF, we derived the non-adiabatic Matsubara dynamics approach. Further making the approximation to drop the imaginary part of the Matsubara Liouvillian and enforce the nuclear momentum integral to be real, we arrived at the non-adiabatic ring-polymer molecular dynamics (NRPMD) approach. We have further justified the capability of NRPMD for simulating the non-equilibrium time-correlation function. This work provides the rigorous theoretical foundation for several recently proposed state-dependent RPMD approaches and offers a general framework for developing new non-adiabatic quantum dynamics approaches in the future.

non-adiabatic_matsubara_dynamics_and_non-adiabatic_ring_polymer_molecular_dynamics

I. INTRODUCTION<!>II. GENERALIZED KUBO-TRANSFORMED TIME-CORRELATION FUNCTION<!>A. Mapping Representation of Electronic States<!>B. Generalized Kubo-Transformed Time-Correlation Functions<!>C. The Quantum Liouvillian<!>D. Time-Correlation Function<!>III. NORMAL MODE REPRESENTATION<!>A. Definition of the Normal Modes<!>B. Time Correlation Function Under the Normal Mode Representation<!>B. Matsubara Approximation<!>C. Matsubara Time-Correlation Function<!>D. Exact Quantum Statistics with Matsubara Modes<!>V. NON-ADIABATIC RING POLYMER MOLECULAR DYNAMICS<!>VI. QUANTUM DETAILED BALANCE<!>VII. TIME-CORRELATION FUNCTIONS WITH ELECTRONIC PROJECTION OPERATORS<!>VIII. NON-EQUILIBRIUM TIME-CORRELATION FUNCTION<!>IX. CONCLUSION AND FUTURE DIRECTIONS<!>X. ACKNOWLEDGEMENT<!>XI. AVAILABILITY OF DATA<!>XII. APPENDIX A: DERIVATION OF THE LIOUVILLIAN<!>XIII. APPENDIX B: QUANTUM BOLTZMANN DISTRIBUTION IN THE NON-ADIABATIC MATSUBARA DYNAMICS<!>XIV. APPENDIX C: DETAILED BALANCE FOR SYSTEM WITH DECOUPLED ELECTRONIC AND NUCLEAR DOF<!>XV. APPENDIX D: CONNECTIONS TO LINEARIZED PATH-INTEGRAL APPROACHES<!>XVI. APPENDIX E: NUMERICAL TEST OF THE NON-ADIABATIC RPMD

<p>Accurately simulating the quantum dynamics of the molecular system remains a central challenge in theoretical chemistry, due to the difficulties of accurately describing electronically non-adiabatic dynamics and nuclear quantum dynamics. Directly performing exact quantum dynamics simulations is computationally demanding, despite exciting recent progress. [1][2][3][4][5][6][7][8][9][10][11] To accurately describe the non-adiabatic dynamics, a large number of these approaches are developed, including the popular trajectory surface-hopping method (mixed quantum-classical approach), [12][13][14][15] the linearized path-integral approaches, [16][17][18][19][20][21][22][23] and the mixed quantumclassical Liouville equation, [24][25][26][27][28] Despite providing accurate electronic non-adiabatic dynamics, these approaches often relies on the Wigner sampling of the initial nuclear distribution and a classical dynamics for propagation. Thus in generally, they do not preserve quantum Boltzmann distribution (QBD) 29,30 or zero point energy (ZPE) associated with the nuclear degrees of freedom (DOF). They often suffer from numerical issues such as ZPE leakage, 31,32 although significant improvements are accomplished through the recent development. [33][34][35][36] To accurately describe nuclear quantum dynamics for electronically adiabatic systems, imaginary-time path integral based approaches [37][38][39] are developed, including the centroid molecular dynamics (CMD) [40][41][42] and the ringpolymer molecular dynamics (RPMD). 43,44 In particular, RPMD resembles the classical MD in the extended phase space, thus provides a convenient way to compute approximate quantum time-correlation functions. 43 The classical evolution of RPMD preserves its initial quana) Electronic mail: schowdh4@ur.rochester.edu b) Electronic mail: pengfei.huo@rochester.edu tum distribution captured by the ring-polymer Hamiltonian, and it is free of the zero-point energy leaking problem. 31,43 Despite its success on describing quantum effects in the condensed phase, RPMD is limited to one-electron non-adiabatic dynamics [45][46][47][48][49] or nuclear quantization, 43,[50][51][52][53] as well as the lack of real-time electronic coherence effects. 45,46 Recently emerged state-dependent RPMD approaches provide a unified description of both the electronically non-adiabatic dynamics and nuclear quantum effects. These state-dependent RPMD methods, such as non-adiabatic RPMD (NRPMD), [54][55][56] mapping variable RPMD (MV-RPMD), [57][58][59] ring-polymer Ehrenfest dynamics, 60 , kinetically-constrained RPMD (KC-RPMD), 48,61,62 coherent state RPMD (CS-RPMD), 63 and ring-polymer surface hopping (RPSH) [64][65][66][67][68] are promising to provide both accurate non-adiabatic dynamics with an explicit description of electronic states, as well as a reliable treatment of nuclear quantum dynamics through the ring polymer path-integral quantization.</p><p>Despite the initial success, all of the above statedependent RPMD approaches are currently viewed as the model dynamics. The Hamiltonians associated with some these approaches (such as MV-RPMD) are derived from quantum partition function, and these Hamiltonians are directly used for dynamics propagation as well as the initial sampling. Thus the fundamental and crucial theoretical question is that can these methods be rigorously justified. If so, not only it will explain the numerical success of these state-dependent RPMD approaches, but also it will offer a general theoretical framework to understand the limitations of these state-dependent RPMD approaches and further improving them. Recent theoretical work on the Matsubara dynamics [69][70][71][72] by Althrope and co-workers indeed provides hope for this, because RPMD (as well as CMD) can be derived as an approximation of the Matsubara dynamics, 70,73 which itself can be rigorously derived. 69 In this paper, we present the non-adiabatic Matsubara dynamics, a general framework for computing the time-correlation function of electronically non-adiabatic systems. This new formalism is derived based on the generalized Kubo-transformed time-correlation function (TCF) formalism 42,69 , using the Wigner representation for both the nuclear DOF and electronic mapping variables. [74][75][76] By dropping the non-Matsubara nuclear normal modes in the quantum Liouvillian, 69 we derived the non-adiabatic Matsubara dynamics, which can be viewed as a generalization of the original (electronically adiabatic) Matsubara dynamics. 69 Further making the approximation that drop the imaginary part of the Matsubara Liouvillian, 73 we arrived at the non-adiabatic RPMD (NRPMD) approach, where the initial distribution coincides with the one in Mapping-Variable (MV)-RPMD 57 , whereas the Liouvillian coincides with the Liouvillian used in the originally proposed NRPMD 54 . We have further justified the capability of NRPMD for simulating the non-equilibrium time correlation function.</p><!><p>We begin by introducing the Generalized Kubo Transformed time-correlation functions for a state-dependent Hamiltonian. We start by expressing the total Hamiltonian operator as follows</p><p>where {|i } is the diabatic basis, T is the nuclear kinetic energy operator, R is the nuclear position operator with the corresponding conjugate momentum operator P . To simplify our discussion, we have assumed that there is only one nuclear DOF in the system. Generalizing the discussion with many nuclear DOF is straightforward. Further, V 0 ( R) is the state-independent potential operator, whereas Ve =</p><p>is the state-dependent potential operator (electronic part of the Hamiltonian) with K total diabatic electronic states. We assume that V ij ( R) is real and symmetric through out this work.</p><!><p>In the electronic part of the Hamiltonian, the K diabatic electronic states can be mapped into K harmonic oscillators' ground and excited states, where (K-1) oscillator in ground state and i th oscillator in the first excited state. It can be formally written as |i → |0 1 , ..., 1 i ..., 0 K = â † i |0 1 , ..., 0 i ..., 0 K ,</p><p>where â † i = 1 √ 2 (q i − ip i ) and âj = 1 √ 2 (q j + ip j ), and q = {q 1 , ...q i , ...q K } and p = {p 1 , ...p i , ...p K } are the mapping position and momentum operator respectively. This mapping formalism is often referred to as the Meyer-Miller-Stock-Thoss (MMST) [74][75][76] mapping representation. With this mapping formalism, the electronic part of the Hamiltonian is expressed as</p><p>Using the above relationships, the electronic part of the Hamiltonian in Eq. 1 can be represented as Ve = 1 2</p><p>This is known as the MMST [74][75][76] mapping Hamiltonian.</p><!><p>We begin by writing the generalized Kubo-transformed time-correlation function (TCF). The conventional Kubo transformed correlation function is defined as</p><p>Tr[e −β N (N −k) Ĥ Âe −β N k Ĥ e i Ĥt Be − i Ĥt ],</p><p>where β = 1/k B T is the inverse temperature, Z = Tr[e −β Ĥ ] is the canonical partition function, Ĥ is defined in Eq. 1, and Tr = Tr n Tr e represents the trace over both electronic and nuclear DOFs. From the first to the second expression, we have converted a definite integral into the discrete Riemann sum through b a f (λ)dλ = lim N →∞ N k=1 f (a + k • ∆λ) • ∆λ, where a = 0, b = β, and ∆λ = β/N = β N . Note that the second line of Eq. 5 is equivalent to the first line under the limit of N → ∞.</p><p>We further insert N − 1 identities of the form e i Ĥt/ e −i Ĥt/ = 1 in Eq. 5 (see details in the Supporting Information), resulting in</p><p>Tr (e −β N Ĥ e</p><p>i Ĥt e − i Ĥt ) N −k−1 × e −β N Ĥ Âe i Ĥt/ e −i Ĥt/ (e −β N Ĥ e i Ĥt/ e −i Ĥt/ ) k−1 × e −β N Ĥ e i Ĥt/ Be −i Ĥt/ ,</p><p>where β N = β/N . Note that Eq. 6 has a symmetric block structure of the form e −β N Ĥ e i Ĥt/ e −i Ĥt/ , where the operator  is evaluated inside a particular block depending on the value of the k index. This type of generalized Kubo-Transformed time-correlation function was first introduced in the work to connect Linearized Path-integral approach and CMD, 42 and later used for the development of the Matsubara dynamics. [69][70][71][72] A path integral representation of Eq. 6 can be obtained by inserting the following identities 1R l ,q l = dR l dq l P|R l , q l R l , q l | (7)</p><p>1R l ,q l = dR l dq l |R l , q l R l , q l | P, (8) where the bead index l = 1...N , {R l , R l } and {q l , q l } are the nuclear and mapping variable position, respectively, with q l = {[q l ] 1 , ...[q l ] i , ...[q l ] K } and similarly for q l for the corresponding momenta. Further, the electronic projection operator is</p><p>which helps to confine the mapping DOFs within the correct SEO subspace. 77 Inserting Eq. 7-8 into the bead-specific positions of Eq. 6, followed by replacing the full trace over the nuclear and electronic DOFs with the phase space integrals, 78 one arrives at the formal mathematical description of Generalized Kubo correlation function 69,78 as follows</p><p>where we introduce the notation dx = N l=1 dx l for x = {R , R , q , q }, and we have used the cyclicsymmetric property and write operator B also into a bead-averaged fashion. A detail derivation of Eq. 10 is provided in the supporting information.</p><p>To proceed, we change the variables (R l , R l , q l , q l ) into the mean (R l and q l ) and difference coordinates (D l and ∆ l ) as follows 42,78</p><p>Noting that the Jacobian of the above transformation is unity for each bead index l. With this transformation, one can re-express the Eq. 10 as</p><p>Next, inserting the following identities</p><p>into Eq. 13 for all l blocks, we have</p><p>with the constant α N = 1/(2π ) (k+1)N , and the operator [e −β Ĥ Â] N is expressed as</p><p>and [ B(t)] N is expressed as follows</p><p>where we have changed the dummy variable from k to k . Integrals over R, P, D are N dimensional, whereas integrals over q, p, ∆ are (N ×K) dimensional. Note that [e −β Ĥ Â] N in Eq. 15 contains a complex structure of the Wigner transform which couples with adjacent beads l and l + 1, whereas [ B(t)] N in Eq. 16 is a simple bead average of the Wigner transform.</p><p>We want to remind the reader that C</p><p>[N ] AB (t) in Eq. 14 should be viewed as a generalized Kubo-Transformed time-correlation function, such that under the N → ∞ limit, it returns to the original definition of the Kubotransformed time correlation function C K AB (t) defined in Eq. 5. With a finite N , even though it is no longer equivalent to C K AB (t), it is still an quantum mechanically exact time-correlation function.</p><!><p>For a Wigner transform of an operator</p><p>where L is the quantum Liouvillian (see Appendix A for detail derivation), which is also commonly referred to as the Wigner-Moyal series. 79,80 Note that [ B(t)] N (Eq. 16) is expressed as the beadaveraged (K+1)-dimensional Wigner transform as follows</p><p>where B(t) = e i Ĥt Be − i Ĥt and [ Bk (t)] W is the Wigner transform of B(t) associated with the k th bead, defined in the above equation. In Eq. 17 we have also changed the dummy variables from D and ∆ to D and ∆. When</p><p>The Quantum Liouvillian L [N ] has the following form</p><p>where Λl is the negative Poisson operator associated with the l th bead expressed as</p><p>with the mapping variables related derivatives defined as</p><p>and ∇ ∇ ∇ T q = ( ∂ ∂q1 , ... ∂ ∂q K ), and likewise for ∇ ∇ ∇ p . Further, [ Ĥl ] W in Eq. 19 is the Wigner transform of MMST mapping Hamiltonian (Eq. 1) associated with the l th bead, which can be shown 25,56,81 as</p><p>The detailed the proof of the above Wigner transform is provided in the Supporting Information. In the above equation, V 0 (R l ) is the state-independent potential evaluated at the l th nuclear bead position R l , and V e (R l , q l , p l ) is the state dependent potential that parametrically depends on R l , q l = ([q l ] 1 , [q l ] 2 , ...[q l ] K ), and</p><p>The above expression is the classical limit of Eq. 4 because the mapping variables are no longer operators. Now using Eq. 19 with the detailed expressions of Λ l (Eq. 20) and [ Ĥl ] W (Eq. 22), the full Liouvillian 78 can be explicitly expressed as follows</p><p>The above Liouvillian was first derived in Ref. [ 78], and the detail of the derivation is provided in the Appendix A and the Supporting Information. Here, V 0 (R l ) and V e (R l , q l , p l ) are defined in [H l ] W (Eq. 22), ∇ ∇ ∇ q l and ∇ ∇ ∇ p l are the gradient operators corresponding to the l th mapping bead's position and momentum, respectively, as defined in Eq. 21. Further, V V V(R l ) is the (K × K) statedependent potential energy matrix, parametrically depends on l th nuclear bead's position as follow</p><p>This generalized Liouvillian L [N ] govern the time evolution of N individual replicas connected through the zero-time quantum Boltzmann distribution. The exact Liouvillian L [N ] in Eq. 24 has three components.</p><p>The first term, denotes as</p><p>∂P l , corresponds to an Ehrenfest-type evolution of the nuclear DOF, 82 with higher-order nuclear derivatives (inside the sin function). The Second term in Eq. 24, denoted as L</p><p>∂P l , describes the electronic evolution with higher-order coupling terms to nuclear motions.</p><p>The third term in Eq. 24, denoted as L</p><p>contains coupled higherorder derivatives of nuclear and electronic motion. 25,81,83 Note that there is no cross-bead interaction terms between electronic DOF and nuclear DOF in Eq. 24. Each individual bead term can only interacts with each other via V(R l ) (Eq. 25). If the system evolves only in a single surface (electronically adiabatic regime) or if V e (R l , q l , p l ) = 0, then Eq. 24 reduce down to conventional Wigner-Moyal series [78][79][80] for N independent replicas as follows</p><p>On the other hand, if there is no nuclear DOF, then the Liouvillian (Eq. 24) becomes</p><p>It has been rigorously proved that the above Liouvillian preserves the electronic Rabi oscillation. 78</p><!><p>With the detail expression of Liouvillian in Eq. 24 one can formally rewrite the Eq. 14 as</p><p>where L [N ] is the Liouvillian in Eq. 24, [e −β Ĥ Â] N is expressed in Eq. 15, and [ B(0)] N = 1 N k [ Bk ] W as shown in Eq. 17. Up to this point, there is no approximation in the expression of C</p><p>[N ] AB (t) in Eq. 28. We can further write [e −β Ĥ Â] N (Eq. 15) into the sym-metric form as</p><p>Using the property of the Wigner transform, 79,84</p><p>Ô2 ] W , with Λ defined in Eq. 20, we can rewrite Eq. 29 as</p><p>where</p><p>and Λl is defined previously in Eq. 20. The multidimentional Wigner transformed Boltzmann operator [e −β Ĥ ] N is expressed as</p><p>The detail derivation of Eq. 30 is provided in the supporting information.</p><p>To simplify our derivation, we further explore various special cases that can further simplify this equation. (i) If operator  is linear in nuclear position or momentum (  = R or  = P ), or (ii) if operator  and the Boltzmann operator [e −β Ĥ ] N act on different DOFs (eg,  contains electronic state (  ∈ P = n |n n|) and in [e −β Ĥ0 ] N , Ĥ0 only contains nuclear DOF), then for both (i) and (ii), only the first term in the cosine expansion will survive and hence we can write [e</p><p>The detailed derivations for both case (i) and (ii) are provided in the Supporting Information.</p><p>From this point and below we assume that both operator  and B are linear function of the nuclear position (R), such that we do not have to worry about the cos( Λl /2) operator in Eq.30. With this simplification, we can write the Eq. 28 as</p><!><p>Here, we briefly introduced the normal mode coordinates of the free ring-polymer. The advantage of using normal modes is that with this set of global coordinates one can conveniently describe the collective motion of individual beads.</p><!><p>The free ring polymer Hamiltonian is defined as</p><p>where there is no external potential. Normal modes are defined as the eigenvectors of the Hessian matrix of H rp . Diagonalizing the Hessian matrix of H rp , one obtain the eigenvalues ω 2 n , which are the square of following normal mode frequency</p><p>where n = 0,...,±(N − 1)/2 is the index of normal modes.</p><p>The same diagonalization process also gives the eigenvectors T ln of the Hessian matrix, which provides the relation between the primitive variables R l , P l and the normal mode variable Q n , P n as follows</p><p>Similar transformation should also apply to {D n } and {D l }. The extra √ N factor in Eq. 36 ensures that the Q Q Q = {Q n } converges in the limit of N → ∞. For an odd N (to simplify our algebra), the transformation matrices are</p><p>Under the normal mode representation, the free ring polymer Hamiltonian (Eq. 34) becomes</p><p>Note that the ring polymer spring terms are now the simplified uncoupled quadratic potentials, hence the normal modes of the free ring polymer can be evolved analytically by simple harmonic motion. 85</p><!><p>It is straightforward to transform Eq. 33 into the normal mode coordinates by using the orthonormal transformations defined in Eq. 36, leading to</p><p>where we used the shorthand notation</p><p>n=−(N −1)/2 dQ n and likewise for dP P P. Here, both the dq and dp are defined as before in beads representation, such as dq = N l=1 dq l and likewise for dp. Only nuclear coordinates R and P are transformed to their corresponding normal mode coordinates Q Q Q and P P P, respectively.</p><p>Using new coordinates ξ ± l defined as follows</p><p>Providing the operators  = B = R, [ Â] N can be expressed in normal mode coordinates as</p><p>and same for [ B] N . In the last line of Eq. 42, we have substitute R k with the corresponding normal mode transformation defined in Eq. 36.</p><p>Finally, we decompose the total Liouvillian L [N ] (Eq. 24) into the following terms</p><p>and express each term with the normal mode coordinates</p><p>Here, normal mode representation of</p><p>Note that in Eq. 50, the We now considering M lowest frequency normal modes for M N , under the limit N → ∞. They are commonly referred to as the Matsubara modes 38,69 of distinguishable particles, with the corresponding Matsubara frequency ωn as follows ωn = lim</p><p>The superposition of these M Matsubara modes produces a smooth and differentiable function 69 in imaginary time τ , such that R l = R(τ ) for τ = β N l with l = 1, ..., N . This means that one can construct the smooth imaginary-time path R(τ ) from the Matsubara modes 69,[86][87][88] as follows</p><p>and the same relations for P (τ ) and D(τ ). The significance of Matsubara modes is that their superposition generate smooth and differentiable functions of the imaginary-time τ . On the other hand, if one consider both the Matsubara (M ) and non-Matsubara(N -M ) modes, then the imaginary-time path is not necessarily differentiable, because the latter give rise to nonsmooth, non-differentiable distribution with respect to τ . It is a well-known fact that the Boltzmann operator guarantees that only the Matsubara modes contribute to the initial Quantum Boltzmann Distribution (QBD) 38,[89][90][91] (for electronically adiabatic systems). Moreover, the previous work on Matsubara dynamics 69,73 have suggest that there is a close connection between the smoothness in imaginary-time and the dynamics that preserves the QBD.</p><p>Similar to the previous work of the Matsubara dynamics, 69 here, we show that (in Section IV. D) one can integrate out all of the non-Matsubara modes from Eq. 39, giving rise to the exact initial quantum statistics (which corresponds to the generalized Kubo transformed correlation function at t=0) in the limit M → ∞, M N 69,[89][90][91] . This suggest that only the smooth and imaginary-time differentiable Matsubara modes contributes to the initial quantum statistics for the electronically non-adiabatic systems as well. This makes one wondering how important are those non-Matsubara modes in quantum dynamics (Liouvillian), since they do not contribute to the initial QBD at all. 69 We will explicitly discuss such an approximation that drops the non-Matsubara modes in Liouvillian 69 in the next section.</p><!><p>We separate each term in Eq. 43 into two parts, one contains the lowest M Matsubara modes, and the other contains (N − M ) non-Matsubara modes</p><p>where the Matsubara Liouvillian is L [M ] , and the non-Matsubara Liouvillian L ] does not contain any derivatives with respect to the mapping variables, hence there is no direct influence from L [N −M ] to the electronic subsystem. The non-Matsubara Liouvillian L [N −M ] , on the other hand, does couple the non-Matsubara modes with the Matsubara mode, where the Matsubara mode couple to the mapping DOFs through the Matsubara Liouvillian L [M ] . Since the non-Matsubara modes do not contribute to the initial quantum statistics, it might also be a good approximation (at least in short time) to ignore their presence in the quantum Liouvillian. 69 The Matsubara dynamics assumes that the time evolution of nuclei are only governed by the smooth Matsubara modes instead of all normal modes. This approximation is achieved by neglecting the non-Matsubara modes in the derivatives of the corresponding Liouvilian terms</p><p>), which effectively produces the decoupling between non-Matsubara modes from the Matsubara modes during the time evaluation. Under the limit N → ∞, one discards the (N − M ) higher frequency, non-smooth modes. 69,71,73 Similar to L [N ] in Eq. 43, we further decompose L [M ] into the following three terms</p><p>where the detailed expressions for each term are</p><p>In Eq. 55 to Eq. 57, we re-express</p><p>and L</p><p>[N ] h</p><p>in Matsubara modes under the N → ∞ and M N limit. It is worth mentioning that the full Matsubara Liouvillian in Eq. 54 contains potential terms U</p><p>, which still depend on all N normal modes. On the other hand, all derivatives only involve M Matsubara modes.</p><!><p>Explicitly applying the Matsubara approximation for the Liouvillian L</p><p>), the exact correlation function in Eq. 39 becomes an approximate correlation function</p><p>AB (0) such that the initial QBD is exactly captured (see Section D). This approximate TCF, which is commonly referred to as the Matsubara TCF, is expressed as follows</p><p>Note that the above expression still depends on the non-Matsubara modes through the potentials in L [M ] and the QBD term [e −β Ĥ ] N , and the integral dQ Q Q and dP P P still include all normal modes (Matsubara and non-Matsubara). However, as non-Matsubara modes are decoupled from the Matsubara modes (because we have dropped L [N −M ] in the Liouvillian), one can analytically integrated out all of the non-Matsubara modes under the limit of N → ∞, M → ∞, and M N . The detailed derivation of this procedure is provided in Appendix C.</p><p>After integrating out the non-Matsubara modes in Eq. 58, we reach to the first key result of this paper as follows</p><p>where the shorthand notations for the integrals are</p><p>dq l , and analogously for dP P P M and dp, and α M is the following constant</p><p>Note that the above correlation function C</p><p>[M ] AB (t) is explicitly depends on the Matsubara modes Q Q Q M and P P P M . The mapping DOFs, on the other hand, are still expressed in the primitive (bead) coordinates with all N copies, as we did not make any approximation on them. The Liouvillian L [M ] has the same expression in Eq. 55-Eq. 57 but with following substitutions U</p><p>), whereas these new potential only contains the Matsubara modes, for example</p><p>and similarly for U</p><p>) by replacing the sum in Eq. 48 and Eq. 49 from originally over all modes to the sum over only the Matsubara modes. Note that the nuclear coordinate</p><p>the former one only contains smooth (and imaginary-time differentiable) Matsubara mode, and later one contains all modes.</p><p>Further, H M (P</p><p>and the Matsubara phase θ M takes the following form</p><p>where ωn = 2nπ/β is the Matsubara frequency (Eq. 51).</p><p>The Γ(Q Q Q M , q, p) term in Eq. 59 corresponds to the QBD originated from the electronic-nuclear interaction, which is expressed as follows</p><p>and I I I is the (K×K) identity matrix. Note that (C l − 1 2 I I I) can be interpreted as the reduced density matrix associated with the l th bead. In addition,</p><p>is the matrix element of the electronic Boltzmann operator expressed as follows</p><p>where</p><p>The expression of Γ was derived in the MV-RPMD partition function expression. 57 Finally, the partition function is expressed as</p><p>Note that under the Matsubara limit N → ∞, M → ∞, and M N , one can further Taylor expand the sine and cosine terms in Eq. 55 to Eq. 57 as follow lim</p><p>From the above analysis, it is clear that in the Matsubara space, the "effective" Planck constant inside the cos term is re-scaled as</p><p>and as</p><p>for the sin term. 70 Thus, it can be made as small as desired by increasing N . Hence, truncating Eq. 55 to 57 to the first order of</p><p>∂Pn in the Matsubara space becomes exact. Further, the L</p><p>[M ] h term (Eq. 57) is on the order of ∼ O( M N ), thus can be completely ignored under the Matsubara limit.</p><p>These effective scaling of the Planck constant is the main advantage of the Matsubara dynamics, compared to the previous approaches (see Appendix D) that rely on the truncation of the Liouvillian based on the argument of small , which may or may not be a good approximation. Also, note that the argument in Eq. 55 and 57 does not work for non-Matsubara modes, as the error term becomes O((N − M ) 3 2 /N 2 ) for Eq. 67 and O((N − M ) 2 2 /N 2 ) for Eq. 68, which are no longer small under the N → ∞ limit. Therefore, under the Matsubara limit, we can exactly expressed the original Matsubara Liouvillian L [M ] (with the expression of Eq. 55-Eq. 57 with Q Q Q M ) into the following equivalent expression</p><p>where</p><p>) are defined analogously as those in Eq. 48 and Eq. 49, respectively, where</p><p>In the above Matsubara Liouvillian, we have explicitly dropped the following higher order term (in L [M ] , Eq. 54)</p><p>where</p><p>∂Qn is a (K×K) matrix, with the matrix el-</p><p>. This term is in the order of O( M N 1 ) and hence can also be made as small as desired under the Matsubara limit. Note that this term accounts for the back action from the electronic subsystem to the nuclear DOF. 81 It has been shown that this term can be transformed into a local potential on nuclear DOF, leading to a non-Hamiltonian Liouvillian in Possion Bracket Mapping Equation (PBME), which can further improve the population dynamics. 83 In the non-adiabatic Matsubara dynamics, this term can be dropped which will not introduce and additional error beyond the order of O( M N 1 ).</p><p>The Matsubara correlation function in Eq. 59 contains an imaginary phase factor θ M , which potentially introduces a sign problem for a system that contains multidimensional nuclear DOF. In addition, Γ will also potentially introduce a sign problem (because it is also a complex quantity) if the system contains many electronic states, or the TCF has a large N . The numerical exploration suggests that this is not a severe problem for a two-state system [57][58][59] with a finite N (for N ≤ 16).</p><p>To eliminate the phase θ M , one can perform the following transformation 73 in P P P</p><p>Note that such transformation on P P P has no effect on the centroid mode (Q Q Q M ), as ωn for the centroid is zero (see Eq. 35 when N = 0).</p><p>Applying the transformation P n → Pn on the Liouvillian L [M ] (see SI for details) in Eq. 70 leads to the following complex Liouvillian (in terms { P</p><p>where we denote the real part of L[M] as the following non-adiabatic RPMD Liouvillian</p><p>and the imaginary part of L[M] as</p><p>Note that there is no mapping related derivative in the above imaginary Liouvillian L</p><p>[M ] I</p><p>, and its impact on the electronic dynamics should only come from its influence on the nuclear dynamics, which in turn couples to the electronic mapping DOF via</p><p>RP . Thus the influence from L</p><p>[M ] I to mapping variables is indirect. Using the above Liouvillian L [M ] , as well as apply the transformation in Eq. 72 to the quantum Boltzmann operator and the phase space integral in Eq. 59, one has the following equivalent expression of the non-adiabatic Matsubara TCF (see SI for details)</p><p>where the original integral</p><p>) is defined previously in Eq. 63 and, H RP M ( P</p><p>is the ring polymer Hamiltonian in the Matsubara domain expressed as follows</p><p>AB (t) in Eq. 76 is exactly equivalent to Eq. 59, with the difference that Eq. 76 has a real Liouvillian and a complex nuclear phase, whereas Eq. 76 has a complex Liouvillian, and a shifted nuclear momentum in the complex plane.</p><!><p>Eq. 76 is perhaps even more difficult to evaluate than Eq. 59 through a trajectory based approach, due to the complex phase space integral and the complex Liouvillian. However, at t = 0, one can analytically perform the integration in the complex phase space. To this end, we use the standard contour integration procedure described by Hele et al. 73 , and shift each imaginary Pn into real axis</p><p>P2 n ] without changing the integration. The details are discussed in the Supporting Information. This procedure allows us to write C</p><p>[M ] AB (t) (Eq. 76) at t = 0 as</p><p>where the original complex phase space integral becomes pure real (by shifting the momentum integral from a complex axis to a pure real axis). Note that at t = 0, hence e Lt = 1, the Matsubara approximation (by discarding L [N −M ] , see Eq. 53) no longer influences the value of</p><p>AB (0) gives the exact QBD (where the non-Matsubara modes can be analytically integrated out and do not influence QBD, as shown in Appendix B). Thus, Boltzmann operator ensures only the Matsubara modes contribute to the exact QBD. This is a well known result for path-integral in the electronically adiabatic case. 38,69,[89][90][91] The C</p><p>[M ] AB (0) expression in Eq. 78 is reminiscent of the mapping variable (MV)-RPMD partition function expression 57 , with the difference that Eq. 78 is expressed in the Matsubara space. On the other hand, one can directly obtain the MV-RPMD formalism from the generalized Kubo-Transformed TCF by taking the t → 0 limit of C</p><p>[N ] AB (t) in Eq. 14, (and explicitly assumes that both  and B are nuclear position related operators). Under these conditions, [ B(t)] N in Eq. 16 becomes</p><p>whereas the thermal Boltzmann operator becomes</p><p>Putting these two expressions back to the time correlation function C</p><p>[N ] AB (t) (eq. 14), and use the relation that</p><p>dP l e i P l D l = δ(D l ), we can explicitly integrate out the dP integrals (which are allowed when  is not a function of P ), resulting in</p><p>where α 0 N = (2π ) −KN . Through standard path-integral techniques (see details in the Supporting Information), one can explicitly show that the above</p><p>where</p><p>] (see Eq. 63 for detailed expressions), and H RP N ( P, R) is the standard ring-polymer Hamiltonian in the primitive nuclear coordinate {R l }</p><p>where the second line is the same ring polymer Hamiltonian expressed in the normal mode representation (through the transformations in Eq. 36, and with the normal mode frequency ω n in Eq. 35). Note that the d P integral in Eq. 82 was reintroduced from a constant (which can be re-expressed as the nuclear momentum Gaussian integral) through the standard path-integral procedure, which does not appear in the Liouvillian of C</p><p>[N ] AB (t) (Eq. 28). On the other hand, the d P P P M in C</p><p>[M ] AB (t) (Eq. 76) is the actual nuclear momentum integral that appears in both initial QBD and the Liouvillian.</p><p>Note that C</p><p>[N ] AB (0) is exactly equivalent to the original MV-RPMD partition function expression, with a slightly different procedure of the derivation compared to those in the original work. 57 Here, we show that this partition function can also be directly obtained from the linked multi-dimensional Wigner transform.</p><p>Further, C</p><p>[N ] AB (0) in Eq. 82 and C</p><p>[M ] AB (0) give the same exact quantum statistics, lim</p><p>such that the same quantum statistics can either be achieved under a large N limit for regular path-integral ring polymer, or under the large M limit for the Matsubara modes. The adiabatic limit of the above relation is a well-known result 38,69,[89][90][91] . Here, we explicitly demonstrate that this is also true for the non-adiabatic scenario. Note that under the adiabatic limit, the convergence of C</p><p>[N ] AB (0) with respect to an increasing N is much faster 38,69 than the convergence of C</p><p>[M ] AB (0) with respect to M (under the N → ∞ as well as M N limit). We have not performed any numerical test to confirm that this is also true for the non-adiabatic scenario, but we conjecture that this is the case.</p><!><p>The analytic continuation procedure performed in C</p><p>[M ] AB (0) (Eq. 78) is not valid when t>0 in general. This is because that when the dynamics is propagated with P P P M and Q Q Q M in the complex plane, one may encounter well-known singularities, 70,92 leading to a diverging e L [M ] t B(Q Q Q M ) such that the function of Pn no longer approaching to 0 when Pn → ±∞ from the real axis and breaks the counter integral trick (outlined in the Supporting Information).</p><p>On the other hand, as proposed in the original Matsubara dynamics work, 73 it is possible to follow a path along which each Pn is partially moved towards the real axis and L</p><p>[M ] I is partially discarded so the contour integration trick remains valid, and at the end of the path, L</p><p>[M ] I has been completely discarded and Pn has reached the real axis. 70 Applying this approximation on the nonadiabatic Matsubara dynamics leads to following nonadiabatic RPMD approach, which is the second key result in the paper as follows</p><p>In the above NRPMD expression of TCF, the initial distribution is governed by e −βH RP M ( P</p><p>), whereas the quantum dynamics is propagated by the Liouvillian L</p><p>[M ] RP (Eq. 74). If we choose to use ring polymer normal mode frequency instead of the Matsubara frequency in the above expression (and denote Pn as P n for simplicity) in Eq. 85, it then gives the non-adiabatic RPMD expression for TCF as follows</p><p>] (see Eq. 63 for detailed expressions), H RP N is the state-independent ring-polymer Hamiltonian (in the initial quantum Boltzmann operator) expressed as</p><p>which corresponds to the following NRPMD Hamiltonian 54 in the primitive nuclear coordinate as follows</p><p>Note that the frequency ω n is the ring polymer normal mode frequency (Eq. 35), whereas ωn in Eq. 85 is the Matsubara frequency (Eq. 51).</p><p>Dropping the imaginary part of the Liouvillian iL</p><p>[M ] I , unfortunately, introduces spurious frequency shift to the non-centroid normal modes, leading to the well-known "spurious resonances" problem in RPMD when there are resonances between ring-polymer frequencies and physical frequencies (such as stretching vibrations). 85,93 This problem can be partially resolved by replacing iL</p><p>[M ] I by an effective white-noise FokkerPlanck operator 94 , leading to the Thermostatting technique for RPMD. 85,93,94 This Thermostatting approach has also been recently incorporated into the NRPMD approach. 56 Note that the NRPMD approach (Eq. 86) in the current work (which can be viewed as an approximation of the non-adiabatic Matsubara dynamics C</p><p>[M ] AB through Eq. 85) samples the same initial distribution of MV-RPMD Γ(Q Q Q, q, p)e −βH RP N (P P P,Q Q Q) , whereas using the NRPMD Liouvillian in Eq. 87 (or NRPMD Hamiltonian 54 in Eq. 88) to propagate the dynamics. For a finite N , NRPMD does not preserve the QBD, 54,55,95 due to the fact that two different effective Hamiltonians are used for the initial sampling (</p><p>) and for the dynamical propagation (H NRP N in Eq. 88), respectively. However, it was conjectured that under the N → ∞ limit, NRPMD will preserve QBD. 54,95 On the other hand, because the MMST Hamiltonian structure is preserved in the dynamics propagation, NRPMD gives the exact electronic Rabi oscillations when the nuclear dynamics is decoupled from the electronic DOF. 54,55,78 This will be discussed further in Section VI as well as in Appendix C.</p><p>Further, our analytical work presented here also provides a theoretical justification for the recent numerical success of NRPMD, 54,55 which was initially proposed as a model non-adiabatic path-integral dynamics. 54,55 Note that the original version of NRPMD does not sample the MV-RPMD initial distribution Γ•e −βH RP N (as described in Eq. 86). Instead, it uses a simple position and momentum mapping variable resolution in the initial QBD, 54 resulting in the original version of the NRPMD TCF 54 as follows</p><p>where φ =</p><p>Thus, the only difference between the original NRPMD 54 and the NRPMD formalism in this work is the expression of the initial Boltzmann operator. Note that Γ is pure real, as oppose to the complex Γ in Eq. 86.</p><p>The recently developed MV-RPMD approach samples the initial distribution with HRP 83), and use the same Hamiltonian to propagate dynamics, resulting in the following MV-RPMD TCF</p><p>where the MV-RPMD Liouvillian is expressed as</p><p>which directly corresponds to the Hamiltonian HRP</p><p>Note that in the original MV-RPMD approach, 57 to facilitate the calculation with real trajectories, it was proposed to replace Γ → [Γ] (only taking the real part of Γ) in both the initial Boltzmann distribution as well as in the above Liouvillian. This argument is based upon the fact that the partition function Z RP N is real, and the operator estimators do not contain any imaginary part, hence the real and the imaginary part of the estimators are completely separated, and the ensemble average of the imaginary part should goes to zero. This is true if both  and B are not related to electronic states (mapping variables). However, there is no rigorous justification why this should also be applied to the Liouvillian. Moreover, for general operators, one should recognize that Γ is indeed complex, and a more rigorous trajectory approach in MV-RPMD should be replacing Γ → |Γ| in the distribution and the Liouvillian, then performing the ensemble average by weighting each trajectory with phase Γ/|Γ|. This will be discussed in detailed in Appendix E for the NRPMD simulation which also uses the MV-RPMD initial distribution.</p><p>To the best of our knowledge, there is no rigorous theoretical justification of the Liouvillian L</p><p>MV , MV-RPMD will not be able to provide the correct electronic Rabi oscillations when the electronic and nuclear DOFs are decoupled. In contrast, non-adiabatic Matsubara dynamics and NRPMD is exact under the electron-nuclear decoupled limit when the nuclear potential is Harmonic. On the other hand, MV-RPMD does preserve the QBD at any given N , because it uses the same Hamiltonian for initial sampling and for dynamics propagation. 57</p><!><p>We want to discuss the quantum detailed balance in our current formalism. For a system under the thermal equilibrium, the quantum expectation value does not change in time, Â(t) = Â(0) . Similarly, one can prove that</p><p>for Kubo-transformed TCF defined in Eq. 5 the above relation is commonly referred to as the condition for satisfying detailed balance.</p><p>The detailed balance condition is also true for the generalized Kubo-transformed correlation function C</p><p>[N ] AB (t) (Eq. 10), such that</p><p>This relation will also be rigorously satisfied for C</p><p>[N ] AB (t) in Eq. 14 (after performing the Wigner transform and replace quantum propagator with the Liouvillian L [N ] ). Since Eq. 14 is quantum mechanically exact, the key to achieve detailed balance condition is</p><p>where L [N ] is the exact Liouvillian in Eq. 24 and [e −βH ] N is the linked Wigner transformed Quantum Boltzmann Operator in Eq. 32, which is also exact quantum mechanically.</p><p>The Matsubara partition function in Eq. 66 is expressed as</p><p>where the new effective Hamiltonian HM is expressed as</p><p>One can prove that for the Matsubara phase θ M (P P P M , Q Q Q M )) is a conserved quantity of the nonadiabatic Matsubara Liouvillian L [M ] in Eq. 70, such that</p><p>where we have used the fact that in the Matsubara domain,</p><p>e ∂τ = 0 for the last equality (with τ as the imaginary time) due to the cyclic symmetry. Also note that the mapping part of the Liouvillian (∇ p l and ∇ q l inside L [M ] does not act on θ M (P P P M , Q Q Q M )). The details of this proof is provided in the Supporting Information, which can be viewed as a generalization of the original proof in the adiabatic Matsubara dynamics. 69 Unfortunately, we do not know whether the nonadiabatic Matsubara Liouvillian L [M ] commutes with the Hamiltonian, i.e., the validity of the the following relation</p><p>HM (P</p><p>thus we are not sure, at this moment, if the non-adiabatic Matsubara dynamics preserves the QBD governed by Z M . On the other hand, when electronic and nuclear DOFs are completely decoupled, such that the Hamiltonian of the system can be written as Ĥ =</p><p>Ve , the QBD is indeed preserved by L [M ] (see detailed discussions in Appendix C), because that the following relation</p><p>is satisfied by both L [N ] and L [M ] , where Γ(q, p) no longer depends upon Q n due to the nuclear position independent electronic potential Ve = i,j V ij |i j|.</p><p>For a general case beyond this special limit, we want to explore the conditions that when non-adiabatic Matsubara dynamics preserves the QBD. By requiring Eq. 99, it leads to the following condition</p><p>where Γ = Γ(Q Q Q M , q, p) in the above relation. Note that the above relation is the sufficient condition for preserving the QBD, whereas the necessary one requires n for all Matsubara modes in the above equation. Of course, for the electronic-nuclear decoupled case, both ∂Γ ∂Qn = 0 and ∂V V V ∂Qn = 0, as well as Eq. 100 is satisfied, hence L [M ] preserves the QBD. Beyond this special case, we do not know if Eq. 101 is always satisfied, and whether the non-adiabatic Matsubara dynamics preserves the QBD remains an open question and subject to future investigations.</p><p>Interestingly, if Eq. 101 is satisfied (i.e., the nonadiabatic Matsubara preserves QBD), then one can show that NRPMD must also preserves QBD. To explicitly demonstrate this, we rewrite the NRPMD time correlation function in Eq. 85 as follows</p><p>where HRP M ( P P P M , Q Q Q M , q, p) is the MV-RPMD Hamiltonian (with the Matsubara frequency instead of the ringpolymer frequency) expressed as follows</p><p>The condition for NRPMD to satisfy detailed balance is L</p><p>[M ] RP HRP M = 0, which resulting in the same condition described in Eq. 101. Hence, if non-adiabatic Matsubara dynamics preserves QBD, then so does NRPMD.</p><!><p>Beside the nuclear position auto-correlation function, the electronic projection correlation function is also an important one. 54,55 For example,  = B = |i i|. When  = |i i|, in general [e −β Ĥ Â] N = [ Â] N [e −β Ĥ ] N based upon Eq. 30. Hence, one needs to write down the generalized Kubo-transformed time correlation function as in Eq. 14. On the other hand, one can follow the same procedure to obtain the normal mode representation of C</p><p>[N ] AB (t) (see Eq. 39) as</p><p>as well as the exact procedure outlined in the previous section by making the Matsubara approximation (via discarding the L [N −M ] , and integrating out the non-Matsubara modes), reaching to the following expression of C</p><p>[M ] AB (t) (see Eq. 59 for a comparison)</p><p>In the above equation, Γ ii is Γ in Eq. 63 projected on |i i| as follows</p><p>where the property of trace ensures each term inside Tr e are identical, hence we can replace the bead average by the last equality. The Matsubara Liouvillian L [M ] is same as expressed in Eq. 70, and Z M is the quantum partition function expressed in Eq. 66.</p><p>Following the same procedure of discarding the imaginary Liouvillian and shifting the momentum integral to the real axis (i.e., the Ring Polymer approximation), and replacing the Matsubara frequency with the normal mode frequency, one can arrive at the following RPMD correlation function</p><p>Here, for B = |i i|, one can use the following estimator</p><p>where the detailed proof is provided in the Supporting Information.</p><p>Alternatively, one could also use the mapping relation</p><p>The above estimator is used in the original NRPMD approach, 54 which has been theoretically justified 55 as well as in MV-RPMD approach for excited state dynamics (which has been derived based on the property of Wigner transform). 58 In fact, C</p><p>[M ] AB (t) is equivalent to the NRPMD population time-correlation function, 54,55 when using Eq. 109 for the estimator of the population at time t, with the exception that the expression of Γ ii is obtained using the Wigner representation for the mapping variable in the current theory, whereas in NRPMD, it is obtained by using a simple integral of both mapping positions and momentum. 54 Our derivation explains the success of the original NRPMD approaches for simulating the population auto-correlation functions. 54</p><!><p>Despite the fact that RPMD was originally developed for equilibrium quantum dynamics simulations, recent theoretical progress has demonstrated that both RPMD and CMD can provide accurate non-equilibrium dynamics upon photo-excitation. 96 Thus, we conjecture that NRPMD is also capable to accurately describe the nonequilibrium TCF, and we will explicitly prove this.</p><p>For a given photo-induced process, we are often interested in the reduced density matrix dynamics upon the initial excitation of the molecular system. The reduced density matrix element can be expressed as</p><p>where the initial density operator ρ(0) is often expressed as a tensor product of the electronic and nuclear DOF as</p><p>, where Z = Tr[e −β Ĥ0 ], and Ĥ0 is the ground state Hamiltonian</p><p>with the ground state potential U g ( R) associated with the ground electronic state |g . The initial density ρ(0) is evolved under the influence of the total Hamiltonian Ĥ (Eq. 1). The reduced density matrix element can also be expressed as the following TCF</p><p>where  = |i i| is the initially occupied electronic state, and B = |i j|. Because  and Ĥ0 commute, [ Â, Ĥ0 ] = 0, hence</p><p>Thus, one can rewrite the time correlation function C AB (t) in Eq. 112 into the Kubo-transformed timecorrelation function C K AB (t) as follows</p><p>(113) The above TCF is not an equilibrium correlation function. Nevertheless, the Kubo-transformed structure allows us to express it as the discrete version of the timecorrelation function as in Eq. 5.</p><p>Following exactly the same derivation we have outlined in the previous sections, we can express C K AB (t) in Eq. 113 without any approximation as follows</p><p>dR dP dq dp ( 114)</p><p>where the Liouvillian L [N ] has the same expression in Eq. 24, and [e −β Ĥ0 ] N has the same expression in Eq. 15 except that Ĥ is replaced by Ĥ0 . There is no approximation in the above expression. Further, to evaluate</p><p>[ Âk ] W is the partial Wigner transformed projection operator (along the mapping DOF) with the following expression</p><p>where we have used the overlap relation</p><p>(π ) K/4 q i e q T k q k and explicitly performing the standard Gaussian integral (see detailed derivation in Supporting Information).</p><p>Similarly,</p><p>where G = e − 1 (q 2 k +p 2 k ) . On the other hand, there are other choices for the Wigner transform of operators. For example, the population can be directly Wigner transformed of â † i âi as shown in Eq. 109, and the Wigner transform of â † i âj as</p><p>This estimator has been proposed in the original NRPMD work 54,55 . It has also been derived in the nonequilibrium TCF with MV-RPMD. 58 Following the Matsubara approximation, we can derive the corresponding expression of the density matrix as follows</p><p>where the H M is expressed in eq. 61, θ M is expressed in Eq. 62, and the Liouvillian L [M ] is expressed in Eq. 70. This is the non-adiabatic Matsubara dynamics expression for the reduced density matrix elements as the third key result of this paper. Further making the RPMD approximation, we can derive the corresponding NRPMD expression of the reduced density matrix as the final key result of this paper</p><p>where H RP N (P P P, Q Q Q) has the same expression as that in Eq. 86, and the Liouvillian L RP N is expressed in Eq. 87, corresponding to the NRPMD Hamiltonian expressed in Eq. 77. This is the NRPMD expression of the reduced density matrix.</p><p>Thus, we explicitly show that NRPMD is capable to simulate non-equilibrium TCF, explaining the recent numerical success of using NRPMD to simulate the nonequilibrium population dynamics. 56 Similar numerical success in simulating non-equilibrium TCF has also be achieved in MV-RPMD. 58</p><!><p>We present the non-adiabatic Matsubara dynamics, a general framework for computing the time-correlation function of electronically non-adiabatic systems. This new formalism is derived based on the generalized Kubotransformed time correlation function, using the Wigner representation for both the nuclear DOF and electronic mapping variables. [74][75][76] By dropping the non-Matsubara nuclear normal modes in the quantum Liouvillian, we derive the non-adiabatic Matsubara dynamics, which can be viewed as a generalization of the original (electronically adiabatic) Matsubara dynamics. 69 The nonadiabatic Matsubara dynamics has two complex phases, one from the nuclear DOF and the other from the electronic DOF. By making a nuclear momentum transformation, one can derive an equivalent expression of nonadiabatic Matsubara dynamics, which has a complex Liouvillian and a complex momentum distribution. Further making an approximation that drop the imaginary part of the Liouvillian, we arrive at the non-adiabatic ringpolymer molecular dynamics formalism. Thus, NRPMD can be viewed as an approximation of non-adiabatic Matsubara dynamics. Interestingly, the initial distribution of NRPMD coincides with that in the Mapping-Variable (MV)-RPMD 57 , whereas the NRPMD Liouvillian coincides with the Liouvillian used in the originally proposed NRPMD 54 (which has a different initial quantum distribution). Our theoretical derivations explain the numerical success of both of these previous approaches. 54,57 We have further proven that the NRPMD is capable to simulate non-equilibrium TCF, hence justifies such simulations and explains the recent numerical success. 56 At this moment, we are not sure whether non-adiabatic Matsubara dynamics preserves the quantum Boltzmann distribution (QBD), except for the special limit when the electronic and nuclear DOFs completely decoupled. Nevertheless, we derived the condition under which the QBD will be preserved by non-adiabatic Matsubara dynamics. Interestingly, if non-adiabatic Matsubara dynamics preserves the QBD, then NRPMD is also guaranteed to preserve the QBD.</p><p>Let us also give more technical summaries of these equations. In this work, we started from an exact formalism of generalized Kubo-transformed TCF C</p><p>[N ] AB (t) (Eq. 28). Making the Matsubara approximation by dropping the L [N −M ] as well as integrate out the non-Matsubara related terms, we have the non-adiabatic Matsubara TCF C</p><p>[M ] AB (t) (Eq. 59 or equivalently, Eq. 76). Note that the exact non-adiabatic QBD is completely governed by these Matsubara modes, and completely irrelevant to the non-Matsubara modes. Further, In the Matsubara domain, is effectively scaled as small as one desired (Eq. 67 and Eq. 68), such that the Liouvillian can be truncated to the linear order of the nuclear derivative Λ without making any approximation. This is in contrast to the Linearized path-integral approaches [16][17][18][19][20][21][22][23] that often need to truncate the Liouvillian up to a certain order of . The Matsubara formalism, on the other hand, keep all orders of within the Matsubara modes and discard all non-Matsubara modes. 69,70 Further making the approximation that drops iL</p><p>Note that both the Matsubara approximation (by dropping L [N −M ] in Eq. 131-133) and the ring polymer approximation (by dropping iL</p><p>[M ] I in Eq. 75) are only related to the nuclear DOF, as these dropped term only contain the nuclear derivatives. For non-adiabatic Matsubara dynamics and NRPMD, we have not make any direct approximations in the mapping DOF. Thus, we only expect an indirect influence of these error terms in Liouvillian on the electronic mapping dynamics.</p><p>There are several interesting limits we would like to discuss as well. (i) Under the limit that the system only contains electronic subsystems, C</p><p>[M ] AB (t) and C NRP AB (t) reduces to the same form of C</p><p>[N ] AB (t) (note that there are still N copies of the mapping DOF for all formalisms), which are all quantum mechanically exact. Hence, for isolated electronic subsystem, both non-adiabatic Matsubara dynamics and NRPMD preserves the exact quantum Rabi oscillations (where an explicit proof can be found in Ref. 78). (ii) Under the single electronic state limit (adiabatic limit), the non-adiabatic Matsubara formalism reduced to the original Matsubara dynamics. 69,70 (iii) Under the decoupled limit of the electronic and nuclear DOF, non-adiabatic Matsubara dynamics rigorously preserves QBD, and gives the exact dynamics for the electronic subsystem, while only give an approximate dynamics for the nuclear subsystem (exact when the potential V 0 is purely harmonic. 69,70 (iv) Under the N = 1 limit, the NRPMD formalism reduced to Linearized Semi-classical Initial Value Representation (LSC-IVR) approach (with a classical nuclear distribution instead of the Wigner distribution, see more explicit discussions in Appendix D).</p><p>The immediate future direction of the current work is to test the numerical performance of the non-adiabatic Matsubara dynamics for computing equilibrium and nonequilibrium TCFs. The current formalism of C</p><p>[M ] AB (t) makes this a challenging task, because the electronic mapping DOFs also have N copies, which are required to take the N → ∞ limit. However, one does not have to use the same number of copies of the mapping DOF and nuclear DOF, and this can be accomplished through the mixed time-slicing technique 97 which has been successfully implemented in a recent work of NRPMD. 55 By using a finite number of the mapping resolution (Eq. 7 and Eq. 8), we expect to make the non-adiabatic Matsubara dynamics practical for system with a few nuclear DOF and a few electronic states. Another direction is using the non-adiabatic Matsubara dynamics framework to theoretically derive other existing state-dependent path-integral approaches, such as the non-adiabatic CMD 98 (through a mean field approximation 73,86 of the non-adiabatic Matsubara dynamics), the coherent-state RPMD 63 approach (through a new non-adiabatic Matsubara dynamics that uses the Husimi representation for mapping variables and Wigner representation of nuclei), or the ring-polymer surface hopping approach [64][65][66][67][68] (through the recently discovered connections between the quantum-classical Liouville equation (QCLE) and the fewest switches algorithm 99,100 ). A third direction is to theoretically explore whether non-adiabatic Matsubara dynamics rigorously preserve quantum Boltzmann distribution (QBD). If so, then non-adiabatic Matsubara will be a trajectory-based approach that can correctly describe electronic Rabi oscillations and preserve the QBD, a method that is currently lacking. 101 We hope that our current work provides a framework and a new paradigm for accurate non-adiabatic quantum dynamics approaches by interfacing the recent development in the field of mapping dynamics [such as new mapping representations [102][103][104][105][106] as well as new estimators (window functions 107,108 and the identity trick 109 )] with the development of accurate nuclear quantum dynamics in the field of path-integral dynamics [such as the Matsubara dynamics [69][70][71][72] and its approximations, including CMD, 40,73 RPMD, 43,44,73 mean-field Matsubara dynamics, 86 quasi-CMD 87 , Planetary model 70,88 , etc]. It allows one to borrow recent developments from each subfield, and facilitate the merger of both sub-fields for developing more accurate non-adiabatic quantum dynamics approaches.</p><!><p>This work was supported by the National Science Foundation CAREER Award under the Grant No. CHE-1845747. P.H. appreciates the support from a Cottrell Scholar award (a program by Research Corporation for Science Advancement). We are indebted to Prof. Stuart Althorpe's help on understanding the Matsubara dynamics and its connection to the ring polymer molecular dynamics. We highly appreciate valuable discussions with Dr. Tim Hele, Dr. Michael Willatt, Dr. Pablo Videla, and Dr. Duncan Bossion.</p><!><p>The data that support the findings of this study are available from the corresponding author upon reasonable request.</p><!><p>We provide the derivation of the exact Liouvillian L [N ] (Eq. 24) as well as the detailed expressions of the non-Matsubara Liouvillian L [N −M ] (Eq. 53)</p><p>We start by differentiating a time-dependent Wigner transformed of a general operator  as follows</p><p>where [ Â(t)] W = dDe</p><p>2m + V ( R), with the nuclear position operator R and corresponding momentum operator P . Further, V ( R) is any general form of the potential energy operator.</p><p>Using the following equality of Wigner transform of two operators 79,84</p><p>where Λ in general is written as Eq. 20), one can rewrite Eq. 120 as</p><p>where in the final step of Eq. 121 we have used the fact that e −i Λ /2 = cos( Λ /2) − isin( Λ /2) and e i Λ /2 = cos( Λ /2) + isin( Λ /2). Using Eq. 121 in Eq. 120 we have</p><p>where the Liouvillian is</p><p>For the case where V is electronically adiabatic, one can explicitly evaluate the detailed expression of the Liouvillian. We first take the Wigner transform of the Hamiltonian 110 as [ Ĥ] W = dDe</p><p>With Eq. 123 and [ Ĥ] W , the Liouvillian is expressed as</p><p>Using the Taylor expansion of sin Λn and the V (R) term, one reach to the well-known Wigner-Moyal series 79,80,84 as follows</p><p>Next, we derive the explicit expression of the Exact Liouvillian of the non-adiabatic Hamiltonian in the generalized Kubo-transformed TCF which has the following expression</p><p>where Λl is defined in Eq. 20, and [ Ĥl ] W is the Wigner transform of the MMST mapping Hamiltonian (Eq. 1) associated with the l th bead, with the expression in Eq. 22.</p><p>Because L [N ] contains N mathematically identical terms (labeled as l ∈ [1, N ]), one only needs to derive the expression of one term and sum them up. Below, we explicitly derive one of this term denoting as L [1] , and we drop the label l for simplicity. With the operator Λ = Λe + Λn defined in Eq. 20, one can rewrite Eq. 126 as</p><p>Explicitly expanding the terms related to the mapping derivatives as cos</p><p>3 ), and note that the V e (R, q, p) term inside the [ Ĥ] W contains up to the second order of p and q (see Eq. 23), such that V e [ Λe ] n = 0 for n ≥ 3, one can rewrite Eq. 127 exactly as follows</p><p>Each term of Eq. 128 can be explicitly evaluated (see the Supporting Information for details), resulting in the exact non-adiabatic Liouvillian as follows</p><p>Adding a total of N mathematically identical terms together (each has the same expression of the above Liouvillian L [1] ), we have L [N ] expressed in Eq. 24.</p><p>Next, we explicitly express the error term of the Liouvillian L [N −M ] when applying the Matsubara approximation. The Matsubara approximation discards (N − M ) Non-Matsubara modes from Eq. 43, where the non-Matsubara Liouvillian is expressed as follows we can express the non-matsubara Liouvillian as follow</p><p>where the operators â and b are defined as â = 2N</p><p>Note that the entire L [N −M ] does not contain derivatives with respect to the electronic mapping variables. Hence, they do not have the direct influence on the electronic dynamics.</p><!><p>To obtain the Quantum Boltzmann Distribution function under Matsubara limit in Eq. 59, we first write down the total Boltzmann factor in normal mode representation as</p><p>which is same as Eq. 41 and ξ ± is previously defined in Eq. 40. The expression in Eq. 136 is equivalent to Eq. 104 of the work by Hele and Ananth 78 , but in normal mode representation.</p><p>Next, we integrate out the non-Matsubara P P P. By the construction of the Matsubara dynamics, there is no functional dependency of non-Matsubara P P P modes in operator B and thus, we can integrate out the non-Matsubara P P P momenta, giving a product of Dirac delta function in non-Matsubara D modes of the form</p><p>which helps to further integrate out the non-Matsubara D [N −M ] modes from Eq. 136. After integrating out the non-Matsubara D [N −M ] modes, [e −β Ĥ ] N (Eq. 136) becomes</p><p>where dD</p><p>n=−(M −1)/2 dD n that only includes the Matsubara modes, whereas the Q n contains all modes (together with all Q Q Q dependent terms in the Liouvillian in the TCF at this moment). Further, η ± l is now expressed as</p><p>noticing that D n only contains the Matsubara modes.</p><p>We further split the Boltzmann operator in Eq. 137 through a symmetric Trotter expansion under the N → ∞ limit, noticing that |η ± l η ± l | commutes with P (as they belong to two different DOFs), and evaluating the nuclear kinetic energy term P 2 2m explicitly through the standard path-integral technique (see SI for details). We can then express Eq. 137 as follows</p><p>Applying the well known trigonometric identities, 69,70 one can explicitly evaluate (η</p><p>where ξ n = nπ N . Note that the first sum includes all Matsubara modes and the second sum includes all non-Matsubara modes. Using the relation in Eq. 140, the quantum Boltzmann operator in Eq. 139 becomes</p><p>Note that under the limit N → ∞, the Gaussian function that involves D −n term have the following form</p><p>Using the properties that f (x + a)δ(x)dx = f (a) δ(x)dx, we can move all terms related to the potential, including exp[−β N V0 (η ± l )] and exp[−β N Ve (η ± l )] related terms outside the dD D D M integral (and set D −n = 0 for η ± terms in Eq 138), resulting in</p><p>where we have explicitly write down each integral for the Matsubara D n mode, R l =</p><p>√ N T ln Q n , and we have defined the following mapping integral as</p><p>We explicitly integrate out dD n in the last line of Eq. 142 (by denoting ξ n = nπ/N ) as follows</p><p>Noticing that the Gaussian term inside the above dD n integral is</p><p>which again allows dD n integral to be evaluated through steepest descent fashion, resulting in</p><p>where</p><p>. Because the dD n integral is in the Matsubara domain, we can further simplify the expression by taking the Matsubara limit (N → ∞ and n ∈ M N ), resulting in cos 2 ξ n → 1, sin 2 ξ n ∼ O(( n N ) 2 ) → 0, and tan ξ n → ξ n = nπ/N . Thus, the final expression of the thermal Boltzmann operator in Eq. 142 is expressed as</p><p>In addition, one can solve the electronic mapping integral in Eq, 143. Recall that P = i |i i| is the projection operator in SEO mapping subspace, and the overlap with electronic states is expressed as q|i = 2 1 (π ) K/4 [q] i e −q T q/2 (148)</p><p>Using Eq. 148 one can re-express Γ as follows</p><p>Rearranging the prefactors of Eq. 149 and grouping terms associated with ∆ ∆ ∆ l (using cyclic property of trace), we have</p><p>Analytically performing the integration over ∆ ∆ ∆ l (which is a standard gaussian integral) leads to the final form of Γ(Q Q Q, q, p) as follows</p><p>where φ = 2 (K+1)N N , with G N = N l=1 (q l q T l + p l p T l ). Further, C l = (q l + ip l ) × (q l − ip l ) T , and I is the (K × K) identity matrix. Note that C l − 1 2 I I I can be interpreted as the reduced density matrix associated with the l th bead. In addition, M ij (R l (Q Q Q)) is the matrix of electronic Boltzmann operator as follows</p><p>is the state dependent potential operator.</p><p>Using the above results, we have the following thermal-Boltzmann operator</p><p>Substituting Eq. 152 into Eq. 58 we get,</p><p>Under the Matsubara limit M → ∞, while M N , one recognises that the Gaussian part of the non-Matsubara normal modes in [e −β Ĥ ] N (P P P M , Q Q Q, q, p) are nascent Dirac delta functions (by noticing the expression of ω n in Eq. 35) as follows lim</p><p>Thus, we can further integrate out the non-Matsubara Q Q Q modes from Eq. 152 (based on a steepest descent argument), leading to the following effective changes inside the nuclear coordinate integral</p><p>where</p><p>is defined in Eq. 61, as well as</p><p>) for the Γ related term. Because these quantities only depend on Matsubara mode Q Q Q M , we can move them outside the integral of the non-Matsubara modes</p><p>where Gaussian integral in the last line of the above equation can be analytically performed, resulting in the constant α M in Eq. 60. The final results of C</p><p>[M ] AB (t) is expressed in Eq. 59.</p><!><p>For the system that has a decoupled electronic-nuclear interaction,</p><p>where V ij is a constant thus ∂V ij /∂R = 0. This case also include two limits (1) electronically adiabatic system ( Ve = 0) or (2) there is only electronic subsystem ( T + V0 = 0).</p><p>For the electronic-nuclear decoupling case, the exact thermal Boltzmann operator is</p><p>where the electronic part Γ(q, p) becomes</p><p>with</p><p>The exact Liouvillian in Eq. 24 becomes</p><p>where the nuclear Liouvillian (the first term) and the electronic Liouvillian (the second term) are completely decoupled. The detailed balance condition L</p><p>for the electronic subsystem, and</p><p>2 D l } = 0 for the nuclear DOF. Under the same decouple limit, the non-adiabatic Matsubara Liouvillian becomes</p><p>and the Hamiltonian HM becomes HM =</p><p>Hence the non-adiabatic Matsubara Liouvillian becomes the separable Matsubara Liouvillian for the nuclear DOF (first line) and the mapping Liouvillian for the isolated electronic DOF (second line). Comparing with the exact Liouvillian in Eq. 161, one can see that the electronic part of the Liouvillian is also exact in L [M ] for this special case.</p><p>Then we can show that under such decoupled limit, non-adiabatic Matsubara dynamics preserves the QBD as follows</p><p>where we have used Eq. 162, as well as L [M ] H M (P P P M , Q Q Q M ) = 0 (which can be easily verified by acting L [M ] on H M (P P P M , Q Q Q M ). Hence, we proved that under the decoupling limit, non-adiabatic Matsubara dynamics preserves the QBD.</p><!><p>We would like to connect the current formalism of C</p><p>[N ] AB (t) with previously developed linearized pathintegral approaches. Most of these approaches are based on the Wigner representations for the mapping and nuclear DOF, which can be formally viewed as various approximate forms of N = 1 case of C</p><p>[N ] AB (t). Since they are extensively used to compute density matrix dynamics, we will mainly focus our discussion on the non-equilibrium correlation function (Eq. 112). On the other hand, the following discussions on Liouvillian are also valid for the thermal-equilibrium time-correlation function (Eq. 14).</p><p>We begin by writing down the non-equilibrium timecorrelation function in Eq. 114 with N = 1 as follows</p><p>, and the Liouvillian is</p><p>The expressions of C AB (t) and the above Liouvillian L [1] are in principle exact, which give rise to the exact quantum dynamics. Note that the mapping variables q i , p i are in the order of O( √ ) (because [p i , q i ] = i , hence there is an 1/2 term in the MMST Hamiltonian in Eq. 4).</p><p>Of course, one can make approximations to L [1] . For example, if one truncate all terms up to O( ) as follows 2 sin 2</p><p>and drop the third line in L [1] which corresponds to a term that is in the order of O( 0 ), the Liouvillian L [1] is reduced to the following form</p><p>which is the Liouvillian used in the Linearized Semiclassical Initial Value Representation (LSC-IVR) approach. Note that the error for the first line in L LSC is O( 2 ), and the error for the second line in L LSC is O( ). However, the dropped term (the third line) in L [1] corresponds to a term of O( 0 ). This ultimately determines the accuracy of L LSC to be up to O( 0 ). On the other hand, if one choose to truncate L [1] up to the linear order of nuclear operator Λn =</p><p>∂P , which is commonly referred to the mixed quantum-classical (MQC) Liouville approximation, 111 then L [1] reduces to the following</p><p>which was first derived in the Possion Bracket Mapping Equation (PBME) approach. 25,81 Note that the last term scales as O( 0 ). Hence the accuracy of L [1] MQC is in principle up to O( ). However, the last term was not straightforward to evaluate. 81 Hence, in the common PBME approach this term is often dropped, and L [1] LSC is used in the PBME calculation. Later, it was shown 81,83 that this term can be equivalently expressed as</p><p>Explicitly including this term and using L [1] MQC for QCLE (which is referred to as the non-Hamiltonian PBME 83 ) indeed improves the accuracy of the population dynamics in spin-boson problems. 83 Note that under the Matsubara limit, this term can be make as small as needed (see Eq. 71)</p><p>If one wants to continue to improve the accuracy of the Liouvillian, the next term to include will be O( ) comes from the quadratic term in the cos expansion in the second line of L [1] . Retaining this O( ) term one can write the following non-Hamiltonian (NH) Liouvillian</p><p>Note that because of the approximation used in QCLE (truncating up to O( Λn ), the third term in L [1] NH does not show up in QCLE. Also, for the spin-boson model (with c and ∆ as the parameters), the state dependent potential matrix V(R) = cR • σ z + ∆ • σ x is purely linear, hence the third in L [1] NH act on V(R) gives strictly zero results. For a general problem, however, both the third and the forth term should be included to fully account the terms of O( ).</p><p>Of course, the form of the Liouvillian is not the only factor that could influence the results of C AB (t). For the population dynamics, how to approximate [ Â] W and [ B] W will also significantly influence the accuracy of these approximated methods. 109 For example, when computing ρ jj (t) = Tr[ρ(0)e i Ĥt |j j|e − i Ĥt ]. The standard LSC-IVR approach 16,18</p><p>where |j j| is the electronic projection operator and â † j âj is the corresponding operator in the mapping representation. Their Wigner transforms are</p><p>The numerical comparison between these two approaches has been extensively discussed in the recent work, 109,112,113 and the recent development on choosing the identity operator 109,113,114 has also shown to significantly improve the population dynamics, even just using a less accurate Liouvillian L [1] LSC . Along the same direction, one can use the mapping action variable's Wigner transform 115 to construct [ Â] W and [ B] W , and engineer various shapes of Window functions for these estimators. 107,108,116 This idea has also lead to significant improvement of the population dynamics. 33 Finally, if one takes the N = 1 for the nuclear DOF and N = 2 for the electronic mapping DOF, as well as making the truncations in Eq. 166-Eq. 167 and dropping the last term in Eq. 165, the exact Liouvillian L [N ] reduces to the partially linearized Liouvillian as follows</p><p>V e (R, q l , p l )]</p><p>which is reminiscent of the Liouvillian used in the Forward-Backward trajectory solution (FBTS) for the QCLE 26,27 , and is also closely related to the equation of motion in the Partial Linearized Density Matrix (PLDM) path-integral approach. 22,117,118 However, we have to cautious to draw any further connections between the correlation function C</p><p>[N ] AB (t) in Eq. 114 and those in FBTS and PLDM, as the latter two approaches use the coherent state representation for the mapping DOFs instead of the Wigner representation used in this work, and hence should be viewed as the Hybrid Husimi (mapping)-Wigner (nuclear) representation for non-adiabatic pathintegral dynamics. In addition, the recently proposed two-oscillator mapping of PBME 119 also adapts the same Liouvillian L PL , even though the two copies of the mapping variables are introduced through the mapping relation of the electronic states.</p><!><p>Finally, we provide preliminary numerical tests one the non-adiabatic RPMD correlation function in Eq. 86. To numerically compute it with a trajectory based approach, we rewrite it as follows</p><p>where the initial distribution |Γ(Q Q Q, q, p)| • e −βH RP N (P P P,Q Q Q) is sampled by Monte-Carlo (based on a simple metropolis algorithm), the dynamics is propagated by e L [N ] RP through a simple numerical integrator, 55 and each trajectory is weighted by a complex phase Γ |Γ| . However, we can further take advantage of the pure real estimators for A and B, as well as the Liouvillian L</p><p>where sgn(ReΓ) is the sign (plus or minus) of the real part of Γ. The above expression is based on the fact that C</p><p>[N ] AB (t) is pure real and the ImΓ part is completely separated from the ReΓ, and does not contribute to the value of the C</p><p>[N ] AB (t). (Note that the ensemble average of ImΓ is 0, but ImΓ for individual trajectory is not).</p><p>We adapt a commonly used model system that contains one nuclear coordinate and two electronic states 54,57</p><p>where ∆ is the electronic coupling, c is the vibronic coupling, and 2ε is the energy bias between the two diabatic states. We choose a reduced unit system such that M = = 1 and ω = β = c = 1. We choose N = 8 beads for model I, and N = 6 beads for model II-IV. A total of 10 5 trajectories are used for tight numerical convergence, even though only 10 3 trajectories are good enough to provide the basic trend. Table I presents the parameters for all of the model systems used in this paper. In particular, Model I is in the adiabatic regime, where ∆ β −1 ; Model II and III are in the non-adiabatic regime, where ∆ β −1 ; Model IV is in the intermediate regime, where ∆ ∼ β −1 . Model III is an asymmetric case with finite diabatic energy bias 2 , and the rest of the model systems are symmetric cases with =0. Fig. 1 presents the nuclear position auto-correlation function computed from NRPMD (black) and the numerical exact method (red) for Models I-IV. Model I in Fig. 1a is in the adiabatic regime. In this case, NRPMD goes back to the standard RPMD, and agrees with the exact result due to the near Harmonic adiabatic potential. Model II in Fig. 1b is in the non-adiabatic regime. This is the most challenging case and the most relevant regime for non-adiabatic electron transfer 46 and proton-coupled electron transfer reactions. 47 In this regime, mean field RPMD starts to break down even at a very short time, as shown in the previous work. 57 NRPMD on the other hand, performs reasonably well compared to exact DVR calculations at the longer time. Models III corresponds to the asymmetric non-adiabatic regime (Fig. 1c) with diabatic energy bias 2 and model IV is in the intermediate regime. In this regime, NRPMD behaves reasonably well.</p><p>We find that the numerical results obtained with the current NRPMD formalism (for the current model systems) are not significantly different than those obtained from the original NRPMD, 54 due to the same Liouvillian used in both formalisms. On the other hand, the correlation function obtained from MV-RPMD 57 starts to oscillate with a different frequency (see Fig 1b) compared to the quantum result at a longer time, especially for model II and III, even though it uses the same initial QBD for NRPMD. This might happen because of the inter-bead couplings for mapping DOF in the Liouvillian (Eq. 92), which starts to contaminate the physical frequency of the system. The same behavior has also been found for population related quantities. 57 On the other hand, MV-RPMD does preserve QBD for any arbitrary number of beads N , whereas for a finite number of N , NRPMD does not preserve QBD. 54,55,95</p>

ChemRxiv

Smart, programmable and responsive injectable hydrogels for controlled release of cargo osteoporosis drugs

Easy-to-prepare drug delivery systems, based on smart, silica gels have been synthesized, characterized, and studied as hosts in the controlled release of bisphosphonates. They exhibit variable release rates and final % release, depending on the nature of bisphosphonate (side-chain length, hydrophilicity/-phobicity, water-solubility), cations present, pH and temperature. These gels are robust, injectable, re-loadable and re-usable.Gel systems have found extensive applications in the medicinal/pharmaceutical field because of their ease of preparation 1 , ability for modifications 2 , and responsiveness to external chemical 3 or physical stimuli 4 . Gels usually act as hosts for active pharmaceutical agents for a variety of pathological conditions 5 . They function as controllers of the release of pharmaceuticals that have proven to be "problematic" because they are either unsuitably insoluble to biological fluids 6 , or they are metabolized unacceptably rapidly 7 .Among the known bone diseases (osteoporosis, osteoarthritis, multiple myeloma, Paget's disease and several others), the most challenging is osteoporosis, which burdens millions of people compromising patients' quality of life 8 . The recommended pharmaceutical treatment is the use of bis-phosphonates (BPs, a.k.a. "-dronates") 9, 10 . Etidronic acid 11 is the first osteoporosis treatment to enter the market (1977), while zoledronic acid 12 is one of the treatments that followed (2007). Studies with N-containing BPs have shown that they are taken up by mature osteoclasts and inhibit farnesyl pyrophosphate synthase, an enzyme of the mevalonate pathway 13 . Their success in mitigating osteoporosis notwithstanding, these "-dronate" drugs present a number of challenges including fast excretion 14 , and numerous side-effects, such as osteonecrosis of the jaw, hypocalcemia, esophageal cancer, ocular inflammation, atrial fibrillation, etc. 15 . Nevertheless, the main drawback of BPs is their limited oral bioavailability (for example, it is ~3% for etidronic acid), which obligates physicians to increase drug intake in order to achieve the therapeutic dosage. It is, therefore, imperative to design and fabricate "smart" systems that allow controlled delivery of the active BP agent, which will depend on the patient's needs and idiosyncrasies.Furthermore, although all injectable BP therapies [eg. ibandronate (a.k.a. Bovina) and zoledronate (a.k.a. Reclast)] have been reported to alleviate the adverse effects of administration via pills/tablets, and are administered as scarcely as once a year, they are given to the patient in liquid form. The low frequency of administration of the above-mentioned BPs correlates with their very high potency as anti-resorption agents.BP controlled release systems are scarce, however, there are some published examples. Wang et al. were the first to graft pamidronate onto a polymeric chain 16 . Their studies led to a new class of hydrogels containing polymeric pamidronate via crosslinking of poly(N-acryl pamidronate-co-N-isopropylacrylamide). A microsphere-based system of polymer-mediated, aledronate-loaded hydroxyapatite was fabricated by Huang et al. 17 as a new releasing device for aledronate in bone repair applications. Chaudhari et al. 18 proposed a new nanoscale targeting system which involves nanoparticles of poly(lactide-glycolide) acid, poly(ethyleneglycol) and zoledronic acid as a nanocarrier-based drug delivery system (DDS). Sundell et al. have prepared a poly[ethylene-g-(vinylbenzyl chloride)] film and then grafted a BP on it ref. 19. Hydrolysis of the polymer initiated the BPs release. Johnston et al. characterized and evaluated refillable polyurethane reservoirs with regard to release of etidronic acid in vitro

smart,_programmable_and_responsive_injectable_hydrogels_for_controlled_release_of_cargo_osteoporosis

<!>Results and Discussion<!>Conclusion<!>Methods<!>Preparation of gels.

<p>into a receptor phase 20 . Kim et al. described bioabsorbable calcium phosphate microspheres that can incorporate alendronate through an in situ loading process and can control the alendronate release rate 21 .</p><p>It is now well-established that silica is not (cyto)toxic and has excellent biodegradation properties [22][23][24][25][26] . Silica has found several uses in diverse scientific and technical application fields, from food additive, to drug excipient. Furthermore, silicic acid (the dissolution product of amorphous silica) is also non-toxic. Moreover, it has been suggested as beneficial to bones 27 . The majority of studies use amorphous xerogels, mesoporous silica (such as MCM-41 or SBA-15) or fumed silica nanoparticles. Specifically, mesoporous silica nanoparticles have specific characteristics appropriate for applications such as DDSs 28 . Usually harsh conditions (elevated temperature, high pressure, and strongly alkaline or acidic solutions) are required for the fabrication of silica and particular mesoporous silica nanoparticles. Prior to loading a drug molecule, a high-temperature (~600 °C) calcination step is necessary, or template extraction using concentrated acid. As a result, these procedures raise additional costs, time and complexity in fabricating DDS 29 . Balas et al. tried to unravel the very attractive application of siliceous ordered mesoporous materials combined with BPs. Two types of hexagonal ordered mesoporous materials, MCM-41 and SBA-15, were used as matrices for alendronate adsorption and release 30 .</p><p>There is also widespread confusion regarding the safe use of mesoporous silica nanoparticles, especially because the interactions at the "nano-bio" interface are unknown. Therefore, it is of paramount importance to design, synthesize and fabricate DDSs for BP controlled release under mild conditions that fit an acceptable biocompatibility and toxicity profile.</p><p>In this paper, we report a detailed study on an easy-to-prepare silica hydrogel-type DDS that can host and incorporate a wide variety of BPs, and subsequently release these in a controlled manner. These hydrogels are sufficiently fluid to be injectable (see Fig. S-1, Supplementary Information, SI). Several factors have been found to influence the controlled release of the active BP, such as cations present in the gel, active groups on the BP backbone, gel density, and temperature. These systems are intended for potential biomedical applications.</p><!><p>For the present study seven BPs (SI, Table S-1) were used for the fabrication of BP-loaded gel DDSs. The general protocols for gel preparation, controlled release and sampling are shown schematically in Fig. 1.</p><p>All studied BPs possess the same substituent environment around the central carbon atom (two phosphonate and one hydroxyl moieties), while the R side chain is variable. Hence, the selection criteria were based on features of R such as polarity, hydrophobicity-philicity, and end-groups. Specifically, three of the BPs possess a systematically elongated hydrophobic alkyl chain (ETID, C3BP, and C5BP), and four display a systematically elongated aminoalkyl chain (PAM, ALE, C4NBP and NER). The importance of the BP side chain on drug effectiveness has been highlighted for commercial BPs 31 . Representative gels ("empty" ie. without BP, loaded with BP and finally after release) have been studied by SEM-EDS (after supercritical drying, see Figs S-2 to S-6, SI) and, as expected, they display amorphous features (by powder XRD). The presence of BP was confirmed by EDS and NMR (see below).</p><p>Solid state NMR (ssNMR) was used to probe the presence of the BPs in the prepared silica gels by observing the 31 P spins. The spectra of the selected pure educts indicate different crystallographic positions of the P-containing functional groups by their isotropic chemical shift (Fig. 2). This is not uncommon, as evidenced by the crystal structures of several BPs and their Na salts 32 . Each chemically and magnetically inequivalent 31 P contributes to a signal in the shown direct excitation MAS spectra (see Fig. 2, and data in Table 1).</p><p>In contrast to the dry educts, the 31 P direct excitation MAS spectra of the BPs exhibit less lines after incorporation into the silica gels (Fig. S-37, SI, and data in Table 1). This might be due to the aqueous environment in the wet gels. Two separate peaks could only be observed for PAM, likely caused by the fast recrystallization of PAM within the silica gel in the rotor. This observation is supported by 1 H-31 P CP MAS measurements. In CP MAS experiments the magnetization transfer efficiency from protons to hetero nuclei is inefficient for non-rigid systems. In case of the BP-loaded silica gels, signals were exclusively found for the PAM sample (Fig. S-38, SI).</p><p>Furthermore, the direct excitation spectra of 29 Si are very similar for all BP-loaded DDS gels. Specifically, there are no significant differences of the peaks caused by Q 2 , Q 3 and Q 4 groups occurring within the different gel samples (see Fig. 3 for selected results). Therefore, the presence of BPs seems to have no impact on the overall bulk structure of the prepared BP-loaded silica hydrogels.</p><p>Additional hydrogel characterization involved their rheological behavior. For this, three representative hydrogels were selected, "empty" (no BP drug present), ETID-loaded, and PAM-loaded. All three samples exhibited time-independent moduli during the course of time sweep measurements over 2000 s. Figure S-7 in the SI compares the subsequent (for each sample) linear viscoelastic spectra for the three samples. First, we note that the three hydrogels are very similar, exhibiting the same frequency dependence. The frequency-independent storage moduli (G'), much lower and frequency dependent (with a minimum) loss moduli (G") and large values of G' (order of 10 4 Pa) indicate that the gels are strong. Typically, many colloidal gels can have much lower moduli (100 or 1000 Pa), and gels with moduli above the kPa range are strong. Secondly, we do observe quantitative differences: the "empty" hydrogel is the strongest. The ETID-loaded gel is slightly softer (note the logarithmic scale), whereas the PAM-loaded gel is clearly softer, with G' being reduced by a factor of 2.5. The fact that "empty" is the strongest gel is corroborated by its lower yield strain (i.e., the fractional deformation needed to break it mechanically). It is about 5%. On the other hand, the softest PAM-loaded gel is more deformable at break, having a yield strain of about 10%, whereas the intermediate ETID-loaded gel has a yield strain of 8%.</p><p>Once BP-loaded hydrogels form they do not absorb additional water. Tests (see Figure S-34 in the SI) indicate weight differences from ~0.5% to ~3.2% in hydrogels exposed to excess water. Interestingly, dried gels do not swell. When exposed to water they do not revert to hydrogels, hence the process from hydrogel to dry powder is irreversible. They also exhibited variable water re-absorption behavior (see Figure S-35 in the SI). For example an "empty" dried gel re-absorbed only ~1.3% water. However, a dried ETID-loaded gel lost ~34.6% of its weight, whereas a dried PAM-gel gained ~12.2% of its weight.</p><p>BP-loaded gels were subjected to 48-hour controlled release experiments (each repeated at least 4 times, with excellent reproducibility, ±3% error). Release quantification was made based on 1 H NMR signals. Figure 4 shows results on controlled release of the family of BPs with hydrophobic side-chains (ETID, C3BP, and C5BP).</p><p>A clear differentiation of the controlled release profiles is evident. ETID (with the shortest methyl side-chain) exhibits the fastest release, followed by C3BP (with n-propyl side-chain). C5BP (with the longest n-pentyl side-chain) demonstrates the slowest release rate. The final % BP released in solution after 48 hours (plateau value) follows the same ranking.</p><p>Figure 5 shows results on controlled release of the family of BPs with hydrophilic, amine-containing side-chains (PAM, ALE, C4NBP, and NER). The results are quite intriguing, when compared to those for non-polar side chain BPs (Fig. 4), as they reveal that the presence of the amine group on the side chain profoundly enhances release rates and the final % plateau value. PAM, with an ethylamine side chain, exhibits the fastest release and final % release (~80%) of all. Side chain elongation in amino-BPs decelerates release and lowers the final % plateau value (eg. for C4NBP it is 70%). Nevertheless, side-chain length increase beyond 3 atoms (two C's, one N) does not induce systematic release reduction, as the results are nearly indistinguishable, thus pointing to a strong effect from the amine end-functionality surpassing that of the chain lengthening.</p><p>The release results presented in Figs 4 and 5 clearly demonstrate some important trends. Both for amino-BPs and non-polar side-chain BPs, their rates and final % release correlate with their aqueous solubility trends (see Figs S-22 and S-23) 33,34 . Interestingly, none of the BP-loaded hydrogels reaches quantitative release (eg. ETID reaches a ~75% plateau after ~24 hours). Hence, it is important to address whether the equilibrium reached after 24 h is final, and whether the remaining BP can be quantitatively delivered. Therefore, step-wise experiments were designed and carried out, in which the supernatant fluid was replaced with "fresh" aqueous medium after each release plateau was reached. The results for three step-wise release stages for the ETID-and PAM-loaded gels are shown in Fig. 6. They strongly support the conclusion that after equilibrium is "reset", BPs' release continues eventually reaching a final, essentially quantitative value. Significantly, there is no detectable BP entrapped in potentially inaccessible hydrogel pores, as indicated by the quantitative final release within 144 hours. "Empty", BP-free hydrogels were evaluated for their ability to be reloaded with drug. Thus, an "empty" (no BP) gel was prepared as "control". This gel, together with a second BP-loaded gel after its release, were exposed to an aqueous supernatant that contained the same content of ETID (as in a regular "loaded" gel) in order to assess whether the ETID will re-enter the hydrogels. Indeed, both gels absorbed ~20% of dissolved ETID. Subsequently, both gels were subjected to the usual release conditions, delivering ~60% of the absorbed ETID (Fig. S-24, SI). Both gels exhibited the same BP re-absorption behavior, thus establishing that both hydrogels ("freshly-prepared" and "used", after release) are robust and re-loadable. Although "re-loadability" may not seem to be necessary in drug/pharmaceutical applications, it may add additional functionality for environmental applications, eg. sustainable pollutant absorption from aqueous systems.</p><p>We have attempted to evaluate several hydrogel features in order to shed light on the factors controlling the BP release. It is important to note that the hydrogel content is principally water (in huge excess), in addition to other ions, such as Na + (from the pH adjustment chemicals and the BP itself). A series of otherwise identical ETID-loaded gels were fabricated, but with variable amounts of Na + ions (ie. [Na + ]:[ETID] molar ratio ranging from 2 to 6), and was found that the Na + ion concentration does not affect BP release (provided the pH is not altered), see In order to assess the influence of nature of the alkali cation present in the hydrogel, we fabricated gels in the presence of larger ionic radius K + and Rb + cations (it was impossible to prepare gels with Li + and Cs + cations). It appears that cation replacement (Na + by K + ) decelerates the release and causes lower BP final % release (Fig. S-28, SI). Unfortunately, the Rb + gel gave complicated results. It is thus unclear how precisely the alkali cations affect release. Gel release curves were generated at different temperatures revealing that the release is enhanced as T increases (Fig. S-28, SI).</p><p>In an effort to further decelerate BP release, more dense hydrogels were prepared, by changing (doubling) the sodium silicate concentration from 6.66% to 13.32%. However, the release rates and final % release plateaus were not systematic (Figure S-33). In the "dense" ETID-loaded gel a higher release rate and final release value were observed. In contrast, in the "dense" PAM-loaded gel a lower release rate and final release value were noted. This is an intriguing point that needs further systematic experimentation.</p><p>All hydrogels collapse upon water loss and eventual drying. Three representative dried gels ("empty" without BP drug present, ETID-loaded, and PAM-loaded) were used for N 2 sorption measurements in order to obtain nitrogen sorption isotherms at 77 K and BET plots (Figures S-8 and S-9). These data revealed that the specific surface areas (SSA) for the three dried gels are: 83.4 m 2 /g for the "empty gel, 41.2 m 2 /g for the ETID-loaded gel and 7.4 m 2 /g for the PAM-loaded gel. The downward trend is consistent with the BP drug partially filling the voids within the gel, a fact that is corroborated by the observation that when BP molecular size increases (from ETID to PAM) the SSA drops from 41.2 m 2 /g to 7.4 m 2 /g. The fact that upon ETID loading the SSA for the "empty" gel drops from 83.4 m 2 /g to 41.2 m 2 /g is consistent with the SEM images shown in the SI. It is interesting to note that both drug-loaded ETID and PAM dried gels exhibit a "burst" release of the active drug (>90%) within a couple hours (results not shown).</p><p>Comparison of BP release (both initial rates and final "plateau" values) between the two families of BPs with non-polar alkyl and aminoalkyl side-chains is quite revealing. Presence of the amine group profoundly enhances release. This is true for similar molecular size BPs, exhibiting "short" 3-atom, and also for "long" 5-atom chain lengths (Figs S-30 and S-31, SI). Phosphonate interactions with silica surfaces have been studied in detail 30,35 . It could be envisioned that the loaded BPs can interact with the gel internal surface through hydrogen-bonding, in which both silanol and phosphonate moieties can participate. Based on 29 Si MAS NMR spectra the Q2 + Q3% content in various gels can range from 22% to 30% (Fig. S-32, SI). In other words, there are sufficient silanol groups to form hydrogen bonds with the oxygen-rich phosphonate groups. We have also attempted to draw useful information on silica-BP interactions from comparisons between FT-IR spectra of dried gels and "authentic" (free) BPs. However, the results are not conclusive. In general, bands due to a variety of vibrational modes of BPs are present in the spectra of dried gels. The main bands assigned to the -PO 3 H − moieties (~900-1100 cm −1 ) overlap with the hydrogel's Si-O-Si bands in the same region. Hence, several fine features of the phosphonate vibrational modes disappear due to this overlap. Furthermore, the bands assigned to the R-(CH 2 ) x -NH 3 + moiety (in the regime >3000 cm −1 ) broaden and are undetectable in the dried gel spectra. Regarding the protonation state of the BPs, it is certain (from literature reports [36][37][38] ) that at pH 7 (the pH of the internal gel regions) each phosphonate moiety is mono-deprotonated (-PO 3 H − ), and (for the aminoalkyl BPs) the amine moiety is protonated (R-(CH 2 ) x -NH 3 + ). The release profiles of the alkyl-BPs and amino-BPs (with the latter demonstrating faster and higher release) are inversely proportional to the aqueous solubility of these two families of BPs. Amino-BPs are much less soluble in water than alkyl-BPs 33,34 , most likely due to the extensive intermolecular H-bonding interactions of the protonated amine moiety. However, our results reveal that they are unexpectedly delivered more readily to the aqueous phase than their alkyl-BP analogs. Assuming that they are anchored to the gel internal surfaces through their bisphosphonate moiety, they leave their protonated amino (R-NH 3 + ) moiety protruding towards the aqueous phase. Strong hydration of the -NH 3 + group augments the eventual detachment of the amino-BPs and their final release from the silica surface. Hydrogen bonds between water and the -NH 3 + group have been reported in the solid state 39 .</p><!><p>In conclusion, we have reported the fabrication, characterization and BP-release properties of silica-based gels. These gels can be prepared in a cost-effective manner from cheap reagents. They possess several attractive features such as injectability, responsiveness to temperature, re-usability, and re-loadability. Furthermore, the BP drug release profiles can be fine-tuned to achieve the desired drug release, by altering several factors, such as temperature, cations present, pH and structural features of the BPs.</p><!><p>Materials. Sodium silicate pentahydrate, Na 2 SiO 3 •5H 2 O, silicic acid (<20 micron, refined, 99.9%) and potassium hydroxide was purchased from Sigma Aldrich. Rubidium hydroxide hydrate was purchased from Alfa-Aesar. ETID (either as solid tetrasodium salt or as acid in aqueous solution) was used as received from Solutia Inc. Deuterium oxide (99.9 atom % D) that contained 0.05 wt. % 3-(trimethylsilyl)propionic-2,2,3,3-d4 acid, sodium salt (TSP) purchased also from Sigma-Aldrich. Deionized water from an ion-exchange resin was used for all experiments and stock solution preparations.</p><p>Instrumentation. Solid state NMR experiments were performed on a Bruker Avance 300 NMR spectrometer with a 7 mm MAS wide bore probe. The operating resonance frequency (rf) of 300.1 MHz for 1 H, 59.6 MHz for 29 Si and of 121.5 MHz for 31 P measurements. 7 mm Zirconia rotors with KEL-F inserts were used. For direct excitation measurements π/2 pulses at rf fields of 34.7 kHz on 29 Si and 65.8 kHz on 31 P were applied. The interscan delay was set to 180 s for 31 P and 120 s for 29 Si. During acquisition TPPM decoupling with rf field at 50.0 kHz was used. Measurements were performed at a sample spinning speed of 7 kHz. For the 1 H-31 P CP MAS experiments a ramped CP by using an rf field of 53.2 kHz for the spinlock on 31 P, an 80-100% ramp on the proton channel, and a contact time of 1.5 ms were applied. The corresponding decoupling was performed using the TPPM decoupling scheme at an rf field of 50.0 kHz during acquisition. The interscan delay was set to 5 s. An AVANCE 300 (Bruker, Karlsruhe, Germany) spectrometer was used for the BP release experiments. SEM data and images collected with a JOEL JSM-6390LV electron microscope.</p><p>Nitrogen sorption measurements. Low-pressure N 2 sorption measurements were carried out on an Autosorb 1-MP instrument from Quantachrome equipped with multiple pressure transducers for highly accurate analyses and an oil-free vacuum system. Ultra-high purity grade N 2 (99.999%) was used for all adsorption measurements. Prior to analysis, each sample was activated by heating at 80 °C under dynamic vacuum for 12 hours. After evacuation, the sample and cell were re-weighed to obtain the precise mass of the evacuated sample. Finally, the cell was transferred to the analysis port of the gas adsorption instrument.</p><p>Bisphosphonate synthesis. ALE and NER were synthesized and characterized as reported elsewhere 29 . C3BP (disodium salt) and C5BP (disodium salt) were synthesized according to the procedure described by Egorov et al. 40 . Further BP synthesis details are given in the SI.</p><!><p>Herein, the synthesis of an ETID-loaded gel is described, as an example. All other BP-loaded gels were prepared in the same manner. The synthesis of each BP-loaded gel was repeated four (4) times using identical shape and diameter borosilicate glass beakers. In a beaker 10 mL of DI water was added. In this a quantity (0.66 g, 3.14 mmol) of sodium metasilicate pentathydrate was dissolved, together with 0.50 g, 1.70 mmol) of tetrasodium ETID, while keeping the solution under stirring. The pH value of this solution was ~12.5. The pH was adjusted to 7.00 with the use 0.75 mL of concentrated HCl (37%). This particular pH value was selected because the polymerization of silicic acid has the highest rate there. Gel formation commences within 10 minutes (see video clip in the SI), however the freshly formed and "loose" gel was allowed to mature for 12 hours, after which a shapely and translucent gel formed. Gel preparation can be reproducibly repeated and can be modified by altering the amount of Na + ions, replacing the alkali ion, or changing the entrapped BP. BP-containing gels for all remaining BPs were prepared in the same manner, using quantities shown in the experimental details in the SI.</p><p>Gel rheological studies. Measurements were performed in a strain-controlled rotational rheometer (ARES 100FRTN1 from TA, USA), with stainless steel parallel plate geometry (diameter 10 mm) at 25 °C. The temperature was controlled by means of a recirculating fluids bath (water/ethylene glycol mixture). Each sample was placed between the plates with a thickness of about 1 mm. It was appropriately trimmed at the edge. To eliminate the risk of solvent (water) evaporation, the region around the samples' edge was effectively sealed with silicon oil (viscosity 5 mPa) by means of a coaxial home-made outer aluminum ring that contained it. Linear viscoelastic measurements were performed in order to characterize the different hydrogels. The protocol involved dynamic time sweeps to ensure equilibration of the samples, strain sweeps to determine linear viscoelastic limit and frequency response to obtain the spectra.</p><p>Controlled release of BPs from gels. On top of the solidified gel (see above), a volume of DI water (50 mL), pre-acidified to pH ~3 was carefully poured. This marked the initiation of the controlled release process (t = 0), which continued for 48 hours. For the initial 6-hour period an aliquot of 0.350 mL was withdrawn from the supernatant every hour. After the 6 th hour and for the next 12 hours, sampling was performed every 3 hours. Finally, after the 18 th hour and until the end of the release experiment (at the 48 th hour) sampling was performed every 8 hours. The withdrawn samples were mixed with 0.150 mL of deuterium oxide (99.9 atom % D) that contained 0.05 wt. % (4.3375 μmol) 3-(trimethylsilyl)propionic-2,2,3,3-d 4 acid, sodium salt, TSP) as standard. 1 H NMR spectra were recorded on a Bruker AVANCE 300 MHz NMR (Bruker, Karlsruhe, Germany) spectrometer at 293.2 K operating at a proton NMR frequency of 300.13 MHz. Standard solvent (D 2 O) was used as internal lock.</p><p>Each 1 H spectrum consisted of 32 scans requiring 3 min. and 39 min. acquisition time with the following parameters: Spectral width = 20.5671 ppm, pulse width (P1) = 15.000 μs, and relaxation delay (D1) = 4.000 seconds. Polynomial 4 th -order baseline correction was performed before manual integration of all NMR spectra. Proton and carbon chemical shifts in D 2 O are reported relative to TSP. The characteristic peaks for each compound were integrated using the integration tool available from the Bruker software (TopSpin 3.2). For each compound we selected the integration value of the sharpest peak. All the integration values were cross-checked in order to ensure the best result for each compound (see spectra in the SI).</p>

Scientific Reports - Nature

SITE-SATURATION MUTAGENESIS OF POSITION V117 IN OXA-1 \xce\xb2-LACTAMASE: THE EFFECT OF SIDE-CHAIN POLARITY ON ENZYME CARBOXYLATION AND SUBSTRATE TURNOVER\xe2\x80\xa0

Class D \xce\xb2-lactamases pose an emerging threat to the efficacy of \xce\xb2-lactam therapy for bacterial infections. Class D enzymes differ mechanistically from other \xce\xb2-lactamases by the presence of an active-site N-carboxylated lysine which serves as a general base to activate the serine nucleophile for attack. We have used site-saturation mutagenesis at position V117 in the class D \xce\xb2-lactamase OXA-1 to investigate how alterations in the environment around N-carboxylated K70 affect the ability of that modified residue to carry out its normal function. Minimum inhibitory concentration analysis of the twenty position 117 variants demonstrate a clear pattern of charge and polarity effects on the level of ampicillin resistance imparted on Escherichia coli (E.coli). Substitutions that introduce a negative charge (D, E) at position 117 reduce resistance to near background levels, while the positively charged K and R residues maintain the highest resistance levels of all mutants. Treatment of the acidic variants with the fluorescent penicillin BOCILLIN FL followed by SDS-PAGE shows that they are active for acylation by substrate, but deacylation-deficient. We used a novel fluorescence anisotropy assay to show that the specific charge and hydrogen-bonding potential of the residue at position 117 affects CO2 binding to K70, which in turn correlates to deacylation activity. These conclusions are discussed in light of the mechanisms proposed for both class D \xce\xb2-lactamases and BlaR \xce\xb2-lactam sensor proteins, and suggest a reason for the preponderance of asparagine at the V117-homologous position in the sensors.

site-saturation_mutagenesis_of_position_v117_in_oxa-1_\xce\xb2-lactamase:_the_effect_of_side-chain_p

<!>Mutagenesis<!>Antibiotic susceptibility<!>Detection of acyl-intermediates<!>\xce\xb2-lactamase expression and purification<!>Kinetic assays<!>Determination of CO2 affinity to OXA-1 and its variants<!>Anisotropy measurements<!>Results<!>Discussion

<p>β-lactam antibiotics provide a tremendous array of therapeutic options for infections caused by both Gram-positive and Gram-negative species. Depending on the different structural features attached to the core β-lactam ring, these drugs can be classified as penicillins, cephalosporins, carbapenems and monobactams (1). Despite this structural diversity, all β-lactams act by inhibiting the transpeptidase enzymes that cross-link the peptidoglycan cell wall. The efficacy of these agents has long been threatened by resistance mechanisms, most notably, the action of β-lactamases. These enzymes can be encoded chromosomally or carried on plasmids, and may be constitutively-expressed or inducible. In the latter case, highly-sensitive signal transduction systems have been identified in Staphylococcus aureus (2) and Bacillus licheniformis (3) for sensing the presence of β-lactam antibiotics and turning on the transcription of β-lactamase genes.</p><p>Chemically, β-lactamase enzymes hydrolyze the β-lactam bond, rendering the drug incapable of transpeptidase inhibition (4). These enzymes are subdivided by sequence homology into four classes (A-D). Classes A, C and D share a similar serine-nucleophile covalent catalysis mechanism, while the structurally unrelated class B enzymes use a Zn+2 ion to activate a water molecule for attack (5). Class D members are named OXA enzymes (ie. OXA-1, OXA-24 etc) due to an unusually high activity against the semi-synthetic penicillin oxacillin. This oxacillinase activity is not robust in all class D enzymes however, and most notably lower in many (though not all) enzymes that have a strong activity against carbapenems (6).</p><p>The active sites of class D β-lactamases contain two unusual features. They are highly hydrophobic compared to those of class A and class C enzymes, with non-polarity conserved at positions 117 (V/I/L), 160 (W) and 161 (L/I)1 among others. Second, a highly conserved lysine (K70) is modified by CO2 to form a carbamate functional group (Figure 1). These two features are related to each other, as a hydrophobic environment is thought to promote the deprotonated state of the lysine side-chain that is necessary for the attack of the lysine amine on the CO2 (7). The carbamate group, normally quite labile, is stabilized by hydrogen bonds with the side-chain of W160 and S67 (and in the case of OXA-1, S120). Functionally, it serves as a general base to activate the alcohol of S67 for attack on the substrate's lactam carbonyl (7). This action leaves the drug transiently attached to the enzyme, but the carbamate subsequently activates a water for hydrolytic deacylation. In support of this, substitutions that lead to destabilization or elimination of the carbamate greatly slow down the acylation and (even more effectively) the deacylation step (8–11).</p><p>The conservation of non-polar branched aliphatic residues at position 117 (12) contrasts to the near ubiquity of asparagine at the homologous position in class A (N132) and C (N152) β-lactamases (5, 13). The asparagine side-chain forms a hydrogen bond with the side-chain amide carbonyl of penicillin and cephalosporin substrates (14, 15). It also hydrogen bonds to carbapenem substrates, but does so through the hydroxyethyl group that takes the place of the penicillin side-chain amide (16, 17). In class D β-lactamases, the valine that is typically found at position 117 also approaches close to both the side-chain amide of penicillins and the hydroxyethyl group of carbapenems, even though it is not able to form a hydrogen bond with either (11, 18–20).</p><p>The importance of having a hydrophobic residue at position 117 in class D β-lactamases has been illuminated by studies of a structurally related but non-enzymatic protein family. The β-lactam sensor proteins such as BlaR (from Bacillus licheniformis) and MecR (from Staphyllococcus aureus) share a similar fold with OXA-1 (21–23), and use an N-carboxylated lysine as a general base for serine-nucleophile attack on β-lactam substrates (24, 25). In line with their function as initiators of signal transduction pathways, sensor proteins differ from β-lactamases in being unable to efficiently deacylate the covalent intermediate. Interestingly, sensor proteins have a polar residue (N or T) in place of the hydrophobic residue typical of position 117 in the otherwise highly similar class D β-lactamases active sites. This notable difference led to the discovery that after acylation by β-lactam substrates, the BlaR sensor's N-carboxylated lysine modification is lost by spontaneous decarboxylation (21, 25, 26). The loss of the general base that normally activates the deacylating water explains the stability of the acyl-enzyme intermediate. In support of this, Cha et al. used a chemical modification approach to replace the N-carboxylated lysine residue of BlaR with a standard carboxylate group, and turned the sensor protein into an enzyme capable of full turnover of β-lactam substrates (27). Conversely, two groups have demonstrated that substitution of polar residues for V117 in two different class D β-lactamases destabilizes the N-carboxylated lysine moiety, yielding various degrees of deacylation-deficiency (11, 20).</p><p>In order to better understand the overall mechanism of class D β-lactamases, and more specifically, the role of V117 in that mechanism, we carried out site-saturation mutagenesis at this site. By revealing the chemical characteristics required at position 117 for proper binding and turnover of various substrates, we expected to illuminate how the wild-type valine regulates carbamate formation and function. Given that substitution for the sequentially-homologous asparagine in class C β-lactamases has been shown to give major alterations in substrate specificity (28), we also aimed to determine whether similar shifts would result in class D enzymes.</p><!><p>Single amino acid mutations were generated using the PCR overlap extension method (29) using the blaOXA-1 gene subcloned into the pHSG398 vector (9). All constructs were sequenced and then retransformed into DH10B E.coli cells for further analysis.</p><!><p>Minimum inhibitory concentration (MIC) analysis was carried out as before, with minor modifications (9). A Steers replicator was used to deliver a 10 μl spot of each of the E.coli strains containing the complete V117X set on Luria Bertani agar plates at pH 7.2 in the absence or presence of 25 mM NaHCO3. The plates contained varying concentrations of one of the antibiotics to be tested (ampicillin: 1–16286 μg/ml; cefotaxime, ceftazidime, doripenem, meropenem: 0.031–32 μg/ml). Each MIC value represents the median of a minimum of three replicate assays.</p><!><p>Acyl-enzyme intermediates were detected using the fluorescent penicillin BOCILLIN FL (30). E.coli cultures (50 ml) containing plasmids expressing wild-type OXA-1, each of the 19 V117 variants, the K70D mutant or a vector control were grown overnight at 37°C. Cells were centrifuged, frozen and lysed in 50 mM NaH2PO4 pH 7.0, 1 mM EDTA, using lysozyme and DNAse I. After clarification by centrifugation, samples (21 μL) of lysate were brought to 100 μM BOCILLIN FL and incubated for 1 min. Reactions were quenched by addition of 10 μL SDS-PAGE loading buffer. Samples were subsequently separated on a 10% SDS-PAGE gel and visualized using 365 nm light. After documenting the fluorescence intensity, the steady-state expression level of OXA-1 protein in each lysate was measured by Western blot following a previous method (9). The separated proteins were transferred to polyvinylidene fluoride (PVDF) membrane, and then probed with 1 μg/ml anti-OXA-1 (9). The reactive protein bands were detected with horseradish-peroxidase-linked Protein G, and visualized by chemiluminescence.</p><!><p>Several V117 variant genes were amplified by PCR (excluding the export sequence), ligated into the NdeI and BamHI sites of pET24a, and transformed into BL21(DE3) E.coli cells. Expression and purification of the mutant proteins were carried out as described previously (9). Purified fractions (> 95% pure) were combined and concentrated using a Centricon ultrafiltration device (10,000 MWCO). Protein concentration was determined from A280 values using an extinction coefficient of 42065 M−1cm−1 (31), and stored at −80°C after snap freezing in liquid nitrogen.</p><!><p>Kinetic analysis was conducted at room temperature in a Beckman DU-800 spectrophotometer. The assays were carried out in 50 mM NaH2PO4 pH 7.4 supplemented with enough NaHCO3 to ensure full carboxylation (25 mM for wild-type, V117K, and V117T; 100 mM for V117N). Initial velocities for ampicillin hydrolysis were determined using the Δε (molar absorption coefficient) of −900 M−1cm−1 (λ =235 nm). The average of three measurements was plotted as a function of substrate concentration. KM and kcat values were determined by non-linear regression of the data to the Michealis-Menten-Henri equation.</p><!><p>Wild-type and V117 mutants were dialyzed against 50 mM sodium acetate, pH 4.5 under vacuum (4°C, > 10 hours) to decarboxylate K70. Samples of each protein were incubated for 10 minutes in degassed 50 mM NaH2PO4, pH 7.4 or the same buffer containing varying concentrations of NaHCO3 up to 100 mM. Enzyme activity was measured after adding ampicillin to these samples up to a concentration of 500 μM. CO2 concentrations were calculated using the Henderson-Hasselbach equation and a pKa adjusted for ionic strength (9). Initial velocities (vo) were measured in triplicate, fit to the following single binding site model, and then normalized to</p><!><p>Anisotropy was measured with a Photon Technology International QuantaMaster7 Fluorimeter in the T-format. BOCILLIN FL (100 nM) was added to 50 mM NaH2PO4, pH 7.4 (either degassed or supplemented with varying amounts of bicarbonate). Anisotropy was monitored over time, and then aliquots of enzyme were added to a final concentration of 400 nM (K70A) or 200 nM (V117D). After an initial rise of anisotropy that lasted ~ 2 minutes, wild-type OXA-1 was added (200 nM) to rapidly hydrolyze any unbound BOCILLIN FL. The decay in anisotropy that followed was fit to a single exponential equation to generate a deacylation time constant.</p><!><p>In order to assess the function of the nonpolar residue V117 and its impact on substrate binding and turnover in OXA-1, we carried out site-saturation mutagenesis at that position. All nineteen substitutions were prepared by PCR in a blaoxa-1 gene cassette in the plasmid pHSG-398 (9). The collection of variants, hereafter referred to as the "V117X set", were individually transformed into the E.coli strain DH10B. Minimum inhibitory concentration values were determined for a number of different penicillin, cephalosporin and carbapenem antibiotics in the absence and presence of 25 mM sodium bicarbonate.</p><p>As expected for a narrow-spectrum penicillinase, low MIC values were observed for cefotaxime (≤ 1 μg/ml), ceftazidime (≤ 0.25 μg/ml), doripenem (≤ 0.06 μg/ml) and meropenem (≤ 0.06 μg/ml) for wild-type OXA-1 and all V117X mutant strains whether or not bicarbonate was present. When challenged with ampicillin, however, the V117X set displayed a wide range of MIC values, suggesting the 117 position has a large effect on substrate binding and/or turnover (Table 1). In the absence of bicarbonate, variants possessing negatively-charged side-chains (aspartate and glutamate) showed the lowest level of resistance with MIC values near background. Conversely, substitution with the positively-charged residues arginine and lysine (but not histidine) led to the highest MIC levels observed (2048 μg/ml) aside from wild-type (8192 μg/ml). High MIC values were also noted for the other two branched chain aliphatic side-chains leucine (1024 μg/ml) and isoleucine (512 μg/ml), as well as the valine isostere threonine (2048 μg/ml). Bulky aromatics and small polar uncharged residues (except T) generally displayed relatively low MIC values, though none as weak as aspartate and glutamate. As observed before for other OXA-1 active site mutations, the addition of 25 mM bicarbonate to the MIC plates raised the MIC level of almost all V117X variants between 2–8 fold (9).</p><p>The positive effect of bicarbonate on MIC levels and the close proximity of the side-chain of V117 to the N-carboxylated K70 is consistent with previous observations that position 117 substitutions can alter the affinity of CO2 for K70. Previously, we and others have observed that class D mutants for which CO2 binds weakly are typically acylation competent, but deacylation deficient (8, 10, 20). We therefore sought to determine which of the V117X substitutions were deacylation-deficient by using the fluorescent penicillin BOCILLIN FL in an SDS-PAGE assay previously developed to observe covalent acyl-intermediates (10). Lysates from overnight cultures of each V117X variant strain, along with wild-type and vector-only controls, were treated with BOCILLIN FL. After incubation for 1 minute, the samples were separated by 10% SDS-PAGE and visualized on a UV-transilluminator. The results, shown in Figure 2, show a striking pattern of the variants that do accumulate acyl-intermediates. Most notably, the two residues which showed the lowest levels of resistance to ampicillin (aspartate and glutamate) were the most strongly labeled by BOCILLIN FL. The asparagine variant was also significantly labeled, while histidine, cysteine, glycine, methionine and serine were weakly modified. Those variants that maintained strong resistance such as lysine, arginine, leucine, isoleucine and threonine all showed no evidence of acyl-enzyme accumulation.</p><p>To ensure that the observed effects were not simply due to differences in expression levels of the various mutants, we carried out Western blot analysis on the same samples used in the BOCILLIN FL assay. While expression levels vary slightly for individual mutants, the differences are not extensive and do not correlate to the fluorescence pattern observed (Figure 2).</p><p>It appeared that polar substitutions that are neutral or negatively-charged lead to deacylation-deficient enzymes that contribute little or nothing to ampicillin resistance in E.coli. Conversely, long, positive side-chains or non-polar branched residues allow OXA-1 to complete the catalytic cycle resulting in significant ampicillin resistance. Given the proximity of position 117 to K70, a carboxylate side chain at 117 would discourage N-carboxylation of K70 by repelling the resulting negative carbamate, and stabilizing the unmodified and presumably protonated amine (21). A neutral side-chain capable of hydrogen bonding (eg. asparagine) would have a similar though possibly weaker effect. Without the active site carbamate, OXA-1 is known to be deacylation-deficient (10). A lysine or arginine at position 117, on the other hand, could possibly stabilize the carbamate, with consequent effects for maintaining higher resistance levels.</p><p>In order to determine how V117 substitutions affected the kinetic properties of OXA-1, and whether or not they affected the stability of N-carboxylated K70, several key mutants were transferred into the expression vector pET24a. OXA-1 mutants V117D, V117N, V117K and V117T were subsequently over-expressed in E.coli strain BL21(DE3) and purified by cation-exchange chromatography. These four variants, along with wild-type OXA-1, were stripped of their carbamate modifications by vacuum dialysis in 50 mM sodium acetate, pH 4.5. This treatment reduced wild-type OXA-1, V117K and V117T ampicillin hydrolysis activity by ≥ 70%, while the V117N and V117D variants were devoid of ampicillinase activity. Titration of bicarbonate at various levels up to 100 mM was used to restore CO2 concentrations, and resulted in a hyperbolic increase in ampicillinase activity for all mutants except V117D (Figure 3). Kd values determined from these plots indicate that CO2 affinity was essentially identical for wild-type and V117K (0.019 mM and 0.020 mM, respectively), slightly reduced for V117T (0.044 mM), and highly reduced for V117N (4.0 mM). Ampicillin KM and kcat values were determined for each variant in the presence of saturating amounts of bicarbonate. Interestingly, KM values remained nearly unchanged for wild-type (30 ± 3 μM), V117K (20 ± 1 μM), V117T (17 ± 4 μM) and V117N (12 ± 1 μM), suggesting that mutation of that position does not affect binding of ampicillin. Values of kcat, however, were greatly reduced even in the presence of bicarbonate, from 564 ± 14 s−1 for wild-type to as low as 4.5 ± 0.1 s−1 for V117N.</p><p>We have previously shown that the loss of N-carboxylation on K70 eliminates catalytic turnover in OXA-1, and that this effect was due to a disproportionate reduction of the deacylation rate (10). We therefore sought to measure deacylation rates for position 117 mutants. Previously, we developed two independent assays to measure deacylation rates. In the first, SDS-PAGE was used to detect covalent acyl-enzyme intermediates between the enzyme and the fluorescent penicillin BOCILLIN FL. While quantitation of the fluorescence signal over time allowed determination of exponential deacylation rates, this assay suffered from low sensitivity and a lack of time-resolution. Deacylation could also be measured by following the quench of OXA-1 tryptophan fluorescence by the dye ligand Cibacron Blue 3GA as the hydrolyzed β-lactam product left the active site (10). This assay greatly increased the time-resolution for measuring deacylation rates, but was dependent on the unusually high environmental-sensitivity of the class D β-lactamase's tryptophan fluorescence to Cibacron Blue. In order to overcome some of these drawbacks, we developed a novel deacylation assay based on the fluorescence anisotropy of BOCILLIN FL. Anisotropy serves as a proxy of the rotational mobility of a fluorescent compound, and is therefore expected to vary markedly as the small BOCILLIN FL molecule binds to and later releases from the large β-lactamase proteins.</p><p>To test this system, we first used OXA-1 K70A, a variant that is known to form highly stable acyl-enzyme intermediates (10). Samples of K70A and BOCILLIN FL were mixed and incubated for 5 min to allow binding and acylation to occur. Deacylation was initiated by the addition of wild-type OXA-1, which rapidly breaks down any free BOCILLIN FL. As shown in Figure 4 (top panel), these complexes had high anisotropy values that decayed over time. This decay fit well to a single exponential curve with a time constant of 0.0042 ± 0.0001 s−1 (white overlay, with residuals shown on the x-axis). That this decay rate is equivalent to the deacylation rate is supported by several lines of evidence. First, the time constant matched closely to that determined previously for this mutant and BOCILLIN FL in the SDS-PAGE assay (0.0056 ± 0.0004 s−1), an assay that presumably measures the deacylation event itself rather than product release (10). Second, the rate of acylation was strongly accelerated by the addition of propionate, which we previously showed to be able to act as a small molecule complement of the N-carboxylated lysine moiety that is missing in this mutant (Figure 4, lower panel). Lastly, the addition of wild-type OXA-1 prior to the addition K70A to BOCILLIN FL, almost completely eliminated the higher anisotropy values observed (Figure 4, lower panel).</p><p>The addition of OXA-1 V117D to BOCILLIN FL resulted in a similarly strong increase in the anisotropy of the β-lactam (Figure 5). If this assay was carried out in buffer that had been degassed by vacuum to remove residual CO2, deacylation was almost non-existent (Figure 5, top panel). If the assay was instead carried out in the presence of 100 mM sodium bicarbonate, the deacylation rate increased to 0.017 ± 0.001 s−1. Interestingly, the rate of the rise in anisotropy at the beginning of the assay (presumably some combination of BOCILLIN FL binding and acylation) is also increased by the presence of bicarbonate. Bicarbonate supplementation also leads to a significant increase in the rate of acylation of the MecR sensor protein from S. aureus (32). In order to further probe the role of polar residues in the active site of the BlaR sensors (which have either an asparagine or a threonine at this position), the anisotropy assay was repeated using OXA-1 V117N and V117T. As expected, BOCILLIN FL deacylation rates on V117N varied with CO2 concentration (Figure 6), with a time constant ranging from 0.00038 ± 0.0001 s−1 for degassed buffer to 0.035 ± 0.001 s−1 for 50 mM NaHCO3 (2.8 mM CO2). No significant rise in anisotropy was observed with V117T, suggesting that the deacylation rate on this enzyme is too fast to isolate the acyl-intermediate on the time-scale of the anisotropy assay (data not shown).</p><!><p>The high degree of non-polar residue conservation observed at position 117 in class D β-lactamases, in contrast to the polar residues found in class A or C, suggests that this position plays an important role in the unique mechanism known to operate in class D enzymes. The results presented in this paper strongly support the hypothesis that the presence of a non-polar residue at position 117 facilitates the formation of the carbamate general base formed when K70 reacts with CO2. The fact that isoleucine and leucine substitutions at this position yield highly functional enzymes fits this assertion well. The strong dependence of activity on the charge at position 117, with negative substitutions having a deleterious effect, and positive substitutions yielding strong activity, gives further insight into the nature of this interaction. The full negative charge of glutamate or aspartate is incompatible with the presence of the carbamate anion, as observed crystallographically with OXA-24 (11). Our bicarbonate titrations demonstrate that the full positive charge of lysine or arginine maintains the very high affinity of CO2 for K70, as would be expected for a stabilizing ionic force. It should be noted, however, that the kcat of the V117K mutant is reduced significantly compared to wild-type, even under saturating bicarbonate conditions. It is likely that the full positive charge of the side-chain in this case decreases the basicity of the carbamate, and thereby reduces its activity as a general base. A similar effect may be responsible for the decreased activity of V117T in OXA-10, a variant in which the threonine side-chain was shown to be hydrogen-bonded to the carbamate (20).</p><p>The identification of several V117 substitutions that result in deacylation-deficiency is broadly useful. The study of β-lactamase/substrate complexes is often dependent on the use of variants that arrest the catalytic mechanism at some mid-point. Indeed, knowledge gained from this study has already led to the structure determination of a class D enzyme (OXA-24 V130D) with the carbapenem doripenem bound as an acyl-enzyme intermediate (11). Such variants that result in loss of carbamate stability may prove to be a more subtle method to achieve deacylation-deficiency than mutation of the N-carboxylated lysine itself.</p><p>When the position homologous to residue V117 is mutated in class A or class C β-lactamases, large changes in substrate specificity are observed (28). It is therefore notable that we did not see any gain of function phenotypes with respect to 3rd generation cephalosporins or carbapenems. It is possible that the reversible nature of the lysine N-carboxylation modification, and the importance of residue 117 in its formation, precludes the types of substitutions that are necessary for changes in substrate binding affinities or turnover rates. A similar situation exists with regard to the residue immediately preceding the catalytic serine, which when mutated yields advanced generation cephalosporin resistance in a class A lactamase (M69) (33). Mutation of the same residue in OXA-1 (D66) results in destabilization of the N-carboxylated K70, and only minor gains in hydrolysis of cefepime and cefotaxime (9). These observations suggest that the carbamate moiety may serve as an "Achilles's heel" for class D β-lactamases—critical for activity, but labile enough that the kind of substitutions necessary for substrate plasticity cripple the enzyme's basic turnover mechanism. In light of this argument, it is interesting to note that point mutations leading to activity against extended spectrum cephalosporins are rare in seven of the nine class D subfamilies (OXA-10 and OXA-2 subclasses being the exceptions) (12).</p><p>The results observed here for polar substitutions of V117 are particularly helpful in understanding how the presence of asparagine or threonine at the homologous position in the BlaR and MecR β-lactam sensor proteins contribute to the deacylation-deficiency that is necessary for their function. It has been proposed that the hydrogen-bonding potential of these polar side-chains in some way induces the post-acylation decarboxylation of the sensors, thus preventing activation of the deacylating water (21, 22). The deacylation-deficiency observed for OXA-1 V117N is consistent with this hypothesis, as is the greatly reduced affinity for CO2 observed for this variant. It is possible that the presence of the amide side-chain of asparagine favors the carbamate nitrogen acting as general base, which leads to a barrier-less decarboxylation event (21, 27). Alternatively, a conformational change that occurs upon acylation could further decrease CO2 affinity, as seen when moxolactam occupies the active site of OXA-10 (20). In either case, the asparagine likely hydrogen-bonds to the decarboxylated amine of K70 as observed for BlaR1 (25).</p><p>A recent report by Kumarasiri et al. describes the results of a similar but converse mutagenesis experiment to what we have reported here. In this study, they introduced an N439V mutation into the BlaR1 sensor protein of Staphylococcus aureus and demonstrated that this variant is cable of full turnover of cephalosporin substrates (34). The presence of the hydrophobic valine side-chain precludes the hydrogen bond that normally forms between the asparagine and the N-carboxylated lysine of BlaR1, and thereby increases the ability of the latter to activate a water for deacylation. It is interesting to note that neither of these two reciprocal substitutions is enough to fully switch the functional profile of the two proteins: BlaR1 N439V mutant gains β-lactamase activity towards cephalosporin substrates (but not penicillins), and the OXA-1 V117N variant, unlike a sensor protein, maintains significant penicillinase activity in the presence of modest levels of bicarbonate. Clearly other amino acid differences between the two protein families will need to be identified to explain their different rates of deacylation. Wilke et al. suggest that the leucine at position 389 in BlaR1 may further destabilize the carbamate moiety and note the presence of polar residues in the spatially-equivalent areas of class D β-lactamases (eg. N73 in OXA-10, S120 in OXA-1) (22).</p><p>It is interesting to note that various polar substitutions at position 117 in OXA-1 yield vastly different biochemical properties, with a rank order of T > N > D for both CO2 affinity and deacylation rate. If deacylation-deficiency is central to the function of β-lactam sensor proteins, why is aspartate never observed at the homologous position in that family? One possibility is that the full negative charge of that residue's side-chain slows down both the deacylation and acylation reaction, as we observed for the V117D mutant of OXA-1. The loss of sensitivity caused by such a slower acylation rate may offset the longer signal life-time achieved with the aspartate. A more important question is, how does the BlaR of Bacillus licheniformis achieve deacylation-deficiency with a threonine at this position? V117T mutations in two different class D enzymes (OXA-10 and OXA-1), while yielding slight decreases in CO2 affinity, maintain strong turnover rates under physiological conditions. The most likely explanation is that the B. licheniformis protein may have additional active site features that augment the slight carbamate-destabilizing effect of threonine. Cha et al. note that different sensor proteins can display vastly different acylation rates, and thus subtle differences in active site residue choice may also simply reflect how different sensor proteins are "tuned" to advantage under particular physiological conditions (32). Ultimately, the more striking deacylation-deficiency displayed by V117N over V117T is probably the reason that the majority of sensor proteins have asparagine at this position.</p><p>Lastly, we would like to note that our anisotropy assay represents a great advance for measuring deacylation rates for β-lactamase enzymes. The assay has much higher time-resolution and is less cumbersome than the SDS-PAGE-based assays previously used for the detection of BOCILLIN FL acyl-intermediates (10). The assay will also be highly suitable for determining the typically slow deacylation rates of penicillin-binding proteins, but could also be adapted to stopped-flow measurements for the faster rates of wild-type β-lactamases.</p>

PubMed Author Manuscript

From Raman to SESORRS: moving deeper into cancer detection and treatment monitoring

Raman spectroscopy is a non-invasive technique that allows specific chemical information to be obtained from various types of sample. The detailed molecular information that is present in Raman spectra permits monitoring of biochemical changes that occur in diseases, such as cancer, and can be used for the early detection and diagnosis of the disease, for monitoring treatment, and to distinguish between cancerous and non-cancerous biological samples. Several techniques have been developed to enhance the capabilities of Raman spectroscopy by improving detection sensitivity, reducing imaging times and increasing the potential applicability for in vivo analysis. The different Raman techniques each have their own advantages that can accommodate the alternative detection formats, allowing the techniques to be applied in several ways for the detection and diagnosis of cancer. This feature article discusses the various forms of Raman spectroscopy, how they have been applied for cancer detection, and the adaptation of the techniques towards their use for in vivo cancer detection and in clinical diagnostics. Despite the advances in Raman spectroscopy, the clinical application of the technique is still limited and certain challenges must be overcome to enable clinical translation. We provide an outlook on the future of the techniques in this area and what we believe is required to allow the potential of Raman spectroscopy to be achieved for clinical cancer diagnostics.

from_raman_to_sesorrs:_moving_deeper_into_cancer_detection_and_treatment_monitoring

Introduction<!>Raman imaging for cancer detection<!>Lipid imaging<!>Alkyne imaging<!>Biomarker detection using SERS-based assays<!>Cancer cell imaging using SERS nanotags<!>pH sensing and imaging<!>Biomarker detection and imaging in cells<!>Raman and SERS analysis during clinical investigation and cancer surgery<!>Gaining depth in cancer detection via SORS and SESORS<!>Raman in the clinic<!>Conclusions and outlook<!>Conflicts of interest

<p>Cancer is the second most common cause of death globally, accounting for an estimated 9.6 million deaths in 2018.1 It results from the abnormal proliferation of normal cells in a multi-stage process, resulting in malignant tumours that can invade other parts of the body.2 Physical, chemical and biological carcinogens are responsible for the onset of the disease, with prevalence increasing with age as risk factors grow and cellular repair mechanisms become less effective. Therapeutics include invasive surgery, chemotherapy and radiotherapy, which can be ineffective in the later stages. Too often, similar symptoms are observed in benign and malignant cases and end with a low positive predictive value (PPV). However, a symptom combined with a positive test result or a relative clinical finding increases the PPV and shortens the time interval between consultation and treatment.3 Despite significant advances in recent years, the early diagnosis and treatment of cancer remains a challenge in medicine. Due to the worldwide prevalence of the disease, and the resulting mortality rates, early detection of cancer is of utmost importance to improve prognosis and patient survival. Non-invasive strategies for early detection, diagnosis and treatment monitoring are therefore urgently needed, and significant progress has been made using Raman spectroscopy and associated enhancement techniques to address these needs.</p><p>Raman scattering is a non-invasive technique that has the ability to specifically determine the chemical composition of samples based on the inelastic scattering of light by molecules.4 This molecular "fingerprinting" can be used for the sensitive and specific detection of biochemical changes that occur in diseases and is therefore a useful tool for cancer detection, diagnosis and treatment monitoring.5,6 Raman spectroscopy in cancer diagnostics has investigated a multitude of different cancer types including lung,7 cervix,8 breast,9 prostate,10 lymph nodes,11 esophagus,12 colon,13 larynx,14 bladder15 and brain.16 It has also been used to distinguish between cancerous and non-cancerous samples in ex vivo biopsies,17–19in vitro biomarker detection,20–22 and in vivo analysis.16,23–26 Advancements in instrumentation, such as the development of Raman microscopy, have allowed the technique to be used to produce high resolution chemical images of a sample by collecting spectra across several points of a defined area. These can then be constructed into false colour images using relative intensities of Raman peaks, or specific spectral regions of certain components, allowing the visualisation of changes in the sample based on their chemical properties.27 This has been exploited extensively for the label-free detection of biochemical changes in cell and tissue samples.28–31 One of the major drawbacks of Raman scattering is that signals are inherently weak due to the small proportion of photons that are inelastically scattered. This often results in poor signal to noise, reducing the sensitivity of the technique. The spectra are also complicated and often require further multivariate analysis techniques to deconvolute the data.</p><p>One method of improving the sensitivity of Raman scattering is the use of non-linear Raman techniques, such as coherent anti-Stokes Raman scattering (CARS)32,33 or stimulated Raman scattering (SRS).34 CARS and SRS are multiphoton systems where two excitation lasers, the "pump" and the "Stokes", are used to excite specific vibrational modes within a sample. In CARS imaging, one laser frequency is fixed (νS) and the other (νP) is tuned to excite a specific molecular vibration. The interaction of the two laser beams results in anti-Stokes photons of frequency νAS = 2νP − νS and the corresponding anti-Stokes signal is detected to produce a CARS image. However, CARS suffers from a non-resonant background that interferes with the resonant vibrational signal and reduces image contrast, leading to distorted line shapes that decrease the amount of chemical information, resulting in data that is difficult to interpret.35</p><p>In SRS, where the difference in frequency between the "pump" and "Stokes" photons matches the frequency of a molecular vibration (νvib = νP − νS), excitation of the vibration is stimulated and small beam intensity changes can be detected to produce images at the selected frequency (νvib). The SRS signal of a molecular species is linearly proportional to its concentration, whereas in CARS it is proportional to the square of the concentration and the laser power. Therefore, SRS has greater potential to be a powerful method for label-free quantitative determination of individual species in a multi-component system. SRS images are also free from non-resonant background and the spectra obtained match the Raman spectra of the sample, making the chemical data easily interpretable.36</p><p>Non-linear Raman techniques offer greater spatial resolution and rapid imaging times in comparison to conventional Raman spectroscopy. Generally, CARS and SRS are used to study cellular components such as lipids, DNA and proteins, and the techniques can be used to obtain detailed images, where cell structure and morphology can be examined. CARS and SRS have been employed to investigate cellular changes for the detection and diagnosis of cancer,37–41 and to monitor the uptake of drugs by cancer cells.42 SRS is also used to image small molecules coupled to vibrational tags such as nitriles or alkynes.43 Alkynes are preferred due to the C <svg xmlns="http://www.w3.org/2000/svg" version="1.0" width="23.636364pt" height="16.000000pt" viewBox="0 0 23.636364 16.000000" preserveAspectRatio="xMidYMid meet"><metadata> Created by potrace 1.16, written by Peter Selinger 2001-2019 </metadata><g transform="translate(1.000000,15.000000) scale(0.015909,-0.015909)" fill="currentColor" stroke="none"><path d="M80 600 l0 -40 600 0 600 0 0 40 0 40 -600 0 -600 0 0 -40z M80 440 l0 -40 600 0 600 0 0 40 0 40 -600 0 -600 0 0 -40z M80 280 l0 -40 600 0 600 0 0 40 0 40 -600 0 -600 0 0 -40z"/></g></svg> C stretching motion that exhibits a substantial change in polarisability, producing a sharp Raman peak in the cell silent region.44 Proteins, DNA and phospholipids have all been tagged with alkynes, introduced into cells and imaged using SRS, offering superb sensitivity, specificity and the biocompatibility required to study complex living systems.45,46</p><p>Further enhancements of weak Raman signals can be achieved by introducing a roughened metal surface. This phenomenon, known as surface enhanced Raman scattering (SERS), occurs when a molecule is adsorbed onto, or held in close proximity to, an enhancing metal surface.47,48 SERS enhancement is a result of the interaction of light with plasmons excited at the surface of the metal, which has been shown to enhance Raman signals up to 1010.49 Nanoparticles of noble metals (most commonly gold and silver) are used as SERS substrates due to their unique optical properties and adaptable synthesis allowing control over size, shape, and morphology, which can be tailored towards diagnostic applications. SERS-active nanoparticles can either be used in a label free (direct) capacity, where the intrinsic scattering from a biomolecule of interest adsorbed onto a nanoparticle surface is obtained, or for labelled (indirect) detection, which is achieved when Raman reporters are added to the nanoparticle surface to create SERS nanotags that can be used to indirectly detect biomolecules.50 Further signal enhancement can also be achieved when the Raman reporter is a chromophore with an electronic transition close in energy to the exciting laser. This increased enhancement is known as surface enhanced resonance Raman scattering (SERRS), which has been reported to increase signals up to 1014.51 SE(R)RS nanotags can also have targeting capabilities by functionalising them with biomolecules, offering further potential for in vivo applications.52 The development of SERS is therefore a significant expansion in the capabilities of Raman spectroscopy for bioanalytical applications and, in particular, for cancer diagnostics and in monitoring the treatment of cancer.53–57</p><p>An additional advantage of SERS is that, due to the sharp peaks present in Raman spectra, the technique is capable of detecting multiple targets simultaneously.58–60 An early example of multiplexing was demonstrated by Faulds et al. who were able to detect 6 DNA sequences corresponding to different strains of the Escherichia coli bacterium that were labelled with different commercially available dye labels.61 A SERS-based assay was also developed for the multiplexed detection and quantification of three bacterial meningitis pathogens with picomolar detection limits,62 and for genotyping human papilloma virus (HPV) from plasmid, cell line and clinical material with the ability to differentiate between six HPV genotypes.63 Furthermore, the simultaneous isolation and detection of three different bacterial pathogens has been achieved using SERS nanotags functionalised with antibodies specific to each target, demonstrating the capability of SERS for providing rapid and sensitive discrimination from a single sample.64 This shows the potential of the technique for advancements in biomedical applications and in future point of care devices, such as lateral flow immunoassays.65 Multiplexed detection using SERS has also been extended to cancer biomarkers.66–69 This includes the sensitive and simultaneous detection of multiple microRNAs associated with lung and breast cancer for the early diagnosis of the disease.70–72 This signifies the capabilities of the technique for cancer detection and diagnosis, where the simultaneous detection of multiple biomarkers is a significant advantage.</p><p>Despite the sensitivity of SERS and its multiplexing capabilities, along with the non-invasive and molecularly specific nature of Raman scattering, Raman is limited by its depth penetration capability and spectra of tissue are dominated by contributions from the subsurface layers, limiting the clinical application of Raman spectroscopy. However, since the transmission of light through tissue is dependent on the wavelength, for example light with a wavelength of 440 nm can only penetrate around 1 mm compared to 5 mm for near infrared (NIR) wavelengths (750 nm),73 the use of longer excitation wavelengths has improved the depth penetration of the technique. Confocal and purpose-designed instrumentation has also increased the applicability of the methods in vivo.74,75 Despite these improvements, it remains challenging to obtain spectral information from below the surface of the skin, without recourse to more invasive approaches such as needle probes.76 By applying an offset between the excitation and collection probes in a Raman experiment, photons scattered below the surface of the sample can be collected. This method, known as spatially offset Raman scattering (SORS),77 allows the collection of Raman spectra from depths significantly greater than those achievable using traditional confocal Raman microscopes, thus improving the potential of the technique for clinical applications.78 This has been validated by demonstrating the non-invasive analysis of bone79–81 and cancer tissue samples.9,82 By introducing nanoparticles into SORS experiments, the depth penetration capabilities of SORS can be combined with the sensitivity of SERS to allow Raman signals to be obtained from significantly increased depths through biological tissues. This alternative approach, known as surface enhanced spatially offset Raman scattering (SESORS), has allowed collection of Raman spectra from depths of around 5 cm through tissue samples.83 The advantages of SESORS for clinical applications have been demonstrated for glucose monitoring,84 detection of neurotransmitters through the skull,85,86 and for in vivo cancer imaging in live mice.87 The capabilities of SESORS for non-invasive detection in vivo is a significant step towards the application of Raman spectroscopy for the clinical diagnosis of cancer and in monitoring the effectiveness of treatment. This is a further demonstration of the versatility of Raman spectroscopy techniques and their potential in medical diagnostics.</p><p>One of the limiting factors in the clinical application of Raman spectroscopy, particularly SERS, is that a standard method is yet to be adopted and results can sometimes be considered irreproducible. Large inter-laboratory studies have recently been undertaken in an effort to overcome these issues,88,89 and recommendations have been published on the key parameters that should be considered to improve comparability of results across laboratories.90 These considerations are essential for clinical translation of the techniques and collaborative studies should continue such that standardised methods can be developed. Additionally, further use of SERS alongside clinical trials is required to prove the capabilities of the technique for cancer detection and diagnosis and so that the full potential of the technique can be realised.91</p><p>This feature article discusses the use of Raman spectroscopy for the detection, diagnosis and treatment monitoring of cancer and the progress of the technique towards clinical application, highlighting the research of our group in this area. The versatility of Raman spectroscopy allows the application of the technique in its various forms to the many approaches of studying cancer, from cellular imaging and biomarker detection to in vivo analysis. Here we discuss some of these approaches, demonstrating advances in Raman spectroscopy that provide benefits for the different methods and improve the potential of the technique for the detection, diagnosis and monitoring of cancer.</p><!><p>To gain insight into the biochemistry of a cancer cell, cellular components can be identified using molecular biology-based approaches such as polymerase chain reaction, electrophoresis and Western blotting.92,93 They offer high levels of chemically specific information but require the cell to be lysed, which can introduce chemical modifications to the results. An attractive alternative is to Raman image cells to provide rapid, non-invasive and high spatial resolution of biochemical and structural information. However, as explored by Butler et al.,94 careful consideration of sample preparation, instrumentation, acquisition parameters and data processing must be taken into account in order to produce high quality data for analysis of biological material. Raman imaging has been used extensively to investigate biological changes. These include classification of different types of liver cancer and their proliferation states,95 investigating the uptake, distribution and metabolism of drugs in colon cancer cells,96 and to help understand the response in cancer cells when exposed to ionising radiation.97</p><!><p>Lipids are an important cellular component whose intracellular uptake, distribution and metabolism are tightly regulated in healthy cells. However, these processes are disrupted in cancer due to the upregulation of de novo lipid syntheses.98 In order to develop new treatments, it is vital that the lipid biochemistry is understood. Raman analysis of prostate and bladder tissues indicated that the relationship between lipids and carcinogenesis could be measured.99 High resolution cellular Raman imaging built on this significantly by showing where these changes to cell biochemistry occurred, and it has been shown to give a detailed, high resolution insight into lipid distribution in cancer cells.28,100 Conventionally, the Raman peak intensities of the lipids is used to create Raman images, however recently ratiometric analysis of Raman peaks from cellular information in the fingerprint region has been shown to reflect lipid/protein abundance across a HEK293T cell.101 The ratiometric images were generated using the intensity ratio of 1448 cm−1 (which is associated with long aliphatic chains present in lipid species) divided by the sum of 1657 cm−1 (amide 1 vibrations) and 1448 cm−1. The images highlighted the nuclear region with lower lipid/protein content compared to the cytoplasmic region, reflective of the nuclear function to store DNA. This work was advanced by Jamieson et al. who used ratiometric values to build Raman images of intracellular lipid distribution of cancerous (PC3) and non-cancerous (PNT2) prostate cells, treated with drugs known to inhibit the enzymes involved in de novo lipid synthesis.30 To create a bivariate descriptor, the ratio between the high wavenumber region, 2851 cm−1 (C–H stretch in CH2 groups) and 2933 cm−1 (C–H stretch in CH3 groups), was selected to correlate lipid abundance. The false colour images of intracellular lipid distribution, shown in Fig. 1A, revealed a difference between cancerous and non-cancerous cells and a uniform distribution of lipids throughout the cytoplasm in PC3 cells compared to lower levels of lipid for PNT2 cells. The cells were then treated with three drugs that interfere with the different stages of de novo lipid synthesis and ratiometric Raman images of the lipid distribution created. Orlistat, an inhibitor of fatty acid synthase, elicited a phenotypic response characteristic of lipid accumulation in both cell lines. CAY10566, an inhibitor of the enzymes stearoyl-CoS desaturase (SCD) which creates mono-unsaturated fatty acids from saturated fatty acids, gave little response. Finally, 5-(tetradecyloxy)-2-furoic acid (TOFA), which inhibits the conversion of acetyl CoA to malonylCoA (one of the first steps in de novo lipid synthesis) induced a decrease in lipids, particularly in the cancer cells. The effect of two control drugs, cyclosporin and propranolol, which are capable of inducing the formation of lipid droplets were also investigated. Interestingly, propranolol showed selectivity towards cancerous cells, indicating it was a strong candidate to be investigated for selective anti-tumour action. It is clear that this non-destructive label free ratiometric analysis, performed using cost effective glass substrates, could revolutionise the understanding of drug–cell response and is a significant step in the monitoring of cancer treatment.</p><!><p>An easily exploited region in the Raman spectra of cells is the 'cell silent' region (1800–2600 cm−1). Designing molecules that give a Raman band in this region can allow their distribution in cells to be easily tracked without any interference from the Raman signal of cellular components. Alkyne tags have become an important functional group as they give a strong band in the silent region (roughly 2120 cm−1).104 The first example of alkyne detection in cells was achieved using EdU (5-ethynyl-2′-deoxyuridine), a thymidine analogue with an alkyne group.46 EdU is incorporated into cellular DNA during DNA replication and accumulates in the nucleus. Due to the presence of the alkyne group, its location can be imaged using Raman mapping, demonstrating the potential of the alkyne moiety as a Raman tag in live-cell imaging of small molecules. The distribution of fatty acids tagged with an alkyne group has also been monitored and relative quantification was achieved, demonstrating how minimally invasive this technique is and how a small Raman tag can produce a large response.101 Recently, we have measured intracellular pH in prostate cancer cells (PC3) by designing low molecular weight oligoyne compounds that exhibit a pH sensitive alkyne stretching frequency.102 To quantitatively determine the pH, calibration within the environment of interest was performed. PC3 cells were treated with the compound and fixed to a discrete pH value. The cells were then Raman mapped and false colour images created using the ratio of the signals at 2221/2210 cm−1 (the bands corresponding to the change in alkyne shift at different pHs). The results from this approach are shown in Fig. 1B. The ratio varied as a function of pH in the cells and the compound was then used to monitor and quantify changes in pH in response to drug treatments. Cells were treated with etoposide, which induced apoptosis and should coincide with a decrease in pH. Over time, the change in ratio indicated that the pH decreased, demonstrating that the compound could effectively monitor and quantify changes in pH of live cells in response to drug treatment. To improve spatial resolution, the live PC3 cells were also imaged using SRS microscopy by selecting 2933 cm−1 (protein), 2951 cm−1 (lipid), 2221 cm−1 (alkyne) and 2321 cm−1 (off resonance) channels. The 2221 cm−1 channel confirmed that the alkyne was distributed in the cytoplasm, demonstrating the compatibility of the probe for intracellular pH sensing. This highlights another approach that could be used for the monitoring of cancer treatment and in enhancing understanding of the disease.</p><!><p>The detection of cancer biomarkers in body fluids overcomes the need for more invasive procedures, such as tissue biopsies. In addition, biomarker detection is more sensitive and specific than traditional morphological characterisation, thus potentially allowing detection of cancer at an earlier stage, increasing the PPV and therefore improving patient prognosis.105 For biomarker detection, SERS offers greater sensitivity than competing techniques and can also be used to detect multiple biomarkers simultaneously, allowing more accurate classification of cancer. With these advantages, it is unsurprising that SERS has been widely studied for the detection of cancer biomarkers and that various approaches have been explored.106–108 Early research in our group demonstrated that nanoparticles functionalised with biomolecules could be used to significantly enhance SERS signals by causing controlled aggregation of nanoparticles following specific biomolecular interactions.109 This nanoparticle assembly approach can be applied to various biomolecules and has thus been exploited extensively for the development of biological detection assays.110–114 Additionally, this method can be used to study biomolecular interactions, yielding significant information that may be useful in understanding cancer pathways. For example, the tumour suppressor protein, p53, plays a key role in many cancers and is regulated by mouse double minute (MDM2) protein. Therefore, understanding the interaction between these proteins could be invaluable in cancer therapeutics. Using a nanoparticle assembly approach with SERS detection, MDM2 interactions were studied in solution, allowing monitoring of the full protein, rather than focusing on only one binding interaction (Fig. 2).115 A p53-mimicking peptide was used to demonstrate the state of MDM2 in solution, while maintaining the biological activity of the protein. This approach validated the ability of SERS to study interactions of full length, unlabelled proteins using biologically driven nanoparticle assemblies, potentially aiding the understanding of biological pathways in diseases such as cancer.</p><p>SERS-based sandwich assays, which use capture antibodies bound to a surface and detection antibodies functionalised to a SERS nanotag, have also been explored in the group to detect low concentrations of clinically relevant biomolecules.116 This format has been exploited to detect cancer biomarkers including the detection of MUC4 expressed in pancreatic cancer,117 as well as the multiplexed detection of breast cancer118 and prostate cancer biomarkers.117,119 Cheng et al. used a SERS-based immunoassay for the simultaneous detection of two prostate specific antigen (PSA) markers and demonstrated the sensitive and specific detection of the biomarkers in clinical serum samples.120 They highlighted the potential applicability of the SERS-based assay for prostate cancer detection by comparing its performance to a current diagnostic assay.</p><!><p>Another popular approach of investigating cancer is to 'tag' biomarkers found on cancer cells followed by optical imaging. Conventionally, fluorescence tags are used as imaging agents; however, nanoparticles offer an attractive alternative due to their photostability, multiplexing capabilities, high spatial resolution, low background and enhanced sensitivity.52,121 By targeting cancer cells with SERS nanotags, cells can be analysed using Raman mapping experiments and the resulting SERS images can be used to differentiate between disease states, detect biomarkers on or within the cell, and assess the effectiveness of treatments.</p><p>Early cancer cellular nanoparticle incubation studies combined with SERS imaging did not target specific events, but focused on bare nanoparticle uptake via endocytosis. The cells were mapped and the resulting SERS spectra were indicative of changes in the chemical environment of the cell.122 In this label free approach, the spectra were complex and the analysis could be simplified by the inclusion of Raman active stains that allowed for faster mapping times. An example of this was demonstrated by Stokes et al., who incubated bone-marrow-derived cells (macrophages) with gold and silver nanoparticles. The cells were then fixed, treated with a dye stain and analysed with line scanning SE(R)RS using biologically active wavelengths. Based on the SE(R)RS images produced by following a major peak of the dye throughout the cell, nanoparticle aggregates could be identified in secondary lysosomes.123 Raman signals of the dye were significantly enhanced due to their close proximity to the nanoparticle surface. However, it should be noted that the signal was only observed in locations where both the dye and the nanoparticle coincided within the cell and was not a true reflection of all the nanoparticles taken up by the cell. To increase the sensitivity of the approach SERS nanotags have been incubated with cells.124–126</p><p>A common predicament in nanoparticle incubation studies combined with SERS imaging is the question of whether the nanoparticles are actually inside the cell or merely bound to the cell surface. To address this, McAughtrie et al. demonstrated the first example of 3D SERS imaging for the simultaneous confirmation of the cellular inclusion and multiple component detection of SERS nanotags.127 Four SERS nanotags, labelled with different thiol-based Raman reporters, aggregated using 1,6-hexamethylenediamine (HMD) to create hotspots, were added to Chinese hamster ovary (CHO) cells. To verify nanotag uptake, the cells were 3D volume Raman mapped and 3D false colour SERS images were constructed by performing multivariate data analysis in the form of direct classical least squares (DCLS). Three out of four nanotags were located within the cells with spatial positioning. To employ SERS tags in cancer detection it is also important to assess their interaction and toxicity in cells. Bhamidipati et al. evaluated the toxicity of gold nanoparticles with different morphologies and surface chemistries and demonstrated that the surface chemistry had the predominant effects on cytoxicity, and that cetrimonium bromide (CTAB) coated gold nanoparticles were the most toxic and polyethylene glycol (PEG) coated gold nanoparticles the least.128</p><p>As well as inferring the location of SERS nanotags in cells, the Raman reporter can be used to investigate a variety of mechanisms that occur within the cell. For example, the activity of beta-galactose, a biomarker overexpressed in cancer, was detected in macrophages by monitoring the change in SERS signal that occurs when the reporter molecule 5-bromo-4-chloro-3-indlyl-beta-d-galactopyranoside, functionalised to gold nanoparticles, was hydrolysed by galactosidase to produce a SERS-active dimerised product.129 The change in SERS signal was visualised when the cells were Raman mapped and the resulting SERS image constructed using the large peak at 598 cm−1. The presence of the dimerised product inside the cells was evident, confirming the abundance of the enzyme. Cleavage of an alkyne Raman reporter, which can be followed ratiometrically, has also been utilised for the detection of caspase 3 in live cells with high sensitivity and good signal reproducibility.130 Caspase 3 plays a key role in apoptosis and has thus been used extensively as a cancer biomarker, particularly in monitoring prognosis.131–133</p><!><p>Homeostasis of intracellular pH is maintained at the organelle level under healthy conditions, but abnormalities can occur in cancer. To detect these changes faster and with increased sensitivities, nanotags and SERS measurements have been used. The pH sensitive Raman reporter 4-mercaptobenzoic acid (4-MBA) functionalised to a nanoparticle is conventionally used to build pH calibration curves based on changing peak ratios or intensities.134 The nanotags are then applied to a cell in numerous ways and the intracellular pH obtained. pH sensitive, SERS active fibre optic nanoprobes combined with Raman measurements were first used to measure the intracellular pH of human prostate cancer cells with no apoptosis nor aggressive lysomal response.135 In this example, the measurements were located to where the fibre optic was placed on the sample and did not give information on the cell as a whole or pH gradients within the cell. Subsequently Kneipp et al. attached 4-MBA to gold nanoaggregates and introduced them into mouse fibroblast cells before Raman mapping the cells.136 False colour plots of the calibrated ratio allowed the various pH values of the cell to be obtained. They displayed the dynamics of pH values in cells at sub-endosomal resolution. This approach has also been used for the SERS mapping of pH in live cells using 4-MBA functionalised to many different nanoparticles including silver clusters,137 gold nanoparticles,138–140 and gold nanostars.141 Building on the existing pH SERS mapping literature, Bando et al. paired pH SERS imaging with 3D nanoparticle tracking to trace the pH dynamics with a spatial accuracy of several tens of nanometres and a temporal resolution of 200 ms.103 By incorporating MBA onto self-assembled silver nanoparticles, nanogaps were designed for local pH sensing with high sensitivity, where the peak intensities of the carboxylate group (1390 cm−1) and CO stretching mode (1690 cm−1) showed a pH-dependent response. The assemblies were added to HeLa cells and time-lapsed SERS imaging showed time and location dependent pH changes in a living cell. This could be used to visualise the dynamic changes in the chemical environment caused by organelle interactions in cancer.</p><!><p>Bioactive SERS nanotags have been successfully used as molecular imaging agents to target a number of biomolecules and chemical interactions specific to cancer. By incorporating a recognition motif onto the surface of a nanoparticle, the tag can target specific moieties on the surface of or inside the cancer cell and the interaction can be monitored by Raman mapping the cell and creating false colour images. For example, lectin-functionalised silver nanoparticles have been used to investigate carbohydrate–lectin interactions on the surface of mammalian cells.142 As there is an increase in sialic acid expression in malignant prostate cells, sialic acid-specific lectin conjugated SERS nanotags were used to discriminate between non-cancerous and cancerous cells. The nanotags were incubated with each cell type and Raman mapped, followed by the construction of false colour images by measuring the intensity of the main SERS peak from the benzotriazole dye. This allowed qualitative differentiation between the SERS signal from the cancerous cells, which produced a large SERS signal due to the lectin and sialic acid interaction and a very low signal on the non-cancerous cells. This successfully demonstrated that glycan expression can be correlated with malignancy using SERS. Various binding interactions have been investigated using this approach, including protein–ligand interactions accomplished using gold nanoparticles coated in RGDFC, a peptide that binds to the αvβ3 integrin and is over expressed in colon cancer cells,143,144 folate receptor interactions on human ovary cancer cells, achieved by conjugating silver nanoparticles with folic acid,145 sentinel lymph nodes that were detected using ratiometric Raman dual-nanotag strategies using folate receptor targeted SERS tags,146 and for the detection of lymphoblastoid cells using silver coated gold nanoparticles conjugated to a DNA aptamer specific to the cell line.147</p><p>The most commonly employed recognition motif conjugated to nanoparticles are antibodies, which have been used to detect specific cancer related biomarkers. For example, the detection and identification of estrogen receptor alpha (ERα), which is one of the main biomarkers present in breast cancer, responsible for increased proliferation and metastasis, is crucial for the clinical diagnosis and correct treatment of the disease. Kapara et al. functionalised ERα specific antibodies to SERS nanotags that were then incubated with breast cancer cells and Raman mapped.148 The nanotags exhibited excellent biocompatibility along with spatial and temporal understanding of the location of the ERα location in breast cancer cell lines with different ERα expression status. To quantify the difference in cell lines, a sophisticated approach based on percentage of SERS response was used to determine that ERα positive breast cancer cells (MCF-7) exhibited a 4.2 times increase in SERS signal area in comparison to ERα negative cells (SKBR-3). This indicated the strong targeting effect of the antibody SERS nanotag towards the ERα. Furthermore, this method was used to investigate the activity of the drug fulvestrant, a selective estrogen receptor degrader (SERD). SERS mapping confirmed a weaker signal was obtained when cells were treated with fulvestrant due to ERα degradation, opening up the possibility of using SERS as a tool for the estimation of ERα expression levels. This work was expanded by employing the ERα specific antibody SERS nanotags for the detection of ERα expression in a 3D tumour model to better understand whether targeted nanotags are required to efficiently target ERα, or whether untargeted uptake by the EPR effect is sufficient.149 Using 2D and 3D SERS measurements, we successfully demonstrated the strong targeting effect of ERα specific antibody SERS nanotags, which had 63% more signal when compared to the non-targeted human epidermal growth factor receptor 2 (HER-2) specific antibody nanotags, confirming the differentiation between targeted and non-targeted nanotags (Fig. 3). Fulvestrant was also investigated in the 3D tumour model and ERα expression was again reduced, as confirmed by the lower SERS signal. This work highlighted the importance of performing assays on 3D cell cultures, which better reflect the tissue architecture and cell-to-cell/cell-to-matrix interactions present in real tumours. It also demonstrates the potential of using SERS nanotags to monitor ERα expression, with potential to be used for developing personalised treatment using primary cancer cells from patients.</p><p>One of the most promising advantages of SERS nanotags in cancer imaging is the multiplexing potential achieved by bio-conjugating SERS nanotags. Detection of multiple biomarkers is possible due to the narrow bandwidths of the reporter molecule, which can be imaged from the Raman maps to indicate the presence and location of multiple biomarkers within or on the surface of a cell. The rapid and sensitive phenotypic markers expressed on the cell surfaces of three different types of breast cancer cell lines have been detected using hollow gold nanospheres conjugated with specific antibodies.67 The results showed a quantitative distribution of the marker proteins as well as the cancer cell phenotypes via the SERS-mapping images. The simultaneous detection of two cancer biomarkers (MUC1 mucin and nucleolin) has also been achieved on the surface of MCF-7 cells using the self-assembly of branched DNA-gold nanoaggregates, providing information on the physiological and pathological states of the cancer cells.150In vivo cancer detection by SERS was first demonstrated by Maiti et al. who used antibody functionalised nanotags to target tumour sites in a mouse.125 The sites were Raman mapped and the resulting images revealed the location and distribution of each nanotag.</p><p>From these examples, it is clear that SERS has increased the sensitivity and selectivity over normal Raman for the detection of cancer in solution, surface and cell-based assays. This demonstrates the potential of SERS to be used as a pre-clinical screening technique that will detect cancer earlier and could fast track patients into treatment.</p><!><p>Raman spectroscopy has been utilised for clinical investigations due to it being non-destructive, non-invasive, and having the ability to monitor changes in molecular composition in a biological sample, which could be indicative of disease. It has a number of other advantages including utilising a back scattering optical configuration, allowing measurements to be taken from below the surface in thick tissue sections without the need for micro-sectioning.5 Water is not a strong Raman scatterer and measurements can be taken in aqueous environments by using visible or NIR excitation to reduce the absorption effects of water. It also provides real-time molecular information at a relatively low cost. However, the technique lacks sensitivity due to the intrinsic weakness of Raman scattering, which can result in long acquisition times.151 Issues can also occur when using visible excitation sources, which decrease the depth of penetration, give rise to tissue autofluorescence and can cause issues due to heat generation.152 In addition, sophisticated data analysis is often required to deconvolute the complex signals acquired.151 New strategies are being developed to overcome some of these limitations, including using NIR excitation sources, carrying out spatially offset measurements and endoscopes combined with SERS measurements.</p><p>Endoscopic imaging is regularly used in clinical diagnostics as it is a minimally invasive method of examining tissues within the body. Endoscopic Raman spectroscopy, as opposed to white light imaging, can provide biomolecular information and enable objective diagnosis to be made.153 Pioneering work by Molckovsky et al. studied the diagnostic potential of NIR Raman spectroscopy of the colon and evaluated its ability to distinguish between adenomatous and hyperplasic polyps using a custom-built, fibre optic, NIR endoscopic system.154 Biochemical monitoring of the human cervix throughout preganacy,155 diagnosis of dysplasia in Barrett's esophagus,156 gastric cancer diagnosis,157 and early lung cancer detection158,159 have also been investigated using endoscopic Raman probes.</p><p>To increase the sensitivity of the approach, it has been combined with nanoparticles and SERS measurements. A Raman endoscopic probe was designed by Zavaleta et al.,160 who inserted the device through a clinical endoscope and demonstrated the multiplexed detection of tumour-targeting nanoparticles. Jeong et al. developed an endoscopic device that combines fluorescence and Raman and used the technique for the simultaneous in vivo detection of cancer biomarkers, HER2 and EGFR, in breast cancer tissue.161 A novel, non-contact, opto-electro-mechanical device was also developed for the rapid imaging of large areas in the human gastrointestinal tract.162 This approach was also capable of detecting multiple SERS nanoparticles simultaneously, and showed potential for cancer diagnosis and treatment monitoring. Evidently, the combination of Raman spectroscopy with endoscopy is a useful approach for investigating cancer and potentially monitoring treatment. Alternatively, Raman spectroscopy can be used during surgery to guide procedures and aid successful resection. Karabeber et al. showed that by injecting tumour-bearing mice with silica-coated gold nanotags, accumulation of the nanoparticles occurred in the brain tumours and could be detected using SERS.163 This allowed imaging of the tumours using a handheld spectrometer that aided the removal of the tumour and showed improved resection when compared to surgical guidance using white light imaging. In a further development, Jermyn et al. developed a handheld Raman probe and demonstrated its use during live human brain surgery.16 Using an NIR laser and placing the fibre probe in contact with the brain tissue, they could differentiate between normal and cancerous cells in the human brain with greater accuracy (92%) than alternative techniques such as microscopy and MRI (73%). Wang et al. applied SERS-active targeting nanotags to freshly excised human breast tissue and obtained quantitative multiplexed molecular imaging in only 15 minutes, indicating that this approach could be used for guidance during breast cancer surgery.164</p><p>These are just some of the examples where the advantages of Raman spectroscopy have been exploited for the detection and diagnosis of cancer in a clinical environment. Evidently, further work is required before the techniques will be adopted in medical clinics; however, the potential of the methods has been demonstrated across several areas of cancer detection, using several different approaches.</p><!><p>The advantages of SERS for sensitive, specific and multiplexed detection can be further driven towards clinical applications by allowing the non-invasive detection of lesions buried beneath the surface of the skin. Spatially offset Raman scattering (SORS) and surface-enhanced spatially offset Raman scattering (SESORS) are novel methods that enable in vivo detection of the molecular changes associated with diseases, such as cancer, by facilitating the ability to obtain signals from depths up to several centimetres below a surface. This allows non-invasive monitoring of signals from tissues in vivo, which could significantly improve early cancer detection and treatment monitoring.</p><p>By offsetting the signal collection probe from the laser excitation probe in Raman spectroscopy, photons scattered from the subsurface medium can be collected, allowing signals to be obtained from below the surface and through barriers, such as tissues, with an increasing offset resulting in signals being obtained from greater depths.165,166 Since the first demonstration of SORS in 2005,77 the technique has been successfully applied for the transcutaneous in vivo analysis of human bone79 and the through tissue analysis of tumours167 and calcifications82 in breast tissue, indicating its potential for non-invasively detecting cancer in its early stages.82 The capabilities of SORS for clinical applications have also been highlighted by demonstrating that signals can be obtained from significant depths, through-barrier, using a handheld spectrometer.166 Although this study focussed on the detection of ethanol through plastic, it showed the potential of using both conventional Raman and SORS in clinics, where handheld spectrometers would be particularly advantageous, and verified that signals can be obtained from greater depths when using SORS than by focussing into the sample using normal Raman optics.</p><p>The potential of SORS is further enhanced by combining its capabilities with the sensitivity of SERS to achieve significantly improved signals from even greater depths, as well as introducing the ability to target specific disease markers using tagged nanoparticles. SESORS was first proposed in 2010,83 when it was established that SERS nanotags could be detected through 25 mm of porcine tissue using transmission Raman, where the collection probe was placed on the opposite side of the sample to the laser. Transmission Raman is an example of an extreme spatial offset, where the angle between excitation and collection is 180°. Silver nanoparticles functionalised with a NIR dye were injected into tissue samples and the potential of the technique for the detection of small tumours was described, indicating the number of nanoparticles that may be required for the detection of lesions of particular sizes. In a further development, SESORS imaging was implemented and four different flavours of SERS nanotag were injected into a porcine tissue block, where their unique signals were non-invasely detected from a depth of 20 mm.168 False colour images were generated using the most intense peak for each flavour of nanotag and the spatial distribution of each nanotag could be observed. Signals were obtained from the nanotags at 47 mm; however, the signal deteriorated at the greater depth, particularly above 1250 cm−1, due to the increased absorption from water and myoglobin from the tissue in this region. In this study, nanotags were encapsulated such that the SERS signal was obtained from Raman reporters rather than target molecules; however, functionalisation of the nanotags with molecules of interest such as cancer biomarkers, cell specific proteins or DNA fragments would allow application of the technology for cancer detection and treatment monitoring. Bisphosphonate-tagged AuNPs were used to target calcium on the surface of bone samples, where the bisphosphonate/calcium binding enabled detection of the nanotags from the surface of the bone using Raman mapping.169 To demonstrate potential for in vivo imaging, bone samples covered in bisphosphonate-functionalised nanotags were covered with 20 mm of porcine tissue to mimic detection of the nanotag-functionalised bone through tissue. Spatially offset Raman maps were collected across the bone samples and principal component analysis (PCA) was used to identify the peaks from the nanotags and the bone. This demonstrated the detection of a fine distribution of NPs from the surface of bone, rather than a concentrated droplet injected into tissue. The use of bone/calcium specific nanotags to obtain a SESORS signal from the surface of the bone, through 20 mm of tissue, showed potential for detection of metastatic breast cancer, as well as bone disease.</p><p>One of the greatest advantages of SESORS is its potential for the non-invasive detection and monitoring of tumours in vivo. Multicellular tumour spheroids (MTS) are used as tumour models to mimic the 3D in vivo environment of tumours. This allows the ex vivo study of cancer, closely mimicking the in vivo environment, without the need for ethical approval and more complicated experiments. Nanoparticles are known to passively accumulate in tumours, allowing SERS imaging to distinguish between cancerous and healthy cells.170,171 MTS can be grown with uniformly distributed NPs to mimic the accumulation in tumours and thus provide a model for ex vivo tumour detection.172 This has been utilised to demonstrate the use of surface enhanced spatially offset resonance Raman scattering (SESORRS) for imaging a live breast cancer tumour model through tissue using a handheld spectrometer.173 SESORRS involves the incorporation of a dye-label with an electronic transition close to the frequency of the exciting laser, to significantly improve the sensitivity and therefore depth penetration of SESORS.174 Human breast cancer cells were incubated with resonant dye-labelled AuNPs resulting in the accumulation of the nanotags within the cells, which were then used to grow MTS.173 The MTS were then transferred to a section of tissue and a 15 mm section of porcine tissue was placed on top of the layer to simulate the detection of SERS nanotags through the tissue using SORS. Spectra were acquired from the MTS models by probing the tissue sample using a handheld SORS instrument, with an 830 nm laser excitation wavelength in backscattering configuration and an 8 mm spatial offset. Peaks in the spectra at 1178 cm−1 and 1592 cm−1 corresponded to the dye label, demonstrating the uptake of the nanotags into the MTS. Spectra were collected every 3 mm to create an image with 7 × 7 pixels and a false colour map was generated based on the intensity of the peak at 1178 cm−1 (Fig. 4(A)). The location of the MTS models is evident in the areas of maximum intensity and the signal from the MTS is clearly distinguishable from the background tissue signal (Fig. 4(B)). This gives an indication that SESORRS imaging could be used to detect functionalised nanoparticles through 15 mm of tissue, thus demonstrating the potential of the technique for in vivo tumour detection. The capability of this approach was further validated by analysing SERS nanotags through 25 mm of porcine tissue (Fig. 4(C)). Again, using an 8 mm offset with the handheld SORS instrument, signal could be obtained from the SERS nanotags through the tissue, and peaks from the dye at 1178 cm−1 and 1592 cm−1 were clearly distinguishable from the tissue reference. Although greater depth penetration was achieved previously,168 this was using a transmission geometry on a benchtop SORS instrument. In the work described here, backscattering geometry was used, where collection is from the same side of the sample as the exciting laser but with a spatial offset applied, rather than collecting from the opposite side of the sample. The use of a handheld spectrometer with backscattering optics signifies the potential of this technique for clinical applications and was a significant step towards the non-invasive detection of tumours.</p><p>In a further development, a similar approach was used to demonstrate the multiplexing capabilities of SESORRS.175 The detection and classification of three nanotags, both individually and as a triplex, was performed through 10 mm of tissue using handheld SESORRS. Spectra were collected from the three individual dyes and from a mixture of the three at equal concentrations, both from a MTS tumour model and from nanotags in solution. Since the Raman spectra of the three dyes were fairly similar, PCA was applied to discriminate between the single nanotags and the triplex. The resulting scores plots gave clear separation into four distinct groups for the three individual dye spectra and the spectra of the triplex, demonstrating the successful identification and discrimination of single and multiplexed SERRS nanotags through 10 mm of tissue using a handheld SORS spectrometer. This highlights the potential to simultaneously detect multiple targets in vivo, which is advantageous for the detection and monitoring of disease, where the sensitive detection of multiple biomarkers is of significant interest to determine cancer phenotype.</p><p>The recent developments in the through tissue detection of live breast cancer tumour models exploited the enhancement in signal that can be achieved by using a dye that is in resonance with the laser excitation wavelength.173,175 This resonance effect allows significant enhancement in SERS signal, which in turn enables greater depth penetration. This concept was further examined by comparing the signals obtained from nanotags functionalised with a non-resonant reporter (SERS tags) to those observed when functionalised with a resonant dye (SERRS tags).174 Observed detection limits were 11 times lower when the resonance effect was exploited and a calculated detection limit of 104 fM was suggested when using SESORRS. Detection of nanotags using handheld instrumentation at this level of sensitivity, through clinically relevant depths, shows the potential of SESORRS for clinical applications and for in vivo detection of cancer.</p><p>An early demonstration of the potential of SESORS for in vivo detection was the transcutaneous detection and quantification of glucose via implanted silver film over nanosphere (AgFON) surfaces.84,176 Using a capture layer of decanethiol/6-mercapto-1-hexanol (DT/MH), glucose was attracted to the AgFON surface, where its Raman signal was enhanced, and could be detected through skin using SESORS. The sensor proved to be functional for 17 days after implantation, with high accuracy and consistency, using laser powers that are safe for skin exposure. The capabilities of SESORS for in vivo detection were further demonstrated by Sharma et al., who obtained spectra of nanotags embedded in tissue through bone.177 This was the first demonstration of through bone detection using SESORS, which demonstrated the potential of the technique to be used for through-skull analysis. This was later proven when SESORS was used for the non-invasive detection of neurotransmitters in a brain tissue mimic through a cat skull.85 The detection of melatonin, serotonin and epinephrine was achieved down to concentrations of 100 μM and PCA was used to demonstrate that unique spectra were obtained from each of the three neurotransmitters.</p><p>Nicolson et al. recently reported the first use of in vivo SESORRS imaging to obtain Raman spectra from brain tumours in mice through the skull.87 Au nanostars were tagged with a resonant Raman reporter to create SERRS nanotags that were then functionalised with a cyclic RGDyK peptide, to enable specific targeting to glioblastoma multiforme (GBM) tumours in vivo. The RGD-SERRS nanotags were injected into the tails of five tumour bearing mice and prior to imaging the mice were anesthetized. SESORRS images were collected from the mice and compared with conventional Raman images. Using the SORS setup, stronger Raman signals were obtained and a greater tumour to background contrast was observed. This allowed tumour location information to be resolved using a lower laser power, indicating the applicability of the technique for in vivo clinical applications.</p><!><p>The work presented in this review indicates how Raman spectroscopy has the potential to become an important clinical tool in cancer detection, diagnosis and treatment monitoring. However, despite great promise, its translation into the clinic for widespread human use is slow. This is due to a number of challenges including safety, cost, sustainability, duration of analysis, laser source and power, and auto-fluorescent tissues.152 Perhaps the largest obstacle is simply demonstrating the added value that Raman analysis can provide over, or in combination with, existing technologies. Although a challenge, many research groups have investigated the potential of Raman spectroscopy for clinical use, mostly in the form of in vivo studies with patient samples, focusing on the sensitivity and specificity of the approach.178,179Ex vivo and in situ measurement on patients have also been achieved to differentiate between cancerous and non-cancerous specimens.180,181</p><p>SERS clinical translation also raises new challenges such as the synthesis of reproducible SERS tags and lack of clinical evaluation when measuring in vitro samples such as serum or blood.182 These issues are being addressed by synthesising SERS tags on a larger scale and introducing protecting agents such as mercaptoundecanoic acid to provide stability.183 More emphasis is also being placed on clinical evaluation involving testing of cohorts of well-characterised patient samples.184–186 Of course, there are also major barriers to administering nanoparticles in vivo, which need to be overcome before they can be safely and routinely used in humans. These include the toxicological effects,187 non-specific binding and formation of protein coronas on the nanoparticle surface,188 circulation time,189 clearance pathways,190 and labelled nanoparticles altering their physicochemical properties.191 In order to progress, biocompatible and biodegradable SERS probes with minimal cytotoxicity are being investigated and several groups have shown minimal toxicity with gold nanoparticles coated in a number of different protective layers such as silica192 and PEG.193,194</p><p>In a recently published review, Xi and Liang retrieved the number of Raman spectroscopy clinical trials being carried out from the International Clinical Trial Registry Platform (ICTRP) search portal using the key word 'Raman'. The registered trials were then screened to exclude non-related Raman records or repeat studies.195 As of 2021, the search produced 55 registered Raman spectroscopy clinical trials, with 36% currently recruiting. The trials can be split into 5 categories: 54.5% are 'observational' aiming to observe patients to measure certain outcomes without intervention; 32.7% are 'interventional' which evaluate one or more particular intentions; 9.1% are 'diagnostic', evaluating diagnostic accuracy; 1.8% are meta-analysis, a statistic process combining findings from individual status; and 1.8% are 'relevant factors research'. There are also 6 SERS clinical trials, one of which is aimed at detecting circulating tumour cells in peripheral blood originating from breast cancer tissue.196 It should be noted that in 83.6% of the trials, the recruitment sample size is less than 200 subjects and it has been suggested that university and research centres need to forge a more collaborative effort with clinicians and industrial sponsors to carry out large-scale, high quality and multicentre registered Raman clinical trials.</p><p>A search of published clinical trials was also performed in PubMed by searching for 'Raman' with the article type restricted to 'clinical trials'. This search resulted in 44 published clinical trials using various Raman techniques, with confocal Raman and transcutaneous Raman being approved to meet the clinical accuracy requirement for the non-invasive detection of glucose in vivo.197,198 However, the majority of these studies had sample sizes of less than 100 and were carried out at single sites. This lack of consistency could deter investors, hamper product development and delay translation.199 To address this, multicentre studies and inter-laboratory 'round robins' need to be implemented to reduce bias, validate the robustness of the technique and generate more convincing evidence. It is evident that there are still barriers that Raman spectroscopy clinical trials need to overcome, however there are strategies that can be employed to produce clear, concise and compelling evidence that Raman should be used in the clinic. Although it is still not the 'gold standard', it is evident that Raman and SERS can offer tremendous gains in cancer detection, diagnosis and treatment, and that the outlook remains positive.</p><!><p>Raman scattering and its enhanced forms offer many advantages for use in cancer detection, diagnosis and treatment monitoring. Each of the different techniques have individual advantages, enabling applicability in the many different approaches of investigating cancer. While Raman spectroscopy can be used to give molecularly specific information that can be useful in determining disease states, variations in the technique can further improve its capabilities. For example, SRS can vastly increase imaging speeds at the cost of spectral molecular information, and the use of SERS significantly improves sensitivity while introducing the capability for targeted assays. Recent modifications, like SORS and SESORS, open up further opportunities for in vivo analysis by allowing spectra to be collected through tissue. Additionally, handheld SORS instruments and probe-based SERS systems have been developed and demonstrated for their potential use in vivo, making the techniques suitable for point of care testing. Despite the advantages and progression towards clinical application, the full potential of Raman spectroscopy is yet to be exploited for medical diagnostics. This is due to several factors mainly pertaining to the use of nanoparticles in the body that can have a toxic effect such as inducing oxidative stress or cellular damage, poor retention times that reduce their targeting properties, and unclear excretion pathways. Other issues, such as cost, analysis time, and difficulty proving the advantages over current standard methods, also reduce the use of unlabelled Raman spectroscopy in the clinic. For these reasons the clinical use of Raman spectroscopy is still limited; however, with recent developments allowing faster imaging speeds, improved sensitivity and greater in vivo potential, instruments with clinically safe laser powers can be used to non-invasively obtain quantitative and detailed information for the detection and diagnosis of cancer. In comparison to current optical imaging techniques, such as MRI or ultrasound, Raman spectroscopy can obtain more detailed biochemical information and is a more quantitative method of analysis. However, these techniques have been employed for many years and are widely accepted as being suitable for cancer detection and medical diagnosis in general. Concerns with, for example, the safety of using lasers in the clinic and toxicity of nanoparticles for human consumption, must be overcome before use in clinical practice will be considered. Therefore, larger studies are required to demonstrate that the instrumentation and methods are safe for clinical use. The various techniques discussed in this review allow the advantages of Raman spectroscopy to be exploited for the detection and diagnosis of cancer in many ways, from in vitro biomarker detection and ex vivo tissue analysis to in vivo tumour detection. This indicates that the different techniques and applications complement each other well and could provide a toolbox for medical applications. Further clinical studies are required to prove the benefits of the techniques but the area is moving in the right direction to achieve this and to move towards clinical translation.</p><!><p>There are no conflicts of interest to declare.</p>

PubMed Open Access

Interfacial Electron Transfer into Functionalized Crystalline Polyoxotitanate Nanoclusters

Interfacial electron transfer (IET) between a chromophore and a semi-conductor nanoparticle is one of the key processes in a dye sensitized solar cell. Theoretical simulations of the electron transfer in polyoxotitanate nanoclusters Ti17O24(OPri)20 (Ti17) functionalized with four para-nitrophenyl acetylacetone (NPA-H) adsorbates, of which the atomic structure has been fully established by X-ray diffraction measurements, are presented. Complementary experimental information showing IET has been obtained by EPR spectroscopy. Evolution of the time-dependent photoexcited electron during the initial 5 fs after instantaneous excitation to the NPA LUMO+1 has been evaluated. Evidence for delocalization of the excitation over multiple chromophoresafter excitation to the NPA LUMO+2 state on a 15 fs timescale is also obtained. While chromophores are generally considered electronically isolated with respect to neighboring sensitizers, our calculations show that this is not necessarily the case. The present work is the most comprehensive study to date of a sensitized semiconductor nanoparticle in which the structure of the surface and the mode of molecular adsorption are precisely defined.

interfacial_electron_transfer_into_functionalized_crystalline_polyoxotitanate_nanoclusters

Introduction<!>Chemicals<!>Data collection<!>EPR spectroscopy<!>Computational Model<!>Electron Transfer Dynamics<!>Electronic Optical Transitions<!>Results<!><!>Binding Mode<!>Charge Separation<!>Interfacial Electron Transfer<!>Electronic Transitions<!>Conclusions<!>

<p>Dye sensitized solar cells (DSSC) promise the environmentally-friendly and cost-effective conversion of solar light into fuels and/or electricity.1–4 DSSCs are driven by photoinduced interfacial electron transfer (IET) which injects photoexcited electrons from the dye sensitizer into the conduction band of the semiconductor substrate. Naturally, the interfaces are critical to performance and much attention has been directed to them via experiment, theory and computational modeling.5–8 However, due to the complexity of the surfaces, including the presence of multiple exposed facets and binding sites as well as surface defects and impurities, the correlation between experiment and simulations based on pristine surfaces remains uncertain. Here, we bridge the gap between theoretical modeling, crystallography, and spectroscopy by studying IET in functionalized polyoxotitanate (POT) nanocrystals Ti17O24(OPri)20 (Ti17) of which the structure has been precisely determined by X-ray diffraction. The precise structural information enables time-dependent electronic structure calculations which predict photoinduced IET in this model system. The sensitization of the nanoparticle by the acetylacetonate-anchored chromophore is confirmed spectroscopically by an increase in the onset of the photoinduced titanium and oxygen EPR signals from approximately 300 nm for the unfunctionalized cluster to over 400 nm in the case of the functionalized nanoparticle. The comprehensive analysis described should contribute to a more detailed understanding of the chemistry and photophysics of sensitized semiconductor interfaces.</p><p>The structure of the interface can greatly influence the rate and efficiency of IET.5,9 In the case of the dye/semiconductor interface, strong electronic coupling between the dye excited state and the conduction band is desirable for ultrafast electron transfer. Fast injection into the semiconductor minimizes losses due to radiative and non-radiative relaxation of the dye excited state. We have recently shown that a new class of derivatized acetylacetonate (acac) linkers can functionalize pure phase TiO2 nanoparticles and be used to anchor photocatalytically active manganese complexes to the surface.10,11 These acac linkers provide robust coupling to the semiconductor substrate under aqueous and oxidative conditions. The characterization of the acac anchor binding mode to the TiO2 surface has been only indirect, based on computational modeling and UV/vis and IR spectroscopy as in previous studies.8,10–16 Here, we structurally resolve TiO2 surfaces functionalized by acac linkers by using X-ray diffraction methods.</p><p>POT clusters functionalized with catechol and isonicotinic acid have been proposed as models for the sensitizer/semiconductor interface in DSSCs.17 POT clusters are attractive analogues for pure phase TiO2 nanoparticles since they possess structural features of bulk crystals, including similar coordination and connectivity, as well as features associated with nanoparticles such as reactive 4- and 5-coordinate Ti4+ centers18 and size-dependent band gaps.19 The clusters are also attractive building blocks due to theirpropensity to assemble into single crystals suitable for X-ray structure determination at atomic resolution.20,21 Here, we focus on the functionalization of the POT cluster Ti17O24(OPri)20 (Ti17) with the model sensitizer para-nitrophenyl acetylacetone (NPA-H) and the subsequent crystallization of the products to gain atomic resolution structural information about the sensitizer/semiconductor interface in these particles. The light-induced charge-separated states in functionalized and unfunctionalized POT clusters are compared by EPR spectroscopy and analyzed by quantum dynamics simulations of IET, providing a detailed description of charge injection for a structurally resolved sensitizer/semiconductor interface.</p><!><p>All reagents and solvents were purchased from commercial sources. Benzene (anhydrous, 99.9%, Alfa Aesar) was degassed prior to transfer and storage in a glove box. All compounds containing titanium were stored and handled in a glovebox under an argon atmosphere. Ti17 was prepared according to previously reported methods.22</p><p>3-(4-nitrophenyl)pentane-2,4-dione (NPA-H):The synthesis of NPA-H was carried out in an analogous manner to the preparation used by Jiang et al.23 to synthesize 3-(3-nitrophenyl)pentane-2,4-dione. A mixture of 1-iodo-4-nitrobenzene (2.0 mmol), 2,4-pentanedione (6.0 mmol), cesium carbonate (8.0 mmol), freshly recrystallized copper (I) iodide (0.20 mmol), and L-proline (0.40 mmol) in dry DMSO (10 mL) was wrapped in aluminum foil to protect from it light and heated at 70 °C under nitrogen atmosphere for 24 hours. The cooled solution was poured into 1M HCl and extracted with ethyl acetate. The organic layer was washed with water, brine, dried over Na2SO4, and the solvent removed in vacuo. The crude residue was purified by silica gel flash column chromatography, using a mixture of hexanes : ethyl acetate (7:3) as eluent to afford 292 mg (66 % yield) of NPA-H as a yellow solid. The product was then recrystallized from hexanes to give a slightly yellow crystalline solid. 1H NMR (400 MHz, CDCl3) δ 16.78 (s, 1H), 8.27 (dd, J = 2.1, 8.9, 2H), 7.39 (dd, J = 2.1, 8.9, 2H), 1.90 (s, 6H). 13C NMR (CDCl3, 500MHz): δ 190.5, 147.3, 144.0, 132.1, 124.0, 113.6, 24.2. HRMS calcd (found) for C11H11NO4 M+: 222.076084 (222.07611).</p><p>Ti17O24(OPri)16(NPA)4 (Ti17NPA4): To a 20 mL vial containing Ti17 (26.9 mg, 11.3 μmol) dissolved in 5.0 mL benzene was added a solution of NPA-H (10.0 mg, 45.2 μmol) dissolved in 5.0 mL of benzene. The vial was loosely capped and allowed to slowly evaporate over a period of 2–3 days. After this time, pale yellow (nearly colorless) crystals suitable for single crystal X-ray diffraction were obtained.</p><p>Crystals obtained from a benzene solution are only stable in oil for 20–40 seconds. After numerous attempts, a single crystal 0.4 × 0.2 × 0.2 mm3 was rapidly mounted on a glass fiber in oil and cooled to 90 K. The rapid decomposition of the crystals is likely due to the presence of seven benzene molecules per Ti17NPA4 in the lattice.</p><!><p>X-ray diffraction data were collected at on a Bruker SMART APEX2 CCD diffractometer installed at a Rigaku RU-200 rotating anode source (Mo Kα, λ=0.71073 Å) and equipped with an Oxford Cryosystems nitrogen gas flow apparatus. Data were collected at 90 K with a crystal to detector distance of 40 mm. Five ω-scans (180°/scan, 0.5°/frame) were collected with an exposure time of 30 seconds per frame.</p><!><p>Samples for EPR spectroscopy were prepared in a glovebox under N2 atmosphere by dissolving 5 mg Ti17 or Ti17NPA4 in 2 mL 2:1 dichloromethane:benzene and transferring approximately 200 µL of this sample to a 4 mm OD quartz EPR tube. Samples were frozen in liquid N2 before being transferred to the cryostat. EPR spectra were measured at 7 K in perpendicular mode on a Bruker ELEXYS E500 spectrometer equipped with an SHQ cavity and Oxford ESR 900 liquid helium cryostat. Samples were illuminated in the cryostat at 7 K using a 1000W Xe arc lamp equipped with a water filter and various longpass filters. Spectra were recorded with the following settings: microwave frequency = 9.391 GHz, microwave power = 1.0 mW, modulation amplitude = 4 G, modulation frequency = 100 kHz.</p><!><p>The crystal structure of [Ti17O24(OPri)16(NPA)4] was used as the initial structure and optimized to the minimum energy configuration as determined by density functional theory (DFT), using the B3LYP exchange-correlation functional, with the LACVP basis set by using the computational chemistry package Jaguar 7.24 The models were simplified by replacing OPri groups by OH groups, giving the model structure [Ti17O24(OH)16(NPA)4]. The RMSD of the Ti atoms between the crystal structure and the relaxed DFT model structure is 0.05 Å, indicating the similarity of the models and confirming that replacing OPri by OH induces only minor structural rearrangements.</p><p>It is well-known that time-dependent density functional theory (TDDFT) yields substantial errors for the excitation energies of charge-transfer (CT) excited states, when approximate standard exchange-correlation (xc) functionals are used, such as SVWN, BLYP, or B3LYP.25–27 Also, the correct 1/R asymptotic behavior of CT states with respect to a distance coordinate R between the separated charges of the CT state is not reproduced by TDDFT employing these xc-functionals.25,28,29 The first failure is due to the self-interaction error in the orbital energies from theground-state DFT calculation, while the latter is a similar self-interaction error in TDDFT arising through the electron transfer in the CT state.30</p><!><p>To characterize the IET time scale, the survival probability, PMOL(t), defined as the probability that the photoexcited electron remains in the adsorbate molecule, NPA, at time t after excitation of the system was computed. PMOL(t) was obtained as the projection of the time-evolved electronic wave function onto the atomic orbitals (AOs) of the molecular adsorbate. The survival probability was computed as PMOL(t)=|ΣiMOLΣjBi*(t)Bj(t)Sij|, where S is the overlap matrix and Bi is a time-dependent expansion coefficient. Computing the time-dependent wave function Ψ(t)=ΣiBi(t)χi, expanded in the basis set of AOs χi, required the propagation of the expansion coefficients Bi(t)=ΣqQiqCqexp[−(i/ℏ)Eqt], where Cq are the expansion coefficients of the initial state. The eigenvectors and eigenvalues Qq and Eq were obtained from the ground state electronic density using the electronic structure package Gaussian 09, Revision A.02,31 with the electronic structure described by DFT at the B3LYP/LANL2DZ level of theory. Initial states, Cq, were defined in terms of the unoccupied orbitals of an isolated NPA-H molecule. Electron transfer dynamics simulations were performed on the previously relaxed [Ti17O24(OH)16(NPA)4] system. It should be noted that out theoretical models do not include losses due to radiative and non-radiative relaxation of the dye excited-state which typically occur on the ns time scale. Likewise, the simulations have been performed in the low temperature (0 K) limit where nuclear motion is much slower than electronic relaxation, thus coupling between the electronic wavefunction and the molecular vibrations has not been included.</p><!><p>To characterize the photoexcited electrons generated by optical transitions, excited states were computed by using time-dependent density functional theory (TDDFT) at theB3LYP/LANL2DZ level, as implemented in Gaussian 09. Optical transitions were computed for both the isolated NPA-H molecule and the [Ti17O24(OH)16(NPA)4] model. The minimum energy configuration of the isolated NPA-H molecule was obtained at the B3LYP/6-31G* level of theory.</p><!><p>The reaction between Ti17 and NPA-H follows the same general scheme as reported for catechol and isonicotinic acid.17 Four equivalents of NPA-H, one for each reactive five-coordinate atom in Ti17, are treated with one equivalent of the POT cluster to give a Ti17 cluster which contains four NPA ligands, Ti17O24(OPri)16(NPA)4 (Ti17NPA4).</p><!><p>Ti17O24(OPri)20 + 4 NPA-H →Ti17O24(OPri)16(NPA)4 + 4 HOPri</p><!><p>Two modes of bidentate attachment of the acac linker to the (101) surface of anatase were investigated previously.10 Bidentate attachment of the acetylacetonate anchor group at an oxygen vacancy on the surface of anatase was predicted to be the most energetically stable mode of attachment. The second bidentate mode resulted in the displacement of the titanium from the surface and the breaking of a sub-surface Ti-O bond, a rearrangement that was energetically disfavored. Given the energetic analysis and the displacement of OPri by the NPA ligand during functionalization, it is natural to expect that the NPA binding motif in Ti17NPA4 closely resembles the bidentate mode in the oxygen vacancy model of anatase.</p><p>Figure 1 shows the geometry of a fragment of Ti17NPA4 and the two models of bidentate attachment to the anatase (101). In the case of the four Ti-O bonds which attach the fragments to the larger nanoclusters, the bond lengths of the experimental geometry (from 1.897 to 1.933 Å) are closer to the values for the oxygen vacancy model (from 1.951 to 1.980 Å) than the surface chelation model (from 2.100 to 2.157 Å). In fact, the Ti-O bond lengths of both the experimental geometry and the oxygen vacancy model are quite similar to that of bulk anatase (1.934 and 1.980 Å). The distance between the two oxygens of the acac anchor are 2.594 Å and 2.544 Å for the experimental geometry and the surface chelation model, respectively. The larger value of 2.787 Å observed in the oxygen vacancy model is a consequence of the fact that the oxygen atom of the anchor which fills the vacancy also interacts with a neighboring titanium atom (not shown).</p><p>While nearly 100 crystal structures containing the acac group bound to titanium have been deposited in the CSD,32 the vast majority contain only one or two titanium atoms. Prior to this report, the largest polyoxotitanate possessing a dione ligand had only five titanium atoms in its core (WIYCOW). In every one of these Ti-acac crystal structures, bidentate chelation of the acac group to a single titanium atom was observed. In several additional crystal structures of larger Ti/O clusters containing acac groups (to be reported), the observed binding is invariably bidentate.</p><!><p>The light-induced charge-separated state was characterized using EPR spectroscopy. Under illumination, electrons in the conduction band are localized on titanium centers, generating d1 Ti3+ species, while holes are localized on oxygen atoms, producing oxygen-centered radicals. At low temperature with continuous illumination, both paramagnetic species are long-lived and easily observed by EPR. The spectra shown for Ti17 and Ti17NPA4 in Figure 2 indicate contributions from both titanium- and oxygen-centered unpaired electrons. The signals at g> 2 are assigned to oxygen anion holes, and those at g< 2 are assigned to titanium electrons, consistent with previous reports of charge separations in TiO2.33–37 Preliminary spectral simulations suggest that both regions of the spectrum are composed of three overlapping signals resulting from contributions of oxygen and titanium centers in different coordination environments. Of the three contributing titanium signals, two axial components are present in both the Ti17 and Ti17NPA4 spectra. A third axial component changes upon coordination of the NPA ligand; suggesting that it derives from the five-coordinate titanium center in Ti17 which becomes six-coordinate upon NPA ligation. Complete analysis of the EPR results will be reported in a future publication.</p><p>The wavelength of light used to illuminate the sample may be varied with longpass filters, used in order from longest to shortest wavelength. Comparison of the wavelength dependent spectra of Ti17 and Ti17NPA4 demonstrates that coordination of NPA shifts the onset wavelength for charge separation. For unfunctionalized Ti17, no oxygen or titanium radicals are seen for wavelengths greater than 345 nm. However, radical formation is observed when a >295 nm filter is used, suggesting that the excitation wavelength for the onset of charge separation lies between 345 nm and 295 nm. By contrast, some of the charge-separated state for Ti17-NPA4 clusters is observed with irradiation above 400 nm, and the EPR signal intensity increases with illumination at shorter wavelengths. The transition responsible for the generation of the EPR signal is significantly shifted by the addition of the NPA chromophore, suggesting that Ti17 clusters are sensitized in a manner analogous to TiO2 nanoparticles.</p><!><p>Before reporting the results from the IET simulations, it is useful to look at the density of states (DOS) and projected density of states (pDOS) of the functionalized nanoparticle. The DOS for an unbound NPA-H molecule is included to examine the effect of ligation on the electronic structure of the sensitizer. Figure 3 shows the DOS of the [Ti17O24(OH)16(NPA)4] model system along with the pDOS of all four NPA adsorbates. The DOS shows that the HOMO-LUMO gap in the model system is ~3.5 eV with the HOMO and LUMO (both are quasi-four-fold degenerate states) populations localized mainly on the NPA adsorbates as indicated by the pDOS. The pDOS also indicates that the majority of the NPA orbitals are energetically unperturbed upon complexation to the Ti17 cluster with the exception of the NPA-H LUMO+1. The ligand LUMO+1 orbitals mix with the first unoccupied levels of the Ti17 cluster to form new ligand-metal hybrid orbitals located at ~4.2 eV above the HOMO.</p><p>Figure 4 shows the electron density of the four initial excited states used in the IET simulations. The initial states correspond to instantaneous photoexcitations to the LUMO, LUMO+1, LUMO+2, and LUMO+3 of a single NPA with energies of 3.72, 4.83, 5.69, and 6.19 eV, respectively. The pDOS indicates that the first excited state (i.e., the NPA LUMO) is located below the unoccupied Ti/O orbitals of the cluster, indicating that IET from that state is unlikely to be observed. The remaining states are aligned with a large density of unoccupied Ti17 cluster levels, especially the LUMO+1, indicating that IET could occur from these states.</p><p>Figure 5 shows the survival probability P(t) for an electron starting from one of the three initial states which overlap energetically with Ti/O orbitals: the LUMO+1, LUMO+2, and LUMO+3 of the NPA adsorbate. The simulation using the NPA LUMO initial state resulted in a survival probability that shows minimal population transfer, <1%, of the photoexcited electron during the first 1 ps of dynamics and was not included in the figure. Thus, no injection can occur from photoexcited electrons populating the NPA LUMO. Charge injection from the NPA LUMO+1 to the Ti/O cluster occurs extremely rapidly as P(t) falls to nearly zero within 1–2 fs (Figs. 5 and 6). Following electronic excitation, strongly coupled non-degenerate quantum mechanical systems generally exhibit Rabi oscillations which are readily observed for IET from the LUMO+1 and LUMO+2 states. The larger oscillations and non-exponential behavior of the IET from the LUMO+1 are a consequence of the finite number of empty Ti17 orbitals available at this energy.38</p><p>The simulation of IET, using the NPA LUMO+2 as the initial state, shows ultrafast IET with an injection time scale of approximately 12 fs. The NPA LUMO+3 initial state shows fast initial IET transferring about 20% of the electron population in the first 10 fs followed by a much slower rate of IET. It is interesting that a large difference in the survival probability is obtained for the third and fourth initial states, since the density of unoccupied Ti17 cluster states is similar at the energy of the initial states.</p><p>Figure 6 shows snapshots of the time-evolved electron density after instantaneous photoexcitation of the functionalized nanocrystal to the NPA LUMO+1 (the second excited state). The NPA LUMO+1 state is electronically coupled to the Ti17 cluster through the d orbitals of Ti chelated by the NPA moiety. The electron flows rapidly into the cluster with the population distributed to almost every Ti ion within 5 fs. Figure S2 shows snapshots of the first 15 fs of electron dynamics after instantaneous photoexcitation to the NPA LUMO+2 on a 3 fs interval. The LUMO+2 appears to couple to the Ti17 cluster through the same Ti d orbital as seen for excitation to the NPA LUMO+1. Similar to the LUMO+1 dynamics, the electron injected from the LUMO+2 state is distributed to all the Ti ions in the cluster. In both cases IET leads to electron density accumulating on a second NPA adsorbed directly across the Ti17 cluster from the NPA adsorbate where the excited electron originated. Such multi-chromophore delocalization arises from indirect coupling of the adsorbate orbitals through near-resonant states in the conduction band. Because the rate of ET is proportional to the square of the electronic coupling and the electronic coupling is determined by the orbital overlap, multi-chromophore delocalization is only observed on longer timescales (beyond ~15 fs).</p><!><p>The simulations of IET indicate that ultrafast injection will occur from the NPA LUMO+1 and LUMO+2 but not from the NPA LUMO. It is important to see if an allowed optical transition will promote an electron into the LUMO+1 or LUMO+2 of NPA (Table S8). The lowest energy transition of the isolated NPA, as computed using TDDFT, is a HOMO-LUMO transition at 349 nm with an oscillator strength of 0.0267 atomic units (au). Unfortunately, as seen from the IET simulations, the NPA LUMO is below the energy level of the unoccupied Ti17 states and this excited state is predicted not to undergo ultrafast IET. While the NPA HOMO to LUMO+1 does contribute to the fourth excited state which is a weak transition at 300 nm, it is the predominant component of the much stronger seventh transition occurring at 258 nm with an oscillator strength of 0.1275 au which also has considerable HOMO-LUMO+2 character. The ninth electronic transition is also a relatively strong transition with an oscillator strength of 0.0938 au and is primarly HOMO to LUMO+2 with considerable HOMO-LUMO+1 character. Even though there are no electronic transitions in the visible region for this model sensitizer, the HOMO-LUMO+1 and LUMO+2 transitions are optically strong and, thus, suitable candidates for charge injection. Furthermore, it is possible that a direct transition from the NPA to unoccupied levels of the Ti17 cluster is allowed with a corresponding excitation wavelength in the visible region, according to the usual type II injection mechanism–i.e., electron injection directly from the ground state of the dye into the semiconductor conduction band without the involvement of excited dye molecular states. In contrast, the so-called type I injection involves electron transfer from an excited state of the dye (e.g., the LUMO or LUMO+1 orbitals localized on the dye close to the interface) populated upon photoexcitation.</p><p>To investigate the possibility of direct transitions, a TDDFT calculation was performed on the [Ti17O24(OH)16(NPA)4] model (Table S7). Due to the large size of the system only the first 10 excited states were obtained. The third and seventh excited states possess relatively large oscillator strengths of 0.1391 and 0.1059 au, repectively. The photoexcited electron population for both states is occupied in the LUMO+2 and LUMO+3 of the [Ti17O24(OH)16(NPA)4] model. These virtual orbitals as well as the LUMO and LUMO+1 are quasi-four-fold degenerate with an energy of 3.76 eV and correspond to the LUMO of the free NPA. Thus, nine of the ten lowest energy transitions are effectively excitations into NPA orbitals which do not overlap with titanium. Only the ninth excited state, corresponding to an excitation wavelength of 357 nm and possessing a relatively small oscillator stength, shows population of the photoexcited electron on the Ti17 cluster, a direct consequence of populating the mixed Ti/NPA orbitals. Given the optical strength of the HOMO-LUMO+1 transition in NPA-H, other strong direct sensitizer-to-Ti transitions undoubtedly occur beyond the first 10 excited states calculated here.</p><p>To the best of our knowledge, this is the first example of an IET simulation involving multiple independent sensitizers in close proximity, attached to a semiconductor surface that is fully structurally characterized by X-ray diffraction. Given the high loading densities, typically observed in DSSCs, it is highly probable that surfaces with chromophores in close proximity exist in these devices. While chromophores are generally considered electronically isolated with respect to neighboring sensitizers, our calculations show that this is not necessarily the case. For the LUMO+2 transition, the excited electron diffuses through the Ti/O core and relocalizes on an NPA molecule on the opposite side of the cluster; density from a single excitation is observed on two sensitizer molecules simultaneously. Delocalization of electron density from an excited sensitizer onto neighboring sensitizer molecules could reduce the ability of the neighboring molecules to absorb light effectively, thereby reducing the efficiency of the device. Furthermore, delocalization of an excitation over multiple chromophores may lead to increased rates of charge transfer to redox shuttles which would also detrimentally impact device performance.</p><p>The strong mixing of the LUMO+1 state with the cluster orbitals also has important implications for light-harvesting systems. An increase in the broadening of a sensitizer injecting state, upon coupling to the substrate, is associated with increased rates of electron transfer.39 Similar behavior is observed in the calculated IET rates for [Ti17O24(OH)16(NPA)4]. While the NPA LUMO+2 and LUMO+3 states do not split or shift upon complexation to the POT cluster (no broadening), the LUMO+1 state does contribute to several new states and is effectively broadened over an energy range of approximately 1 eV. Initial rates of IET from the LUMO+2 and LUMO+3 states are similar at 0.02 and 0.03 fs−1, respectively, and are considerably less rapid than the rate of injection from the LUMO+1 state which is nearly an order of magnitude faster at 0.19 fs−1.</p><p>Given the fast injection and the low energy of the transition relative to the LUMO+2 and LUMO+3 states, injection from the NPA LUMO+1 appears to be ideal. In addition, direct injection through a type II mechanism should be possible. However, the formation of new energetically isolated orbitals below the quasi-conduction band could mean that direct excitation into these orbitals might not lead to injection into the semiconductor substrate. Instead, the excitation may remain localized on these orbitals and such states may even act as low energy sinks for higher-energy excitations.</p><!><p>As part of our project of single-crystal analyses of functionalized Ti/O nanoclusters,17,19 the structure of the polyoxotitanate nanoclusterTi17O24(OPri)20 functionalized with four para-nitrophenyl acetylacetone (NPA-H) adsorbates has been resolved by single crystal X-ray diffraction methods revealing the exact way in which the acac anchor group attaches to the semiconductor surface via bidentate chelation, as previously predicted by computational modeling.10 The crystalline nature of the nanoclusters has allowed us, for the first time, to perform both spectroscopy and simulations for the same model structures of photosensitizer dyes on semiconductor surfaces. We find evidence of photoinduced IET, even at λ>400 nm, as revealed by EPR spectroscopy and the analysis of the electronic structure at the DFT level. Theoretical calculations based on the functionalized cluster are used to predict the electronic properties relevant to light-harvesting devices. The density of states plot reveals that many of the unoccupied states of the free NPA-H are unaffected upon binding to the Ti17 cluster; however, the NPA-H LUMO+1 states mix with cluster orbitals near the quasi-conduction band edge resulting in new states which are lower in energy by approximately 0.5 eV and have both NPA and Ti character. This mixing is expected to facilitate direct IET and a significant increase in the rate of IET. Delocalization of the excitation over multiple chromophores is also identified.</p><!><p>Supporting information: Additional figures of Ti17NPA4 complex and IET simulations, tables of crystallographic information, CSD refcodes for titanium acetylacetone structures, TD-DFT results, the full reference 31, and .cif file for Ti17NPA4·7 C6H6.</p>

PubMed Author Manuscript

Sex Pheromone of the Alfalfa Plant Bug, Adelphocoris lineolatus: Pheromone Composition and Antagonistic Effect of 1-Hexanol (Hemiptera: Miridae)

The sex pheromone composition of alfalfa plant bugs, Adelphocoris lineolatus (Goeze), from Central Europe was investigated to test the hypothesis that insect species across a wide geographical area can vary in pheromone composition. Potential interactions between the pheromone and a known attractant, (E)-cinnamaldehyde, were also assessed. Coupled gas chromatography-electroantennography (GC-EAG) using male antennae and volatile extracts collected from females, previously shown to attract males in field experiments, revealed the presence of three physiologically active compounds. These were identified by coupled GC/mass spectrometry (GC/MS) and peak enhancement as hexyl butyrate, (E)-2-hexenyl butyrate and (E)-4-oxo-2-hexenal. A ternary blend of these compounds in a 5.4:9.0:1.0 ratio attracted male A. lineolatus in field trials in Hungary. Omission of either (E)-2-hexenyl-butyrate or (E)-4-oxo-2-hexenal from the ternary blend or substitution of (E)-4-oxo-2-hexenal by (E)-2-hexenal resulted in loss of activity. These results indicate that this Central European population is similar in pheromone composition to that previously reported for an East Asian population. Interestingly, another EAG-active compound, 1-hexanol, was also present in female extract. When 1-hexanol was tested in combination with the ternary pheromone blend, male catches were reduced. This compound showed a dose-response effect with small doses showing a strong behavioral effect, suggesting that 1-hexanol may act as a sex pheromone antagonist in A. lineolatus. Furthermore, when (E)-cinnamaldehyde was field tested in combination with the sex pheromone, there was no increase in male catch, but the combination attracted both males and females. Prospects for practical application are discussed.Supplementary InformationThe online version contains supplementary material available at 10.1007/s10886-021-01273-y.

sex_pheromone_of_the_alfalfa_plant_bug,_adelphocoris_lineolatus:_pheromone_composition_and_antagonis

Introduction<!>Insects for Experiments<!>Volatile Collection from Live Females<!>Coupled Gas Chromatography-Electroantennography (GC-EAG)<!>Identification of EAG-Active Compounds<!>Chemicals<!>Field Experiment with Live Virgin A. lineolatus<!>Field Experiments with Synthetic Compounds<!><!>Statistics<!><!>Discussion<!>

<p>Plant bugs (Heteroptera: Miridae) represent the most species-rich family of true bugs. Several species are pests, and some have an extremely wide spectrum of hosts (e.g., Holopainen and Varis 1991). Due to new pest control technologies and recent changes in regulation, there has been a marked and continuous decrease in the use of broad-spectrum insecticides in agriculture. As a consequence, pests considered previously to be of minor importance can become more important, as observed for genetically engineered lepidopteran-resistant crops, such as Bt-cotton (Lu et al. 2010). Furthermore, this effect may reach beyond the intended crop. Lu et al. (2010) found that broad-spectrum insecticide sprayings may result in 'sink' populations of a particular pest that, without such treatments, can reach high abundance and create 'source' populations, resulting in higher levels of damage in other crops. Adelphocoris species are among those pests that have gained increasing economic importance with the decrease in broad-spectrum insecticide use (Lu et al. 2008).</p><p>The alfalfa plant bug, Adelphocoris lineolatus (Goeze), occurs widely in the Palearctic, where it is a major pest of alfalfa, Medicago sativa L., Fabaceae (Benedek et al. 1970); several other potential hosts have also been reported (Golledge 1944; Peterson et al. 1992). Currently, the most serious economic impact of Adelphocoris spp., including A. lineolatus, is the damage caused to Bt-cotton in China (Lu et al. 2008; Wu et al. 2002).</p><p>Partially due to their increased economic importance, several reports on the chemical ecology of Adelphocoris species have been published recently, including pheromone identification of major pests of Bt-cotton in China, such as A. fasciaticollis Reuter (Zhang et al. 2015b), A. suturalis (Jakovlev) (Zhang et al. 2016) and A. lineolatus (Zhang et al. 2015a). The aim of such work is to develop species-specific detection and monitoring traps to aid pest management. Among these species, A. lineolatus has the widest distribution in the Palearctic, and has also been introduced to the Nearctic. Zhang et al. (2015a) identified hexyl butyrate, (E)-2-hexenyl butyrate and (E)-4-oxo-2-hexenal as components of the female sex pheromone of an east Asian population. As found in other pest insects with a wide distribution, pheromone composition can vary throughout a pest's range. One example of this is Agrotis segetum (Denis and Schiffermüller), in which sex pheromone composition of populations in different geographic regions consist of different combinations of components (Tóth et al. 1992). Therefore, we wished to analyze a central European population of A. lineolatus to determine whether its sex pheromone differed from that of the east Asian population.</p><p>In addition, we wanted to test the behavioral effects of other chemicals on a central European population. (E)-4-Oxo-2-hexenal is a common component of plant bug pheromones that can be degraded by environmental conditions, including heat, light and oxidation. In previous studies on the chemical ecology of Miridae, special caution has been taken with its use. For instance, in one study, the compound was applied in separate bait dispensers and replaced on a daily basis to maintain activity (Byers et al. 2013). Yasuda and Higuchi (2012) reported that the level of (E)-4-oxo-2-hexenal decreased quickly in dispensers but that an increased dose attracted more male Stenotus rubrovittatus (Matsumura), another pestiferous plant bug species. Thus, in our study, we decided to test the compound in two dosages. We also tested (E)-2-hexenal, a more stable compound than (E)-4-oxo-2-hexenal, for possible analogous activity. Plant volatiles can affect sex pheromone production and activity in some insect species (Landolt and Phillips 1997). For example, in a closely related plant bug, Lygus rugulipennis (Poppius), host plant odors elicited increased sex pheromone production in females (Frati et al. 2009). Therefore, we wished to assess potential interactions between the sex pheromone and (E)-cinnamaldehyde, a floral volatile that attracts A. lineolatus (Koczor et al. 2012).</p><p>Overall, the major aims of this study were: 1) determine the pheromone composition of A. lineolatus in a population from central Europe, 2) test responses to two dosages of (E)-4-oxo-2-hexenal, 3) test responses to (E)-2-hexenal for potential analogous activity, and 4) assess responses to combinations of sex pheromone and (E)-cinnamaldehyde.</p><!><p>Virgin A. lineolatus males and females were reared in the laboratory at 18:6 light:dark photoperiod, 26 °C and ca. 40% RH. Nymphs were collected by sweep-netting at alfalfa fields in Halásztelek, Pusztazámor and Tököl (Hungary), and taken to the laboratory, where they were reared on green bean pods in 12.5 × 17.5 cm glass jars covered with fine mesh. Freshly molted adults were removed from the rearing containers, identified, sexed and kept separate to ensure they were virgin when used. Adult bugs were kept in the same conditions as nymphs.</p><!><p>As field cage experiments with live bugs indicated the presence of a female-produced sex pheromone, headspace collections were performed with single A. lineolatus females on green bean pods, and with green bean pods alone as a control, for 1 d (20–24 h) or 3 d (71–72 h). For preparation of headspace collections, two methods were used. The bugs and green bean pods were placed in 200 ml glass containers of a closed-loop stripping apparatus (CLSA, Boland et al. 1984), equipped with a DC12/16NK vacuum pump (Erich Fürgut GmbH, Tannheim, Germany) with an airflow of ca. 5.0 l.min−1 and a collection filter containing 5 mg activated charcoal (Brechbühler AG, Schlieren, Switzerland). Trapped volatiles were eluted from the charcoal filter with 25 μl dichloromethane (Merck KGaA, Darmstadt, Germany).</p><p>To determine pheromone emission, dynamic headspace collection (air entrainment) (Birkett 2010) was carried out with single A. lineolatus females on green bean pods for 24 h under a 14:10 L:D photoperiod, 20 °C and ca. 50% RH. The material to be sampled was placed in a 380 ml glass jar, and activated charcoal-filtered (Capillary-Grade Hydrocarbon Trap; Thames Restek Ltd., High Wycombe, UK) air pumped (Pye volatile collection kits, Kings Walden, UK) through the inlet port at 600 ml.min−1. Air subsequently passed over the material in the jar and headspace volatiles were adsorbed on Porapak Q filters (0.05 g, 50/80 mesh; Supelco) on the outlet port, through which air was drawn at 500 ml.min−1. All connections in the air entrainment setup used PTFE tubing. Prior to entrainment, Porapak Q filters were washed with diethyl ether and conditioned at 132 °C for 2 h with an activated charcoal-filtered nitrogen stream. Entrained volatiles were eluted with 750 μl redistilled diethyl ether and stored in 1.1 ml glass microvials at −20 °C until analysis. Glass jars were washed with detergent (Teepol), acetone and distilled water, and baked overnight at 140 °C. Sampling was replicated four times.</p><!><p>Female air entrainment extracts were tested for electroantennographic activity on male antennae by coupled GC-EAG using an Agilent 6890 N gas chromatograph equipped with a DB-WAX column (30 m × 0.32 mm i.d.). Helium was the carrier gas and injection was performed in the splitless mode. The column oven temperature program started at 60 °C and increased to 220 °C at 10 °C.min−1. The effluent was split between the flame ionization detector (FID) and a heated transfer line to the EAG apparatus. For each test, we co-injected 1 μl aliquots of air entrainment extracts and 10 ng tetradecyl acetate (internal standard) in 1 μl of dichloromethane. For EAG, the male antenna was freshly cut at the base from a live bug, and the tip of the last segment excised to ensure a good connection. The antenna was mounted between two glass capillaries containing Ringer solution. One electrode was grounded, while the other was connected to a high-impedance DC amplifier (IDAC-232, Ockenfels Syntech GmbH, Kirchzarten, Germany). A compound was defined as EAG-active if it evoked an antennal response, distinguishable from background noise, in at least three runs.</p><!><p>For identification of electrophysiologically active compounds from air entrainment samples, a Hewlett-Packard 5890 series II GC, fitted with a capillary DB-WAX column (30 m × 0.32 mm i.d., 0.5 μm film thickness; J&W Scientific, Folsom, CA) and a cool on-column injector, was coupled to a mass spectrometer (Hewlett-Packard 5972). Ionization was by electron impact at 70 eV. The column oven temperature was maintained at 40 °C for 1 min and then increased at 5 °C.min−1 to 250 °C (hold time 17.2 min). The carrier gas was helium. Tentative identification by gas chromatography/mass spectrometry (GC/MS) was confirmed by comparing retention indices of peaks with those of synthetic standards, and by peak enhancement on GC by co-injection with authentic compounds (Pickett 1990) using an Agilent 7890A GC equipped with a cool on-column injector, FID and a 30 m × 0.32 mm i.d., 0.52 μm film thickness DB-WAX column. The oven temperature was maintained at 30 °C for 0.5 min and then programmed at 5 °C.min−1 to 150 °C for 0.1 min, then at 10 °C.min−1 to 230 °C for 25 min. The carrier gas was hydrogen.</p><p>Quantification of compounds was achieved using the multiple-point external standard method, generating calibration curves from synthetic standards.</p><!><p>Hexyl butyrate, (E)-2-hexenyl butyrate, (E)-cinnamaldehyde and 1-hexanol (≥96% purity as per the manufacturer) were obtained from Sigma-Aldrich Kft (Budapest, Hungary). (E)-4-Oxo-2-hexenal was synthesized as follows. To a solution of 2-ethylfuran (10.00 g, 104.03 mmol) in a mixture of THF (100 ml), acetone (80 ml) and water (40 ml), cooled to −15 °C under nitrogen, was added N-bromosuccinimide (27.78 g, 156.04 mmol), followed by pyridine (16.8 mL, 208.06 mmol). The reaction mixture was stirred for 30 mins before being warmed to 0 °C for a further 3 h. The reaction mixture was poured into 0.5 M HCl and extracted with EtOAc. The combined organics were washed with water, dried (MgSO4) and concentrated under vacuum. The crude product was purified on silica gel (20% EtOAc in pet ether) to give (E)-4-oxo-2-hexenal (4.42 g, 37% yield) as an orange oil.1H-NMR (CDCl3, 500 MHz): 9.79 (d, 1H, J = 7.2 Hz), 6.90 (d, 1H, J = 16.2 Hz), 6.80 (dd, 1H, J = 16.2 and 7.2 Hz), 2.75 (qu, 2H, J = 7.3 Hz), 1.18 (t, 3H, J = 7.2 Hz). 13C-NMR (CDCl3, 500 MHz): 200.38, 193.46, 144.78, 137.30, 34.54 & 7.55. Due to its inherent instability, the compound was stored as a 1:1 solution in dichloromethane at −80 °C until required.</p><!><p>This experiment (Experiment 1) was performed at Pusztazámor, Hungary, at the edge of an alfalfa field from July 15 to August 8, 2013. Four different treatments were applied: three virgin females on a green bean pod, three virgin males on a green bean pod, a green bean pod without insects and a blank control. Traps consisted of a plastic roof (27 × 24 cm) with the upper side covered with aluminum foil (I.S.X.-TRADE Kft., Budapest, Hungary) to prevent insolation. On the roof, a transparent sticky PVC sheet (23 × 36 cm) was attached with pegs, its sticky side facing inward. The bugs and pods were placed in 9.5 × 4 cm cylindrical containers made of transparent PVC foil, and closed at both ends with fine mesh. The containers were fixed to the underside of the roof. At each inspection, bean pods and bugs were replaced with fresh ones. One replicate of each treatment was incorporated into a block, within which individual treatments were 5–8 m apart in a randomized arrangement. The distance between blocks was 10–15 m. The experiment was run with 4 blocks. Traps were inspected twice weekly, with insects caught in the sticky insert removed and taken to the laboratory for identification.</p><!><p>Ternary pheromone baits were prepared as follows: hexyl butyrate, (E)-2-hexenyl butyrate and (E)-4-oxo-2-hexenal were formulated into 0.7 ml polyethylene vials (No. 730, Kartell Co., Italy) in a 5.4:9.0:1.0 ratio, respectively, and capped. Total load of baits was kept at 50 mg. Binary combinations in Experiment 2 were prepared with the same load of the respective compounds. For Experiment 3, 0.1, 1 or 10 mg of 1-hexanol was added to the ternary pheromone blend in a vial. The dispensers were attached to 8 × 1 cm plastic handles for easy handling when assembling the traps. The dispensers were kept in the shade under the roof of traps and equipped with loosely applied aluminum foil to provide protection from light, since (E)-4-oxo-2-hexenal is light-sensitive (Fountain et al. 2014). For Experiment 4, 10 mg of 1-hexanol was added to the ternary pheromone blend either in the pheromone bait or in a separate polyethylene vial. The latter was closed but no shading was added.</p><p>(E)-Cinnamaldehyde, a known attractant for A. lineolatus (Koczor et al. 2012), was also tested as a positive control. Baits were prepared as follows: 100 mg (E)-cinnamaldehyde was loaded onto a 1 cm piece of dental roll (Celluron®, Paul Hartmann AG, Heidenheim, Germany), and placed in a polyethylene bag (ca 1.0 × 1.5 cm, 0.02 mm polyethylene foil, (FS471–072, Phoenixplast BT, Pécs, Hungary). Dispensers were heat-sealed and attached to 8 × 1 cm plastic handles for handling when assembling traps. In the field, polyethylene vial dispensers were replaced at intervals of 4–5 weeks and polyethylene bag dispensers replaced every 3–4 weeks. Previous experience has shown that dispensers do not lose attractiveness over this period (Koczor et al. 2012, 2015).</p><p>For storage, all dispensers were wrapped singly in pieces of aluminum foil and stored at −18 °C until used. For field testing, CSALOMON® VARL+ funnel traps were used (produced by the Plant Protection Institute, CAR, Budapest, Hungary), which have been proven to be suitable for catching plant bugs (Koczor et al. 2012). A small piece (1 × 1 cm) of household anti-moth strip (Chemotox®, Sara Lee; Temana Intl. Ltd., Slough, UK; active ingredient 15% dichlorvos) was placed in the containers to kill captured insects. The experiments were performed in a randomized complete block design, i.e., one replicate of each treatment was incorporated into a block, with individual treatments 5–8 m apart in a randomized arrangement. Distance between blocks was 10–15 m. To avoid positional effects, trap positions were changed fortnightly. As a rule, traps were inspected weekly, and catches brought to the laboratory, where individuals were sexed and determined to species level.</p><!><p>Experiment 2: We tested ternary and binary combinations of hexyl butyrate, (E)-2-hexenyl butyrate and (E)-4-oxo-2-hexenal (Table 1). Traps were set at the edge of an alfalfa field in the vicinity of Cegléd (Hungary) from 12 July to 24 September, 2018, with 4 blocks.</p><p>Experiment 3: In this experiment, we added 1-hexanol to the ternary pheromone blend, containing hexyl butyrate + (E)-2-hexenyl butyrate + (E)-4-oxo-2-hexenal. 1-Hexanol was loaded in a 0.1, 1, or 10 mg dose in the same bait dispensers (Table 1). Traps were set at the edge of an alfalfa field in the vicinity of Cegléd (Hungary) from 12 July to 24 September, 2018, with 4 blocks.</p><p>Experiment 4: We added 1-hexanol to the pheromone blend in the same or in separate dispensers to assess if inhibition of A. lineolatus catches by 1-hexanol was a result of chemical interactions with pheromone constituents (Table 1). Traps were set at the edge of an alfalfa field in Érd-Elvira major (Hungary) from 15 July to 19 September, 2019, with 5 blocks.</p><p>Experiment 5: In this experiment we tested the effect of the addition of (E)-cinnamaldehyde to the ternary pheromone blend (Table 1) on A. lineolatus catches. (E)-Cinnamaldehyde was added in a separate dispenser. Traps were set at the edge of an alfalfa field in the vicinity of Cegléd (Hungary) from 12 July to 24 September, 2018, with 4 blocks.</p><p>Experiment 6: We tested two doses of (E)-4-oxo-2-hexenal in the pheromone blend, as well as whether (E)-4-oxo-2-hexenal could be substituted with (E)-2-hexenal (Table 1). Traps were set at the edge of an alfalfa field in Érd-Elvira major (Hungary) run from 15 July to 19 September, 2019, with 5 blocks.</p><p>Treatments for Experiments 2–6</p><p>*HB hexyl butyrate, E2HB (E)-2-hexenyl butyrate, E4O2H (E)-4-oxo-2-hexenal;'+' indicates the presence of a treatment in an experiment</p><!><p>Trap catch data were tested for normality by Shapiro-Wilk tests. Since experimental data were not normally distributed, nonparametric tests were used. Inspections with low catches (i.e., accounting for <5% of total catches in an experiment) were excluded from the analysis. Catch data were analyzed by Kruskal-Wallis tests, and differences between treatments were evaluated by pairwise Wilcoxon tests with Benjamini-Hochberg correction. Statistical procedures were conducted using R (R Core Team 2016).</p><!><p>Catches of Adelphocoris lineolatus males in traps baited with live virgin A. lineolatus males on green bean pods, live virgin A. lineolatus females on green bean pods, green bean pods alone or unbaited. Treatments marked with the same letter are not different (Kruskal-Wallis test, pairwise comparison by Wilcoxon test with Benjamini-Hochberg correction at p = 0.05) ∑ = total number of A. lineolatus males caught in the experiment (box plot diagram indicating median, minimum, maximum, the 1st and 3rd quartiles of catches of the respective treatments)</p><p>Coupled gas chromatography-electroantennogram (GC-EAG) analysis of a female Adelphocorus lineolatus headspace extract, with bioactive peaks labeled. The extract used for GC-EAG shows a ratio of pheromone constituents different from that from air entrainment samples, which were used for quantitative analysis. FID = flame ionization detector</p><p>Catches of Adelphocoris lineolatus males in traps baited with ternary and binary combinations of hexyl butyrate, (E)-2-hexenyl butyrate and (E)-4-oxo-2-hexenal and unbaited. Treatments marked with the same letter are not different (Kruskal-Wallis test, pairwise comparison by Wilcoxon test with Benjamini-Hochberg correction at p = 0.05) ∑ = total number of A. lineolatus males caught in the experiment (box plot diagram indicating median, minimum, maximum, the 1st and 3rd quartiles of catches of the respective treatments)</p><p>Catches of Adelphocoris lineolatus males and females in traps baited with a ternary pheromone blend and different doses of 1-hexanol (total catch = 45 A. lineolatus)</p><p>*Treatments marked with the same letter are not different (Kruskal-Wallis test, pairwise Wilcoxon test with Benjamini-Hochberg correction at P = 0.05)</p><p>Catches of Adelphocoris lineolatus males in traps baited with a ternary pheromone blend, with addition of 1-hexanol, and unbaited. The 1-hexanol was added either in the same or a separate dispenser. Treatments marked with the same letter are not different (Kruskal-Wallis test, pairwise comparison by Wilcoxon test with Benjamini-Hochberg correction at p = 0.05) ∑ = total number of A. lineolatus males caught in the experiment (box plot diagram indicating median, minimum, maximum, the 1st and 3rd quartiles of catches of the respective treatments)</p><p>Catches of Adelphocoris lineolatus males and females in traps baited with a ternary pheromone blend, (E)-cinnamaldehyde, their combination and unbaited. Treatments marked with the same letter are not different (Kruskal-Wallis test, pairwise comparison by Wilcoxon test with Benjamini-Hochberg correction at p = 0.05) ∑ = total number of A. lineolatus males/females caught in the experiment (box plot diagram indicating median, minimum, maximum, the 1st and 3rd quartiles of catches of the respective treatments)</p><p>Catches of Adelphocoris lineolatus males in traps baited with a ternary pheromone blend with a standard dose of (E)-4-oxo-2-hexenal, with a 5-fold increased dose of (E)-4-oxo-2-hexenal, with (E)-2-hexenal substituted for (E)-4-oxo-2-hexenal and unbaited. Treatments marked with the same letter are not different (Kruskal-Wallis test, pairwise comparison by Wilcoxon test with Benjamini-Hochberg correction at p = 0.05) ∑ = total number of A. lineolatus males caught in the experiment (box plot diagram indicating median, minimum, maximum, the 1st and 3rd quartiles of catches of the respective treatments)</p><!><p>Our results on central European populations of A. lineolatus confirm the identity of hexyl butyrate, (E)-2-hexenyl butyrate and (E)-4-oxo-2-hexenal as female-produced pheromone components of A. lineolatus, as reported previously from east Asian populations (Zhang et al. 2015a). The relative importance of the compounds identified was also similar in the present study, as binary blends from which either (E)-2-hexenyl butyrate or (E)-4-oxo-2-hexenal were missing were not active, whereas binary combination of these compounds and the ternary blend were. Thus, it appears that populations from central Europe and east Asia are similar with respect to the chemistry of their pheromonal communication.</p><p>The above three compounds are sex pheromone components of several other plant bug species, including L. rugulipennis (Innocenzi et al. 2005) and L. pratensis (Linnaeus) (Fountain et al. 2014), which may occur in the same habitats as A. lineolatus. Fountain et al. (2014) reported that the ratios of the same three sex pheromone components for the closely related Lygus, Lygocoris and Liocoris species, were less important. Thus, it is likely that other means of communication may also be important in mate recognition in A. lineolatus, as found for Lygocoris pabulinus (Linnaeus) (Drijfhout and Groot 2001) and L. rugulipennis (Koczor and Cokl 2014).</p><p>Based on the findings of Yasuda and Higuchi (2012) on S. rubrovittatus, we tested a higher dosage of (E)-4-oxo-2-hexenal in the pheromone blend; however, this did not result in catches of more males. Since a compound may have multiple functions and (E)-4-oxo-2-hexenal is thought to be important in defense (Moreira and Millar 2005), we tested blends in which (E)-4-oxo-2-hexenal was substituted by (E)-2-hexenal, a more stable compound: the substituted blend did not attract males.</p><p>1-Hexanol was also found in air entrainment samples of female A. lineolatus and elicited EAG responses from male antennae. Surprisingly, when the compound was tested in combination with the ternary pheromone blend, it resulted in a decrease in male catch. Subsequent field experiments, in which the compound was tested in the same or separate dispensers showed that this was likely due to a biological response and not to a chemical reaction in the lure. However, we cannot rule out that the compounds might react in the air. In their laboratory study on host plant volatiles, Sun et al. (2013) found that more A. lineolatus adults chose a solvent control over 1-hexanol in Y-tube olfactometer tests, indicating a repellent-like effect, supporting our finding.</p><p>The ecological role of 1-hexanol for A. lineolatus is uncertain. Host plant volatiles are known to affect sex pheromone production and activity in insects (Landolt and Phillips 1997); for instance, in L. rugulipennis a closely related plant bug species, Frati et al. (2009) found that host plant odors evoked increased sex pheromone production in females. Thus, it is possible that a compound indicating unfavorable conditions of a host may negatively affect attraction of males to sex pheromone. Another potential explanation is that the antagonistic effect could have been functionally important in the past. For instance, if an ancestor of A. lineolatus was using 1-hexanol as a pheromone component, the compound could have become antagonistic during speciation as indicative of the ancestral species. Interestingly, 1-hexanol was found in gland extracts of a closely related eastern Asian species, A. suturalis (Zhang et al. 2014). Nevertheless, as air entrainment extracts in this study were prepared from live bugs on green bean pods, the compound may also be connected to other activities, such as feeding.</p><p>Several reports have demonstrated the synergistic effect of plant volatiles on insect attraction to sex pheromones (Landolt and Phillips 1997). This, however, was not the case for A. lineolatus males as addition of the known floral attractant, (E)-cinnamaldehyde, to the pheromone had no significant effect on catch. On the other hand, the presence of the sex pheromone did not affect catch of females to (E)-cinnamaldehyde. This lack of interaction between the chemicals may open up opportunities for monitoring both sexes using a combination of sex pheromone and (E)-cinnamaldehyde.</p><p>Pheromones are important in monitoring or direct control (e.g., mating disruption) of insect pests (Witzgall et al. 2010). Whereas monitoring of plant bugs may prove to be an important tool in agriculture, mating disruption may not be economically feasible, as suggested by Yasuda and Higuchi (2012). In the case of A. lineolatus, one problem could be the instability of (E)-4-oxo-2-hexenal affecting its storage and bait longevity. A further problem for health and safety could be the irritating property of this compound. Substitution of this compound with a more stable alternative may be a solution; however, as our study has shown, more work is needed to screen for feasible substitutes. Finally, 1-hexanol may be suitable for use as a sex pheromone antagonist in, for example, mating disruption. Experiments are underway to assess its potential.</p><!><p>(PDF 115 kb)</p>

PubMed Open Access

Optimization of a Cyclic Peptide Inhibitor of Ser/Thr Phosphatase PPM1D (Wip1)\xe2\x80\xa0

PPM1D (PP2C\xce\xb4 or Wip1) was identified as a wild type p53-induced Ser/Thr phosphatase that accumulates after DNA damage and classified into the PP2C family. It dephosphorylates and inactivates several proteins critical for cellular stress responses, including p38 MAPK, p53, and ATM. Furthermore, PPM1D is amplified and/or overexpressed in a number of human cancers. Thus, inhibition of its activity could constitute an important new strategy for therapeutic intervention to halt the progression of several different cancers. Previously, we reported the development of a cyclic thioether peptide with low micromolar inhibitory activity towards PPM1D. Here, we describe important improvements in the inhibitory activity of this class of cyclic peptides and also present a binding model based upon the results. We found that specific interaction of an aromatic ring at the X1 position and negative charge at the X5 and X6 positions significantly increased the inhibitory activity of the cyclic peptide, with the optimized molecule having Ki = 110 nM. To the best of our knowledge, this represents the highest inhibitory activity reported for an inhibitor of PPM1D. We further developed an inhibitor selective for PPM1D over PPM1A with Ki = 2.9 \xce\xbcM. Optimization of the cyclic peptide and mutagenesis experiments suggest that a highly basic loop unique to PPM1D is related to substrate specificity. We propose a new model for the catalytic site of PPM1D and inhibition by the cyclic peptides that will be useful both for the subsequent design of PPM1D inhibitors and for identification of new substrates.

optimization_of_a_cyclic_peptide_inhibitor_of_ser/thr_phosphatase_ppm1d_(wip1)\xe2\x80\xa0

<!>Chemical Reagents<!>Peptide synthesis<!>Amino acid composition analysis and determination of concentration<!>Protein expression and purification<!>Phosphatase activity<!>Inhibition mechanism<!>Molecular Modeling<!>Circular Dichroism (CD) Spectroscopy<!>Phe substitution at position X1 increases the inhibitory activity of the cyclic peptide against PPM1D<!>The chirality of the Ile residue at position X3 affects PPM1D inhibition<!>Negative charges at positions X5 and X6 are preferred for inhibition<!>Binding of the cyclic peptides to PPM1D is competitive with substrate<!>Substitution of pS with phospho-homoserine confers specificity towards PPM1D over PPM1A<!>Molecular Modeling<!>Arg243 and Lys247 in PPM1D are crucial for phosphatase activity<!>Discussion<!>

<p>Kinases and phosphatases are important regulators of protein function in biological systems and thus constitute good targets for the development of new drugs. While the human genome encodes 518 kinases (1), there are estimated to be only 147 phosphatases; of those, only 40 are serine/threonine phosphatases (1–3). The PP2C family in humans consists of seven monomeric serine/threonine phosphatases (4, 5). This includes PPM1D (also called PP2Cδ or Wip1), which was first identified as induced by wild type p53 after DNA damage (6). Consistent with other members of the PP2C family, PPM1D is a monomeric enzyme that requires divalent cations, either Mn2+ or Mg2+, for catalytic activity and is insensitive to oakadaic acid (7). This phosphatase is composed of two major domains: a highly conserved N-terminal phosphatase domain and a less-conserved, non-catalytic domain at the C terminus (7). The known substrates of PPM1D include several proteins critical for cellular stress responses, namely: p38 MAPK (8), Chk1 (9), Chk2 (10–12), ATM (13), and p53 (9). Dephosphorylation of each of these proteins by PPM1D results in its inactivation. PPM1D is amplified and/or over-expressed in a number of human cancers, such as breast cancer (14–16), neuroblastoma (17), medulloblastoma (18), ovarian clear cell adenocarcinoma (19), and pancreatic adenocarcinoma (20). In addition, PPM1D-null mice show a dramatic tumor-resistant phenotype (21). Thus, inhibition of PPM1D activity could constitute an important new strategy for therapeutic intervention to halt the progression of several different cancers.</p><p>PPM1D dephosphorylates phosphoserine (pS) or phosphothreonine (pT) as a part of two different peptide motifs: pT-X-pY (22) and pS/pT-Q (23). In a study of the pT-X-pY motif, we observed that PPM1D preferentially dephosphorylates pT from a diphosphorylated sequence compared to a monophosphorylated one and that amino acids adjacent to the motif do not significantly affect the substrate specificity (24). Additionally, it was found that pS substitution of the pT in the pT-X-pY sequence from p38 MAPK resulted in PPM1D inhibition. This result raised the possibility that a pS-substituted peptide could be developed as an effective inhibitor of PPM1D phosphatase activity. After extensive optimization, a cyclic thioether peptide of sequence M-pS-I-pY-VAC was identified with a Ki of approximately 5 μM (Figure 1).</p><p>Combining this result with mutagenesis studies of the protein and a NMR solution structure of the cyclic peptide, we were able to propose a structural model of the complex at the active site (24). For this, we developed a homology model of PPM1D from the crystal structure of the related PPM1A (PP2Cα) protein in humans (25). Although this model incorporated the pS and pY residues of the cyclic peptide in key charge-charge interactions with the protein, it did not provide obvious roles for the Met, Ile and Ala residues. We suggested that at least some of them may interact with a relatively long loop adjacent to the active site (residues 237–268) that is unique to PPM1D. Unfortunately, due to its absence in the PPM1A template, the loop could not be included in our homology model. Although the precise start of the insertion is uncertain, a sub-segment of it (residues 245–268) has been dubbed the "B-loop" by Chuman et al. (26) because of the preponderance of positively charged amino acids. Interestingly, the B-loop sprouts from a conserved sub-domain (residues 165–194 of PPM1A) that has been designated a flap, presumably due to its different conformations observed in the crystal structures of both eukaryotic and prokaryotic homologues (27–33) (Supporting Information Figure S1). It has been postulated that movement of the flap regulates enzymatic activity through modulation of the binding of a third metal ion, substrate recognition and/or steric availability to the catalytic site. Thus, that the B-loop is part of the flap sub-domain in PPM1D supported the idea that it plays a role in regulation.</p><p>In the current work, we describe the development of a new cyclic peptide with significantly enhanced inhibitory activity. While keeping the pS-I-pY core residues, the sidechains at the other positions were extensively modified to identify an optimized molecule with a Ki for PPM1D of approximately 100 nM, reflecting a 50-fold improvement over the starting peptide. Combining the results of the optimization with the results of strategic mutations in the B-loop enabled further refinement of the model of the PPM1D catalytic domain. We also describe improvements obtained in selectivity for PPM1D over PPM1A by modification of the pS residue of the cyclic peptide. These results further demonstrate the possibility of developing selective inhibitors for this neglected class of phosphatase targets.</p><!><p>Resins and N-α-Fmoc-protected amino acids, including phosphorylated amino acids and Fmoc succinimide, were purchased from Novabiochem (San Diego, CA). N-α-Fmoc-protected unusual amino acids, including L-2-amino-4-phosphono-4,4-difluorobutyric acid (F2Pab), were obtained from Anaspec (San Jose, CA). Chloroacetic anhydride, solvents, and the amino acid standard solution for amino acid analysis were obtained from Sigma-Aldrich (Milwaukee, WI).</p><!><p>Peptides were synthesized by the solid phase method utilizing 9-.fluorenylmethoxycarbonyl (Fmoc)/tert-butyl or tert-butoxycarbonyl (Boc)/benzyl chemistry. Trityl (Trt) and p-methoxybenzyl groups were utilized as Cys sidechain protecting groups for Fmoc- and Boc-chemistry, respectively. Peptides were assembled on NovaPEG Rink amide resin or Wang resin for Fmoc chemistry or MBHA resin for Boc chemistry. The amino group of O-sulfo-L-serine was protected by an Fmoc group (34). Phosphorylated Ser and Tyr were incorporated as Fmoc-Ser[PO(OBzl)OH]-OH and Fmoc-Tyr(PO3H2)-OH, respectively. In the synthesis of peptides 46 and 48, O-Sulfo-L-serine and homoserine (Hse) were assembled as Boc-Ser(SO3H)-OH and Fmoc-Hse(Trt)-OH on MBHA resin, respectively. The coupling reactions were carried out by means of the HBTU-HOBt method. To introduce a phosphate group onto the hydroxyl group of the Hse residue in peptide 48, the trityl group was removed by treatment with trifluoroacetic acid/dichloromethane/triisopropylsilane (1/94/5 v/v) after the amino group of the Phe residue at position X1 was protected with a Boc group. Phosphorylation was achieved using dibenzyl-N,N′-diisopropylphosphoramidate and anhydrous tert-butylhydroperoxid. The N terminus of each peptide was chloroacetylated using chloroacetic anhydride for downstream cyclization. Cleavage of the peptide from the resin was achieved with trifluoroacetic acid/water/triisopropylsilane (92.5/5/2.5 v/v) or trifluoromethanesulfonic acid/trifluoroacetic acid/water/triisopropylsilane (10/82.5/5/2.5 v/v) for 2 h at room temperature. After removing the resin by filtration, the filtrate was concentrated by flushing with nitrogen gas and crude peptides were precipitated by diethyl ether. The crude peptides were then dissolved in 1% triethylamine-containing water (pH 8–9, approximately 3 mM) and stored at room temperature overnight to allow cyclization to occur. The cyclization reaction was quenched by acidifying using acetic acid. Crude peptides were purified using reversed-phase high-performance liquid chromatography (RP-HPLC) on a preparative C4 column (BioAdvantage Pro 300, Thomson Liquid Chromatography) with a water-acetonitrile solvent system containing trifluoroacetic acid. Purified peptides were characterized by matrix-associated laser desorption ionization time-of-flight mass spectrometry (MALDI micro MX, Waters) and RP-HPLC on an analytical C18 column (Eclipse XDB-C18, Agilent). The purity of all peptides was found to be > 95%.</p><!><p>Amino acid composition analysis was carried out using the phenylthiocarbamyl method (35, 36). Peptide stock solutions in water (ca. 5 mM) were transferred to a glass tube for hydrolysis, resulting in ca. 10 nmol per glass tube. The peptide was hydrolyzed with 6 N hydrogen chloride solution at 105 °C in a sealed tube for 24 h. After cooling, the solution was evaporated with nitrogen gas and a water bath. A solution (10 μl) of ethanol:water:triethylamine (2:1:1 v/v) was added to each tube. After removing the solution in vacuo, a phenylisothiocyanate-containing solution (ethanol:water:triethylamine:phenylisothiocyanate, 7:1:1:1 v/v) was added to the lyophilized sample. The tube was left for 30 min at room temperature. Subsequently, the sample was redried in vacuo to remove excess reagent. The dried sample was dissolved in 1 ml 60 mM sodium acetate buffer (pH 6.0) and 10% of the total was analyzed by RP-HPLC as follows: BioAdvantage Pro 300 C18 5 μm (4.6 × 250 mm) column; solution A, 60 mM sodium acetate buffer (pH 6.0); solution B, acetonitrile; gradient, 5–50%B/0–30 min. Identification and quantitation of each amino acid in the solution was performed on the basis of retention times and peak area integration as compared with that of an amino acid standard solution. Peptide concentration was calculated on the basis of a standard curve made by the peptide stock solution described above.</p><!><p>The cloning and purification of the human PPM1D catalytic domain (residues 1–420) containing an N-terminal His-tag was performed as previously described (22). The plasmid containing the catalytic domain was used for construction of mutants using the QuickChange protocol as described by the manufacturer (Stratagene, La Jolla, CA). The sequences of the oligonucleotides used for PPM1D R243A were as follows: 5′-GAGTGTAATGAACAAGTCTGGGGTTAACGCTGTAGTTTGGAAACGACCTCGACTC -3′ and 5′-GAGTCGAGGTCGTTTCCAAACTACAGCGTTAACCCCAGACTTGTTCATTACACTC-3′. An HpaI restriction enzyme site was created in the oligonucleotides (underlined) to facilitate screening of mutant clones. For PPM1D K247A, the following oligonucleotides were used: 5′-CTGGGGTGAATCGTGTAGTTTGGGCCCGACCTCGACTCACTCACAATGGACCTG-3′ and 5′-CAGGTCCATTGTGAGTGAGTCGAGGTCGGGCCCAAACTACACGATTCACCCCAG-3′. An ApaI site (underlined) was created in the oligonucleotides for mutant screening.</p><p>Individual mutant plasmids were co-transformed with pACYC- Duet (Novagen, Madison, WI) expressing Skp and DsbC proteins to improve solubility of the proteins into BL21-star competent cells (Invitrogen, Carlsbad, CA). Transformed colonies were selected on LB plates containing ampicillin (100 μg/mL) and chloramphenicol (15 μg/mL). Cells were grown (800 mL) to mid-log phase in Circle Grow (Q-Biogen, Irvine, CA) at 37 °C before transfer to 16 °C. After incubation for approximately 15 min, cells were induced with 0.5 mM isopropyl 1-thio-α-D-galactopyranoside for 16 h. Cells were harvested and the His-tagged mutant PPM1D proteins purified by metal affinity chromatography, as described for the wild type protein (22). Activity of the purified protein was confirmed by measuring the kinetic properties for dephosphorylation of a substrate peptide from ATM ([S1981pS] human ATM(1976–1986)-GY).</p><!><p>Phosphatase activities of wild type and mutant PPM1D were measured by a malachite green/molybdate based assay (Millipore), as described previously (22). A substrate peptide ([S1981pS] human ATM(1976–1986), AFEEG-pS-QSTTIGY or [T180pT, Y182pY] human p38α MAPK(175–185), TDDEM-pT-G-pY-VAT) was incubated with 60 ng of PPM1D in 50 mM Tris-HCl pH 7.5, 0.1 mM EGTA, 0.02% 2-mercaptoethanol, 40 mM NaCl, and 30 mM MgCl2 for 7 min at 30 °C, and the amount of free phosphoric acid released from the substrate was detected by measuring the absorbance at 650 nm from the molybdate:malachite green:phosphate reaction complex Determination of the kinetic parameters Km and Vmax were performed with a range of peptide concentrations.</p><p>Inhibitory activity of the cyclic peptides was measured as previously described (22). PPM1D was incubated with 30 μM substrate peptide ([S1981pS] human ATM(1976–1986) or [T180pT, Y182pY] human p38α MAPK) and a cyclic peptide under the same conditions as the phosphatase activity assay described above. For most assays, peptides were included at 10 μM, with some peptides tested also at 1 μM. Data was expressed as mean ± S.D. Statistical analysis of the inhibitory activity was performed by Student's t-test with p<0.05 considered statistically significant. The determination of the apparent inhibitory constant Ki was performed with a range of cyclic peptide concentrations. For most peptides, the concentration was determined from the powder weight of the cyclic peptide; the peptide concentration of peptides 1 and 37 was calculated using amino acid composition analysis.</p><p>PPM1A phosphatase activity was measured using the malachite green/molybdate assay described above. Cyclic peptides were incubated with 12 ng or 24 ng of PPM1A under the same conditions described above. For assays of PPM1A inhibition, cyclic peptide 40 (50 μM) was used as the substrate.</p><!><p>The mechanism of inhibition of cyclic thioether peptide 37 was estimated from the initial velocities of dephopshorylation measured at different concentrations of cyclic peptide and [S1981pS] human ATM(1976–1986) peptide, as described above. The observed initial velocities were graphed on double-reciprocal plots and fitted by the Lineweaver-Burk equation (eq. 1) using GraphPad Prism 4 (GraphPad Software, La Jolla, CA).</p><p>In equation 1, V is the reaction velocity, Km is the Michaelis–Menten constant, Vmax is the maximum reaction velocity, and [S] is the substrate concentration.</p><!><p>The molecular model of PPM1D was updated from that used previously (24). A new sequence alignment with the PPM1A structural template is presented in Supporting Information Figure S2, which also depicts the 7 regions of conservation defined by Chuman et al. (26). The only changes to the previous model were to regions I and VII, which now make all regions consistent with the model of Chuman et al. (26). Although changes to the model in region I involve residues that comprise the active site, the sequence changes are conservative and the impact on the predictive properties of the model are minimal. Specifically, in the previous model PPM1D residue Arg76 was at the position of the functionally important, phosphate-binding Arg33 residue of PPM1A (25, 37). In the current model, this is replaced by PPM1D residue Arg18. The change also added metal-binding PPM1D residue Glu22, which substitutes for PPM1A residue Glu37. However, this residue is on the opposite side of the active site from which the substrates approach and bind. Addition of PPM1D residues Lys19 and Tyr20 to positions proximal to (above) the metal center is significant, as they feature in the optimization of the cyclic peptide inhibitor (see Results).</p><p>Models of the cyclic peptide inhibitor bound to PPM1D were energy minimized with the CHARMM software package (38). The backbone of the cyclic peptide was the same as in the structure determined by NMR in our previous study (24). Residue topology files for the different sidechains tested on the inhibitor were developed manually using the "all22_prot" force field of MacKerell et al. (39). Since our focus was on the interactions with substrate or inhibitors, we did not attempt to model the inserted loops and deletions of the PPM1D homology model distant from the active site.</p><p>Structural similarity searches of the Protein Data Bank (40) were conducted with the VAST conformational search engine (41). Structural alignments, conformational adjustments and figures were generated with the UCSF Chimera molecular modeling software package (42). Sequence alignments and figures were made with BioEdit Sequence Alignment Editor (43).</p><!><p>The CD spectra of wild type and mutant PPM1D were recorded on a JASCO J-715 spectropolarimeter (Easton, MD) with a cylindrical cell of 1 mm path length at room temperature. The CD cell was washed with an aqueous NaOH solution before each new measurement to remove any material that might have adhered to the inner surface. Sample solutions (1.8–2.1 μM PPM1D in 50 mM Tris-HCl, 30 mM MgCl2, 5% glycerol, pH 7.5) were prepared 5 min before measurement. All spectra shown are the average of eight repeated measurements obtained by collecting data from 260 to 190 nm at 0.2 nm intervals with a response time of 2 sec for each point. The results are expressed as the mean residue ellipticity.</p><!><p>Previously, we identified a cyclic thioether peptide (M-pS-I-pY-VAC) (peptide 1) (Figure 1) that functions as an inhibitor of the Ser/Thr phosphatase PPM1D (24). In that peptide, only the X3 position between the pS and pY residue was optimized. Here, we explored the structure-activity relationship for each of the other positions of the peptide with the goals of improving the inhibitory activity and learning more about PPM1D substrate specificity. Initially, position X1 (Figure 1) was varied by replacing the Met of peptide 1 with different amino acids, and then the inhibitory activity of the analogs were determined. Inhibitory activity was calculated from the ability of the cyclic peptide to suppress dephosphorylation of pS in a peptide derived from ATM (residues 1976–1986 with phosphorylation of Ser1981) using a malachite green/molybdate assay to detect the amount of free phosphoric acid in solution. Performing this assay in the presence of 10 μM peptide 1, we observed a 30% decrease in PPM1D dephosphorylation activity (Figure 2A), resulting in a Ki value for peptide 1 of 5.0 ± 1.7 μM (Figure 2B). This value is 5-fold reduced as compared with the previous report of this inhibitor (24). However, in that study the peptide concentration was calculated from pY absorbance at 280 nm with 640 M−1 cm−1 as the molar extinction coefficient (24). As reported, the molar extinction coefficients for pY range from 600–700 M−1 cm−1 at pH 7.0 (44) and they are strongly pH dependent (45). In this study, we determined the cyclic peptide concentration from the dry weight, which by amino acid composition analysis we found to be more accurate. The concentration of peptide 1 determined in this way was approximately 6-fold higher than the concentration calculated from the molar extinction coefficient used previously, resulting in overestimation of the Ki value. Due to the difficulty associated with purifying large quantities of PPM1D, each new inhibitor were tested at 10 μM, with some also used at 1 μM for greater discrimination of the inhibitors.</p><p>In our original model of peptide 1 bound to PPM1D, the Met sidechain of peptide 1 was proximal to Arg76 (24). Thus, we initially hypothesized that mutation of the Met residue to an acidic residue would allow formation of a charge-charge interaction that could improve the inhibitory activity of the cyclic peptide. In contrast to the hypothesis, a peptide in which Glu was substituted at position X1 (peptide 2) abrogated inhibitory activity of the peptide at 10 μM and also functioned as a weak substrate for PPM1D (Figure 2A). Further substitution at the X1 position with a Lys residue (peptide 3) or a Pro residue (peptide 4) also did not significantly alter the inhibitory activity compared to peptide 1. In contrast, a Phe-substituted analog (peptide 5) significantly decreased PPM1D phosphatase activity in the presence of 10 μM peptide (Figure 2A). From measurements at several concentrations of peptide 5, we determined that the Ki value of peptide 5 was 1.1 ± 0.3 μM (Figure 2B), about 5-fold more potent than the original peptide 1. A second cyclic peptide terminating in a carboxylic acid group (peptide 6) rather than an amide as in peptide 5 showed a similar inhibitory activity (Ki = 0.73 ± 0.18 μM, Figure 2B) as peptide 5, indicating that a negative charge at the extreme C terminus does not affect inhibitory activity.</p><p>To further explore the role of the Phe residue in the X1 position for inhibitory activity, a series of peptides was synthesized containing Phe derivatives and their inhibitory activities were determined (Figure 2C). Substitution of position X1 with halogenated phenylalanine derivatives (peptides 7, 8, 14, and 15) did not significantly change the inhibitory activity as compared with peptide 5. Likewise, substitution of the Phe at position X1 with 4-nitrophenylalanine (peptide 11), Tyr (peptide 12), Trp (peptide 13), and 4-ethoxyphenyalanine (peptide 21) did not greatly improve the inhibitory activity relative to peptide 5. These results indicate that addition of a functional group to the 3- and/or 4-position of the aromatic ring does not affect the interaction between the cyclic peptide and PPM1D. In contrast, peptides 9 and 10, containing basic functional groups on the Phe (4-aminophenylalanine and 4-guanididophenylanalnine, respectively) showed decreased inhibitory activities at 10 μM (Figure 2C). These results indicate that a basic functional group at position X1 disrupts the interaction of the cyclic peptide with PPM1D.</p><p>To determine if the position of the aromatic ring of the Phe had an effect on inhibitory activity, position X1 was substituted with N-benzylglycine (peptide 17), homophenylalanine (peptide 18), and L- and D-Tic [(S)- and (R)-1,2,3,4-tetrahydroisoquinoline-3-carboxylic acid] (peptides 19 and 20, respectively) (structures in Supporting Information Figure S3). The inhibitory activities of these peptides were reduced relative to peptide 5 at 10 μM (Figure 2C). Peptide 16, containing beta-cyclohexylalanine (Cha), which has a cyclohexyl ring instead of the aromatic ring of Phe, also showed decreased activity as compared with the Phe-substituted peptide, with approximately 50% inhibition of dephosphorylation observed in the presence of 10 μM peptide 16. Thus, these results suggest that a simple phenyl ring at position X1 of the peptide is important for inhibition, but additional groups attached to the aromatic ring do not further enhance activity.</p><!><p>Our previous work showed that the Ile residue at position X3 is critical for inhibition (24). However, the importance of the stereochemistry of the Ile was not investigated. Therefore, D-Ile and L-allo-Ile (Figure 3A) substituted peptides (22 and 23, respectively) were synthesized, and their inhibitory activity was measured. The D-Ile peptide 22 showed greatly reduced inhibitory activity (Figure 3B). In contrast, the Ki value of the L-allo-Ile analog peptide 23 was 1.1 ± 0.3 μM, similar to that of peptide 5, which has L-Ile (Figure 3B). These results indicate that the chirality at the beta position of the Ile is not important for inhibition, while that of the alpha position is important.</p><!><p>Figure 4 shows the inhibitory activities of X5- and X6-substituted cyclic peptides in which the Val and Ala residues were modified to aromatic, acidic, basic, neutral, and aliphatic amino acids. In position X5, the substitution of Val in the original peptide with Lys (peptide 25) or Phe (peptide 26) reduced the inhibitory activity of the cyclic peptide (Figure 4A). Substitution with Leu (peptide 27), Met (peptide 28), Ile (peptide 29), or Cha (peptide 30) did not have any significant effects (Figure 4A). In contrast, a peptide containing Glu (peptide 24) at position X5 showed improved inhibitory activity, decreasing PPM1D phosphatase activity by >90% and > 60% at 10 and 1 μM, respectively.</p><p>Substitutions at position X6 with hydrophobic amino acids, as in peptides 33 (Phe), 34 (Leu), and 35 (Met), either did not have any significant effect or significantly reduced inhibitory activity as peptide 34 (Figure 3B), while incorporation of Lys at position X6 (peptide 32) also significantly reduced inhibitory activity (Figure 4B). In contrast, increased inhibitory activity was observed when a Glu residue was put in position X6 (peptide 31), with PPM1D dephosphorylation activity decreasing by 95% and 80% when the cyclic peptide was used at 10 μM and 1 μM, respectively. Combined with the results of the screen at position X5, these data suggest that an acidic residue is preferred at positions X5 and X6 for PPM1D inhibition. Indeed, a cyclic peptide containing Gln at position X6 (peptide 36) did not have improved inhibitory activity relative to peptide 5 (Figure 4B). Moreover, the improvement associated with introduction of an acidic residue was moderately larger in the X6 position as compared with the X5 position.</p><p>Finally, we determined the effect of having acidic residues at both X5 and X6 on inhibitory activity. As expected, inclusion of Glu residues at both positions X5 and X6 further increased the inhibitory activity of the peptide (peptide 37). The Ki value of peptide 37 was 0.20 ± 0.02 μM, an improvement of 4-fold relative to peptide 5 (Figure 4C). A similar Ki value for this peptide was observed in the presence of a different substrate derived from human p38 MAPK (residues 175–185, [T180pT, Y182pY]) (Figure 4E), demonstrating that PPM1D inhibition was not dependent on the substrate motif. To evaluate whether any acidic residue would improve the inhibitory activity or if a Glu residue was specifically required, peptides were tested in which X5 or X6 were individually substituted with Asp instead of Glu (peptides 38 and 39, respectively), X5 and X6 were both Asp (peptide 40), and X5 and X6 were both homoglutamic acid (Aad, peptide 41). All peptides containing an acidic group at one or both of positions X5 and X6 showed similar inhibitory activities (Figure 4C), although peptide 40, containing two Asp residues, did show slightly improved activity relative to peptide 37, which has two Glu residues. The Ki value for peptide 40 was found to be 110 nM, with 10- and 50-fold increased relative to peptides 5 and 1, respectively (Figure 4D). In contrast, when positions X5 and X6 were substituted with Gly residues (peptide 42), the inhibitory activity was not affected as compared with peptide 5 (Figure 4C). Thus, a negative charge at positions X5 and X6 is critical, although the carbon chain length of the sidechain with the acidic residues does not seem to affect inhibitory activity.</p><p>To investigate if the stereochemistry affects the inhibitory activity, a cyclic peptide containing D-Asp at positions X5 and X6 (peptide 43) was synthesized and its inhibitory activity was estimated (Figure 4C). The Ki value of peptide 43 was found to be 1.8 ± 0.8 μM, approximately 20-fold lower than peptide 40, containing L-Asp at positions X5 and X6 (Figure 4C). Thus, the chirality at positions X5 and X6 affects the inhibitory activity, however the effect is not as large as that observed at position X3.</p><!><p>Having identified an optimized inhibitor, we next investigated its mechanism of inhibition to better understand its binding site on PPM1D. Figure 5 shows the double-reciprocal plot of PPM1D dephosphorylation of human ATM(1976–1986) [S1981pS] in the absence or presence of cyclic peptide inhibitor 37. The inverse of V, the rate of the PPM1D-mediated reaction, is shown on the vertical axis, while the inverse of the substrate concentration is plotted on the abscissa. Fitting of the data using the Lineweaver-Burk equation provided the maximum reaction velocity Vmax from the inverse of the intercept of the vertical axis and the Michaelis constant Km, which indicates the binding affinity of the substrate to PPM1D, from the inverse of the intercept on the abscissa. The slope of the best fit line is equal to the value of Vmax/Km. The data show that the value of Vmax/Km changed at different cyclic peptide concentrations, whereas the Vmax value was independent of the cyclic peptide concentration. These results indicate that the presence of the cyclic peptide decreased the substrate binding affinity in PPM1D but not the rate of reaction. Thus, these results demonstrate that the cyclic peptide is a competitive inhibitor of PPM1D and, as such, binds to the catalytic site.</p><!><p>As specific inhibitory activity is critical for drug development, we assessed whether peptide 40 was specific for PPM1D. Previously, we found that peptide 1 was recognized as a substrate by PPM1A (PP2Cα) (Km = 11 μM) (24). Similarly, PPM1A dephosphorylated 40 with Km = 115 ± 32 μM, 10-fold less than peptide 1 (Figure 6A). We synthesized a series of peptides containing non-hydrolyzable mimetics of phosphoserine to try to identify a peptide that would not be recognized by PPM1A as either a substrate or an inhibitor. We first substituted the pS with L-2-amino-4-phosphono-4,4-difluorobutyric acid (F2Pab), in which the bridging oxygen of the phosphate group in pS is replaced with a CF2-moiety. In contrast to previous results in which a F2Pab-substituted peptide was partially active against PPM1D (5-fold weaker than the parent peptide) (24), peptide 44, cyclo(F-F2Pab-I-pY-DDC)-amide, did not show inhibitory activity against PPM1D at 25 μM (Figure 6B). Peptide 45, containing Aad, also did not have any inhibitory activity towards PPM1D (Figure 6B). Similarly, peptides containing Ser(SO3H) in place of pS (peptide 46) and the retro-inverse peptide 47, cyclo[(D-Asp)-(D-Asp)-(D-pY)-(D-I)-(D-pS)-(D-F)-C]-amide, showed no significant inhibitory activity toward PPM1D at 25 μM (Figure 6B). In contrast, when the pS was substituted with phosphorylated homoserine (pHse, peptide 48), the resultant peptide showed good inhibitory activity toward PPM1D, with Ki = 2.9 ± 0.5 μM (Figure 6B and 6C). Importantly, peptide 48 did not show any activity, either as a substrate or as an inhibitor, towards PPM1A (Km > 100 μM) (Figure 6D). Thus, although the Ki value of peptide 48 was approximately 30-fold reduced in comparison to peptide 40, it was highly selective. One possible reason for the decreased inhibitory activity of peptide 48 is that the longer sidechain of pHse may disrupt or weaken critical stabilizing interactions between PPM1D and other resides on the cyclic peptide. To determine if this was occurring, we examined cyclic peptides containing (4-Cl)Phe (peptide 49, cyclo[(4-Cl)Phe-Hse(PO3H2)-I-pY-DDC]-amide) or homophenylalanine (Hph, peptide 50, cyclo[Hph-Hse(PO3H2)-I-pY-DDC]-amide) with the hypothesis that the increased length of the Phe sidechain would improve inhibitory activity. While the Ki value of peptide 49 was to be found 4.7 ± 0.8 μM, similar to that of peptide 48, peptide 50 showed decreased inhibitory activity (Ki = 19 ± 2 μM) (Figure 6C). These results suggest that the orientation of the aromatic ring at position X1 of peptide 48 is still critical for inhibition, as described earlier.</p><!><p>To understand the physical basis for the enhanced binding affinity, we undertook to model one of the best cyclic peptides (37) bound to the active site of PPM1D. We knew from our previous experimental and computer docking studies with substrates and peptide inhibitor 1 that the interaction is dominated by electrostatic interactions involving the two phosphate groups (22, 24). Specifically, the phosphate group of the pS residue interacts with the metal center and Arg18 and the phosphate of pY interacts with the sidechains of Lys218 and Lys238 (this latter confirmed by mutagenesis studies (24)). Thus, we maintained these interactions as the core of the updated model. As seen in Figure 7 (and the alternate view in Supporting Information Figure S4), this locates the phenylalanine at position X1 of peptide 37 (F1), next to the non-polar regions of the Lys19 and Tyr20 of PPM1D, thus forming a stabilizing, hydrophobic interaction. This explains why, as demonstrated in Figure 2, highly charged sidechains at X1 lead to reduced inhibition. The fact that Tyr, Trp and some derivatized Phe residues were also favored, and not proline and cyclohexane (peptide 16), suggests that the stabilization is enhanced by aromatic, pi-pi interactions. This is also consistent with the inhibitor being selective for PPM1D over PPM1A, since the corresponding residues in the latter are the non-aromatic Val34 and Glu35. This conformation also buries the sidechain of the X5 peptide residue (E5) on top of PPM1D His107. This would explain why, as shown in Figure 4A, the best inhibition was obtained with negatively-charged residues at this position. Given the electrostatic influence of the carboxylate group, the histidine can be expected to be protonated, thus forming a stabilizing salt-bridge.</p><p>Unfortunately, the effects of substitutions at the X3 and X6 positions (i.e., preference for an isoleucine and a negatively charged residue, respectively) are more complicated to understand. Given the conformation of the bound cyclic peptide, these two sidechains project away from the conserved, catalytic core of PPM1D and are either solvent exposed or interact with the unique B-loop insert of the flap subdomain (24, 43). In an attempt to find a structural template for the B-loop, the Protein Data Bank (40) was searched for all structures similar to PPM1A, the prototypic PP2C family member (41). The results were clustered into 14 unique structures by visual inspection (Supporting Information Table S1). These were then superimposed to identify the range of conformational changes and sequence insertions/deletions in the flap subdomain. The results are displayed in Figure 8A, which aligns the sequence of PPM1D with that of PPM1A and the subset of structures close enough to be structurally aligned, and with homologous human sequences that lack known structure. The most prominent feature in the B-loop region is the conservation of a RV motif (residues Arg243 and Val244 in PPM1D) among all the sequences. It should be noted that human PPM1H (SwissProt: Q9ULR3) and PPM1J (SwissProt: Q5JR12) proteins also have an insertion in the flap region with RV motifs, but the sequences are too divergent to align unambiguously. While this procedure failed to provide a clear template for the B-loop, it did highlight the importance of the RV motif, so as an initial hypothesis, we chose the simplest option of keeping the RV residues of PPM1D structurally aligned with that of PPM1A. As seen in Figure 7, this places Arg243 in an ideal position to interact with the X6 residue (E6) of the cyclic peptide. This explains the preference for a negatively-charged sidechain at this position of the cyclic peptide and the decreased inhibitory effect of positively-charged residues (Figure 4B). Although of a greater distance, this position of Arg243 also allows for interaction with the pY sidechain of the cyclic peptide. In contrast, structurally aligning the similarly conserved Asn254 and Gly255 residues of PPM1D positions them too distant from the active site to appear relevant to regulation. We also attempted automated computer modeling of the B-loop to investigate how it might fold over the active site and interact with the I3 peptide sidechain. Unfortunately, this was unsuccessful, and we are still unable to credibly position these additional B-loop residues in the model.</p><p>Finally, of all the homologous structures identified, only PP2Ctg from Taxoplasma gondii (2I44.pdb) (31) had a B-loop insertion similar to PPM1D. This is shown in Figure 8B, which aligns the flap subdomain of PPM1D with that of the structurally aligned PP2Ctg and PPM1A. In addition to being of similar lengths, they have the highlighted similarity of the WXR(246–248), RRS(258–260) and DQ(264–265) subsequences of PPM1D. However, because the overall degree of conservation is relatively low and it would reposition the highly conserved RV residues away from the location in the PPM1A structure, we disfavored the use of PP2Ctg as a template for the B-loop. Nonetheless, the structural overlay of PP2Ctg and PPM1A shown in Supporting Information Figure S5 provides a good, general hypothesis for how the B-loop emerges from the conserved flap subdomain.</p><!><p>The model of PPM1D in complex with the cyclic peptides, combined with the increased inhibitory activity of cyclic peptides containing acidic groups at positions X5 and X6, suggests the importance of the basic residues in the B-loop for PPM1D activity. Among these amino acids, Arg243 is proposed by the model to interact with the acidic amino acid at position X6; Lys247, located proximal to Arg243, may also interact with acidic amino acids in the cylic peptide. To investigate if such interactions were occurring, we individually mutated these residues to alanine and looked at the effect of the mutations on PPM1D activity. We first used CD spectroscopy to ensure that the mutations did not alter the PPM1D structure. The CD spectrum of wild type PPM1D was characterized by double minima of negative ellipticity at approximately 200 and 225 nm and positive ellipticity at shorter wavelengths in the presence of Mg2+. The two mutants, [R243A]PPM1D and [K247A]PPM1D, had very similar CD spectra (Figure 9A), indicative of similar structures for the wild type and mutant proteins. Next, we estimated the phosphatase activities of the two mutants using the human ATM substrate peptide. Suprisingly, both mutants showed no phosphatase activity. These results demonstrate that Arg243 and Lys247 are crucial for PPM1D catalytic activity, likely through recognition of the phosphorylated substrate.</p><!><p>In this study, the activity of a cyclic peptide inhibitor of PPM1D was improved 50-fold relative to previously-reported inhibitors of this type (24). Alteration of specific sidechains in the cyclic peptide resulted in marked improvements in activity and also revealed new structural information about the catalytic site and substrate specificity of PPM1D. At position X1, an aromatic residue (especially Phe) generally improves the inhibitory activity. At position X3, only the L-enantiomer of the Ile alpha carbon leads to good inhibitory activity, while either enantiomer at the beta position is tolerated. At positions X5 and X6, a negatively-charged residue increases the inhibitory activity dramatically. Combined, the optimal cyclic peptide inhibited PPM1D with Ki = 110 nM.</p><p>Previously, we found that a Leu at position X3 converts the cyclic peptide into a substrate for PPM1D, whereas Val- and Cha-substituted analogs had similar inhibitory activities as the Ile-substituted analog (24). Branching at the beta position is required for inhibitory activity, possibly by providing steric hindrance that prevents the pS sidechain from appropriately positioning in the binding pocket for hydrolysis. Furthermore, hydrophobic interactions in this region appear to be required, as a Gly-substituted linear peptide is a much weaker inhibitor as compared with the Ile-substituted linear peptide (24). The observation that the peptides had similar activity indicates that the methyl group at the beta position does not bind in one specific pocket of PPM1D.</p><p>As described above, substitution of acidic residues at positions X5 and X6 greatly improved inhibitory activity, showing a 10-fold increase relative to peptide 5 (Figure 4C, D). This observation is consistent with previous research on PPM1D substrate specificity, which reported that mutation of two Glu residues to alanines in the substrate sequence of ATM led to a 3-fold decrease in Km relative to the wild type sequence (23). A separate study investigating p53 substrate peptides reported that the introduction of acidic residues around p53 Ser15 resulted in high substrate affinity to PPM1D (26). Combined, these results demonstrate the importance for negative charge in the substrate or inhibitor for binding to PPM1D, likely through formation of electrostatic interactions with the enzyme. PPM1D is the only known PP2C family members to have a highly-basic loop inserted into the flap sub-domain, which may confer a preference for acidic substrates. Our molecular model suggests that Arg243 could interact with the Glu at position X6 on the cyclic peptide if it has the same position as in the PPM1A template. The interaction of the Glu at X6 is further suggested by the similar inhibitory activities of other peptides containing an L-acidic residue at this position and the poor inhibitory activities of those with basic residues. Thus, the uniqueness of the B-loop makes it an interesting target for development of a highly specific inhibitor.</p><p>As PPM1D is a member of a family of phosphatases, it is important to understand the selectivity of our inhibitors. We previously demonstrated that peptide 1 had no activity towards PP2A, but was a substrate for PPM1A (24). Likewise, although peptide 40 showed significantly improved PPM1D inhibition, it was dephosphorylated by PPM1A, albeit 10–fold less than peptide 1. Therefore, to obtain an inhibitor selective for PPM1D, five different peptides were synthesized, four containing a pS mimetic and the fifth a retro-inverse peptide. This process proved difficult, as four of the five showed weak to no PPM1D inhibitory activity. Only peptide 48, in which pS was substituted with pHse, was found to be active against PPM1D, although its inhibitory activity was approximately 30-fold decreased relative to peptide 40 (Figure 6C). This peptide was not dephosphorylated by PPM1A, nor was it an inhibitor of it. These results indicate that the inhibitory activity of PPM1D is strongly dependent on the carbon chain length of the pS sidechain. The dramatic changes in PPM1A dephosphorylation that resulted from relatively minor chemical variations likely reflect changes to the conformation of the complex that determine whether the phosphate bond is positioned correctly for hydrolysis by the enzyme's metal center. Thus, we show it is possible to develop a selective inhibitor of PPM1D, but detailed structural information will likely be required to further improve the inhibitory activity.</p><p>For quite a while, the structure of PPM1A from Das et al. (25) was the only one available for this class of enzymes. However, recently published related structures have revealed that the so-called "flap sub-domain" adjacent to the active site is likely not a static unit, but rather changes conformation to accommodate the binding of different substrates (27–30, 32, 33) (Supporting Information Figure S1). This makes accurate modeling of substrate and inhibitor binding much more difficult. This is especially true for PPM1D, since it has the added complication of the relatively long B-loop. Despite this difficulty, we can surmise a number of structural features important for activity from the available data. The first is the conservation of the Arg243-Val244 sequence of PPM1D in all eukaryotic PP2C sequences and some prokaryotic ones (Figure 8A). The second is the correspondence of His107 in PPM1D with His62 in PPM1A (Supporting Information Figure S2), which has been implicated as the acid which protonates the leaving group oxygen and is thus crucial for the phosphatase activity. A corresponding histidine residue is conserved in most of the other eukaryote homologues (except PP2CH/J) and in some prokaryotic ones. In the model of the PPM1D-inhibitor complex, a negatively-charged residue at the cyclic peptide X5 position forms a stabilizing salt bridge and sterically blocks His107 of PPM1D, thus enhancing inhibition.</p><p>As the B-loop is unique to PPM1D in the PP2C family, it is important to know how it confers the unique substrate specificity of this enzyme and, thus, its physiological functions. The elimination of activity observed here for the R243A and K247A point mutations is consistent with the results of Chuman et al. (26), who substituted B-loop residues 245–268 with the corresponding residues from the PPM1A sequence (NGS). In this case, both the Km and kcat of the mutant were approximately 3-fold reduced relative to wild type PPM1D. The difference between complete abrogation and 3-fold reduction may reflect the different substrates used (ATM vs. p53 phosphopeptides) and their differing patterns of negatively charged residues. This raises the question of the purpose of these and the other positively charged residues in the B-loop. Presumably they function to attract and correctly position the highly negatively charged mono- and di-phosphorylated substrates for catalysis. Fundamental to this understanding is to know exactly where the B-loop starts in the flap subdomain sequence. As seen in Figure 8, PPM1D and the other proteins display a range of insertions before the RV motif compared to PPM1A. Thus, it is uncertain whether the highly conserved RV residues are located at the same position in the structure as in PPM1A. If the locations are conserved, as in the model presented here and as defined by Chuman et al. (26), then the arginines corresponding to Arg243 likely play similarly important roles in all the other PP2C proteins that share this motif. If not, then Arg243 plays a unique role in PPM1D along with Lys247 and potentially some or all of the other positively charged residues in the B-loop. Further research is needed to discover how these residues coordinate in a possibly mobile flap-subdomain mechanism.</p><p>Here, we report a cyclic thioether peptide (F-pS-I-pY-DDC-amide) characterized by good inhibitory activity (Ki=110 nM). To date, only a few compounds have been reported as PPM1D inhibitors (26, 43, 46, 47), and the cyclic peptides presented here show the highest inhibitory activity, to the best of our knowledge. We further identified a second cyclic peptide (F-pHse-I-pY-DDC-amide) that showed selectivity for inhibition of PPM1D over PPM1A. Although the two phosphoric acid groups on the cyclic peptide severely limit the cellular bioavailability of the compounds, their substitution with non-phosphorus-based mimetics, such as a tetrazole scaffold, a sulfhydantoin scaffold, or a pyridine-based scaffold (48), could be an effective way to overcome this issue. In addition, although the Lipinski rules suggest that a cyclic peptide would have poor absorption or permeation into a cell because of its large molecular weight, use of a drug delivery system, such as nanoparticles, including liposomes (49–52), and cell-penetrating peptide conjugation (53–55), could overcome this issue. In the absence of detailed structural information, the findings in this study will be valuable in designing the next generation of small molecule inhibitors of PPM1D and in understanding the substrate specificity of PPM1D as a serine/threonine phosphatase.</p><!><p>This research was supported by the Intramural Research Program of the National Cancer Institute and National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health.</p><p>Supporting Information Available: Supporting Figures 1–6 show the full alignment of PPM1A and PPM1D along with additional views of the PPM1D model and Supporting Table 1 contains PDB accession numbers and references for proteins used in modeling. This material is available free of charge via the Internet at http://pubs.acs.org.</p><p>homoglutamic acid</p><p>tert-butoxycarbonyl</p><p>circular dichroism</p><p>beta-cyclohexylalanine</p><p>9-fluorenylmethoxycarbonyl</p><p>L-2-amino-4-phosphono-4,4-difluorobutyric acid</p><p>L-homophenylalanine</p><p>L-homoserine</p><p>phosphoserine</p><p>phosphothreonine</p><p>reversed-phase high-performance liquid chromatography</p><p>1,2,3,4-tetrahydroisoquinoline-3-carboxylic acid</p><p>trityl</p><p>Schematic representation of the cyclic thioether peptide. The thioether bond as formed between the acylated N-terminal residue and the cysteine sidechain at the C terminus.</p><p>An aromatic residue at position X1 of the cyclic peptide is important for inhibitory activity. (A) Inhibitory activity of cyclic peptides (10 μM) substituted at position X1 in the presence of [S1981pS] human ATM(1976–1986) peptide (30 μM) as a substrate (*p<0.05 as compared with peptide 1). (B) Concentration dependence of inhibition by cyclic peptides: peptide 1, cyclo(M-pS-I-pY-VAC)-amide (closed circle); peptide 5, cyclo(F-pS-I-pY-VAC)-amide (open circle); peptide 6, cyclo(F-pS-pY-I-VAC)-OH (open triangle). (C) Inhibitory activity of cyclic peptides used at 10 μM and 1 μM, as indicated (*p<0.05 as compared with peptide 5).</p><p>The enantiomer of the Ile at position X3 is important for inhibition of PPM1D phosphatase activity. (A) Chemical structure of isoleucine analogs. (B) Concentration dependence of the inhibition of PPM1D by cyclic peptide [F-pS-(L-Ile)-pY-V-A] (peptide 5, open circle), [F-pS-(D-Ile)-pY-VAC] (peptide 22, cross), and [F-pS-(L-allo-Ile)-pY-VAC] (peptide 23, closed circle) in the presence of [S1981pS] human ATM(1976–1986) peptide (30 μM) as a substrate.</p><p>Acidic residues at positions X5 and X6 increase the inhibitory activity of the cyclic peptides. Inhibitory activity of cyclic peptide (F-pS-I-pY-X5-X6-C) in the presence of [S1981pS] human ATM(1976–1986) peptide (30 μM) as a substrate. Inhibitory activities of (A) cyclo(F-pS-I-pY-X5-AC)-amide and (B) cyclo(F-pS-I-pY-V-X6-C)-amide peptides were measured at 10 μM (closed bar) and 1 μM (open bar) (*p<0.05 as compared with peptide 5). (C) Ki values of cyclic thioether peptides substituted by acidic residues at both position X5 and X6. (D) Concentration dependence of cyclic thioether peptide (M-pS-I-pY-VAC)-amide (peptide 1, closed circle), (F-pS-I-pY-VAC)-amide (peptide 5, open circle), (F-pS-I-pY-DDC)-amide (peptide 40, closed triangle). (E) Inhibitory activity of cyclic peptide 37 in the presence of different Wip1 substrates (open circle, [T180pT, Y182pY]Human p38(175–185); closed circle, [S1981pS]Human ATM(1976–1986)).</p><p>The cyclic peptide is a competitive inhibitor of PPM1D. Double-reciprocal plot of PPM1D phosphatase activity against [S1981pS]human ATM(1976–1986) peptide concentration during the inhibition of PPM1D with different concentrations of cyclic thioether peptide (F-pS-I-pY-EEC)-amide (peptide 37) (cross, absent; open circle, 0.2 μM; closed circle, 0.5 μM).</p><p>Peptide 48 is a selective inhibitor of PPM1D. (A) Dephosphorylation of peptide 40 by PPM1A. (B) Inhibitory activity of cyclic peptides (25 μM) substituted at position X2 in the presence of [S1981pS] human ATM(1976–1986) peptide (30 μM) as a substrate (*p<0.05 as compared with a no-inhibitor control). †Retro-inverse peptide. (C) Concentration dependence of cyclic thioether peptides 48 (open circle), 49 (open triangle), and 50 (closed circle). (D) Phosphatase activity of PPM1A against cyclic peptides 1, 40 and 48 at 100 μM. Inset: PPM1A phosphatase activity against peptide 40 (50 μM) in the presence (open bar) or in the absence of peptide 48 (closed bar).</p><p>Model of cyclic peptide inhibitor 37 bound to the active site of PPM1D. Residues discussed in the text are labeled. Highlighted residues of the protein are colored cyan, blue, red and white, for carbon, nitrogen, oxygen and hydrogen. The metal ions are magenta. The unique colors of the cyclic peptide are green, yellow and orange, for carbon, sulfur and phosphorus.</p><p>Putative alignment of the B-loop region of PPM1D. (A) Comparison with PPM1A and structurally aligned PP2C homologues with similar flap sub-domain conformations, and with human PP2C proteins with unknown 3D structure. References for the structures are given in Supporting Table 1. The Swiss-Prot database (http://expasy.org/sprot/) accession numbers for PPM1D and the other human sequences are: PPM1D_HUMAN (O15297), PPM1A_HUMAN (P35813), PPM1B_HUMAN (O75688), PPM1E_HUMAN (Q8WY54), PPM1F_HUMAN (P49593), PPM1G_HUMAN (O15355), PPM1K_HUMAN (Q8N3J5), PPM1L_HUMAN (Q5SGD2). (B) Comparison with the structurally aligned sequences of PPM1A and PP2Ctg.</p><p>Characterization of the structure and activity of PPM1D mutants. (A) CD spectra of wild type and mutant PPM1D; wild type (black), [R243A]PPM1D (red), [K247A]PPM1D (blue). Protein concentration was 1.8–2.1 μM in 50 mM Tris-HCl, 30 mM MgCl2, and 5% glycerol (pH 7.5). (B) Km values and Vmax values for wild type and mutant PPM1D. Phosphatase activities were measured using [S1981pS]human ATM(1976–1986) substrate peptide (closed circle, wild type; open circle, [R243A]PPM1D; open triangle, [K247A]PPM1D).</p>

PubMed Author Manuscript

C-2-Aryl O-substituted HI-236 derivatives as non-nucleoside HIV-1 reverse-transcriptase inhibitors

Several novel thiourea derivatives of the NNRTI HI-236 substituted at the C-2 oxygen of the phenyl ring have been synthesized and evaluated for their inhibitory activity against HIV-1 (IIIB) replication in MT-2 cell cultures. The compounds were synthesized in order to fine-tune the activity of HI-236 as well as to gain insight into spatial characteristics in the pocket pertaining to the positional choice of tether in the design of [NRTI]-tether-[HI-236] bifunctional inhibitors. Two of the thiourea derivatives bearing a butynyl (6c) or hydroxyethyl tether (6n) were endowed with improved anti-HIV activity compared to HI-236. NNRTI activity was confirmed by a cell-free RT assay on six of the derivatives in which 6c returned an IC50 of 3.8 nM compared to 28 nM for HI-236, establishing it as an improved lead for HI-236. The structure-activity profile is discussed in terms of potential interactions in the NNRTI pocket as suggested by a docking model using AutoDock, which have a bearing on the bifunctional drug design.

c-2-aryl_o-substituted_hi-236_derivatives_as_non-nucleoside_hiv-1_reverse-transcriptase_inhibitors

1. Introduction<!>2. Chemistry<!>3. Biological results, modeling, and discussion<!>4.1. Docking aspects<!>4.2. General procedures for synthesis<!>4.3. Procedure for tosylates 5a and 5b<!>4.3.1. N-(tert-Butoxycarbonyl)-2-[5-methoxy-2-(2-p-toluenesulfonyloxyethoxy)phenyl]ethylamine (5a)<!>4.3.2. N-(tert-Butoxycarbonyl)-2-{5-methoxy-2-[2-(2-p-toluenesulfonyloxyethoxy)ethoxy]phenyl}ethylamine (5b)<!>4.4. Procedure for the N-Boc O-alkylated derivatives 4<!>4.4.1. N-(tert-Butoxycarbonyl)-2-(5-methoxy-2-propoxyphenyl)ethylamine (4a)<!>4.4.2. N-(tert-Butoxycarbonyl)-2-(5-methoxy-2-propargyloxyphenyl)ethylamine (4b)<!>4.4.3. N-(tert-Butoxycarbonyl)-2-[2-(3-butynyl-1-oxy)-5-methoxyphenyl]ethylamine (4c)<!>4.4.4. N-(tert-Butoxycarbonyl)-2-[5-methoxy-2-(4-pentynyl-1-oxy)phenyl]ethylamine (4d)<!>4.4.5. N-(tert-Butoxycarbonyl)-2-[2-(2-butynyl-1-oxy)-5-methoxyphenyl]ethylamine (4e)<!>4.4.6. N-(tert-Butoxycarbonyl)-2-(2-allyloxy-5-methoxyphenyl)ethylamine (4f)<!>4.4.7. N-(tert-Butoxycarbonyl)-2-(2-benzyloxy-5-methoxyphenyl)ethylamine (4g)<!>4.4.8. N-(tert-Butoxycarbonyl)-2-[2-(2-benzyloxyethyl-1-oxy)-5-methoxyphenyl]ethylamine (4h)<!>4.4.9. N-(tert-Butoxycarbonyl)-2-[5-methoxy-2-(2-propargyloxyethoxy)phenyl]ethylamine (4i)<!>4.4.10. N-(tert-Butoxycarbonyl)-2-{5-methoxy-2-[2-(2-propargyloxyethoxy)ethoxy] phenyl}ethylamine (4j)<!>4.4.11. N-(tert-Butoxycarbonyl)-2-(2-methoxycarbonylmethyloxy-5-methoxyphenyl)ethylamine (4k)<!>4.4.12. N-(tert-Butoxycarbonyl)-2-(2-cyanomethoxy-5-methoxyphenyl)ethylamine (4l)<!>4.4.13. N-(tert-Butoxycarbonyl)-2-[2-(3-cyanopropyl-1-oxy)-5-methoxyphenyl]ethylamine (4m)<!>4.4.14. N-(tert-Butoxycarbonyl)-2-[2-(2-hydroxyethoxy)-5-methoxyphenyl]ethylamine (4n)<!>4.4.15. N-(tert-Butoxycarbonyl)-2-[2-(3-hydroxypropyl-1-oxy)-5-methoxyphenyl]ethylamine (4o)<!>4.5. Procedure for the synthesis of thioureas 6a\xe2\x80\x93o<!>4.5.1. N-(5-Bromo-2-pyridinyl)-N\xe2\x80\xb2-[2-(5-methoxy-2-propyl-1-oxyphenyl)ethyl]-thiourea (6a)<!>4.5.2. N-(5-Bromo-2-pyridinyl)-N\xe2\x80\xb2-[2-(5-methoxy-2-propargyloxyphenyl)ethyl]-thiourea (6b)<!>4.5.3. N-(5-Bromo-2-pyridinyl)-N\xe2\x80\xb2-{2-[2-(3-butynyl-1-oxy)-5-methoxyphenyl]ethyl}-thiourea (6c)<!>4.5.4. N-(5-Bromo-2-pyridinyl)-N\xe2\x80\xb2-{2-[5-methoxy-2-(4-pentynyl-1-oxy)phenyl]ethyl}-thiourea (6d)<!>4.5.5. N-(5-Bromo-2-pyridinyl)-N\xe2\x80\xb2-{2-[2-(2-butynyl-1-oxy)-5-methoxyphenyl]ethyl}-thiourea (6e)<!>4.5.6. N-(5-Bromo-2-pyridinyl)-N\xe2\x80\xb2-[2-(2-allyloxy-5-methoxyphenyl)ethyl]-thiourea (6f)<!>4.5.7. N-(5-Bromo-2-pyridinyl)-N\xe2\x80\xb2-[2-(2-Benzyloxy-5-methoxyphenyl)ethyl]-thiourea (6g)<!>4.5.8. N-(5-Bromo-2-pyridinyl)]-N\xe2\x80\xb2-{2-[2-(2-Benzyloxyethyl-1-oxy)-5-methoxyphenyl]ethyl}-thiourea (6h)<!>4.5.9. N-(5-Bromo-2-pyridinyl)]-N\xe2\x80\xb2-{2-[5-methoxy-2-(2-propargyloxyethoxy)phenyl]ethyl}-thiourea (6i)<!>4.5.10. N-(5-Bromo-2-pyridinyl)-N\xe2\x80\xb2-{2-[5-methoxy-2-(2-(2-propargyloxyethoxy)ethoxy)phenyl]ethyl}-thiourea (6j)<!>4.5.11. N-(5-Bromo-2-pyridinyl)-N\xe2\x80\xb2-[2-(2-methoxycarbonylmethyloxy-5-methoxyphenyl)ethyl]-thiourea (6k)<!>4.5.12. N-(5-Bromo-2-pyridinyl)-N\xe2\x80\xb2-[2-(2-cyanomethyloxy-5-methoxyphenyl)ethyl]-thiourea (6l)<!>4.5.13. N-(5-Bromo-2-pyridinyl)]-N\xe2\x80\xb2-{2-[2-(3-cyanopropyl-1-oxy)-5-methoxyphenyl]ethyl}-thiourea (6m)<!>4.5.14. N-(5-Bromo-2-pyridinyl)]-N\xe2\x80\xb2-{2-[2-(2-hydroxyethoxy)-5-methoxyphenyl]ethyl}-thiourea (6n)<!>4.5.15. N-(5-Bromo-2-pyridinyl)]-N\xe2\x80\xb2-{2-[2-(3-hydroxypropyl-1-oxy)-5-methoxyphenyl]ethyl}-thiourea (6o)<!>4.6. Anti-HIV evaluation<!>4.7. Steady-state IC50 determination<!>

<p>The HIV-1 reverse-transcriptase enzyme is responsible for converting the genomic single-stranded RNA of HIV into a double-stranded DNA and continues to be a major target for anti-HIV drug discovery.1 Inhibitors fall into two distinct classes as nucleoside analogues (NRTIs) and non-nucleoside analogues (NNRTIs), and their modes of action are distinct and well documented.2 Thus, while NRTIs act as competitive substrates at the substrate-binding site, the NNRTIs work allosterically in an adjacent pocket, interfering with reverse transcription by altering the conformational mobility of RT, which results in non-competitive inhibition. To date, more than 30 structurally diverse NNRTIs have been identified, which have been comprehensively reviewed by several workers.3 NNRTIs are vulnerable to HIV's high mutation rate, and to circumvent this, they are currently used in combination with NRTIs. An alternative strategy that has been investigated by a number of groups over the last decade or so has been to combine an NRTI and an NNRTI into a single bifunctional molecule.4 Such inhibitors may be classified into two types: cleavable5 and non-cleavable.6 The former involve NRTI–linker–NNRTI systems that are designed to release the two drugs into the cytoplasm via enzymatic hydrolysis in the hope of promoting synergistic inhibitory action. Conversely, the latter, based on an original suggestion by Nanni7 and coworkers, are designed to promote inhibition at both sites cooperatively through the individual drugs in the same entity, on the basis that the NRTI and NNRTI drug–target sites are in close proximity of 10–15 Å.8 The latter, ambitious strategy has produced some inventive combinations, but true bifunctional character has yet to be comprehensively established. In the design of such bifunctional drugs, it is important to identify attachment points on each drug as well as to choose an appropriate tether regarding size and functional character.</p><p>We have been interested in developing non-cleavable bifunctional NRTI/NNRTI compounds involving d4T as the NRTI and the PETT (phenethylthiazolyl) derivative HI-236 as the NNRTI. These compounds were selected in view of their excellent profiles as antiviral agents, and it was decided to attach the linker on the NRTI side of the molecule to C-5 of the pyrimidine base in view of anticipated low interference with base pairing.9 On the HI-236 side, it was decided to use the C-2 phenolic oxygen in view of studies conducted by Uckun10 regarding the binding mode of HI-236 in the RT pocket. Our recently published11 first prototype, [d4T]-butyne-[HI-236], returned an EC50 of 250 nM against HIV-1 (IIIB) in MT-2 cell culture using an MTT assay. This result compares favourably with Ladurée's bifunctional,12 also based on d4T but conjugated to a PETT derivative through a glycylsuccinyl cleavable linker, which turned out to be inactive (Fig. 1).</p><p>PETT (phenethylthiazolyl) derivatives such as those shown in Figure 1 were first introduced by the Eli Lilly group in 1995.13 Changing the thiazole ring to a pyridine ring as N-(5-bromo-2-pyridinyl)-N′-(2-pyridylethyl)-thiourea (named as Trovirdine), resulted in enhanced activity and a detailed structural analysis of it carried out by the Uckun group revealed abundant sterically usable space surrounding the pyridyl group as well as the ethyl linker.14,15 They postulated that an efficient use of this space would lead to more potent anti-HIV agents with higher affinity for the NNRTI binding pocket, and this resulted in several more potent derivatives including the highly potent HI-236,10,15 with a much superior inhibition profile in which the pyridyl ring of trovirdine was replaced by a 2,5-dimethoxyphenyl group, Figure 2.</p><p>Structural studies around HI-236 revealed, as with other PETT derivatives, that the thiourea portion interacts in Wing 1 towards the front of the pocket via H-bonding with K101.10,15 Conversely, the C-5-methoxy group forges cooperative C–Hπ interactions with the highly conserved W229, while the phenyl ring interacts with Y188. The methoxy group at C-2 was found to lie underneath the ethyl linker and was considered to occupy some of the vacant space near Y188 in Wing 2, an issue which assumes a greater weight of importance for drug-resistant strains where there is an increase in pocket size due to mutated residues of smaller size such as Y181C and Y188L. The realization of the [d4T]-butynyl-[HI-236] bifunctional prototype mentioned previously11 prompted a study of the influence of the nature of the C-2 tether of HI-236 on its biological activity. In this paper, we present the synthesis and biological activity of a small library of HI-236 derivatives involving substitution of the C-2 phenolic methyl group, the objective being not only to shed light on NNRTI attachment at the C-2 oxygen for the bifunctional design, but also to probe the possibility of further binding in the region of the pocket just mentioned in an attempt to improve the activity of HI-236 even further. Tethers attached to the C-2 phenolic oxygen were chosen to probe length, functionality and polarity. Bearing in mind that a Sonogashira connection16 between C-5 of d4T and a terminal alkyne was used for the bifunctional connection chemistry, some, but not all of the derivatives studied were selected with an alkyne terminus.</p><!><p>Synthesis of the tethered derivatives utilized phenolic amine 1 as a key intermediate that was readily available in multigram quantities using chemistry reported by Glennon17 via a three-step sequence from commercially available 2-hydroxy-5-methoxybenzaldehyde, Scheme 1. In our case, it was found it easier to isolate 1 as an N-Boc carbamate 2. Yields for the various steps in the sequence to carbamate 2 were high and spectroscopic data was in accordance with literature values for all of the known compounds, Scheme 1.</p><p>Thereafter, following benzyl group deprotection to 3, the various O-alkylations at the C-2 phenolic oxygen could be carried out with introduction of the thiourea moiety last for each case. A more convergent approach involving converting amine 1 first into the thiourea derivative, debenzylating and then carrying out the C-2 phenolic alkylation reaction to generate a small library of targets was deemed to be unattractive in view of the anticipated incompatibility of the thiourea functionality with the alkylation chemistry.</p><p>Alkylation of carbamate 3 (Scheme 2 and Table 1) to afford a library of alkylated N-Boc derivatives 4 proceeded efficiently with bromide or tosylate electrophiles using K2CO3 in acetonitrile at reflux for around 20 h and proved to be superior to a Mitsunobu reaction with the corresponding alcohol. Yields of chromatographed products were generally in excess of eighty percent (see Section 4). In most cases, the electrophile was readily available either commercially or via a one-step derivatisation.</p><p>The exceptions came for the alkynyl tethers used to probe the question of the influence of tether length and aqueous solubility of the bifunctional compounds. These involved alkynyl PEG (ethylene glycol) chains and were synthesized by the following sequence shown in Scheme 3 in which the tether was developed divergently. Thus, standard phenolic alkylation of a monobenzylated glycol bromide afforded the alkylated product, which was committed to a two-step sequence involving hydrogenolytic debenzylation followed by tosylation to afford tosylates 5a and 5b. Nucleophilic displacement with propargyloxide anion then furnished the products 4i and 4j. The latter18 turned out to be superior to hydroxyl-group alkylation with propargyl bromide. Similarly, alcohols 4n and 4o were obtained from hydrogenolysis of their benzyl ethers obtained from alkylation of 3 as described.</p><p>Table 1 summarises the various products of C-2 phenolic alkylation of Boc carbamate 3.</p><p>All derivatives 4a–o returned acceptable NMR spectra together with acceptable combustion analysis data (solids) and/or HRMS mass spectral data. Notable in the NMR were the triad of signals for the three aromatic protons in the 1H NMR spectrum integrating correctly against the N-Boc tert-butyl singlet. The 13C NMR spectra returned the correct number of carbon singlets in each case.</p><p>Finally, synthesis of the target thioureas 6 was accomplished via a Boc-deprotection, condensation sequence shown in Scheme 4.</p><p>Thus, each derivative 4 was reacted with trifluoroacetic acid (TFA) in DCM at 0 °C for 1–3 h until TLC revealed complete deprotection to a polar amine spot. In view of the amine's anticipated water solubility, the final step was conducted without using an aqueous work-up. Thus, following complete removal of all volatiles including any excess TFA, the residue was redissolved in THF, Hünig's base (EtN(i-Pr)2) was added to liberate the free amine and the thiocarbonyl reagent 7 added according to the original procedure. As described by the Eli Lilly group,13 the latter could be readily prepared by reacting 2-amino-5-bromopyridine with 1,1′-thiocarbonyldiimidazole in acetonitrile at room temperature for 12 h to afford a precipitate that was used without purification. Attempted recrystallization of 7 from methanol resulted in a product with imidazole substituted by methoxy (Scheme 5).</p><p>Condensation between the amine and 7 could be realized in DMF at 100 °C as described by Bell and coworkers13 to afford the targets 6a–o in an overall yield of around 30% for the two steps. However, later on, it was discovered that condensation could be carried out under much milder conditions in THF or DMF at room temperature to increase the two-step yield to around 60%. All of the targets were isolated by silica-gel column chromatography as crystalline solids that were crystallized to a constant, sharp melting point to return acceptable combustion analysis data. A full spectroscopic analysis using 1H and 13C NMR spectroscopy was also carried out on each derivative to return acceptable data. Notably, the aromatic region in the 1H NMR spectra provided convenient markers for the aromatic and heteroaromatic rings to demonstrate that condensation had taken place. The thiourea N–H's could be discerned downfield as two separate resonances. The thiourea thiocarbonyl carbon could be identified using 13C NMR, resonating at around 179 ppm. The retention of bromine in the pyridine ring was confirmed by 13C NMR (δC–Br ∼112.6 ppm) as well as microanalytical and IR data. Table 2 summarises yields for the two-step sequence of 4–6 in Scheme 4.</p><!><p>The inhibitory activity of compounds 6a–o was measured against HIV-1 (IIIB) replication in MT-2 cell culture using an MTT assay.19 HI-236 was included as a reference compound. The results are shown in Table 3.</p><p>In order to lend support that these HI-236 derivatives act as NNRTIs, six of the derivatives (6c, 6d, 6i, 6k, 6n, and 6o) covering the most (6n) to the least active (6k) were subjected to an in vitro steady-state RT inhibition assay, the results from which are shown in Table 4. The results present strong support for all of the derivatives in this study to be acting as NNRTIs since they have no possibility of being phosphorylated for NRTI activity against RT. This is also in keeping with the published findings on HI-236. The activities ranged from 3.8 to 100 nM, and as expected were superior to those from the cell-culture results except for 6n, which was slightly lower. Such improvements likely reflect the lipophilicity of the derivatives and their poor solubility in the aqueous cell-culture medium. Notably, the ester 6k, which was effectively inactive in cell culture (12 μM), revealed a 120-fold improvement to 100 nM in the RT assay, possibly due to ester hydrolysis in the cell-culture medium.20 Importantly, the alkyne 6c returned a value of 3.8 nM in the RT assay, which was more active compared to that of the alcohol 6n (19 nM), in spite of the reverse being true in cell culture (26 nM vs 12 nM, respectively). The most active derivative 6c was then evaluated in cell culture against the Y181C resistant strain and retained activity much better than HI-236 (8-fold decrease against 25 for HI-236).</p><p>Regarding a general comparison of all derivatives, the results from the cell culture shown in Table 3 were taken to reveal important trends within classes of different tether (polar vs lipophilic). Four of the compounds returned activities equal to HI-236 (6b and 6m) or greater (6c, 2-fold increase; 6n, 4-fold increase). In order to assist with interpretation of the results, docking studies were carried out using AutoDock 3.0521 based on the recently published HIV-1 RT protein crystal structure of N-(5-chloro-2-pyridinyl)-N′-[2-(4-ethoxy-3-fluoro-2-pyridinyl)ethyl]-thiourea as template.22 The conformation of the PETT derivatives studied was set to accommodate the well documented14,22,23 intramolecular hydrogen bond between the hydrogen of the thiourea nitrogen attached to the alkyl side and the nitrogen of the pyridinyl ring to form a flat six-membered pseudo-ring. This protocol resulted in a strong preference for one low energy conformation which dominated 153 of the 250 possible conformations. HI-236 was docked first, and we were able to establish parity of result with that obtained by Uckun. The two structures are shown in Figure 3.</p><p>Thus, as established by Uckun, the inhibitor adopts the characteristic butterfly-shaped orientation with the thiourea moiety embedded in Wing 1 in which hydrogen bonding between the thiourea hydrogen on the N-pyridyl side and the carbonyl oxygen of K101 is clearly visible. Wing 2 accommodates the phenyl ring of HI-236 close to the Y181 and Y188 region with the 5-methoxy group interacting with W229. The 2-alkoxy substituent positions into a well-containing residues Y181, Y188, F227, V106, and V179. This feature had significance for the aims of this study. The only difference between our structure and that of Uckun's was the orientation of W229 relative to Y181. In Uckun's model, the six-membered ring of the indole of W229 overlapped Tyr181, whereas the model obtained from our calculations had the five-membered ring of the indole overlapping Tyr181. This could have been due to differences in the static protein structure used in our docking calculations as compared with the structure of the HIV-1 RT binding pocket used in the studies by Uckun. Attention was then focused on docking two of the derivatives, chosen as the ester 6k and the alcohol 6o. The parity of 6k and 6o with our docked HI-236 in terms of overall topology in the pocket was first investigated, as illustrated in Figure 4 for 6k. The elongation of the 2-alkoxy substituent can be seen but the positioning of the rest of the PETT structure essentially stayed the same, revealing that EC50 differences are likely to be based on interactions in the Wing 2 region as postulated. Alcohol 6o gave a similar result.</p><p>Examination of the way in which the 2-alkoxy substituents of 6k and 6o positioned in the hydrophobic region of Wing 2 as shown in Figures 5 and 6 taken from each end of the pocket for both cases, gave insights into the possible interactions responsible for fluctuation of biological activity.</p><p>Both structures suggest the possibility of interaction with V109 or, to a lesser extent V179, towards the floor of the pocket. A cooperative hydrogen bond for 6o would nicely explain the much enhanced activity of alcohol 6n (EC50 = 0.012 μM) with a shorter (two-carbon) chain, in which presumably the distance between hydrogen-bond donor and acceptor is optimal. 6n returned an activity four times higher than HI-236 in cell culture, and almost twice that in the RT assay and thus presents itself as an interesting candidate for the problematic V106A mutation if the hydrogen bonding postulate is correct.</p><p>Similarly, for the alkyl and alkynyl compounds 6a–f, the results show that the hydrophobic pocket is accommodating but with the only significant cooperative interaction being with the triple bond. In this regard, the butynyl chain of 6c, rather than the propargyl (6b) or pentynyl substituents (6d) appears to be optimal, with 6c indicating a 2-fold increase in activity compared to HI-236 in cell culture and 7-fold in the RT assay. Although this classical force field does not include an explicit π–π term, the charge distribution on the aromatic rings mimics this fundamentally quantum effect surprisingly well. Therefore we suggest that the increased activity of 6c is an optimal cooperative π–π face-to-face interaction between the aromatic residues of Y181, Y188, and the triple bond. In addition, a much lower fold reduction in activity for 6c (8×) was observed against the Y181C24 mutant strain compared to that of HI-236 (25×), (see Table 4). The PEG derivatives predictably returned lower activities, but not overly so given their bulk. Thus 6j, with 10 atoms in the substituent, returned an activity of 0.39 μM as only 8-fold less potent than HI-236. By comparison, the benzyl derivative 6h is too bulky to be satisfactorily accommodated. Finally, the nitrile derivative 6m indicated some level of cooperation in the pocket being as active as HI-236, in spite of its five-atom substituent. Once again, π–π and/or hydrogen bonding possibilities appear to be likely explanations.</p><p>In summary, it is important to note that of the 15 compounds tested in cell culture, two (6c and 6n) were more active than HI-236 and two (6b and 6m) were as active, while this picture improved in the RT assay in which of the six derivatives tested, two (6c and 6n) were more active and two were as active as HI-236 (6d and 6o). Such results endorse the conclusions drawn from the study by Uckun15 that the PETT derivatives have unused available pocket volume with good potential for drug-development, and have identified tethered butynyl derivative 6c as an advanced lead. Further fine-tuning is worthwhile pursuing on developing side chains that can cope with mutated residues contained in resistant strains,25 and the modelling results suggest that this might be possible to achieve by adding further substitution at the C-3 position of the aromatic ring ortho to the C-2 O-tether. In addition, the study has generated important insights regarding the choice of the C-2 oxygen as the attachment point for the tether in the bifunctional compounds, and the likelihood of a tether at this position providing a route from the pocket to the NRTI binding site. In this regard, a comprehensive study of elongated alkylated bifunctional double-drugs in order to shed light on the origin of biological activity for the prototype in Figure 111 will be communicated in a forthcoming paper.</p><!><p>The binding conformations of HI-236 (1) and its ester (6k) and alcohol (6o) derivatives bound to HIV-1 Reverse Transcriptase (RT) were modelled using AutoDock 3.0521 based on the published HIV-1 RT protein crystal structure of N-(5-chloro-2-pyridinyl)-N′-[2-(4-ethoxy-3-fluoro-2-pyridinyl)ethyl]-thiourea.22 Each ligand was built in Gaussview,26 and optimised using Gaussian 98 to relax bond lengths and angles that were not varied in the docking simulation. Polar hydrogen atoms were added to the protein and Kollman united-atom partial charges assigned using the AutoDockTools package. For each ligand, Gasteiger-Marsili27 partial charges were assigned, as implemented in AutoDockTools. Autodock 3.05 utilises an empirical scoring function21 to calculate binding free energies, which incorporates five energy terms, including a Lennard–Jones 12-6 term and a directional 12-10 hydrogen bond term, for which the default parameters distributed with AutoDockTools were used. Electrostatic interactions were calculated using a distance-dependent dielectric constant.28 Atomic solvation parameters and fragmental volumes were assigned using the AddSol utility, from which the desolvation contribution to the binding free energy is calculated. A 61 × 61 × 61 grid map was used in all docking calculations, with a grid spacing of 0.375 Å. Given the known location of the NNRTI binding site, the grid was centred on the coordinates for the equivalent atom of PETT-lig corresponding to the default selected root atom of each ligand investigated. Docked conformations were generated using the Lamarckian genetic algorithm (LGA) with an initial population size of 150 structures. Translation, quaternion and torsional step sizes were set to 2 Å, 5.0° and 5.0° respectively for HI-236 and 0.1 Å, 1.0° and 1.0° for ligands 9k and 9o. Further parameters were set to their default values. A total of 250 runs were performed for each ligand, and the resulting conformations clustered using a root mean-squared deviation criterion of 0.5 Å in x, y, z positional coordinates. Bond rotation from the pyridinyl ring across to the thiourea moiety was disallowed, and the conformation was set to accommodate the well documented14,21,22 intramolecular hydrogen bond between the hydrogen of one of the thiourea nitrogens and the nitrogen of the pyridinyl ring to form a flat six-membered pseudo-ring. This protocol resulted in a strong preference for one low energy conformation which dominated 153 of the 250 possible conformations. Although using a static protein structure in the docking simulation and limiting the rotation of certain bonds, the binding conformation of HI-236 in this study compared well with that obtained by Uckun.15</p><!><p>Microanalyses were obtained with a Fisons EA 110 CHN Elemental Analyser. Infrared (IR) absorptions were measured on a Perkin-Elmer Spectrum One FT-IR spectrometer. 1H NMR spectra were recorded on a Varian Mercury Spectrometer at 300 MHz and a Varian Unity Spectrometer at 400 MHz with Me4Si as internal standard. 13C NMR spectra were recorded at 75 MHz on a Varian Mercury Spectrometer or at 100 MHz on a Varian Unity Spectrometer with Me4Si as internal standard. High resolution mass spectra were recorded on a VG70 SEQ micromass spectrometer. Melting points were determined using a Reichert-Jung Thermovar hot-stage microscope and are uncorrected. Analytical thin-layer chromatography (TLC) was performed on silica-gel 60 F254 (Merck). Column chromatography was performed with Merck silica-gel 60 (70–230 mesh). HI-23610 was synthesized by the same sequence as for derivatives 6 and returned acceptable NMR data.</p><!><p>The phenol 3 was O-alkylated with the appropriate PEGbromide as its monobenzyl ether according to the procedure below for formation of the alkylated derivatives 4. The alkylated products were subjected to hydrogenolysis in ethanol using Pd–C (10 mol%) at atmospheric pressure for 18 h. Following filtration through Celite, the alcohols were purified by column chromatography in yields in excess of 80% before being tosylated with p-toluenesulfonyl chloride (1.5 equiv) in dichloromethane using triethylamine (2 equiv) and DMAP (cat). Following a conventional work-up, the product was isolated by column chromatography. Solid products were generally recrystallized from ethyl acetate/hexane mixtures.</p><!><p>76% Yield as a colourless oil; IR (CHCl3) νmax 3690, 3451, 1706, 1367, 1164 cm−1; 1H NMR (400 MHz, CDCl3) δ: 7.75 (2H, d, J = 8.2 Hz), 7.28 (2H, d, J = 8.2 Hz), 6.62 (3H, m), 4.77 (1H, br s), 4.30 (2H, t, J = 4.7 Hz), 4.05 (2H, t, J = 4.7 Hz), 3.68 (3H, s), 3.22 (2H, q, J = 6.8 Hz), 2.65 (2H, t, J = 6.8 Hz), 2.38 (3H, s), 1.38 (9H, s); 13C NMR (100 MHz, CDCl3) δ: 155.9, 154.1, 150.2, 145.0, 132.9, 129.9, 129.3, 127.8, 116.7, 112.0, 111.9, 78.8, 68.4, 66.4, 55.5, 40.5, 30.9, 28.4, 21.5; EI-HRMS: m/z; found: 465.18130 (M+). C23H31NO7S (M+) requires 465.18212.</p><!><p>83% Yield as a colourless oil; IR (CHCl3) νmax 3693, 3453, 1706, 1367, 1168 cm−1; 1H NMR (400 MHz, CDCl3) δ: 7.76 (2H, d, J = 7.9 Hz), 7.28 (2H, d, J = 7.9 Hz), 6.73 (1H, d, J = 8.8 Hz), 6.68 (2H, m), 4.80 (1H, br s), 4.17 (2H, t, J = 4.7 Hz), 3.98 (2H, t, J = 4.7 Hz), 3.75 (4H, m), 3.73 (3H, s), 3.29 (2H, q, J = 6.7 Hz), 2.74 (2H, t, J = 6.7 Hz), 2.39 (3H, s), 1.39 (9H, s); 13C NMR (100 MHz, CDCl3) δ: 155.9, 153.9, 150.9, 144.8, 133.1, 129.8, 129.3, 127.9, 116.6, 113.0, 112.0, 78.8, 70.0, 69.3, 68.8, 68.3, 55.6, 40.6, 31.0, 28.4, 21.5; EI-HRMS: m/z; found: 509.20675 (M+). C25H35NO8S (M+) requires 509.20834.</p><!><p>The alkylating agent as a tosylate or bromide (2.0 mmol) in dry acetonitrile (5 mL) was added dropwise over 1 h to a refluxing and stirring mixture of the phenol 3 (1.0 mmol) and anhydrous potassium carbonate (4.0 mmol) in dry acetonitrile (15 mL). The reaction was refluxed for 20 h. The mixture was filtered, the acetonitrile evaporated, and the residue subjected to silica-gel column chromatography (10% EtOAc/pet ether) to afford the product generally as a colourless solid. Entries 4n and 4o were derived via hydrogenolysis of their corresponding benzyl ethers, while entries 4i and 4j were obtained via substitution of tosylates 5a and 5b respectively using propargyloxide ion in refluxing THF (5 h). Yields cited in the text refer to the final step in each case.</p><!><p>95% Yield, mp 52–54 °C; IR (CHCl3) νmax 3684, 3452, 3002, 2977, 1706, 1502 cm−1; 1H NMR (400 MHz, CDCl3) δ: 6.76 (1H, d, J = 9.2 Hz), 6.70 (2H, m), 4.78 (1H, br s), 3.88 (2H, t, J = 6.4 Hz), 3.75 (3H, s), 3.35 (2H, q, J = 6.6 Hz), 2.78 (2H, t, J = 6.6 Hz), 1.80 (2H, m), 1.42 (9H, s), 1.04 (3H, t, J = 7.4 Hz); 13C NMR (100 MHz, CDCl3) δ: 155.9, 153.5, 151.3, 129.5, 116.8, 112.3, 112.0, 79.8, 70.2, 55.7, 40.9, 30.9, 28.4, 22.8, 10.7; EI-HRMS: m/z; found: 309.19401 (M+). C17H27NO4 (M+) requires 309.19400. Anal. Found: C, 66.13; H, 8.80; N, 3.91. C17H27NO4 requires; C, 65.99; H, 8.80; N, 4.53.</p><!><p>75% Yield, mp 49–50 °C; IR (CHCl3): νmax 3691, 3454, 3308, 3022, 2124, 1707, 1501 cm−1; 1H NMR (300 MHz, CDCl3): δ 6.90 (1H, d, J = 9.3 Hz), 6.72 (2H, m), 4.65 (3H, d, J = 2.4 Hz), 3.75 (3H, s), 3.35 (2H, q, J = 6.8 Hz), 2.79 (2H, t, J = 6.8 Hz), 2.47 (1H, t, J = 2.4 Hz), 1.42 (9H, s); 13C NMR (75 MHz, CDCl3): δ 155.9, 154.3, 149.8, 129.6, 116.7, 113.6, 112.0, 79.0, 79.0, 56.7, 55.6, 40.6, 30.9, 28.4; EI-HRMS: m/z; found: 305.16244 (M+). C17H23NO4 (M+) requires 305.16271. Anal. Found: C, 66.90; H, 7.54; N, 4.54. C17H23NO4 requires C, 66.86; H, 7.59; N, 4.59.</p><!><p>61% Yield, mp 76–77 °C; IR (CHCl3): νmax 3455, 3309, 2413, 1707, 1602 cm−1; 1H NMR (300 MHz, CDCl3): δ 6.75 (3H, m), 4.70 (1H, br s), 4.05 (2H, t, J = 6.8 Hz), 3.75 (3H, s), 3.36 (2H, q, J = 6.6 Hz), 2.79 (2H, t, J = 6.6 Hz), 2.66 (2H, dt, J = 2.7, 6.8 Hz), 2.03 (1H, t, J = 2.7 Hz), 1.42 (9H, s); 13C NMR (75 MHz, CDCl3): δ 155.9, 153.9, 150.6, 129.3, 116.8, 112.8, 112.0, 80.7, 78.9, 69.8, 66.8, 55.6, 40.7, 30.9, 28.4, 19.7; EI-HRMS: m/z; found: 319.17756 (M+). C18H25NO4 (M+) requires 319.17836. Anal. Found: C, 67.10; H, 7.66; N, 3.67. C18H25NO4 requires C, 67.89; H, 7.89; N, 4.39.</p><!><p>93% Yield; mp 69–71 °C; IR (CHCl3) νmax 3696, 3452, 3307, 2980, 2344, 1707, 1502, 1218, 1164 cm−1; 1H NMR (300 MHz, CDCl3) δ: 6.79 (1H, d, J = 9.6 Hz), 6.70 (2H, m), 4.72 (1H, br s), 4.02 (2H, t, J = 6.5 Hz), 3.75 (3H, s), 3.34 (2H, q, J = 6.5 Hz), 2.78 (2H, t, J = 6.5 Hz), 2.41 (2H, td, J = 2.7, 6.5 Hz), 1.99 (2H, m), 1.96 (1H, t, J = 2.7 Hz), 1.43 (9H, s); 13C NMR (75 MHz, CDCl3) δ: 155.9, 153.7, 151.0, 129.0, 116.7, 112.5, 112.0, 83.4, 79.3, 68.9, 66.9, 55.7, 40.8, 30.9, 28.4, 28.4, 15.4. Anal. Found: C, 68.44; H, 8.16; N, 4.20. C19H27NO4 requires; C, 68.53; H, 8.28; N, 3.87.</p><!><p>92% Yield; mp 53–55 °C; IR (CHCl3) νmax 3681, 3449, 3014, 2246, 1703, 1501, 1205 cm−1; 1H NMR (300 MHz, CDCl3) δ: 6.90 (1H, d, J = 9.6 Hz), 6.71 (2H, m), 4.65 (1H, br s), 4.60 (2H, q, J = 2.3 Hz), 3.76 (3H, s), 3.35 (2H, q, J = 6.0 Hz), 2.79 (2H, t, J = 6.0 Hz), 1.83 (3H, t, J = 2.3 Hz), 1.43 (9H, s); 13C NMR (75 MHz, CDCl3) δ: 155.9, 154.0, 150.1, 129.4, 116.5, 113.6, 111.9, 83.3, 78.6, 74.5, 57.3, 55.6, 40.6, 30.6, 28.4, 3.6. Anal. Found: C, 67.39; H, 7.97; N, 4.12. C18H25NO4 requires C, 67.69; H, 7.89; N, 4.23.</p><!><p>99% Yield; mp 54–56 °C; IR (CHCl3) νmax 3682, 3449, 3201, 1502, 1703, 1210 cm−1; 1H NMR (300 MHz, CDCl3) δ: 6.78 (1H, d, J = 9.0 Hz), 6.71 (2H, m), 6.05 (1H, ddt, J = 5.1 Hz, 10.5 Hz, 17.3 Hz), 5.39 (1H, dq, J = 1.6 Hz, 17.3 Hz), 5.26 (1H, dq, J = 1.6 Hz, 10.5 Hz), 4.71 (1H, br s), 4.50 (2H, dt, J = 1.6 Hz, 5.1 Hz), 3.76 (3H, s), 3.40 (2H, q, J = 6.5 Hz), 2.80 (2H, t, J = 6.5 Hz), 1.43 (9H, s); 13C NMR (75 MHz, CDCl3) δ: 156.0, 153.7, 150.8, 133.6, 129.2, 117.1, 116.7, 112.9, 112.0, 79.0, 69.5, 55.7, 40.8, 30.9, 28.4. Anal. Found: C, 66.41; H, 8.21; N, 4.40. C17H25NO4 requires C, 66.43; H, 8.20; N, 4.56.</p><!><p>89% Yield; mp 103-104°C; IR (CHCl3): νmax 3691, 3453, 1707, 1503 cm−1; 1H NMR (CDCl3, 400 MHz): 7.41 (5H, m), 6.84 (1H, d, J = 8.1 Hz), 6.71 (2H, m), 5.03 (2H, s), 4.70 (1H, br s), 3.76 (3H, s), 3.37 (2H, q, J = 6.4 Hz), 2.83 (2H, t, J = 6.4 Hz), 1.42 (9H, s); 13C NMR (CDCl3, 100 MHz) δ: 155.9, 153.9, 151.0, 137.4, 128.6, 128.0, 127.9, 127.3, 116.8, 113.0, 112.0, 79.0, 70.9, 55.7 (OCH3), 40.8 (C-1), 30.9 (C-2), 27.8 (OC(CH3)3); EI-HRMS: m/z; found: 301.13383 [(M+–tert-butyl) + H]. C21H27NO4 requires 301.13409 [(M+ – tert-butyl) + H]. Anal. Found: C, 70.50; H, 7.62; N, 3.86. C21H27NO4 requires C, 70.56; H, 7.61; N 3.92.</p><!><p>85% Yield; mp 56–58 °C; IR (CHCl3) νmax 3673, 3449, 3000, 2398, 1701, 1501 cm−1; 1H NMR (300 MHz, CDCl3) δ: 7.40 (5H, m), 6.79 (1H, d, J = 8.8 Hz), 6.70 (2H, m), 4.78 (1H, br s), 4.63 (2H, s), 4.10 (2H, t, J = 4.8 Hz), 3.82 (2H, t, J = 4.8 Hz), 3.75 (3H, s), 3.37 (2H, q, J = 6.5 Hz), 2.80 (2H, t, J = 6.5 Hz), 1.42 (9H, s); 13C NMR (75 MHz, CDCl3) δ: 155.9, 153.7, 151.0, 138.1, 129.2, 128.4, 127.7, 127.6, 116.6, 112.9, 111.9, 78.7, 73.2, 68.7, 68.4, 55.5, 40.7, 30.9, 28.4; EI-HRMS: m/z; found: 401.22044 (M+). C23H31NO5 (M+) requires 401.22022. Anal. Found: C, 68.90; H, 7.80; N, 3.37. C23H31NO5 requires C, 68.80; H, 7.78; N, 3.49.</p><!><p>74% Yield from tosylate 5a with propargyloxide; colourless oil; IR (CHCl3): νmax 3674, 3452, 3307, 2121, 1703 cm−1; 1H NMR (300 MHz, CDCl3): δ 6.78 (1H, d, J = 7.8 Hz), 6.69 (2H, m), 4.79 (1H, br s), 4.26 (2H, d, J = 2.3 Hz), 4.10 (2H, m), 3.88 (2H, m), 3.75 (3H, s), 3.35 (2H, q, J = 6.7 Hz), 2.80 (2H, t, J = 6.7 Hz), 2.45 (1H, t, J = 2.3 Hz), 1.42 (9H, s); 13C NMR (75 MHz, CDCl3) δ: 156.0, 153.9, 150.9, 129.4, 116.7, 113.0, 112.0, 79.6, 78.9, 74.7, 68.4, 68.3, 58.5, 55.7, 40.8, 31.0, 28.4; EI-HRMS: m/z; found: 349.18937 (M+). C19H27NO5 (M+) requires 349.18892.</p><!><p>91% Yield from tosylate 5b with propargyloxide; colourlesss oil; IR (CHCl3): νmax 3693, 3607, 3453, 3308, 3012, 2980, 2934, 2120, 1706, 1502; 1H NMR (400 MHz, CDCl3): δ 6.74 (1H, d, J = 8.6 Hz), 6.66 (2H, m), 4.90 (1H, br s), 4.17 (2H, d, J = 2.4 Hz), 4.05 (2H, t, J = 4.9 Hz), 3.80 (2H, t, J = 4.9 Hz), 3.71 (3H, s), 3.70 (4H, m), 3.31 (2H, q, J = 6.7 Hz), 2.76 (2H, t, J = 6.7 Hz), 2.42 (1H, t, J = 2.4 Hz), 1.39 (9H, s); 13C NMR (100 MHz, CDCl3): δ 156.0, 153.8, 151.0, 129.4, 116.6, 113.1, 112.0, 79.6, 78.7, 74.6, 70.5, 69.8, 69.1, 68.5, 58.3, 55.6, 40.7, 31.0, 28.4; EI-HRMS: m/z; found: 393.21448 (M+). C21H31NO6 (M+) requires 393.21514.</p><!><p>98% Yield; mp 66–68 °C; IR (CHCl3) νmax 3681, 3449, 1758, 1704, 1501, 1227 cm−1; 1H NMR (300 MHz, CDCl3): δ 6.67 (3H, m), 4.84 (1H, br s), 4.61 (2H, s), 3.78 (3H, s), 3.74 (3H, s), 3.40 (2H, q, J = 6.5 Hz), 2.90 (2H, t, J = 6.5 Hz), 1.41 (9H, s); 13C NMR (75 MHz, CDCl3) δ: 169.6, 156.0, 154.3, 150.1, 129.5, 116.9, 112.4, 112.0, 78.8, 66.0, 55.6, 52.1, 40.7, 30.9, 28.4. Anal. Found: C, 60.37; H, 7.36; N, 3.91. C17H25NO6 requires C, 60.16; H, 7.42; N; 4.13.</p><!><p>96% Yield; mp 98–102 °C; IR (CHCl3) νmax 3681, 3449, 2240, 1703, 1501, 1226 cm−1; 1H NMR (300 MHz, CDCl3) δ: 6.89 (1H, d, J = 9.6 Hz), 6.74 (2H, m), 4.73 (2H, s), 4.61 (1H, br s), 3.76 (3H, s), 3.35 (2H, q, J = 6.8 Hz), 2.80 (2H, t, J = 6.8 Hz), 1.43 (9H, s); 13C NMR (75 MHz, CDCl3) δ: 155.9, 155.4, 148.9, 130.1, 117.1, 115.4, 114.0, 112.2, 79.2, 55.6, 54.9, 40.5, 31.4, 28.4. Anal. Found: C, 62.87; H, 7.09; N, 8.30. C16H22N2O4 requires C, 62.73; H, 7.24; N, 9.14.</p><!><p>96% Yield; mp 69–71 °C; IR (CHCl3) νmax 3681, 3449, 3014, 2246, 1703 cm−1; 1H NMR (300 MHz, CDCl3) δ: 6.74 (3H, m), 4.73 (1H, br s), 4.03 (2H, t, J = 5.8 Hz), 3.75 (3H, s), 3.32 (2H, q, J = 6.8 Hz), 2.77 (2H, t, J = 6.8 Hz), 2.60 (2H, t, J = 7.0 Hz), 2.14 (2H, m), 1.42 (9H, s); 13C NMR (75 MHz, CDCl3) δ: 155.9, 154.0, 150.6, 128.9, 119.3, 116.8, 112.5, 112.0, 79.5, 66.3, 55.7, 40.8, 31.2, 28.4, 25.7, 14.4; EI-HRMS: m/z; found: 334.1924 (M+). C18H26N2O4 (M+) requires 334.1893. Anal. Found: C, 63.62; H, 7.82; N, 8.06. C18H26N2O4 requires C, 64.65; H, 7.84; N, 8.38.</p><!><p>93% Yield via hydrogenolysis of 4h; mp 90–91 °C; IR (CHCl3) νmax 3681, 3615, 3449, 3021, 2405, 1700, 1501, 1165 cm−1; 1H NMR (400 MHz, CDCl3) δ: 6.74 (3H, m), 4.87 (1H, br s), 4.00 (4H, m), 3.88 (1H, br s), 3.75 (3H, s), 3.31 (2H, q, J = 6.7 Hz), 2.78 (2H, t, J = 6.7 Hz), 1.42 (9H, s); 13C NMR (100 MHz, CDCl3) δ: 156.5, 153.9, 151.4, 128.9, 117.1, 112.6, 112.0, 79.7, 70.6, 61.6, 55.9, 41.0, 32.6, 28.6; EI-HRMS: m/z; found: 311.17278 (M+). C16H25NO5 (M+) requires 311.17327. Anal. Found: C, 61.70; H, 8.03; N, 4.14. C16H25NO5 requires C, 61.72; H, 8.09; N, 4.50.</p><!><p>100% and 95% Yields in the alkylation and hydrogenolysis steps respectively; mp 44–46 °C; IR (CHCl3) νmax 3681, 3623, 3449, 2399, 1700, 1505, 1205 cm−1; 1H NMR (300 MHz, CDCl3) δ: 6.80 (1H, d, J = 9.6 Hz), 6.70 (2H, m), 4.78 (1H, br s), 4.06 (2H, t, J = 5.9 Hz), 3.87 (2H, t, J = 5.9 Hz), 3.75 (3H, s), 3.34 (2H, q, J = 6.6 Hz), 2.76 (2H, t, J = 6.6 H), 2.37 (1H, br s), 2.03 (2H, quin, J = 5.9 Hz), 1.41 (9H, s); 13C NMR (100 MHz, CDCl3) δ: 155.9, 153.6, 151.1, 128.9, 116.8, 112.4, 112.0, 79.0, 66.9, 65.9, 55.7, 40.7, 31.0, 29.9, 28.4. Anal. Found: C, 62.97; H, 8.35; N, 4.13. C17H27NO5 requires C, 62.75; H, 8.36; N, 4.30.</p><!><p>Trifluoroacetic acid (0.2 mL) was added to a solution of the N-Boc carbamate 4 (1 mmol) in CH2Cl2 (2 mL) at 0 °C, and the solution stirred for 2 h. Diisopropylethylamine (0.4 mL) was added, the solvent evaporated in vacuo and the crude amine dried under vacuum for 1 h. Thiocarbonyl reagent 7 (1.3 mmol) was added to the crude amine in DMF (5 mL) and the mixture stirred at 100 °C for 16 h. The mixture was then poured into ice-cold water (5 mL) and stirred for 30 min. The precipitate formed was filtered and washed with cold water (2× 5 mL) or alternatively extracted into ethyl acetate and the residue purified by column chromatography using ethyl acetate/light petroleum mixtures as eluent. Alternatively, the condensation reaction with 7 could be carried out in THF (5 mL) at room temperature for 20 h. Following evaporation of solvent, the residue was subjected directly to column chromatography to afford thioureas 6.</p><!><p>Using THF/rt, (65%); mp 162–163 °C; IR (CHCl3) νmax 3696, 3413 (NH), 3167, 2963 (C–H), 1512, 1472 (C=S), 1212 (C–N) cm−1; 1H NMR (400 MHz, CDCl3) δ: 11.18 (1H, br s, NH), 8.95 (1H, br s, NH), 8.08 (1H, d, J = 2.4 Hz), 7.67 (1H, dd, J = 2.4 Hz, 8.8 Hz), 6.81 (1H, d, J = 3.2 Hz), 6.78 (1H, d, J = 8.8 Hz), 6.74 (1H, dd, J = 3.2, 8.8 Hz), 6.72 (1H, d, J = 8.8 Hz), 4.02 (2H, m), 3.87 (2H, t, J = 6.4 Hz), 3.75 (3H, s), 2.99 (2H, t, J = 6.4 Hz), 1.80 (2H, m), 1.04 (3H, t, J = 7.4 Hz); 13C NMR (100 MHz, CDCl3) δ: 179.0 (C=S), 153.3, 151.7, 151.5, 146.8, 141.1, 128.5, 117.7, 113.2, 112.6, 112.2, 111.5, 70.2, 55.6, 45.8, 30.0, 22.8, 10.7. Anal. Found: C, 50.77; H, 5.21; N, 9.50; S, 7.14%. C18H22BrN3O2S requires C, 50.95; H, 5.23; N, 9.90; S, 7.55.</p><!><p>Using DMF/100 °C, (17%); mp 121–122°C; IR (CHCl3) νmax 3691, 3416 (NH), 3307 (≡CH), 3174, 1602, 1137 cm−1; 1H NMR (400 MHz, CDCl3) δ: 11.13 (1H, br s, NH), 8.29 (1H, br s, NH), 8.12 (1H, d, J = 2.4 Hz), 7.69 (1H, dd, J = 2.4, 8.8 Hz), 6.95 (1H, d, J = 8.8 Hz), 6.82 (1H, d, J = 3.1 Hz), 6.76 (1H, dd, J = 3.1, 8.8 Hz), 6.59 (1H, d, J = 8.8 Hz), 4.67 (1H, d, 1H, J = 2.2 Hz), 4.01 (1H, q, J = 6.7 Hz), 3.76 (3H, s), 3.01 (2H, t, J = 6.7 Hz), 2.47 (1H, t, J = 2.2 Hz); 13C NMR (75 MHz, CDCl3) δ: 179.1 (C=S), 154.2, 151.7, 150.0, 146.7, 141.1, 129.2, 117.6, 113.5, 113.3, 112.6, 111.5, 79.0, 75.3, 56.8, 55.5, 45.7, 29.8. Anal. Found: C, 51.23; H, 4.50; N, 8.62%. C18H18BrN3O2S requires C, 51.44; H, 4.32; N, 10.00.</p><!><p>Using DMF/100 °C, (25%); mp 155–156 °C; IR (CHCl3) νmax 3691, 3415, 3308 (≡CH), 3176, 3048, 2360 (C≡C), 1591, 1511, 1138 cm−1; 1H NMR (300 MHz, CDCl3) δ: 11.13 (1H, br s, NH), 8.44 (1H, br s, NH), 8.10 (1H, d, J = 2.4 Hz), 7.68 (1H, dd, J = 2.4, 8.7 Hz), 6.76 (3H, m), 6.61 (1H, d, J = 8.7 Hz), 4.03 (4H, m), 3.75 (3H, s), 3.00 (2H, t, J = 6.6 Hz), 2.68 (2H, td, J = 2.6, 6.9 Hz), 2.04 (1H, t, J = 2.6 Hz); 13C NMR (75δMHz, CDCl3) δ: 179.2 (C=S), 153.8, 151.6, 150.9, 146.9, 141.2, 128.9, 117.8, 113.1, 112.8, 112.7, 111.6, 80.8, 69.9, 66.9, 55.6, 45.8, 30.0, 19.8. Anal. Found: C, 52.01; H, 4.46; N, 9.00%. C19H20BrN3O2S requires C, 52.54; H, 4.64; N, 9.67.</p><!><p>Using THF/rt, (69%); mp 140–141 °C; IR (CHCl3) νmax 3684, 3415 (NH), 3308 (≡CH), 3152, 3018, 2956, 2245 (C≡C), 1512, 1468 (C=S), 1226 (C–N) cm−1; 1H NMR (400 MHz, CDCl3) δ: 11.18 (1H, br s, NH), 8.83 (1H, br s, NH), 8.10 (1H, d, J = 2.4 Hz), 7.68 (1H, dd, J = 2.4 Hz, 8.8 Hz), 6.81 (1H, d, J = 9.0 Hz), 6.81 (1H, d, J = 3.0 Hz), 6.75 (1H, dd, J = 3.0, 9.0 Hz), 6.70 (1H, d, J = 8.8 Hz), 4.01 (4H, m), 3.75 (3H, s), 2.98 (2H, t, J = 6.8 Hz), 2.41 (2H, td, J = 2.4, 6.8 Hz), 2.02 (2H, m), 1.96 (1H, t, J = 2.4 Hz); 13C NMR (100 MHz, CDCl3) δ: 179.1 (C=S), 153.5, 151.7, 151.2, 146.7, 141.1, 128.5, 117.7, 113.2, 112.6, 112.3, 111.5, 83.5, 68.9, 66.9, 55.6, 45.7, 30.0, 28.4, 15.4. Anal. Found: C, 53.54; H, 4.78; N, 7.81; S, 5.62%. C20H22BrN3O2S requires C, 53.58; H, 4.95; N, 9.37; S, 7.15.</p><!><p>Using THF/rt, (60%); mp 149–150 °C; IR (CHCl3) νmax 3690, 3410 (NH), 3018, 2239 (C≡C), 1510, 1469 (C=S), 1207 (C–N) cm−1; 1H NMR (400 MHz, CDCl3) δ: 11.12 (1H, br s, NH), 8.58 (1H, br s, NH), 8.11 (1H, d, J = 2.4 Hz), 7.68 (1H, dd, J = 2.4 Hz, 8.8 Hz), 6.93 (1H, d, J = 8.8 Hz), 6.81 (1H, d, J = 3.2 Hz), 6.75 (1H, dd, J = 3.2 Hz, 8.8 Hz), 6.65 (1H, d, J = 8.8 Hz), 4.61 (2H, q, J = 2.4 Hz), 4.01 (2H, m), 3.75 (3H, s), 3.00 (2H, t, J = 6.8 Hz), 1.83 (3H, t, J = 2.4 Hz); 13C NMR (100 MHz, CDCl3) δ: 179.1 (C=S), 153.9, 151.6, 150.4, 146.8, 141.1, 129.1, 117.5, 113.6, 113.1, 112.6, 111.6, 83.4, 74.5, 57.5, 55.6, 45.8, 29.8, 3.7. Anal. Found: C, 52.51; H, 4.53; N, 8.95; S, 6.31%. C19H20BrN3O2S requires C, 52.54; H, 4.64; N, 9.67; S, 7.38.</p><!><p>Using THF/rt, (67%); mp 153–154 °C; IR (CHCl3) νmax 3681, 3413 (N–H), 3041, 1509 (C=C), 1422 (C=S), 1212 (C–N) cm−1; 1H NMR (400 MHz, CDCl3) δ: 11.11 (1H, br s, NH), 8.13 (1H, br s, NH), 8.10 (1H, d, J = 2.4 Hz), 7.69 (1H, dd, J = 2.4, 8.7 Hz), 6.82 (1H, d, J = 2.9 Hz), 6.80 (1H, d, J = 9.0 Hz), 6.74 (1H, dd, J = 2.9, 9.0 Hz), 6.55 (1H, d, J = 8.7 Hz), 6.05 (1H, ddt, J = 5.1, 10.5, 17.3 Hz), 5.39 (1H, dq, J = 1.6, 17.3 Hz), 5.26 (1H, dq, J = 1.6, 10.5 Hz), 4.49 (2H, dt, J = 1.6, 5.1 Hz), 4.01 (2H, m), 3.76 (3H, s), 3.01 (2H, t, J = 6.6 Hz); 13C NMR (100 MHz, CDCl3) δ: 179.1 (C=S), 153.5, 151.5, 151.0, 146.9, 141.1, 133.6, 128.7, 117.7, 117.0, 113.0, 112.8, 112.6, 111.5, 69.5, 55.6, 45.8, 30.0. Anal. Found: C, 51.32; H, 4.62; N, 9.56%. C18H20BrN3O2S requires C, 51.18; H, 4.78; N, 9.95.</p><!><p>Using THF/rt, (37%); mp 141–142 °C as a yellow solid; IR (CHCl3) νmax 3691, 3416, 1602, 1505, 1138 cm−1; 1H NMR (300 MHz, CDCl3) δ: 11.18 (1H, br s, NH), 8.80 (1H, br s, NH), 8.03 (1H, d, J = 2.6 Hz), 7.65 (1H, dd, J = 2.6, 8.7 Hz), 7.37 (5H, m, ArH), 6.86 (1H, d, J = 9.0 Hz), 6.84 (1H, d, J = 3.2 Hz), 6.74 (1H, dd, J = 3.2, 9.0 Hz), 6.67 (1H, d, J = 8.7 Hz), 5.02 (2H, s), 4.03 (2H, m), 3.75 (3H, s), 3.03 (2H, t, J = 6.6 Hz); 13C NMR (75 MHz, CDCl3) δ: 179.0 (C=S), 153.6, 151.6, 151.1, 146.7, 141.1, 137.3, 128.8, 128.5, 127.8, 127.2, 117.7, 113.2, 112.9, 112.6, 111.5, 70.7, 55.6, 45.7, 30.0; HRMS (EI) m/z; found: 471.0594 (M+), C22H22BrN3O2S (M+) requires 471.0616. Anal. Found: C, 55.48; H, 4.68; N, 8.59; S, 6.17%. C22H22BrN3O2S requires C, 55.94; H, 4.69; N, 8.90; S, 6.79.</p><!><p>Using THF/rt, (58% yield); mp 110–111 °C; IR (CHCl3) νmax 3692, 3416 (NH), 3175, 3016, 1506, 1475 (C=S) cm−1; 1H NMR (400 MHz, CDCl3) δ: 11.09 (1H, br s, NH), 8.38 (1H, br s, NH), 8.04 (1 H, d, J = 2.4 Hz), 7.66 (1H, dd, J = 2.4 Hz, 8.8 Hz), 7.31 (5H, m, ArH), 6.81 (1H, d, J = 8.8 Hz), 6.80 (1H, d, J = 3.2 Hz), 6.74 (1H, dd, J = 3.2 Hz, 8.8 Hz), 6.58 (1H, d, J = 8.8 Hz), 4.63 (2H, s), 4.11 (2H, t, J = 4.8 Hz), 4.03 (2H, m), 3.84 (2H, t, J = 4.8 Hz), 3.75 (3H, s), 3.01 (2H, t, J = 6.6 Hz); 13C NMR (100 MHz, CDCl3) δ: 179.0 (C=S), 153.6, 151.8, 151.3, 146.6, 141.0, 138.1, 128.8, 128.5, 127.7, 127.7, 117.7, 113.4, 112.8, 112.9, 111.6, 73.4, 68.9, 68.5, 55.6, 45.7, 30.0. Anal. Found: C, 56.07; H, 4.86; N, 7.37; S, 5.80%. C24H26BrN3O3S requires C, 55.82; H, 5.07; N, 8.14; S, 6.21.</p><!><p>Using DMF/rt, (62%); mp 128–130 °C; IR (CHCl3) νmax 3693, 3416, 3307, 3165, 2961, 1506, 1475, 1223 cm−1; 1H NMR (300 MHz, CDCl3) δ: 11.19 (1H, br s, NH), 9.24 (1H, br s, NH), 8.07 (1H, d, J = 2.6 Hz), 7.66 (1H, dd, J = 2.6, 8.7 Hz), 6.81–6.77 (3H, m), 6.72 (1H, dd, J = 2.9, 9.2 Hz), 4.25 (2H, d, J = 2.4 Hz), 4.09 (2H, t, J = 4.8 Hz), 4.01 (2H, q, J = 6.6 Hz), 3.88 (2H, t, J = 4.8 Hz), 3.74 (3H, s), 2.99 (2H, t, J = 6.6 Hz), 2.45 (1H, t, J = 2.4 Hz); 13C NMR (75 MHz, CDCl3) δ: 178.9 (C=S), 153.6, 151.7, 151.1, 146.6, 141.0, 128.8, 117.6, 113.4, 112.9, 112.6, 111.5, 79.5, 74.7, 68.4, 68.2, 58.5, 55.5, 45.6, 29.9; HRMS (EI) m/z; found: 463.0578 (M+). C20H22BrN3O3S (M+) requires 463.0565. Anal. Found: C, 51.76; H, 4.72; N, 8.66; S, 6.86%. C20H22BrN3O3S requires C, 51.73; H, 4.78; N, 9.05; S, 6.91.</p><!><p>Using DMF/rt, (60%); mp 91–92 °C; IR (CHCl3) νmax 3691, 3416, 3307, 3166, 2935, 1506, 1475, 1266 cm−1; 1H NMR (300 MHz, CDCl3) δ: 11.11 (1H, br s, NH), 8.67 (1H, br s, NH), 8.07 (1H, d, J = 2.4 Hz), 7.67 (1H, dd, J = 2.4, 8.7 Hz), 6.81–6.78 (2H, m), 6.73 (1H, dd, J = 3.0, 8.7 Hz), 6.67 (1H, d, J = 8.7 Hz), 4.21 (2H, d, J = 2.1 Hz), 4.07 (2H, t, J = 4.9 Hz), 4.01 (2H, q, J = 6.6 Hz), 3.85 (2H, t, J = 4.9 Hz), 3.73 (7H, m), 2.99 (2H, t, J = 6.6 Hz), 2.42 (1H, t, J = 2.1 Hz); 13C NMR (75 MHz, CDCl3) δ: 179.1 (C=S), 153.6, 151.6, 151.2, 146.8, 141.1, 128.8, 117.7, 113.2, 112.8, 112.6, 111.5, 79.6, 74.6, 70.6, 70.0, 69.2, 68.4, 58.5, 55.6, 45.7, 30.0; HRMS (EI) m/z; found: 507.0821 (M+). C22H26BrN3O4S (M+) requires 507.0827. Anal. Found: C, 52.02; H, 4.10; N, 7.87; S, 6.24%. C22H26BrN3O4S requires C, 51.97; H, 5.15; N, 8.26; S, 6.31.</p><!><p>Using THF/rt, (50%); mp 161–164 °C; IR (CHCl3) νmax 3681 (N–H), 3015, 1519, 1469 (C=S), 1212 (C–N) cm−1; 1H NMR (400 MHz, CDCl3) δ: 11.12 (1H, br s, NH), 8.32 (1H, br s, NH), 8.11 (1H, d, J = 2.4 Hz), 7.69 (1H, dd, J = 2.4, 9.1 Hz), 6.83 (1H, d, J = 2.4 Hz), 6.71 (2H, m), 6.61 (1H, d, J = 9.1 Hz), 4.62 (2H, s), 4.05 (2H, m), 3.78 (3H, s), 3.74 (3H, s), 3.06 (2H, t, J = 6.8 Hz); 13C NMR (100 MHz, CDCl3 ppm) δ: 179.2 (C=S), 169.6, 154.2, 151.5, 150.3, 146.9, 141.2, 129.2, 117.8, 113.0, 112.7, 112,6, 111.6, 66.3, 55.5, 52.2, 45.7, 29.9; HRMS (ES) m/z; found: [M+H]+, 454.0437. Calcd. for C18H21BrN3O4S (M+H), 454.0436.</p><!><p>Using THF/rt, (68%); mp 155–156 °C; IR (CHCl3) νmax 3681, 3413, 3022 (NH), 2239 (C≡N), 1505, 1469 (C=S), 1216 cm−1; 1H NMR (400 MHz, CDCl3) δ: 11.20 (1H, br s, NH), 8.72 (1H, br s, NH), 8.10 (1H, d, J = 2.4 Hz), 7.70 (1H, dd, J = 2.4 Hz, 8.8 Hz), 6.92 (1H, d, J = 8.8 Hz), 6.85 (1H, d, 3.2 Hz), 6.78 (1H, dd, J = 3.2, 8.8 Hz), 6.69 (1H, d, J = 8.8 Hz), 4.75 (2H, s), 3.99 (2H, q, J = 6.8 Hz), 3.76 (3H, s) 3.01 (2H, t, J = 6.8 Hz); 13C NMR (100 MHz, CDCl3) δ: 179.4 (C=S), 155.4, 151.6, 149.0, 146.7, 141.3, 129.8, 117.9, 115.4, 114.0, 113.3, 112.8, 111.9, 55.6, 54.9, 45.6, 29.6. Anal. Found: C, 48.38; H, 4.04; N, 12.38; S; 6.76%. C17H17BrN4O2S requires C, 48.46; H, 4.07; N, 13.30; S, 7.61.</p><!><p>Using THF/rt, (53%); mp 165–166 °C; IR (CHCl3) νmax, 3413 (NH), 3167, 2964 (C–H), 2246 (C≡N), 1505, 1465 (C=S), 1239 (C–N) cm−1; 1H NMR (400 MHz, CDCl3) δ: 11.26 (1 H, br s, NH), 9.34 (1H, br s, NH), 8.08 (1H, d, J = 2.4 Hz), 7.67 (1H, dd, J = 2.4 Hz, 8.8 Hz), 6.82 (1H, d, J = 3.2 Hz), 6.81 (1H, d, J = 8.8 Hz), 6.77 (1H, d, J = 8.8 Hz), 6.74 (1H, dd, J = 3.2 8.8 Hz), 3.99 (4H, m), 3.74 (3H, s), 2.98 (2H, t, J = 6.8 Hz), 2.62 (2H, t, J = 7.0 Hz), 2.14 (2H, m); 13C NMR (75 MHz, CDCl3) δ: 179.0 (C=S), 153.7, 151.7, 150.6, 146.5, 141.1, 128.4, 119.2, 117.7, 113.4, 112.7, 112.2, 111.5, 66.0, 55.5, 45.6, 29.9, 25.6, 14.4. Anal. Found: C, 50.61; H, 4.78; N, 12.37; S, 6.92%. C19H21BrN4O2S requires C, 50.78; H, 4.71; N, 12.47; S, 7.13.</p><!><p>Using THF/rt, (65%); mp 162–163 °C; IR (CHCl3) νmax 3623, 3413, 3181 (NH, OH), 2928, 1505 (C–H), 1472 (C=S) 1227 (C–N) cm−1; 1H NMR (400 MHz, CDCl3) δ: 11.16 (1H, br s, NH), 8.47 (1H, br s, NH), 8.12 (1H, d, J = 2.4 Hz), 7.70 (1H, dd, J = 2.4, 8.8 Hz), 6.81 (1H, d, J = 2.8 Hz), 6.77 (1H, d, J = 8.8 Hz), 6.73 (1H. dd, J = 2.8, 8.8 Hz), 6.66 (1H, d, J = 8.8 Hz), 3.99 (6H, m), 3.77 (3H, s), 3.03 (2H, t, J = 7.2 Hz), 2.66 (1H, br s); 13C NMR (75 MHz, CDCl3) δ: 179.4 (C=S), 153.7, 151.6, 151.3, 146.7, 141.3, 128.3, 117.4, 113.4, 112.8, 112.5, 111.7, 70.3, 61.5, 55.6, 45.7, 30.3. Anal. Found: C, 47.96; H, 4.61; N, 9.77; S, 7.01%. C17H20BrN3O3S requires C, 47.89; H, 4.75; N, 9.86; S, 7.52.</p><!><p>Using THF/rt, (51%), mp 163–165 °C; IR (CHCl3) νmax 3681 (N–H), 3406 (O–H) 3014, 1509, 1472 (C=S), 1216 (C–N) cm−1; 1H NMR (400 MHz, CDCl3) δ: 11.15 (1H, br s, NH), 8.46 (1H, br s, NH), 8.10 (1H, d, J = 2.4 Hz), 7.68 (1H, dd, J = 2.4, 8.8 Hz), 6.81 (1H, d, J = 8.8 Hz), 6.81 (1H, d, J = 2.9 Hz), 6.74 (1H, dd, J = 2.9, 8.8 Hz), 6.62 (1H, d, J = 8.8 Hz), 4.05 (2H, t, J = 5.9 Hz), 4.00 (2H, m), 3.87 (2H, q, J = 5.9 Hz), 3.76 (3H, s), 2.98 (2H, t, J = 6.6 Hz), 2.05 (2H, quin, J = 5.9 Hz), 2.04 (1H, br s, OH); 13C NMR (101 MHz, CDCl3 ppm) δ: 179.1 (C=S), 153.5, 151.6, 151.3, 146.8, 141.2, 128.4, 117.7, 113.2, 112.7, 112.4, 111.6, 65.9, 60.1, 55.6, 45.8, 32.3, 30.0. Anal. Found: C, 49.37; H, 4.88; N, 9.08; S, 6.60%. C18H22BrN3O4S requires C, 49.09; H, 5.04; N, 9.54; S, 7.28.</p><!><p>The following reagents were obtained through the NIH AIDS Research and Reference Reagent Program, Division of AIDS, NIAID, NIH: MT-2 cells and Nevirapine-Resistant HIV-1 (N119) from Dr. Douglas Richman and HTLV-IIIB/H9 from Dr. Robert Gallo.</p><p>Antiviral activity and cellular toxicity were determined using the MTT colorimetric method.19 MT-2 cells29 at a concentration of 1 × 105 cells per millilitre were infected with wild-type HIV-IIIB30 or Y181C mutant virus31 at a multiplicity of infection (MOI) of 0.1. Infected and mock-infected cells were incubated in growth medium (RPMI 1640, 10% dFBS, kanamycin) for 5 days with varying concentrations of each compound being tested in triplicate in a 96-well plate. MTT, a cell-permeable tetrazolium dye was then added to each well. After 5 h, acidified isopropanol was added to lyse the cells and stop the reaction. The plates were gently shaken overnight, and the absorbance measured at 595 nm on a plate reader. The average of these triplicate samples was then plotted versus inhibitor concentration to generate dose–response curves. The 50% effective concentration (EC50) and 50% cytotoxic concentration (CC50) of the compounds were defined as the concentrations required to inhibit viral replication and to reduce the number of viable cells by 50%, respectively.</p><!><p>6 nM RT (active sites based on pre-steady-state active site determination) was pre-incubated for at least 15 min with 1 μM 5′-radiolabeled primer/template prior to mixing with appropriate concentrations of inhibitor and allowed to incubate for a minimum of 15 additional minutes on ice. DMSO concentrations were kept constant at less than 2%. DMSO alone was added as a no inhibitor control for each set of experiments. Reactions were initiated with the addition of 5 μM dTTP and 10 mM MgCl2 and were quenched after 15 min at 37 °C with 0.3 MEDTA. All concentrations represent final concentrations after mixing. Reaction products were subjected to 20% denaturing polyacrylamide gel-electrophoresis and quantitated on a Bio-Rad Molecular Imager FX. Product formation was plotted as a function of inhibitor concentration and fitted to a hyperbola to generate IC50 curves. IC50 values are defined as the concentration of inhibitor that inhibits steady-state single nucleotide incorporation by 50%.</p><!><p>Structures of two recent bifunctional inhibitors based on PETT compounds with d4 T.</p><p>Structures of HI-236 and Trovirdine.</p><p>HI-236 docked into the HIV pocket.</p><p>The similarities in orientation between the AutoDock models of HI-236 and 6k.</p><p>Docking of ester 6k.</p><p>Docking of alcohol 6o.</p><p>Reagents and conditions: (i) BnBr, K2CO3, EtOH, reflux, 16 h; 95%; (ii) CH3NO2, NH4OAc, 70 °C, 14 h; 75%; (iii) LiAlH4, THF, reflux, 4 h; (iv) (Boc)2O, Et3N, CH3CN, rt, overnight; 76% (2 steps); (v) H2, Pd/C, EtOH, rt, 5 h; 66%.</p><p>Reagents and conditions: (i) K2CO3, CH3CN, 80 °C, 20 h or NaH, DME, 80 °C, 20 h with ROTs.</p><p>Reagents and conditions: (i) BnPEGBr (n = 0 or 1), K2CO3, CH3CN, 80 °C, 20 h or NaH, DME, 80 °C, 20 h; (ii) H2, Pd/C, EtOH, rt, 18 h (iii) TsCl, Et3N, DMAP, CH2Cl2, 0 °C–rt, 20 h (iv) propargyl alcohol, NaH, THF, 70 °C, 5 h.</p><p>Reagents and conditions: (i) CF3COOH, CH2Cl2, 0 °C, 1 h and then DIEA, 0 °C, 10 min, followed by 8, THF, rt, 2 h.</p><p>Reagents and conditions: (I) 2-amino-5-bromopyridine, CH3CN, rt, 120.</p><p>O-Alkylation of 3 to afford 4 as precursors of C-2 substituted HI-236 derivatives.</p><p>Yields for converting 4–6.</p><p>DMF at 100 °C; the others involved THF at rt.</p><p>Antiviral activity and cytotoxicity assaya results for 6a–o against HIV-1 (IIIB) in MT-2 cell culture.</p><p>Ref. 19; MOI was 0.1.</p><p>Effective concentration that inhibits viral-mediated T-cell death by 50%, determined by averaging samples of each concentration in triplicate.</p><p>Concentration that kills 50% of the T-cells, also determined by averaging triplicate samples.</p><p>In vitro therapeutic index (CC50/EC50).</p><p>Comparison of cell culture versus in vitro RT inhibition (nM) for selected derivatives of 6.</p>

PubMed Author Manuscript

One-Step Selective Exoenzymatic Labeling (SEEL) Strategy for the Biotinylation and Identification of Glycoproteins of Living Cells

Technologies that can visualize, capture, and identify subsets of biomolecules that are not encoded by the genome in the context of healthy and diseased cells will offer unique opportunities to uncover the molecular mechanism of a multitude of physiological and disease processes. We describe here a chemical reporter strategy for labeling of cell surface glycoconjugates that takes advantage of recombinant glycosyltransferases and a corresponding sugar nucleotide functionalized by biotin. The exceptional efficiency of this method, termed one-step selective exoenzymatic labeling, or SEEL, greatly improved the ability to enrich and identify large numbers of tagged glycoproteins by LC\xe2\x80\x93MS/MS. We further demonstrated that this labeling method resulted in far superior enrichment and detection of glycoproteins at the plasma membrane compared to a sulfo-NHS-activated biotinylation or two-step SEEL. This new methodology will make it possible to profile cell surface glycoproteomes with unprecedented sensitivity in the context of physiological and disease states.

one-step_selective_exoenzymatic_labeling_(seel)_strategy_for_the_biotinylation_and_identification_of

INTRODUCTION<!>Synthesis of Modified Sugar Nucleotides and Enzymology<!>Labeling and Trafficking of Cell Surface Glycoproteins of Living Cells<!>Proteomic Analysis of Labeled Glycoproteins Following Lysosomal Disruption<!>Proteomic Analysis of Cell Surface Proteins Labeled by SEEL and Amino-Reactive Reagents<!>DISCUSSION<!>Methods<!>Materials<!>Cell Lines and Culture<!>Metabolic Labeling, Two-Step SEEL, or One-Step SEEL<!>SEEL and Chase Experiment of HeLa Cells with Chloroquine Treatment<!>Cell Staining and Imaging<!>Cell Surface Labeling with EZ-Link Sulfo-NHS-LC-Biotin<!>Immunoblotting, Immunoprecipitation, and Silver Staining<!>Proteomic Analysis

<p>The bioorthogonal chemical reporter strategy offers exciting possibilities to interrogate biomolecules of living cells that are not encoded by the genome.1 In this approach, the biosynthetic machinery of a cell is hijacked by feeding a metabolic precursor functionalized with a chemical reporter for incorporation into a target class of biomolecules, such as lipids or glycoconjugates. Next, a bioorthogonal reaction is performed to modify the reporter with a biophysical probe for visualization or enrichment. To date, azido modification is the most widely employed bioorthogonal chemical reporter because of its small size and inertness to most components in a biological environment. It can for example be tagged by Staudinger ligation using modified phosphines,2 copper(I)-catalyzed cycloaddition with terminal alkynes (CuAAC),3 or strain-promoted alkyne–azide cycloaddition (SPAAC).4 In particular, SPAAC is attractive because it avoids the cellular toxicity associated with copper ions without compromising alkyne reactivity.</p><p>In an alternative approach, chemical reporters can be introduced into biomolecules of living cells by performing glycosyltransferase reactions at the cell surface.5 This process, for which we coined the term SEEL (selective exoenzymatic labeling), takes advantage of recombinant glycosyltransferases and a corresponding functionalized nucleotide sugar to install chemical reporters on cell surface acceptor glycans. For example, we have demonstrated that recombinant ST6Gal1 sialyltransferase and CMP-Neu5Ac9N3 (1) can be exploited for the selective labeling of N-linked glycans of living cells with azido-modified sialic acid. 5g,6 The attraction of the SEEL approach is that it only labels a specific class of cell surface molecules (e.g., N- vs O-glycans) and does not rely on feeding with metabolic substrates that must compete with natural sugar precursor pools. In light of the direct nature of this method, we explored whether SEEL can be accomplished in a single step by employing CMP-Neu5Ac modified by biotin.7 Such an approach would circumvent possible limitations associated with bioorthogonal reactions, such as side reactions,8 and decrease the time needed for labeling.</p><p>We have found that a one-step SEEL procedure, employing exogenously administered CMP-Neu5Ac analogues modified at C-5 or C-9 with biotin (compounds 2 and 3, Figure 1),5d,7a not only is feasible but dramatically improves cell surface labeling of glycoconjugates compared to two-step SEEL or metabolic labeling. The new methodology offers exciting possibilities to track, capture, and identify subsets of cell surface glycoconjugates with unprecedented sensitivity in whole cells. We have exploited the high efficiency labeling with compound 2 to enrich and identify tagged glycoproteins by LC–MS/MS, and could monitor the internalization and degradation of labeled cell surface glycoproteins over time, and identify a subset of glycoproteins that are subject to accumulation upon lysosomal disruption by chloroquine. Lastly, the new labeling strategy was compared with a commonly employed kit for cell surface biotinylation and was found to be vastly superior for enrichment of cell surface glycoproteins. Many diseases exhibit lysosomal dysfunction and altered recycling of cell surface glycoproteins.9 For example, Niemann-Pick type C (NPC) disease is characterized by impaired cholesterol efflux from late endosomes and lysosomes and secondary accumulation of lipids. Previously, we employed the chemical reporter strategy to demonstrate an unrecognized accumulation of glycoconjugates in endocytic compartments of NPC1-null and NPC2-deficient fibroblasts.9c The endosomal accumulation of sialylated glycoproteins was attributed to impaired recycling as opposed to altered fusion of vesicles. It is our expectation that identification of altered cell surface residency of glycoproteins in diseases such as NPC using the new SEEL methodology will uncover the molecular mechanisms that cause the phenotypes of these diseases, which in turn may lead to the design of new therapeutic strategies.</p><!><p>It is known that sialyltransferases tolerate modifications at C-5 and C-9,7b and it has for example been demonstrated that a CMP-sialic acid derivative having biotin at C-9 can be transferred to LacNAc.5d,7a Therefore, CMP-sialic acid derivatives 2 and 3 were prepared having biotin at the C-5 and C-9 positions. Thus, condensation of C-5 and C-9 azido modified sialic acid 5a and 5b with CTP in the presence of the recombinant CMP-sialic acid synthetase from Neisseria meningitis10 and the inorganic pyrophosphatase from Saccharomyces cerevisiae gave readily CMP-Neu5Ac derivatives 6 and 1, respectively. A CuAAC3a of 6 and 1 with alkyne modified biotin 7 in the presence of CuSO4, ascorbic acid, and TBTA gave, after purification by size exclusion column chromatography over Biogel P2, target compounds 2 and 3, respectively (Scheme 1). An alternative approach in which compounds 5a and 5b were modified by 7 followed by condensation with CTP in the presence of the recombinant CMP-sialic acid synthetase did not give the required products because the enzyme did not tolerate the bulky substituents at C-5 or C-9.</p><p>Incubation of N-acetyllactosamine and CMP-sialic acid derivatives 1, 2, or 3 in the presence of ST6Gal1 led to almost quantitative formation of Neu5AcRα(2,6)Galβ(1,4)GlcNAc 8S, 9S, and 10S, respectively highlighting that, in addition to azide, the C-5 and C-9 triazole linked biotin moieties are tolerated by the transferase. Kinetic analysis of the enzymatic transformation showed that the modifications of 2 and 3 had no or a marginal impact on the Km value (2, 3, and CMP-Neu5Ac, Km = 53, 108, and 54 μM, respectively) with no appreciable influence on Vmax.</p><!><p>HeLa cells were incubated with CMP-Neu5Ac derivatives 1, 2, and 3 (compounds shown in Figure 1) in the presence of ST6Gal1 for 2 h at 37 °C to examine the efficiency of one-step SEEL in comparison to the previously reported5g two-step SEEL method. The length of labeling and the chosen enzyme and substrate donor concentrations were optimized in order to achieve maximal labeling with minimal effects on cell viability. The labeling was performed with and without prior treatment with Vibrio cholerae sialidase. In addition, cells were metabolically labeled with azido containing sialosides by feeding peracetylated N-azidoacetylmannosamine (Ac4ManNAz).2a The cells modified by azido-containing glycoconjugates were biotinylated by exposure to sulfated dibenzocyclooctynylamide containing biotin (S-DIBO-biotin 4, 30 μM, Figure 1)11 for 1 h. The efficiency of the various surface-labeling procedures was determined by SDS–PAGE of cell lysates followed by Western blotting using an anti-biotin antibody conjugated with HRP. These experiments demonstrated that prior neuraminidase treatment of the cells increased the efficiency of labeling for both one- and two-step SEEL (consistent with the presence of abundant highly sialylated N-glycan structures in these cells which reduce the abundance of SEEL acceptors) (Figure 2a). Remarkably, the one-step labeling procedure using 2 and 3 gave a robust biotin signal even after exposure of the blot for only seconds, whereas the two-step protocol with compound 1 or metabolic labeling with ManNAz gave a faint biotin signal and longer exposure time was required to make the glycoproteins visible. Furthermore, labeling with 2 was more efficient than with 3 probably because it is a better substrate for ST6Gal1. The different labeling methods (metabolic, two-step, and one-step SEEL) were compared in three other cell lines (Figure S1) to gauge the generality of this striking improvement in labeling with one-step SEEL. In each case, the one-step SEEL procedure gave by far more robust labeling.</p><p>Next, we performed proteomic analysis on HeLa cells labeled by two-step SEEL or one-step SEEL (with ST6Gal1) to compare the labeling efficiency of the two methods (Table S1). 294 proteins were assigned at <1% false-discovery rate using one-step SEEL, whereas 174 proteins were assigned with two-step SEEL (Figure 2b). One-step SEEL detected 140 glycoproteins not observed using the two-step method, while two-step SEEL identified only 20 unique glycoproteins. The total spectral count of all assigned proteins from one-step SEEL was 2.1 times higher than that from two-step SEEL. Out of the 154 commonly found proteins in both methods, 85 proteins showed more than 2-fold increase in their spectral counts, while 57 proteins showed more than 3-fold increase and 25 proteins showed more than 5-fold increase in one-step SEEL. The relative spectral counts of the top 20 most abundant commonly assigned glycoproteins were compared in Figure 2c. Among these 20 proteins, PTPRF (752%), BSG (423%), ROBO1 (359%), and IGF2R (342%) showed at least a 3-fold improvement in assigned spectral counts. These data demonstrate the remarkably superior labeling eficiency of sialoglycoproteins at the cell surface when one-step SEEL is utilized.</p><p>Next, we examined whether the biotin containing sialosides at the cell surface are desialylated by neuraminidase. If the neuraminidase hydrolyzes only the natural form of sialosides but not the biotinylated sialosides, the one-step SEEL procedure can be done in 2 h with concurrent treatment of neuraminidase. Fibroblasts labeled with 2 and 3 were treated with Vibrio cholerae (VC) sialidase or Arthrobacter ureafaciens (AU) sialidase, and the results were compared with untreated cells by Western blotting. Only the C-9 compound was sensitive to treatment with VC sialidase whereas the C-5 analogue was resistant (Figure S2, panel A). In contrast, both modifications were resistant to AU sialidase treatment (panel B and data not shown). Treatment of the C-9 azide-modified sialyllactosamine (8S) with VC sialidase led to the removal of the modified sialic acid moiety within a period of 24 h whereas a similar treatment of C-9 biotin modified sialoside 9S resulted only in partial cleavage of the sialoside, and in this case an extended period of time was required for full removal. A sialoside modified at C-5 with biotin (10S) was resistant to VC sialidase treatment. None of the sialosides could be cleaved by AU sialidase. These findings confirm the cell-based results and also suggest that the AU sialidase can be used concurrently with SEEL labeling as a means to remove existing natural sialic acids without affecting the addition of the biotin-modified sialic acids.12</p><p>Next, attention was focused on the use of the new compounds to visualize trafficking of N-linked glycoconjugates in normal and chloroquine-treated cells. HeLa cells were enzymatically labeled with ST6Gal1 and CMP-sialic acid derivative 2, and the resulting biotin-modified N-glycoproteins were visualized by confocal microscopy following incubation with a fluorophore conjugated anti-biotin antibody (Figure 3a). As expected, robust staining at the cell surface was observed. Labeled cells were then incubated with or without chloroquine for 16 h to allow labeled glycoproteins to be internalized from the cell surface. Chloroquine is known to disrupt lysosomal pH and prevent eficient catabolism within this compartment. In the absence of chloroquine, a clear decrease in the intensity of labeling at the cell surface was detected, indicating that the SEEL-tagged glycoproteins can be internalized and/or degraded. In the chloroquine-treated cells, the labeled glycoconjugates instead accumulated within intracellular vesicles, consistent with late endosomes/lysosomes.</p><!><p>The high labeling efficiency achieved by the one-step SEEL procedure provides opportunities to enrich and identify cell surface glycoproteins and investigate how different glycoproteins respond to biological processes such as lysosomal disruption. For enrichment of tagged glycoproteins, immunoprecipitation was performed with an anti-biotin antibody, which was followed by SDS–PAGE and silver staining (Figure 3b). The results clearly show a loss of labeled N-linked glycoproteins following incubation in the absence of chloroquine (indicative of endocytosis and degradation of these glycoproteins) but a substantial reduction in the degradation of many labeled proteins in the presence of chloroquine. This observation was confirmed by Western blotting analysis of the same samples using an anti-biotin antibody (Figure 3b). Next, immunoprecipitated glycoproteins were subjected to tryptic digestion followed by shotgun proteomics (LC–MS/MS). The efficiency of labeling and enriching cell surface proteins is demonstrated by the identification of 255 membrane proteins out of the 282 proteins assigned (>90%) in the initial time zero sample (Table S2). Alterations in protein abundance were then examined in samples incubated with or without chloroquine for 16 h. A total of 173 proteins (161 of which are membrane proteins) were present in all three incubation conditions with less than a 1% false-discovery rate (Table S2).</p><p>Quantification via spectral counts revealed that 40 glycoproteins decreased in abundance by at least 5-fold (or became undetectable) after incubation of untreated cells for 16 h, but were detected in at least a 3-fold greater abundance when chloroquine was present during the 16 h incubation period (Figure 3c, Table S2). These glycoproteins, which include many receptors, such as the cation-independent mannose 6-phosphate receptor (IGF2R) and ephrin B receptor (EPH2A), likely represent those that internalize and turn over readily when lysosomal function is normal but will accumulate inside cells when the function of this organelle is compromised by chloroquine treatment. The data had been analyzed using stringent conditions for the quantification removing from consideration those glycoproteins that were detected in non-SEEL labeled control samples. Relaxation of this criterion, such as also considering proteins which were enriched at least 10-fold compared to control, captured additional proteins that exhibit the above-described internalization behavior such as CD44 and EGFR (Table S2). Interestingly, another set of glycoproteins including caveolin-1, mucin-1, and several cell surface channel proteins were not or only marginally affected in abundance after the chase period in the presence or absence of chloroquine (Figure 3c, Table S3).</p><p>To validate the proteomic findings, the steady-state level of one identified glycoprotein (IGF2R) was assessed by immunoprecipitation and Western blotting. IGF2R, also known as the cation-independent mannose 6-phosphate receptor, is primarily localized to post-Golgi compartments but also cycles to the plasma membrane where it can bind a variety of ligands.13 It is estimated that approximately 10–15% of the total IGF2R pool in cells localizes to the plasma membrane at steady state. Glycoproteins labeled with 2 were immunoprecipitated using an anti-biotin antibody and the resulting samples resolved by SDS–PAGE and Western blotting using an IGF2R protein-specific antibody. When the 6-fold higher protein load in the eluted fraction is taken into account, about 15% of the total IGF2R was labeled by SEEL, consistent with prior estimates of its cell surface distribution under steady state conditions. It can be seen in Figure 3d that incubation for 16 h resulted in a significant decrease in the abundance of the tagged glycoprotein whereas incubation in the presence of chloroquine showed less degradation of the labeled IGF2R. This closely mirrors the proteomic findings and supports the fidelity of the approach to monitor cell surface glycoprotein levels and dynamics, in particular for proteins that actively cycle between the plasma membrane and the endosomal system.</p><!><p>Existing strategies for the enrichment of cell surface glycoproteins utilize reagents such as sulfo-NHS-activated biotin to tag lysine residues on available proteins.14 These reagents suffer from lack of selectivity and often isolate cytosolic proteins. Using non-adherent HEK293F cells, a direct comparison of the SEEL and amino labeling method was performed. As shown in Figure 4a, labeling with the sulfo-NHS-biotin reagent resulted in the detection of numerous proteins across a broad molecular weight. On the other hand, a more restricted profile of proteins was visualized by one-step SEEL labeling with ST6Gal1. Analysis of the total proteins detected (vs total cell surface glycoproteins detected), however, revealed a clear enrichment in the detection of cell surface glycoproteins using SEEL. Nearly 95% of the total proteins detected following SEEL with ST6Gal1 (367) were bona fide cell surface glycoproteins whereas only 18% of the total proteins detected using the sulfo-NHS-biotin reagent were cell surface glycoproteins (Figure 4b, top panel). Analysis of total spectral counts recovered using the two methodologies further enforced the obvious advantage of using SEEL-based labeling to enrich and detect cell surface glycoproteins (Figure 4b, bottom panel).</p><!><p>The findings reported here demonstrate a remarkable labeling efficiency of cell surface glycoproteins by employing CMP-sialic acid modified by biotin and an appropriate sialyltransferase. The approach made it possible to track, capture, and identify subsets of cell surface glycoconjugates with unprecedented sensitivity in whole cells. The efficiency of the enzymatic transfer by ST6Gal1 is not substantially altered by the presence of the biotin (or azide) tag, and therefore we believe that the inferior labeling with the two-step SEEL method lies in the bioorthogonal reaction step. It is possible that the alkyne-containing compounds are prone to side reactions,8 but very little background is observed when cells are incubated with DIBO or S-DIBO alone. This observation would suggest instead that steric hindrance or electrostatic effects impact the efficiency of labeling at the cell surface. The use of the biotin conjugated CMP-sialic acid derivatives also decreases the total labeling time which makes it suitable for SEEL on cell types that may be sensitive to the conditions needed for this type of labeling. We did note that metabolic labeling most efficiently labeled a different subset of glycoproteins than SEEL (Figure 2). This may reflect the ability of metabolic labeling to label glycoproteins with high rates of turnover and biosynthesis. Likewise, it is anticipated that metabolic labeling can label intracellular glycoproteins (e.g., Golgi enzymes, etc.). The SEEL approach requires galactosyl acceptors at the cell surface, which in situ can be generated by performing the reactions in the presence of a bacterial neuraminidase that is capable of removing the natural sialic acids but not the biotin-bearing sialic acids. The ability to effectively label in the presence of the neuraminidase is advantageous since it may limit the effects that desialylation would have on the cell surface residence or endocytosis of glycoproteins.</p><p>Immunopurification of labeled glycoconjugates followed by tandem mass spectrometry demonstrated the power of the one-step SEEL approach to efficiently enrich and confidently identify cell surface glycoproteins, as approximately 300 known cell surface glycoproteins were detected. Almost twice as many proteins were identified compared to the two-step SEEL method (Figure 2b), and the new approach allows for many lower abundance glycoproteins to be uniquely identified. The 2.0-fold increase in spectral counts with one-step SEEL is far less dramatic than the increased labeling efficiency of glycoproteins detected by Western blotting. We believe that this difference arises from the fact that only one glycan of a glycoprotein that may have multiple glycans needs to be labeled for enrichment and subsequent proteomic analysis. In contrast, labeling multiple glycans of a single glycoprotein will greatly influence Western blot detection of a given glycoprotein as each biotin contributes independently to the overall signal.</p><p>A commonly performed procedure for analyzing cell surface proteomes is by biotinylation with reagents such as sulfo-NHS activated biotin followed by affinity capture and mass spectrometric analysis.14 This widely applied approach lacks selectivity and often leads to the isolation of cytosolic proteins likely due to cell permeability of the reagent. A number of alternative methods have been explored;15 however, all lack specificity required for comprehensive analysis of the surface membrane proteome.16 A more selective approach relies on mild oxidation of glycans with NaIO4 to install aldehyde functions, followed by tagging with a hydrazine containing bifunctional linker, proteolysis, capture of glycopeptides, enzymatic release of the N-linked glycans, and identification of the resulting peptides by LC–MS/MS.17 Although elegant, the approach requires a large number of steps and relies on rather inefficient hydrazone formation, and protein identification is based on the characterization of mainly one peptide.</p><p>The SEEL methodology described here is highly selective for cell surface proteins because the employed enzymes and reagents cannot cross the cell membrane. The high labeling efficiency and the selectivity for cell surface proteins achieved is best underscored by the impressive enrichment of cell surface glycoproteins vs total proteins detected (95% using SEEL; 18% using NHS-biotin). Moreover, the procedure is technically simple and allows glycoprotein trafficking and turnover to be easily investigated though of course it is limited to the glycoproteins that can be substrates for the enzyme. Furthermore, the one-step SEEL approach can readily be expanded to the labeling of specific classes of glycoproteins by employing alternative glycosyltransferases that have unique glycosyl acceptor specificities.18</p><p>The new strategy allowed the identification of glycoproteins whose steady state levels decrease following internalization as well as those that respond differently to incubation in the absence or presence of chloroquine. Many proteins such as IGF2R and EPHA2 were greatly reduced following the 16 h chase period for untreated cells but were more stable to degradation when chloroquine was present during the chase period. This profile indicates that these glycoproteins are subject to internalization and turnover or loss of label within the endolysosomal system. The ability of the new methodology to monitor glycoprotein dynamics in the context of lysosomal dysfunction will be particularly advantageous for the investigation of lysosomal storage disorders and the mechanisms that might explain the pathophysiology of these diseases. Abnormal recycling of glycoproteins is a known feature of these disorders that may lead to the intracellular accumulation of proteins that are required at the cell surface for normal health of cells such as neurons.9b,19 Methods that provide an accurate assessment of the cell surface glycoproteins have potential to uncover new targets for investigation and will no doubt find numerous applications in other areas of biology and biomedicine. For example, the global characterization of glycoproteins expressed on a cell surface and how this proteome responds to different stimuli can offer a better understanding of many disease processes, lead to new biomarkers for diagnosis and early detection of disease, and accelerate drug development.</p><!><p>Methods, associated references, and additional information and data are available in the Supporting Information.</p><!><p>Recombinant rat α-(2,6)-sialyltransferase (ST6Gal1) was prepared as reported.20 CMP-Neu5Ac9N3, 4, and Ac4ManNAz were synthesized as previously described.5g,11 Vibrio cholerae neuraminidase was purchased from Sigma-Aldrich (N6514). Arthrobacter ureafaciens neuraminidase was purchase from NEB (P0722). Alkaline phosphatase (FastAP) was purchased from Thermo Scientific (EF0651). HRP conjugated anti-biotin antibody (200-032-211), Alexa Fluor 488 conjugated anti-biotin antibody (200-542-211), and unconjugated anti-biotin antibody (200-002-211) were purchased from Jackson ImmunoResearch Laboratories. HRP conjugated β-actin antibody was from Abcam (ab20272). Anti-CI-MPR (IGF2R) polyclonal antibody was a kind gift of Dr. Peter Lobel (CABM-Rutgers). Protease inhibitor cocktail tablet (88666), EZ-Link Sulfo-NHS-LC-Biotin (21335), and mass spectrometry compatible silver staining kit (24600) were from Thermo Scientific. Protein G beads were from Sigma-Aldrich (Protein G sepharose, Fast Flow, P3296).</p><!><p>HeLa, human fibroblast (CRL-1509; Coriell Cell Repository, Camden, New Jersey), HepG2, or HEK293T cells were cultured in DMEM medium with high glucose (4.5 g/L) and l-glutamine, Jurkat cells were cultured in RPMI 1640 medium with l-glutamine, and all those media were supplemented with 10% fetal bovine serum (FBS, BenchMark) and penicillin (100 IU/mL)/ streptomycin (100 μg/mL, MediaTech). HEK293F cells were cultured in suspension with serum-free Freestyle 293 expression medium. All cell lines were cultured in a 5% CO2 atmosphere, 37 °C humid incubator.</p><!><p>Metabolic labeling of HeLa cells was done with 50–70% confluent cells in 12 well dishes by incubating with 30 μM Ac4ManNAz in DMEM supplemented with 10% FBS for 24 h in the cell culture incubator. Metabolic labeling of HepG2 and HEK293T cells was carried out by culturing 50–70% confluent cells in 10 cm culture dishes with 60 μM Ac4ManNAz for 24 h. Jurkat cells (~2.0 × 107 cells in suspension in a 10 cm dish) were also treated with 60 μM Ac4ManNAz for 24 h. Cells were washed with DPBS, further labeled with 4 (30 μM) in 2% FBS containing DPBS for 1 h at room temperature, and then lysed and analyzed by Western blot.</p><p>Two-step SEEL or one-step SEEL was done with confluent HeLa cells in 12 well dishes. Cells were pretreated with (+) or (−) Vibrio cholerae (VC) neuraminidase (50 mU/mL) for 2 h at 37 °C in serum free DMEM. After cells were washed with DPBS three times, cells were incubated in SEEL reaction mixture in 37 °C for 2 h. The SEEL reaction mixture (300 μL) was prepared using serum free DMEM with ST6Gal1 (42 μg/mL), CMP-sialic acid derivative (100 μM), 2 μL of BSA (2 mg/mL), alkaline phosphatase (2 μL), and 46 μL of 3 M sucrose. After cells were washed with DPBS, the cells with two-step SEEL were further labeled with 4 (30 μM) in 2% FBS containing DPBS for 1 h at room temperature and then lysed and analyzed by Western blot. Following one-step SEEL reaction, cells were directly lysed and analyzed by Western blot.</p><p>Two-step or one-step SEEL of Jurkat, HepG2, or HEK293T cells was carried out similarly except that they were labeled in suspension in Eppendorf tubes. In the case of HepG2 or HEK293T, cells confluent in 10 cm dishes were detached by pipetting after brief incubation (5–10 min) with DPBS and collected in Eppendorf tubes. For Jurkat cells, ~3.0 × 107 cells in Eppendorf tubes were prepared. Cells were pretreated with Arthrobacter ureafaciens (AU) neuraminidase (2 μL, 1:1 diluted in 50% glycerol) in 300 μL of serum free medium containing BSA (2 μL) for 90 min. Next, two-step SEEL, one-step SEEL, or 4 labeling was carried out similarly in 300 μL total volume in Eppendorf tubes.</p><!><p>For chase experiment, HeLa cells were labeled with the one-step SEEL method without sucrose since the sucrose-treated HeLa cells did not grow well over the chase period. The labeled cells were either (1) directly lysed for the following Western blot or immunoprecipitation or fixed and then stained with anti-biotin-488 for confocal imaging (0 h), or (2) further incubated in DMEM containing 10% FBS with or without chloroquine (50 μM) treatment (16 h). After the chase, the resulting cells were lysed or fixed depending on the desired analysis.</p><!><p>For staining after SEEL, HeLa cells were cultured on glass cover slides (gelatin coated) in 12 well dishes. For staining, one-step SEEL reaction (300 μL) was performed with ST6Gal1 and compound 2 (C-5 tetrazole) with cotreatment of Arthrobacter ureafaciens (AU) neuraminidase (2 μL/mL, 1:1 diluted in 50% glycerol) in the absence of sucrose. For 0 h analysis, cells were directly fixed with 3.7% formaldehyde for 15 min and then stained with anti-biotin-Alexa Fluor 488 (1:500) for 1 h. For chase, cells labeled by the SEEL reaction were further cultured with or without 50 μM chloroquine in DMEM containing 10% FBS for 16 h, followed by the fixing and staining with anti-biotin-Alexa Fluor 488. Stained cells were visualized using an Olympus FV1000 laser scanning confocal microscope.</p><!><p>We followed the manufacturer's protocol for the Sulfo-NHS-LC-Biotin reagent with modification in labeling temperature (4 °C instead of room temperature) in an attempt to reduce internalization of the compound. HEK293F cells (2.2 × 107 cells each for biotinylation or control) cultured in suspension were collected in Eppendorf tubes and pelleted. The cells were then washed three times with cold DPBS (1.0 mL, pH 8.0) and then resuspended in 500 μL of cold DPBS containing (+) or (−) 2 mM EZ-Link Sulfo-NHS-LC-Biotin. The resulting mixture was rotated at 4 °C for 30 min. Next, the cells were then washed three times with DPBS containing 100 mM glycine followed by three washes with DPBS at 4 °C. These biotinylated cells were lysed in the RIPA buffer for further analysis.</p><!><p>Labeled HeLa cells were lysed on plate by scraping in RIPA buffer (50 mM Tris-HCl buffer pH 8.0, 150 mM NaCl, 1.0% NP-40, 0.1% SDS, 0.5% sodium deoxycholate) supplemented with protease inhibitor cocktail on ice and collected in an Eppendorf tube. For the HEK293F cells labeled in suspension, cells were pelleted and lysed in an Eppendorf tube with the same lysis buffer. After vortexing for 20 s followed by incubation for 30 min on ice, the resulting tubes were spun down at 20000g for 10 min to clear nuclei. Protein concentration was determined by using BCA Protein Assay Kit (Thermo Science Pierce). Lysates were analyzed by SDS–PAGE and immunoblot using anti-biotin antibody conjugated with HRP typically at 1:50,000–1:100,000 dilution.</p><p>For immunoprecipitation, protein G beads (Sigma-Aldrich, St. Louis, MO) coated with anti-biotin antibody were prepared by incubating anti-biotin antibody (unconjugated) with protein G beads (protein G beads: antibody = 3:2 volume ratio) in the immunoprecipitation buffer (= RIPA buffer without protease inhibitor cocktail). We found that the anti-biotin antibody gave less background and higher specificity for enrichment compared to streptavidin-bound agarose. Cell lysates were precleared by incubating with protein G beads for 2 h at 4 °C. The precleared lysate was collected and then incubated with the antibody-coated protein G beads overnight at 4 °C. Next, the beads were washed 5 times with the RIPA buffer and then eluted with 2× sample loading buffer containing 10 mM dithiotheitol by boiling for 10 min. For proteomic analysis, one-step SEEL was performed in large scale without using sucrose for both HeLa and HEK293F cells, and the concentration of recombinant sialyltransferase and biotinylated CMP-sialic acid was the same as in the smaller scale reactions. HeLa cells were cultured confluent (~1.0 × 107 cells) in a 10 cm dish, and SEEL reaction was made on the same dish in 2 mL total reaction volume after 2 h of pretreatment of VC neuraminidase. In the case of HEK293F cells, cells cultured in suspension (1.7 × 107 cells) were collected in an Eppendorf tube and then SEEL reaction was performed with 530 μL total reaction volume in the presence of AU neuraminidase as concurrent treatment. 1.0 mg of each lysate was incubated with 50 μL of the protein G-beads (precoated with anti-biotin antibody) for complete pull-down of cell surface sialoglycoproteins. Eluted proteins were resolved by SDS–PAGE (8%), and the resulting gel was silver stained for in-gel trypsin digestion followed by proteomic MS analysis.</p><!><p>Each lane of the silver-stained SDS–PAGE gel was cut into 4 parts above 50 kDa and subsequently processed for in-gel digestion. Briefly, destained gel bands were denatured by incubating with 10 mM dithiothreitol at 56 °C for 1 h and alkylated by 55 mM iodoacetamide for 45 min in the dark prior to digestion with trypsin overnight. The resulting peptides were extracted, dried, and reconstituted in 0.1% formic acid. The peptides were separated on a 75 μm (i.d.) × 15 cm C18 capillary column (packed in house, YMC GEL ODS-AQ120 ÅS-5, Waters) and eluted into the nano-electrospray ion source of an Orbitrap Fusion Tribrid mass spectrometer (Thermo Fisher Scientific) with a 180 min linear gradient consisting of 0.5–100% solvent B over 150 min at a flow rate of 200 nL/min. The spray voltage was set to 2.2 kV, and the temperature of the heated capillary was set to 280 °C. Full MS scans were acquired from m/z 300 to 2000 at 120k resolution, and MS2 scans following collision-induced fragmentation were collected in the ion trap for the most intense ions in the Top-Speed mode within a 3 s cycle using Fusion instrument software (v1.1, Thermo Fisher Scientific). The raw spectra were searched against the human protein database (UniProt, Oct. 2014) using SEQUEST (Proteome Discoverer 1.4, Thermo Fisher Scientific) with full MS peptide tolerance of 20 ppm and MS2 peptide fragment tolerance of 0.5 Da, and filtered using ProteoIQ (v2.7, Premier Biosoft) at the protein level to generate a 1% false-discovery rate for protein assignments. Proteins detected at 1% false-discovery rate in the negative controls (no recombinant enzyme added, data not shown) were excluded from the final lists of proteins identified in respective conditions. UniProt was used to define cellular localization. Quantification was performed by normalizing the spectral counts generated by ProteoIQ (v2.7, Premier Biosoft).</p>

PubMed Author Manuscript

Non-metal to metal transition of magnesia supported Au clusters affects the ultrafast dissociation dynamics of adsorbed CH3Br molecules

The detection of intermediate species and the correlation of their ultrafast dynamics with the morphology and electronic structure of a surface is crucial to fully understand and control heterogeneous photoinduced and photocatalytic reactions. In this work, the ultrafast photodissociation dynamics of CH3Br molecules adsorbed on variable size Au clusters on MgO/Mo(100) is investigated by monitoring the CH3 + transient evolution using a pump-probe technique in conjunction with surface mass spectrometry. Furthermore, extreme-ultraviolet photoemission spectroscopy in combination with theoretical calculations are employed to study the electronic structure of the Au cluster on MgO/Mo(100). Changes in the ultrafast dynamics of CH3 + fragment are correlated with the electronic structure of Au as it evolves from monomers to small nonmetallic clusters to larger nanoparticles with a metallic character. This work provides a new avenue to a detailed understanding of how surface photoinduced chemical reactions are influenced by the composition and electronic structure of the surface.

non-metal_to_metal_transition_of_magnesia_supported_au_clusters_affects_the_ultrafast_dissociation_d

<p>Gold clusters and nanoparticles have sparked intense interest over the years because of their size dependent electronic, optical, and chemical properties, 1-2(and ref. therein) with applications in sensors, 3 probes, 4 biological systems and medicine, [5][6][7] as well as energy conversion, which includes solar cells, [8][9][10] catalysis, 11-13 (and ref. therein) and photocatalysis. 14 Not surprisingly, theoretical and experimental studies of small clusters have been performed to investigate their electronic [15][16][17][18][19] and geometric structure, 17,[20][21][22][23][24][25][26][27][28][29][30][31][32] as well as their catalytic properties, 11 with a vast amount of investigations focusing on the CO oxidation reaction. 19,[33][34][35][36][37][38] Large gold nanoparticles were often studied and employed in various applications because of their optical properties as facilitated by the localized surface plasmon resonance effect. 3, 7-8, 14, 39 Despite the large number of experimental and theoretical investigations that explored properties of free and supported gold nanoparticles, which have led to various scientific breakthroughs and technological applications, 3,[39][40] there is a lack of knowledge on how small gold clusters that do not present localized surface plasmon resonances could influence photoinduced chemical reactions at surfaces. With the above in mind, the present investigation explores photoinduced chemical reactions on oxide surfaces decorated with gold clusters, in the size range close to the nonmetal-to-metal transition.</p><p>To investigate the molecular photodissociation dynamics of the CH3Br molecules on magnesia supported gold clusters, a technique is employed that combines femtosecond (fs) laser pump-probe spectroscopy with multiphoton ionization a n d time-of-flight mass spectrometry. 41 The technique relies on the dissociation of the adsorbed CH3Br molecules via two-photon pump excitation of the A-band followed by three-photon probe resonance multiphoton ionization at the surface to detect the CH3 fragments by mass spectrometry of the desorbed ions. 42 The center wavelength of the pump and probe laser beams are 266 nm and 333 nm, respectively. Details about the experimental setups and the pump-probe technique used in this investigation are presented in the Supplementary Information (see section 1 and 2 in the Supplementary Information). For reference, in each case a CH 3 + transient was recorded from the bare magnesia surface just before gold deposition (blue circles in Figure 1). The methyl bromide coverage, which is 0.25 ML, and the laser parameters are kept constant for all recorded transient signals.</p><p>On the bare magnesia thin film, the transient methyl signal starts with an initial delay of 150 ±50 fs and exhibits a single exponential rise with a time constant of 320 ±60 fs, which are in a perfect agreement with the previous CH3Br photodissociation measurements on MgO thin films. [43][44][45][46] The initial coherent delay of 150 ±50 fs reflects the minimal time needed for the liberation of the methyl fragments from the molecular force field and from the force field of the magnesia surface. This time is longer than the A-band dissociation time of CH3Br in the gas phase, i.e. 116 ±25 fs, 47 due to the molecular adsorption geometry in which the C-Br axis is almost parallel to the MgO substrate, which facilitates collision of the CH3 fragment with the adjacent molecule just before being released into gas phase. (45 and refs. therein) The subsequent growth of the methyl signal with a time constant of 320 ±60 fs, which is obtained from data fitting with a 'delayed exponential rise' function convoluted with the pump-probe cross correlation function 48 (red curve in Figure 1a) is attributed to the average lifetime of all trajectories leading to the release of the methyl fragment.</p><p>The transient methyl signal obtained from the photodissociation of CH 3 Br (dosed at 100 K) on magnesia covered by 0.065 ML Au, deposited previously at 100 K, displayed in Figure 1a (black squares) is essentially similar to the one obtained from the bare MgO surface (Figure 1a blue circles). After deposition of 0.1 ML Au on the magnesia surface, the transient methyl signal, which is displayed in Figure 1b, also exhibits an exponential rise, but in addition it displays a new peaked structure with a maximum at 240 fs. Fitting the experimental data with a model that consists of an exponential 'rise and decay' in conjunction with a 'delayed exponential rise', convoluted with the laser cross correlation 48 (red curve in Figure 1b) gives 190 ±30 fs for the rise and decay of the peaked structure, attributed below to gold particles on the surface. For the subsequent exponential rise at longer delay times, the time constant matches with the one obtained from the bare MgO surface.</p><p>Figure 1c shows the transient methyl signal obtained from the photodissociation of CH 3 Br on magnesia covered by 0.13 ML Au. In this transient signal, the peaked structure at early delay times with a maximum intensity at 270 fs is even more apparent and is followed by an exponential rise. The best fit with the same model as in Figure 1b to the experimental data (Figure 1c -black squares) gives similar time constants of 200 ±30 fs for the rise and the decay components of the observed peaked structure in time. For the subsequent exponential rise at longer delay times, the time constant matches again with the one obtained from bare MgO surface. The CH 3 + transient peaked structure observed in Figure 1 b,c appears only on magnesia substrates prepared with gold particles (coverages between 0.1 and less than 0.15 ML). The photodissociation of CH 3 Br on MgO/Mo(100) covered by 0.15 ML Au leads to a dramatic increase of the methyl cation signal intensity (factor of six) compared to the signal obtained from the bare magnesia surface (cf.</p><p>Figure 1d). However, no time dependence of the CH 3 + signal is observed any more. These new features are all assigned to the dissociation of methyl bromide molecules adsorbed on the gold clusters.</p><p>For the interpretation of the transient data in Figure 1(b-d), the size dependent structure of the gold clusters on MgO is considered, which was previously investigated via scanning tunneling microscopy (STM) and electron paramagnetic resonance (EPR) spectroscopy by Freund and coworkers. [49][50] According to the STM and EPR investigations, at low gold coverages up to 0.06 ML on 8 ML MgO/Mo(100), at 100 K, mostly gold atoms are observed on the surface.</p><p>Consequently, it can be assumed that in the present investigation, isolated gold atoms and very small clusters are present on the MgO surfaces for an Au coverage of 0.065 ML (cf. Figure 1a), and these do not lead to a CH 3 + transient peaked structure. This conclusion is further supported by an experiment in which Au atoms produced by a mass-selected cluster source, coupled to the investigation chamber, were soft landed on 10 ML MgO/Mo(100) surface (Au coverage of about 0.03 ML). In this experiment, no CH 3 + transient peaked structure is observed (not shown here). STM investigations of Freund and coworkers demonstrate that large three dimensional Au particles are formed via sintering, if MgO samples decorated with gold atoms (Au coverages below 0.06 ML) are annealed to 300 K. 49 Therefore, in order to confirm the hypothesis that large particles are responsible for the change of the methyl signal at coverages above 0.15 ML Au (Figure 1d), the magnesia sample with 0.13 ML Au (Figure 1c) was briefly heated to 800 K to induce the formation of larger Au particles via sintering. Subsequently, CH3Br molecules are dosed on the surface at 100 K, and a CH 3 + transient signal is recorded. The result is displayed in the inset of Figure 1c. Indeed, the peak structure and exponential rise measured before the annealing (Figure 1c) are not observed anymore. Instead, an intense methyl signal is obtained that does not exhibit a time dependence, similar to the case of 0.15 ML coverage in Figure 1d. This result supports the hypothesis that the presence of a considerable fraction of large Au nanoparticles are responsible for the different signal in Figure 1d. surface. 42 As mentioned above, two photons at 266 nm are necessary to induce the CH3Br photodissociation on a 10 ML MgO/Mo(100) surface, while three photons at 333.3 nm are necessary to detect the CH3 fragments through (2+1) REMPI. 43 On an extended Au surface, power dependence measurements reveal that a single pump photon at 266 nm is absorbed, while the CH3 ionization-detection is accomplished by two probe photons at 333.3 nm. 43 The reduced number of photons necessary for the photoexcitation of CH3Br and detection of CH 3 + from an Au surface is attributed to a red-shift of the CH3Br A-band by about 1.5 eV, 42 as also observed for CH3Br adsorbed on Ag(111). 52 The CH 3 + transient signal, i.e. peaked structure, recorded on the Au surface is attributed to the following pump-probe schema: (i) a single pump photon at 266 nm excites the CH3Br molecule into the dissociative A-band; (ii) before the molecule dissociates, two-probe photons at 333 nm further excite the molecule from the A-band into a cationic dissociative state which decomposes and leads to the CH 3 + transient signal. 42 It is assumed that the CH substrate. [53][54] The photoemission onset of Mo(100) coincides with the photoemission onset of 0.15 ML-1.00 ML Au/MgO/Mo(100), which is located at 0.08 eV above the Fermi level.</p><p>When 10 ML MgO are grown on Mo(100) the photoemission onset shifts to -3.4 eV below the EF (larger negative numbers refer to greater binding energies), as determined by the intersection point where the extrapolation of the linear portion of the MgO rising edge intersects the straight line describing the background (see red dashed line in Figure 2a). [55][56] The occupied electronic states below -3.4 eV are attributed to the photoemission from the O-2p states that make up the valence band of MgO. No photoemission peaks are observed at binding energies between 0 eV and -3.4 eV, which corresponds to the portion of the MgO band gap that can be accessed via photoemission spectroscopy, indicating that the MgO film is fully oxidized.</p><p>When gradually increasing amounts of Au are evaporated on the 10 ML MgO film, the photoemission progressively shifts to lower binding energy. The shift of the photoemission to lower binding energy as the amount of gold is increased (cf Figure 2(a)), is attributed to a lowering of the work function of the combined system, i.e. Au/MgO/Mo(100). This shift is expected to be due to the electrostatic interaction between the photoemitted electrons and the negative charges that accumulate at the cluster-oxide interface, which increases with the number of Au atoms in direct contact with the oxide surface. 57 To explore the change in the electronic structure of the Au particles as their size is increased, and to eliminate the influence of the substrate, in Figure 2 The peak at -3.30 eV is attributed to the photoemission from large Au particles, since this peak is visible only at high Au coverages. The distinct features at -2.2 eV, -1.35 eV, and -0.75 eV are attributed to photoemission from Au atoms and clusters composed of very few atoms on MgO/Mo(100). Since the photoemission feature at -2.2 eV displays the highest intensity at Au coverages of 0.01-0.06 ML, the result suggests that this photoemission feature originates from Au atoms on MgO/Mo(100). This is in agreement with the STM and EPR investigations, [49][50] which reveal that mostly gold atoms are present on the MgO/Mo(100) surface decorated with Au coverages lower than 0.06 ML. Moreover, theoretical investigations of Pacchhioni and coworkers</p><p>show that Au atoms on 3 ML of MgO(100) have the Au-5d density of states located between -1.5 eV and -2.5 eV, below the Fermi level. 58 In a different theoretical investigation, by Häkkinen and coworkers, it was shown that the Au atoms at hollow sites on 3 ML MgO/Mo(100) have the Au 5d electrons centered at -2 eV below the Fermi level. 59 Therefore, the feature at -2.2 eV in Figure 4(b) is attributed to the photoemission from Au-5d states belonging to Au atoms on MgO/Mo(100). The peak at -1.35 eV might be due to the Au-5d electrons of Au clusters composed of a few atoms, since this peak is visible between 0.06-0.12 ML Au and is barely visible at coverages below 0.06 ML Au. The low intensity feature at -0.75 eV is attributed to photoemission from the Au-6s states of Au atoms. This photoemission feature matches the predicted energetic position and relatively low intensity of the Au 6s electrons of Au atoms. [58][59] Furthermore, the photoemission feature initially centered at -0.75 eV systematically expands toward lower binding energies, as the amount of gold is increased above 0.1 ML, lowering the onset energy of the photoemission spectra.</p><p>For Au coverages up to 0.12 ML the photoemission onsets are located below the Fermi level, indicating that the Au clusters have a non-metallic character, while for a Au coverage of 0.15 ML, the photoemission onset shifts to +0.08 eV above the Fermi level, similar to the photoemission from metals. The photoemission onset does not change for Au coverages larger than 0.15 ML and coincides with the photoemission onset of the Mo(100) substrate. This clearly indicates that Au particles formed at Au coverages of 0.15 ML or greater have a metallic character.</p><p>Consequently, the transition of Au clusters from nonmetal to metal on a 10 ML MgO film grown on Mo(100) occurs at Au coverages of about 0.15 ML.</p><p>The static fs-XUV photoemission data obtained from various amounts of Au on 10 ML MgO/Mo(100) (cf. Figure 2) strongly suggest that the dramatic increase in the CH3 + signal observed when 0.15 ML Au is evaporated on the surface is due to the formation of Au particles that have a metallic character. The photodissociation mechanism of the CH3Br molecule on metallic Au particles is expected to be broadly similar to the photodissociation mechanism on an extended Au surface, 42 due to the red-shift of the CH3Br A-band by the metal structure, as discussed above. Certainly, the morphology of various Au particles that have a metallic character might also play a role in the photodissociation of CH3Br, however, the changeover of Au from non-metal to metal plays the essential role. Information (see Figures S1 and S3). The theoretical investigation reveals that no occupied electronic states are present in the vicinity of the Fermi level for Au13, Au19 and Au47. However, for Au clusters containing 55 or more atoms, occupied electronic states are detected at the Fermi level (Figure S3), indicating that these clusters have a metallic character. Therefore, from the photoemission experiments (cf. Figure 2) and theoretical investigations (cf. Figure S3) we can conclude that the evaporation of 0.</p>

ChemRxiv

Quantitative Analysis of Adulterations in Oat Flour by FT-NIR Spectroscopy, Incomplete Unbalanced Randomized Block Design, and Partial Least Squares

This paper developed a rapid and nondestructive method for quantitative analysis of a cheaper adulterant (wheat flour) in oat flour by NIR spectroscopy and chemometrics. Reflectance FT-NIR spectra in the range of 4000 to 12000 cm−1 of 300 oat flour objects adulterated with wheat flour were measured. The doping levels of wheat flour ranged from 5% to 50% (w/w). To ensure the generalization performance of the method, both the oat and the wheat flour samples were collected from different producing areas and an incomplete unbalanced randomized block (IURB) design was performed to include the significant variations that may be encountered in future samples. Partial least squares regression (PLSR) was used to develop calibration models for predicting the levels of wheat flour. Different preprocessing methods including smoothing, taking second-order derivative (D2), and standard normal variate (SNV) transformation were investigated to improve the model accuracy of PLS. The root mean squared error of Monte Carlo cross-validation (RMSEMCCV) and root mean squared error of prediction (RMSEP) were 1.921 and 1.975 (%, w/w) by D2-PLS, respectively. The results indicate that NIR and chemometrics can provide a rapid method for quantitative analysis of wheat flour in oat flour.

quantitative_analysis_of_adulterations_in_oat_flour_by_ft-nir_spectroscopy,_incomplete_unbalanced_ra

1. Introduction<!>2.1. Collection and Preparation of Samples<!>2.2. Data Preprocessing and Multivariate Calibration<!>3. Results and Discussion<!>4. Conclusions

<p>Food adulteration and fraud have been a common problem in food production since ancient times. Food adulteration is economically motivated and is performed by the addition, substitution, or removal of food ingredients, for example, replacing or diluting high-cost ingredients with cheaper ones [1, 2]. It is an issue that concerns not only consumers, but also food producers, sellers, regulatory agencies, and even the entire food industry chain. For consumers, food adulterations have caused growing concern about health risks, as well as the food quality and nutrition value. A notorious and common phenomenon is the adulterations of raw food materials, which can not only influence the quality of raw materials but also cause potential crisis in further processed foods [3, 4].</p><p>Oat is widely utilized for human consumption and food industrial uses. Due to its high nutritional value and characteristic flavor, oat flour plays an important role in the breakfast cereals group and other processed foods as an alternative or supplement to the ordinary wheat flour [5]. As a nonstaple cereal, the yield of oat is much less than wheat, so oat flour is more expensive than wheat flour. For producers and sellers, it is economically profitable to add wheat flour to oat flour. Because the appearances and physical and chemical properties of wheat and oat flours are very similar, rapid and effective methods are required to analyze the adulterations.</p><p>For food analysis and quality control, NIR spectroscopy has demonstrated some advantages, including less sample treatment, reduced analysis time and cost, and the feasibility for nondestructive analysis and online analysis. NIR spectroscopy has been widely used for analysis of grains and cereals. Hurburgh et al. [6] combined NIR spectroscopy and principal component analysis to discriminate transgenic grains and nontransgenic grains. Munck et al. [7] applied the NIRS technology to distinguish barley flour with different levels of lysine amino acids. NIR was also successfully used to distinguish corn samples of different genotypes [8, 9]. For quantitative analysis, NIR technology has provided a rapid tool for analysis of different constituents or quality parameters in grain products, including corn dry-milling quality [10], protein content in wheat kernels [11], the ratio of starch amylose content to total grain in corn [12], undried rough rice constituent content [13], kernel rots and mycotoxins in maize [14], and protein, moisture, dry mass, hardness, and other residues of wheat [15] and so on [16].</p><p>Considering the large number of samples in market shelf and small private retailers, NIR is a convenient and economic technique for analysis of adulterations in oat flour. This paper was aimed at developing a rapid method for analysis of potential wheat flour added to oat flour using NIR spectrometry and chemometrics. Considering the composition variations of oats from different producing areas, an incomplete unbalanced randomized block design [17] was performed to ensure the generalization performance of multivariate calibration models.</p><!><p>Pure oat flour and wheat flour objects were made from intact grains. A set of oat and wheat kernels harvested in 2013 were collected from domestic markets. Oat kernels were produced in Hebei (15), Henan (17), Gansu (17), Shanxi (18), and Qinghai (13). Wheat kernels were produced in Henan (11), Hebei (10), Jiangsu (8), Anhui (10), Shandong (11), Shanxi (7), and Heilongjiang (8). The kernels were dried in the sun and milled by a crusher. All the particles were filtered through a 200-mesh sieve.</p><p>Adulterated oat flour samples were made by mixing the oat flour with different levels of the wheat flour. The doping levels were 5%, 10%, 15%, 20%, 25%, 30%, 35%, 40%, 45%, and 50% (w/w). In order to obtain a representative nut not too large sample set, an incomplete and unbalanced randomized design [17] was performed. Because this paper was focused on quantitative analysis of wheat flour in oat flour, the producing areas of oat and wheat were considered to be two blocking factors. In this way, 120 adulterated objects were prepared for developing calibration models with 10 doping levels each having 12 objects. Another 100 adulterated objects (10 doping levels each having 10 objects) were prepared for model validation by mixing oat and wheat flour objects that were different from those used for preparation of training samples.</p><p>The NIR diffuse reflectance spectra of impacted powders were measured on a Bruker-TENSOR37 FTIR spectrometer (Bruker Optics, Ettlingen, Germany). The working range of spectrometer was 4000–12000 cm−1. The spectra were measured using a PbS detector with an internal gold background as the reference. The instrument resolution was 4 cm−1 and the scanning interval was 1.929 cm−1, so each spectrum contained 4148 wavelengths. Each spectrum was the average of 64 scans. For each object, three spectra were measured by stirring the powder and the average spectrum was taken.</p><!><p>All the data preprocessing and chemometrics models were performed using MATLAB 7.0.1 (Mathworks, Sherborn, MA). Smoothing was used to remove random noise in the data and improve the signal-to-noise ratio (SNR). In this work the S-G polynomial fitting algorithm [18] was used for smoothing for its simplicity and effectiveness. Taking second-order derivative (D2) of spectra was performed to enhance spectral resolution and remove linear baseline shifts. The D2 spectra were also computed by S-G polynomial fitting algorithm because this method can avoid degradation of SNR compared with direct differencing. Standard normal variate (SNV) [19] transformation was performed to reduce the spectral variations caused by scattering and uneven sizes of particle.</p><p>Partial least squares (PLS) models were developed using the raw and preprocessed spectra. An important problem when performing PLS is the overfitting of models. In this paper, Monte Carlo cross-validation (MCCV) [20] was used to select the number of PLS components. MCCV can avoid the risk of overfitting by multiple random splitting of the training objects and having a higher percent of leave-out objects for prediction.</p><!><p>The raw NIR spectra of 120 adulterated oat flour objects are shown in Figure 1. Some of the absorbance peaks can be assigned as follows [21]: (1) the peak at 4318 cm−1 caused by the combination absorbance of –CH2 deformation and various C–H stretching; (2) the wide peak at 4748 cm−1, overlapping of combination of C=O stretching and peptide group deformation and combination of N–H stretching and peptide group deformation; (3) the peak at 5167 cm−1, combination of O–H stretching and O–H deformation; (4) 5629 cm−1, the first overtone of symmetrical and asymmetrical C–H stretching in –CH2; (5) the wide peak at 6802 cm−1, overlapping of the first overtone of O–H stretching (~6900 cm−1) and the first overtone of N–H stretching (~6500 cm−1); (6) 8329 cm−1, the second overtones of C–H stretching in various groups; and (7) 9970 cm−1, the second overtones of N–H stretching or the third overtones of C–H stretching. The spectral interval 9000–12,000 cm−1 has no significant peaks, so this interval was not used for developing calibration models.</p><p>Smoothed, D2, and SNV spectra were shown in Figure 2. Seen from Figure 2, the D2 spectra can remove most of the backgrounds and the peak resolution was largely improved by taking D2 spectra. D2 spectra also obtained much detailed and high-frequency information. SNV transformation can remove most unwanted variations. Multivariate calibration models were developed with PLS to predict the levels of wheat flour. The number of PLS components was estimated using F-test of MCCV. In this work, random splitting of the training set was performed for 100 times and each time 70% of the training objects were used for developing a PLS model and 30% for prediction. The pooled predicted residual sum of squares (PRESS) was computed using different numbers of PLS components. Finally, F-test was performed to select the fewest PLS components with a PRESS value not significantly higher than the minimum PRESS value. As recommended by the original literature, the significance level of the F-test was set to be 0.25 [22, 23].</p><p>Based on differently preprocessed spectra (9000–4000 cm−1) the calibration and prediction results of PLS were demonstrated in Table 1. Seen from Table 1, the model complexity of D2-PLS and SNV-PLS was reduced by one component compared with the PLS model using raw data. Moreover, seen from the model complexity of PLS models (3 or 4 components), the mixtures of wheat flour and oat flour investigated in this work were not a simple two-component system and spectral variations caused by different geographical origins had made it necessary to perform multivariate calibration. The root mean squared error of MCCV (RMSEMCCV) value was slightly reduced by smoothing (2.274), taking D2 (1.921), and SNV transformation (1.981) compared with PLS with raw data (2.388). For prediction, the lowest root mean squared error of prediction (RMSEP) of 1.975 was obtained by D2-PLS model. The training and prediction results by D2-PLS are demonstrated in Figure 3. For all of the PLS models, the differences between RMSEP and RMSEMCCV values were insignificant, indicating that both the training and test sets obtained by IURB design were representative to include composition variations caused by different origins of oat and wheat flour.</p><!><p>Multivariate calibration models were developed by PLS for analysis of wheat flour in oat flour from different geographical origins. The results demonstrated that a three- or four-component PLS model can accurately predict the levels of wheat flour in oat flour. Moreover, IURB design was shown to be useful to obtain representative training and test sets to include the composition variations caused by different producing areas. The developed PLS models will have a good generalization performance and are useful for quantitative analysis of oat flour in domestic market.</p>

PubMed Open Access

Excited-State Dynamics of a Substituted Fluorene Derivative. The Central Role of Hydrogen Bonding Interactions with the Solvent

Substituted fluorene structures have demonstrated unusual photochemical properties. Previous reports on the substituted fluorene Schiff base FR0-SB demonstrated super photobase behavior with a \xce\x94pKb of ~14 upon photoexcitation. In an effort to understand the basis for this unusual behavior, we have examined the electronic structure and relaxation dynamics of the structural precursor of FR0-SB, the aldehyde FR0, in protic and aprotic solvents using time-resolved fluorescence spectroscopy and quantum chemical calculations. The calculations show three excited singlet states in relatively close energetic proximity. The spectroscopic data are consistent with relaxation dynamics from these electronic states that depend on the presence and concentration of solvent hydroxyl functionality. These results underscore the central role of solvent hydrogen bonding to the FR0 aldehyde oxygen in mediating the relaxation dynamics within this molecule.

excited-state_dynamics_of_a_substituted_fluorene_derivative._the_central_role_of_hydrogen_bonding_in

INTRODUCTION<!>Materials.<!>Synthesis of FR0.<!>Steady-State Emission Spectroscopy.<!>Time-Resolved Fluorescence Spectroscopy.<!>COMPUTATIONAL DETAILS<!>RESULTS AND DISCUSSION<!>CONCLUSIONS

<p>Chemical reactions are typically performed in the liquid phase or at the interface between phases with reactants in their ground states. Under such conditions there is little opportunity for temporal or spatial control of the reaction, although photochemical activation of one reactant is often best suited to achieve this. Nonetheless, the use of the latter strategy has historically been limited. Therefore, by gaining broader utility in photochemical reactions one affords new opportunities for temporal and spatial control of reaction chemistry, properties that are central to the emerging fields of precision chemistry and high-speed chemical sensing.</p><p>Most chemical reactions are either acid–base (proton transfer) or redox (electron transfer) reactions. The ability to alter the acidity or basicity of certain compounds through photoexcitation has been explored extensively for photoacids, which are photoactivated proton donors,1,2 but the development of their photobase counterparts, which are able to abstract available protons from their local environment upon photoexcitation, has been much more limited.3,4 Recently, we have focused on the design and characterization of the super photobase FR0-SB5–8 that is capable of exhibiting a remarkable increase in Kb of 14 orders of magnitude (or ΔpKb ~ 14) upon photoexcitation. The behavior of FR0-SB (Figure 1a) in protic and aprotic media has been reported recently,5–7 providing fundamental insights into the proton abstraction reaction that occurs in protic solvents. Among the issues that require further investigation is the chemical structural basis for such a large change in the basicity of the imine nitrogen upon photoexcitation. In an effort to understand more completely the structural properties that lead to super photobase behavior, we have examined the excited-state properties of the aldehyde precursor of FR0-SB, namely, FR0 (Figure 1b).</p><p>A comprehensive understanding of the reactivity and excited-state population relaxation dynamics of such precursor molecules is a key step toward the rational design of super photobases with controlled properties for specific applications. In this work, we report on the reactivity and excited-state relaxation dynamics of the FR0 precursor molecule in protic and aprotic solvent media using steady-state and time-resolved spectroscopy in conjunction with quantum chemical computational results.</p><p>The spectroscopic properties of the super photobase FR0-SB make the study of its photoreactivity relatively straightforward. Emission bands for unprotonated and protonated forms of the photoexcited Schiff base FR0-SB* are well resolved, and the time evolution of those band intensities provides direct insights into solvent proton abstraction by FR0-SB*. The data reported here for FR0 also exhibit spectral dynamics, although the resolution of individual bands is not as facile. The emission spectrum of FR0 exhibits time-dependent changes in the band shape and position in protic solvents, suggesting a significant role of the FR0 chromophore structure in the unusual photobase behavior of the corresponding Schiff base. We report on the presence of the overlapped electronic singlet state manifolds within the FR0 emission band and population exchange between those singlet states in protic solvents. The excited-state relaxation dynamics of FR0 depend on the identity of the (protic) solvent. Spectral reconstruction from the time-resolved emission data reveals the details of band evolution and demonstrates that the relaxation dynamics of the FR0 precursor molecule are mediated by hydrogen-bonding interactions between FR0 and the surrounding solvent medium.</p><!><p>n-Propanol (99.7%, anhydrous), n-pentanol (≥99%, ACS grade), and dimethyl sulfoxide (DMSO, ≥99.9%, anhydrous) were all purchased from Sigma-Aldrich and used as received. n-Pentanol, which was not available in an anhydrous form, was stored over molecular sieves (type 3A).</p><!><p>The synthesis of FR0 followed the procedure reported previously8,9 and is described briefly here. The synthesis commenced with fluorene, which was converted to 2-bromofluorene by treating with N-bromosuccinimide in propylene carbonate at room temperature. Dimethylation of the C9 position was accomplished with the treatment of the brominated fluorene with excess iodomethane in the presence of NaOH. Fuming nitric acid was used for the nitration of the nonbrominated aryl ring to yield 2-bromo-9,9-dimethyl-7-nitro-9H-fluorene. Subsequent reduction of the nitro group with iron powder suspended in aqueous ammonium chloride solution provided the corresponding amine. The resulting product was heated with ethyl iodide and potassium carbonate generating the diethyl amino substituent. Metal/halogen exchange promoted by the addition of n-BuLi, followed by the addition of dimethylformamide led to the formation of FR0. Purification of the crude on a silica-gel column (hexane/ethyl acetate eluant) provided the title compound as a yellow solid. 1H NMR (500 MHz, CDCl3): δ (ppm) = 9.98 (d, J = 0.87.90−7.85 (m, 1H)), 7.78 (dt, J1 = 7.8 Hz, J2 = 1.1 Hz, 1H), 7.68−7.59 (m, 2H), 6.73−6.66 (m, 2H), 3.46 (q, J = 7.1 Hz, 4H), 1.50 (s, 6H), 1.30−1.20 (m, 7H). 13C NMR (125 MHz, CDCl3): δ (ppm) = 192.05, 157.33, 153.19, 148.99, 146.90, 133.41, 131.27, 125.20, 122.47, 122.35, 118.12, 110.87, 105.06, 46.62, 44.71, 27.24, 12.59.</p><p>The compound was stored in ethanol (1 mM). To prepare samples of FR0 in each solvent, aliquots of FR0 were brought to dryness using nitrogen and reconstituted with the solvent system to be studied at a concentration appropriate for the measurement, typically ca. 5 × 10−6 M.</p><!><p>Emission spectra for FR0 in each of the solvents were collected using a Hitachi F-4500 fluorescence spectrometer and Spex Fluorolog 3 emission spectrometer. Quartz cuvettes (1 cm) were used for all the measurements and spectra acquired using an excitation wavelength of 440 nm, with an excitation and emission spectral resolution of 1 nm.</p><!><p>Time-resolved fluorescence measurements were collected using a time-correlated single photon counting (TCSPC) instrument that has been described in detail elsewhere,10,11 and we provide only a brief summary here. The light source is a continuous-wave passively mode-locked Nd:YVO4 laser (Spectra Physics Vanguard) that produces 13 ps pulses at 1064 nm (80 MHz repetition rate). The output of this laser was at the second harmonic (532 nm) and the third harmonic (355 nm), with2.5 W of average power at both wavelengths, with the same pulse duration. For the experiments reported here, the cavity-dumped dye laser (Coherent 702–2) was excited by the third harmonic output of the pump laser. The dye used was Stilbene 420 (Exciton) and the dye laser output at 440 nm was 5 ps full-width at half-maximum (fwhm) pulses at a repetition rate of 4 MHz. The pulses from the dye laser were divided with approximately half going to a reference channel diode (Becker & Hickl PHD-400-N) and the remaining light going to the sample. The vertically polarized excitation light was focused on the sample cuvette and a reflecting microscope objective (40×, Ealing) was used to collect emission. The collected emission was passed through a polarization-selective beam splitter and sent to two identical detection channels. Each detection channel had a subtractive double monochromator (Spectral Products CM-112) and a microchannel plate photomultiplier tube (PMT) detector (Hamamatsu R3809U-50). Signals from each detector were sent to TCSPC detection electronics (Becker & Hickl SPC-132). A typical instrument response function for this system is ca. 35 ps fwhm. The TCSPC acquisition electronics, PMT bias, and monochromator wavelengths were controlled using a computer program written in-house using LabVIEW (National Instruments) software. For these experiments, time-resolved fluorescence data were acquired from 470 to 620 nm, in 10 nm increments. All raw time-domain data were exported to Microsoft Excel (Microsoft Office 365, Microsoft Corporation, Redmond, WA). Data analysis, including the extraction of fluorescence lifetime decay constants and band fitting, was performed using Microcal Origin (OriginPro 9.0, OriginLab Corporation, Northampton, MA).</p><!><p>The quantum chemistry calculations performed in this work were based on the computational protocol that has been described in detail in our earlier work.5 We started by optimizing the geometry of the isolated FR0 chromophore in its ground electronic state S0 using the Kohn–Sham12 formulation of density functional theory,13 where we employed the CAM-B3LYP14 functional and the 6–31+G* basis set.15–17 Subsequently, we used the equation-of-motion (EOM) extension18 of the coupled-cluster (CC) theory19 to the excited electronic states to determine the vertical excitation energies characterizing the transitions from the ground state (S0) to the four lowest excited singlet states (Sn, n = 1–4) of the FR0 molecule using the composite formula (1) ωn(EOMCC)=ωn(EOMCCSD/6−31+G*)+[ωn(δ-CR-EOMCC(2,3)/6−31G)−ωn(EOMCCSD/6−31G)] </p><p>The first term on the right-hand side of eq 1 denotes the vertical excitation energy obtained using the EOM extension of the CC method with singles and doubles (CCSD),20 designated as EOMCCSD,18 in conjunction with the 6–31+G* basis set. The next two terms on the right-hand side of eq 1 correct the EOMCCSD/6–31+G* results for the higher order many-electron correlation effects due to triple excitations, represented in this work by the δ-CR-EOMCC-(2,3) approach,21–23 computed at the smaller 6–31G basis set.15 We also used the CCSD/6–31+G* and EOMCCSD/6–31+G* one-electron reduced density matrices to determine the Mulliken atomic charges, permanent dipole moments, and total electron densities characterizing the ground and excited states, as well as the transition dipole moments (TDMs) and oscillator strengths (OSs) associated with the vertical Sm → Sn excitations involving the lowest four excited singlet states of FR0.</p><p>All electronic structure calculations for the FR0 molecule reported in this work were performed using GAMESS.24,25 The relevant CCSD, EOMCCSD, and δ-CR-EOMCC(2,3) computations using the restricted Hartree–Fock (RHF) determinant as a reference and the corresponding left-eigenstate CCSD and EOMCCSD calculations, required to obtain the triples corrections of δ-CR-EOMCC(2,3) and the one-electron properties of interest, were performed using the CC/EOMCC routines developed by the Piecuch group,22,26–28 which form part of the GAMESS code. In all the post-RHF calculations, the core orbitals corresponding to the 1s shells of the C, N, and O atoms were kept frozen. In the calculations employing the 6–31+G* basis set, we used the spherical d-type polarization functions. We utilized VMD software29 to visualize the FR0 species.</p><!><p>As mentioned in the Introduction section, the super photobase FR0-SB exhibits an anomalously large change in its excited-state pKb5,7 and among the issues that remain to be resolved is the role that the fluorene derivative chromophore structure plays in producing this behavior. In this work, we investigate the precursor to FR0-SB, the aldehyde FR0 (Figure 1b), which exhibits unusual fluorescence relaxation dynamics in protic solvents. We present the time-resolved spectra for FR0 in n-propanol (Figure 2a), n-pentanol (Figure 2b), and DMSO (Figure 2c). These spectra were extracted from the time-domain experimental data with the integrated emission intensity normalized to the steady-state fluorescence spectrum.</p><p>For the time-resolved spectral evolution seen with FR0-SB, there were well-resolved emission bands corresponding to unprotonated and protonated species,5–7 but for FR0 there is no corresponding imine and, consequently, no analogous spectrally resolved features. It is thus important to consider first whether the data shown in Figure 2a,b represent spectral shifts of a single electronic state in time or the evolution of populations in multiple spectrally overlapped bands. There are several factors that point to the latter explanation being correct.</p><p>There is a substantial literature related to spectral evolution and its relationship to solvation dynamics.30–34 One of the few probe molecules that exhibit this effect is Coumarin 153, which has been studied extensively and found to exhibit a time-resolved fluorescence Stokes shift31–37 that is mediated by solvent relaxation.30,38 The data for the time-resolved Stokes shift seen for Coumarin 153 resemble, at least qualitatively, the results shown in Figure 2a,b. The observed spectral relaxation for Coumarin 153 correlates with the solvent Debye relaxation time at a scale of ca. 50–100 ps, which is somewhat shorter than that of the spectral shift we observe but not necessarily at odds with the relaxation times in normal alcohols.39 More importantly, the solvent relaxation model used in the interpretation of C153 dynamics presumes that the spectral dynamics are mediated by relaxation along a featureless reaction coordinate, with the time constant for the relaxation being determined by the solvent surrounding the dipolar excited state accommodating the change in the dipole moment orientation and magnitude resulting from excitation.</p><p>If other intramolecular processes such as IVR or relaxation between multiple conformers or electronic state manifolds can be shown to contribute, then the spectral relaxation process cannot be assigned to solvent relaxation alone.37,40–49 Indeed, the interpretation of these data is not a matter of either solvent relaxation or intramolecular relaxation determining the observed spectral dynamics. It is fully expected that both intermolecular and intramolecular relaxation contribute to the observed spectral behavior and, thus, the issue is which of these contributions influences the FR0 spectral response most prominently. The intermolecular component is reflected in the solvent-dependent dynamics we report, while the intramolecular component is supported by quantum chemical computations and emission wavelength-dependent rotational diffusion data (vide infra). The wavelength dependence of the rotational diffusion data is a consequence of different electronic states interacting with the solvent environment according to their different electron density distributions. There are several pieces of information that we consider below, both experimental and computational, that point to the spectral dynamics we report here being understood in the context of intramolecular relaxation dynamics that are mediated by specific intermolecular interactions. In addition to the analysis provided in the following sections, we have taken the time-resolved emission data at several different wavelengths across the FR0 emission band and have extracted the rotational diffusion time constants from different emission wavelengths. The reorientation time of FR0 in alcohols exhibits a measurable wavelength dependence. We will return to a more thorough discussion of this point below, but wavelength-dependent reorientation times require the existence of multiple excited electronic states, each with a finite radiative lifetime. We assert that the data result from relaxation from multiple electronic states rather than temporal evolution along a solvent-mediated relaxation coordinate for a single electronic state.</p><p>Computational chemistry has demonstrated its importance in understanding molecular-scale phenomena and spectroscopic dynamics of the super photobase FR0-SB.5–7 We have applied the protocol described in the Computational Details section to investigate the electronic structures of the ground state S0 and four lowest singlet excited states Sn, n = 1–4, of FR0. The results of these calculations reveal several interesting features that help us understand the complex spectral dynamics shown in Figure 2. The S0 → Sn vertical excitation energies and permanent dipole moments of the ground and four lowest excited singlet states of FR0 are reported in Table 1, while the TDMs and OSs characterizing the Sm → Sn vertical excitations involving all the calculated states are reported in Table 2. The Mulliken charges are shown for the ground state S0 (Figure 3a) and excited states S1 (Figure 3b), S2 (Figure 3c), and S3 (Figure 3d), and the comparison of the magnitude and direction of the permanent dipole moments in each state, along with the electron density differences characterizing the S0 → Sn (n = 1–3) vertical excitations, are shown in Figure 4. There are several interesting observations that stem from our calculations. The first one is that, as shown in Table 1, the three lowest excited electronic singlet states are in relatively close energetic proximity, which suggests a facile population relaxation from S3 to S2 or S1 and from S2 to S1. The second important result of our calculations is that the S0, S1, and S3 states exhibit relatively similar electron densities for the aldehyde moiety, with the carbonyl oxygen carrying a Mulliken charge of ca. −0.4, and, not surprisingly, the dipole moments are oriented along similar axes (see Figures 3 and 4). The S2 state, however, exhibits a significantly different electron density distribution for the aldehyde group, with the carbonyl oxygen having a Mulliken charge of ca. −0.1 and the aldehyde carbon being substantially more negative (cf. Figure 3c). These differences in Mulliken charges are reflected in the small permanent dipole moment in S2 relative to those characterizing the remaining four electronic states considered in our calculations (see Table 1 and Figure 4). These findings are consistent with our experimental data, as discussed below.</p><p>The existence of several excited electronic states in close energetic proximity combined with the unusual spectral dynamics in alcohols suggests that these experimental results are accounted for by the population relaxation between the excited electronic states and radiative relaxation from each state, rather than the temporal evolution of a single band. We propose a kinetic model that is consistent with the observed spectroscopic behavior (Figure 5), as defined below: (2) dS3dt=−(k30+k31+k32)S3,dS2dt=k32S3−(k20+k21)S2, anddS1dt=k31S3+k21S2−k10S1 </p><p>In this model, the transitions associated with k30, k20, and k10 are all considered to be radiative, while the relaxation phenomena described by k32, k31, and k21 are non-radiative. We also consider the excitation to be instantaneous. The temporal evolution of the populations of each state is given by (3) S3(t)=S3(0)exp(−(k30+k31+k32)t),S2(t)=(S2(0)+∫0tk32  exp((k20+k21)x)⋅S3(x)dx)exp(−(k20+k21)t), andS1(t)=(S1(0)+∫0texp(k10x)⋅(k21S2(x)+k31S3(x))dx)exp(−k10t) </p><p>With the time-resolved intensities of the three bands determined, it is possible to estimate the rate constants for the model shown in Figure 5 and eq 2. Comparing the kinetic model described above to the experimental data is not as straightforward as fitting individual time-resolved fluorescence intensity decays at specific wavelengths. As noted above, the emission spectrum of FR0 is broad and relatively featureless, and the presence of multiple radiative decays will be complicated by the spectral overlap of unresolved bands. We have fitted the steady-state emission spectra of FR0 in the solvents used in this work to three Gaussian bands (Figure 6). Using these fitted band energies, we have taken time slices of the normalized time-domain data and have fitted those time-resolved spectra to determine the band intensities of each of the fitted bands as a function of time. In that manner, we extract the time-evolution of each of the three spectral features. Based on the transition energies, we assign the highest, intermediate, and lowest energy bands to be the S3 → S0, S2 → S0, and S1 → S0 emissions, respectively. While not without precedent, it is noteworthy that S2 → S0 emission appears prominently,50,51 even though the calculated TDM and OS characterizing the S0 → S2 vertical excitation reported in Table 2 are very low. We have given considerable thought to this apparent incongruity. One possible explanation is vibronic intensity borrowing between the closely spaced electronic states,52–54 which is a well-established phenomenon, known to depend sensitively on the details of solvent–solute interactions that mediate coupling between the electronic state manifolds. Although the details of solvent–solute interactions are beyond the scope of our quantum chemical computations, such intermolecular interactions may play a determining role in defining the functional form of the experimental data. Based on the computational results shown in Tables 1 and 2, the S2 state is in close proximity to S1 and S3, with the energy differences being ΔE(S2 − S1) = 2017 cm−1 and ΔE(S3 − S2) = 1129 cm−1, and the OSs for the relevant S0 → S1 and S0 → S3 transitions of 0.84 and 0.14, respectively, are significant, thereby producing a condition where intensity borrowing is likely. Furthermore, because of the anti-Hermitian nature of the absorption and emission vibronic coupling terms,53,55 the OS for the emission transitions can be significantly different than those for the absorption transitions and dependent on the details of the solvent mediation of the vibronic transitions.</p><p>In addition to the limitations associated with the spectral overlap and modest signal-to-noise ratio, extracting the rate constants from the time-resolved spectral data is inherently underdetermined because we are trying to evaluate six rate constants using the decay profiles of the three bands. For this reason, we apply the following constraints to obtain physically realistic estimates of our rate constants. The first constraint is that all rate constants must be greater than or equal to zero. The second constraint is that the lifetime of the first excited singlet state, S1, is on the order of several nanoseconds (i.e., k10 ≤ 109). With these basic constraints, we have used a Monte Carlo fitting routine (coded in-house) to estimate the rate constants for the temporal evolution of the fitted emission bands (Table 3). The uncertainties (±1σ) are based on at least 10 individual determinations for each rate constant. The fits of eq 2 to the experimental data using the rate constants in Table 3 are shown in Figure 7. The agreement between the fitted and experimental data is not exact for reasons of the unaccounted for spectral overlap and the limited signal-to-noise ratio of the measured data. We note that the spectral dynamics we have discussed to this point are seen in protic solvents but not in aprotic solvents, such as DMSO. Based on this information, we assert that the spectral dynamics we observe are associated with hydrogen bonding between the solvent molecules and the FR0 chromophore.</p><p>While the ability of the solvent to form hydrogen bonds with FR0 is related to the observed spectral dynamics, it is important to consider whether solvent viscosity or [OH] mediate the dynamics. Based on the data shown in Table 3, it is not possible to determine whether the observed solvent-dependent changes in rate constants are dependent primarily on the hydroxyl concentration or the viscosity of the solvent. An alternative way to phrase this uncertainty is whether or not large amplitude molecular motion, such as the rotation of the diethylamino moiety, mediates the observed spectral relaxation dynamics. We can resolve this issue experimentally.</p><p>We performed a series of measurements in binary solvent systems of DMSO and n-propanol in which the amount of each solvent is controlled (Figure 8). The reason for the choice of DMSO and n-propanol as the two solvents is twofold: first, the spectral dynamics seen in n-propanol are not seen in DMSO and second, the two solvents have essentially the same bulk viscosity (ηDMSO = 1.99 cP and ηn-propanol = 1.94 cP).56,57 Thus, the binary solvent systems used will allow control over the hydroxyl group concentration ([OH]) under constant viscosity conditions. The data in Figure 8 show the decreasing extent of resolvable spectral dynamics with decreasing [OH]. We report the fitted rate constants for 100, 75, 50, and 25% n-propanol in Table 4. These data are shown graphically in Figure 9, and they reveal several interesting features. Specifically, the rate constants k32 and k31 decrease with increasing [OH], k20 increases with [OH] and the rate constants k30, k21, and k10 are independent of [OH]. One conclusion from these data is that the spectral dynamics we observe are associated with [OH] and not large amplitude molecular motion. This finding is consistent with the computational results (Figure 3) showing no significant accumulation of the positive charge on the amino nitrogen in any of the excited electronic states when compared to the ground state, precluding any contribution from charge-separated species.</p><p>The [OH] dependence of k32 and k31 indicates that the formation of a hydrogen bond between FR0 and the solvent alcohol mediates the intramolecular relaxation from S3 to both S2 and S1. The formation of a hydrogen bond between FR0 and an alcohol does not, however, influence k30 or k10. It is possible that the formation of a hydrogen bond between the alcohol proton and the FR0 carbonyl oxygen serves to reduce the dipole moment of the excited FR0, thereby diminishing the dipolar coupling between the excited FR0 and DMSO. This explanation is based on the observation that k3x (x = 0–2) is faster than what we can measure with our instrumentation in DMSO, implying either facile quenching of S3 by DMSO or dipolar enhancement of internal conversion in FR0. Similar to our previous observations in the computational studies of FR0-SB in its S0 and S1 states hydrogen bonded to alcohol solvent molecules,6 the formation of the hydrogen bond between the alcohol proton and FR0 carbonyl oxygen in S2 increases the magnitude of the dipole moment relative to the gas phase value, enhancing dipolar interactions between DMSO and FR0 in the S2 state. An implication of these explanations for our experimental observations is that there may be observable state-dependent solvent interactions with the FR0 chromophore.</p><p>We can test this hypothesis by examining the rotational diffusion behavior of FR0 in n-propanol, n-pentanol, and DMSO as a function of emission wavelength, which we show in Table 5. The zero-time fluorescence anisotropy and the orientational relaxation time data in Table 5 demonstrate state-dependent reorientation dynamics for FR0 in protic solvents and state-independent reorientation dynamics in DMSO. The reorientation time constants for FR0 in n-propanol and n-pentanol exhibit three distinct, wavelength-dependent values, with the reorientation time being fastest in S3 and slowest in S2, with an S1 reorientation being intermediate in value. There is no state dependence for FR0 in DMSO outside the experimental uncertainty, consistent with the rapid relaxation of the higher excited singlet states in this dipolar solvent. State-dependent reorientation times have been observed previously for oxazine dyes in alcohols, where the state dependence was found to be due to the strong association between the alcohol solvent hydroxyl group and the non-bonding lone pair on the heterocyclic nitrogen.58–61 While FR0 does not contain an analogous nitrogen, it does manifest state-dependent changes in the electron density distributions that are central to solvent–solute interactions. While it may be tempting to speculate on the state-dependent changes in solvent interactions with FR0 based on the measured time constants, there are several factors (e.g., local heating, hydrogen bond lifetime in each state, and competing solvent–solvent interactions) that we cannot determine with sufficient certainty. Regardless, the changes in zero-time anisotropies (seen most clearly for n-pentanol, determined by the angle between the excited and emitting TDM) and reorientation times (related to the volume of the reorienting entity) at different emission wavelengths demonstrate that the solvent–solute interactions of FR0 vary in a manner that is consistent with the results of our quantum chemical computations, that is, FR0 exhibits state-dependent rotational diffusion dynamics. For this to be the case, the lifetimes of the excited states involved must be on the order of hundreds of ps or more, consistent with the fitted results for k30, k20, and k10.</p><p>The spectral dynamics data, taken in conjunction with the wavelength-dependent reorientation data, support our interpretation of the spectral dynamics data in the context of multiple emissive states relaxing at different rates rather than as a single electronic state exhibiting a spectral shift. The computational results indicate that the S2 state is characterized by an electron density distribution that differs significantly from the electron density distributions seen in S1 and S3 (see Figures 3 and 4). The singlet state S2 is characterized by a permanent dipole moment that is substantially smaller in magnitude than those for S1 and S3 (Figure 3) because the change in total electronic density associated with the S0 → S2 transition is localized significantly on the aldehyde carbonyl functionality (Figure 4). The rate constant data and its dependence on the concentration of [OH] in solution points to the importance of dipole–dipole interactions in determining the coupling between the excited singlet states in FR0. Although the calculated results do not take solvent H-bonding into account, the data are consistent with the hydrogen-bonding interactions between the solvent alcohol proton and the FR0 carbonyl oxygen modulating the dipole moment in each excited state and thereby controlling the efficiency of solvent–solute dipole–dipole interactions. It is these interactions that appear to facilitate the relaxation between the excited singlet states.</p><!><p>We have examined the excited state relaxation dynamics of the substituted fluorene derivative FR0. The experimental data in conjunction with the quantum chemical calculations reveal that relaxation between three excited singlet electronic states accounts for the spectral relaxation dynamics observed for FR0 in protic solvents. The relaxation dynamics are related to the concentration of hydroxyl functionality in the solvent system, indicating that hydrogen bonding between the aldehyde oxygen of FR0 in the S2 state and the solvent hydroxyl proton diminishes the role of dipolar solvent–solute coupling, reducing the relaxation rates for some of the relaxation pathways. Further experimentation will be required to address in detail the dependence of the electronic state coupling and its dependence on solvent interactions on chemical structures.</p>

PubMed Author Manuscript

Umbrella Sampling Simulations Simultaneously Measure Switch Peptide Binding and Hydrophobic Patch Opening Free Energies in Cardiac Troponin

The cardiac troponin complex (cTn) is an important regulatory protein in heart contraction. Upon binding of Ca 2+ , cTn undergoes a conformational shift that allows the troponin I switch peptide (cTnISP) to be released from the actin filament and bind to the troponin C hydrophobic patch (cTnCHP). Mutations and modifications to this complex can change its sensitivity to Ca 2+ and alter the energetics of the transition from the Ca 2+ -unbound, cTnISP-unbound form to the Ca 2+ -bound, cTnISP-bound form. We utilized targeted MD (TMD) to obtain a trajectory of this transition pathway, followed by umbrella sampling to estimate the free energy associated with the cTnISP-cTnCHP binding and the cTnCHP opening events for wild-type (WT) cTn. We were able to reproduce experimental values for the cTnISP-cTnCHP binding event and obtain cTnCHP opening free energies in agreement with previous computational measurements of smaller cTnC systems. This excellent agreement for WT cTn demonstrated the strength of computational methods in studying the dynamics and energetics of cTn complex. We then introduced mutations to the cTn complex that cause cardiomyopathy or alter its Ca 2+ -sensitivity and observed a general decrease in the free energy of opening the cTnCHP. For these same mutations, we observed no general trend in the effect on the cTnISP-cTnCHP binding event. Our method sets the stage for future computational studies on this system that predict the consequences of yet uncharacterized mutations on cTn dynamics and energetics.

umbrella_sampling_simulations_simultaneously_measure_switch_peptide_binding_and_hydrophobic_patch_op

Introduction<!>Model Preparation for TMD<!>TMD Optimization and Simulations<!>Umbrella Sampling<!>WHAM Free Energy Estimation<!>TMD Simulations Generate a Transition Pathway for Umbrella Sampling<!>WT Free Energies of Patch Opening and Switch Peptide Binding Agree with Previous Experiments<!>HCM and RCM Mutations Generally Decreased cTnCHP Opening Free Energy<!>No Observed Trend in cTnISP-cTnCHP Binding Free Energy<!>Thr143 Phosphorylation May Increase Ca 2+ -sensitivity of cTnC<!>Conclusions

<p>Contraction of the thin filament in heart muscle is regulated by Ca 2+ binding to the cardiac isoform of the troponin (cTn) complex. 1 Upon binding of Ca 2+ to cTn, a conformational shift occurs that releases cTn from actin, exposing myosin binding sites that allow muscle contraction to occur. The exact mechanism of the cTn conformational transition and how mutations cTn effect his process is still not fully understood.</p><p>The cTn complex is comprised of 3 subunits: the Ca 2+ -binding subunit (cTnC), the inhibitory subunit (cTnI), and the tropomyosin binding structural subunit (cTnT) (Figure 1). cTnC can be divided in two regions: N-terminal cTnC (NcTnC, residues 1-89) contains Ca 2+ -binding sites I and II, and C-terminal cTnC (CcTnC, residues 90-161) contains Ca 2+ -binding sites III and IV. Under normal physiological conditions, sites III and IV are always occupied, and site I is never occupied by Ca 2+ . Thus, site II of NcTnC is predominately responsible for sensing Ca 2+ for muscle contraction. When Ca 2+ binds to site II, it induces a conformational and dynamical shift that exposes a hydrophobic patch (cTnCHP) in NcTnC that is able to bind the cTnI switch peptide (cTnISP, residues 149-164). The cTnISP (along with the cTnI inhibitory peptide) interacts with actin in the Ca 2+ -unbound state, and its dissociation from actin and subsequent binding to the cTnCHP in the cTn Ca 2+ -bound state allows for the thin filament interaction with myosin necessary for muscle contraction.</p><p>Mutations throughout the cTn complex subunits have been shown to cause cardiomyopathy, a disease that can lead to heart failure or sudden cardiac death. 2 3 types of cardiomyopathy caused by these mutations include: Hypertrophic (HCM), Restrictive (RCM), and Dilated (DCM). HCM causes thickening of the interventricular septum, leading to decreased ventricular filling and impaired diastolic function of the heart. RCM is a rarer disease but causes similar problems with decreased ventricular filling without thickening of the heart muscle. Both HCM and RCM mutations have been shown to increase Ca 2+ -sensitivity in cTnC. DCM has been shown to decrease cTnC Ca 2+ -sensitivity, with physiological affects including an increase in the size of the left ventricle, or both ventricles, and impaired systolic function. 3 Structurally, cTn has been well characterized. A partial crystal structure of the Ca 2+ -bound cTn complex, published by Takeda et al. 4 in 2003, constituted a major advancement of our structural understanding of the complex. More recently, a cryo-EM structure of the entire cTn complex in the context of the thin filament was published by Yamada et al. 5 This study was able to elucidate structures of the cTn complex, anchored onto the thin filament, in the Ca 2+ -bound and Ca 2+unbound states. Although this breakthrough was able to provide snapshots of the cTn complex and thin filament in both states, it was unable to provide information on the conformational transition and thermodynamic pathway between the two. A quantitative assessment of these processes is desirable for our understanding of heart muscle biology and disease, drug discovery and protein design.</p><p>Computational methods can facilitate the study of these processes. 6 Previous efforts have been able to utilize molecular dynamics (MD) and free energy methods to study Ca 2+ -binding to cTnC site II. [7][8][9][10][11] MD methods have also been able to investigate disruptions in dynamics upon the introduction of mutations that alter site II's Ca 2+ -sensitivity or mutations that have been linked with disease. [12][13][14] Enhanced sampling methods, umbrella sampling, and long timescale simulations have been utilized to study the energetic pathway for the transition between Ca 2+ -unbound and Ca 2+ -bound forms. [15][16][17][18] Simulations have also incorporated cTnI into the system to elucidate how other subunits in the cTn complex can affect the dynamics of cTnC. [19][20][21][22][23][24] Computer aided drug discovery has also proven effective in helping identify potential therapeutics for heart disease by targeting the cTn complex. [25][26][27][28][29] How mutations located in regions not involved in either the coordination of Ca 2+ to site II or the cTnISP-cTnCHP binding event can alter the overall dynamics of muscle contraction, remains underexplored. One hypothesis on how some of these mutations, particularly those in the cTnI inhibitory peptide, may affect troponin function was that they alter the availability, or effective concentration, of the cTnISP to the cTnCHP. Our lab previously explored this hypothesis by determining the volume sampled by the cTnISP during MD simulations. We saw no significant difference in cTnISP effective concentration between WT and mutations in the cTnI inhibitory peptide (residues 137-148), 19 leading us to search for other computational methods to explore the effects of mutations on cTnC dynamics.</p><p>Another interesting interaction that has been, at least computationally, underexplored is the binding of the cTnISP to the cTnCHP region. Experimentally, this has been studied before by Li et al. in an NMR kinetic study of cTnI147-163 titrated into a Ca 2+ -bound NcTnC. The authors estimated a binding KD of 154µM. 30 Another experiment by Tikunova et al. estimated the binding affinity to be 200nM using IAANS fluoresence, 31 measuring an almost three orders of magnitude as compared to the first study. The difference is likely explained by the different sequences of cTnI used, where the NMR experiment used residues cTnI147-163, and the fluorescence experiment used cTnI128-180. Because residues C-terminal of the cTnISP are largely structural, 6 the additional 17 residues located on the C-terminal end of the cTnISP may significantly increase the binding affinity of cTnISP to the cTnCHP. In order for the cTnISP to bind to the cTnCHP, cTnC first needs to undergo an opening event. Although this event is difficult to measure experimentally, 32 it has been studied computationally before by Bowman et al. using a model of just the N-terminal region of cTnC. 16 We therefore set out to simultaneously estimate the free energy associated with the cTnISP-cTnCHP binding and the opening of the cTnCHP using a combination of MD-based simulations and free energy methods. To obtain a trajectory of conformations between the Ca 2+ -unbound, cTnISPunbound structure and the Ca 2+ -bound, cTnISP-bound structure, we employed targeted molecular dynamics (TMD). This specific type of MD simulation can steer a starting model towards a target model by applying an additional potential energy function. 33 The implementation of this MD method has been shown recently to help elucidate transition pathways of other proteins. 34,35 Using the conformations along this transition, we performed umbrella sampling to determine the free energy associated with two separate events: opening of the cTnCHP and cTnISP binding to cTnCHP. We also applied this same method to mutations known to be linked with HCM or RCM, one mutation that was experimentally designed to increase Ca 2+ -sensitivity of cTnC, and the phosphorylation of cTnI Thr143. Given that Ca 2+ -binding causes opening of the cTnCHP, we hypothesized that mutations linked with HCM or RCM and/or known to experimentally increase Ca 2+ -sensitivity would lower the free energy associated with the cTnCHP opening event. For the cTnISP-cTnCHP interaction, we hypothesized that these same mutations may cause the free energy of cTnISP-cTnCHP binding to become less favorable due to the mutations potentially disturbing important residue-residue contacts necessary for binding.</p><p>We were able to make two important observations from these simulations. First, our TMD method followed by umbrella sampling was able to calculate values for both the cTnCHP opening and cTnISP binding free energies in agreement with previous experimental and computational studies. This agreement allowed us to conclude that our method is a reliable way to study the free energy profile of the transition between Ca 2+ -unbound and bound conformations of cTn. Second, we saw a general decrease in free energy of the cTnCHP opening upon introduction of mutations linked with HCM and RCM, as well as the Ca 2+ -sensitizing mutation L48Q. However, we did not observe a general weakening or strengthening of the cTnISP-cTnCHP binding upon introduction of these same mutations. In addition to these findings, our simulations suggest that cTnI Thr143 phosphorylation may cause Ca 2+ -sensitization of cTnC due to a decrease in free energy of cTnCHP opening when compared to WT.</p><!><p>To run TMD, a start and end point for the cTn complex simulations needed to be obtained. For the starting model, chains a, b, and c from PDB entry 6KN7 5 were extracted (Figure 1A). This model provides atom positions for all three subunits of cTn in the Ca 2+ -unbound, cTnISP-unbound form in the context of the thin filament. For the end point of the TMD simulation, chains a, b, and c from PDB entry 6KN8 5 were extracted (Figure 1B). This model provides atom positions for all three subunits of cTn in the Ca 2+ -bound, cTnISP-bound form in the context of the thin filament. Since TMD requires identical atoms for the start and end models, the following residues for both extracted models were used in the simulations: cTnC residues 2-161, cTnI residues 41-166, and cTnT residues 199-272. Neither of these two PDBs provided information on the position of Ca 2+ ions, so Ca 2+ was added to sites II, III, and IV of the cTnC in both models using AutoDock Vina. 36 Although the 6KN7 model represents a Ca 2+ -unbound, cTnISP-unbound form of the cTn complex, Ca 2+ binding initiates the conformational change to the cTnISP-bound form of the cTn complex, making the addition of Ca 2+ to site II of the start model necessary to simulate the conformational change appropriately.</p><!><p>For the TMD simulations, the structures were solvated with explicit TIP3W water molecules and NaCl counterions were added to neutralize the entire system and establish a final salt concentration of 150mM. The system was minimized over two sequential 10,000 minimization steps to optimize the positions of the solvent molecules around the protein and to subsequently relax the positions of residue side chain atoms. The system was then slowly heated up to 310K over an initial equilibration of 190,000 steps, followed by a final equilibration of 10,000 steps to bring the system into the final production run conditions. All preparation steps and TMD simulations were conducted using NAMD 2.13, 37 the Charmm36 force field, 38 and a 2 femtosecond timestep. The TMD simulations were conducted using an NPT ensemble at 310K with Langevin temperature and pressure dampening. All bonds with hydrogens were constrained using the ShakeH algorithm which allowed for a 2 femtosecond timestep, with structures being saved every 2 picoseconds.</p><p>Targeted Molecular Dynamics (TMD) adds an additional potential energy term (UTMD) to specified atoms to guide them to a target structure. UTMD is defined as: 39</p><p>Where RMSD(t) is the instantaneous best-fit root mean square distance of the current coordinates from the target coordinates, and RMSD*(t) evolves linearly from the initial target RMSD at the first TMD step to the final target RMSD at the last TMD step. The spring constant k (kcal/mol/Å 2 ) is scaled down by the number of targeted atoms, N. 39 Five different values for the spring constant, k (50, 100, 200, 500, 1000), were tested at five different lengths of TMD simulation time (20ns, 40ns, 60ns, 80ns, 100ns) , creating 25 unique combinations of TMD parameters. These simulations were carried out on the Owens Cluster of the Ohio Supercomputer Center 40 using 28 processors on 1 node, for 500,000 steps per nanosecond.</p><p>After the simulation was completed, the system was stripped of all water molecules and Na + /Clions. Then, each saved structure was aligned to the first frame to create a new DCD file. For each frame in the resulting DCD file, the all-atom RMSD to the target structure was determined using VMD, and the interhelical angle of cTnC helices A (residues 14-28) and B (residues 40-48) was evaluated using interhlx. 41 The interhelical angle reported by interhlx is 180 degrees minus the angle theta reported by the program VGM. 42 After the best parameters for the spring constant and simulation length had been determined by evaluating the opening of the cTnCHP and the helical nature of the cTnISP during TMD simulations, 3 independent trials of TMD simulations with a spring constant of 200 kcal/mol/Å 2 were run for 60ns to obtain trajectories of the transition from Ca 2+ -unbound, cTnISP-unbound to the Ca 2+ -bound, cTnISP-bound form of cTn.</p><p>The DCD files for the 3 TMD simulation trials were stripped and aligned exactly as described above. For each frame in these trajectories, two values were determined: interhelical distance between cTnC helices A and B, and the distance between the cTnISP and the cTnC hydrophobic patch. Interhelical distance was used as a proxy for interhelical angle to evaluate the degree of openness of the hydrophobic patch by measuring the distance between residues 14 and 48 of cTnC. The position of residues 14 and 48 in each frame, respectively, was determined by averaging the position of the residues' N, C, and CA atoms. The distance between the cTnISP (residues 149-164) and the cTnCHP (residues 20, 23, 24, 26, 27, 36, 41, 44, 48, 57, 60, 77, 80, 81) was determined by calculating the average position of the CA atoms in the collective group of residues for each region and measuring the distance between the two centers of mass. The trial that exhibited the most sampling close to the line of regression was selected as the representative trajectory from which frames were extracted to run umbrella sampling.</p><!><p>Specific frames from the representative TMD trajectories were selected as windows for umbrella sampling (US) simulations (Table S1). The timestep for these simulations was set to 1fs. For each US window, two separate collective variables were applied to restrain the US windows to coordinates along the investigated conformational transition. The variables were defined as shown in Table S1 and a harmonic force constant of 5 kcal/mol was applied. Interhelical distance was restrained by evaluating the distance between the average position of atoms N, C, and CA in residues 14 and 48 of cTnC. cTnISP-cTnCHP distance was restrained by evaluating the average position of the CA atoms in the collective group of residues for each region (see TMD section above). Each of the 27 windows was then run for 5ns in 3 independent simulations, at 310K under an NPT ensemble, similar to the TMD simulations.</p><p>Simulations that used cTn structures with cardiomyopathic mutations, phosphorylation, or Ca 2+sensitizing mutations were created by modifying the windows described above. For cardiomyopathic or Ca 2+ -sensitizing mutations, the mutations were introduced to each window using the Mutagenesis Wizard function available in PyMOL. 43 Mutations introduced to cTnC include: A8V, L29Q, A31S, L48Q, and C84Y. Mutations introduced to cTnI include: R141Q, L144P, L144Q, R145G, R145Q, R145W, A157V, R162P, R162Q, and R162W. The cTnI Thr143 phosphorylated (T143p) windows were created using the PyTMs plug-in available in PyMOL. 44 Each of the modified windows were also run for 5ns in 3 independent simulations, at 310K under an NPT ensemble.</p><!><p>For each frame in each US trajectory, the values of the collective variables were evaluated. The data for each window was combined across the three independent simulations to provide 15ns of sampling in each window. The Weighted Histogram Analysis Method (WHAM) 45 was used to create a free energy profile for each of the collective variables, respectively, for each simulated system using all 27 simulation windows. This allowed us to estimate free energy profiles of the cTnCHP opening transition coordinate and cTnISP association coordinate, respectively, using the same simulations. The minimum and maximum values provided to WHAM for the interhelical distance collective variable were 14.5Å and 28.5Å, respectively. The minimum and maximum values provided to WHAM for the cTnISP-cTnCHP distance collective variable were 9.58Å and 16.15Å, respectively. The histogram analysis for both profiles was divided into 27 bins, with a tolerance of 0.001. A force constant of 10 kcal/mol was used after properly adjusting the 5kcal/mol spring constant applied in the charmm36 force field, as described in "An implementation of WHAM: the Weighted Histogram Analysis Method" by Alan Grossfield. 45 Monte Carlo bootstrap error analysis was performed for each window using 10 fake data sets generated with a decorrelation time of 50 steps.</p><!><p>Targeted MD simulations were used to obtain a trajectory of the transition from a Ca 2+ -unbound, cTnISP-unbound cTn complex to a Ca 2+ -bound, cTnISP-bound cTn complex. We optimized two variables in the potential energy function for the TMD term: the force constant, k (kcal/mol/Å 2 ), and the length of the simulation, t. TMD simulations were run for five separate force constants, at increasing simulation lengths. For each simulation, the interhelical angle between cTnC helices A and B for all frames captured from the trajectories was determined. TMD trajectories run for t = 20ns and t = 40ns exhibited abnormally high interhelical angles for the lowest three force constants (k = 50kcal/mol/Å 2 , k = 100kcal/mol/Å 2 , k = 200kcal/mol/Å 2 ), while higher force constants (k = 500kcal/mol/Å 2 , k = 1000kcal/mol/Å 2 ) led to a rapid cTnCHP opening (Figure S1). Additionally, visual inspection for all these simulations revealed denaturation of the cTnISP into a disordered region before interaction with the cTnCHP. For the t = 60ns simulations, the trajectories with k = 50kcal/mol/Å 2 and k = 100kcal/mol/Å 2 force constants again experienced unnaturally high interhelical angles. The k = 200kcal/mol/Å 2 and k = 500kcal/mol/Å 2 trajectories exhibited sufficient sampling of both the closed and open states of the cTnCHP while also sampling a smooth transition between the two (Figure 2). Visual inspection of the cTnISP region revealed that cTnISP maintained its helical nature for both force constants throughout the 60ns simulation. We therefore chose to move forward with the parameters t = 60 ns and k = 200 kcal/mol/Å 2 to avoid denaturation of parts of the Tn complex. Simulations ran for t = 80 ns and t = 100 ns across all force constants also sampled the anticipated ranges of interhelical angles and exhibited gradual cTnCHP opening events, but we decided to use the shorter t = 60ns simulations for production runs.</p><p>3 independent trials of TMD of the WT cTn complex were run using these parameters (k=200 kcal/mol/Å 2 and simulation length of 60ns) and the sampling along the cTnCHP patch opening coordinate was investigated. The correlation between interhelical angle and interhelical distance for helices A/B can be seen in Figure S2. Furthermore, the data for each frame in the trajectories was analyzed with respect to the interhelical and cTnISP-cTnCHP distances, with a line of regression being created along the patch opening event (Figure 3A). To obtain full coverage of the transition coordinate between the closed and open hydrophobic patch, target values for the interhelical distance were set every 0.5Å between 15Å and 28Å, creating 27 target interhelical distances. This corresponded to interhelical angles ranging from approximately 95° to 130°. Each of these values was paired according to the regression with a target cTnISP-cTnCHP distance to obtain full coverage of the transition coordinate between an unbound cTnISP to a bound cTnISP. This resulted in 27 windows with pairs of target values, ranging from window 1 with a closed cTnCHP and unbound cTnISP (interhelical distance = 15Å, cTnISP-cTnCHP distance = 15.733Å) to window 27 with an open cTnCHP bound to the cTnISP (28Å, 10.273Å) (Figure 3B). Figures 3C, 3D, and 3E show the starting conformation of the cTn complex for windows 1, 14, and 27, respectively. Window 1 shows the cTnISP near a closed cTnCHP, window 9 contains the cTnISP in a primed position to bind to the cTnCHP (while the patch is only in a semi-open conformation), and window 27 corresponds to the cTnISP completely bound to an open cTnCHP.</p><!><p>We simulated the 27 extracted windows using umbrella sampling and evaluated the results with WHAM. Analysis of the collective variables used in the umbrella sampling revealed that the evenly spaced windows exhibited sufficient overlap and that we were able to sample the entire reaction coordinate (Figure S3). From this, the estimated free energy of the cTnISP-cTnCHP binding event was determined to be -5.61 ± 0.10 kcal/mol. Figure 4A shows the calculated free energy per window as a function of the distance between the cTnISP and cTnCHP. The free energy was measured from the peak of the free energy curve in window 9 (the unbound state; at a cTnISP-cTnCHP distance of about 14.053Å) to the free energy estimation in the last window (the bound state; at a cTnISP-cTnCHP distance of about 10.3Å; window 27). This was done because the cTnISP was not in proximity to the cTnCHP binding site until window 9, when the patch opening reached a semi-open conformation suitable for cTnISP-cTnCHP binding to commence.</p><p>An NMR kinetic study of cTnI147-163 titrated to a Ca 2+ -bound NcTnC estimated a binding KD of 154 ± 10 µM. 30 This experimental study is the most comparable to our simulation data because a highly similar region for the cTnISP was used. At T = 303.15 Kelvin (30° C), this data corresponds to an experimentally determined free energy of -5.29 ± 0.05 kcal/mol, a value within 6% of our computational value. Other experimental studies explored the binding of longer cTnI segments to cTnC. However, since residues located C-terminally of the cTnISP were not resolved in the 6KN8 PDB structure, we were unable to test the effect of these residues on the switch peptide binding affinity. Given this small deviation, we hypothesized that our method of using TMD to produce windows representative of a transition coordinate, in combination with umbrella sampling to measure the free energy along that coordinate, can reliably reproduce experimentally determined free energy values of events associated with the cTn conformational transition. Applying the same strategy to the cTnCHP opening transition coordinate, we estimated the free energy of the opening event to be 15.80 ± 0.47 kcal/mol. The WHAM determined free energy as a function of the opening coordinate can be seen in Figure 4B. Although the free energy of this event has not been studied experimentally, the result agrees with previous computational simulations that measured the free energy of cTnCHP opening in NcTnC to be 13.8 ± 2.2 kcal/mol using a combination of steered MD and umbrella sampling. 16</p><!><p>After developing a reliable method to simultaneously measure the free energy of the cTnCHP opening and cTnISP-cTnCHP binding events, we subsequently introduced mutations and posttranslational modifications in cTn to determine how these changes affect the free energies of patch opening and switch peptide binding. We tested 12 mutations that have been linked with HCM, two mutations linked with RCM, one designed mutation that increases Ca 2+ -sensitivity without causing disease, and phosphorylation of cTnI Thr143. All modifications tested were located in the NcTnC region, or between residues 141-162 of cTnI. Data for the estimated free energy of patch opening and switch peptide binding for all 16 modified models can be seen in Table 1. 11 of the modifications had a significant decrease in the free energy of the cTnCHP opening, while the other five (cTnC L29Q, cTnC A31S, cTnI R145G, cTnI R162P, cTnI R162W) caused no significant change in the free energy. We did not observe a significant increase in the free energy of the cTnCHP opening event for any of the modified models.</p><p>Three mutations had a drastic effect on the cTnCHP opening, all of which are in NcTnC: A8V (9.67 kcal/mol), L48Q (6.56 kcal/mol), and C84Y (7.72 kcal/mol). The A8V and C84Y mutations have been shown experimentally to increase Ca 2+ -sensitivity of force generation in skinned cardiac muscle fibers. 46 Interestingly, the mutation that caused the biggest change in free energy of this opening event was L48Q, a designed cTnC Ca 2+ -sensitizing mutation which has not been linked with disease. 47,48 This mutation also caused the most favorable cTnISP-cTnCHP interaction (-7.06 kcal/mol) amongst all mutations studied (see next section).</p><p>Mutations in the cTnI inhibitory peptide and cTnISP regions all caused either a slight decrease in the cTnCHP opening free energy or did not affect the free energy significantly. Prior experiments have suggested that mutations that cause a more severe increase in Ca 2+ -sensitivity lead to RCM, whereas mutations that cause a less severe increase in Ca 2+ -sensitivity lead to HCM. 49 Our data would support this theory since mutations linked with RCM in the cTnI inhibitory peptide and cTnISP regions (L144Q, R145W) generally had a larger effect on the cTnCHP opening free energy, with only the R162Q mutation causing a similar effect to the RCM mutations. Currently, no RCM mutations have been identified in cTnC, 2, 50 so we were unable to test this trend amongst the cTnC mutations.</p><!><p>We were unable to observe any consistent trend in the free energy of cTnISP-cTnCHP binding. Four mutations caused a more favorable interaction, eight mutations caused a less favorable interaction, and three mutations, plus the T143p modification, had no measurable effect on cTnISP-cTnCHP binding. The effect of cTn mutations on cTnISP-cTnCHP binding has been scarcely studied before.</p><p>A study by Tikunova et al. that estimated the binding affinity of cTnI128-180 to cTnC in the presence of Ca 2+ -sensitizing, non-cardiomyopathic mutations observed a slight increase in KD when compared to the WT. 31 We would therefore hypothesize to see a less favorable interaction between the cTnISP and the cTnCHP given mutations linked with HCM or RCM or those that have been experimentally shown to cause a Ca 2+ -sensitizing effect. Although we did see this trend in most of the mutated models, for the one Ca 2+ -sensitizing, non-cardiomyopathic mutation that was studied (L48Q), we observed a significantly more favorable cTnISP-cTnCHP interaction (-7.06 kcal/mol) as compared to WT. A previous experiment that focused on the A8V mutation indicated that it causes an increase in the affinity of cTnISP to the cTnCHP, 51 which we corroborated here with a more favorable binding free energy of cTnISP-cTnCHP for A8V (-6.21 kcal/mol) as compared to WT.</p><!><p>Phosphorylation of cTnI has been shown as a mechanism for increasing the heart rate under stress by allowing the heart muscle to contract at a faster rate as a consequence of a decrease in Ca 2+sensitivity from phosphorylation of cTnI Ser 22/23. 52 However, studies on the phosphorylation of cTnI Thr143 have actually shown the adverse effect, namely that phosphorylation of this site may cause an increase in Ca 2+ -sensitivity. 53 Our phosphorylated model of cTnI Thr143 (cTnI T143p) slightly decreased the cTnCHP opening free energy (13.22 kcal/mol) while having no effect on the free energy of cTnISP-cTnCHP binding. As a decrease in free energy of cTnCHP opening was observed for other cTn Ca 2+ -sensitizing mutations, we therefore hypothesize that cTnI T143p may cause an increase in Ca 2+ -sensitivity, as reported by Wang et al., 53 without having any measurable effect on the interaction between the cTnISP and cTnCHP.</p><!><p>In this study, we modeled the transition of cTn from the Ca 2+ -unbound to Ca 2+ -bound states by using targeted molecular dynamics (TMD). Using the trajectory of this transition, we subsequently performed umbrella sampling simulations to determine the free energy associated with two separate events: opening of the cTnCHP and cTnISP binding to cTnCHP. Results obtained for the cTnISP-cTnCHP interaction in the WT model agree with previous experimental results, 30 while our cTnCHP opening free energy estimation agrees with previous computational efforts. 16 This excellent agreement for WT cTn demonstrates the important role of advanced computational methods in quantitatively studying cTn dynamics and muscle contraction. The methodology described here, using TMD followed by umbrella sampling, could also be applied to other protein systems given the availability of structures in two different states.</p><p>We further applied the methodology to cTn mutations and post-translational modifications. Our results showed that mutations linked with HCM or RCM cause a Ca 2+ -sensitizing effect, generally lowering the free energy associated with the cTnCHP opening. We also predict that the cTnI T143p modification will cause a Ca 2+ -sensitizing effect due to the modification lowering the free energy of the cTnCHP opening. We observed that most of the studied mutations cause a less favorable free energy associated with cTnISP -cTnCHP binding, although this trend did not hold for all HCM and RCM mutations. Our method could impact the future of computational studies on this system by predicting the consequences of unknown mutations on cTn energetics. Additionally, this method could prove helpful in rationale protein design. Overall, we showed that our protocol of utilizing umbrella sampling to measure the free energy of transitions sampled by TMD is a reliable method for studying the energetics of the cTn complex. Table 1. Free Energy of cTnCHP opening and cTnISP binding events for all models. All data shown in units of kcal/mol.</p>

ChemRxiv

Theoretical and Experimental Studies of Tyrosyl Hydroperoxide Formation in the Presence of H-bond Donors

Oxidative damage to biomolecules such as lipids, proteins, nucleotides and sugars has been implicated in the pathogenesis of various diseases. Superoxide radical anion (O2\xe2\x80\xa2\xe2\x88\x92) addition to nitrones bearing an amide N-H has been shown to be more favored compared to other nitrones (Villamena, F. A., et al., J. Am. Chem. Soc., 2007, 129, 8177\xe2\x80\x938191). It has also been demonstrated by others (Winterbourn, C. C., et al., Biochem. J. 2004, 381, 241\xe2\x80\x93248) that O2\xe2\x80\xa2\xe2\x88\x92 addition to tyrosine to form hydroperoxide is favored in the presence of basic amino groups but the mechanism for this observation remains obscure. We, therefore, hypothesized that the \xce\xb1-effect resulting from the interaction of O2\xe2\x80\xa2\xe2\x88\x92 with N-H can play a crucial role in the enhancement of hydroperoxide formation. Understanding this phenomenon is important in the elucidation of mechanisms leading to oxidative stress in cellular systems. Computational (PCM/B3LYP/6-31+G**//B3LYP/6-31G level of theory) as well as experimental studies were carried out to shed insights into the effect of amide or amino N-H on the enhancement (or stabilization) of hydroperoxide formation in tyrosine. H-bond interaction of amino acid group with O2\xe2\x80\xa2\xe2\x88\x92 results in the perturbation of the spin and charge densities of O2\xe2\x80\xa2\xe2\x88\x92. Similar phenomenon has been predicted for non-amino acids bearing H-bond donor groups. Using FOX assay, tyrosyl hydroperoxide formation was enhanced in the presence of H-bond donors from amino acids and non-amino acids. The role of H-bonding in either stabilizing the hydroperoxide adduct, or facilitation of O2\xe2\x80\xa2\xe2\x88\x92 addition via \xce\xb1-effect was further theoretically investigated, and results show that the latter mechanism is more thermodynamically preferred. This study provides new mechanistic insights in the initiation of oxidative modification to tyrosyl radical.

theoretical_and_experimental_studies_of_tyrosyl_hydroperoxide_formation_in_the_presence_of_h-bond_do

Introduction<!>a. General procedure<!>b. Computational Method<!>a. Tyrosyl hydroperoxide formation with amino acids<!>b. Tyrosyl hydroperoxide formation with non-amino acids<!>c. Alpha effect versus stabilization of the tyrosyl hydroperoxide formation in the presence of urea<!>Conclusion<!>Supporting Information Available

<p>Reactive oxygen species, such as superoxide radical anion (O2•−), have been shown to play a crucial role in modulating cell function, signaling, and immune response (1). However, production of O2•− can be induced through various chemical, enzymatic, or biological means (2–4) and in unregulated concentrations, O2•− can be a major source of the most highly oxidizing species known to exist in biological systems such as peroxynitrite (ONOO−), oxidized glutathione radical anion (GSSG•−), hypochlorous acid (HOCl), carbonate radical anion (CO3•−), or hydroxyl radical (HO•) (1). Superoxide is not highly reactive in spite of its free radical nature but its selective reactivity with other radical species (e.g., NO, tyrosyl radical) and transition metal ions such as Fe(II) (5) makes O2•− one of the toxic radical species in biological system.</p><p>In our efforts to develop spin traps with improved properties for analytical and therapeutic applications (6–11), we have demonstrated that nitrones with an amide substituent, e.g., 5-carbamoyl-5-methyl-pyrroline N-oxide (AMPO), exhibit higher reactivity towards O2•− compared to other known spin traps such as 5,5-dimethyl-pyrroline N-oxide (DMPO), 5-diethoxyphosphoryl-5-methyl-pyrroline N-oxide (DEPMPO) and 5-ethoxycarbonyl-5-methyl-pyrroline N-oxide (EMPO). This high reactivity towards O2•− has been rationalized to be due to a combination of electrostatics and intra-molecular H-bonding interaction of the O2•− with the amide-H at the transition state of the adduct (10). This observation has given rise to more questions about the possibility that this process could also be happening in protein systems in which amide moiety is abundant, and hence, can have significant ramification in the initiation of oxidative damage to biomolecules.</p><p>Oxidative damage is prevalent in protein systems and oxidative modification has been shown to lead to loss of protein function (2, 12–14). The addition of O2•− to the phenoxyl (PhO•) radical leading to the formation of hydroperoxide suggests a similar oxidative modification may occur in peptides or proteins with tyrosyl radical (TyrO•) group (15). Superoxide has the ability to preferentially interact with certain amino acids in biological systems such as the TyrO• through an addition reaction to produce hydroperoxide (16–19). In addition, the formation of hydroperoxide adduct prevails over the formation of tyrosine dimers, or phenol and O2 via electron transfer mechanism (18, 19). In peptides, the efficiency of the reaction of TyrO• to O2•− has been proposed to be dependent on the proximity of the tyrosyl moiety to the amino or amide groups (17). Thus, it has been suggested that hydroperoxides such as tyrosyl hydroperoxide and tyrosine dimers can be used as biomarkers of oxidative stress in a number of pathophysiological condition such as cardiovascular disease (17). TyrO• is part of the catalytic cycle of ribonucleotide reductase (20–22), prostaglandin synthase and photosystem II (23), and is being formed from myoglobin (24) and peroxidases (25) in the presence of hydrogen peroxide. Furthermore, since modification of tyrosine moieties in proteins has been shown to result in enzyme deactivation (13), studies involving O2•− reaction to TyrO• are relevant.</p><!><p>Tyrosyl hydroperoxide formation was generated according to the method previously described (19). In a typical experiment, a solution composed of horseradish peroxidase (HPO) (10 µg/mL), xanthine oxidase (XO) (0.05 U/mL), tyrosine (1 mM), amino acid (1 mM), and acetaldehyde (1 mM) in 10 mM phosphate buffer solution was allowed to sit at room temperature for 30–40 minutes then catalase (20 µg/mL) was added to consume any extra hydrogen peroxide. The final solution with a volume of 700 µL was allowed to sit for another 10 minutes and analyzed for hydroperoxide formation using FOX assay (18, 26, 27). FOX reagent is composed of 0.4 M sorbitol, 1 mM Fe(NH4)SO4, 0.4 mM xylenol orange in 0.2 M H2SO4, and 250 µL of this reagent was added to the 700 µL of reaction mixture. The solution was allowed to sit at room temperature for 45 minutes and absorbance reading was obtained at 560 nm. Peroxide equivalents were calculated based on cumene hydroperoxide as standard and were corrected by subtracting the absorbance readings from the mixture in the absence of tyrosine. Similar experiments were performed using non-amino acid in the presence and absence of H-bond donor groups. All experiments were done in duplicate or triplicate. The amino acids used were arginine, lysine, glycine, tryptophan, histidine, asparagine and phenylalanine. Non-amino acids used include benzylamine, aniline, acetamide, 3-amino benzoic acid, urea, benzoic acid, phenyl acetic acid, benzyl alcohol, ethanol, N,N-dimethyl formamide, N,N,N′,N′-tetramethylurea, triethylamine and 3-dimethylamino-methyl benzoate. As controls, each reaction was repeated by adding superoxide dismutase (SOD) (20 µg/mL) solution to the mixture in order to consume any superoxide that was formed and peroxide equivalents were also calculated in the presence of 2 mM amino acid or non-amino acid alone and in the absence of tyrosine. Time dependence studies. Same as above but 1 mM of urea was added during the 30 min incubation period at different time points, i.e., 0, 5, 10, 15, 20, 25, and 30 min. After incubation, catalase was added and the final solution was quantified for hydroperoxide formation using FOX assay.</p><!><p>Conformational search was carried out using Spartan 04 (28) via a Monte Carlo method coupled with the MMFF-94 force field. Variations in energetics as a function of basis set and method was assessed using B3LYP/6-31G* geometries and single-point energies at B3LYP/6-31+G**, B3LYP/6-311++G**, B3LYP/aug-cc-pVDZ, BHandHLYP/6-31+G**, mpw1pw91/6-31+G** and PBE1PBE/6-31+G** levels of theory. Reaction energy at the B3LYP/6-31+G**//B3LYP/6-31G* is within ∼1 kcal/mol difference of the energetics calculated at the B3LYP/aug-cc-pVDZ and BHandHLYP/6-31+G** levels (see Table S1 of Supporting Information), therefore, B3LYP/6-31+G**//B3LYP/6-31G* offers a reasonable approximation to the higher-level basis set and BHandHLYP/6-31+G** which is effective in approximating barrier height energies (29, 30). Density functional theory (DFT) (31, 32) was applied in this study to determine the optimized geometry, vibrational frequencies, and single-point energy of all stationary points (33–36). All calculations were performed using Gaussian 03 (37) at the Ohio Supercomputer Center. Single-point energies were obtained at the B3LYP/6-31+G** level based on the optimized B3LYP/6-31G* geometries, and the B3LYP/6-31+G**//B3LYP/6-31G* wave functions were used for Natural Population Analyses (NPA) (38). The effect of aqueous solvation was also investigated using the polarizable continuum model (PCM) (39–43). These calculations used six Cartesian d functions. Stationary points for all the optimized compounds have zero imaginary vibrational frequency as derived from a vibrational frequency analysis (B3LYP/6-31G*). A scaling factor of 0.9806 was used for the zero-point vibrational energy (ZPE) corrections for the B3LYP/6-31G* level (44). Spin contamination for all of the stationary point of the radical structures was negligible, i.e., 〈S2〉 = 0.75.</p><!><p>Tyrosyl hydroperoxide formation in the presence of glycine, phenylalanine, asparagine or tryptophan as well as basic amino acids such as arginine, lysine or histidine, was experimentally investigated. Figure 1 shows that, in general, the formation of hydroperoxide was enhanced in the presence of equimolar concentrations of an amino acid (1 mM) and tyrosine (1 mM) regardless of their pKa. Tyrosine (2 mM) alone in the absence of added amino acid resulted to hydroperoxide formation which is significantly higher compared to using other amino acids (2 mM) in the absence of tyrosine but did not exhibit the same degree of enhancement as 1 mM tyrosine with 1 mM of an amino acid (Figure 1 and Figure S1). Based on the concentrations of the amino acids used in the experiments, the amount of hydroperoxide formed from the mixture of amino acids is not additive of the hydroperoxides formed from individual amino acids, and this is further demonstrated in Figure S1 for tyrosine and arginine. These results also suggest that hydroperoxide formation can be influenced not only by the basic group of an amino acid but perhaps by any H-bond donor group originating from the amino acid moiety. The false positive results obtained from amino acids in the absence of tyrosine could arise from catalase, SOD or HRP in the mixture, but its nature is unclear at the moment and warrants further investigation. The enhanced formation of hydroperoxide in the presence of tyrosine as well as in the absence of SOD further indicate that the hydroperoxide originates from the reaction of TyrO• and O2•− (Figure 1). From the seminal work of Jin et al. (45), O2•− was shown to react with TyrO• to form the hydroperoxide adduct with a rate constant of 1.5 × 109 M−1 s−1 and its subsequent decay (t1/2 = 4.2 h) leads to the formation of the bicyclic compound 1. Evidence for hydroperoxide formation has also been previously reported (13, 15–17, 19, 45).</p><p> </p><p>Figure 2 shows that the amount of hydroperoxide formed increases with increasing concentration of tyrosine, acetaldehyde or arginine. The transition state structure of the hydroperoxide formed could involve participation of three major species, i.e., TyrO•, O2•− and an H-bond donor, since the amount of hydroperoxide formed is dependent on tyrosine, acetaldehyde and arginine concentration, respectively. Winterbourn, et al. (19) had proposed that the increased hydroperoxide formation in the presence of basic amino acid is due to the formation of a stable amino hydroperoxide via Michael addition reaction of the amino group to the tyrosyl peroxide intermediate.</p><p>In a separate study, we observed a similar phenomenon in which there is an increased favorability of O2•− addition to nitrones in the presence of amide substituent compared to methyl, ester or phosphoryl substituents (9). We later concluded that factors such as intramolecular H-bonding of O2•− in the transition state and electrostatic effects facilitate the O2•− adduct formation (10). This increased rate in hydroperoxide formation in nitrones bearing an amide moiety was suggested to be due to the α-effect (46) as shown by the presence of intramolecular H-bonding interaction between the amide hydrogen and the superoxide oxygen in the transition state. Alpha-effect is exhibited by a class of nucleophiles which have an electronegative atom (with one or more lone-pair of electrons) adjacent to the nucleophilic center called α-nucleophiles (46, 47). Alpha-nucleophiles tend to be very strong electron donors, and yet are very weak bases due to the inductive effect of the heteroatom adjacent to it. This high activity is due to the repulsive interactions between the unshared electron pairs on adjacent atoms between the lone-pair of electrons on the α-atom and those on the nucleophilic center. This makes an α-nucleophile unstable and hence more reactive (46).</p><p>We therefore hypothesized that the presence of H-bond donors, in general, may affect the rate of O2•− addition to TyrO• in which the O2•− H-bonding interaction with amino acids may modify its reactivity similar to that of hydroperoxyl radical (HO2•). Table 1 shows the spin and charge density distributions on the O2•− and HO2•, and their short O-O bond distances of 1.33 Å are characteristics of a pi-radical. By virtue of symmetry, the charge and spin density distribution on the two oxygen atoms of O2•− are equivalent, but addition of a proton to one of the oxygens to form HO2• perturbs the electronic and charge distribution between the two oxygen atoms resulting in higher spin density on the terminal oxygen atom (73%) compared to the oxygen atom (35%) bound to the hydrogen atom. This polarization in the charge and spin density distribution in HO2• can have significant effect on its reactivity as shown by the difference in the reduction potentials between O2•− and HO2• of Eo′ = 0.94 and 1.06 V, respectively, in which the latter is more oxidizing than the former (48).</p><p>Based on the premise that the O2•− interaction with an H-bond donor would perturb its electronic property, and hence, would assume a similar reactivity to that of HO2• (see Table 1 for the spin and charge densities of O2•− versus HO2•), we calculated the thermodynamics of O2•− complex formation with various amino acids at the PCM/B3LYP/6-31+G**//B3LYP/6-31G* level of theory. Figure 3 shows the various modes of intermolecular interaction of O2•− with arginine and lysine exhibiting strong H-bond interactions from the amino, carboxylic acid, guanidinium hydrogens (see Figure S2 for other amino acid-O2•− complexes). These H-bonding interactions result in significant perturbation of the spin and charge densities of the O2•− as shown in Table 1, except for His---O2•− and Arg---O2•− complexes (in which the charge densities were perturbed but not the spin densities) due perhaps to the competing H-bond interaction of the amino group with the carboxylic acid for the O2•−. Nevertheless, the low endoergicity of the free energies and exoergicity of the enthalpies of complexation of O2•− with amino acids may translate to a favorable complex formation in solution. The perturbation of the electronic properties of O2•− could be due to acid-base reaction since proton transfer reaction from the ammonium and carboxylic acid to O2•− was observed for Lys---O2•− and Gly---O2•− complexes, respectively. However, proton transfer was not observed for other amino acid---O2•− complexes but still shows the same perturbation in the electronic properties of O2•− suggesting that H-bond interaction plays a major role in altering the electron density distribution on O2•− (see Figure S2).</p><p>A similar complexation phenomenon was observed for 1-methyl-1-carbamoylcyclopentane (MCCP) and O2•− with a calculated ΔGrxn,298K,aq and ΔHrxn,298K,aq of −4.1 and 4.7 kcal/mol, respectively, and an experimental rate constant of 2.0 M−1 s−1 in DMF (10) which is far less than the rate constant observed for the tyrosyl radical dimer formation of 4.5 × 108 M−1 s−1 (45). Therefore, the rate of O2•− complexation may not be competitive towards TyrO• dimerization but upon consideration of the relative concentrations of the H-bond donating amino acids and non-amino acids used in this study versus the TyrO• concentration present in solution, the latter is expected to be in the orders of magnitude lower than the mM concentrations of the H-bond donors due to the short half-life of TyrO•. Moreover, based on the typical yield of the tyrosyl hydroperoxide (< 2 µM) and the rate constant of O2•− addition to TyrO• of 1.5 × 109 M−1 s−1 (45), the initial concentration of TyrO• formed is expected to be in the same concentration range as the tyrosyl hydroperoxide. Since the half life of TyrO• in ribonucleotide reductase is in days (49) the low rate constant for the -NH----O2•− complex formation can facilitate O2•− addition to TyrO• similar to the electrostatic guidance that Koppenol (50) had suggested to explain the efficiency of superoxide dismutase (SOD) selectivity for O2•− (51). Moreover, Fridovich (51) had pointed out that ionic strength effect (52, 53) and site-specific modification (54) can affect the specificity and activity of SOD, respectively. It was also previously shown (19) that there is an increase in tyrosyl hydroperoxide formation in peptides containing an amino group such as lysine that is adjacent to the tyrosyl group. The same enhancement in tyrosyl hydroperoxide formation was observed in the presence of free lysine or ethanolamine (19). It is therefore, reasonable to assume that -NH----O2•− complex formation may play an important role in the facilitation and selectivity of O2•− addition to TyrO• in protein systems especially when these H-bond donors are in close proximity to the TyrO• group.</p><p>Figure S3 of the Supporting Information shows a comparison of the energetics of the various modes of O2•− and HO2• addition to TyrO•, and results show that the ortho addition was the most exoergic with ΔGrxn,aq,298K of −6.8 kcal/mol (Figure 4) compared to the endoergic para and meta addition with ΔGrxn,aq,298K of 16.3 and 19.9 kcal/mol, respectively. Addition of HO2• however to TyrO• at the ortho and para positions are the most exoergic with ΔGrxn,aq,298K of −11.5 and −11.8 kcal/mol, respectively, with the meta addition being the least favorable with ΔGrxn,aq,298K of 19.1 kcal/mol. Ortho and para addition of O2•− to TyrO• have been previously proposed (45) but the former is more likely to be formed in solution based on their relative energetics. Figure 4 shows the spin and charge distribution on the TyrO• indicating that the carbon atom para to the amino acid group has the most positive charge (0.37 e) while the carbon to which the amino group is attached to has the highest spin density distribution of 0.35 e. Because the ortho carbon of TyrO• has relatively high spin density and negative charge compared to other carbon atoms in the aromatic ring (Figure 4), and that the preferred site of O2•− and HO2• addition to TyrO• is at the ortho position, suggest that the nature of O2•− and HO2• addition to TyrO• is electrophilic in nature.</p><p>As shown in Figure 1, the significantly higher hydroperoxide formed from 2 mM tyrosine alone compared to using 2 mM of an amino acid in the absence of tyrosine could be due to the direct addition of O2•− to TyrO• or HO2• to TyrO• according to Scheme 1, Reactions A and G-H, respectively. Proton abstraction by O2•− from the carboxylic acid moiety to give HO2• (Reaction G) has ΔGrxn,aq,298K of 0.0 kcal/mol and that direct addition of HO2• to the unprotonated (Reaction H) and protonated (Reaction F) TyrO• gave exoergic reaction energies of −6.8 and −11.5 kcal/mol, respectively. Since the hydroperoxide formation experiments were performed in neutral pH, and that the known pKa for HO2• is 4.8 (55) and 4.4 (56) and that of tyrosine is 2.2, the addition of HO2• to unprotonated TyrO• via Reactions G-H is more likely to occur than Reaction F. However, since Reactions G-H which involve initial acid-base reaction to form HO2•, hydroperoxide formation from this mechanism should be independent of the nature of the amino acid and should yield similar amounts of hydroperoxide from using 2 mM of tyrosine alone to that using 1 mM of tyrosine in the presence of added 1 mM amino acid. Since tyrosine in the presence of added amino acid yielded hydroperoxide that is significantly higher than tyrosine alone, Reaction A could be a more plausible mechanism and that the added amino acid plays a role in hydroperoxide formation most likely via α-effect as mentioned above. Other pathways for O2•− reaction to TyrO• was also considered and are shown in Scheme 1. The oxidation of O2•− by TyrO• via electron transfer mechanism (Reaction D) is exoergic with ΔGrxn,aq,298K = −5.7 kcal/mol compared to its endoergic reduction to peroxide (Reaction E) with ΔGrxn,aq,298K = 82.3 kcal/mol with the former only leading to the formation of phenoxide and oxygen. Although Reaction D is favorable, the formation of hydroperoxide is competitive enough to be observed experimentally.</p><!><p>To further verify the role of H-bonding on the formation of hydroperoxide, non-amino acids with a variety of H-donor groups were used. Compounds with -NH2 group (e.g., benzylamine, aniline, acetamide, and urea) as well as compounds with –COOH group (e.g., benzoic acid, phenylacetic acid) as potential H-bond donors were assessed for their ability to enhance hydroperoxide formation in the presence or absence of SOD, as well as in the absence of tyrosine. Similar to the observation made in the presence of added amino acid, an increase in tyrosyl hydroperoxide formation was observed compared to when these non-amino acids are not present as shown in Figure 6. Moreover, hydroperoxide formation is minimal in the absence of tyrosine. Compounds with –OH group such as ethanol and benzyl alcohol also exhibited enhancement of hydroperoxide formation but not to the same degree of enhancement as observed for the –NH2 and –COOH bearing compounds. Figure 6 therefore suggests that regardless of the nature of the H-bond donors, i.e., amine, amide, carboxylic acid, or alcohol, that the formation of hydroperoxide is dependent on the presence of H-bond donors.</p><p>The high concentration of tyrosyl hydroperoxide formed in the presence of H-bond donors led us to rationalized that H-bond interaction with O2•− can increase the amount of tyrosyl hydroperoxide formed due to the perturbation of O2•− electron density (Figure 6 and Figure 7). We therefore further focused our investigation on urea due to its biological relevance. Theoretical calculation at the PCM/B3LYP/6-31+G**//B3LYP/6-31G* level of theory was performed on the O2•− complex formed with urea, acetamide and N,N-dimethylacetamide to investigate if H-bonding will perturb the electronic property of O2•−. Figure S4 shows that O2•− complex formation via H-bonding with urea and acetamide is less endoergic (ΔGrxn,aq,298K = 5–;7 kcal/mol) compared to O2•− complex formation with N,N-dimethylated acetamide (ΔGrxn,aq,298K = 11.4 kcal/mol). Urea and acetamide can polarize the charge and spin density distribution of O2•− upon complexation provided that only one end of the O2•− is H-bonded. The same behavior was also predicted for benzyl alcohol and ethanol in which their complex formation with O2•− gave ΔGrxn,aq,298K = 7.2−7.3 kcal/mol (Figure S4) with polarization of the O2•− electronic distribution. However, there was negligible change in the spin and charge densities on O2•− and N,N-dimethyl acetamide complex which is supported by the experimentally observed low hydroperoxide formation in the presence of N,N-dimethyl acetamide (Figure 7). Hydroperoxide formation using non-H-bond donating compounds (Figure 7) is significantly lower relative to using H-bond donating compounds which further suggests that intermolecular H-bond interaction plays a critical role in the enhanced formation of tyrosyl hydroperoxide similar to that observed in the presence of amino acids.</p><!><p>In order to investigate the effect of urea on the thermodynamics of O2•− addition to TyrO•, the free energy of reaction of urea- O2•− complex to TyrO• was calculated and compared to the energetics of addition of O2•− to TyrO• in the absence of urea (Scheme 1, Reactions A versus C). Figure 5 shows that the most preferred mode of urea---O2•− addition is at the ortho position with ΔGrxn,aq,298K = −3.7 kcal/mol compared to the para- and meta- addition with ΔGrxn,aq,298K of −1.7 and 23.0 kcal/mol, respectively (Figure S5). The preference for urea---O2•− to add at the ortho position is similar to that predicted for O2•− addition to TyrO• alone. However, addition of urea---O2•− is exoergic by −3.7 kcal/mol but is less exoergic by 3.0 kcal/mol than the addition of O2•− alone, indicating that α-effect from the amino acid moiety may already have played a role in facilitating O2•− addition but experimental results show that 1 mM tyrosine in the presence of 1 mM urea gave significantly higher hydroperoxide yield compared to 2 mM tyrosine (Figure 7) suggesting that urea plays a role in facilitating hydroperoxide formation. A more obvious effect of urea can be seen on the urea---O2•− addition at the para position with a ΔGrxn,aq,298K of −1.7 compared to 16.3 kcal/mol in the absence of urea (Figure S5).</p><p>To further prove that the amino acid group plays a role in facilitating O2•− addition, the effect of amino acid group was eliminated by investigating the energetics of addition of O2•− versus urea---O2•− to the phenoxyl radical only (Figure 5 and Figure S6). The free energies of O2•− addition to PhO• at the para, meta and ortho positions are 1.8, 19.5, and 7.0 kcal/mol, respectively, while addition of urea---O2•−gave more favorable energies with ΔGrxn,aq,298K of 0.5, −15.5 and 3.5 kcal/mol for para, meta and ortho additions, respectively. Therefore, a more consistent trend in the energetics of O2•− addition versus urea---O2•− can be observed in the absence of amino acid moiety. Moreover, the thermodynamics of reaction of O2•− versus urea---O2•− to methylated-TyrO• (ArO•) are shown in Scheme 2, Reactions I and J, respectively. The O2•− addition to Tyr-3 and Tyr-4 (Reaction I) only yielded dioxetane products but not in the presence of urea (Reaction J) which suggests the need for H-bond donors in the preferential formation of hydroperoxide. In general, a more consistent trend in relative reactivity of methylated– TyrO• was observed similar to PhO• in which urea---O2•− gave less endoergic (or more exoergic in the case of Tyr-2) free energies of reaction to Tyr-1, Tyr-3 and Tyr-4 relative to O2•− alone. These results further suggest that urea as well as carboxylic acid group can play a significant role in the facilitation of O2•− addition via α-effect.</p><p>Figure 8 shows the increase in hydroperoxide formation as a function of urea concentration consistent to that observed using arginine as H-bond donor (see Figure 2), while Figure 9 demonstrates the inverse relationship of the time of addition of urea on hydroperoxide formation. Although the yield of hydroperoxide with 1 mM tyrosine alone in Figure 8 is relatively lower, i.e ∼ 0.3 µM, compared to ∼ 0.6 µM using 1 mM of tyrosine as shown in Figure 2c, the plot shows a qualitative trend of the dependence of the amount of hydroperoxide formed as a function of urea concentration. The dependence of hydroperoxide formation on the time of urea addition as well as on the concentration of urea further supports that urea can have a stabilizing effect on the hydroperoxide adduct or that the α-effect can enhance hydroperoxide formation. To shed more insights on to whether the enhanced production of hydroperoxide is mainly due to α-effect from urea, and/or stabilization of the TyrO-O2H adduct by urea, the thermodynamics of urea addition to TyrO-O2H was investigated. The addition of urea to TyrO-O2H (Scheme 2, Reaction L), is less favored with ΔGrxn,aq,298K = 13.0 kcal/mol compared to the addition of urea---O2•− to TyrO• with ΔGrxn,aq,298K = −3.7 kcal/mol (Scheme 2, Reaction J). Given the endoergicity of urea addition to TyrO-O2H (Reaction L) versus urea---O2•− to TyrO• (Reaction J), the latter is the most preferred pathway for urea-hydroperoxide complex formation. This trend is also true for all the methylated-TyrO• analogues (see Scheme 2, Reaction J versus L). Based on the energetics of tyrosyl hydroperoxide formation presented thus far, it is therefore reasonable to assume that the hydroperoxide formed can be facilitated by H-bond donor via α-effect.</p><!><p>The role of α-effect on the enhancement of tyrosyl hydroperoxide formation has been experimentally and computationally explored. Superoxide exhibits various modes of H-bond interaction with carboxylic acid, amine and amide groups of amino acids as well as non-amino acids. Initial H-bonding of carboxylic acid and/or amine groups to O2•− causes polarization of the spin and charge density distribution around the O2•− resulting to a more favorable O2•− addition reaction with PhO• or TyrO• as shown by the less positive (or more negative) free energies of this reaction compared to in the absence of H-bond donors. Addition of O2•− to TyrO• at the ortho-position is the most thermodynamically preferred mode of addition in the presence or absence of H-bond donor such as urea. The role of urea in either stabilizing the tyrosyl hydroperoxide adduct via H-bonding, or enhancing hydroperoxide formation through α-effect was computationally investigated using TyrO• and its methylated analogues and results indicate that the latter is the more preferred mechanism. Therefore, in enzyme systems containing TyrO•, the presence of acidic moieties or H-bond donors at close proximity to TyrO• may lead to the facilitation of hydroperoxide formation and hence potential enzyme inactivation. Site-directed mutagenesis may be able to shed more insights in to this phenomenon and merits further investigation. Moreover, the presence of exogenous H-bond donors such as urea or alcohols in elevated concentrations may induce hydroperoxide formation, and therefore, can play a pro-oxidant role in the initiation of oxidative stress in biological systems.</p><p>The models used in this study have provided new insights into the understanding of the molecular mechanism of oxidative damage as mediated by strong interaction of O2•− with H-bond donors. The implication of α-effect in hydroperoxide formation in various enzyme systems (23) containing protein radicals such as TyrO• needs further investigation since there has been growing evidence on the pro-oxidant role of urea and its derivatives in the initiation of oxidative stress and damage in biological systems such as the inactivation of the R2 subunit of ribonucleotide reductase by hydroxyurea (20). Furthermore, increased production of reactive oxygen species, DNA damage, protein carbonylation levels, or increased expression of the oxidative stress-responsive transcription factor, Gadd153/CHOP, at the mRNA and protein levels in cultured renal cells have been observed during hyperosmolality caused by elevated concentrations of NaCl and urea, or during urea treatment (57). Further studies is also necessary on the potential implication of this hydroperoxide formation from tryptophan radical found in other enzyme systems (23).</p><!><p>Cartesian coordinates for all the compounds are available as supporting information and are available free of charge at http://pubs.acs.org.</p>

PubMed Author Manuscript

Patterning Graphene Film by Magnetic-assisted UV Ozonation

Developing an alternative method for fabricating microscale graphene patterns that overcomes the obstacles of organic contamination, linewidth resolution, and substrate damaging is paramount for applications in optoelectronics. Here we propose to pattern chemical vapor deposition grown graphene film through a stencil mask by magnetic-assisted ultraviolet (UV) ozonation under irradiation of a xenon excimer lamp. In this process, the paramagnetic oxygen molecules and photochemically generated oxygen radicals are magnetized and attracted in an inhomogenous external magnetic field. As a consequence, their random motions convert into directional, which can greatly modify or enhance the quality of graphene patterns. Using a ferromagnetic steel mask, an approximately vertical magneticfield-assisted UV ozonation (B Z = 0.31 T, ∇B Z = 90 T • m −1 ) has a capability of patterning graphene microstructures with a line width of 29 μm and lateral under-oxidation less than 4 μm. Our approach is applicable to patterning graphene field-effect transistor arrays, and it can be a promising solution toward resist-free, substrate non-damaging, and cost effective microscale patterning of graphene film.Patterning graphene film is a significant step in fabricating graphene-based elements for both fundamental studies and industrial applications [1][2][3][4][5][6][7][8][9][10][11][12][13][14][15][16] . Besides taking direct bottom-up fabrication routes 1,2 , various top-down etching solutions are used to cut graphene film into certain patterns [3][4][5][6][7][8][9][10][11][12][13][14][15][16] . Electron-beam lithography and photolithography techniques are widely used for their high resolution capability to pattern micro-/nanostructures. However, they both incur resist contamination, and as a consequence inevitably degrade graphene quality after the lithography process [3][4][5][6][7] . In order to circumvent this obstacle, direct focused ion beam and laser writing techniques are employed [8][9][10][11][12] . As for the focused ion beam, it has a capability of patterning graphene with nanometer scale resolution. However, coexistence of high expense, low productivity, and damage to the supporting substrate limits its applications 8,9 . As a contrast, direct laser writing is popularly used to pattern large-area chemical vapor deposition (CVD) grown graphene film due to its high productivity and resist-free characteristics, even though inevitable drawbacks like coarse edges and serious damage to the supporting substrates still exist [10][11][12] .In order to overcome these problems in graphene patterning, some alternative solutions have been proposed in the past few years, like TiO 2 -based photocatalysis 13 , resist-free reactive ion etching (RIE) with oxygen and argon plasmas 14,15 , and UV ozonation 16 . Weak oxidation of TiO 2 -based photocatalysis and the subsequent complicating disposals prevent its development 13 . As for RIE plasma etching technique, the positively charged ions are electrically accelerated to acquire a directional motion toward substrate, in which way the quality of graphene patterning is improved 17 . Even so, severe lateral under-oxidation up to ten micrometers was induced for graphene underneath the mask due to diffusion of those highly dynamic gaseous reactants 14,15 . UV ozonation, a kind of mild oxidation compared with oxygen plasma, showed to be too weak to pattern CVD grown graphene film in the previous study, though it could cut graphene oxide into 2-μ m-wide strips when assisted with the ultrasonic wave treatment and high-temperature annealing 16 . Even for a high-temperature enhanced UV ozonation 18 , seeking a solution to making the electrically neutral oxygen radicals move directionally as that of the positively charged ions in RIE process, would be crucial to attain high-quality graphene patterning.In this study, we propose to pattern CVD grown graphene film through stencil masks by magnetic-assisted UV ozonation. An external inhomogenous magnetic field can magnetize the paramagnetic oxygen molecules and photochemically dissociated oxygen radicals, and induce strong attractive magnetic forces [19][20][21][22][23][24] . As a

patterning_graphene_film_by_magnetic-assisted_uv_ozonation

<!>Results<!>Working mechanism of magnetic-assisted UV ozonation.<!>Patterning graphene through a sapphire mask by a vertical magnetic-field-assisted UV ozonation.<!>Discussion<!>Summary<!>Methods

<p>consequence, random motions of the oxygen molecules and radicals turn into directional, and it can tremendously enhance the quality of graphene patterning when a vertical magnetic field (relative to graphene film) is applied. This magnetic-induced directional and enhanced photochemical reaction is consistent with the report on macroscopic deflection of the flow of oxygen gas when they were put in an inhomogeneous magnetic field (B = 1 T, ∇ B = 100 T • m −1 ) 20,21 . Using a ferromagnetic steel mask, the vertical magnetic-field-assisted UV ozonation (B Z = 0.31 T, ∇ B Z = 90 T • m −1 ) has a capability of patterning 29-μ m-wide conformal graphene microstructures with the lateral under-oxidation less than four micrometers, and this approach is applicable to patterning graphene field-effect transistor (FET) arrays.</p><!><p>Strategy for patterning graphene by magnetic-assisted UV ozonation. A home-designed UV ozonation vacuum machine is used for graphene patterning with a xenon excimer lamp installed on top inside the chamber (Supplementary Figure 1). The electric power supplied for the UV source is fixed at 1.5 kW. The distance between the UV lamp and graphene film (i.e., working distance) is fixed at 20 mm. To acquire a strong photochemical oxidation, the chamber is filled with a small fraction (10 Pa) of oxygen gas in a nitrogen (N 2 ) atmosphere with the total pressure of 1 atm. A cube permanent magnet is placed underneath the SiO 2 /Si substrate in order to exert a vertical inhomogenous magnetic field (Supplementary Figure 2a).</p><p>Graphene patterns are formed through different stencil masks at room temperature after four cycles of consecutive UV ozonation treatments (10 min/cycle, with the same initial oxidation parameters). Nonmagnetic sapphire mask, and magnetic masks made of nickel grid (the commercial 400 mesh for transmission electron microscope) and molybdenum doped steel grid are used individually for patterning graphene film. The magnetization of nickel mask is along its surface when put in a magnetic field, and as a result it cannot be used to pattern graphene film in a vertical magnetic-field-assisted UV ozonation (see Supplementary Figure 2a). However, an approximately horizontal magnetic field provided by a stack of bar magnets can make it stable (see Supplementary Figure 2b). The steel mask is magnetized along a direction perpendicular to its surface. As a consequence, its contact with graphene film can be improved due to strong magnetic attraction close to one pole of the cube magnet (see Supplementary Figure 2c).</p><!><p>UV ozonation is a type of photochemical oxidation characterized by random etching of the oxidative reactants 16,18,[25][26][27][28] . As schematically shown in Fig. 1, oxygen molecules are primarily dissociated to ground-state O( 3 P) atoms, or called oxygen radicals under irradiation of the xenon excimer lamp centered at 172 nm. Then, these oxygen radicals oxidize graphene into carbon monoxide (CO) and carbon dioxide (CO 2 ) molecules, or they combine with oxygen molecules to form instable ozone (O 3 ). As confirmed by the X-ray photoelectron spectroscopy analyses (Supplementary Figure 3 and Supplementary Note 1), no extra contamination is introduced to graphene except for the adsorbed carbonyl (-C= O) and epoxide (C-O) oxygen functional groups 18,28 .</p><p>In the UV ozonation process, oxygen molecule behaves as a strong paramagnetic substance, while the reactive products (O 3 , CO and CO 2 molecules) and the prefilled gaseous nitrogen behave as weak diamagnetic substances [29][30][31][32] . The magnetization capability of all these gases is characterized by the molar magnetic susceptibility (20 °C, 1 atm) as listed in Table 1, where we can see that the magnetic susceptibility of oxygen molecule is two orders higher than those of the diamagnetic molecule gases [29][30][31][32] . For comparison, the referred volume magnetic susceptibility for oxygen and nitrogen molecules are converted into molar susceptibility as elaborated in detail in Supplementary Note 2 23,29 .</p><p>When an inhomogenous magnetic field (B Z = 0.31 T, ∇ B Z = 90 T • m −1 ) is applied to UV ozonation, the paramagnetic oxygen molecule and radical, which have the magnetic moment μ B equal to 2.0 and 1.67 Bohr magnetons, respectively, are magnetized as shown in Fig. 1 23,24 . The magnetic forces F Z exerted on one oxygen molecule and radical is deduced to be both in the order of 10 −22 N toward graphene 24 from the formula F Z = g J μ B ∇ B Z , where g J is the Landé g-factor taking values between 1 and 2. The paramagnetic-induced attractive magnetic force converts the random motions of oxygen molecules and radicals into directional ones as demonstrated by the longer orange arrows pointing to graphene film. On the contrary, all the diamagnetic molecules (N 2 , O 3 , CO 2 , CO) experience weak repulsive magnetic forces away from graphene surface due to their two-order smaller negative magnetic susceptibilities 21,[29][30][31][32] , which may further facilitate ongoing of the photochemical reaction. As a consequence, UV ozonation turns into directional, and in principle it can improve the quality of graphene patterning.</p><p>In the above discussion, we have simplified the theoretical model by assuming that all these gaseous substances work at 1 atm at room temperature in the magnetic-assisted UV ozonation. As a matter of fact, except for nitrogen gas, all other gaseous substances, including the paramagnetic oxygen molecules and radicals (10 Pa or less), have much lower partial pressures. As a consequence, it may further enhance directionality due to decreasing probability of intermolecular collisions adjacent to graphene surface in UV ozonation.</p><p>Patterning graphene through a nickel mask by UV ozonation. UV ozonation is used to pattern graphene film through a ten-micrometer-thick nickel mask without applying any magnetic field. The mask has a honeycomb structure of hexagonal holes with the lattice constant of 62 μ m and rib width of 26 μ m as shown in the SEM image of Fig. 2a. The patterned holes deform into circular and appear over-etched, especially in the region away from corners when compared with the mask profile as partially outlined by the white hexagons in Fig. 2b. The minimum rib width connecting two adjacent holes decreases to 21 ± 1 μ m, a few micrometers narrower than that of the mask. Raman spectroscopy is used to evaluate the quality of graphene microstructures 14,33 . Figure 2c shows the defect band (D band) map in the region denoted by the red rectangle in Fig. 2b. From the color variation, we can see the D band shows up in the region along graphene edges. The corresponding Raman spectrum evolution for the green-outlined dots (Fig. 2d) shows that the lateral under-oxidation across the graphene edges is 4-5 μ m. Herein, we use a mechanically exfoliated high-quality monolayer graphene, which is testified to be free of defect bands in the edge area, as reference (see Supplementary Figure 4). A horizontal magnetic-field-assisted UV ozonation (B Y = 40 mT, ∇ B OY = 2 T • m −1 ) can completely modify the graphene pattern even though the same nickel mask is used (see Supplementary Figure 5 and Supplementary Note 3 for more information). The results indicate that the paramagnetic oxygen radicals and molecules turn into directional motions in the horizontal magnetic field, and as a result it makes the photochemical oxidation directional. In spite of the unwanted lateral under-oxidation, a properly designed and well controlled magnetic field may provide a solution to intentionally modifying graphene patterns.</p><!><p>Using a non-magnetic 316-μ m-thick sapphire mask shown in Fig. 3a, a vertical magnetic-field-induced directional photochemical reaction (B Z = 0.31 T, ∇ B Z = 90 T • m −1 ) can be intuitively demonstrated when patterning graphene by UV ozonation. Comparing the optical images in Fig. 3b and c, we can see that most of the multilayer graphene nucleuses disappear in the structure patterned without assistance of the vertical magnetic field. Further micro-Raman maps of the D band in Fig. 3d and its corresponding Raman spectrum evolution Fig. 3e indicate that the nonmagnetic-assisted UV ozonation induces severe lateral under-oxidation throughout the graphene pattern. This phenomenon stems from UV penetration through the transparent sapphire mask and the subsequent photochemical reaction propelled by diffusion of ozone, oxygen radicals and molecules. When the vertical magnetic field is applied, the lateral under-oxidation decreases to 11 μ m as indicated by Raman spectrum characterizations in Fig. 3f and g. For the nonmagnetic sapphire mask, its contact with graphene is independent of the vertical magnetic field. Therefore, it is the magnetic field that induces the directional motions of oxygen molecules and radicals toward graphene film, and then reduces their lateral diffusion underneath the mask 19,21 .</p><p>Patterning graphene through a steel mask by a vertical magnetic-field-assisted UV ozonation.</p><p>Using a magnetic steel mask, the quality of graphene patterning can be improved in the vertical magnetic-field-assisted UV ozonation (B Z = 0.31 T, ∇ B Z = 90 T • m −1 ). Figure 4a shows the optical image of a steel mask composed of a hexagonal lattice of holes with the constant of 220 μ m and rib width (at surface) of 29 ± 2 μ m. As indicated by the high-resolution SEM image in the inset, these holes have rough sidewalls with protrusions fluctuating within four micrometers. The etched graphene pattern conforms well to the mask profile as shown by the optical image in Fig. 4b. Further high-resolution SEM imaging (Fig. 4c) indicates that there still exist some bright micro/nanofilaments along edges, similar to but sparsely distributed compared to those in Fig. 2b. These residues originate from wrinkles formed during CVD growth and the following transfer of graphene film (see Supplementary Figure 6 for more information) 34,35 . Analyses of the Raman map of D band (Fig. 4d) and the corresponding Raman spectrum evolution (Fig. 4e) indicate that the lateral under-oxidation aroused by dissipation of oxygen radicals underneath the mask is 3-4 μ m, decreased compared to that using a sapphire mask. The gas-diffusion induced graphene oxidation could be related to a combination of the non-ideal vertical distribution of the magnetic field and the imperfect contact between the mask and graphene.</p><p>When a weak magnetic field (B Z = 19 mT, ∇ B Z = 4.5 T/m) is applied, the quality of patterned graphene microstructure deteriorates rapidly after the same UV ozonation treatment (see Supplementary Figure 6). Further, when no magnetic field is applied, such pattern cannot form due to severe diffusion and dissipation of oxidative reactants underneath the mask.</p><p>The success of graphene patterning by magnetic-assisted UV ozonation is attributed to two major factors. Firstly, the directional motion of oxygen molecules and radicals toward graphene surface can enhance directionality of the photochemical etching process. Secondly, magnetic-induced contact improvement between the stencil mask and graphene film is critical for attaining high quality graphene patterning. Besides the directional oxygen radicals and molecules, the photo-generated weak diamagnetic ozone molecules, which is instable and subject to decomposing into oxygen radicals and molecules, may diffuse underneath the mask and then contribute to the lateral under-oxidation. This opinion is supported by the feasibility of patterning graphene film through a high-quality nickel mask in Fig. 2 when no magnetic field is applied.</p><p>Table 2 summaries the traits of patterning graphene film by UV/ozonation when different masks and magnetic fields are used. As can be seen, the best graphene patterning can be obtained when a magnetic steel mask is used in the strong vertical magnetic-field-assisted UV ozonation.</p><p>Patterning graphene FET arrays by the vertical magnetic-field-assisted UV ozonation. The capability of patterning highly improved graphene microstructures for the vertical magnetic-field-assisted UV ozonation (B Z = 0.31 T, ∇ B Z = 90 T • m −1 ) allows it to fabricate graphene electronic circuits (see Method and Supplementary Figure 7). Figure 5a demonstrates a graphene FET array that consists of three devices with the same channel width of 70 μ m and two different channel lengths of 390 μ m (named as S1 and S2) and 586 μ m (named as L1). The contacts are made of Cr/Au film (5/90 nm thick). Consistent with that in Fig. 4, the lateral under-oxidation remains 3-4 μ m as indicated by micro-Raman measurement (see Supplementary Figure 8). Before electrical measurement, the neutral (Dirac) points are shifted close to zero in a high vacuum of 4.5 × 10 −4 Pa under in situ illumination at 254 nm from a low-pressure mercury lamp in a double-chamber UV ozonation machine 18,36 . The relationship between source-drain current and the applied bias voltage V SD is linear for all graphene FET elements at different back-gate biases ranging from − 45 to + 45 V. Figure 5b demonstrates the linear dependence for L1 device, which confirms the ohmic contact between Cr/Au electrodes and the underneath graphene film.</p><p>At a fixed source-drain voltage (V SD = 0.1 V), the conductivity curves varying with the gate bias for these three FETs are shown in Fig. 5c. The transfer length method (TLM), which is used for attaining accurate density dependent mobility and contact resistance at relatively high carrier density, is not applicable to the hundreds-of-micrometers-long non-uniform CVD graphene devices 37 . Using a fitting method proposed by Kim, we obtain the highest hole and electron mobilities of ~1682 cm 2 1,2 , for the S2 graphene FET device. For the other S1 and L1 devices, the conduction decreases and their hole mobilities are lower than electron mobilities 33,38 . Our extra measurements further confirm that the magnetic-assisted UV ozonation is not the only element that determines the conductivity and electron-hole asymmetry conduction. A combination of the neutrality point misalignment caused by non-uniformity due to randomly distributed wrinkles, cracks, multilayer nucleuses, and contamination in the CVD grown graphene film can all contribute to such variation for each individual FET device 34,35,[39][40][41] . When lacking a vertical magnetic field, the above discussed 70-μ m-wide graphene FET array cannot be successfully patterned by UV ozonation due to severe dissipation of oxidative reactants underneath the mask. Instead, using a 168-μ m-wide steel mask, we can only obtain 129-μ m-wide channels (Supplementary Figure 9a). Meanwhile, the lateral under-oxidation across graphene channel reaches 40 μ m (Supplementary Figure 9b). Further electrical measurements (Supplementary Figure 9c,d) show that the electrical current varies linearly with the source-drain voltage under different back-gate biases (from − 60 V to 60 V), and that its conductivity degrades compared to those of the FET devices in Fig. 5c. This electrical deterioration mainly stems from severe lateral under-oxidation across channels by the randomly moving oxidative reactants when lacking a vertical magnetic field.</p><!><p>Compared with the laser writing or reactive ion etching (RIE), the magnetic-assisted UV ozonation has the following characteristics [10][11][12]14,15 . Firstly, the unique directional photochemical etching mechanism explains the feasibility of highly improved graphene patterning by the vertical magnetic-field-assisted UV ozonation. Meanwhile, no observable damage is induced to the substrate in the photochemical process. Increasing the magnetic field and its gradient can further enhance the dynamic energies of oxygen radicals and molecules, and its impact on the substrate and the quality of patterned graphene needs to be explored. Secondly, unlike the direct laser writing, its etching productivity does not depend on the area that needs to be etched off since the patterning is a photochemical reaction. Thirdly, it can be applicable to patterning high-quality large-area graphene film provided that the most critical factor, the magnetic field, can be scaled up and well controlled. As known, the other two critical factors, the stencil mask and the fourteen-inch-long xenon excimer lamp, can be readily scaled up.</p><p>The magnetic-assisted UV ozonation approach manifests good sample-to-sample consistency and reproducibility for patterning graphene microstrustures. For seeking applications in the field of nanotechnology, it is important to explore the minimum line width that UV ozonation can realize to pattern graphene film. Put aside the quality of graphene film, a high-quality magnetic mask etched with micro/nano-structures may provide a solution to further improving graphene patterning. Besides, a precisely designed and controlled external magnetic field would facilitate improving (such as eliminating wrinkle-incurred residues and diffusion-induced lateral under-oxidation) or intentionally modifying graphene patterns.</p><!><p>In summary, we have proposed and demonstrated a new strategy to pattern CVD grown graphene film by magnetic-assisted UV ozonation. By virtue of the paramagnetic property of oxygen molecules/radicals, we can pattern 29-μ m-wide graphene microstructure with the lateral under-oxidation less than four micrometers under irradiation of a xenon excimer lamp. The vertical magnetic-field-assisted UV ozonation approach is applicable to patterning graphene FET arrays, and it provides a resist-free, substrate non-damaging, and cost-effective solution to microscale graphene patterning.</p><!><p>Preparation of CVD graphene film on SiO 2 /Si substrate. A "PMMA-mediated" approach was used to transfer CVD grown monolayer graphene on a copper foil onto the thermally grown SiO 2 film (300 nm) on a highly doped p-type silicon substrate (0.001~0.004 ohm • cm) as follows 33,39,40 . Firstly, a 200 nm thick PMMA 950 A5 layer was spin-coated on the graphene/copper substrate and then baked for 2 min at 160 °C. Secondly, we removed the copper foil in an etchant of 0.5 M FeCl 3 aqueous solution after 3 h immersion and then obtained the PMMA/graphene stack layer. Thirdly, the stack was etched by dipping in H 2 O/H 2 O 2 /HCl (20:1:1) and H 2 O/ H 2 O 2 /NH 4 OH (20:1:1) solutions successively to remove possible metal residues. After each etching, it was rinsed sufficiently by deionized water and then scooped out onto a clean SiO 2 /Si substrate. Monolayer graphene, predominantly monolayer with randomly distributed multilayer flakes less than 5%, was finally obtained by solving the PMMA in acetone. In order to remove possible organic residues and enhance its contact with the SiO 2 /Si substrate, an extra disposal of annealing in a flow of gas mixture (Ar:H 2 = 200 sccm:100 sccm) at 290 °C was carried out for three hours.</p><p>Characterizations and electronic measurements. An optical microscope (Leica DM 4000) was used for morphology imaging of the patterned graphene microstructures. A scanning electron microscope (SEM, Zeiss Ultra Plus) under 5 kV and 3 kV biases was used to obtain highly resolved topographical images of masks and graphene patterns, respectively. A confocal micro-Raman spectroscopy (Senterra R200-L) was used to map graphene patterns under excitation of 532 nm laser (50X objective, ~1.2-μ m spot size) with the scanning step size of 1 μ m. Relative to the sample positioning platform, there exists a shift of ~3-μ m upward and ~0.5-μ m rightward for the laser positioning system. The ex situ XPS spectra were collected using a Kratos Axis Ultra DLD spectrometer (equipped with a monochromatic Al Kα X-ray source) with the anode power of 150 W. A gaussmeter was used to measure the strength and direction of a magnetic field adjacent to graphene surface. Its gradient was approximated by the ratio of the magnetic difference to a certain distance within one millimeter. All the electrical measurements were carried out in a high vacuum chamber (4.5 × 10 −4 Pa) with a combination of Keithley 6430 and 2400 systems.</p>

Scientific Reports - Nature

Sponge-like nanostructured conducting polymers for electrically controlled drug release

An electrically controlled drug release (ECDR) system based on sponge-like nanostructured conducting polymer (CP) polypyrrole (PPy) film was developed. The nanostructured PPy film was composed of template-synthesized nanoporous PPy covered with a thin protective PPy layer. The proposed controlled release system can load drug molecules in the polymer backbones and inside the nanoholes respectively. Electrical stimulation can release drugs from both the polymer backbones and the nanoholes, which significantly improves the drug load and release efficiency. Furthermore, with one drug incorporated in the polymer backbone during electrochemical polymerization, the nanoholes inside the polymer can act as containers to store a different drug, and simultaneous electrically triggered release of different drugs can be realized with this system.

sponge-like_nanostructured_conducting_polymers_for_electrically_controlled_drug_release

1. Introduction<!>2.1. Chemicals<!>2.2. Apparatus<!>2.3. Preparation of nanostructured PPy films loaded with drugs<!>2.4. Electrically controlled drug release<!>3.1. Morphology characterization of the nanostructured film<!>3.2.1. Effect of Potential Amplitude<!>3.2.2. Drug Release Comparison of Different Systems<!>3.2.3. Simultaneous Release of Different Drugs<!>4. Conclusion<!>

<p>For the creation of "smart" implantable devices that can monitor health status and provide therapeutic treatment, a key step is to build controllable drug release devices that can be miniaturized1–4. Drug release systems based on CPs are excellent potential candidates, as CPs can be flexibly synthesized on conductive surfaces even at the nanoscale5–7, and their inherent electroactivity provide the mechanism for electrically triggered molecule release8–11. In the past, different molecules have been incorporated into CP films and then released upon electrical stimulation12–15. In these cases, drug release is based on the electrical switching of the polymer redox states, accompanied by the movement of dopant ions in and out of the material16. Several limitations have prevented the broad use of CPs in controlled release systems. For instance, the drug molecules to be delivered were incorporated into the bulk of the CPs, which has limited capacity for drug loading. In addition, the range of drugs was restricted due to the charge and size requirements of dopant molecules.</p><p>In our previous work, we have significantly improved the drug release efficiency of a CP controlled release system by template-synthesizing high surface area nanoporous polypyrrole17. Here we report a further improved ECDR system based on sponge-like nanostructured PPy. With this system, drugs can be loaded during the PPy polymerization into the bulk, and extra drug or a different drug, not necessarily being a dopant, can be loaded into the nanopores inside the polymer film and sealed by a thin layer of polypyrrole on top. This design provides a CP based drug release system with higher loading capacity for a wider range of drugs.</p><!><p>Pyrrole (98%), Fluorescein (Flu) sodium salt and Dexamethasone (Dex) 21-phosphate disodium salt were purchased from Sigma-Aldrich, and pyrrole was distilled under vacuum and stored frozen. Polystyrene (PS) nanobeads (mean diameter 46±2.0 nm) were obtained from Duke Scientific Corporation. Milli-Q water from a Millipore Q water purification system was used throughout.</p><!><p>Electrochemical experiments were performed on a Gamry Potentiostat, FAS2/Femtostat (Gamry Instruments) using a three-electrode system, with a glassy carbon (GC, diameter of 3.0 mm) electrode as the working electrode, a platinum wire as the counter electrode, and a Ag/AgCl electrode as reference electrode. SEM was performed with an XL30 SEM instrument (FEI Company). The concentration of Dex and Flu solutions were measured with a microplate reader SpectraMax M5 (Molecular Devices). The ultraviolet absorption of Dex was measured at 242 nm, and the fluorescence of Flu was measured at the excitation and emission wavelength of 405 and 538 nm, respectively.</p><!><p>The ECDR system was constructed by forming a thin PPy layer on top of the nanoporous PPy film, as shown in Scheme 1. Nanoporous PPy films loaded with Flu were prepared using a template method as previously described17. For comparison, conventional PPy films incorporated with Flu were synthesized similarly, but without the PS templates.</p><p>The prepared nanoporous PPy modified electrode was soaked in ethanol and water alternatively for several times to make the nanoporous film more hydrophilic, and 5.0 µL solution containing 0.01 M Flu was then dropped onto the electrode surface and dried in air. The added drugs can penetrate throughout the hydrophilized nanoporous PPy films, as the nanoholes resulted from closely packed PS nanobeads are interconnected18. To prevent the stored drugs from leaking, a protective thin PPy layer was electropolymerized on top of the nanoporous PPy. The electrode was immersed in the solution containing 0.2 M pyrrole and 0.01 M Flu, and electropolymerization was carried out immediately by cycling the potential from 0.5 V to 1.2 V at the scan rate of 50 mV s−1 for one cycle, which was optimized to form a uniform thin PPy layer.</p><p>A similar method was used for the preparation of the binary drug release system, in which Dex (0.02 M, 5.0 µL) was added instead of Flu. The resulted drug release system is denoted as NPPy-Flu/Dex/PPy-Flu.</p><!><p>The ECDR was carried out in an electrochemical cell containing 1.6 mL 0.01 M PBS (pH 7.4). Before the drug release, the electrodes were immersed in 30 mL 0.01 M PBS under magnetic stirring for 15 min to remove any physically adsorbed drugs. The electrical stimulus for drug release was voltage application of −2.0 V or −0.5 V for 5s followed by 5 second of 0V period (for PPy recovery). Such stimulus was found to be most efficient in driving the drug out while minimizing the electrochemical damage to the polymer film.</p><!><p>SEM images of the PS nanobead template and the nanoporous PPy film are shown in Figure 1a and 1b. It is clear that the template is composed of tightly packed PS nanobeads, and the sponge-like PPy film is composed of numerous nanoholes resulting from the removal of the PS nanobeads.</p><p>Figure 1c shows the subsequently polymerized PPy protective layer in which the top layer nanoholes of the porous PPy film were clearly covered. Similar results have also been reported by Bajpai et al.19, and they found that the opened mouths of PPy micro-containers could be sealed through further polymerization of pyrrole. As can be seen from the cross-section image (Figure 1d), a thin layer of dense PPy lays on top of the nanoporous structure. This image verified that the second PPy thin film layer was formed on top of the nanoporous PPy film, and the nanoporous morphology of the bottom layer remained unchanged.</p><!><p>The effect of the potential amplitude was studied. As shown in Figure 2a, with each stimulus, the drug release at −2.0 V was significantly more than at −0.5 V. The drug release is indeed electrically controlled and the amount released is dependent on the strength of the stimulation. Interestingly, within the 30 stimuli tested, the amount of released drug after each stimulus decreases rapidly when the stimulation potential is high, while it was considerably steady at the lower potential. Therefore, higher potential can be chosen to release drugs more rapidly and lower potential allows more linear and steady drug release over a longer period of time.</p><!><p>The drug release of different systems was tested (Figure 2b) to investigate the mechanisms of drug loading and release. Comparing the amount of drug released from the nanoporous PPy film (NPPy) and the NPPy with added drug (NPPy/Drug), there was no significant increase in drug release. When a protective PPy layer was formed (NPPy/Drug/PPy), the amount of released drug was more than doubled. Two possibilities could lead to this effect. First, the additional layer of PPy may have contained additional drug thereby increasing the drug release. Second, the top layer of PPy helps to prevent the added drug from leaking out. The first possibility has been proven to be unlikely because adding a layer of PPy without the intermediate step of loading extra drug did not result in obvious increased drug release. Therefore, the main reason for the enhanced drug release is that the nanopores in the polymer film acted as containers to hold more drugs while the protective layer on top prevented the leakage. The nanoporous morphology is critical for extra drug load as adding extra drug and a protective layer to the conventional PPy film did not result in any increase in drug release. Without the nanopores, drug dropped on surface was easily washed out during the next step of polymerization.</p><p>In past studies, drug release from CPs is primarily based on the electrochemical reduction and dedoping process16. Recent efforts have been taken to design release systems utilizing the actuation effect of CPs, which relies on volume changes of the polymer matrix associated with the movement of ions and water under electrochemical switching20–24. For example, Abidian et al.25 have reported the release of Dex from poly(3,4-ethylenedioxythiophene) nanotubes based on actuation. In our system, we hypothesize that both mechanisms of release may be at play. Upon electrical stimulation, drug molecules incorporated in the backbone as dopants are released via the dedoping process, while those loaded in the nanopores may be squeezed out due to the actuation of the sponge-like PPy film. Although the protective layer was able to prevent the passive drug diffusion, the squeezing pressure and the shrink of the protective layer itself might allow drug solution to leach out.</p><!><p>As the nanoholes inside the nanoporous PPy film were used as nanocontainers, the porous film could theoretically be loaded with any drugs. Simply adding Dex solution to the nanoholes instead of Flu solution, a binary drug release system (NPPy-Flu/Dex/PPy-Flu) was constructed. As shown in Figure 2c, both Flu and Dex were released after electrical stimulation, and there was negligible diffusion of Flu, while the diffusion of Dex was not detectable. This experiment result is interesting in three ways. First, we verified that the drug added after nanoporous film synthesis was indeed loaded and released via electrical stimulation. Secondly, we can differentiate the amount of drug released from the PPy film as dopants vs. those stored in the pores. Thirdly, we have demonstrated the proof of concept in delivering two different drugs simultaneously.</p><!><p>A controlled release system based on sponge-like nanostructured PPy film is developed, which can load drugs not only within the PPy bulk, but also into the nanoholes. Different from the conventional CPs based drug release, choices of drug added to the nanoholes in this sponge like system are not limited by the charge and size of the molecules. Furthermore, this system can be used for simultaneous release of multiple drugs. With one kind of drug incorporated in the polymer bulk as dopants, another kind of drug can be stored in the nanoholes inside the polymer film. This simultaneous drug release system may find applications in cases where combined drug delivery is necessary, for example the delivery of an enzyme and its cofactor, or the delivery of a drug and its adjuvant.</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

The interplay between fluorescence and phosphorescence with luminescent gold(<scp>i</scp>) and gold(<scp>iii</scp>) complexes bearing heterocyclic arylacetylide ligands

The photophysical properties of a series of gold(I) [LAu(C^CR)] (L ¼ PCy 3 (1a-4a), RNC (5a), NHC (6a))and gold(III) complexes [Au(C^N^C)(C^CR)] (1b-4b) bearing heterocyclic arylacetylide ligands with narrow band-gap are compared. The luminescence of both series are derived from an intraligand transition localized on the arylacetylide ligand (pp*(C^CR)) but 1a-3a displayed prompt fluorescence (s PF ¼ 2.7-12.0 ns) while 1b-3b showed mainly phosphorescence (s Ph ¼ 104-205 ms). The experimentally determined intersystem crossing (ISC) rate constants (k ISC ) are on the order of 10 6 to 10 8 s À1 for the gold(I) series (1a-3a) but 10 10 to 10 11 s À1 for the gold(III) analogues (1b-3b). DFT/TDDFT calculations have been performed to help understand the difference in the k ISC between the two series of complexes. Owing to the different oxidation states of the gold ion, the Au(I) complexes have linear coordination geometry while the Au(III) complexes are square planar. It was found from DFT/TDDFT calculations that due to this difference in coordination geometries, the energy gap between the singlet and triplet excited states (DE ST ) with effective spin-orbit coupling (SOC) for Au(I) systems is much larger than that for the Au(III) counterparts, thus resulting in the poor ISC efficiency for the former. Timeresolved spectroscopies revealed a minor contribution (<2.9%) of a long-lived delayed fluorescence (DF) (s DF ¼ 4.6-12.5 ms) to the total fluorescence in 1a-3a. Attempts have been made to elucidate the mechanism for the origins of the DF: the dependence of the DF intensity with the power of excitation light reveals that triplet-triplet annihilation (TTA) is the most probable mechanism for the DF of 1a while germinate electron-hole pair (GP) recombination accounts for the DF of 2a in 77 K glassy solution (MeOH/EtOH ¼ 4 : 1). Both 4a and 4b contain a BODIPY moiety at the acetylide ligand and display only 1 IL(pp*) fluorescence with negligible phosphorescence being observed. Computational analyses attributed this observation to the lack of low-lying triplet excited states that could have effective SOC with the S 1 excited state.

the_interplay_between_fluorescence_and_phosphorescence_with_luminescent_gold(<scp>i</scp>)_and_gold(

Introduction<!>Synthesis and characterization<!>X-Ray crystallography<!>Electrochemical properties<!>UV-vis absorption spectroscopy<!>Steady-state emission spectroscopy<!>Time-resolved spectroscopies for the gold(I) complexes<!>Intersystem crossing rate<!>Computational study<!>Calculations on 1a and 1b<!>Calculations on 4a and 4b<!>General remarks on the photophysical properties<!>Intersystem crossing in gold(I) and gold(III) complexes<!>Mechanism for the generation of delayed uorescence<!>Conclusion

<p>Phosphorescence is a distinctive photophysical property of transition-metal complexes, which has widespread applications in diverse areas. As it is derived from the 'forbidden' radiative relaxation of a triplet-excited state to the singlet ground state, it is featured by long emission lifetime (in ms) and reduced emission energy compared to uorescence commonly encountered in organic luminophores. Electronic transitions associated with a change of spin are prohibited by the spin-selection rule. However, transition-metal ions that have high atomic number and hence, large spin-orbit coupling constant (x), usually lead to efficient spin-orbit coupling (SOC) that relaxes the spin selection rule. Fast intersystem crossing (ISC) in the sub-picosecond to picosecond time regime 1 leads to rapid depletion of a singlet excited state to a triplet excited state instead of uorescence as uorescence radiative lifetime is typically in the nanosecond range. Thus, in transition-metal complexes, phosphorescence normally prevails in their luminescence spectra.</p><p>However, in recent years, there are an increasing number of reports on transition-metal complexes which display slow ISC rate with lifetimes ranging from hundreds of ps to ns. For instance, 2,5-bis(arylethynyl)rhodacyclopentadiene complexes (x Rh ¼ 1260 cm À1 ) 2 were reported to display exclusively prompt uorescence with high emission quantum yields of 0.3-0.7 and lifetimes of 1-3 ns, corresponding to ISC rate constants of $10 8 s À1 . 3 In addition, transition-metal complexes containing fused aromatic systems such as perylene, perylene diimide, pyrene and tetracene also show ligand-dominated uorescence (see Fig. 1). 4 Hence, it has become clear that the presence of heavy elements does not guarantee fast ISC rate; the molecular structure and the nature of the ligands may play more critical roles in determining the ISC rate.</p><p>Luminescent Au(I) complexes are well documented to display rich photophysical properties. Although Au(I) complexes generally display phosphorescence owing to the large SOC constant of Au(I) ion (x Au $ 5100 cm À1 ), 2 ligand-centered uorescence has also been reported in a number of gold(I) complexes. For example, as revealed by the luminescence of [TEE(AuPCy 3 ) 4 ] and [TEB(AuPCy 3 ) 3 ] (TEE ¼ tetraethynylethene; TEB ¼ 1,3,5-triethynylbenzene), subtle changes in the electronic structure of the bridging alkynyl ligand leads to intense phosphorescence (F em ¼ 0.46, s ¼ 285 ms) in the latter but solely uorescence (F em ¼ 0.22, s < 0.05 ms) in the former. 5 In both cases, the luminescence originates from the ligand-centered transition mainly localized on the bridging alkynyl ligands. Che and co-workers also reported a series of Au(I)-conjugated acetylides, [(Cy 3 P)Au(C^C-C 6 H 4 ) nÀ1 (C^CPh)] (n $ 2), which display dual uorescence (prompt and delayed) and phosphorescence. 6 Both the F em and ratio of uorescence versus phosphorescence were found to depend on the conjugation length (number of repeating units n) and the substitution pattern of arylacetylide ligands.</p><p>As a continuous effort to elucidate the ligand effects on the photophysics of luminescent gold complexes, heterocyclic arylacetylide ligands containing narrow band-gap moieties (benzothiadiazole (L1), coumarin (L2), naphthalimide (L3) and borondipyrromethene (referred to as Bodipy) (L4); Chart 1) were chosen in this study. A series of Au(I) heterocyclic arylacetylide complexes, 1a-4a, were synthesized. Tricyclohexylphosphine (PCy 3 ) was used as the auxiliary ligand in these complexes because (1) it is optically transparent at wavelength >250 nm so that it is not involved in the emissive excited states in the UV-visible spectral region and (2) its steric bulkiness would prevent the gold ions from coming into close contact that could lead to low-lying excited states originated from metal-metal and p-p interactions. It is worth mentioning that several recently reported Au(I) alkynyl complexes bearing similar benzothiadiazole, 7 coumarin 8 and naphthalimide 9 derivatives also show similar luminescence properties as our Au(I) complexes. 10,11 The effects of auxiliary ligands on the photophysical properties of Au(I) complexes were also studied by comparing 1a with two derivatives containing 2,6-dimethylphenyl isocyanide (RNC, 5a) and 1,3-dimethylimidazol-2-ylidene (NHC, 6a) instead of the phosphine auxiliary ligand, respectively.</p><p>The effects of the oxidation state of the metal ion on the photophysical behaviours of transition-metal complexes are relatively unexplored. Herein, an analogous series of Au(III)acetylides supported by the cyclometalated [C^N^C] ligand (1b-4b; HC^N^CH ¼ 2,6-diphenylpyridine) were also prepared and their photophysical properties were compared with those of the Au(I) counterparts. The photophysical properties of both Au(I) and Au(III) complexes were investigated by steadystate and time-resolved spectroscopic measurements. DFT/ TDDFT calculations were performed on the pairs (1a, 1b) and (4a, 4b) in order to understand the origin of the dramatic difference in ISC efficiencies between these Au(I) and Au(III) complexes.</p><!><p>The gold(I) alkynyl complexes 1a-6a were synthesized in 53-79% yields following the protocol of base deprotonation (NaOMe) of terminal alkynes and substitution of chloride ion of the corresponding Au(I) precursors. 5a,6a,b,12 As these complexes were observed to show signs of decomposition on SiO 2 column, column chromatography was not used for their purication. Analytically pure 1a-6a were obtained by recrystallization from CH 2 Cl 2 /hexane mixtures. The Au(III) complexes 1b-4b were synthesized by copper-catalyzed Sonogashira coupling between terminal alkynes and [Au(C^N^C)Cl] using deoxygenated CH 2 Cl 2 as the solvent, with NEt 3 added to initiate the deprotonation of alkynes. 13 These complexes were puried by chromatography on SiO 2 column using dichloromethane and hexane as eluent. The yields were 51-84%.</p><p>All complexes have been characterized by 1 H and 13 C NMR, mass spectrometry (FAB+) and elemental analyses. Ligands L1-L4 were characterized by 1 H NMR and MS-EI. The complexes are stable in the solid state and in solution under ambient conditions. Complexes 1a-6a are highly soluble in CH 2 Cl 2 and THF but are less soluble in alcoholic solvents such as MeOH. Complexes 1b-4b have lower solubility compared with their Au(I) counterparts. All of these gold complexes appear as yellow or orange solids except for 4a and 4b that are purplish red. The</p><!><p>Crystals of 1a, 4a and 2b were obtained by layering hexane over concentrated CH 2 Cl 2 solutions. Their crystal data and selected bond lengths and angles are given in ESI. † Fig. 2 shows the structures of 1a and 4a (top panel). The P1-Au1-C(acetylide) angles of 1a and 4a are 175.0(13) and 178.1(2) and Au1-C^C angles are 170.9(4) and 177.6(5) , respectively, revealing slight deviation from linear coordination geometry. The Au1-C(acetylide) distances of 2.049(4) and 2.001(5) Å and C^C distances of 1.146 (7) and 1.191(8) Å for 1a and 4a, respectively, are comparable with those of other reported gold(I) acetylide complexes. 6a,b,12a,b The crystal packing diagrams of 1a and 4a are shown in ESI (Fig. S1 †). In both cases, there are no short intermolecular contacts; the closest Au/Au distances are 5.9715(5) and 5.2726(4) Å for 1a and 4a, respectively.</p><p>The crystal structure of 2b (Fig. 2, bottom) shows a slightly distorted square-planar geometry with C1-Au1-C17 angle of 162.44(14) . The Au1-C(acetylide) and C^C distances are 1.969(4) and 1.197(5) Å, respectively. These parameters are similar to those found in related cyclometalated Au(III) arylacetylide complexes. 13 The torsional angle between the Au(C^N^C) and arylacetylide planes is approximately 72.6 . This non-planarity gives rise to negligible p-p stacking between molecules as shown in Fig. S2 in ESI. †</p><!><p>The electrochemical properties of selected complexes, 1a-4a and 1b-4b, were investigated by cyclic voltammetry. The electrochemical data are summarized in Table 1. The cyclic voltammograms of the Au(I) complexes and their Au(III) counterparts are shown in Fig. S3, ESI. † Except for 1a and 2a, both classes of complexes display both irreversible oxidation (E pa ¼ +0.6 to +1.4 V) and quasi-reversible/irreversible reduction waves (E pc ¼ À1.5 to À1.8 V) attributed to the redox process localized on the arylacetylides. For the pairs [2a, 2b] and [4a, 4b], the E pa occur at relatively low potential of ca. +0.6 and +0.7 V, respectively, suggesting higher HOMO level of the conjugated coumarin and Bodipy than the other heterocyclic moieties. For the Au(III) complexes, other than the redox reactions occurring at the arylacetylide ligands at potentials similar to the Au(I) counterparts, there are also irreversible reduction waves at ca. À1.9 to À2.0 V attributable to reduction of the [C^N^C] ligands.</p><!><p>All photophysical data of the gold complexes and the free ligands L1-L4 are listed in Table 2. Electronic absorption of Au(I) complexes 1a-6a. Fig. 3 (top le) shows the absorption spectra of 1a-4a. The lowest energy absorption bands of 1a-4a are at 379, 410, 397 and 553 nm, respectively, and their molar absorptivities (3) fall in the range 6.9 Â 10 3 to 4 Â 10 4 mol À1 dm 3 cm À1 . These lowest energy absorption bands have spectral features resembling those of L1-L4 (Fig. S4 in ESI †) and are attributable to the dipole-allowed intraligand transitions of the arylacetylide ligands ( 1 pp*(C^CR)) with some charge-transfer character. Similar assignments were also made for other Au(I) alkynyl complexes in the literature. 6,7,8a Bathochromic shis of 1 IL transitions of arylacetylides are observed upon coordination of the arylacetylides to the Au(I) ion and are ascribed to p-interaction between Au(I) 5d orbitals and the ligand p-orbitals (see MO surfaces in Fig. 9 and 10).</p><p>Replacing the neutral auxiliary ligand PCy 3 in 1a with 2,6dimethylphenyl isocyanide (RNC, 5a) and 1,3-dimethylimidazol-2-ylidene (NHC, 6a) results in a slight change in l max of the lowest energy absorption band (l max ¼ 371 nm (5a, RNC) and 383 nm (6a, NHC) cf. l max ¼ 379 nm (1a, PCy 3 )) (Fig. S6 in ESI †).</p><p>Electronic absorption spectra of Au(III) complexes 1b-4b.</p><!><p>All of the complexes are luminescent in degassed CH 2 Cl 2 at room temperature and in 77 K glassy solutions (EtOH : MeOH ¼ 4 : 1) upon excitation at the corresponding lowest-energy absorption l max . As depicted in Table 2, there is a distinct difference between the two classes of complexes: the Au(I) complexes 1a-6a display predominantly uorescence while the Au(III) complexes, 1b-3b, exhibit exclusively weak phosphorescence. Complex 4b, on the other hand, shows uorescence only.</p><p>Emission of 1a-4a and 5a-6a. In dichloromethane solutions, structureless emission bands are observed at l max ¼ 467, 466, 439 and 553 nm for 1a-4a, respectively (Fig. 3, top right). The corresponding excitation spectra of 1a-4a can be found in the ESI (Fig. S8 †). The emission quantum yields for 1a-3a are high (F em ¼ 0.91, 0.70 and 0.78, respectively). In the case of 4a, its emission quantum yield is low (F em ¼ 0.04). Emission lifetimes of 1a-4a are in the nanosecond time regime: 0.8-12 ns. As the emissions of these Au(I) complexes resemble those of the corresponding free ligands L1-L4, they are attributable to 1 pp*(C^CR) excited states, with some charge transfer character, which probably arise from mixings of metal-to-ligand charge-transfer (MLCT) character. Solvent effects on the emissions of 1a-3a can be found in Fig. S10, ESI. † There is no discernible phosphorescence for 1a-4a under steady-state conditions in solutions at either room temperature or 77 K (Fig. S9, ESI †).</p><p>Comparing the three Au(I) complexes bearing the benzothiadiazole moiety, the emission energies (l max ¼ 467, 456 and 476 nm for 1a (PCy 3 ), 5a (RNC) and 6a (NHC), respectively (Fig. S11 in ESI †)) and emission lifetimes (s PF $ 11-14 ns) are similar, indicating that the auxiliary ligand plays an insignicant role in modication of the electronic structures of the excited states.</p><p>Emission of 1b-4b. Emission spectra of the Au(III) complexes are depicted in Fig. 3 (bottom right). Contrary to the Au(I) analogues where the emission proles are structureless, the emission spectra of complexes 1b-3b are vibronically structured with l max at 630, 592 and 603 nm and quantum yields of 0.003, 0.01 and 0.04, respectively. The emission lifetimes are of hundreds of microseconds ($100 ms for 1b and 2b; $200 ms for 3b). Taking into account the large Stokes shis (between 6300 and 10 400 cm À1 ), structured emission proles, and long emission lifetimes, the emissions of 1b-3b could be attributed to 3 pp*(C^CR) excited states with negligible mixings of MLCT and LLCT character (LLCT ¼ ligand-to-ligand charge transfer). Solvent effects on the emissions of 1b can be found in Fig. S12, ESI. † On the contrary, 4b shows emission with a small Stokes shi of 920 cm À1 and emission lifetime of only 2.1 ns. Thus, the emission of 4b is derived from uorescence with 1 pp*(C^CBodipy) parentage.</p><!><p>Nanosecond time-resolved emission (ns-TRE) spectra of the Au(I) complexes in degassed CH 2 Cl 2 solutions at 298 K (5 Â 10 À5 M) are measured at different time delays and are presented in Fig. 4 (1a) and ESI (2a-3a, 5a-6a; Fig. S13 †). There are two components in the emission decay: a major component which decays within nanoseconds (s 1 ¼ 12.0 (1a), 2.7 (2a), 2.8 (3a), 11.0 (5a) and 14.0 ns (6a)) and a minor component with microsecond decay lifetime (s 2 ¼ 11.9 (1a), 4.6 (2a), 2.6, 12.5 (3a), 6.8 (5a) and 7.4 ms (6a)). For each of these Au(I) complexes, both decay components have identical emission prole and peak energy and so, the short-lived one (s 1 ) is assigned to be prompt uorescence (PF) while the long-lived one (s 2 ) is delayed uorescence (DF) of 1 pp*(C^CR) character. In the case of 4a, only PF (s PF ¼ 0.8 ns) is observed. The proportion of DF and PF constituting the total emission of 1a-3a have been estimated (Table 3): the intensity of DF is minute (<3%) when compared with that of PF (>97%). It is noted that delayed uorescence in the microsecond time regime is indicative of the emission generated from a long-lived excited state. This is further supported by nanosecond transient absorption (ns-TA) measurements that reveal the presence of long-lived absorbing species in the microsecond timescale (vide infra).</p><p>Weak phosphorescence bands were observed for 1a-3a under different conditions. For dilute CH 2 Cl 2 solutions (1 Â 10 À5 M) at 298 K, dominant emissions were observed in the spectral region of 440-470 nm, which correspond to uorescence (Fig. S14, le panel in ESI †). In addition, weak emission peaks at ca. 600 nm become discernible for 2a and 3a and the lifetimes measured are 13.6 and 61.9 ms, respectively (Fig. S13, right panel in ESI †). Cooling to 77 K gives more resolved phosphorescence bands with vibrational progression spacings of 1300-1400 cm À1 for all three complexes (Fig. 5 and 6). For 1a, contrary to the ns-TRE spectra recorded in degassed CH 2 Cl 2 at room temperature (Fig. 4 (5 Â 10 À5 M); Fig. S14, ESI † (1 Â 10 À5 M)) where only DF could be observed over the time range 1-46 ms, in 77 K glassy solution, phosphorescence at 630 nm is dominant and the weak DF at 467 nm vanishes aer 80 ms (Fig. 5a). The phosphorescence band decays with rstorder kinetics at s phos ¼ 109 ms. Similarly, the low-temperature ns-TRE spectra of 3a is dominated by phosphorescence at 609 nm and DF vanishes aer 200 ms (Fig. 5b). The phosphorescence band also decays mono-exponentially with s phos ¼ 530 ms. The photodynamics of 2a at 77 K, however, is different from that of 1a and 3a: both DF and phosphorescence of 2a are of comparable intensities initially ($1 ms) in the 77 K ns-TRE spectra (Fig. 6); in addition, DF and phosphorescence do not follow rst-order kinetics but decay according to the power law (I f t À1 ) in the time interval 1 ms to 1.2 ms (inset of Fig. 6). The thermally induced Stokes shis (DE s ¼ E 00 (77 K) À E 00 (298 K)), being $0 (2a) and $107 cm À1 (3a), are small, thus supporting that the phosphorescence bands are originated from 3 IL. 15 Moreover, as the emission energies and proles of the lowenergy bands are similar to those of the steady-state emission a % PF and % DF are estimated by integrating the emission intensity of degassed CH 2 Cl 2 (5 Â 10 À5 M) in the spectral region of l ¼ 350-700 nm over the time range: 0-500 ns and 800 ns to 999 ms, respectively (l exc ¼ 355 nm). spectra of the Au(III) analogues, the low-energy emission bands of 1a-3a are assigned to be from phosphorescence decay of the 3 pp*(C^CR) excited state.</p><p>Nanosecond transient absorption (ns-TA) difference spectra of 1a-3a (Fig. 7) and 5a-6a (Fig. S15 in ESI †) have been recorded in deoxygenated CH 2 Cl 2 at a gate delay of 1 ms aer excitation at l ¼ 355 nm. The ns-TA spectra are characterized by an intense positive signal due to excited-state absorption (ESA) within the spectral range 400-700 nm. The decay time constants of the lowest-energy ESA (s ESA ) are 20.3 (1a), 39.2; 304 (2a), and 13.3; 70.9 ms (3a) (insets of Fig. 7). Changing the auxiliary ligand from PCy 3 (1a) to RNC (5a) and NHC (6a) results in negligible changes in the ns-TA spectra and s ESA (Fig. 7 vs. S15 †), suggesting that auxiliary ligand has little effect on the photophysics of the gold(I) arylacetylide complexes.</p><p>Time-resolved spectroscopies for the gold(III) complexes ns-TRE and ns-TA difference spectra of 1b-3b have been recorded in degassed CH 2 Cl 2 solutions at 298 K at a gate delay of 1 ms. The ns-TRE spectra of 1b-3b (Fig. S16 in ESI †) have the same emission proles and peak positions as the corresponding steadystate phosphorescence spectra and exhibit single exponential decay lifetimes of 20.8 (1b), 9.7 (2b) and 18.3 ms (3b). For the ns-TA difference spectra of 1b-3b (Fig. 8, bottom panel), a broad positive ESA band was observed in the spectral region 450-800 nm; this ESA signal follows rst-order kinetics with lifetimes determined to be 23.8 (1b), 13.6 (2b) and 25.4 ms (3b), in reasonable agreement with the phosphorescence decay lifetimes determined from their respective ns-TRE spectra, thus indicating that the broad ESA is derived from T 1 / T n absorption.</p><p>To probe the early excited state dynamics of the gold(III) complexes, in particular the events associated with ISC, femtosecond time-resolved uorescence (fs-TRF) and transient absorption difference spectra (fs-TA) of 1b-3b have been recorded. Fig. 8 depicts the fs-TRF (top panel) and fs-TA spectra (middle panel) of complexes 1b-3b in CH 2 Cl 2 solution at various time intervals aer 400 nm excitation at 298 K. Promptly (<2 ps) aer photo-excitation, an unstructured uorescence band peaking at 461 (1b), 473 (2b) and 459 nm (3b) appears and decays completely within 100 ps. As the TRF emission peaks and proles closely resemble those of their Au(I) analogues, 1a-3a, these TRF spectra are suggested to be originated from the 1 pp*(C^CR) excited state. Fitting of the kinetic traces at their peaking wavelengths reveals that bi-exponential functions are required for 1b-3b with s 1 and s 2 being 1.28 and 13.8 ps for 1b, 0.95 and 9.04 ps for 2b, and 0.74 and 5.22 ps for 3b.</p><p>In the fs-TA of 1b-3b (Fig. 8, middle panel), all three complexes displayed similar spectral transformations: the initially formed ($1.4-2.5 ps) excited state absorption peaking at $490 nm (ESA1) decays with a concomitant growth of a broad band covering a spectral region 450-800 nm (ESA2) and is fully developed within 40 ps and persists up to 2.7 ns (the longest time recorded in the fs measurements). Clear isosbestic points could be observed at $500 nm (1b), 530 nm (2b) and $500 and 700 nm (3b) during the temporal evolution. Such kind of spectral conversion points to a precursor-successor relationship between ESA1 and ESA2. Kinetic analyses at representative wavelengths of these TA spectra reveals that ESA1 of 1b and 3b decay bi-exponentially with s 1 and s 2 being 0.80 and 13.2 ps for 1b and 0.63 and 3.49 ps for 3b, respectively, whereas ESA1 of 2b decays with a single exponential time constant of s 2 ¼ 8.38 ps. ESA2, on the other hand, grows with rst-order kinetics for all three complexes 1b-3b with time constants s ESA2 ¼ 9.93 (1b), 4.64 (2b) and 5.73 ps (3b). Given the similar decay time constants between the fs-TRF and ESA1 in fs-TA of 1b-3b, the spectral dynamics for both time-resolved spectra should be originated from the same S 1 excited state, namely, the 1 pp*(C^CR) excited state as revealed in the fs-TRF. On the other hand, comparing the ESA2 in fs-TA spectra at the longest time recorded with the corresponding ns-TA spectra for each Au(III) complex (Fig. 8, bottom panel), the two spectra are similar, indicating that ESA2 is derived from T 1 / T n absorption. Because there is a precursor-successor relationship between the ESA1 (S 1 / S n absorption) and ESA2 (T 1 / T n absorption), s 2 of ESA1 is assigned to ISC from the S 1 excited state to a receiving triplet excited state, which then internally converted to the T 1 excited state with an ultrafast time scale. Thus, s ISC ¼ 13.2 ps (1b), 8.38 ps (2b) and 3.49 ps (3b). The short s 1 ¼ 0.80/1.28 (1b), 0.95 ps (2b) and 0.63/0.74 ps (3b) of ESA1/TRF may likely correspond to the S 1 vibrational relaxation.</p><!><p>The spectroscopically determined intersystem crossing rate constants (k ISC ) and the corresponding time constants (s ISC ) for both Au(I) and Au(III) complexes studied herein are tabulated in Table 4. For 1a-3a and 5a-6a, assuming that the major non-</p><p>The estimated k ISC for the gold(I) complexes are 7.5 Â 10 6 to 1.1 Â 10 8 s À1 and the intersystem crossing time constants (s ISC ) are 9.0-133 ns. These s ISC are much larger than those of many phosphorescent transition-metal complexes (s ISC in the femtosecond to picosecond timescale). For 1b-3b, the s ISC values are more than three orders of magnitude faster than their gold(I) analogues; these ISC rates, nevertheless, are comparable to other transition-metal complexes where S 1 / T 1 ISC is mediated by a higher-lying T n triplet excited state. 16,17</p><!><p>The different luminescence behaviors between the Au(I) and Au(III) systems were investigated by DFT/TDDFT calculations. The pair (1a, 1b) was chosen as a representative example to examine why the Au(I) complexes studied herein display only uorescence while the Au(III) counterparts exhibit exclusively phosphorescence. As the Bodipy-functionalized complexes give uorescence for both Au(I) and Au(III) complexes, DFT/TDDFT calculations were also performed on the pair, (4a, 4b). To save computational time, the cyclohexyl groups of the phosphine ligands in 1a and 4a were replaced by methyl groups.</p><!><p>The frontier MO diagrams of 1a and 1b are shown in Fig. 9. The HOMO and LUMO for both complexes 1a and 1b are predominantly localized on the arylacetylide ligand. For 1b, a considerable contribution (17%) from the C^N^C moiety to the LUMO is also noted. The energy gap between HOMO and HÀ1 in 1a is approximately 0.88 eV. For 1a, the HÀ1 is comprised of the antibonding combinations of Au(d xy ) and p(C^C) orbitals with little involvement of the phosphine ligand. For 1b, HÀ1 is composed of Au(d xz ) and the p(C^N^C) orbitals; the energy gap between HOMO and HÀ1 in 1b is only 0.2 eV. The HÀ2 of 1b is made up of an antibonding combination of the Au(d xy ), p(C^C) and s(C^N^C) orbitals, with a HOMO/HÀ2 orbital energy gap of only $0.4 eV. Clearly, the cyclometalated [C^N^C] ligand has a role in destabilizing the Au(d) orbitals. Therefore, the HOMO and HÀ1/HÀ2 energy gaps in 1b are much smaller than that in 1a.</p><p>The energies of the singlet and triplet excited states and the associated nature and composition for 1a and 1b at their respective optimized singlet ground state geometries are obtained by TDDFT and are shown in Table S5 and S6 in ESI. † For 1a, there is only one triplet excited state (T 1 ) which is more than 10 000 cm À1 below S 1 . In addition, both S 1 and T 1 excited states are of the same parentage and are derived from HOMO / LUMO transition ($90%) and thus, there would be no effective SOC between them. The triplet excited states above S 1 were also considered; the closest lying T m excited state with efficient SOC is when m ¼ 4, which is derived from a HÀ1 to LUMO transition (90% HÀ1 / L). However, the energy separation DE(S 1 -T 4 ) is À3180 cm À1 , which is too large to be overcome by thermal activation.</p><p>On the other hand, for 1b, there are four triplet excited states which are lower-lying than S 1 , of which the closest-lying T 4 excited state is only $70 cm À1 below the S 1 excited state. Thus, thermal energy at room temperature assists facile ISC, even though SOC is small between the S 1 and T 4 excited states (|<S 1 | H SOC |T 4 >| 2 $ 1.5 cm À2 ). In addition, among the triplet excited states above S 1 , there is a close-lying T 5 excited state derived from the HÀ2 / LUMO transition (79%) which lies only 390 cm À1 above the S 1 excited state and ISC from S 1 to T 5 could be thermally activated. Besides, owing to the different orientations of the d-orbitals in HOMO and HÀ2, the S 1 and T 5 excited states could have effective SOC (|<S 1 |H SOC |T 5 >| 2 $ 4.1 Â 10 3 cm À2 ).</p><!><p>The Frontier MOs for 4a and 4b are shown in Fig. 10. Relative to 1a and 1b, the HOMO is destabilized and the LUMO is stabilized Fig. 9 Frontier MOs of 1a and 1b at the optimized S 0 geometries. Orbital energies are also given in eV. for 4a and 4b. Even for 4b, which contains a cyclometalated [C^N^C] ligand, the LUMO is predominantly localized on the Bodipy-functionalised arylacetylide ligand. The HOMO is composed of an antibonding combination of the Au(d yz ) and p(C^CBodipy) orbitals. The energies and compositions of the singlet and triplet excited states of 4a and 4b at their respective optimized singlet ground state geometries were obtained by TDDFT and are collected in Tables S7 and S8 in ESI. † For this pair, (4a, 4b), the two triplet excited states, T 1 and T 2 , are more than 2000 cm À1 below the S 1 excited state and are all composed of Au(d yz ) orbitals. As SOC between the coupling singlet and triplet excited states would be ineffective with d-orbitals of the same orientation, ISC from S 1 to T 2 (or T 1 ) for 4a and 4b would be sluggish.</p><p>For 4a, the T 3 excited state is the closest-lying triplet excited state that could have effective SOC with the S 1 excited state due to a minor contribution of the HÀ5 / LUMO transition to the T 3 excited state (HÀ5 is composed of the Au(d z 2 ) orbital); however, DE(S 1 -T 3 ) is À2955 cm À1 (negative sign indicates that T 3 lies above S 1 ) which is much larger than the thermal energy. For 4b, the T 3 excited state is also the closest-lying triplet excited state that could have effective SOC with the S 1 excited state due to a minor contribution of HÀ1 / LUMO transition in the T 3 excited state (the d-orbitals of the Au(III) ion at the HOMO and HÀ1 of 4b are of different orientations, Fig. 10). However, the singlet-triplet gap, DE(S 1 -T 3 ) ¼ À1192 cm À1 , is also much larger than the thermal energy. Thus, the pair (4a, 4b) is expected to have slow ISC rates, when taking into consideration both the singlet-triplet energy gaps and SOC.</p><!><p>The emissions of the Au(I) complexes, 1a-6a are attributable to 1 IL pp*(C^CR) excited states. Pronounced red shis in emission l max of arylacetylides can be observed upon their coordination to Au(I) ion (e.g. l max ¼ 412 nm (L1) vs. 467 nm (1a)).</p><p>Considering the complexes 1a, 5a and 6a, which have different neutral auxiliary ligands (phosphine (PCy 3 , 1a), isocyanide (RNC, 5a) and N-heterocyclic carbene (NHC, 6a)) but the same acetylide ligand with a benzothiadiazole moiety, the lowest energy emission l em red shis with the auxiliary ligand from 456 nm (RNC) to 467 nm (PCy 3 ) to 476 nm (NHC). A rationalization would be that the NHC, being the strongest s-donor ligand among the auxiliary ligands in the three complexes, destabilizes the Au(d) orbital to the greatest extent. From DFT calculations, the HOMO is comprised of a Au(d) orbital and p(C^CR) (Fig. 9). Thus, the more electron-donating the auxiliary ligand, the more destabilized the HOMO, and hence, the smaller the HOMO-LUMO gap and the 1 pp*(C^CR) energy. A similar trend in the lowest energy absorption l abs can also be observed on changing the auxiliary ligand from RNC (371 nm) to PCy 3 (379 nm) and NHC (383 nm). Most of the reported luminescent cyclometalated Au(III) complexes display phosphorescence that comes from the 3 pp* IL excited state localized on the cyclometalated ligands. 13 For the Au(III) complexes studied herein, 1b-3b, the lowest-energy triplet excited states are of 3 pp*(C^CR) in nature, with l max ¼ 630, 592 and 603 nm respectively. These complexes, however, exhibit rather weak phosphorescence, with F em values in the range of 0.003-0.04. It is noted that in 77 K glassy solutions, the phosphorescence lifetimes are signicantly increased compared with those obtained in degassed CH 2 Cl 2 at RT (e.g. 205 ms at RT to 2.2 ms at 77 K for 3b). Since low temperature and rigid glassy matrix can impede structural distortion, the lifetimes obtained at 77 K could reect the intrinsic radiative lifetime of the complexes. The especially long emission lifetimes can reect the predominant localization of the emitting T 1 excited state on the arylacetylide ligand, i.e. 3 pp*(C^CR) with little participation of the metal ion. This is also corroborated by the small thermally induced Stokes shis (DE s ¼ E 00 (77 K) À E 00 (298 K)) of less than 600 cm À1 (Table 2).</p><!><p>ISC is usually fast in transition-metal complexes with time constants (s ISC ) in the fs to ps time regimes. In the literature, there are numerous examples which show ultrafast ISC, 1,18-25 e.g.</p><p>[M(bpy) 3 ] 2+ (M ¼ Ru or Fe, s ISC ¼ 30 fs), 19a,b [Re(L)(CO) 3 (bpy)] (s ISC ¼ 100-140 fs), 20 [Ir(piq) 3 ] (piq: 1-phenylisoquinoline; s ISC ¼ 70 fs), 21b [Pt(PBu 3 ) 2 (C^CPh) 2 ] (s ISC ¼ 70 fs), 22 and [Cy 3 PAu(2-naphthyl)] (s ISC ¼ 230 fs) 23 etc. These s ISC correspond to rates of intersystem crossing (k ISC ) in the range of 10 12 to 10 13 s À1 . The fast k ISC in transition-metal complexes is traditionally attributed to a large spin-orbit coupling (SOC) constant inherited from the heavy metal atom. However, there are increasing number of reports revealing slow ISC rate (k ISC $ 10 8 s À1 ) in spite of the presence of heavy transition metal, such as the cases of Rh(I)and Ir(III)-bis(arylethynyl)cyclopentadiene, 3 Au(I)-pyrene, 4b Pt(II)-perylene/ 4c tetracene, 4e and Pd(II)-perylene diimide; 4f all these complexes contain highly conjugated ligand systems and display ligand-dominated 1 pp* uorescence. There are also cases where comparable k ISC and k r of S 1 / S 0 leads to the observation of dual uorescence-phosphorescence under steady-state condition, e.g., [Pt(L)(acac)] and [Ir(L)(acac)] (L ¼ 2-(oligothienyl)pyridine); 26 [Os(L)(CO) 3 X] (L ¼ 8-quinolinolate 27 or isoquinoline-triazole), 28 and [Bu 4 N] 4 [Pt 2 (m-P 2 O 5 (BF 2 ) 2 ) 4 ], 29 etc.</p><p>In this work, the luminescence behaviour of the Au(I) and Au(III) complexes are drastically different, even though they have the same metal and arylacetylide ligands. Ligand-dominated uorescence has been observed with the Au(I) complexes, 1a-3a and 5a-6a, with k ISC estimated to range from 7.5 Â 10 6 to 1.1 Â 10 8 s À1 . The Au(III) complexes 1b-3b, on the other hand, display phosphorescence, with k ISC estimated to be larger than 10 10 s À1 . The major difference between the two series of gold complexes is the oxidation state of Au ion, that dictates the coordination geometry, i.e. a linear geometry for the Au(I) complexes, 1a-6a, and a square-planar geometry for Au(III) complexes, 1b-4b. The coordination geometry has a signicant impact on the relative energies of the frontier orbitals (specically, the d-orbital energies) and hence the relative energies of the singlet and triplet excited states, which subsequently affect the k ISC .</p><p>The two factors that determine the k ISC are (1) the SOC matrix element <S n |H SOC |T m >, and (2) the energy gap (DE ST ) between the coupling singlet (S n ) and triplet (T m ) excited states. The larger the H SOC and the smaller the energy gap (DE ST ), the faster will be k ISC . For effective SOC, this requires the metal d-orbitals of the coupling singlet and triplet excited states to have different orientations. For example, if S n is derived from a metal-to-ligand charge transfer (MLCT) excited state where Au(d xz ) orbital is involved, H SOC would be zero if the triplet excited state is also an MLCT state that involves Au(d xz ) orbital because of symmetry reasons.</p><p>The pair (1a, 1b) has been chosen as a representative example to illustrate the different photophysical properties exhibited by the Au(I) and Au(III) arylacetylide complexes studied in this work. From the DFT/TDDFT calculations, it is revealed that owing to the inherent linear coordination geometry of the Au(I) complex, the dorbitals of the gold(I) ion is mainly destabilized by the arylacetylide ligand (Fig. 9 and Table S9 in ESI †). On the other hand, as Au(III) complexes are assumed to have a square-planar four-coordinated geometry, thus, in addition to the antibonding interactions with the arylacetylide ligand, the d-orbitals of gold(III) ion could also be destabilized by the cyclometalated [C^N^C] ligand (both p-type, e.g. HÀ1, and s-type, e.g. HÀ2 in 1b; Fig. 9 and Table S9 in ESI †); these latter interactions result in smaller d-orbital splittings in the Au(III) series than the Au(I) series. In effect, S 1 and S 2 excited states are $4200 cm À1 apart for 1a while the analogous splitting (between S 1 and S 3 excited states) is only $1300 cm À1 for 1b. As S 2 of 1a is derived from 1 ( 1 MLCT/ 1 ILCT/ 1 LLCT), i.e., both are of charge-transfer type excited states, the singlet-triplet energy gaps for this type of transitions are small (DE(S 2 -T 4 ) $ 1000 cm À1 for 1a and DE(S 3 -T 5 ) $ 900 cm À1 for 1b) (T 4 (1a) and T 5 (1b) are the triplet counterpart of S 2 (1a) and S 3 (1b) respectively). As depicted in Fig. 11, the S 1 /T 5 energy gap for 1b is small but the S 1 /T 4 energy gap for 1a is large. In other words, the oxidation state of the gold ion affects the coordination geometry of the complex, which in turn change the interactions between the metal d-orbitals and ligand orbitals, giving rise to different d-orbital splitting and subsequently the singlet-triplet splitting (DE ST ) of the two coupling excited states in the gold complexes.</p><p>Moreover, DFT/TDDFT calculations also revealed that there is a triplet excited state (T 4 ) almost isoenergetic with the S 1 excited state (<70 cm À1 below the S 1 excited state) in 1b such that even though the SOC between S 1 and T 4 is small due to the similar d-orbital orientations involved in both excited states, thermal energy could promote facile ISC. With 1a, the closest triplet excited state (T 2 ) to the S 1 excited state is more than 500 cm À1 above the S 1 excited state, which is more than twice the thermal energy at room temperature and SOC is also small between these two excited states as the d-orbitals involved are also of the same orientations. Thus, taken together both the SOC and DE ST , 1b should have a much faster k ISC than 1a.</p><p>On the other hand, for the Bodipy-functionalized complexes, 4a and 4b, only 1 pp*(C^CBodipy) uorescence with no long-lived species are observed under ns-TRE and ns-TA measurements. The photophysical behavior of the Bodipyfunctionalized complexes can be attributed to the intrinsically small band-gap of the Bodipy moiety. Due to the highly conjugated structure of Bodipy, the HOMO is much destabilized and there is a wide orbital energy gap between the HOMO and other occupied MOs, even in the case of 4b which contains a [C^N^C] ligand. As a result, the HOMO/HÀx orbital energy gap is the largest among the four arylacetylide ligands studied herein (HÀx is the other occupied orbitals lower in energy than the HOMO; x ¼ 1, 2, .). In effect, the closest T m excited state that could have effective SOC with S 1 is more than 1000 cm À1 above the S 1 excited state. With such a large DE(S 1 -T m ), thermal energy would be insufficient to promote ISC. Therefore, similar to the scenario in the case of 1a (Fig. 11, le), ISC is sluggish for Au(I) and Au(III) arylacetylide complexes bearing Bodipy.</p><!><p>From the ns-TRE measurements of 1a-3a, DF contributes to the total uorescence, though only a minute proportion (<3%, Table 3). In general, the mechanism of DF could be inferred from the dependence of the DF intensity (I DF ) with the power of excitation light. 30 According to Bässler, a quadratic dependence of excitation power with the DF intensity indicates that the mechanism of the DF is TTA with dominant phosphorescence. 30b On the other hand, a linear dependence of DF intensity with excitation power could be due to three possible mechanisms: TTA with dominant delayed-uorescence, TADF, and GP-recombination. As depicted in Fig. 12, the plot of I DF against excitation intensity in double-logarithm scale gave a slope of 1.71 z 2 for 1a; this nearly quadratic dependence is most consistent with the TTA mechanism with dominant phosphorescence. However, for 2a and 3a (slope ¼ 0.916 and 1.10), both display nearly linear dependence between I DF and excitation intensity. Therefore, it is not possible to conrm the mechanism for DF in the case of 2a and 3a by solely considering the excitation power dependence.</p><p>Time-dependence of I DF and phosphorescence intensity (I P ) could also give hints to the DF mechanism. 30,31 For TTA with dominant DF, phosphorescence intensity decays with a power law, I P f t À1 while I DF is approximately constant at short time and I DF f t À2 at longer time. For the GP-recombination mechanism, both DF and phosphorescence decay in accordance with the power law, I DF f t À1 , at both short and long times. 30a In the case of 2a in 77 K glassy solution, both DF and phosphorescence decayed according to the power law: I f t À1 over the time intervals investigated (1 ms to 1.2 ms) (Fig. 6, inset), suggesting that the DF mechanism under this condition is most likely the GP-recombination mechanism. As for 3a, there is no power law decay relation with both DF and phosphorescence and so it seems unlikely that GP-recombination is the mechanism for the generation of DF in 3a. There is still not enough information to conclude on the DF mechanism for 3a.</p><!><p>A series of gold complexes with different oxidation states, gold(I) complexes [LAu(C^CR)] and gold(III) complexes [Au(C^N^C)(C^CR)] bearing the same heterocyclic arylacetylides with narrow bandgap were synthesized and characterized. The photophysical behaviors with the gold ion in different oxidation states are strikingly different: uorescence dominates the luminescence of the Au(I) complexes while phosphorescence takes over in the Au(III) complexes. Detailed computational studies by DFT/TDDFT have accounted for these phenomena as a result of different coordination environments inherited from the gold ion in a particular oxidation state: a linear coordination geometry for Au(I) and a square-planar coordination geometry for Au(III). This difference in coordination geometry subtly affects the energy separation between the coupling singlet and triplet excited states, leading to smaller DE ST of the Au(III) complexes than the Au(I) complexes and hence, larger k ISC in the Au(III) complexes than the Au(I) complexes. For the complexes bearing Bodipy-functionalized acetylide ligand, they only display prompt uorescence. Computational analyses revealed that, due to the especially narrow bandgap of Bodipy, the DE ST is still large even in the Au(III) complex so that k ISC could not compete with uorescence radiative decay. Additionally, the mechanisms for the generation of DF in Au(I) complexes have been explored. To the best of our knowledge, this is the rst report which systematically studies the effects of the metal ion oxidation state on the photophysical behaviours of transition-metal complexes.</p>

Royal Society of Chemistry (RSC)

A mature ROMANCE: a matter of quantity and how two can be better than one

Capillary electrophoresis coupled to mass spectrometry (CE-MS) is increasingly gaining momentum as an analytical tool in metabolomics, thanks to its ability to obtain information about the most polar elements in biological samples. This has been helped by improvements in peak robustness by means of mobility-scale representations of the electropherograms (mobilograms). As a necessary step towards the use of CE-MS for untargeted metabolomics data, the authors previously developed and introduced the ROMANCE software, with the purpose of automating mobilogram generation for large untargeted datasets while offering a simple and self-contained user interface. In natural continuation ROMANCE has been upgraded to its v2 to read other types of data (targeted MS data and 2D UV-like electropherograms), offer more flexibility in the transformation parameters (including field ramping delays and the use of secondary markers), more measurement conditions (depending on detection and ionization modes), and most importantly tackle the issue of quantitative CE-MS. To prepare the ground for such an upgrade, we present a review of the current theoretical framework with regards to peak reproducibility and quantification, and we develop new formulas for multiple marker peak area corrections, for anticipating peak position precision, and for assessing peak shape distortion. We then present the new version of the software, and validate it experimentally. We contrast the multiple marker mobility transformations with previous results, finding increased precision, and finally we showcase an application to actual untargeted metabolomics data.

a_mature_romance:_a_matter_of_quantity_and_how_two_can_be_better_than_one

Introduction<!>Theory<!>Effective electrophoretic mobility 2.1.1 Migration under non-constant fields<!>Two-marker formulas<!>Summary<!>Area-preserving transformation<!>Effects on peak shape<!>Ionization mode<!>Summary<!>Marker influence on precision<!>Software<!>Experimental validation 4.1 Material and methods<!>Chemicals and reagents<!>Standard solution preparation<!>Cell culture preparation<!>BGE and sheath-liquid preparation<!>Validation analyses<!>Untargeted metabolomics<!>Data processing<!>Ramp and 2-marker formulas<!>Peak area precision<!>Metabolomics application<!>Concluding remarks

<p>There is a vast diversity of molecules presenting potential interest as the subject of metabolomic analyses. Due to the different nature of their chemical properties, it is unfeasible to explore their behaviours in biological systems except when using a panel of complementary separation and detection techniques. Each analytical platform shall be able to grasp the most relevant information about one of more families of chemically-related compounds. In the case of capillary zone electrophoresis (CZE), the mechanism driving the separation makes it particularly well suited for the analysis of polar and charged molecules [1]. Even though a significant fraction of the most commonly studied metabolites falls within this category, the use of CE still remains quite limited in the field of metabolomics [2,3].</p><p>Compared to other techniques such as LC or GC, where the retention of the analyte can be used to identify the unknows along with other properties such as the accurate mass or the fragmentation pattern, the use of migration times in CE presents the disadvantage of being a much less robust parameter. Many factors account for this lack of robustness, such as different experimental setups, matrix adsorption on the capillary walls, or changes in the suction caused by ESI sources at the capillary outlet when hyphenated to MS. To mitigate this issue, some alternatives to migration times have been proposed. Relative migration times expressed as ratios to a reference compound can be useful, but they are unable to adequately cope with the effect of constant parameters such as the application of a pressure to assist the electroosmotic flow (EOF) during the separation. Effective electrophoretic mobility is a much more robust parameter, and it constitutes a better option for the identification of metabolites since it is a molecular property which only depends on the nature of the chosen BGE and separation temperature [4,5].</p><p>In spite of its convenience, calculation of electrophoretic mobilities is a tedious process, and so we recently developed ROMANCE, a software allowing to automate the process of converting batches of CE-MS files into effective electrophoretic mobility scale [6]. As it has been already described, files using this x-axis instead of time scale, present several advantages such as constant peak position for a given compound, no need for alignment and better shareability of experimental and reference data between different labs. Nevertheless, our previous article was focused on how the use of the electrophoretic mobility scale could improve metabolite annotation, while paying little attention to the influence of the time-toelectrophoretic mobility conversion on the peak area.</p><p>As a natural continuation to the introduction of ROMANCE in our previous article, we herein present ROMANCE v2. This new version of the software has been developed to take into account new operational scenarios such as non-constant electric fields, the use of more than one reference compound, and the influence of the conversion or the ionization regime on the peak area. These and other relevant fundamental considerations are described in detail in the provided theoretical framework before illustrating their utility and performance using a panel of reference compounds used for validation and then, a metabolomics study conducted on a small set of cell culture samples.</p><!><p>In this section we will derive formulas to transform migration times to effective electrophoretic mobilities, taking into account non-constant fields and the possibility of using more than one marker in the spirit of [7,8]. We will review the effect of these transformations on peak areas generalizing the results in [9]. Finally, we will study their impact on peak shapes giving quantitative measures of peak displacement and deformation, and from them we will propose a priori rules for the optimization of precision of peak position.</p><p>Each derivation is accompanied by a summary subsection recollecting simplified versions of the most important formulas, for readers that may wish to skip the corresponding mathematical details or just have a quick reference.</p><!><p>To make the connection between the kinetics of electrophoretic migration and the parameters of the system, start by recalling that the migration speed v of a particular analyte in CE is</p><p>where µ is the effective electrophoretic mobility of the analyte, E is the applied electric field, and v BGE the speed of the background electrolyte (BGE) flow. We assume this can be written as</p><p>where µ BGE is the electroosmotic mobility of the BGE and v p a possible constant term generated typically by the application of a pressure gradient along the capillary.</p><p>Our first modification to the standard story is the presence of a ramp time t R over which the electric field will change up to a final constant value. This ramp time will always be smaller than the migration times of the analytes, t M ,</p><p>On the one hand, it is crucial to have formulas that can be written in closed form, without relying on numerical integration for every transformation. Without such closed formulas it would be impossible to derive the correction factors for the peak areas that will be presented later in the article. On the other, this is already the case in virtually every CE setup. Contrast with LC, where gradients of retentivity have an intrinsic positive effect on separation. In CE, a high electric field will normally just induce a better separation, as it linearly affects the migration speeds. The "ramp phase" of the electric field is often required for mostly instrumental reasons. This allows us to parametrize an arbitrary electric field profile E(t) by</p><p>where t R is the time over which the field is allowed to vary, and E m the final, constant applied field. The function</p><p>describes the shape of the ramp, giving the fraction of the maximum applied field E m as a function of the fraction of time between 0 and t R . For example, a simple linear ramp has a shape function given by S linear (τ ) = τ,</p><p>resulting in a linear ramp field profile</p><p>as shown in Figure 1.</p><p>During the separation, the analyte must travel the length of the capillary, L. This length is obtained from integrating the speed between the start of the separation at t = 0 and the migration time of the analyte, t M .</p><p>Since we assumed that t M > t R , we can split the integral into three terms: the contributions of pressure, of the ramping field, and the constant field</p><p>The only dependence on the shape of the ramp is in the second term, which in turn does not depend on the migration time. We get</p><p>The whole contribution of the shape can be summed up in a shape parameter, defined by</p><p>This is just the area over the ramp shape curve. As a sanity check, a constant, maximal shape function (S R (τ ) = 1) produces a shape factor of λ = 0, or that is, no contribution from the ramp. A linear ramp like in Figure 1 on the other hand will have</p><p>The shape factor for other ramp profiles can be easily calculated with formula (11).</p><p>Our objective is to transform the migration times t M into effective mobilities µ, so we solve for the latter in (10),</p><p>Here we face the well known problem of CE -the electroosmotic mobility µ BGE is highly variable, and needs to be determined run by run. Being just an offset, by measuring also the migration time t A of a substance of known effective mobility µ A we arrive at</p><p>A convenient choice for A is the migration time of the electroosmotic flux, t EOF , which has effective mobility µ EOF = 0 (this should not be mistaken for the electroosmotic mobility µ BGE , which is not zero). If additionally there is no ramp time, we have just the usual formula</p><p>However, when a ramp is present there is another unknown parameter, v p . This depends in a non-trivial manner on the applied pressure, but also indirectly on harder to control variables such as temperature that influence, e.g., the viscosity of the fluid. Just as for µ BGE (and unlike L or E m ) we cannot expect to know its value on a run by run basis.</p><!><p>Instead we can apply the same reasoning as for µ BGE : eliminate an unknown instrumental parameter with a measurable marker property. The idea goes back to [8,10], where it is applied to other sources of time shift, although without considering the implications on peak area and always assuming one of the markers to be the EOF.</p><p>In short, take a second marker (t B , µ B ) and consider the quotient</p><p>The only instrumental parameter this relation depends is the ramp profile (through λ and t R ). Solving for µ, we arrive at our two-marker effective mobility formula,</p><p>Intuitively, this is an interpolation between the mobilities of the two markers, weighted (nonlinearly) by their respective distances to an analyte with migration time t. The formula simplifies considerably if one of the two markers is the EOF, say the second one µ B = µ EOF = 0,</p><p>In this form, it interpolates between zero mobility, for analytes arriving with the EOF (t M t EOF ), and mobility µ A , for analytes arriving with A. As in the general two-marker formula, the interpolation is not linear. We can see an example of the shape of µ as a function of t in Figure 2. The two-marker formula in (18) has an additional advantage, already seen in [8]. In the search to eliminate the instrumental parameter v p that appeared as a consequence of the field ramp, we have also gotten rid of the field and capillary length, E and L. Because of this, even in absence of a field ramp, the two-marker formula can actually be an improvement in terms of precision with respect to the known one-marker formula (due to inexact values of the effective capillary length, parasitic resistances affecting the actual value of the field applied to the capillary, etc.).</p><!><p>The mobility of a peak with migration time t M , in a capillary of length L, with a maximal electrical field E m and a linear field ramp of length t R is</p><p>as a function of the electroosmotic flow time t EOF , and neglecting contributions from the backpressure. If another marker with mobility µ A and time t A is used, the formula is</p><p>2.2 Electropherogram peak shapes</p><!><p>Recall that the purpose of transforming migration times into effective mobilities is to allow peak-picking software to have a robust identifying parameter (the mobility of the compound) to annotate features extracted from electropherograms. The obvious approach, as implemented in [6], is to use formula (14) or (18) to map t → µ(t). An electropherogram is measured as a list of couples (t i , I i ) of times t i and intensities/counts I i . Therefore the mobilogram should be computed by</p><p>In [6] we observed that this performs well for the annotation of features. However, if the intensities I i represent a concentration profile that should be integrated over time, this change poses a problem. In Figure 3 we show an example. We suppose an EOF arriving at t EOF = 10, a marker with t A = 1 and effective mobility µ A = 100, and no ramp time. We then place two peaks with similar widths σ 1 = 0.2 and σ 2 = 0.25 and exactly the same area A 1 = A 2 = 1, at times t 1 = 2.5 and t 2 = 7. Because of the nonlinearity of the t → µ transformation, the peak areas are completely changed. Although gaussian-like peaks will remain essentially gaussian as long as they do not overlap the EOF (whatever distortion may be introduced by the transformation is negligible next to the inherent variations in CE peak shapes, as we will see), peak widths are changed notably. This was already observed in [7], and the necessary correction was derived in [9] for the simple case of (15). We will redo the calculation to include the ramp factor and the secondary marker. Essentially, having a closed formula such as (17) for µ(t) allows us to compute a correction factor. In the limit of a continuous intensity profile I(t), the area under the curve of a peak in the electropherogram is given by</p><p>Changing variables from t → µ(t), the integration measure changes by This has a simple expression as a function of t, which can be obtained from (17),</p><p>Mobilograms should take into account this factor, giving a corrected mobilogram intensity I mob ,</p><p>The areas obtained by integrating the electropherogram (t i , I i ) and the corrected mobilogram (µ(t i ), I mob i ) will therefore be the same. The right factors in (25) do not depend on the specific point of the mobilogram, and it is a function of only L, E m , t R and v p . This means that when comparing runs performed under the same instrumental conditions (field magnitude and ramp, capillary length, and pressure), we can do without such overall factor, simply using</p><p>Of course, in doing so we ignore potential variability arising from the instrumental parameters. This may be important for the precision of migration times vs. effective mobilities. But the intrinsic variability in both the detection's response to actual concentration and the numerics of peak integration makes instrumental parameter variability barely relevant for peak areas.</p><!><p>With (26) -or more correctly (25)-we can ensure that peak areas in the mobilogram stay faithful to those in the electropherogram. While our previous derivation, or that of [9], focus on this aspect, there is another issue that can severely affect peak integration: peak shape. Both the correction factor and the transformation itself may introduce distortions. Figure 4 provides a few examples. The larger peak widths have been intentionally chosen to exaggerate the effects of the transformation, and real electropherograms will display milder effects. We can see that peaks closer to the EOF (case B) show, as expected, a greater increase in relative width in the mobilogram (corrected through their reduced height), but their shape remains largely Gaussian. Peaks of very short migration times (high mobilities, case A) present little broadening, but may display asymmetry if the peak width is comparable to the migration time. This asymmetry will complicate peak detection by peak-picking software. Nonetheless, note that case A2 is quite extreme, with its relative peak width being 40% of its migration time. Let us first focus on the asymmetry introduced only by the transformation. Suppose we have a peak centered in the electropherogram at t 0 , spanning from t − = t 0 − ∆t to t + = t 0 + ∆t. By solving</p><p>with equation ( 13) one can easily arrive at</p><p>These would be the (asymmetric) widths of the peak in the mobilogram. The terms on the right side of the expression are essentially (25), and correspond to the width change introduced by the transformation we have already studied. We can define a dimensionless relative asymmetry by</p><p>which from (28) turns out to be equal to</p><p>having a remarkably simple form. In short, equation (30) says that the relative asymmetry of a peak in the mobilogram equals precisely its electropherogram relative width with respect to its migration time. The latter will normally be low anyway, if we are to have a separation between the different analytes at all. Except for compounds of extremely high mobility (and thus short migration times), the impact of the transformation on peak shape should not affect peak integration. However, as we can see in case A2 in Figure 4, the asymmetry will also shift the position of the peak's maximum away from the correct value. While a 5% asymmetry originating from a 5% relative peak width will not be a problem for the detection of the peak shape, a potential 5% shift in peak position would definitely render the advantages of the transformation useless. On the other hand, the centre of mass of the peak seems to be roughly in the same place, as the asymmetry gives a slightly heavier tail on the right side of the mobilogram. Let us quantify this effect.</p><p>In the electropherogram, the center of mass position t is given by</p><p>The mobility corresponding to this center of mass is µ( t). In the case of an infinitely thin peak, this would also be the mobilogram's center of mass. In general, it will be μ = µ I mob (µ) dµ</p><p>We defined I mob in (25) so that</p><p>and so</p><p>We can expand µ(t) in a Taylor series around the electropherogram's center of mass,</p><p>and inserting it into the expression for the center of mass of the mobilogram,</p><p>Critically, the linear term vanishes, and the quadratic one is simply the variance of the electropherogram's peak, σ 2 . Using (18) to find the second derivative of µ with respec to t, we end with</p><p>This means that the center of mass in the mobilogram, μ is shifted with respect to the true mobility of the center of mass in the electropherogram. The term depending on t EOF is just an artefact of considering the relative shift -compounds close to the EOF have near zero effective mobility, so any variation will induce a large relative shift. The important part is the dependence on σ -quadratic on the ratio between peak width and peak position.</p><p>As an example, for a peak mid-way through the electropherogram, t = 0.5 t EOF , and a relatively large peak width, σ = 5% • t, the relative shift between the mobilogram's center of mass and the true mobility is</p><p>If the peak width is σ = 1% • t, this quickly descends to</p><p>much less than the variability on effective mobility that will stem from other sources. The lesson is that as long as relative peak widths are kept below reasonable limits (< 5% of the migration time) the transformation and correction should not negatively affect peak detection, and if center of mass integration is used, it will also not affect peak position whatsoever.</p><!><p>The previous section has been devoted to ensuring that areas in mobilograms map faithfully to areas in electropherograms. Without further treatment, this assumes that in turn electropherogram areas represent faithfully the amount of substance in the sample. This is highly dependent on the detection type. In the context of CE-MS, the critical element is the electrospray ionization (ESI) that nebulizes the output of the capillary to be fed into the MS. This ionization has two well known regimes of operation, the so-called mass and concentration modes [11]. In mass mode, the ESI is able to ionize (nearly) all the substance coming from the capillary. If the capillary flow is increased, more substance will be ionized per unit time and the MS will simply register more counts. By contrast, when in concentration mode the ESI is saturated. It outputs an amount of ions proportional to the volumetric concentration of substance arriving from the capillary, independently of time. Crucially, increasing the flow in the capillary does not increase the number of counts.</p><p>The increase in width between the two flow regimes is strictly proportional to the exit speed of the analyte from the capillary -this does not depend on the ionization mode, simply on the fact that a fixed length in the capillary spends twice as much time exiting it if it is moving at half the speed. One can emulate the height reduction observed in mass mode on data measured in concentration mode by simply multiplying each peak by its speed. Equation (1) relates the speed of an analyte to its mobility, which in turn can be derived from its migration time from (13) or (17). This provides the exit speed v exit as a function of migration time t M ,</p><p>As a side note, detection from UV absorbance and similar methods work exactly in the same way concentration mode ESI does in CE-MS. The absorbance is not a function of the flow rate, so total peak areas will depend on it. The same correction applies to them.</p><p>If one has an electropherogram measured in concentration mode (or from UV measurements), given by couples (t i , I conc. i ), the equivalent mass-mode electropherogram can be computed with</p><p>Just like for the mobility transformation correction (26), when comparing runs under the same experimental conditions one may as well ignore the constant factor and use</p><!><p>The need to apply the corrections shown in the previous sections is selected by the following two questions:</p><p>1. Integration: does the peak-picking software integrate the area under the curve, or does it sum the number of counts?</p><p>The former scenario requires correction (26) due to the distortion of the mobility transformation on the time scale. On the other hand, if each point in the electropherogram is understood as the number of counts since the previous acquisition, no correction needs to be made, since the timing between points becomes irrelevant.</p><p>2. Detection: is the signal proportional to the total mass of substance, or to its concentration? Mass mode ESI requires no correction, while concentration mode ESI or UV absorbance measurements require (42).</p><p>The combined correction to be applied in each case to the electropherogram, prior to its conversion to mobilities, is summarized in Table 1. Of course, if there are non-negligible contributions from the ramp, the full formulas ( 25) and (41) should be used.</p><!><p>The choice of EOF as a marker is quite natural, since it is normally easily identifiable. This begs the question of which criterion to optimize when choosing a secondary marker in the two-marker formula. Intuitively, one can already see that choosing a marker too close to the EOF will yield bad quality results, since any small error in the determination of the marker's position would have important effects when extrapolated away. Let us start with the formula for an analyte's mobility using the EOF and a marker (18). Assume there is some variation δt A in the determination of the marker time. This can reflect either instrumental variability, or uncertainty in peak integration. Similar to the calculations for peak asymmetry, we consider the variation this would induce in the transformed mobility,</p><p>and equation (18) we get that the relative uncertainty in mobility caused by an uncertainty δt A in the marker time is</p><p>The approximation we just made is valid for as long as the uncertainty is smaller than the distance between the secondary marker and the EOF. In such case, the relative mobility uncertainty is directly proportional to the marker position uncertainty. When t A is too close to either t A → t EOF or t A → λt R the relative error induced in the computed mobility becomes very large. We can also compute the mobility uncertainty induced by that of t EOF , which we denote δt EOF . A similar computation yields</p><p>Unlike the uncertainty from t A , this depends also on the the migration time of the analyte t M -analytes too close to the EOF will suffer more from uncertainty in its determination.</p><p>Assuming that both δt A and δt EOF are independent random variables, the variance of their combined effects will just be the sum of the variances. The expected total uncertainty is δµ µ A+EOF = δµ µ In Figure 5 we plot this combined effect, as a function of t A and t M . We take t EOF = 15 min, and consider uncertainties of 1 s and 2 s. From the influence of t A , the lesson is that the secondary marker should be chosen at the center between t = 0 and t EOF . However, if uncertainty in the determination of the EOF's time is larger (this can happen if the EOF marker is a relatively wide or non-Gaussian peak), compounds with low mobilities -close to the EOF-will be highly affected by a choice of t A with high mobility, further away from the EOF. Overall, the recommendation is to choose the secondary marker as close as possible to the midpoint towards the EOF, erring on the side of slightly lower mobilities only if necessary.</p><!><p>We have hitherto presented a theoretical derivation of a two-marker mobility formula, the possibility to handle field ramping, and the necessary area corrections to ensure that mobilograms represent consistently the amount of analyte in the samples. These features were not available in the version of the ROMANCE software introduced in [6]. With the publication of this article, we have updated ROMANCE to its v2, to include them. The software is still publicly available at https://ispso.unige.ch/labs/fanal/romance and still developed in the Scala language, to take advantage of parallelization and multiplatform support. Summarizing the main changes and additions, it now offers:</p><p>• Possibility to choose between the old instrumental parameters (E, L) and a secondary marker as in (17).</p><p>• Ramp time correction.</p><p>• Visual peak assessment windows for both markers, if applicable.</p><p>• Mobilogram area correction, ionization mode area correction, and inter-sample area normalization.</p><p>• Support for untargeted MS (spectra-based) mzML files, targeted MS (electropherogrambased) mzML files, and plain CSV files for UV-style data.</p><p>This updated version has been used to perform the conversions of experimental data studied in the following sections.</p><!><p>To validate the formulas derived in the previous section, we have selected 15 compounds, which can be found in the supplementary material, Table A. They were chosen first for having medium to high mobilities, to focus on those that can in principle suffer the highest distortions due to the transformation, as seen in section 2.2.2. In second place, by displaying good (i.e. relatively Gaussian) peak shapes in the electropherogram, to reduce as much as possible the variability in peak area stemming from peak integration, and again focus on the effects of the transformation. The compounds were prepared as a mixture of standards.</p><!><p>Acetic acid and formic acid were purchased from Biosolve (Dieuze, France). Water, methanol, isopropanol (iPrOH), and acetonitrile (ACN) were purchased from Fisher Scientific (Loughborough, United Kingdom). Standard compounds were purchased from Sigma Aldrich (Buchs, Switzerland).</p><!><p>Individual stock solutions of compounds were prepared in 5% v/v ACN and 0.1 % v/v FA at a final concentration of 1 mg/mL and stored at -80 • C. Mix stock solutions were prepared in 5% v/v ACN and 0.1% v/v FA at 10 µg/mL and stored at -80 • C. Mix stock was extemporaneously diluted to 500, 250, 125 and 62.5 ng/mL with water.</p><!><p>Four replicates of 2D-cell cultures of astrocytes were grown in the presence of different natural neuro-inflammatory triggers at different concentrations, namely interleukin 1β (IL-1β) at 30 ng/mL, tumoral necrosis factor α (TNFα) at 30 ng/mL, and lipopolysaccharide (LPS) at 10 µg/mL. The control cell culture was grown in parallel in absence of any inflammatory trigger. After two weeks of growth, cell cultures were snap-frozen in liquid nitrogen. Protein precipitation was achieved by adding 1 mL of a cold solution (-20 • C) of MeOH:H 2 O (80:20 v/v), scraping, and by vortexing during 1 minute. Samples were centrifuged at 14000 g during 15 minutes at 4 • C, the supernatant was collected and then evaporated to dryness before resuspension in 100 µL of a solution made of MeCN:H 2 O (50:50 v/v). Quality control (QC) samples were prepared by pooling the same volume from each sample after reconstitution. Volumes of 10 µL of individual cell culture extracts, QCs and diluted QCs were evaporated to dryness using a SpeedVac (ThermoFisher, Langenselbold, Germany). Before injection, samples were reconstituted with 10 µL of an aqueous solution containing paracetamol, procaine and ethyl-sulfate at a concentration of 50, 5 and 5 µg/mL respectively.</p><!><p>Through the study, 10% v/v acetic acid in water was used as BGE. The sheath-liquid was composed of isopropanol-water-acetic acid (50:50:1). To ensure a large metabolome coverage for comprehensive metabolomics profiling, the sheath-liquid was composed of acetic acid (5 mM) in an isopropanol-water (50:50) solution. Purine and HP-0921 were purchased from Agilent technologies (Santa Clara, CA, USA, P/N: G1969-8001) and used as lock masses after being spiked into the sheath-liquid to yield final concentrations of 50 and 25 nM, respectively.</p><!><p>A triple quadrupole platform was used for the study and validation of the formulas derived in the theory section. The separation was carried out with a G7100 capillary electrophoresis (CE) system from Agilent Technologies (Waldbronn, Germany). Separations were performed using a fused silica capillary purchased from BGB technologies (Boeckten, Switzerland) with a total length of 70 cm and an internal diameter of 50 µm. Prior to its first use, the capillary was conditioned with MeOH, H 2 O, NaOH 1M, H 2 O, HCl 1M, H 2 O, HCl 0.1M, H 2 O, and BGE at 5 bar during 1 minute each. Injections were performed hydrodynamically by application of 50 mbar during 12 s, using ∼1% of the capillary total length, circa 14 nL. Injected volumes were calculated with Zeecalc v1.0b (https://ispso.unige.ch/labs/fanal/zeecalc). Separation was performed by application of + 30 kV. Before each analysis, the capillary was washed with MeOH and BGE at 5 bar during 1 minute. To avoid temperature inhomogeneities between the capillary parts inside and outside the CE instrument, the CE thermostat was set at room temperature (∼23 • C).</p><p>The CE system was hyphenated with an Agilent 6490 triple quadrupole mass spectrometer (QqQ MS, Agilent Technologies, Santa Clara, CA, US) equipped with an ESI source via a coaxial sheath-flow interface with a standard triple-tube sprayer (P/N G1607B) from Agilent Technologies. The sheath liquid was delivered at a flow rate of 3 µL/min, using a 2300 Series isocratic pump purchased from Agilent Technologies (Waldbronn, Germany) equipped with a 1:100 split. Electrospray ionization was operated in positive mode, and spectra were acquired via SRM measurements. The pressures and injection volumes used during the validation are described in section 4.2. The precursor and productions monitored for each compound and the collision energies are reported in Table B in the supplementary information. The following source parameters were used: the nebulizing gas pressure was set at 0 psi and the sheath gas at 11 L/min and 150 • C. The capillary voltage was adjusted to 5500 V. The ion funnel voltages were set at 150 V for the high-pressure funnel and 60 V for the low-pressure one. The EMV voltage was set at 400 V and the cell accelerator voltage at 5 V. For all transitions, precursor and product ion selection was performed with a resolution of 1.2 and 0.7 m/z, respectively. Data acquisition and instrument control were performed using MassHunter version B.08.00 (Agilent, Santa Clara, US).</p><!><p>The CE setup for the untargeted metabolomics profiling was the same as described in section 4.1.5 for the validation. The CE system was in turn hyphenated to a maXis-3G QTOF MS from Bruker (Bremen, Germany), equipped with an ESI source via a coaxial sheath-flow ESI interface with a standard triple-tube sprayer (Agilent P/N G1607A) and a platinum needle. The sheath liquid was delivered at a flow rate of µL/min, using a 2300 Series isocratic pump purchased from Agilent Technologies (Waldbronn, Germany) equipped with a 1:100 split.</p><p>For cationic profiling, ESI was operated in positive mode with the following MS parameters: nebulizer and sheath gas were set to 0 bar and 10 L/min, 100 • C, respectively. Capillary, end-plate and funnel voltages were respectively adjusted to 6000, 400 and 300 V. For anionic profiling, ESI was performed in negative mode with the following source parameters: nebulizing gas and sheath gas were set to 0.3 bar and 4 L/min, 150 • C respectively. Capillary, end-plate and funnel voltages were adjusted to 4000, 400 and 300 V, respectively. MS acquisitions were performed at a frequency of 1 Hz, with a mass range going from 50 to 1000m/z.</p><!><p>The raw data files were converted to the mzML format [12] using ProteoWizard msConvert [13]. They were subsequently converted to the effective electrophoretic mobility scale with ROMANCE v2, when applicable. Finally, the peak positions and areas were extracted with Skyline [14] for the targeted analyses on standards, and with Progenesis QI v.2.4 for the untargetd metabolomics data (Nonlinear Dynamics, Newcastle upon Tyne, UK).</p><!><p>In this section we will study the effect of the ramp correction and 2-marker formula (18) on the determination of the mobilities of the analytes. We will compare the obtained mobilities with values from a previous library [6] (restricting ourselves to the 10 substances for which a mobility was there given), and their precision within our set of experiences.</p><p>The samples were separated with a linear ramp of t R = 60 s, and run in triplicate once applying a 0 mbar pressure, and once applying a 50 mbar pressure. This had the purpose of ensuring a considerable spread of the migration times. Each run was transformed with ROMANCE v2 to the mobility scale under each of the following four modes:</p><p>1. One marker, no ramp: using the classical formula (15), neglecting the ramp.</p><p>2. One marker, with ramp: using (14), neglecting the pressure-induced speed v p .</p><p>3. Two markers, no ramp: using (18), neglecting the ramp. 4. Two markers, with ramp: using (18).</p><p>We chose as a secondary marker choline, the compound with migration times closest to mid-point towards the EOF, following the conclusions of section 2.3.</p><p>In Figure 6 plot the results of the comparison against the previously known values for the mobilities of these compounds. For each of them, the mobility was computed for each run, and the maximum relative deviation with respect to the known value amongst all six replicates (three at each pressure) was taken as an indicator of the maximal potential deviation. These maximum deviations per compound were then gathered in the shown box-plots. In the case of a single marker ignoring the ramp, the variation reaches 20%. Simply including the ramp correction reduces the median to around 4% even while neglecting the correction due to pressure. Finally, using two markers and the ramp correction lowers the median to a maximal 2% deviation with respect to the known values.</p><p>At that point, the deviation may come as much from inaccuracies in the present determination as from inaccuracies in the reference values, the latter derived with the traditional single-marker formula. In Figure 7 we plot the coefficient of variation (standard deviation over average) of the mobility of each compound over the six replicates (again, three at each pressure). The spread in migration times is high, which is to be expected from running each half of the replicates at different pressures. The conversion to mobilities, even with the single marker formula and no ramp correction reduces the variability to little more than 2%. This does not change with the addition of the ramp correction, meaning that the large deviation in Figure 6 is caused by a systematic shift, as one would expect. But the addition of a second marker reduces the variability further to less than 0.5% between the six runs at different pressures. In this light, the deviations of ∼ 2% of the two-marker formulas in Figure 6 are most likely due to variability in the original determination. These new mobility values for the chosen standards are available in the supplementary material. Notice still that this variability happens within the same set of experiences, and somewhat higher variability should be expected in inter-laboratory comparisons.</p><p>As one would hope from the elimination of all instrumental parameters from the formula, the use of the second marker improves the precision of the mobilities of the compounds by about a factor of ∼ 4. For this reason, if a reliable second marker is present in the sample, we strongly recommend using two markers to determine the mobility of compounds separated with CE.</p><!><p>Our second aim is to assess the suitability of the mobility transformation for quantitative CE. All experiences under this section used the 15 selected compounds, and a milder ramp of t R = 6 s.</p><p>First, to make the choice of correction as per Table 1, we determined the regime of the ESI source. The mix of 15 compounds was analysed under four different pressures (30, 50, 70, and 90 mbar), each run in triplicate, resulting in a array of migration times t c,p,i of migration times for each compound c, pressure p and replicate i, and another one of peak areas A c,p,i . To follow each compound along the different pressure, the replicates were averaged out,</p><p>and in order to compare the compounds against each other, normalized by their own mean over all pressures,</p><p>The same transformations were applied to obtain an array of normalized areas per compound and pressure, A</p><p>. These normalized values track only the variation between the free parameter (in this case, the pressure) relative to the compound's overall mean, so that if for some compound t our case (×t 2 , following Table 1) the CVs remain the same between the peaks in the electropherograms and the ones in the mobilograms (∼ 5%). If the wrong correction is made, by assuming that the ESI operates in concentration mode, leading to a factor of ×t, the peak areas show about three times more variability. It is only made worse by making no correction whatsoever.</p><p>Figure 11 shows the electropherogram for one of the runs at 10 mbar, and the corrected mobilogram obtained after conversion by ROMANCE, showcasing how the relative changes in width are compensated with the peaks' height.</p><p>In summary, choosing the right correction is critical to obtain reproducible areas, in which case the transformed mobilograms will perform just as well as the electropherograms. Of course, while having the advantage of permitting the identification of peaks by their position, using libraries of known mobilities.</p><!><p>The development of simplified assays for safety assessment is a key element within the changes taking place over the last decade in the field of chemical toxicology testing. By moving from the classical observation of apical endpoints in animals towards cheaper and faster assays performed on cell cultures, it is possible to cut costs and save time, paving the way to increased-throughput testing [15]. When it comes to toxicity assessment of molecules with neurotoxic potential, astrocytes are an appealing model system, since their activation upon exposure to different neuroinflammatory triggers can take them to either neurotoxic or neurotrophic states [16]. To check this approach, and as a first proof-of-concept, 2Dcultures of astrocytes exposed to different natural neuroinflammatory triggers, namely interleukin 1β (IL-1β), tumoral necrosis factor α (TNFα), and lipopolysaccharide (LPS). In order to study polar metabolites involved in these processes, and as a showcase of the full ROMANCE workflow on actual metabolomics data, we have analyzed the astrocyte samples using the previously described CE-MS approach. First, raw Agilent .d files were transformed to the open format mzML. Then the files were either directly imported into the Progenesis QI peak-picking software, or transformed (and area-corrected) into the effective mobility scale by ROMANCE prior to the peak-picking step. Peaks were manually reviewed to ensure correct identifications against an in-house library, producing a set of 38 identified features in ESI+ mode and 28 in ESI− mode, common to both electropherogram and mobilogram peak extraction. To compensate sample amount variability in the samples, probabilistic quotient normalization (PQN) was applied to both datasets, a widely used normalization technique in metabolomics [17]. Drift and other analytical effects were in turn corrected with the inclusion of quality control (QC) samples [18] to ensure analytical consistency, used to apply a principal component based correction to cancel out the sources of variability between the QCs [19].</p><p>Running a principal component analysis on the peaks extracted from the electropherograms, we obtain the score plot shown in the top half of Figure 12. We can observe that each sample set is well clustered, and that the first component captures the largest part of the inflammatory triggers' effect on the metabolic status of the astrocytes, distinguishing all three treatments from the control group. Additionally, the second component finds an effect separating the TNFα from the other two groups (IL-1β and LPS), which remain clustered together.</p><p>The lower half of Figure 12 shows a PCA of the same samples after conversion by RO-MANCE, using mass-mode area correction. Following the expectations from the results of Figure 10 on standards, the score plots are fundamentally equivalent before and after conversion. The corresponding identified metabolites, together with their loadings, are available in the supplementary material. Of course, mobilograms offer the advantage of allowing reliable identification based on external libraries, enlarging the amount of metabolites that can be identified without needing to resort to the evaluation of in-house libraries of standards.</p><p>The list of identified peaks and their corresponding loadings from the mobility data are available in Table C in the supplementary information.</p><!><p>We have seen that the transformation of CE data to the electrophoretic mobility scale not only improves peak identification, as was already known [6], but it also allows quantitative information to be extracted from mobilograms.</p><p>ROMANCE v2 has been introduced to perform these corrections, and also give more control to the user over the transformation parameters including the possibility of using multiple markers, field ramps, and selecting different ionization and detection regimes. We have validated the theoretical framework by studying peak position and area precision under the several transformation formulas shown in the article, showing the need to use the right area transformation to have reliable quantitative data. Finally, we have seen that with the current version of ROMANCE the worfklow is ready for multivariate analysis of real metabolomics data, achieving a significant milestone in the path to make CE-MS part of the metabolomics toolkit.</p>

ChemRxiv

The effect of quench agent on urine bioassay for various radionuclides using Quantulus\xe2\x84\xa21220 and Tri-Carb\xe2\x84\xa23110

Following a radiological or nuclear incident, the National Response Plan has given the Department of Health and Human Services / Centers for Disease Control and Prevention the responsibility for assessing population\xe2\x80\x99s contamination with radionuclides. In the public health response to the incident, valuable information could be obtained in a timely and accurate manner by using liquid scintillation counting techniques to determine who has been contaminated above background for alpha and beta emitting radionuclides. The calibration plays a major role in this process therefore, knowing the effect of quench agents on calibration is essential.

the_effect_of_quench_agent_on_urine_bioassay_for_various_radionuclides_using_quantulus\xe2\x84\xa212

Introduction<!>Reagents and materials<!>Instrumentation and labware<!>Sample preparation and LSC analysis<!>Results and Discussion<!>Quench (efficiency) curves and LSC activities results of urine spikes<!>Conclusion

<p>Normally the signal in counts per minute (CPM) is converted into activity units of becquerels per liter (Bq/L) by means of quench (efficiency) curve for a given radionuclide. The quench curve connects the quenching factor (SQP(E) - the Spectral Quench Parameter of the External standard or tSIE – transformed Spectral Index of the External standard, depending on the LSC instrument) with the instrument efficiency, that is the observed activity (CPM) divided by the added activity (DPM). The typical quench agents are nitromethane, nitric acid, carbon tetrachloride, and toluene [1,2]. Nitromethane is a good quench agent for high energy beta emitters as Sr-90 and alpha emitters as Am-241. However, in urine bioassay measurements for low energy radionuclides such as tritium (H-3), nitromethane is not the optimum choice. Tritium is more sensitive to quenching. Some researchers used carbon tetrachloride with yellow food dye [3] for urine tritium bioassay or tried to reduce the color quenching by ultraviolet photolysis [4]. Both approaches have limitations: carbon tetrachloride with yellow food dye is not the best match for urine while quenching decrease using ultraviolet photolysis creates an additional step in sample preparation as well as photoluminescence peak in spectra. For tritium bioassay we evaluated other possible quench agents such as black tea [5] by itself and black tea with addition of either urea or nitromethane. The choice and the amount of a quenching agent depends on the instrument: the Quantulus™1220 requires more quenching agent than Tri-Carb™3110. This work is devoted to the study of the effect of quench agent on the urine bioassay for such radionuclides as Am-241, Sr-90/Y-90, H-3, P-32 using Quantulus™1220 and Tri-Carb™3110.</p><!><p>For gross alpha/beta analysis we used the Ultima Gold® AB cocktail (UGAB) from PerkinElmer Company; 99% Nitromethane from ACROS Organics; black tea solution (Lipton black tea from any grocery store, 1 regular tea bag steeped in 200 mL of boiling water for 10 min, the cooled solution was used for quenching); and 99% Urea from ACROS Organics. Deionized (DI) water was used for all solutions (≥18 MΩ∙cm, from an Aqua Solutions Ultrapure Water System, Aqua Solutions, Inc.). "Base urine" was collected through anonymous human donations (according to Centers for Disease Control and Prevention Institutional Review Board protocol 3994) and acidified to 1% HNO3. All radioactive source solutions were traceable to the National Institute for Standards and Technology (NIST) (Gaithersburg, MD, USA). Urine gross alpha/beta quality control (QC) materials U-GAB_Low_2015 and U-GAB_High_2015 were purchased from Eckert & Ziegler Analytics, Inc. They are base urine samples spiked with Am-241 and Sr-90/Y-90 at low and high levels (see Table 4). Reference Material (RM) and High Calibration Range material (HCR) were prepared in our laboratory by spiking base urine with NIST traceable reference solutions of Am-241 and Sr-90/Y-90 (for gross alpha/beta analysis), P-32 (for P-32 analysis) or H-3 (for tritium analysis). QC materials for tritium analysis (LU12318 and MU12319) were prepared in our laboratory by spiking base urine with H-3 NIST traceable reference solution at low and high levels (see Table 6). All reference solutions used for spiking were purchased from Eckert & Ziegler Analytics, Inc.</p><!><p>For this study we used two ultralow level liquid scintillation spectrometers Quantulus™1220 (#2 and #3) and two Tri-Carb™3110 (#1 and #2) (all from PerkinElmer Company) for Am-241, Sr-90/Y-90, H-3, and P-32 analysis in alpha/beta mode; 20-mL LSC plastic vials (PerkinElmer Company) for LSC analysis; a high precision analytical balance capable of accuracy weighing 0.0001 gm (Mettler-Toledo, LLC); 15-mL and 50-mL conical polypropylene tubes (Becton Dickinson Company) for solution preparation; a Brinkman bottletop dispenser with capacity from 5 mL to 25 mL (Brinkman Instruments, Inc.) for cocktail dispensing; and four electronic pipettes with total volume range from 5 μL to 5 mL (Eppendorf, Inc).</p><!><p>First, we determined the optimal PSA (pulse shape analysis) or PDD (pulse decay discriminator) settings for a urine matrix using base urine [6]. Next we built quench curves according to procedure [6] for each nuclide at optimal PSA or PDD settings on each instrument using quench agents such as nitromethane, DI water and nitromethane added, black tea and 10% urea mixture, and black tea and 5% nitromethane mixture as described in Tables 1, 2. For the Am-241 and Sr-90/Y-90 quench sets we used 20 mL of UGAB cocktail and 15 mL of UGAB cocktail for P-32 and H-3 quench sets. Then we optimized such parameters as sample analysis time, external standard analysis time, type of cocktail, and sample/cocktail volume for 20 mL vial geometry [6]. In addition, a region of interest (region in which the given nuclide will be counted) was optimized based on spectra for each nuclide from each instrument, since each instrument needs its own optimization depending on the nuclide of interest. Finally, for sample preparation we mixed 5 mL of urine sample with 15 mL of UGAB cocktail in 20-mL LSC plastic vials till a uniformed state was reached. Then LSC vials with solutions were placed on the LSC counter tray and LSC analysis was performed using parameters described in Table 3.</p><!><p>In this study we present the analytical results obtained using the optimal quench agents for each nuclide of interest (Tables 1 and 2). Black tea and urea were used as they are a better match for urine by color and chemical quenching. This is important for low energy nuclides which are more sensitive to quenching. The criteria for choosing the quench agent were activity results by LSC from known urine spikes.</p><!><p>We chose Quantulus™1220 #2 and Tri-Carb™3110 #1 to show the examples of quench curves for each type of instrument.</p><p>Figures 1 and 2 represent quench curves collected from Quantulus™1220 instrument using 20 mL of UGAB cocktail and nitromethane (from 0 to 0.5 mL) as the quench agent. Both quench curves are fitted to polynomial equation third degree. Figures 3 and 4 represent quench curves collected from Tri-Carb™3110 using 20 mL of UGAB cocktail and nitromethane (from 0 to 0.3 mL) as a quench agent. The quench curve for Sr-90 is fitted to polynomial equation third degree while the quench curve for Am-241 is fitted to exponential equation.</p><p>Gross Alpha/Beta Quality Control Materials (QC), Reference Material (GAB-RM) and High Calibration Range Material (GAB-HCR) results were used to evaluate the gross alpha/beta analysis using the built quench curves. The results by Quantulus™1220 and Tri-CarbTM3110 are presented in Table 4. All activities are shown in Bq/L. The use of a nitromethane quench agent with Ultima Gold® AB cocktail for quench curve preparation gave the activity correlation in the range of ± 5% for gross alpha nuclides and ± 2% for gross beta nuclides when comparing the observed and target value data. The change of the quench agent for others to include black tea and nitromethane or water and nitromethane did not give any benefit for both types of instruments.</p><p>However, for P-32 the best quench agent was nitromethane with 5 mL of DI water for Quantulus™1220. An example of a P-32 quench curve is presented on Figure 5. This quench curve was produced on Quantulus ™1220 using 15 mL of UGAB cocktail, DI water (5 mL – 4.7 mL), and nitromethane (0 mL – 0.3 mL) as a quench agent. For Tri-Carb™3110 the best quench agent for P-32 was black tea - 5% nitromethane mixture diluted with water at different ratios (total volume of 5 mL) mixed with 15 mL of UGAB cocktail. An example of a P-32 quench curve built on Tri-Carb™3110 is presented on Figure 6. Both quench curves for P-32 are fitted to polynomial equation third degree. Table 5 represents the results of P-32 urine spikes analysis using Quantulus and Tri-Carb instruments' quench curves with the optimal quench agents for each instrument. Base urine (BU) is not spiked and the solutions BU-P32–5K through BU-P32–1M are base urine spiked with P-32 in the range of 5 000 Bq/L – 1 000 000 Bq/L. The bias between found and target activity is in the range of ± 3% for all instruments. These results confirm the optimal choice of quench agents for P-32 urine bioassays for both types of instruments.</p><p>Figures 7 and 8 demonstrate the H-3 quench curves using the best quench agent: mixture of black tea - 10% urea diluted with DI water (total volume 5 mL), and UltimaGold®AB cocktail (15 mL) as shown in Tables 1 and 2. Both quench curves for H-3 are fitted to polynomial equation third degree. The Table 6 represents the characterization results for QC, RM, and HCR materials using optimal quench curves on both types of instruments. Both types of the instruments show the correlation between found and target activity in the range of ± 4%, which means the black tea with 10% urea imitates urine better than nitromethane with water or just black tea.</p><!><p>This work demonstrates the importance of the proper selection of appropriate quench agents used for urine bioassay with careful evaluation of the nature of nuclide and type of the instrument. We showed that for our gross alpha/beta urine screening, nitromethane in Ultima Gold® AB cocktail is optimal for radionuclides such as Sr-90/Y-90 and Am-241 while using either the Quantulus™ 1220 or Tri-Carb™3110 instrument. For the P-32 urine bioassay, DI water with nitromethane is the most appropriate quench agent (Table 1) while using the Quantulus™1220 whereas DI water, black tea, and nitromethane as a quench agent (Table 2) yields better results on the Tri-Carb™3110. For a tritium urine bioassay the best quench agents are DI water, black tea, and urea for both instrument types.</p>

PubMed Author Manuscript

Rhodium Catalyzed Stereoselective Mono-alkenylation of Aryl sp 2 C-H Bond via C-N Bond Cleavage: N-allylbenzimidazole as Strategic Alkenylating Agent

A Rh-catalyzed C(sp 2 )-H alkenylation has been achieved by taking N-allylbenzimidazole as an allylamine congener. This distinctive transformation has been observed for the first time which is attributed to the rigid benzimidazole unit. Lewis acid assisted cleavage of C(sp 3 )-N bond by coordinating to the N3 of N-allylbenzimidazole has been established. Thus, herein we have demonstrated an unprecedented protocol of domino C-N bond cleavage followed by aryl C(sp 2 )-H alkenylation. Further, detailed mechanistic studies, control experiments have been conducted to understand the mechanism. The rhodacycle-intermediates involved in the reaction have been isolated and characterized through NMR, HRMS, and single crystal X-ray.

rhodium_catalyzed_stereoselective_mono-alkenylation_of_aryl_sp_2_c-h_bond_via_c-n_bond_cleavage:_n-a

INTRODUCTION<!>2<!>RESULTS AND DISCUSSION

<p>Method development for the formation of C-C bond has been the foremost topic of research in organic synthesis. Transition metal catalyzed methodologies involving organo-halides, alcohols, alkanes, olefins have played prominent role for the construction of new C-C bonds. [1][2][3] In this context, the transitionmetal catalyzed Tsuji-Trost reaction 4a has evolved as an efficient methodology for allylation of organo-nucleophiles by using allyl halides, 5 allyl alcohols, 6 and allyl ester derivatives 7 as electrophilic component (Figure 1a). [4][5][6][7] Here, the nucleophile attacks the metal(π-allyl) intermediate, which was formed after the cleavage of C-X (X=-halogen/-OR) bond. It has been observed that, the allyl alcohol derivatives with a better leaving group shows a positive cooperation for the allylation reactions. 7 As compared with C-O/C-halo bond, the C-N bond is thermodynamically more stable, which is attributed to its high bond dissociation energy. 8 As the cleavage of C-N bond is very difficult and challenging, allylamines are less explored as electrophilic component. Different strategies have been employed in order to activate the robust C-N bond of allylamines. 9 The most commonly used strategies are (i) strong Lewis acid catalysis, and (ii) hydrogen bonding interaction (Figure 1b). In both of these strategies, once the allylic cation is generated, it coordinates to the metal and forms metal π-allylic cation. This in the presence of active nucleophile delivers the allylated product (Figure 1b). Tian and co-workers have exploited the Lewis acid catalysis for the coupling of allylic amines, and boronic acids. 10 In their reaction, boric acid plays a crucial role triggering the C-N bond cleavage of allylamines. Further, this concept has been extended to the synthesis of wide range of structurally diverse chiral sulfones. 10 In 2011, the Zhang research group had discovered a Pd-catalyzed α-allylation of aldehydes and ketones effectively via the C-N bond cleavage assisted by hydrogen-bonding interaction in protic solvent. 11 This methodology worked efficiently with primary, secondary as well as tertiary amines. Substrate having active methylene, and methine unit were subjected for allylation smoothly from allylamine derivatives via C-N bond cleavage. 12</p><!><p>During the last few decades, transition metal catalyzed directed C-H bond functionalization has evolved as a powerful tool for a step and atom economic transformations. 13 However, a parallel C-H activation and C-N bond cleavage for the C-C bond forming reaction is still in its infancy. 14 In 2018, the Wang research group have explored the allylation of 2-phenylpyridine via C-N bond cleavage of allyamines, where the protic solvent trifluoroethanol (TFE) was observed to trigger the C-N cleavage via hydrogen bonding interaction (Figure 1c). 14a Recently, Gooben research group have successfully achieved the orthoallylation of benzoic acid using N,N-dialkylallylamines as the allylating agent (Figure 1c). 14b Here, protic solvent was found to be compatible, enhancing the reactivity. An interesting feature in the catalysis field is that, a slight change in the electronics of the substrate and the reaction condition could deliver a completely different product. Therefore, we decided to study the reactivity of N-allylbenzimidazole 1a as an allylamine congener (Figure 1d). The basic difference between 1a and the reported allylamine is that, in the case of allylamines the non-bonded electron pair on N1 is readily available for protonation/hydrogen bonding with protic solvent. Whereas, the non-bonded electron pair over N1 of 1a is not available for hydrogen bonding or bonding with Lewis acids as it is a part of the aromatic sextet. However, the non-bonded electrons over N3 atom of 1a could be used for this purpose in lieu of N1 atom. Upon activation (Lewis acid/H-bonding) of 1a, the aromaticity of the non-benzenoid ring would get perturbed (Figure 1d). Now it's interesting to see, whether this disturbed non-benzenoid aromatic unit which is attached with C1, will show the same allylation chemistry or different? With these basics, we anticipated to study the chemistry of N-allylbenzimidazole for directed C(sp 2 )-H functionalization. We carried out the reaction between 2-arylpyridine derivatives 2 and N-allylbenzimidazole 1a under rhodium catalysis. To our delight, we observed selectively C(sp 2 )-H alkenylation as opposed to the allylation. Salient features of this methodology are (i) orthogonal reactivity by N-allylbenzimidazole, (ii) selective mono-alkenylation instead of allylation, (iii) first report on C-N cleavage of N-allyl benzimidazole with Rh-catalysis, (iv) use of N-allylbenzimidazole as alkenyl surrogate, (v) detailed mechanistic study, (vi) characterization of Rh-intermediate, and (vii) exclusive trans alkenylation.</p><!><p>Our investigation began with the reaction of N-allylbenzimidazole 1a and 2-(4-chlorophenyl)pyridine 2f (Table 1). We were delighted to find that 5 mol % of Cp*Rh catalyst, 2 equiv of LiClO4 in combination with 1.5 equiv of Zn(OAc)2 gave the desired mono-alkenylated product 3af in 75% of yield (Table 1, entry 1). The use of cationic Rh-complex resulted in 31% yield of 3af (Table 1, entry 2), whereas Rh(OAc)2 dimer and Wilkinson's catalyst failed to deliver the product (Table 1, entries 3-4). When solvents other than TFE were screened for the reaction, lower yields were observed (Table 1, entries 5-7). This suggest that the protic solvent TFE playing a crucial role in the reaction. It was observed in the literature that the use of water could enhance the hydrolysis of C-N bond. 15 Therefore, to enhance the rate of C-N bond cleavage of N-allylbenzimidazole, 1:1 ratio of TFE:H2O was used (Table 1, entry 8). Instead of improved yield, we observed only a trace amount of product suggesting the need of moisture-free condition for the above transformation. Further, the rate of the reaction is highly affected by the temperature; an exponential increase in the reaction yield was observed with increasing temperature (Table 1, entries 9-11). We are surprised to observe that, LiClO4 works wonderfully for this designed protocol, replacing costly silver additives such as AgSbF6 and AgOAc which results in no reaction (Table 1, entries 12, and 13). In addition to that, use of NaIO4 in place of LiClO4 resulted in 20% yield of the product 3af (Table 1, entries 12-13). Varying the equivalents of N-allylbenzimidazole resulted in lower yields (Table 1, entries 15-16). This might be due to N-allylbenzimidazole is going through a multi-step process, that is C-N bond cleavage and isomerization. Table 1. Optimization of Reaction Conditions a Reaction conditions: 2f (1 equiv, 0.06 mmol), 1a (3 equiv, 0.18 mmol), [Cp*RhCl2]2 (5 mol %, 0.003 mmol), LiClO4 (2 equiv, 0.12 mmol), Zn(OAc)2 (1.5 equiv, 0.09 mmol), TFE (0.1 M, 0.6 mL), 130 °C, N2, b Isolated yield. c (20 mol%, 0.2 equiv) of silver additives were used, d Isolated yield after 12 h. Further evaluation of lewis acid/acid additive such as Zn(OTf)2, PivOH, and Cu(OTf)2 didn't result in an improved yield of 3af (Table 1, entries 17-19). To know the effect of time, three parallel reactions were performed, and it was observed that after 4 hours the product starts to decompose under the reaction conditions (Table 1, entries 20-22). Finally, the control experiments confirmed the necessity of catalyst [Cp*RhCl2]2, additive LiClO4, and Zn(OAc)2 (Table 1, entries 23-25). From the experiments it is clear that the reaction is triggered by the addition of Lewis acid. Thus, it is confirmed that the role of LiClO4 is crucial for this reaction and Zn(OAc)2 acts as a promoter. With the optimized condition in hand, we proceeded to study the electronic influence of the N-allyl coupling partner, for C-H alkenylation of 2-arylpyridines. When, 1a containing electron donating group (-OMe) 1b, and electron withdrawing group (-NO2) 1c were screened, similar reactivity was observed yielding 50% of 3af. It indicates that, -OMe/-NO2 substituent in the benzenoid system has no remarkable impact for this transformation. When 2-methyl-N-allylbenzimidazole 1d was taken as an alkenylating source, we got 41% yield of 3af. Further, to check the influence of benzenoid ring, N-allylimidazole 1e and allylamine 1f, were taken instead of 1a, but inferior result was observed in both cases. This implies that the presence of benzene ring is necessary for this transformation. Disubstituted alkenes (1g and 1h) could not deliver the respective alkenylated products. It indicates that, alkene insertion in to the C-Rh bond is sterically controlled and occurs prior to the C-N bond cleavage. To check the role of N3 nitrogen atom of 1a, N-allylindole 1i was employed as the coupling partner. In this case, we did not observe any product 3af, which implies that the reaction is facilitated by the chelation of Lewis acid at N3 atom of 1a. Further, N-Allyl indoline 1j was also tested and found to be ineffective for this transformation. When more electron deficient Nallyl phthalimide 1k, and N-allyl isatin 1l were chosen as alkenylating agents they failed to deliver the product 3af. The use of 1,3-diallylbenzimidazole 1m and N-allyl-4-bromopurine 1n gave the mixture of alkenylated as well as allylated products in poor yields. In contrast to imidazole 1e, N-allyl pyrazole 1o did not give the product 3af. Moreover, aryl pyridine 2a was subjected to the standard reaction condition with the more frequently used allylating reagents such as allyl alcohol 1p, and allyl ethyl carbonate 1q, but none of them could produce C-H allylation or C-H alkenylation product. All these studies discussed above confirms the efficiency and selectivity of N-allylbenzimidazole 1a for this transformation. To test the generality of this methodology by using N-allylbenzimidazole 1a as alkenylating surrogate, various substituted 2arylpyridines were tested. The scope of 2-arylpyridines were outlined in Scheme 2a. The aryl unit containing both electron donating groups EDGs (-CH3, -C2H5, -OMe, -F/Cl) and electron withdrawing groups EWGs (-CHO, -COCH3, -CF3, -CO2Me) were well tolerated under this condition delivering moderate to very good yields of the respective C-H alkenylated products. It has been observed that, the substrates with EDGs were giving lesser yields (Scheme 2, 3ab-3af, 3ak, and 3al) as compared to the substrates with EWGs (Scheme 2, 3ag-3aj). Interestingly, sensitive functional groups such as -formyl 3ai and -ester 3aj were retained in the final product. The unsymmetrical substrate bearing dioxolane ring selectively gave 3al in 67% yield, by activating ortho-hydrogen from more sterically hindered site. The origin of this selectivity might be due to the additional stability gained from the chelation of oxygen atom in the cyclometallated intermediate. 16 Further, the scope of the reaction has been extended to the heterocycles such as 2-arylpyrimidines, and 2arylpyrazoles (Scheme 2, b and c). These heterocycles were found very cooperative to deliver their corresponding monoalkenylated products without any variation in the standard reaction conditions. The substrates bearing EDGs or EWGs worked smoothly, giving products in good yields (Scheme 2, 7aa-7ae, and 8aa-8ae). Purine unit has a special role being as a nucleobase, and a core unit in nucleic acid. Transition metal-catalyzed purine directed C-H alkenylations have been reported using phenylacetylene or vinylcarboxylic acids by Yu 17a and Xu group 17b respectively. We envisioned that, our protocol could install selective alkelyl unit in this system. Gratifyingly, this reaction condition was found viable for purine directed alkenylated products 9aa, and 9ab in good yields (Scheme 2d). The trans-stereochemistry was confirmed unambiguously from the single crystal X-ray analysis of product 9ab (CCDC 2077201). In order to get better understanding about the influence of electronics on the substrate, intermolecular competition experiments were conducted between different arylpyridines (Scheme 3). The results indicate that, electronically poor substrates were reacting relatively faster than electronically rich substrates with the reactivity trends 2g>2a>2d (Scheme 3a and 3b). In order to check the feasibility of di-alkenylation under the standard reaction condition, 3aa was employed as a substrate. However, we did not observe any di-alkenylation product, rather 94% of 3aa was recovered (Scheme 3c). It shows the highly selective induction of mono-alkenyl group into the substrate. To gain further insight into the mechanism, we conducted several mechanistic experiments (Scheme 3d-3k). When 2f was allowed to react with deuterium source D2O or CD3OD in absence of 1a, 13% and 11% of deuterium exchange were observed with D2O and CD3OD respectively (Scheme 3d). Additionally, the reaction of 2f and D2O in presence of 1a shows 30% H/D-scrambling at the ortho-position of 3af (Scheme 3e). Both of the experiments together indicate that, the C-H bond metalation step might be reversible. 18 The reaction of 2f with stoichiometric amount of [Cp*RhCl2]2 under the standard reaction condition in absence of 1a yielded rhodacycle Int-1 in 70% yield, which was characterized by NMR spectroscopy, and HRMS (Scheme 3f).</p><p>Similarly, rhodacycle Int-2 was synthesized from the substrate 2l, which was confirmed by NMR spectroscopy, HRMS, and X-ray crystallography (Scheme 3g) (See supporting information). The active involvement of Int-1 in the catalytic cycle was confirmed when 5 mol% of Int-1 was used as catalyst for the reaction of 2f with 1a, afforded 61% yield of 3af (Scheme 3h). There are several reports on transition metal-catalyzed in situ isomerization of terminal alkene to internal alkene. 19 Thus, we envisaged whether internal alkene 1a' is an active coupling partner in the course of this reaction or not. To rule out this possibility, a reaction has been performed employing 1a', resulted in no reaction (Scheme 3i). This confirms that, the terminal alkene 1a is participating in the reaction not the internal alkene 1a'. The formation of 3aa was observed even in the presence of stoichiometric amount of radical scavenger BHT and TEMPO in 75% and 46% respectively; which rule out the involvement of radical mechanism (Scheme 3j). Furthermore, to check whether the reaction is proceeding through 2'-allylphenylpyridine 3aa 1 as an initial product, it was subjected to the standard condition, which resulted in 67% of 2a and 18% of 3aa (Scheme 3k). These results are well supported by the literature reports. 20 As, 2-phenyl pyridine 2a is the major product not 3aa under the rhodium catalyzed condition, it shows that, 2'-allylphenylpyridine 3aa 1 is not a key intermediate in the reaction. This has been further verified by performing a series of parallel experiments and the reaction progress has been checked in NMR at different time (Figure 2, for details see Supporting information). Throughout the series of experiments, the peaks correspond to the allyl group.</p>

ChemRxiv

Using Small Molecules to Dissect Mechanisms of Microbial Pathogenesis

Understanding the ways in which pathogens invade and neutralize their hosts is of great interest from both an academic and clinical perspective. However, in many cases genetic tools are unavailable or insufficient to fully characterize the detailed mechanisms of pathogenesis. Small molecule approaches are particularly powerful due to their ability to modulate specific biological functions in a highly controlled manner and their potential to broadly target conserved processes across species. Recently, two approaches that make use of small molecules, activity-based protein profiling (ABPP) and high-throughput phenotypic screening, have begun to find applications in the study of pathways involved in pathogenesis. In this review we highlight ways in which these techniques have been applied to examine bacterial and parasitic pathogenesis as well as discuss possible ways in which these efforts can be expanded in the near future.

using_small_molecules_to_dissect_mechanisms_of_microbial_pathogenesis

Introduction<!>Using chemical genetic screens to study components of pathogenesis<!>Bacteria<!>Parasites<!>Using activity-based probes to survey components of pathogenesis<!>Bacteria<!>Parasites<!>Clinical benefits of using small molecules to study microbial pathogenesis<!>Conclusion

<p>Microbial pathogenesis, the collection of mechanisms through which an infectious agent causes disease, is a dynamic process involving concerted interactions between the etiological organism and host. Due to the highly regulated nature of the infection cycle, small molecules that modulate or monitor enzyme activity levels are ideal tools for studying pathogenesis.</p><p>In microbial pathogenesis distinct environmental cues often trigger key events such as host cell invasion that are required for disease progression. Small molecules often act rapidly with remarkable spatial and temporal control, allowing these events to be dissected in real-time. Genetic manipulation of targets, on the other hand, often is not possible on a timescale that is sufficient to monitor distinct processes with a high degree of temporal resolution. Even conditional knock down techniques such as RNA interference are limited by the speed of protein translation and degradation before activity levels can be sufficiently changed. In addition, many pathogens lack the machinery necessary for such methods to be used. Together, these limitations underscore the benefits of using pharmacological approaches to study pathogenesis.</p><p>In addition to being used to perturb protein function, small molecule probes can be used to profile the activity levels of enzymes involved in different stages of virulence processes. These compounds, known as activity-based probes (ABPs), give an indirect readout of enzymatic activity within a complex system without the need for biochemical purification of individual components. Because many enzymes involved in processes such as host-pathogen interactions are difficult to purify or express recombinantly, this method is extremely powerful for interrogating the precise function of a given factor in pathogenesis. Additionally, many proteins are post-translationally regulated, making it impossible to monitor their activity by examining transcript or protein expression levels alone. Therefore, ABPs are capable of providing a more complete view of a pathogenic mechanism.</p><p>Although small molecules have proven to be valuable for studies of a wide range of basic biological processes, they have primarily been used in eukaryotic systems. This is partially because genetic tools are particularly robust in many prokaryotes systems. However, some clinically relevant bacteria are either difficult to genetically manipulate or are genetically intractable. One such pathogen, the anaerobic, Gram-positive bacterium Clostridium difficile, is highly antibiotic-resistant and frequently causes a severe infection of the colon in hospitalized patients. Another is the obligate intracellular bacterium Chlamydia trachomatis, a causative agent of the sexually transmitted disease that bears its name. Similarly, protozoan parasites are often obligate intracellular pathogens and are not easily manipulated by genetic means. Some species have poor DNA transfection efficiencies and homologous recombination rates and/or are not amenable to RNA interference for gene knockdown. Therefore, there is a clear need for alternative methods in a number of biologically and medically relevant organisms.</p><p>While small molecules have been used in the past to study microbes, they primarily have been used to kill the organism. However, with the development of more sophisticated screening methods, it is now clear that specific small molecules can be used to target discrete steps in pathogenic mechanisms of interest. Thus, small molecules must also be viewed as tools that can provide new information regarding the biochemical aspects of virulence. Small molecules therefore have great potential to shed light on the details of pathogenesis that will likely help guide our efforts to treat infectious diseases.</p><!><p>In order to functionally dissect biological systems, small molecules must be able to selectively modulate the activities of specific pathway components. This is the principle behind chemical genetics, wherein chemical agents perturb the function of a given gene product with the same degree of specificity as traditional genetic manipulation. To discover suitable small molecules, high-throughput screening methods are traditionally employed. These screens typically can be divided into two classes: forward and reverse chemical genetic screens (Figure 1) (for review see (1)).</p><p>In a forward chemical genetic screen, libraries of small molecules are added to a relevant biological system to search for a desired change in phenotype (Figure 1a). In the case of pathogenesis, examples include changes in the kinetics of invasion, evasion, or host cell death. Once a desired compound is identified based on its activity, the target of this small molecule is then identified. Forward chemical genetic screens, also known as phenotypic screens, have some notable advantages. First, no assumptions are made regarding the importance of individual elements in a pathway; the screen identifies the critical components. Second, if whole organisms are used in the screen, the "hits" are pre-selected to be cell permeable and nontoxic, thereby facilitating their use for subsequent biological assays. However, the most significant downside of phenotypic screens is that identifying and validating the target of the small molecule is often an arduous task. In an effort to address this pitfall, several techniques have been developed to aide in target identification. One recent example combines the mass spectrometry approach SILAC (stable isotope labeling with amino acids in cell culture) with binding assays (2), while another uses massively parallel sequencing to identify causative mutations in bacteria resistant to compounds of interest (3).</p><p>Reverse chemical genetic screens begin with a protein of interest and involve screening for compounds that modulate its function (Figure 1b). This often requires the target proteins to be recombinantly expressed in relatively large quantities and the development of a suitable in vitro assay to monitor its activity. Lead compounds identified in these screens are then applied to whole organisms to assess whether alterations of the protein function produce specific phenotypes. If the small molecule inactivates the protein, this technique is analogous to creating a genetic knockout. One of the primary advantages of reverse chemical genetic screens is that the target of the small molecule is already known and therefore a direct link between that protein and a cellular phenotype can be made. However, some of the most significant drawbacks of this type of screen are that lead compounds may lack cell permeability and/or specificity for the target protein, thus making in vivo phenotypic studies difficult.</p><!><p>Pathogenic mechanisms are often conserved in many related bacterial species. Therefore, compounds discovered using chemical genetic screens in microbes can often be used to identify and study basic virulence mechanisms in related organisms with minimal effort. One such mechanism that has been studied using small molecules is the type III secretion system (T3SS), which is used by many Gram-negative species of bacteria to inject pathogenic effectors into a host cell.</p><p>To identify compounds that could be used to study T3SS function, a luciferase-based high-throughput phenotypic screen was developed in Yersinia pseudotuberculosis. This screen identified a group of salicylidene acylhydrazides that block the expression and secretion of specific T3SS effector proteins at different phases of the pathway (see Table 1) (4). These compounds were subsequently used to study T3SS in Yersinia, Chlamydia, Salmonella, and Shigella spp., illustrating how the small molecule approach can be applied to study conserved mechanisms in related organisms (4-8). In contrast, a similar approach using genetic techniques would have required the creation of mutants in each organism of interest.</p><p>The salicylidene acylhydrazides identified in this chemical screen have helped elucidate specific functions of the T3SS beyond its known general role in pathogenesis. In the obligate intracellular bacterium Chlamydia, inhibiting the T3SS at different stages of the infection cycle partially blocked host invasion, stopped effector secretion, and also slowed intracellular chlamydial development and replication (5). Host invasion was also inhibited in Salmonella and Shigella (6, 8). These findings were made possible by the high degree of temporal control afforded by small molecule techniques. Additionally, in Salmonella and Yersinia the inhibitors disrupt motility, reaffirming that the T3SS is related to conserved flagellar export (6, 8). However, it is unclear if the T3SS is directly involved in flagellar processing or if the compounds are also targeting the highly related flagellar secretion apparatus. For all three bacterial species, determining if and how these diverse phenotypic effects are interconnected through the T3SS will require deciphering the exact mechanisms by which each compound acts. To date some of the molecules have been shown to inhibit the T3SS by disrupting needle assembly (8). This explains the observation that, in some cases, these compounds are incapable of inhibiting pathogenesis unless they are preincubated with the bacteria during growth.</p><p>Another common pathogenic mechanism used by many bacterial species is the deployment of toxins to disrupt host processes. Toxin pathogenesis is an intricate process involving toxin expression, secretion, host uptake and trafficking, and ultimately toxin activity leading to alterations in host cell function. High-throughput chemical genetic screens can be used to discover compounds capable of disrupting this regulated process at different stages. In one such example, a phenotypic screen was used to find small molecules that attenuate toxin expression in Vibrio cholerae, a Gram-negative bacterium that results in severely infectious gastroenteritis (9). The authors found a compound (subsequently named Virstatin; see table 1) that post-translationally inhibits the activity of the virulence regulator ToxT. Because ToxT controls the transcription of the genes encoding cholera toxin (CT) and toxin coregulated pilus (TCP), Virstatin blocks the production of both proteins. Interestingly, Virstatin is capable of preventing V. cholerae from effectively colonizing mice even though the compound fails to kill bacteria in culture. This emphasizes that Virstatin targets a pathogenic mechanism that is distinct from processes required for general viability of the bacterium. More recently a luciferase-based high-throughput screen that assayed toxin-mediated inhibition of protein synthesis was used to discover compounds that inhibit the trafficking of CT and related toxins in host cells (10). These compounds modify the trafficking of the toxins in several ways, and have highlighted the distinct retrograde flow of toxins through the Golgi in the host.</p><p>In some cases, pathogenic mechanisms used by bacteria are particularly amenable to manipulation by synthetic compounds because the pathogen itself uses small molecules as key pathway regulators. This is the case for quorum sensing (QS), a method of communication used by many Gram-negative proteobacteria to sense their population density. Several species use mechanisms to trigger virulence only when appropriate colony numbers have been reached in order to maximize the chance of overwhelming the host. This communication is accomplished using N-acylated-L-homoserine lactones, and there has been a significant body of research attempting to manipulate and understand QS using chemical biology. For example, high-throughput screening was used to discover a structurally-unrelated agonist of quorum sensing in the pathogenic strain Pseudomonas aeruginosa that acts through one of the endogenous QS receptors, LasR (11). In another example, focused libraries of synthetic structural analogues of these compounds were screened to discover modulators of quorum sensing in both Pseudomonas aeruginosa and Agrobacterium tumefaciens (12). Interestingly, only minor changes to these molecules can result in both potent inhibitors as well as compounds that act as superagonists of quorum sensing (see table 1). Additionally, some compounds behave as antagonists at low concentrations and agonists at higher concentrations. Several of the identified inhibitors were found to block the production of an essential virulence factor in P. aeruginosa (12). More recently inhibitors of a QS transcription factor have also been discovered, adding to the arsenal of compounds that can be used to dissect this pathogenic mechanism (13).</p><!><p>Diseases caused by obligate intracellular protozoan parasites are often the result of multiple rounds of host cell invasion, parasite replication, and host cell lysis. Understanding the mechanisms underlying each of these events is therefore paramount for dissecting pathogenesis. High throughput screens have been utilized successfully to this end. A forward chemical genetic screen involving a microscopy-based assay was used to identify compounds that modulate host invasion of Toxoplasma gondii (14). T. gondii is an highly prevalent protozoan parasite that can cause neurological damage in neonates and immunocompromised individuals.</p><p>In order to identify tools that could be used to dissect the process of T. gondii invasion, Carey and co-workers developed a screen in which fluorescent parasites were added to host cells and surface labeling of extracellular parasites was used to quantify efficiency of host cell invasion using microscopy (14). Compounds were selected for their ability to alter the ratio of extracellular to intracellular parasites. Interestingly, this high-throughput phenotypic screen not only identified compounds that inhibited host cell invasion but also compounds that enhanced the process (see table 1). Further mechanistic studies demonstrated that the compounds inhibited or activated distinct aspects of invasion and that the vast majority had some effect on parasite gliding motility. The compounds also were able to uncouple the regulatory mechanisms behind constitutive and pathogenesis-associated secretion of adhesion proteins involved in host recognition and attachment.</p><p>Although the compounds identified in this screen will be valuable new tools, the identification of the molecular targets has proven difficult. Recent efforts have included modification of one of the inhibitors in an effort to create a biotinylated version suitable for target isolation by affinity purification (15). As was the case for several of the screens discussed in bacterial systems, the small molecules identified in this study were also broadly applicable, with several of the compounds inhibiting invasion by the related apicomplexan parasite, Plasmodium knowlesi.</p><p>Small molecules have also been used to study the pathogenesis of Plasmodium falciparum, the causative agent of the most deadly form of malaria. P. falciparum multiplies within red blood cells in a compartment called the parasitophorous vacuole. In order to escape and continue the infectious cycle, P. falciparum must lyse the vacuole and subsequently the host cell. In an effort to characterize host cell rupture by this pathogen, two groups conducted forward and reverse chemical genetic screens independently. In the forward chemical genetic screen, a fluorescence-activated cell scanning (FACS) method was used to identify compounds that prevented parasites from escaping the host red blood cell (16). The authors identified a chloroisocoumarin and two peptide vinyl sulfones that were capable of inhibiting this key process (see table 1). Biotin-conjugated versions of these covalent inhibitors were used to identify the subtilisin-family serine protease PfSUB1 and dipeptidyl peptidase 3 (DPAP3), an ortholog of the cysteine protease cathepsin C as the relevant targets of the compounds. In the reverse chemical genetic screen, inhibitors of PfSUB1 were identified and used to demonstrate that inhibition of PfSUB1 activity attenuated the parasite's ability to rupture erythrocytes (17). Using their respective inhibitors, both groups identified the serine repeat antigen SERA5 as a downstream target of PfSUB1. Based on these findings the mechanisms leading to SERA5 activation and subsequent parasite egress are beginning to take shape and both proteins may represent valid therapeutic targets.</p><p>Reverse chemical genetic screens are often useful for studying the function of essential genes that cannot be knocked out. This was the case for an enzyme known as calcium-dependant protein kinase 1 (PfCDPK1) that is critical for Plasmodium falciparum pathogenesis (18). The authors attempted to knock out the kinase to determine its biological function, but it was found to be essential for viability. Gene expression patterns linked PfCDPK1 to a cluster of motility genes, and subsequently a high-throughput assay was used to identify inhibitors of the protein. A 2,6,9-trisubstituted purine named Purfalcamine (see table 1) was found to potently inhibit both recombinant PfCDPK1 activity and parasite proliferation. Cultures of P. falciparum treated with Purfalcamine arrested in the late schizont stage, and the authors speculate that the compound is most likely affecting motor processes. Similarly, incubation of the compound with T. gondii cultures resulted in blockade of parasite invasion, once again demonstrating the cross-species applicability of these compounds. Thus, using a combination of genomic analysis and chemical perturbation PfCDPK1 was found to be associated with parasite motility, which is essential for host cell invasion.</p><p>In a slightly different take on a reverse chemical genetic screen, high-throughput screens have been used to determine the substrate specificity of a target enzyme. This information can then be used to rationally design a specific inhibitor for the target of interest. This technique was used in Entamoeba histolytica, a virulent protozoan parasite that causes liver abscesses and intestinal lesions. The parasite secretes a cysteine protease termed EhCP1 that is capable of cleaving extracellular matrix in host tissues. EhCP1 is also important for host immune system evasion, as it can cleave IgG and the complement protein C3. Using recombinantly expressed EhCP1 Melendez-Lopez and co-workers used a combinatorial library of fluorogenic peptide substrates to determine the substrate specificity of this protease (19). Based on the results of this screen, a selective peptidic vinyl sulfone inhibitor named WRR483 (see table 1) was synthesized and used to monitor the activity of EhCP1 in a human colon xenograft model. Blockade of EhCP1 resulted in a complete inhibition of amebic invasion, demonstrating the necessity of this cysteine protease for parasite virulence.</p><!><p>Activity-based probes (ABPs) are a rapidly emerging tool for profiling the regulation of enzyme activity levels in the context of a native cellular environment (for review see (20)). ABPs make use of chemically reactive functional groups to covalently modify the active site of a target enzyme (Figure 2). They are composed of three components: an electrophilic warhead, a region that determines specificity, and a tag for visualization and/or target pulldown. Alternatively, a small surrogate tag containing a bioorthogonal reactive group can be used that allows the bulky visualization or affinity moiety to be attached in a secondary step following target labeling by the probe (21). Depending on the tag employed, enzyme activity can be profiled using a wide variety of methodologies including autoradiography, gel electrophoresis, fluorescent scanning, microarrays, LCMS, etc. (21-23). Thus far ABPs have primarily been developed for several subclasses of hydrolases, although some ligases and transferases that rely on nucleophilic residues in catalysis have also been targeted with ABPs. Because there are numerous proteases involved in both parasitic (24, 25) and bacterial (26, 27) virulence, ABPs have a great deal of potential for profiling microbial pathogenesis.</p><p>Several of the properties of ABPs make them ideal tools for studying pathogenesis. General ABPs can be used to simultaneously monitor changes in the activity of entire superfamilies of enzymes during different stages of pathogenesis, such as during microbial growth and interaction with the host. This can lead to the identification of new protein targets that show regulatory patterns that correlate with virulence (Figure 3a). ABPs are also useful for target validation, including identification of the targets of small molecules discovered in forward chemical genetic screens such as those discussed above. Fluorescently quenched versions of probes can also be used to visualize the activation of enzymes in real-time, making them extremely valuable for examining the spatially- and temporally-regulated process of pathogenesis (28, 29).</p><p>In addition to assaying activity directly, ABPs can be used to develop selective inhibitors for specific enzymes using a method known as competitive activity-based protein profiling (ABPP), in which an ABP with broad specificity is added after inhibitor treatment to show residual activity of all labeled proteins (Figure 3b) (20, 30). In this way, one can determine the selectivity of an inhibitor by monitoring the disappearance of a targeted protein band after pre-treatment with the compound of interest.</p><!><p>There are currently relatively few examples of the use of activity-based probes to study bacterial virulence. This is likely to change as microbiologists and chemists realize the value of these tools to elucidate the roles of different enzymes in pathogenic mechanisms. For example, one highly relevant target that has not yet been explored in this way is the multifunctional autoprocessing repeats-in-toxin (MARTX) family of toxins (27). All MARTX toxins contain a cysteine protease domain (CPD) involved in autoproteolytic activation. These CPDs are activated after binding the small molecule inositol hexakisphosphate (InsP6) in the cytosol of eukaryotic host cells. One could therefore imagine using ABPs to monitor this interesting mode of activation more closely (31). Recently, inhibitors of the MARTX CPD have been identified that may provide a starting point for ABP development (31). Notably, the large glucosylating toxins of Clostridium difficile also contain this CPD and may therefore be able to be studied using the same or similar probes.</p><p>One example of the use of an ABP to monitor aspects of bacterial pathogenesis involved following the activity of the host cysteine protease cathepsin B (CatB) in macrophages infected with Salmonella typhimurium (32). For this study the authors synthesized an ABP with an epoxide warhead, azido-E-64 (see table 2), based on a previously constructed probe for cysteine proteases, DCG-04 (33). Azido-E-64 was used to examine the localization of active CatB in endocytic compartments following infection with Salmonella. The authors discovered that active CatB was absent from vacuoles containing Salmonella, suggesting that the bacterium is capable of inhibiting CatB activation. The authors surmise that this could be part of Salmonella's method evading host defense mechanisms (32).</p><p>An ABP was also used to help identify a factor critical for the growth of Chlamydia trachomatis within host cells (34). Two matrix metalloprotease (MMP) inhibitors were found to block chlamydial growth, and subsequent gene sequencing of resistant mutants revealed a point mutation in the promoter of peptide deformylase (PDF). PDF is an enzyme that uses zinc to remove an N-terminal formyl group from bacterial proteins after they are synthesized. A hydroxamate-based ABP termed AspR1 was used to confirm that these inhibitors were indeed targeting PDF, as incubation with the inhibitors competed with activity-based labeling of the enzyme. This study identifies a new target with clinical potential for battling C. trachomatis infection.</p><p>In another application, β-lactones and common antibiotic β-lactam scaffolds were used to create ABPs to label different bacterial enzymes in vivo (see table 2) (35-37). Competitive ABPP was subsequently employed to create selective inhibitors for one of these enzymes, ClpP (caseinolytic protein protease), a widely conserved serine protease associated with virulence in bacterial strains such as Staphylococcus aureus (35, 37). S. aureus is a widespread Gram-positive bacterium that can result in food poisoning or "toxic shock syndrome," which is potentially fatal. A refined ClpP inhibitor was able to block ClpP activity in S. aureus in cell culture. Using the inhibitor, invasive proteolytic and hemolytic activities were confirmed to be regulated directly by ClpP in wild-type and methicillin-resistant (MRSA) strains (35). In addition, these techniques were used to demonstrate that ClpP is upstream of several critical virulence factors in pyrogenic toxin superantigen (PTSA)-producing strains of S. aureus. These included the enterotoxins SEB and SEC3 as well as toxic shock syndrome toxin (TSST-1) (37). Because this data was collected using actual clinical isolates, these studies are highly relevant to the development new therapeutic strategies.</p><!><p>Several insights regarding parasitic pathogenesis have been made in the last decade using the broad-spectrum cysteine protease probe DCG-04 (33). These studies are therefore a testament to the value of activity-based protein profiling in studying microbial pathogenesis when the proper tools are available. DCG-04 has an epoxide warhead for covalently labeling the active site cysteines of papain fold proteases. In addition, the probe has a biotin affinity-tag for affinity purification and a tyrosine residue that can be iodinated for visualization (see table 2).</p><p>DCG-04 was first used in parasites to discover a role for the cysteine protease falcipain 1 in host cell invasion by P. falciparum (38). This enzyme is a good candidate for profiling with an ABP, as it cannot easily be recombinantly expressed and its transcription, and translation levels do not correlate with its activity. The authors used DCG-04 to profile the regulation of different cysteine proteases throughout the P. falciparum lifecycle. The highest levels of falcipain 1 activity were found in the merezoite stage, which is responsible for host invasion by the parasite. Subsequent competitive ABPP lead to the development of selective inhibitors for falcipain 1. When synchronized parasites were treated with the resulting falcipain 1 inhibitor, there was a decrease in new ring stage parasites but no effect on host cell rupture, implying falcipain 1 has a role in the invasion of new host cells (38).</p><p>Competitive ABPP was also used to develop selective inhibitors for DPAP1 and DPAP3 in P. falciparum (16). The authors used the broad-spectrum probes 125I-DCG-04 and FY01 to screen a small library of inhibitors (see table 2) (30, 39). One of the resulting compounds, SAK1 (see table 1), was then used to show DPAP3 likely plays a role in regulation of PfSUB1 maturation.</p><p>Recently DCG-04 was used to demonstrate a role for the host calcium-dependent proteases calpain-1 and -2 in host cell rupture by the apicomplexan parasites P. falciparum and T. gondii (40). Synchronized P. falciparum cultures arrest in the late schizont lifecycle stage immediately before rupture when incubated with DCG-04, and the authors used the probe to identify calpain-1 and show that it is active and membrane-associated with the correct timing for egress. Subsequently they used a variety of techniques to confirm that both P. falciparum and T. gondii cannot rupture and escape infected host cells that do not have active calpains. The authors suggest a model wherein these parasites co-opt the host proteases by triggering a calcium cascade that activates the calpains and causes them to remodel host cellular components that enable egress.</p><p>It is also possible to use multiple ABPs simultaneously to classify enzymes in complex proteomes involved in pathogenic processes. For example, such a strategy was used to identify secreted proteases involved in host skin penetration by Schistosoma spp. larvae, known as cercariae (41). Shistosoma, or blood flukes, cause Shistosomiasis, which is second only to malaria among globally relevant parasitic diseases. The authors used the probe b-nLeu-Val-Pro-Leu-P(OPh)2 (see table 2) to profile serine protease activity and 125I-DCG04 to profile cysteine protease activity in the secretomes of cercariae from different Schistosoma species. S. mansoni cercariae were found to primarily secrete serine proteases such as cercarial elastase. In contrast, S. japonicum cercariae secrete clan CA cysteine proteases with cathepsin B-like activity. This information is highly relevant for developing new classes of anti-parasitic agents for each of these pathogens.</p><!><p>Employing small molecules to dissect mechanisms of pathogenesis gives rise to the potential for additional clinical benefits beyond providing new insights in this medically relevant field of study. It is possible that using therapeutics that target virulence pathways instead of microbial viability will put less selective pressure on a pathogen and therefore reduce the chance of resistance developing to a treatment (42-44). This approach should still be efficacious in patients with active immune systems, as virulence factors have been shown to be necessary for productive colonization of the host. It should be noted, however, that this strategy requires regulators of pathogenesis that are not directly involved in cell viability. For many bacteria, pathogenicity is encoded by mobile genetic elements and is independent from essential pathways (45). However, the benefits of targeting virulence may not apply to obligate intracellular organisms such as T. gondii and P. falciparum where pathogenesis is an innate part of the lifecycle and would likely be subject to as much selective pressure as other essential survival pathways. Nonetheless, there are potential clinical benefits to using small molecules to study pathogenesis. By using approaches such as high-throughput phenotypic screens, efforts are necessarily directed towards "druggable" targets that can modulate pathogenic phenotypes (46, 47). This head start may shorten the timeline between laboratory research and the development of successful therapeutics.</p><!><p>Many aspects of chemical manipulation are ideally suited for examining the highly spatially- and temporally-regulated processes involved in microbial pathogenesis. Small molecules are therefore the perfect complement to genetic techniques in these studies. Although currently there are relatively few examples of compounds being used to functionally dissect mechanisms of microbial virulence, this is an area poised for rapid growth. This will undoubtedly occur as more chemists and microbiologists realize the potential value of high-throughput chemical genetic screening and activity-based protein profiling in microbial systems, and the potential value of the results.</p>

PubMed Author Manuscript

Synthesis and characterization of modified nucleotides in the 970 hairpin loop of Escherichia coli 16S ribosomal RNA

The synthesis of the 6-O-DPC-2-N-methylguanosine (m2G) nucleoside and the corresponding 5\xe2\x80\xb2-O-DMT-2\xe2\x80\xb2-O-TOM-protected 6-O-DPC-2-N-methylguanosine phosphoramidite is reported [DPC, diphenyl carbamoyl; DMT, 4, 4\xe2\x80\xb2-dimethoxytrityl; TOM, [(triisopropylsilyl)oxy]methyl]. The availability of the phosphoramidite allows for syntheses of hairpin RNAs with site-selective incorporation of 2-N-methylguanosine modification. Four 18-nt hairpin RNA analogues representing the 970-loop region (helix 31 or h31; U960\xe2\x80\x93A975) of Escherichia coli 16S rRNA were synthesized with and without modifications in the loop region. Subsequently, stabilities and conformations of the singly and doubly modified RNAs were examined and compared with the corresponding unmodified RNA. Thermodynamic parameters and circular dichroism spectra are presented for the four helix 31 RNA analogues. Surprisingly, methylations in the loop region of helix 31 slightly destabilize the hairpin, which may have subtle effects on ribosome function. The hairpin construct is suitable for future ligand-binding experiments.

synthesis_and_characterization_of_modified_nucleotides_in_the_970_hairpin_loop_of_escherichia_coli_1

1. Introduction<!>2.1. Synthesis of N2-methylguanosine and the corresponding phosphoramidite<!>2.2. Synthesis and characterization of RNA analogues<!>2.3. Effects of modifications on the stability of h31<!>2.4. Effects of modifications on the secondary structure of h31<!>3. Conclusions<!>4.1.1. General<!>4.1.2. 2\xe2\x80\xb2, 3\xe2\x80\xb2, 5\xe2\x80\xb2-Tri-O-acetylguanosine (1)<!>4.1.3. 2\xe2\x80\xb2, 3\xe2\x80\xb2, 5\xe2\x80\xb2-Tri-O-acetyl-2-N-(p-methylphenylthiomethyl)guanosine (2)<!>4.1.4. 2\xe2\x80\xb2, 3\xe2\x80\xb2, 5\xe2\x80\xb2-Tri-O-acetyl-2-N-methylguanosine (3)<!>4.1.5. 2-N-Methyl-6-O-(diphenylcarbamoyl)guanosine (5)<!>4.2.1. 5\xe2\x80\xb2-O-(4,4\xe2\x80\xb2-Dimethoxytrityl)-2-N-methyl-6-O-(diphenylcarbamoyl)guanosine (6)<!>4.2.2. 5\xe2\x80\xb2-O-(4,4\xe2\x80\xb2-Dimethoxytrityl)-2\xe2\x80\xb2-O-[[(triisopropylsilyl)oxy]methyl]-2-N-methyl-6-O-(diphenylcarbamoyl) guanosine (7)<!>4.2.3. 5\xe2\x80\xb2-O-(4,4\xe2\x80\xb2-Dimethoxytrityl)-2\xe2\x80\xb2-O-[[(triisopropylsilyl)oxy]methyl]-2-N-methyl-6-O-(diphenylcarbamoyl) guanosine 3\xe2\x80\xb2-(2-cyanoethyl diisopropylphosphoramidite) (8)<!>4.3. RNA oligonucleotide synthesis<!>4.4. RNA deprotection and purification<!>4.5. UV melting studies<!>4.6. Circular dichroism (CD) spectroscopy<!>

<p>An intriguing question about the structure and function of the ribosome is the role played by the naturally occurring modified nucleotides.1–3 The range of modifications includes methylation on both the nucleoside base and 2′-hydroxyl group of the ribose moiety, pseudouridylation (uridine isomerization), as well as more elaborate alterations.4 Although the rRNA modifications have only been mapped for a small number of organisms, it appears that most, if not all, ribosomes contain modified nucleotides. It has been suggested that the modified nucleotides of the small subunit have important roles in ribosome assembly and protein synthesis, despite the fact that unmodified ribosomes are able to carry out key biological functions in vitro.5–8 The Escherichia coli (E. coli) small subunit RNA (16S rRNA) contains one pseudouridine and ten methylated nucleotides (http://library.med.utah.edu/RNAmods); m7G527, m2G966, m5C967, m2G1207, m4Cm1402, m5C1407, m3U1498, m2G1516, m26A1518, and m26A1519.9–11 In 3D models of the 30S subunit, based on high-resolution X-ray crystal structures, eight of the ten methylated residues cluster in the subunit cleft near the decoding region.3,12,13</p><p>The 970 loop (helix 31, h31) of E. coli 16S rRNA is located near the ribosomal P site and therefore believed to be intimately involved in translation.14–17 E. coli h31 contains two modified nucleotides, N2-methylguanosine at position 966 (m2G966) and 5-methylcytidine at position 967 (m5C967) (Figure 1). Among the nucleotides present in h31, residue 965 is the least conserved, with a U present in 52% of known sequences, and A (37%), C (9%) or G (2%) in the remaining sequences.18 In contrast, nucleotide 966 is the most conserved, with G occurring in 97.5% of the sequences.18 Similarly, nucleotide 967 is a C in 85% of the sequences and an A in 13%.18 Mutations at either G966 or C967, as well as loss of methylation at G966, do not affect the growth rate of E. coli;19,20 however, a single-base deletion at C967 leads to a dominant lethal phenotype.19 Recently, a saturation mutagenesis study on the 970 loop revealed that mutations at positions 966 and 969 significantly affect protein synthesis by the ribosome.21 It was also shown that ribosomes with single mutations at positions m2G966 and m5C967 are capable of producing more protein than the wild-type ribosome.21 These studies suggest that modifications in the 970 loop influence the function of the ribosomal machinery.</p><p>There is a considerable amount of structural information for h31 from ribosome X-ray crystal structures.12,15–17,22,23 In E. coli 70S ribosomes, a three-base stacking interaction is observed between residues 966, 967, and 968.23 In the crystal structure of T. thermophilus 30S ribosomes, stacking is observed between those same residues, as well as between positions 965, 969, and 970.22 In addition, the stacked arrangement formed by residues 965, 969, and 970 shows interactions with the ribosomal protein S13.22 Residue G971 makes six hydrogen bonds within a pocket formed by residues 949, 950, 1363, 1364, and 1365 in the 16S rRNA.22 In the ribosome, m2G966 is also protected by the P-site-bound tRNA from chemical probes.24 Direct contacts between residues of the tRNA anticodon loop, the P site, and h31 of 16S rRNA were observed in X-ray crystal structures of the 70S ribosome.15,16 These structures suggest that the anticodon loop of the P-site tRNA is held in place by two stacking interactions; one between m5C1400 of 16S rRNA and G34 of the tRNA, and a second between m2G966 of h31 and ribose 34 of tRNA.16,25 The m22G966 residue is flipped out in the 70S ribosomal complex of T. thermophilus containing a model mRNA and two tRNAs.16 In the E. coli 70S ribosome crystal structure with a tRNA analogue and mRNA, the corresponding m2G966 residue resides in a tight binding pocket with the anticodon and the decoding site, which differs from the 30S structure.22 These data suggest that the 970 loop is dynamic and may be involved in proper positioning of tRNA in the ribosomal P site. Modeling studies by Aduri et al.13 suggested that methylation of residues 966 and 967 increases the surface area for stacking and also improves van der Waals contacts with the hydrophobic portion of Arg128 of the S9 protein in the ribosome.</p><p>We were interested in developing a better understanding of the structural and possible stabilizing roles of the modifications in h31 of 16S rRNA. Currently, information is limited regarding the effects of nucleotide modifications on rRNA structure and function. Until recently, the chemical properties of the modified nucleotides had not yet been found to influence specific functional roles in the ribosome.26 Nevertheless, even minor changes in chemical composition of the RNA nucleotides can lead to altered steric properties, hydrogen-bonding interactions, local base-stacking potential, van der Waals interactions, or structural rigidity.2 The goals of this research were to synthesize N2-methylguanosine and its corresponding phosphoramidite, and then utilize the amidite in the syntheses of h31 model sequences. The syntheses were then followed by biophysical studies in order to understand the influence of base methylations on structure and stability of h31 in a systematic manner. The structures and stabilities of four model h31 RNAs were examined by using circular dichroism (CD) and thermal melting studies.</p><!><p>The methylated guanosine nucleoside, m2G, was synthesized by using a combination of several published procedures (Scheme 1). The first step involved the acetylation of guanosine to give compound 1.27 The second and third steps to generate N-methylated intermediate 3 were adapted from Bridson and Reese's method using a p-thiocresol intermediate.28 The generation of compounds 2 and 3 was accomplished with 75% yields for each step. To avoid solubility problems typically encountered during the phosphoramidite synthesis of guanosine derivatives, we used a strategy that involves protection at the lactam function of the guanine residue according to a procedure devised by Kamimura29 and used by others.30 Diphenylcarbamoyl (DPC) protection and acetyl deprotection under mild basic conditions gave compound 5 in good yield (55%, two steps). In addition to the successful O6-protection of the guanine residue, the presence of DPC improved the solubility and chromatographic purification properties of the resulting derivative. The corresponding phosphoramidite was generated by using a similar approach as Höbartner and coworkers,31 but employing the DPC-protected precursor (compound 5). In our hands, compound 5 gave higher yields in the m2G phosphoramidite (compound 8) synthesis (Supplementary material) than the methods reported previously.31,32 The synthesis could be performed on a reasonably high scale to produce sufficient amounts of amidite 8 for multiple couplings.</p><!><p>Figure 1 shows the RNA analogues representing the 970 stem-loop region (h31) of E. coli 16S rRNA. The numbering system is based on the full-length E. coli rRNA sequence. A terminal G-C base pair was added to each stem (denoted by lower case g-c) in order to stabilize the hairpin structure. Residues in all four constructs are numbered from g1 to c18 for the ends and U960 through A975 for the component representing the natural E. coli h31 sequence. Two h31 RNAs (ECh31UNMOD and ECh31M5C) were obtained using commercial amidites. The two remaining h31 analogues (ECh31M2G and ECh31WT) were synthesized using the 5′-O-DMT-2′-O-TOM-6-O-DPC-2-N-methylguanosine phosphoramidite (compound 8) along with the commercially available amidites. The doubly modified h31 analogue (ECh31WT) represents the wild-type sequence of E. coli 16S rRNA helix 31. The incorporation of m2G and m5C residues was confirmed by MALDI-TOF mass spectrometric analysis of purified full-length RNA and by P1 nuclease digestion and calf intestinal phosphatase treatment of the RNAs, followed by reverse-phase HPLC analysis of the enzyme digest products (see Supplementary material).</p><!><p>Absorbance versus temperature profiles (melting curves) were obtained at pH 7.0 for all four RNA constructs. They were analyzed in terms of ΔG°37, ΔG°50, ΔH°, ΔS°, and melting temperature (Tm).33,34 The normalized absorbance plots at single RNA concentrations are shown in Figure 2. The melting curves are biphasic for four RNAs (see Supplementary material) with transitions at 0 to 35 °C and 40 to 80 °C. The low temperature transitions were concentration dependent, suggesting the formation of a bimolecular complex, such as a duplex or loop–loop interaction. The higher melting transitions were concentration independent, consistent with unimolecular unfolding of a hairpin structure.35 The corresponding thermodynamic parameters for the four RNAs are given in Table 1. The data indicate slight destabilizing effects (ΔΔG°37 values of 0.2–0.5 kcal/mol) of the modifications on helix 31. The observed order of stability of the RNAs is ECh31UNMOD > ECh31M2G > ECh31M5C ≥ ECh31WT. The RNA samples were analyzed as five different dilutions per experiment and UV melts were also performed in triplicate for each construct. Experiments in K+ buffer (15 mM KCl, 20 mM cacodylic acid, 20 mM Tris [basic form], 0.5 mM Na2EDTA, pH 7.0) gave similar results (Supplementary material).</p><p>The subtle destabilizing effect of the methylations was somewhat unexpected. In X-ray crystal structures of E. coli 70S ribosomes23 and T. thermophilus 30S ribosomes,22 m2G and m5C are involved in a three-base stacking interaction with residue 968. One might expect the methyl groups in the modified nucleosides to facilitate stacking interactions;13 however, both m2G and m5C are destabilizing relative to standard nucleotides G and C within the given sequence context. The presence of the h31 methylations might serve another purpose, such as stabilizing the ribosome through hydrophobic interactions with Arg128 of the S9 protein.13 Furthermore, the slight destabilization caused by the base methylation may facilitate base flipping of m2G966, which has been observed in ribosome crystal structures and likely plays an important role in protein synthesis.16 The small energetic penalty (ΔΔG°37 = 0.5 kcal/mol) would then be overcome by contacts with various ribosome components such as 16S rRNA helices, ribosomal proteins, and tRNA.16 Although the effects on stability may appear to be quite small, this level of destabilization could account for an approximate two-fold difference in binding to a ribosomal RNA component, which may be important for fidelity of protein synthesis.</p><!><p>The CD spectra of the helix 31 analogues were obtained to analyze the effects of modifications on the folded structure (Figure 3). The data indicate that the unmodified, singly modified, or doubly modified RNAs contain A-form stem regions, and they display similar conformations. They all have peak maxima at 270 nm and minima at 240 nm, similar to other A-form RNAs. A difference spectrum was obtained using Equation 1 in order to determine if the structural changes induced by the modified nucleotides are additive. In Equation 1, the difference spectrum for the singly modified RNAs and unmodified RNA is set equal to the difference spectrum of fully modified (wild-type) and unmodified RNA. If the effects of the modifications on the structure are additive, then the total difference spectrum should be equal to zero (Equation 2).</p><p>As shown in Figure 3D, the total difference spectrum is close to zero except for slight changes in the lower wavelength range (<260 nm). Therefore, the data are interpreted as the presence of modifications in h31 having subtle effects on the RNA structure. Studies in K+- and Mg2+-containing buffers showed similar results (Supplementary material).</p><!><p>Several conclusions can be drawn from these studies with respect to the relative stabilities and structures of RNAs with single and multiple modified nucleotides. Small reproducible differences in the free energy values for the E. coli h31 variants reveal slight destabilizing effects of the modifications on helix 31. A recent X-ray crystal structure of the T. thermophilus 70S ribosome complexed with a model mRNA and two tRNAs revealed that the positions of the 16S rRNA P-site nucleotides in the vacant ribosome superimpose well with those in the tRNA-containing complex, with the exception of m22G966.16 Residue m22G966 (m2G966 in E. coli ribosomes) is flipped out in the crystal structure of the T. thermophilus 70S ribosome containing a model mRNA and two tRNAs (Figure 4A)15,16 or remains stacked in the 30S ribosomal subunit from T. thermophilus crystal structure (Figure 4B).22 The interaction with the anticodon loop of the P-site-bound tRNA appears to be stabilized by stacking interactions involving m22G966 with ribose 34. Hence, the flipped-out base has been suggested to facilitate correct positioning of the tRNA during translation.16,25 Therefore, the slight destabilizing effects of modifications in h31 may be important for facilitating the flipping movement of residue 966; but at the same time, stacking interactions with the tRNA are stabilized through the methyl group. Positioned in the middle of the three stacked bases, the m5C967 residue has a greater destabilizing effect than m2G966 (ECh31M5C vs. ECh31M2G). This result may be due to a greater disruption of stacking by the methylated base of m5C967.</p><p>The CD data indicate that the unmodified, singly modified, and fully modified RNAs all contain A-form stem regions and display similar conformations. Minor differences between the CD spectra of the fully modified and unmodified h31 constructs indicate possible differences in the loop regions. These differences could arise from modification-dependent changes in the loop, such as altered base stacking at positions 966, 967, and 968. The UV melting data reveal, however, that the presence of modifications at specific locations does not influence the ability of the constructs to form stable hairpin loop structures.</p><p>The exact functional role of the modifications at positions 966 and 967 of h31 is still unknown. Ribosomes carry out the essential biochemical process of translation, which requires an exquisite array of highly specific interactions between rRNA, mRNA, tRNAs, and ribosomal proteins. Modifications are believed to help fine-tune ribosome function.3 Since proper ribosome function depends on the correct balance of speed and accuracy of tRNA binding, peptide-bond formation and tRNA release, methylations in h31 could play a role in maintaining proper interactions within the ribosome.36,37 Mutational analyses revealed that a loss of methylation at either position 966 or 967, leads to increased protein production by the mutant ribosomes.21 Our data would therefore suggest that methylations destabilize h31 in order to maintain the proper interactions with tRNA, rRNA, or proteins. A lack of modification at residues 966 or 967 in h31 could reduce the ability of base 966 to flip and regulate tRNA affinity, positioning, or accuracy. Thus, it will be of great interest to explore in greater detail the relationship between G966 and C967 methylation and translational fidelity. Furthermore, the availability of a suitable method for synthesizing m2G and its corresponding phosphoramidite (commercially not available) will allow RNAs containing m2G modifications to be generated in sufficiently large quantities for use in additional biophysical and ligand-binding studies.</p><!><p>Most reagents and solvents were either purchased from Aldrich (St. Louis, MO) or Acros (Morris Plains, NJ) and used as received. Anhydrous pyridine was purchased in a sure-seal bottle from Aldrich. Methylene chloride (CH2Cl2) was distilled over CaH2. Methanol and triethylamine were purchased dry from Aldrich in a sure-seal bottle. Moisture sensitive reactions involved flame-drying equipment (syringes, round-bottom flasks, stir-bars, etc.) under vacuum and performing reactions under dry argon. TLC analyses were accomplished with precoated silica gel (0.25 mm thickness) glass plates. Reactions were monitored by visualizing the TLC plates under UV light and/or by staining with phosphomolybdic acid (PMA) solution (10% w/v in absolute ethanol) and heating with a hot plate. Flash chromatography was performed with silica gel 60 (0.038–0.063 mm). Flash columns were neutralized with 0.5–1% triethylamine (TEA) prior to purification of pH sensitive intermediates/compounds. 1H NMR and 13C NMR spectra were recorded on either a Varian Unity 300, Mercury 400, or Varian Unity 500 spectrometer and referenced to tetramethylsilane as an internal standard. ESI spectra were recorded on a Quattro LC (Bruker Daltonics) in the positive ion mode unless otherwise noted.</p><!><p>Compound 1 was generated according to a literature procedure.27</p><!><p>This compound was generated using a procedure described in the literature.28</p><!><p>Compound 3 was prepared using a procedure described in the literature.38</p><!><p>To a solution of 2′, 3′, 5′-tri-O-acetyl-2-N-methylguanosine 3 (3.0 g, 7.1 mmol, 1.0 eq) in dry pyridine (50 mL) were added diphenylcarbamoyl chloride (3.45 g, 15 mmol, 2.1 eq) and diisopropylethylamine (1.9 mL, 11.4 mmol, 1.6 eq). The dark brown reaction mixture was then stirred at room temperature for 1 h to obtain 2′, 3′, 5′-tri-O-acetyl-6-O-(diphenylcarbamoyl)-2-N-methylguanosine 4. TLC showed complete conversion to product at this stage of the reaction. The reaction mixture was then diluted with pyridine (15 mL) and EtOH (30 mL). To this solution cooled at 0 °C was added 2 M NaOH (25 mL), which was also precooled to 0 °C. The reaction mixture was stirred for 10 min at 0 °C. Acetic acid (c.a. 5 mL) was then added to neutralize the solution. Extraction with CH2Cl2 followed by chromatography on silica gel afforded 5 after two steps in 55% yield (1.92 g). TLC (CH2Cl2: MeOH, 9:1 v/v): Rf = 0.5; ESI-MS (ES+) calculated for C24H24N6O6 492.1757, found 493.2 (M+H+), 515.2 (M+Na+), 1007.4 (2M+Na+).</p><!><p>Compound 5 (0.51 g, 1.05 mmol, 1.0 eq) and 4, 4′-dimethoxytritylchloride (0.39 g, 1.14 mmol, 1.09 eq) were azeotroped three times with toluene for ~5 h. To the dried compound 5, anhydrous pyridine (5 mL) was added. The mixture was stirred at room temperature under Ar atmosphere for 4 h. 4-Dimethylaminopyridine (0.09 g, 0.7 mmol, 0.7 eq) was subsequently added and stirring was continued for 17 h. The reaction was quenched with methanol (1 mL) and evaporated to dryness. The crude residue was dissolved in 50 mL of dichloromethane and washed with 5% sodium bicarbonate followed by saturated sodium chloride. The organic layer was dried over sodium sulfate and filtered. The product was then purified by silica gel chromatography using a solvent mixture of 90% dichloromethane, 9% methanol, 1% triethylamine to give 6 as a light yellow crystalline solid (0.56 g, 67%). TLC (CH2Cl2: MeOH, 9:1 v/v): Rf =0.4; ESI-MS (ES+) calculated for C45H42N6O8 794.3, found 795.7 (M+H+), 303.4 ([(MeO)2Tr]+).</p><!><p>Di-tert-butyltindichloride (0.26 g, 0.846 mmol, 1.2 eq) was added to a solution of dry 1,2-dichloroethane (6.5 mL) containing compound 6 (0.56 g, 0.705 mmol, 1.0 eq) and ethyldiisopropylamine (0.36 mL, 2.82 mmol, 4.0 eq). The reaction mixture was heated to 70 °C for 15 min under refluxing conditions. Then, the mixture was allowed to cool to room temperature. Upon cooling down, the mixture became cloudy and light brown in color. The crude reaction mixture was then stirred with [(triisopropylsilyl)oxy]methylchloride (0.18 mL, 0.776 mmol, 1.1 eq) for 3 h at room temperature. After 3 h, the mixture was evaporated to dryness. The resulting crude residue was dissolved in 20 mL of dichloromethane and washed with saturated sodium bicarbonate followed by saturated sodium chloride. The organic layer was dried over sodium sulfate and filtered. The product was then purified by silica gel column chromatography using a solvent mixture of dichloromethane: methanol (20:1) and triethylamine (0.5%) to give 7 as a light yellow oil (0.49 g, 70%). TLC (CH2Cl2: MeOH, 9:1 v/v): Rf =0.7; ESI-MS (ES+) calculated for C55H64N6O9Si 980.4, found 981.9 (M+H+), 1003.9 (M+Na+), 1019.9 (M+K+).</p><!><p>Compound 7 (0.25 g, 0.25 mmol, 1.0 eq) was dried extensively under vacuum. Then, it was dissolved in 5 mL of anhydrous dichloromethane. Next, N,N-diisopropylethylamine (0.3 mL, 2.5 mmol, 10 eq) and 2-cyanoethyldiisopropylchlorophosphoramidite (0.08 mL, 0.38 mmol, 1.5 eq) were added and the mixture was stirred for 2 h at room temperature. The reaction was quenched with 5% aqueous sodium bicarbonate, and then it was extracted with 2 × 50 mL of dichloromethane. Combined extracts were dried over anhydrous sodium sulfate and evaporated. The crude mixture was purified by silica gel column chromatography (hexane: ethyl acetate, 3:1 and triethylamine, 0.5%) to yield 8 as a white form (0.285 g, 95%). TLC (hexane: EtOAc: Et3N=75%: 24.5%: 0.5%): Rf =0.23; 1H NMR ((CD3)2CO, 400 MHz) (mixture of diastereoisomers) 0.89–1.05 (2m, 60H), 1.18–1.30 (2m, 10H), 1.45 (t, 4H), 2.49 (br.s, 2H), 2.79–2.84 (m, 8H), 3.39–3.41 (m, 2H), 3.66–3.75 (m, 14H), 4.17–4.19 (m, 2H), 4.69 (m, 2H), 5.12–5.14 (m, 8H), 6.11 (d, J = 5.2 Hz, 2H), 6.79–6.83 (m, 10H), 7.17–7.50 (2m, 36H), 8.04 (s, 2H); 13C NMR ((CD3)2CO, 400 MHz) 12.56, 18.04, 18.08, 20.7, 20.75, 24.76, 24.82, 24.89, 24.95, 28.97, 43.82, 43.91, 43.96, 44.06, 55.45, 59.0, 59.16, 59.89, 60.02, 64.39, 72.59, 84.4, 84.7, 87.14, 87.71, 90.03, 90.36, 113.86, 127.54, 128.54, 128.58, 128.91, 128.98, 129.91, 130.85, 130.93, 130.99, 136.57, 136.69, 143.38, 145.96, 151.43, 157.33, 159.58, 160.82, 206.15; 31P NMR ((CD3)2CO, 400 MHz) (mixture of diastereoisomers) 150.96, 151.14; High resolution ESI-MS (ES+) calculated for C64H81N8O10PSi 1180.5583, found 1181.5634 (M+H+), 1203.5439 (M+Na+), 1219.5358 (M+K+).</p><!><p>The four RNA hairpins used in this study (Figure 1) were chemically synthesized at the W. M. Keck Foundation at Yale University, New Haven, Connecticut, USA. For RNAs containing m2G, 50 μmoles of the corresponding phosphoramidite were provided. The m5C phosphoramidite was purchased from Glen Research. The sequences of the four synthetic RNAs are as follows (U, C, G, A are nucleotides representing the natural E. coli sequence; g, c are added nucleotides to enhance hairpin stability):</p><p>5′-gUU CGA UGC AAC GCG AAc-3′ (ECh31UNMOD)</p><p>5′-gUU CGA UG(m5C) AAC GCG AAc-3′ (ECh31M5C)</p><p>5′-gUU CGA U(m2G)C AAC GCG AAc-3′ (ECh31M2G)</p><p>5′-gUU CGA U(m2G)(m5C) AAC GCG AAc-3′ (ECh31WT)</p><!><p>Upon completion of coupling on an automated synthesizer, the CPG-bound RNA was cleaved from the solid support and deprotected with 1:3 (v/v) EtOH/NH4OH and TBAF (tetrabutylammonium fluoride solution, 1 M in THF) as described in the literature.29,39 The RNAs were desalted over Poly-Pak II cartridges (Glen Research), then purified by HPLC on an XTerra MS C18 column (2.5 μm, 10 × 50 mm, Waters) in which the eluent was 0.1 M TEAA (triethylammonium acetate) buffer, pH 7.0, with a 5 – 15% linear gradient of acetonitrile over 25 min at a flow rate of 4.0 mL/min. After HPLC purification, each oligomer was further desalted by ethanol precipitation and dialysis for 3 days against RNase-free, deionized water using a 1000 molecular weight cut-off membrane (Spectra-Por). RNA concentrations were calculated using Beer's law and a single-stranded extinction coefficient (ε) of 176,900 M−1cm−1.40 The same extinction coefficients were used for guanosine and m2G (1.4 × 104 M−1cm−1 at pH 7.0) and cytidine and m5C (9.1 × 103 M−1cm−1 at pH 7.0).</p><!><p>The absorbance versus temperature profiles were obtained on an Aviv 14DS UV-visible spectrophotometer with a five-cuvette thermoelectric controller. Microcuvettes with two different pathlengths, 0.1 and 0.2 cm (60 and 120 μL volumes, respectively), were employed. Each measurement was taken in triplicate. The buffer used in each experiment contained 15 mM NaCl, 20 mM sodium cacodylate, and 0.5 mM Na2EDTA (pH 7.0), unless noted otherwise. Each oligomer was dissolved in a specific volume to yield an absorbance reading just below 2.0 in a 0.1 cm pathlength cuvette. The RNA concentrations were determined from the absorbance values (260 nm) at 95 °C. The absorbance data were collected at 280 nm from 0 to 95 °C with a constant heating rate of 0.5 °C/min.41 Thermodynamic parameters were obtained from the absorbance versus temperature profiles using the MELTWIN v. 3.5 melting curve program.33 This program performs a van't Hoff analysis, assuming a two-state model for the transition between a native and a denatured (random coil) structure of a hairpin loop.</p><!><p>CD spectra were obtained on an Applied Photophysics Chirascan circular dichroism spectrometer (220–320 nm) at 25 °C in 15 mM NaCl, 20 mM sodium cacodylate, and 0.5 mM Na2EDTA at pH 7.0. The RNA concentrations were maintained at 2.5–3.0 μM for all CD experiments. Based on the RNA strand concentration, the measured CD spectra were converted to molar ellipticity (Δε),42 which denotes the moles of RNA molecules rather than moles of individual residues present in the sequence.</p><!><p>The structures of N2-methylguanosine (m2G) and 5-methylcytidine (m5C) are shown. Secondary structure representations of E. coli helix 31 of the small subunit ribosomal RNA with indications of the modification sites and the synthetic RNA hairpins derived from residues 960–975 are shown (unmodified (ECh31UNMOD), fully modified (wild-type or ECh31WT, containing m2G966 and m5C967), m5C967 singly modified (ECh31M5C), and m2G966 singly modified (ECh31M2G)). The terminal G-C base pair (g1-c18) was added to stabilize the short stem region.</p><p>Representative normalized UV melting curves of the modified RNAs compared to the unmodified RNA, taken in 15 mM NaCl, 20 mM sodium cacodylate, 0.5 mM Na2EDTA (pH 7.0), are shown. The melting curves for the unmodified RNA (ECh31UNMOD, solid grey lines in panels A – C) are compared to those for the modified RNAs (dashed lines): (A) ECh31WT, (B) ECh31M2G, and (C) ECh31M5C. All of the melting curves were normalized at 95 °C and absorbance measurements were taken at 280 nm.</p><p>CD spectra of the modified and unmodified RNA constructs are shown. Each spectrum is an average of five scans. The CD spectrum of the unmodified analogue (ECh31UNMOD, solid grey line) is shown in panels A–C with overlays (dashed lines) of the ECh31M5C (A), ECh31M2G (B), and ECh31WT (C) RNAs. The molar ellipticities are normalized to RNA concentrations. The total difference spectrum (ECh31M5C + ECh31M2G − ECh31UNMOD − ECh31WT) is shown in panel D (solid black line).</p><p>Structures of h31 showing the flipped out (A) and stacked (B) conformations of residue G966 (PDB accession IDs - 2J0015 and 1FJF22).</p><p>Synthesis of the 6-O-DPC-2-N-methylguanosine 5: (i) acetic anhydride, DMAP, Et3N, acetonitrile, 0.5 h; (ii) p-thiocresol, acetic acid, formaldehyde, ethanol, reflux; (iii) NaBH4, DMSO, 100 °C; (iv) DPCCl, diisopropylethylamine, pyridine; (v) a) 2 M NaOH, 20 min; b) acetic acid.</p><p>Thermodynamics of the four RNA analogues</p><p>Each measurement was taken in triplicate. Best fits were obtained by assuming a hairpin formation. The buffer conditions were 15 mM NaCl, 20 mM sodium cacodylate and 0.5 mM Na2EDTA, pH 7.0.</p><p>A conservative estimate of the standard error for ΔG°50 is 3% (± 0.2 kcal/mol).43</p>

PubMed Author Manuscript

Providing the 'best' lipophilicity assessment in a drug discovery environment

Lipophilicity is a fundamental structural property that influences almost every aspect of drug discovery. Within Pfizer, we have two complementary high-throughput screens for measuring lipophilicity as a distribution coefficient (LogD) -a miniaturized shake-flask method (SFLogD) and a chromatographic method (ELogD). The results from these two assays are not the same (see Figure 1), with each assay being applicable or more reliable in particular chemical spaces. In addition to LogD assays, the ability to predict the LogD value for virtual compounds is equally vital. Here we present an in-silico LogD model, applicable to all chemical spaces, based on the integration of the LogD data from both assays. We developed two approaches towards a single LogD model -a Rule-based and a Machine Learning approach. Ultimately, the Machine Learning LogD model was found to be superior to both internally developed and commercial LogD models.

providing_the_'best'_lipophilicity_assessment_in_a_drug_discovery_environment

Introduction<!>Methods<!>Lipophilicity models<!>Summary

<p>Assessment of lipophilicity is fundamental to understanding the properties of a molecule. Within the drug discovery environment, additional consideration of the ionization of the compound in the aqueous phase, especially at pH 7.4, is necessary. Hence, distribution coefficient (LogD) is the lipophilicity descriptor of choice. While there are many methods for determining LogD (Kempinska, Chmiel et al. 2019), partitioning between n-octanol and an aqueous buffer (i.e. shake flask) remains the most direct method, though chromatographic methods have become popular indirect experimental approaches. To meet the demands of thousands of LogD determinations, rapid, high capacity and compound-sparing assays needed to be developed.</p><p>Within Pfizer, we have two assays for assessing LogD -a miniaturized shake flask method (SFLogD) (Stopher and McClean 1990) and a chromatographic method (ELogD) (Lombardo, Shalaeva et al. 2000, Lombardo, Shalaeva et al. 2001). Each assay has limitations that affect which assay is the most appropriate for a particular compound.</p><p>For our SFLogD assay, measured LogD values > 3 appear to be an underestimation, which has also been noted by others (Low, Blasco et al. 2016). The SFLogD underestimation is evident when comparing experimental values of SFLogD and ELogD for compounds within the Pfizer sample collection where SFLogD < ELogD when SFLogD > 3 (Figure 1). There may be multiple causes of this lower than anticipated value including cross contamination of the extracted aqueous sample by n-octanol with high compound concentration due to increased lipophilicity. Inadequate shaking/equilibration and the presence of DMSO for compound solubilization are other proposed causes of this underestimation (Low, Blasco et al. 2016).</p><p>For our ELogD assay, while it is an indirect estimation of LogD, it does not appear to suffer from high LogD limitations. However, it is limited in its inability to estimate LogD for acids or zwitterions due to specific interaction of anions with the stationary phase (Lombardo, Shalaeva et al. 2001). Internally, acids are excluded from this assay.</p><p>Given the complementary nature and limitations of these two high-throughput (HTS) assays, Pfizer has implemented a screening paradigm where compounds are first submitted to the SFLogD assay. Those with SFLogD > 3 are automatically submitted to the ELogD assay for assessment of underestimation by SFLogD. It is acknowledged that highly lipophilic acids, where only SFLogD values are obtained, may be underestimated in this paradigm.</p><p>In addition to the availability of HTS lipophilicity assays, a predictive lipophilicity model for use during the design phase is equally critical. We initially built Machine Learning models for each assay to complement the two LogD assays. However, both models suffer from the same limitations as their respective assays, namely underprediction for the SFLogD model and inability to predict acidic compounds for the ELogD model. Therefore, users of the models are stuck with the same wet data quandary when looking at the models, i.e. which model should I trust for this compound? Introduction of a single model for both assays became paramount. Analysis of commercial programs (i.e. ACD or MoKa) indicated that they were inadequate for serving as a reliable single source of predictions that matched our internal assay output. Ultimately, we launched two hybrid models, where each can provide users with a single best estimate of LogD. Development of these two hybrid models will be the focus of this paper. For differentiation and for the remainder of this paper, reference to measured will contain a 'w' (i.e. wet) as prefix, while predicted lipophilicity will contain a 'c' (i.e. computed) as prefix.</p><!><p>The ELogD and SFLogD models are non-linear regression models built on an internal database of compounds. The internal database contains approximately 200k unique structures with wSFLogD data and approximately 80k unique structures with wELogD data. Models on both datasets are constructed using descriptors generated by MoKa (Molecular Discovery Ltd 2020) , alvaDesc (Mauri 2020), and counts of a set of internally developed SMARTS fragments (calculated using OpenEye OEChemTK (OpenEye Scientific Software 2020). For a MoKa pKa prediction where no value is given (e.g., neutrals), the value of ApKa is set to 14 and BpKa is set to 0. Compounds that failed descriptor generation due to problematic substructures are removed from the dataset prior to model construction. Datasets are averaged by unique structure and are modeled as LogD values (i.e. untransformed). All error metrics are also calculated in the LogD space. Models were constructed using XGBoost (Chen, T.; Guestrin, C. In XGBoost: A Scalable Tree Boosting System). Hyperparameters and descriptors are optimized using a 5-fold cross-validation scheme, where 80% of the data was selected for training and the remaining 20% is predicted. Model statistics are generated from these cross-validated prediction values. A final model, using all training data and the selected parameters and descriptors, is constructed and published internally for LogD predictions. Model optimization is performed only once and subsequent newly measured data is added to the training set of the published model.</p><p>To create a single LogD calculator, two variants that integrate data from the ELogD and SFLogD assays were developed -Rule-based PFLogD and Machine Learning PFLogD.</p><p>The Rule-based PFLogD model generates a prediction based on a set of heuristic rules for integrating predictions from the ELogD and SFLogD models. The basic rules are depicted in Figure 2 and were created with cooperation from assay scientists and medicinal chemists. Supplementing these rules is set of exceptions to handle corner cases that have been identified during application of this heuristic model.</p><p>The Machine Learning (ML) PFLogD model is built on a collection of LogD values that is derived from integrating predictions from the ELogD and SFLogD models. The integration follows the rules shown in Figure 2. The cSFLogD and cELogD are from 5-fold CV predictions. Only structures with measured LogD in one of the two assays are included in the training set. The supplementary exceptions used in the Rule-based PFLogD are not considered, hence measured LogD values are not directly used in the integration algorithm. A model, based on this integrated set of predictions as input, is optimized using the same methodology as described above for the SFLogD and ELogD models.</p><!><p>The lipophilicity model (cSFLogD or cELogD) developed for each assay (wSFLogD or wELogD) is able to reasonably predict its respective assay as shown in Figures 3 and 5. However, the models are limited in their ability to predict the complementary assay (i.e. cELogD to predict wSFLogD or cSFlogD to predict wELogD) as shown in Figures 4 and 6. The inability of cELogD to predict acids is evident in Figure 4 where the acidic compounds (colored in red) have significantly higher cELogD compared to wSFLogD. The underestimation of SFLogD, whether experimental or predicted, is evidenced by cELogD > wSFLogD or wELogD > cSFLogD for highly lipophilic compounds, as shown in Figures 4 and 6, respectively. The performance of these models is shown in Table 1. Statistics of the models on their own experimental data (i.e., cSFLogD on wSFLogD) are based on 5-fold cross-validation. Statistics against the complementary assay (i.e., cSFLogD on wELogD) are external predictions. There is sufficient disagreement between the experimental datasets that neither model is able to successfully predict the complementary dataset.</p><p>Analysis of commercial lipophilicity models' ability to predict our assays are shown Figures 7 -10. For prediction of wSFLogD by ACD or MoKa, while there are no apparent non-linear deviations, both appear to overpredict at very high and underpredict at very low wSFLogD values (Figures 7 and 8). For wELogD, while the overall prediction pattern for these commercial models is closer to unity, their prediction accuracies for this dataset is poor as indicated by the low percentage within +/-0.5 (see Table 2).</p><p>From a modeling perspective, it is more important to model our internal assays instead of an external data source -i.e. higher prediction accuracy of our internally generated LogD values even if there is a discrepancy between our internal and published values. Ultimately, the role of the model is to provide an equally accurate prediction of the experiment for all compounds and their effectiveness is measured by how well they predict the assay values. As an example, even though the underprediction of very low wSFLogD by ACD and MoKa may be justified by correlation to an external data source, for internal structure activity relationship (SAR) development, it is more important for our model to predict the eventually acquired experimental value.</p><p>With the goal of introducing a single LogD model that is able to predict both wSFLogD and wElogD, we developed two variants that integrate data from these two assays -Rule-based PFLogD and Machine Learning PFLogD.</p><p>The Rule-based PFLogD model generates a prediction based on a set of heuristic rules for integrating predictions from the ELogD and SFLogD models. The rules are described in Figure 2. Supplementing these rules was an evolving and ever-increasing set of exceptions to handle the integration of measured LogD values at different LogD ranges. The initial advantage of this model is in its simplicity and the integration of available experimental data. However, there is a high performance and management cost with this approach. Firstly, when experimental data was integrated became a critical question for the model. Batch differences, structure normalization, and data update frequency all cause significant maintenance problems and confusion. Second, constant re-analysis and refinement of these "expert" rules were required to determine if they are still valid or need to be updated.</p><p>As an alternative, a machine learning model -ML-PFLogD, was developed. This model is more consistent with current modeling approaches where structure-based descriptors of the input structure are used for the prediction. The training set for this model is derived from the cSFLogD and cELogD predictions that are integrated using the rules shown in Figure 2. The cSFLogD and cELogD are from 5fold CV predictions. No measured values are directly used in the integration and hence, none of the exceptions used in the Rule-based PFLogD are considered. All the compounds in the training set must contain a measured LogD value in one of the two assays, and the size of the training set for this model increases as compounds are tested in these assays.</p><p>Predictions of our wSFLogD and wELogD assays by ML-PFLogD are shown in Figures 11 and 12. The desired non-linear deviation for high lipophilicity compounds in wSFLogD is present, addressing the underestimation of the wSFLogD assay. However, the underprediction of low wSFLogD compounds, present with ACD and MoKa, is absent with ML-PFLogD (Figure 11). The deviation of PFLogD prediction of wELogD is linear across the range of lipophilicity and appears to be smaller than those of ACD and MoKa (Figure 12). The prediction statistics of these models against both datasets are summarized in Table 2. While ACD and MoKa are inadequate at predicting either internal assay, the ML-PFLogD model performs well against both datasets.</p><!><p>Pfizer has implemented two complementary high-throughput lipophilicity assays -wSFLogD and wELogD, and a screening paradigm to address the limitations of each assay -i.e. underestimation of high wSFLogD, and the overestimation of acidic compounds in wELogD. We have also built LogD models for each of these two assays. While each model performs well against its own assay, it performs poorly against the complementary assay. We developed two hybrid models -Rule-based PFLogD and Machine Learning (ML) PFLogD, in order to provide users with a single, best LogD prediction that was applicable for all compounds. The input for these hybrid models is a blend of experimental and/or predicted SFLogD and ELogD values, where the integration follows a set of heuristic rules. The Rule-based PFLogD, while simplistic in its prediction algorithm, has a high cost in performance and management. The machine learning based ML-PFLogD was found to be superior to commercial models (e.g. ACD or MoKa) and our internal models for predicting the LogD values generated by our assays, as well as simpler to maintain.</p>

ChemRxiv

Design, Synthesis, and Biological Evaluation of (E)-N'-((1-Chloro-3,4-Dihydronaphthalen-2-yl)Methylene)Benzohydrazide Derivatives as Anti-prostate Cancer Agents

Prostate Cancer (PCa) is the most frequently diagnosed cancer in men in their late '50s. PCa growth is mainly due to the activation of the androgen receptor by androgens. The treatment for PCa may involve surgery, hormonal therapy, and oral chemotherapeutic drugs. A structural based molecular docking approach revealed the findings of (E)-N'-((1-chloro-3,4-dihydronaphthalen-2-yl)methylene)benzohydrazide derivatives, where the possible binding modes of the compounds with protein (PDB ID: 3V49) are shown. The compounds (6a-k) were synthesized and characterized by using conventional methods. The compounds, 6g, 6j, and 6k were reconfirmed through single crystal X-ray diffraction (XRD). Further, the compounds (6a-k) and standard drug were evaluated against human prostate cancer cell lines, LNCaP and PC-3 and the non-cancerous cell line, 3T3. Among these compounds, 6g and 6j showed higher cytotoxicity, and 6g exhibited dose-dependent activity and reduced cell viability. The mechanism of action was observed through the induced apoptosis and was further confirmed by western blot and ELISA. Molecular dynamics simulation studies were carried out to calculate the interaction and the stability of the protein-ligand complex in motion. ADME properties were predicted for all the tested compounds. These findings may give vital information for further development.

design,_synthesis,_and_biological_evaluation_of_(e)-n'-((1-chloro-3,4-dihydronaphthalen-2-yl)methyle

Introduction<!><!>Introduction<!>In silico Molecular Docking<!>Chemistry<!>General Procedure for the Preparation of Benzoic Acid Hydrazide (3a-3k)<!>Preparation of 1-chloro-3,4-dihydronaphthalene-2-carbaldehyde (5)<!>General procedure for the Preparation of Compounds (6a-k)<!>Experimental Data<!>(E)-N'-((1-chloro-3,4-dihydronaphthalen-2-yl)methylene)benzohydrazide (6a)<!>(E)-2-chloro-N'-((1-chloro-3,4-dihydronaphthalen-2-yl)methylene)benzohydrazide (6b)<!>(E)-N'-((1-chloro-3,4-dihydronaphthalen-2-yl)methylene)-4-nitrobenzohydrazide (6c)<!>(E)-N'-((1-chloro-3,4-dihydronaphthalen-2-yl)methylene)-3-nitrobenzohydrazide (6d)<!>(E)-3-chloro-N'-((1-chloro-3,4-dihydronaphthalen-2-yl)methylene)benzohydrazide (6e)<!>(E)-N'-((1-chloro-3,4-dihydronaphthalen-2-yl)methylene)-4-methylbenzohydrazide (6f)<!>(E)-N'-((1-chloro-3,4-dihydronaphthalen-2-yl)methylene)-4methoxybenzohydrazide (6g)<!>(E)-4-bromo-N'-((1-chloro-3,4-dihydronaphthalen-2-yl)methylene)benzohydrazide (6h)<!>(E)-4-chloro-N'-((1-chloro-3,4-dihydronaphthalen-2-yl)methylene)benzohydrazide (6i)<!>(E)-N'-((1-chloro-3,4-dihydronaphthalen-2-yl)methylene)-4-hydroxybenzohydrazide (6j)<!>(E)-3-bromo-N'-((1-chloro-3,4-dihydronaphthalen-2-yl)methylene)benzohydrazide (6k)<!>Single Crystal X-Ray Diffraction Studies<!>Anti-proliferation Assay<!>Acridine Orange (AO)/Ethidium bromide (EtBr) Dual Staining for Apoptosis<!>Western Blot<!>DNA Methyl Transferase Enzyme Inhibitor (DNMT) Assay<!>Pharmacokinetic Properties Prediction (ADME)<!>Molecular Dynamics<!>Molecular Docking<!><!>Chemistry<!><!>Biological Evaluation<!><!>ADME<!>Molecular Dynamic Simulations<!><!>Conclusion<!>Data Availability<!>Author Contributions<!>Conflict of Interest Statement

<p>Cancer is a group of heterogeneous diseases leading to abnormal cell growth and dysfunction which proliferates to other parts of the body. Prostate cancer (PCa) is the second leading cause of cancer deaths among men in the United States. The American Cancer Society has estimated that 174,650 men will be diagnosed and there will be 31,620 deaths due to PCa in the United States in 2019 (American Cancer Society, 2019). Androgens, Dihydrotestosterone (DHT), and testosterone bind with the ligand binding domain of the androgen receptor (AR) and stimulate the growth of prostate cancer (Ferlay et al., 2010; Elancheran et al., 2015). The AR agonist and antagonist bind to the same binding pocket, but the AR antagonist is made of bulkier molecules than the AR agonist. Therefore, the AR agonist prevents the closing of helix-12 toward the ligand binding domain (Bohl et al., 2005). Thus, AR antagonists such as bicalutamide, enzalutamide, and flutamide prevent the activation of the androgen receptor. The mutation at amino acids Threonine 877 and Tryptophan 741 were frequently identified in AIPC patients (Jeanny et al., 2007). There are a few available drugs, such as bicalutamide, enzalutamide, cabazitaxel, abiraterone acetate, and so on, used in the treatment of PCa, but still, newly developed incipient candidates were needed with high activities and lower drawbacks (Elancheran et al., 2016). The Food and Drug Administration (FDA) has recently approved the potent second-line agents, abiraterone, and enzalutamide. Abiraterone is a CYP17A1 inhibitor (blocking both 17α hydroxylase and 17,20-lyase, two enzymes important for androgen synthesis), while enzalutamide is a potent AR ligand binding domain (LBD) competitive antagonist that blocks nuclear translocation and AR-dependent gene transcription (Charles et al., 2018; Moses et al., 2018) and demonstrated efficacy against metastatic castrated resistance prostate cancer (CRPC) and late-stage clinical trials with advanced PCa (Jung et al., 2010; Scher et al., 2010). Apalutamide-509 is used for a high therapeutic index and has a protective effect for the treatment of PCa (Clegg et al., 2012; Rathkopf et al., 2013). Sipuleucel-T is the only immunotherapy currently available to prevent prostate cancer (Sims, 2012; Wesley et al., 2012). The structural modification and important interactions derived from the groups of flutamide (i), bicalutamide (ii), enzalutamide (iii), N-(4-(4-hydroxyphenoxy)-3-methylphenyl)benzamide (iv), and the designed molecule (6a-k) were shown in Figure 1.</p><!><p>Androgen receptor targeting therapeutics for the treatment of PCa.</p><!><p>Several studies are ongoing for the development of specific target-based drugs to minimize drug resistance, toxicity, dosage, and more (Veeramanikandan and Benita Sherine, 2015). In vitro anticancer activity of the synthesized hydrazide derivatives was determined against human colorectal (HCT116) cancer cells line. Recently, hydrazide derivative was reported as a potent and selective inhibitor for antibacterial–antifungal (Somashekhar, 2013; Popiołek, 2017), anti-inflammatory (Todeschini et al., 1998), antimalarial (Melnyk et al., 2006), and anti-tuberculosis activities (Bedia et al., 2006), as an Entamoeba histolyica (Afreen et al., 2016), a cruzipain inhibitor (Cerecetto and Gonzalez, 2010) and as Epidermal growth factor receptor (EGFR) Kinase Inhibitor (Wang et al., 2016). Thus, benzohydrazides are important moieties that exhibit more effective inhibitory activity against various cancer cell lines such as A549, MCF-7, HeLa, and HepG2 (Wang et al., 2016). The current study on benzohydrazide reveals it as an effective inhibitor for prostate cancer and its physicochemical properties. Some newly synthesized substituted fused pyrazolo, triazolo, and thiazolo pyrimidine derivatives were shown potent anti-prostate cancer activities with low toxicity comparable to bicalutamide as a reference drug (Bahashwan et al., 2014; Elancheran et al., 2017; Ferroni et al., 2017). The structure-based drug design approach for analyzing the structure-activity relationships (SAR) of molecules followed SAR studies (Antonella et al., 2013), which provided rules for the selection of new, potentially active compounds (Shvets and Dimoglo, 1999), which can be synthesized. In this study, we have reported the design and development of benzohydrazide derivatives as novel AR antagonists. The two combined substructures of the R-dihydronaphthalene-2-carbaldehyde ring and benzohydrazides might show synergistic anticancer effects. All these facts encouraged us to integrate these two moieties and screen new benzohydrazide derivatives as potential anti-prostate cancer agents. The compounds (6a-k) were synthesized and characterized by IR, NMR, and Mass Spectral techniques. Further, 6g, 6j, and 6k molecule structures were confirmed by single crystal XRD. The in vitro anticancer activities of the compounds (6a-k) were tested against LNCaP and PC3 cell lines (Divakar et al., 2017). The above remarkable considerations and pharmaceutical and industrial applications prompted us to synthesize a new series of benzohydrazide derivatives via the Schiff base route.</p><!><p>The protein structure of AR (3V49) was obtained from Protein Data Bank (PDB) with the resolution of 1.70 Å bound with the ligand 4-[(4R)-4-(4-hydroxyphenyl)-3,4-dimethyl-2,5-dioxoimidazolidin-1-yl]-2-(trifluoromethyl)benzonitrile, and prepared using Protein Preparation Wizard in Maestro 11.2. OPLS-2005 was used for the optimization and minimization until the root mean square deviation reached 0.3 Å. Then, Grid was generated using Grid generation Wizard for docking studies. The ligands were drawn and imported 2D structures to the Maestro project table. The 3D structures were prepared, geometrically refined, energy minimized, and assigned appropriate protonation state at pH 7.0 ± 2.0 using LigPrep module. The docking was carried out using GLIDE XP (extra precision method) (Maestro, 2016).</p><!><p>All the reagents and solvents were obtained from commercial sources like Spectrochem, Merck, Alfa Aesar, Sigma-Aldrich, and used without further purification. The reactions were monitored by Merck silica gel 60 F254 thin layer chromatography (TLC) and visualized in a UV light chamber. Column chromatography was performed on Merck silica gel (100–200 mesh). Melting points were recorded by the open capillary method using Raagaa melting point apparatus. The FT-IR spectra using KBr pellets was recorded on a Thermo Scientific FT-IR spectrophotometer. In FT-IR spectra, the compounds exhibited the absorption bands at 3,250–3,150 cm−1 due to the presence of a secondary amine (NH) stretching, at 1,670–1,630 cm−1 due to amide (C = O) stretching, and at 1,605–1,590 cm−1 due to imine (C = N) stretching, which confirmed the formation of the compounds (6a-k). Also, we found C-Cl stretching around 750–700 cm−1 and nitro stretching around 1,550 and 1,350 cm−1. The 1H and 13C NMR spectra were recorded by Bruker AVANCE II at 400 and 100 MHz, respectively using DMSO d6 solvent at room temperature. In 1H NMR spectra, the compounds showed the amide proton signal at δ 11.9–12.3 ppm, CH = N- signal at δ 8.7–8.9 ppm, aromatic proton signals at δ 7.2–8.0 ppm, and 2 aliphatic CH2 protons signals at δ 2.5–3.0 ppm. In 13C NMR spectra, carbonyl (NH-C = O) signal was observed at δ 160–165 ppm, aromatic and CH = N- signals at 120–150 ppm, and aliphatic carbon signals at 20–30 ppm. Mass spectra were recorded using a JEOL GCMATE II LC-Mass spectrometer. The compounds showed significant parent ion peaks. High Performance Liquid Chromatography (HPLC) was used to identify the purity of the compounds, 6g and 6j as shown in Figures S3, S4. The melting point and yield of the compounds were illustrated in Table S1.</p><!><p>The substituted benzoic acid (0.01 mol) was taken in ethanol (10 ml) which had 2–3 drops of Conc. Sulfuric acid added to it. The reaction mixture was refluxed in 80°C for 7 h and monitored by TLC. After the completion, the reaction mixture was neutralized using sodium bicarbonate solution then extracted with ethyl acetate. The organic layer was evaporated and dried well. The product was purified by column chromatography using Ethyl acetate /Hexane as an eluent. To this, ester (2) was added with hydrazine hydrate (0.02 mol) in the presence of ethanol and refluxed in 80°C for 6 h. The completion of the reaction was monitored by TLC. Then, ethanol was removed, and ice-cold water was added. The white color precipitate was formed. Further, the precipitate was filtered and recrystallized from a suitable solvent (3a-3k).</p><!><p>To the solution of 4-dihydronaphthalen-1(2H)-one (0.174 mol), the dry DMF and POCl3 (0.171 mol) was slowly added at 0°C and stirred for 30 min. The mixture was heated to 80°C for 2 h and transferred into the beaker containing 25% cold sodium acetate, then extracted with ether. The organic layer was evaporated and dried well to get residual oil, which solidified on cooling.</p><!><p>Benzohydrazide (0.01 mol) and 1-chloro-3,4-dihydronaphthalene-2-carbaldehyde (0.01 mol) were taken in ethanol (5 ml) with a few drops of acetic acid, refluxed for 8 h, then cooled to room temperature. The excess ethanol was removed through a high vacuum, and the residue was quenched with ice. The precipitate was filtered, dried, and further purified by column chromatography using ethyl acetate and hexane as eluent.</p><!><p>The synthesized compounds (6a-k) were well-characterized by IR, NMR, and HRMS, and all the data are in accordance with the proposed structures as described below.</p><!><p>Yield: 90%; mp: 168–170°C, pale white solid. FT-IR (KBr) ν max: 3204 (NH), 2938, 2851 (aromatic C-H), 1638 (amide C = O), 1592 (imine C = N) cm−1. 1H NMR (DMSO d6, 400 MHz, ppm): δ 12.10 (s, 1H, NH), 8.77 (s, 1H), 7.87 (d, J = 7.2, 2H), 7.64–7.25 (m, 7H), 2.808–2.764 (m, 4H). 13C NMR (DMSO d6, 100 MHz, ppm): δ 163.34, 143.77, 144.5, 136.97, 132.72, 132.63, 131.65, 130.73, 129.80, 129.49, 128.16, 128.00, 127.41, 124.94, 29.50, 26.54, 23.65, 21.51. Mass: m/z Calcd for C18H15ClN2O [M+H]+: 310.08; Found: 310.9758.</p><!><p>Yield: 88%; mp: 168–170°C, pale white solid. FT-IR (KBr) ν max: 3184 (NH), 2940, 2894 (aromatic C-H), 1652 (amide C = O), 1594 (imine C = N) cm−1. 1H NMR (DMSO d6, 400 MHz, ppm): δ 12.9 (s, NH), 12.16 (s, NH), 8.70 (s, 1H, CH = N), 8.48 (s, 1H, CH = N), 7.34–7.72 (m, 16H, Ar H), 2.84–2.94 (m, 4H, (CH2)2), 2.74 (t, J = 4.8 Hz, 2H, CH2), 2.371 (t, J = 8 Hz, H2, 2H CH2). 13C NMR (DMSO d6, 100 MHz, ppm): δ 168.1, 159.5, 159.2, 158.8, 158.5, 157.3, 148.8, 133.9, 128.8, 127.9, 125.6, 121.2, 118.2, 115.2, 108.2, 58.9. Mass: m/z Calcd for C18H14Cl2N2O [M+H]+: 344.05; Found: 344.9302.</p><!><p>Yield: 85%; mp: 242–245°C; Green solid; FT-IR (KBr) ν max: 3165 (NH), 2918, 2849 (aromatic C-H), 1667 (amide C = O), 1601 (imine C = N) cm−1. 1H NMR (DMSO d6, 400 MHz, ppm): δ 12.39 (s, 1H), 8.845 (s, 1H), 8.39 (d, J = 8.8 Hz, 2H), 8.1735 (d, J = 8.4 Hz, 2H), 7.681–7.659 (m, 2H), 7.367–7.321 (m, 2H), 2.885–2.814 (m, 4H); 13C NMR (DMSO d6, 100 MHz, ppm): δ 161.86, 149.82, 147.9, 139.29, 138.10, 133.58, 132.52, 131.59, 129.7, 128.06, 127.47, 125.09, 29.49, 26.90, 23.61. Mass: m/z Calcd for C18H14ClN3O3 [M+H]+: 355.07; Found: 354.0069.</p><!><p>Yield: 86%; mp: 213–215°C; pale yellow solid. FT-IR (KBr) ν max: 3166 (NH), 2954, 2853 (aromatic C-H), 1669 (amide C = O), 1591 (imine C = N), 1536, 1348 (NO2) cm−1. 1H NMR (DMSO d6, 400 MHz, ppm): δ 12.333 (s, 1H), 8.843 (s, 1H), 8.783 (s, 1H), 8.469–8.446 (m, 1H), 8.4205 (d, J = 8 Hz, 1H), 7.875–7.835 (m, 1H) 7.676–7.653 (m, 2H), 7.360–7.315 (m, 2H), 2.882–2.815 (m, 4H); 13C NMR (DMSO d6, 100 MHz, ppm): δ 161.34, 148.27, 146.95, 138.05, 134.99, 134.67, 135.51, 132.53, 131.59, 130.82, 130.00, 128.04, 127.45, 126.90, 125.07, 122.77, 26.90, 23.62. Mass: m/z Calcd for C18H14ClN3O3 [M+H]+: 355.07; Found: 354.0234.</p><!><p>Yield: 90%; mp: 208–210°C; pale pink solid. FT-IR (KBr) ν max: 3189 (NH), 2950, 2887 (aromatic C-H), 1648 (amide C = O), 1597 (imine C = N) cm−1. 1H NMR (DMSO d6, 400 MHz, ppm): δ 12.079 (s, 1H), 8.820 (s, 1H), 7.984 (s, 1H), 7.8965 (d, J = 7.6 Hz, 1H), 7.689–7.651 (m, 2H), 7.5885 (d, J = 7.6 Hz, 1H) 7.353–7.307 (m, 3H), 2.873–2.807 (m, 4H); 13C NMR (DMSO d6, 100 MHz, ppm): δ 162.03, 146.59, 138.59, 135.59, 133.79, 133.23, 132.55, 132.22, 131.65, 131.03, 129.94, 128.02, 127.79, 127.44, 127.03, 125.03, 26.91, 23.62. Mass: m/z Calcd for C18H14Cl2N2O [M+H]+: 344.05; Found: 344.9384.</p><!><p>Yield: 92%; mp: 128–130°C; pale brown solid. FTIR (KBr) ν max: 3209 (NH), 2933, 2852 (aromatic C-H), 1635 (amide C = O), 1593 (imine C = N) cm−1. 1H NMR (DMSO d6, 400 MHz, ppm): δ 11.945 (s, 1H), 8.828 (s, 1H), 7.849 (d, J = 7.2, 2H), 7.663–7.642 (m, 1H), 7.344–7.296 (m, 5H), 2.831 (m, 4H), 2.088 (s, 3H); 13C NMR (DMSO d6, 100 MHz, ppm): δ 161.34, 145.77, 142.3, 137.97, 132.72, 132.63, 131.85, 130.73, 129.80, 129.49, 128.16, 128.00, 127.41, 124.96, 29.50, 26.94, 23.63, 21.51. Mass: m/z Calcd for C19H17ClN2O [M+H]+: 324.10; Found: 324.9966.</p><!><p>Yield: 86%; mp: 190–192°C, Appearance—Cream solid. FTIR (KBr) ν max: 3222 (NH), 2952, 2840 (aromatic C-H), 1639 (amide C = O), 1605 (imine C = N) cm−1. 1H NMR (DMSO d6, 400 MHz, ppm): δ 11.928 (s, 1H), 8.812 (s, 1H), 7.9345 (d, J = 8.4, 2H), 7.657–7.636 (m, 1H), 7.340–7.295 (m, 3H), 7.068 (d, J = 8.8, 2H), 3.840 (s, 3H), 2.852–2.814 (m, 4H); 13C NMR (DMSO d6, 100 MHz, ppm): δ 162.31, 162.62, 145.41, 137.94, 132.65, 132.51, 131.89, 130.10, 129.75, 127.98, 127.45, 125.64, 124.91, 114.21, 55.9, 26.95, 23.67. Mass: m/z Calcd for C19H17ClN2O2 [M+H]+: 340.10; Found: 340.9898.</p><!><p>Yield: 84%; mp: 238–240°C, Appearance—Light yellow solid. FTIR (KBr) ν max: 3215 (NH), 2955, 2842 (aromatic C-H), 1642 (amide C = O), 1595 (imine C = N) cm−1. 1H NMR (DMSO d6, 400 MHz, ppm): δ 12.098 (s, 1H), 8.814 (s, 1H), 7.8895 (d, J = 8.4, 2H), 7.7565 (d, J = 8.4, 2H), 7.6455 (d, J = 4.4 1H), 7.3105 (d, J = 10.8, 3H). 2.857–2.801 (m, 4H); 13C NMR (DMSO d6, 100 MHz, ppm): δ 162.59, 146.39, 136.01, 133.16, 132.68, 132.57, 132.02, 131.69, 131.22, 130.24, 129.89, 129.69, 128.01, 127.42, 126.22, 125.01, 26.33, 23.63. Mass: m/z Calcd for C18H14BrClN2O [M+H]+: 388.00; Found: 388.9875.</p><!><p>Yield: 85%; mp: 238–240°C, Appearance—Light yellow solid. FTIR (KBr) ν max: 3210 (NH), 2945, 2849 (aromatic C-H), 1648 (amide C = O), 1592 (imine C = N) cm−1. 1H NMR (DMSO d6, 400 MHz, ppm): δ 12.086 (s, 1H), 8.801–8.790 (m, 1H), 7.944 (s, 2H), 7.6045 (d, J = 5.2, 3H), 7.310–7.270 (m, 3H), 2.805 (d, 4H); 13C NMR (DMSO d6, 100 MHz, ppm): δ 162.45, 146.36, 137.98, 137.24, 133.14, 132.55, 132.29, 131.66, 130.05, 129.87, 129.06, 128.25, 127.99, 127.39, 124.99, 26.90, 23.61. Mass: m/z Calcd for C18H14Cl2N2O [M-H]+: 344.05; Found: 343.0401.</p><!><p>Yield: 86%; mp: 233–235°C, Appearance—Light yellow solid. FTIR (KBr) ν max: 3220 (NH), 2946, 2842 (aromatic C-H), 1652 (amide C = O), 1598 (imine C = N) cm−1. 1H NMR (DMSO d6, 400 MHz, ppm): δ 11.857 (s, 1H), 10.216 (s, 2H), 8.795 (s, 1H), 7.834 (d, 2H), 7.650–7.628 (m, 1H) 7.330–7.269 (m, 3H), 6.8775 (d, J = 8.4, 2H), 4.428–4.403 (m, 1H), 2.847–2.85 (m, 4H); 13C NMR (DMSO d6, 100 MHz, ppm): δ 163.18, 161.31, 145.17, 137.91, 132.67, 132.37, 131.92, 130.25, 129.70, 127.96, 127.38, 124.89, 124.08, 115.52, 26.96, 23.68. Mass: m/z Calcd for C18H15ClN2O2 [M-H]+: 326.08; Found: 325.0741.</p><!><p>Yield: 86%; mp: 218–220°C, Appearance—Pale Yellow. FTIR (KBr) ν max: 3220 (NH), 2946, 2842 (aromatic C-H), 1652 (amide C = O), 1598 (imine C = N) cm−1. 1H NMR (DMSO, 400 MHz, ppm): δ 2.860–2.793 (m, 4H), 12.099 (s, 1H), 8.805 (s, 1H), 7.343–7.284 (m, 3H), 7.526–7.486 (m, 1H), 7.660–7.638 (m, 1H), 8.115 (s, 1H), 7.9411–7.922 (m, 1H), 7.826–7.796(m, 1H). 13C NMR (DMSO, 100 MHz, ppm): δ 161.97, 146.58, 138.02, 135.77, 135.10, 133.27, 132.55, 131.64, 131.26, 130.62, 129.32, 128.01, 127.42, 125.03, 122.24, 26.91, 23.62, 21.22. Mass: m/z Cal.cd for C18H15ClN2O [M+H]+: 388.00; Found: 388.9874.</p><!><p>The single crystal X-ray diffraction data of 6g, 6j, and 6k was recorded at the University of Mysore, India (Bruker axs kappa apex2, CCD diffractometer using graphite monochromated MoK radiation). The structure was solved by SHELXS-97 and refined by full-matrix least square methods in SHELXL-97. All non-hydrogen atoms were refined anisotropically and the hydrogen atoms were refined isotropically.</p><!><p>The in vitro antiproliferation activities were performed by using prostate cancer cell lines (LNCaP & PC-3) obtained from the National Center for Cell Science (NCCS), Pune, India. The cells were cultured in RPMI medium with 10% FBS and incubated at 37°C with 5% CO2. Approximately 5,000 cells/well were seeded into a 96 well cell culture plate. The antiproliferation activities of the compounds were tested by adding different concentrations and incubated for 48 h to evaluate the activities. Then, 10 μL of MTT (5 mg/ml) was added to each well and incubated for an additional 4 h, and the absorbance was measured at 560 nm. The experiments were performed in triplicates, and the standard deviation was calculated (Hayon et al., 2003; Chen, 2011).</p><!><p>For conventional AO/EtBr apoptosis assay 5X105 cells/well were seeded in a 6 well plate and incubated at 37°C and 5% CO2 for overnight. The cell lines were treated with the test compound for 48 h. The concentration of dye was prepared as a 1:1 ratio. Each dye concentration was 100 μg/ml. Cells were stained with 8 μL of dye for 5 min, and images were taken immediately by fluorescence microscopy (Maliheh et al., 2012; Tayyaba et al., 2016).</p><!><p>The LNCaP cells were cultured in medium supplemented with 10% charcoal-stripped FBS. The cells were treated for 24 h with ARA3 (1 μM) or bicalutamide (10 μM) in the presence or absence of 1 nM DHT. The cells were lysed by RIPA buffer supplemented with protease and phosphatase inhibitor cocktail. The lysate was centrifuged at 16,000 g for 20 min. The supernatant equivalent to 40 μg of protein was subjected to SDS PAGE gel electrophoresis. The resolved proteins were transferred to a PVDF membrane. The transferred proteins were then hybridized with antibodies, AR (sc816) or p-AR213 (sc135635) or procaspase-3 (sc7148) or akt1 (sc5289) or p-akt1 473 (sc293125) antibodies. The expression of β-actin was measured as an internal control using β-actin antibody (ab8227). The proteins were detected by horseradish peroxidase-labeled anti-rabbit or anti-mouse IgG monoclonal or polyclonal antibody, and a chemiluminescence detection kit (ECL, GE healthcare). The intensities of the PCR product in the agarose gel 402 were scanned with G: BOX (Syngene) image scanner.</p><!><p>A DNMT inhibition assay was carried out using EpiQuik DNA methyltransferase activity/inhibition screening assay kit (Epigentek, Brooklyn, NY, USA) according to the manufacturer's instruction. Two different concentrations of each test compound were screened using the Enzyme-linked immunosorbent assay (ELISA) kit. Their enzyme inhibitory activity was quantified by colorimetric assay method. The assay was performed in duplicates for each concentration (Robertson, 2002).</p><!><p>The two-dimensional structures of the compounds (6a-k) were imported to the Maestro project table and minimized using LigPrep. The program, Qikprop40 module (Schrödinger software) was used for the in silico determination of pharmacokinetic properties. A preliminary test of the drug-likeness of the compounds was calculated in accordance with Lipinski's rule of five. To obey Lipinski's rule of five and rule of three, the compounds required molecular weight (mol_MW) of less than 500 amu, not more than 5 and 10 hydrogen bond donors (Donor HB), hydrogen bond acceptors (Accpt HB), the partition coefficient between octanol and water (QPlogPo/w) to be <5, (QPlogS > −5.7), (QPPCaco > 22 nm/s) and, respectively, primary metabolities < -5.7. The compounds which have more than one violation of these rules are not considered as orally active molecules. The properties such as IC50 value for the blockage of HERG K+ channels, the percentage of human oral absorption, blood-brain barrier permeability, and aqueous solubility were also predicted to identify the bioactivity of the compounds.</p><!><p>Molecular Dynamic (MD) stimulation studies were carried out by Desmond (2012) module of Schrödinger software. The docked conformer of benzohydrazides with a good Glide-score was chosen for the Molecular Dynamics simulations study with an OPLS-2005 force field. The protein-ligand complex was bounded with a predefined TIP3P water model (Jorgensen et al., 1983) in orthorhombic box. The volume of the box was minimized, and the overall charge of the system was neutralized by adding Na+ and Cl− ions. The pressure and temperature were kept constant at 300 K and 1.01325 bar using Nose-Hoover thermostat (Hoover, 1985) and Martyna-Tobias-Klein barostat (Martyna et al., 1994) methods. The simulations were performed using NPT ensemble by considering number of atoms, pressure, and timescale. During simulations, the long-range electrostatic interactions were calculated using the Particle-Mesh-Ewald method (Essmann et al., 1995). The Root Mean Square Deviation (RMSD) was used to measure the average change in displacement of a selection of atoms for a particular frame concerning a reference frame. This was calculated for all structures in the trajectory.</p><!><p>A molecular docking approach was followed for designing of benzohydrazide derivatives by using Schrödinger (Maestro 11.2) software (Maestro, 2016). The 3D structure of the human androgen receptor alpha ligand binding domain with the Selective Androgen Receptor Modulators (SARM) inhibitor (PDB ID: 3V49) was obtained from the Protein Data Bank, refined the structure and was used for the study (Nique et al., 2012; Lakshmithendral et al., 2019). Docking studies were carried out to find the potential binding affinity and the interaction between the compounds (6a-k) and AR protein, as shown in Figure S1. Compounds (6a-k) have van der Waals interactions with surrounding hydrophobic residues TRP741, LEU873, ARG752, MET747, MET749, VAL746, MET745, LEU873, MET742, PHE876, VAL903, ILE899, ILE898, LEU701, LEU704, LEU707, PHE891, and form hydrogen bonds through a hydrazide group with two residues LEU704, HIE 876 in the Helix 12. The highest scoring pose of compound 6j and the drug bicalutamide from docking studies is shown in Figure 2 with receptor residues. From docking studies, we found that the compound 6j showed the highest glide score (−11.776 kcal/mol) and glide energy of −52.02 kcal/mol, respectively. Compounds (6a-k) also showed binding affinity to AR, and their docking scores are shown in Table 1. The compounds (6b, 6d, 6a, 6g) also have higher binding energy in the range of −10.767 to −9.893 kcal/mol and glide energy vary from −49.591 to −44.96 kcal/mol. The standard Bicalutamide glide score was −11.064 kcal/mol and binding energy was −44.712 kcal/mol, respectively. These studies were conventional, and the results were reported.</p><!><p>Predicted binding mode of compound 6j (A,C) and standard Bicalutamide (B,D) in AR LBD (PDB: 3V49) highlighting the H-bond, Pi-Pi stacking, and specific hydrophobic interactions.</p><p>Predicted binding energy and mode of the compounds, 6a-k.</p><p>Glide evdw, van der Waals interaction energies; Glide ecoul, Coulomb interaction energies. Bold values indicate the highest docking score.</p><!><p>Molecular docking scores intend the synthesis of the compounds (6a-k). The series of compounds were synthesized, as shown in Scheme 1. The substituted benzoic acid was first converted to their corresponding ester (esterification) followed by refluxing with hydrazine hydrate in ethanol at 80°C for 6–7 h, which resulted in the formation of acid hydrazides (3a-k) (Chidananda et al., 2012). 1-chloro-3,4-dihydronaphthalene-2-carbaldehyde (5) was prepared by using alpha-tetralone (4) under 0–5°C with reagents DMF and POCl3 (Perumal et al., 2012). Further, the compounds (6a-k) were obtained by the reaction of compounds (3a-k) with 1-chloro-3,4-dihydronaphthalene-2-carbaldehyde (5) refluxing in the presence of ethanol for 14–16 h (Rapartia et al., 2009; Sirisoma et al., 2009). All the compounds were purified by column chromatography, and their structures were confirmed with spectral techniques. The melting point and yield of the compounds were shown in Table S1. The compounds, 6g, 6j, and 6k, were crystallized by slow evaporation at room temperature in DMSO/chloroform, ethanol/acetonitrile solvents (Sheldrick, 1997). The compound 6g belongs to the monoclinic system with space group belonging to P21/a, a = 8.431(3) Å, b = 15.653(5) Å, c = 12.771(4) Å, α = 90°, β = 97.522(12)°, γ = 90°, Z = 4. The compound 6j belongs to the monoclinic system with space group belonging to P21/c, a = a = 5.2090 (4) Å, b = 25.128 (2) Å, c = 12.7739 (10) Å, α = 90°, β = 96.138(2)°, γ = 90°, Z = 2. Similarly, the compound 6k belongs to a Tetragonal system with space grouping I 41/a, a = 22.136 (3) Å, b = 22.136 (5) Å, c = 14.5289 (4) Å, α = 90°, β = 90°, γ = 90°, Z = 16 (Sheldrick, 1990, 2015). The ORTEP diagram of the compounds 6g, 6j, and 6k were illustrated in Figure 3. The unit cell diagram of the compounds 6g, 6j, and 6k are shown in Figure S2. The details of the crystal data and structure refinement parameters were summarized in Tables S2–S4. The purity of compounds 6g and 6j were analyzed using HPLC. The compound 6g showed the purity of 99.42%, and 6j had the purity of 99.90% as illustrated in Figures S3, S4. All the spectroscopic measurements confirm the compounds' structures and purity.</p><!><p>Synthesis of benzohydrazide derivatives.</p><p>ORTEP structure of Compounds 6g, 6j, and 6k.</p><!><p>The compounds (6a-k) were evaluated for the anticancer activities against the prostate cancer cell lines (PC-3, LNCaP) and non-cancerous cell line (3T3), as compared with the standard, Bicalutamide by using MTT assay (Divakar et al., 2017). The results demonstrated that the compounds (6a-k) showed significant cytotoxic potential in PC-3 and LNCaP cell lines in the IC50 range of 7–38 μM, as compared with Bicalutamide (IC50-11.06), listed in Table 2. Of these compounds, the 6g and 6j exhibited significant dose-dependent activities in LNCaP cell lines (IC50 7.17 ± 1.87 and 10.45 ± 0.7, respectively) and PC-3 (IC50 32.09 ± 0.86 and 44.65 ± 0.32, respectively). The percentage of cell viability of the compound, 6g in LNCaP and PC-3 cell lines were shown in Tables S5, S6. The anticancer activities were shown in Figures S5, S6 by plotting percentage cell viability against the concentration of the compounds. These studies suggests that the methoxy group at R influenced the anticancer activities (Plouvier et al., 1995). In addition, the replacement of the hydroxyl group at R slightly decreased the IC50 value. Similarly, the nitro group at R showed less IC50 value, when compared with the hydroxyl group. The compounds, 6g and 6j did not show the toxicity against 3T3 cells line, which reliably represents normal human mammary cells. From the series, 6g and 6j showed more promise for anti-prostate cancer activity in LNCaP and PC-3 cell lines. It is evident that the groups, especially OCH3, OH, and NO2 substitutions, showed more anticancer activities in the cell proliferation assays. The compound 6g exhibited dose-dependent activity in LNCaP cell lines. Analysis of the AO/EtBr staining revealed that the synthetic compound 6g reduced cell viability of LNCaP cell lines and induced the apoptosis. Chromatin condensation and Cell membrane blebbing is the first phase of cell disassembly during apoptosis, which was observed in the treatment group. Live cells had normal nuclei staining, which presented green chromatin with organized structures. Apoptotic cells containing condensed or fragmented chromatin (green or orange) are shown in Figure 4. As described earlier, the western blotting analysis was carried out and is shown in Figure S7. The androgen significantly increased the protein expression of AR (~1.2 fold) and p-AR 213 (~1.4 fold), compared to the control (Saravanan et al., 2017). Bicalutamide and 6g significantly decreased the androgen-stimulated AR and p-AR protein expression. The notable difference between bicalutamide and 6g is that 6g could significantly decrease the protein expression of p-AKT (~0.7 fold) and procaspase-3 (~0.6 fold), while bicalutamide treatment did not show a significant effect on those proteins. The significance (P < 0.05) was obtained by applying one-way Analysis of Variance (ANOVA) followed by a post-hoc turkey test, as shown in Figure 5. The compound 6g showed DNMT1 enzyme inhibitory activity at 1 and 100 μM concentration, as shown in Table 3. A concentration gradient response was observed, and the high concentration exhibited 23.9% in the enzyme inhibition assay. So, this compound can be used for DNMT1 isoform targeted therapies to treat prostate cancer.</p><!><p>Bioactivity of test and standard compounds.</p><p>IC50 of the compounds stimulated by 1nM DHT.</p><p>The values are the mean ± standard deviation (SD) of three independent experiments performed in triplicate.</p><p>Positive control.</p><p>Bold values indicate the highly active compound.</p><p>The morphological changes in 6g and Zebularine treated in prostate cancer cell line. The treated and untreated cells were stained with Acridine orange and Ethidium bromide and observed under an inverted fluorescent microscope. Live cells appear green, late apoptotic cells are appearing orange in color, and necrotic cells are colored red.</p><p>AR: DHT significantly (#) increased the AR expression compared to control. Bicalutamide and 6g significantly (*) reduced the DHT stimulated expression of AR. p-AR: DHT significantly (##) increased the p-AR 213 expression compared to control. Bicalutamide and 6g significantly (*) reduced the DHT stimulated expression of p-AR 213. AKT: Expression of AKT1 protein. There is no significant difference with the treatments. p-AKT: 6g significantly (*) reduced the DHT stimulated expression of p-AKT1 473. Procaspase-3: 6g significantly decreases the expression of procaspase-3 compared to control. The experiments were conducted in duplicates and the average value was represented as mean ± S.D. The significance (P < 0.05) was obtained by applying one way ANOVA followed by post-hoc turkey test. (***P < 0.001, **P < 0.01, * P < 0.05).</p><p>DNMT inhibition assay.</p><!><p>Several drugs are failed during clinical trials due to poor adsorption, distribution, metabolism, and elimination properties, so ADME was paid large attention in drug discovery. It was utilized for the screening and optimization of active compounds (6a-k). The ADME has an important role in the assessment of drug-likeness and pharmacokinetic properties. We have analyzed all the predicted ADME parameters, such as hydrogen bond acceptors (Acceptor HB), hydrogen bond donors (Donor HB), molecular mass, and so on, as shown in Table S7. All the compounds fell under Lipinski's rule of five and pharmacokinetic properties. There were no more than 5 hydrogen bonds donors, 10 hydrogen bond acceptors, and all molecules masses had <500 Daltons. All the compounds had a good percentage of human oral absorption. The IC50 value of HERG K+ (channel blockage (−6.2 to −6.9), (QPlogHERG) of all the tested compounds showed a good range of values (Schrödinger., 2019 -1). In this study, the compounds obeyed the recommended drug-likeness properties and were orally active. All the pharmacokinetics results were in good agreement with previous reports.</p><!><p>The benzohydrazide-3V49 complex was immersed in the orthorhombic box with a TIP3P water solvent model used for simulations using OPLS-2005 force field. After the solvent system was formed, the simulation was equilibrated for 5 ns by Dynamic simulation (Anand et al., 2015). The quality study of the simulations was performed as depicted in Table 4. The root mean square deviation (RMSD) plots are shown in Figure 6 indicating the benzohydrazide−3V49 complex reached its stable form. Protein RMSD started from 0.8 Å and later stabilized at 1.4 Å, whereas, in ligand RMSD started from 2.4 Å and stabilized around 1.5 Å. Changes of the order of 1–3 Å are perfectly acceptable for small, globular proteins. Changes much larger than that, however, indicate that the protein is undergoing a large conformational change during the simulation. It is also important that simulation converges; the RMSD values stabilize around a fixed value. The nitrogen atom of the hydrazide ring displayed hydrogen bond interaction with LEU704 (68%) and ASN705 (20%), and bridged with H2O. The current geometric criteria for the protein-ligand H-bond is: a distance of 2.5 Å between the donor and acceptor atoms (D—H•••A); a donor angle of 120° between the donor-hydrogen-acceptor atoms (D—H•••A); and an acceptor angle of 90° between the hydrogen-acceptor-bonded atoms (H•••A—X). The π-π stacking interaction was also observed with TRP 741 (74%) and PHE764 (73%) in the binding pocket of 3V49. The 2D interaction poses, and the histogram chart are depicted in Figure 7.</p><!><p>Summary of simulation quality analysis after equilibration.</p><p>The RMSD plot obtained for compound 6g and 3V49 complex.</p><p>The histogram chart and the percentage of interactions in molecular dynamic simulations.</p><!><p>In this present study, a series of molecules were designed by the molecular docking approach for androgen receptors. The binding affinities of the ligand and the precise interactions, molecular dynamics simulations validate the results of molecular docking. Further, the compounds (6a-k) were synthesized and demonstrated anti-prostate cancer activity. The compounds 6g and 6j were the most promising among the series, especially against LNCaP cell lines, IC50 values 7.17 ± 1.87 and 10.45 ± 0.7 μM, respectively. Also, the compounds 6a-j were screened against 3T3 cells line for the toxicity. The binding energy of compounds 6g and 6j with the androgen receptor were significantly noticeable (−49.311 and −52.02 kcal/mol). Moreover, it is evident from the results that the methoxy and hydroxy substitutions at R have better binding affinities and anticancer activities than others. Also, the compound 6g significantly decreased the mRNA expression of AR-stimulated genes and induced apoptosis. We intend to further investigate the target site and study the in vivo anticancer activity of the active compounds.</p><!><p>All datasets generated for this study are included in the manuscript and/or the Supplementary Files.</p><!><p>HA, RE, and SK conceived and designed the experiment. HA and NM performed the experiment. HA, RE, and KL analyzed the data. HA and RE wrote the manuscript. MR and AB have done a critical revision of the manuscript for important intellectual content. NL carried out the analysis of single crystal XRD structures. All authors have contributed to the final version and approved the final manuscript.</p><!><p>The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.</p>

PubMed Open Access

Carnitine is a pharmacological allosteric chaperone of the human lysosomal α-glucosidase

AbstractPompe disease is an inherited metabolic disorder due to the deficiency of the lysosomal acid α-glucosidase (GAA). The only approved treatment is enzyme replacement therapy with the recombinant enzyme (rhGAA). Further approaches like pharmacological chaperone therapy, based on the stabilising effect induced by small molecules on the target enzyme, could be a promising strategy. However, most known chaperones could be limited by their potential inhibitory effects on patient’s enzymes. Here we report on the discovery of novel chaperones for rhGAA, L- and D-carnitine, and the related compound acetyl-D-carnitine. These drugs stabilise the enzyme at pH and temperature without inhibiting the activity and acted synergistically with active-site directed pharmacological chaperones. Remarkably, they enhanced by 4-fold the acid α-glucosidase activity in fibroblasts from three Pompe patients with added rhGAA. This synergistic effect of L-carnitine and rhGAA has the potential to be translated into improved therapeutic efficacy of ERT in Pompe disease.

carnitine_is_a_pharmacological_allosteric_chaperone_of_the_human_lysosomal_α-glucosidase

Introduction<!><!>Introduction<!>Reagents<!>Enzyme characterisation<!>Thermal stability of rhGAA<!>Fibroblast cultures<!>Incubation of fibroblasts with rhGAA and GAA assay<!>Immunofluorescence analysis and confocal microscopy<!>L-CAR improves rhGAA stability in vitro<!><!>L-CAR improves rhGAA stability in vitro<!>Effect on rhGAA stability by the combined action of allosteric and active-site directed PCs<!><!>Effect on rhGAA stability by the combined action of allosteric and active-site directed PCs<!><!>Effect on rhGAA stability by the combined action of allosteric and active-site directed PCs<!>Effect of L-CAR in PD fibroblasts<!><!>Effect of L-CAR in PD fibroblasts<!><!>Effect of L-CAR in PD fibroblasts<!><!>Effect of L-CAR in PD fibroblasts<!><!>Discussion<!>

<p>Glycogen storage disease type 2, or Pompe disease (PD, OMIM 232300) is an inborn metabolic disorder caused by the functional deficiency of the acid lysosomal α-glucosidase (GAA, acid maltase, E.C.3.2.1.20), the enzyme hydrolysing α-1,4 and α-1,6-glucosidic bonds in glycogen and belonging to family GH31 of the carbohydrate-active enzyme (CAZy) classification (www.cazy.org1). GAA deficiency results in glycogen accumulation in lysosomes and in secondary cellular damage, with mechanisms not fully understood2–5. In PD, muscles are particularly vulnerable to glycogen storage, and disease manifestations are predominantly related to the involvement of cardiac and skeletal muscles. However, central nervous system involvement is emerging as part of the clinical spectrum in infantile-onset patients6.</p><p>It is assumed that to obtain positive therapeutic effects it is enough that the enzymatic activity of GAA is rescued at about 10% of the wild type, meaning that a relatively small increase in activity can mitigate the clinical course2. Therapeutic strategies include the supply of wild type enzymes, such as enzyme replacement therapy (ERT), gene therapy, or small-molecule drugs able to adjust cellular networks controlling protein synthesis, folding, trafficking, aggregation, and degradation, thus facilitating the escape of mutated proteins from the endoplasmic reticulum-associated degradation (ERAD) machinery7–10.</p><p>Since 2006, enzyme replacement therapy (ERT) with recombinant human α-glucosidase has been approved and is currently considered the standard of care for the treatment of PD, improving survival by stabilising the disease course6,11–13. However, limitations are also known, in fact, despite treatment, some patients experience little clinical benefit or show signs of disease progression14. Several factors concur in limiting the therapeutic success of ERT, including the age at the start of treatment15,16, the immunological status of patients17, the insufficient targeting of the enzyme to skeletal muscle18, the possible instability at neutral pH of the recombinant enzyme during the transit to lysosomes19–21, the relative deficiency of the cation-independent mannose-6-phosphate receptor, required for enzyme uptake in muscle cells22,23, and the build-up of the autophagic compartment observed in myocytes24–26.</p><p>For all the reasons pointed above, alternative treatments, like pharmacological chaperone therapy (PCT), would be highly desirable. This approach, which has been designed for the treatment of protein misfolding diseases (PMD), exploits small-molecule ligands that may bind directly to the defective enzymes, templating the folding of proteins in the most stable conformation(s) and preventing their recognition and disposal by the ERAD machinery27–30.</p><p>Most pharmacological chaperones (PC) proposed or used for the treatment of lysosomal storage diseases (LSD) are reversible competitive inhibitors of the target enzymes. Compared to ERT, small-molecule chaperones have important advantages in terms of biodistribution, oral availability, and reduced impact on patients' quality of life. Recent studies have shown that 1-deoxynojirimycin, N-butyl-deoxynojirimycin (DNJ, 1 and NB-DNJ, respectively, Figure 1), and 1-deoxygalactonojirimycin (DGJ, 2), may also potentiate the effects of the enzymes used for ERT in Pompe31 and Fabry diseases, respectively21,32. However, these active-site-directed PCs interfere with the activity of the targeted enzymes5,33. The paradox that an inhibitor can increase the enzymatic activity is explained by the fact that therapeutic levels can be reached at sub-inhibitory intracellular concentrations and that the high concentrations of the natural substrate accumulated in the lysosome or the acidic conditions within the organelle may displace the PC inhibitor from the active site.</p><!><p>Pharmacological chaperones for lysosomal storage diseases. deoxynojirimycin (DNJ) (1), 1-deoxy-galactonojirimycin (DGJ) (2), N-acetylcysteine (NAC) (3), N-acetylserine (NAS) (4), N-acetylglycine (NAG) (5), L-carnitine (L-CAR) (6), D-carnitine (D-CAR) (7), and acetyl-carnitine (A-D-CAR) (8).</p><!><p>An ideal chaperone should be able to protect the enzymes from degradation without interfering with its activity, be largely bioavailable in tissues and organs, reach therapeutic levels in cellular compartments where its action is required, show high specificity for the target enzyme with negligible effects on other enzymes, and have a good safety profile. The extensive search for new PCs is currently being performed by high-throughput screening of chemical libraries34 for molecules specific for GAA35, β-glucoceramidase (GCase)36–38, and β-hexosaminidase39, or by biochemical characterisation of known inhibitors40. However, a reason of major concern on the clinical use is that the majority of PCs identified so far for the treatment of LSD are active-site directed competitive inhibitors33.</p><p>We focussed our search on drugs already approved for human therapy for their rapid clinical translation without the need for phase I clinical trials. We found that N-acetylcysteine (NAC, 3, Figure 1), a known pharmaceutical drug, and the related aminoacids N-acetylserine (NAS, 4) and N-acetylglycine (NAG, 5), structurally unrelated to known inhibitors of GAA, behave like novel allosteric PCs for this enzyme41. These molecules stabilise rhGAA at non-acidic pH, enhanced the residual activity of mutated GAA, and improved the efficacy of rhGAA used for ERT in Pompe disease41,42. The high-resolution 3 D-structure of rhGAA in complex with NAC allowed to identify two binding sites for this PC in regions distant from the active site, and to explain the chaperoning activity of NAC43.</p><p>Following the same approach, here we report on the results of screening for other putative allosteric chaperones, already approved as drugs or nutraceuticals. We found that L-carnitine (L-CAR, 6 in Figure 1), D-carnitine (D-CAR, 7), and the related compound acetyl-D-carnitine (A-D-CAR, 8) can stabilise rhGAA at non-lysosomal pH and improve the activity of GAA in PD patient's fibroblasts. Therefore, these molecules are novel potential pharmacological chaperones with excellent perspectives for the treatment of Pompe disease alone and in combination with ERT.</p><!><p>rhGAA (α-glucosidase, Myozyme), was from Genzyme Co, Cambridge, MA, USA. As a source of enzyme, authors used the residual amounts of the reconstituted recombinant enzyme prepared for the treatment of PD patients at the Department of Translational Medical Sciences of the University of Naples, "Federico II". D-CAR, A-D-CAR were from Sigma-tau; L-CAR, DNJ, and 4NP-Glc were purchased from Sigma–Aldrich.</p><!><p>The standard activity assay of rhGAA was performed in 200 µL by using 0.2 µM at 37 °C in 100 mM sodium acetate pH 4.0 and 20 mM 4NP-Glc. The reaction was started by adding the enzyme. After 2 min incubation time the reaction was blocked by adding 800 µL of 1 M sodium carbonate pH 10.2. Absorbance was measured at 420 nm at room temperature and an extinction coefficient of 17.2 mM−1 cm−1 was used to calculate enzymatic units. One enzymatic unit is defined as the amount of enzyme catalysing the conversion of 1 μmol substrate into the product in 1 min, under the indicated conditions.</p><p>The effect of pH on the rhGAA stability was measured by preparing reaction mixtures containing 6.8 µM of enzyme in the presence of 50 mM sodium phosphate, pH 7.4. After incubations at 37 °C, aliquots were withdrawn at the times indicated in Results and the residual α-glucosidase activity was measured with the standard activity assay described above. To test the effect on the pH stability of rhGAA by the chaperons, experiments were performed as described above by adding to the reaction mixtures the amounts of the different compounds indicated in the Results.</p><!><p>Thermal stability experiments of rhGAA were performed as described in Porto et al.41 and the dissociation constant of L-CAR was measured as described in Roig-Zamboni et al.43. Briefly, 0.9 µM of the enzyme were incubated in the absence and in the presence of L-CAR, D-CAR A-D-CAR, NAC, and DNJ in 25 mM sodium phosphate buffer, pH 7.4, and 150 mM NaCl.</p><p>The effect of L-CAR on rhGAA stability was tested by analysing the specific activity. L-CAR at various concentrations was incubated with rhGAA and the enzymatic specific activity was measured after 5 h of incubation at pH 7.4.</p><p>Thermal stability scans were performed at 1 °C/min in the range 25–95 °C in a Real-Time LightCycler (Bio-Rad). Differential Scanning Fluorimetry (DSF) scans were performed at ten concentrations of L-CAR (from 2 to 20 mM) and changes in the fluorescence of SYPRO Orange dye were monitored as a function of temperature at pH 7.4. Thermal scans were performed in triplicate and melting temperatures were calculated according to Niesen et al.44. For the determination of the dissociation constant (KD) of L-CAR experimental data were best fitted according to a simple cooperative model equation reported in Vivoli et al.45 by using the software GraphPAD Prism (GraphPad Software, San Diego, CA, USA). The melting temperature values were plotted as a function of ligand concentration.</p><!><p>Fibroblasts from PD patients were derived from skin biopsies after obtaining the informed consent of patients. Normal age-matched control fibroblasts were available in the laboratory of the Department of Paediatrics, Federico II University of Naples. All cell lines were grown at 37 °C with 5% CO2 in Dulbecco's modified Eagle's medium (Invitrogen, Grand Island, NY, USA) and 20% foetal bovine serum (Sigma–Aldrich, St Louis, MO, USA), supplemented with 2 mM/L glutamine, 100 U/ml penicillin and 100 µg/ml streptomycin.</p><!><p>To study the rhGAA uptake and correction of GAA activity in PD fibroblasts, the cells were incubated with 50 µM rhGAA for 24 h, in the absence or in the presence of 10 mM L-CAR. Untreated cells were used for comparison. After the incubation, the cells were harvested by trypsinization and disrupted by 5 cycles of freezing and thawing.</p><p>GAA activity was assayed by using the fluorogenic substrate 4-methylumbelliferyl-α-D-glucopyranoside (4MU) (Sigma–Aldrich) according to a published procedure31. Briefly, 25 µg of cell homogenates were incubated with the fluorogenic substrate (2 mM) in 0.2 M acetate buffer, pH 4.0, for 60 min in incubation mixtures of 100 µl. The reaction was stopped by adding 1 ml of glycine-carbonate buffer, 0.5 M, pH 10.7. Fluorescence was read at 365 nm (excitation) and 450 nm (emission) on a Promega GloMax Multidetection system fluorometer. Protein concentration in cell homogenates was measured by the Lowry assay.</p><!><p>For immunofluorescence studies, cells (human fibroblasts) grown on coverslips were fixed using methanol (5 min at −20 °C to study the colocalization GAA-LAMP2), permeabilized using 1% PBS (phosphate-buffered saline)—Triton 0,1% and blocked with 0.05% saponin, 1% BSA diluted in 1% PBS (blocking solution) at room temperature for 1 h. The cells were incubated with the primary antibodies anti-GAA rabbit polyclonal antibody (PRIMM) and anti-LAMP2 mouse monoclonal antibody (Santa Cruz Biotechnology), and diluted in blocking solution overnight at 4 °C. Then, cells were washed with 1% PBS and then incubated with appropriate autofluorescent secondary antibodies (anti-rabbit or anti-mouse antibodies conjugated to Alexa Fluor 488 or 596) and DAPI (4′,6-diamidino-2-phenylindole, Invitrogen) in 0.05% saponin, 3% BSA, 1% PBS. Samples were then washed, mounted with Mowiol (Sigma), and examined with a Zeiss LSM700 confocal microscope. Colocalization and quantitative analysis were performed with Fiji (ImageJ, NIH, USA) software.</p><!><p>In the framework of our search for novel PCs, which led to the identification of NAC/NAS/NAG (3–5 in Figure 1)41, we embarked on the search of molecules already approved as pharmaceutical drugs and/or nutraceuticals that can be rapidly introduced in therapeutic treatments without the need of long and expensive clinical trials. Among the molecules considered, L-carnitine (6) was identified as a possible target. In addition, D-carnitine and the related compound acetyl-D-carnitine were also analysed. L-CAR is a well-known conditionally essential micronutrient and nutraceutical46,47, whereas for D-CAR side effects have been documented, including toxicity in patients treated with dialysis, and in rats and fishes48–50. In fact, most of the study was performed on L-CAR.</p><p>To test this molecule on GAA, we analysed its effect on the pH stability of the enzyme similarly to previous studies on lysosomal enzymes20,41,51. In particular, we analysed rhGAA residual activity on 100 mM 4-nitrophenyl-α-D-glucopyranoside (4NP-Glc) in 100 mM sodium acetate buffer, pH 4.0. These assays, in which rhGAA is optimally active and stable for up to 24 h, were used to test the enzyme stability. In fact, at acidic or neutral pHs (pH 3.0 and 7.0, respectively), which are lower and higher, respectively, to the one of the lysosomal compartments, the enzyme halved its activity in about 5 h41.</p><p>L-CAR, already at the concentration of 10 mM, rescued the activity of rhGAA on 4NP-Glc after 5 h of incubation at pH 7.4 (Figure 2(a)). The stabilising effect on the rhGAA activity was maintained even after 48 h of incubation in the presence of 20 mM L-CAR (Figure S1(a)). No effect on the specific activity of rhGAA was observed when L-CAR at any concentration was included in the α-glucosidase assay, indicating that it did not interact with the active site of the enzyme (Figure S1(b)).</p><!><p>Comparison of the effect of L-carnitine on the stability of rhGAA. (a) Effect of L-CAR on the rhGAA stability; (b) Effect of L-CAR on the structural stability of rhGAA; (c) Summary of the Tms measured by DSF; (d) Determination of the KD rhGAA-L-CAR by DSF.</p><!><p>Interestingly, L-CAR increased in a dose-dependent manner also the structural stability of rhGAA as analysed by DSF (Figure 2(b)). The variations of the melting temperature (ΔTm) increased by about 2 °C at every 2 mM increment of L-CAR concentration (Figure 2(c)).</p><p>The dissociation constant of L-CAR for rhGAA was measured by DSF according to Vivoli et al.45 (Figure 2(d)). L-CAR showed a KD similar to that of the allosteric chaperone NAC (9.16 ± 1.02 and 11.57 ± 0.74 mM, respectively)43. As expected for molecules that do not bind to the rhGAA active site, these values are higher than the typical Ki of 3.4 µM exhibited by active-site directed molecular chaperones, such as the DNJ inhibitor41.</p><!><p>Similar stabilising effects were also observed with the related compounds D-CAR and A-D-CAR (Figure 1, 7 and 8, respectively). Both compounds rescued the activity of rhGAA on 4NP-Glc after 5 h of incubation at pH 7.4 (Figure S2(a)). Again, no effect on the specific activity of rhGAA at 0.1–10 mM concentrations was observed (Figure S2(b)), indicating that D-CAR and A-D-CAR also did not interact with the active site of the enzyme. Compared to the L-isomer (compare 6 and 7 in Figure 1), D-CAR showed a complete rescue of rhGAA activity already at 10 mM concentration vs. 20 mM of L-CAR (Figure S2(c)), maintaining the stabilising effect even after 24 h of incubation (Figure S1(a)). DSF analysis showed that D-CAR increased the structural stability of rhGAA in a dose-dependent manner (Figure S2(d)) and that the ΔTm increased by about 2 °C at every 2 mM increment of D-CAR concentration (Figure S2(e)).</p><p>To test if carnitine L- and D-enantiomers had additive effects, we analysed rhGAA stability in the presence of equimolar amounts of D- and L-CAR. As shown in Figure 3, when rhGAA was incubated with 10 mM total concentration of the two enantiomers (resulting from L-CAR 5 mM + D-CAR 5 mM), the ΔTms of 9.4 ± 0.8 °C corresponds to the sum of the ΔTms measured when the enzyme was incubated with either L- or D-CAR at 5 mM concentration (ΔTms of 4.3 ± 0.2 and 4.9 ± 0.1 °C, respectively). A similar additive effect was observed when the concentration of each enantiomer was increased to 10 mM of D- and L-CAR (Figure 3(b)).</p><!><p>Effect of a racemic mixture of D/L-CAR on the structural stability of rhGAA. (a) DSF analysis. L-CAR and D-CAR were incubated with rhGAA either alone (5 and 10 mM) or in combination (at 5 or 10 mM each). (b) Summary of the Tm measured by DSF.</p><!><p>The combined effect on rhGAA by L-CAR and other active-site directed or allosteric chaperones is shown in Figure 4. At a concentration of 10 mM, L-CAR increased the Tm of rhGAA by 9.0 ± 0.3 (Tm 58.6 ± 0.2 vs. 49.6 ± 0.1 °C of rhGAA alone) a value similar to that obtained with NAC at the same concentration (9.6 ± 0.2 °C), but slightly lower than that of the active-site directed pharmacological chaperone DNJ (Figure 1: 1) at a concentration of 0.1 mM (+12.1 ± 0.3 °C) (Figures 4(a,b)). To understand the mechanism of stabilisation towards rhGAA we combined these molecules in DSF experiments. L-CAR was mixed at 10 mM concentration in equimolar ratios with NAC (Figure 4(a)) or with 0.1 mM DNJ (Figure 4(b)). The stabilising effect of L-CAR in the presence of 10 mM equimolar amounts of NAC (20 mM total) was identical to the effect observed when each of the allosteric PCs was used individually at 20 mM concentration (Figure 4(a)). However, when L-CAR and NAC were combined, the ΔTm of 14.4 ± 0.2 °C was almost identical to those observed when L-CAR and NAC were used singularly at 20 mM concentration each (14.3 ± 0.2 and 14.3 ± 0.1 °C, respectively) (Figure 4(a)).</p><!><p>Comparison of the effect of allosteric and non-allosteric chaperones on the stability of rhGAA. (a) Analysis of the synergistic effect of L-CAR and NAC. rhGAA was incubated with L-CAR either alone (10 or 20 mM) or in combination with NAC, at 10 mM each. (b) Analysis of the synergistic effect of L-CAR and DNJ. L-CAR was incubated with rhGAA either alone (10 or 20 mM) or in combination with DNJ (0.1 mM).</p><!><p>The ΔTms obtained with either 10 mM L-CAR combined with 0.1 mM of the active-site directed allosteric chaperone DNJ, were exactly additive with ΔTm of +9.0 ± 0.3, +12.1 ± 0.3, and +23.2 ± 0.2 °C with L-CAR, DNJ, and L-CAR + DNJ, respectively, confirming that these PCs interact with different sites of rhGAA (Figure 4(b)).</p><!><p>We studied the effect of L-CAR on mutant GAA activity in cultured fibroblasts from three PD patients carrying different mutations and with early-onset phenotypes (see Table 1). Fibroblasts were incubated in the presence of 0.1 to 10 mM L-CAR for 24 h and the GAA activity was compared to that obtained in untreated cells. The chaperone had negligible and non-significant effects on endogenous residual activity in the cells from patients 1 and 2, while significantly enhancing effects were seen in cells from patient 3, homozygous for the p.L552P mutation, that had been already reported to be responsive to the active site-directed chaperones DNJ and NB-DNJ (Figure 5)52. Significant increments in activity were observed in a range of L-CAR concentrations between 1 and 10 mM, with a 2.8-fold increase at 2 mM.</p><!><p>Effect of L-CAR in PD fibroblasts. (a) Effect of L-CAR on the residual activity of mutated GAA in fibroblasts. Fibroblasts derived from three PD patients were incubated in the presence and in the absence of 0.1–10 mM L-CAR before being harvested and used for GAA assay. The untreated cells (UT) were used as a control. The chaperone has significant effects on endogenous residual activity in the cells from patient 3.</p><p>Mutants used in this study.</p><!><p>It has been shown previously that active site-directed chaperones enhance the activity of recombinant enzymes used for ERT in PD and Fabry disease31,32, with a synergistic effect. In PD fibroblasts the iminosugar NB-DNJ enhanced rhGAA efficacy by ∼1.3- to 2-fold. An enhancing effect on correction of GAA deficiency by rhGAA and a better enzyme processing was also demonstrated with the allosteric chaperone NAC41.</p><p>We tested whether the allosteric PC L-CAR also shows a similar effect in combination with ERT in the three cell lines indicated above. We first studied the optimal conditions to evaluate this effect. We compared a protocol based on pre-incubation of cells with L-CAR for 24 h, followed by co-incubation of L-CAR and rhGAA for an additional 24 h, with a protocol based on co-incubation of L-CAR and rhGAA for 24 h (Figure 6(a)). The results of both protocols were compared with those obtained in cells treated with rhGAA alone. The second treatment protocol gave the best results and was selected to evaluate the optimal L-CAR concentration for rhGAA enhancement.</p><!><p>Synergy between L-CAR and rhGAA in PD fibroblasts. (a) Setting the conditions for evaluation of synergy between L-CAR and rhGAA. Different treatment protocols were evaluated: (i) pre-incubation of cells with L-CAR for 24 h, followed by co-incubation of L-CAR and rhGAA for an additional 24 h; (ii) co-incubation of L-CAR and rhGAA for 24 h. (b) Setting the optimal L-CAR concentrations for evaluation of synergy between L-CAR and rhGAA. Fibroblasts were incubated with rhGAA and different L-CAR concentrations (1–20 mM). GAA activity enhancements were observed at 5, 10, and 20 mM L-CAR concentrations with the highest and statistically most significant enhancements at 10 and 20 mM. (c) Effect of L-CAR on rhGAA processing in PD fibroblasts. Cells were incubated for 24 h with rhGAA alone or with rhGAA in combination with 10 mM L-CAR. In the cells treated with the combination of rhGAA and L-CAR the amount of the 70–76 kDa mature GAA active peptides were dramatically improved, as indicated by quantitative analysis by western blot. Glyceraldehyde 3-phosphate dehydrogenase (GAPDH) is the loading control. (d) GAA activities measured in PD fibroblasts. The increase of GAA activity confirms the enhancing effect of L-CAR.</p><!><p>With the co-dosing of rhGAA and L-CAR (1–20 mM) GAA activity enhancements were observed at 5, 10, and 20 mM L-CAR concentrations (Figure 6(b)). The highest and statistically most significant enhancements were obtained at 10 and 20 mM. Higher L-CAR concentrations (up to 50 mM) were toxic for fibroblasts (not shown). Thus, we selected the concentration of 10 mM for further experiments, as this concentration appeared to combine efficacy and safety for cells.</p><p>We next studied the effect of L-CAR on rhGAA processing in PD1 and PD2 fibroblasts. For enzyme replacement therapy rhGAA is provided by the manufacturer as a 110 kDa precursor. Once internalised by cells through the mannose-6-phosphate receptor and the endocytic pathways, the enzyme is converted into an intermediate of 95 kDa and the active molecular proteoforms of 76 and 70 kDa. Cells were incubated for 24 h with rhGAA alone or with rhGAA in combination with 10 mM L-CAR. In the cells treated with the combination of rhGAA and L-CAR the amount of the 70–76 kDa mature GAA active peptides was dramatically improved (Figure 6(c)). The corresponding GAA activities measured in PD1 and PD2 cells (Figure 6(d)) confirmed the enhancing effect of L-CAR and were in line with those observed in previous experiments.</p><p>We also studied the kinetics of GAA enhancements at different time points in PD fibroblasts treated with rhGAA alone or in combination with 10 mM L-CAR. GAA activity increased progressively over time and an enhancing effect of co-incubation with L-CAR was already detectable at 2 h and became progressively more pronounced up to 24 h (Figure 7(a)). The amounts and the processing of rhGAA, analysed by western blot, also improved over time (Figure 7(b)).</p><!><p>Kinetics of GAA enhancements at different time-points in PD fibroblasts treated with rhGAA alone or in combination with 10 mM L-CAR. (a) GAA activity increased progressively over time and an enhancing effect of co-incubation with L-CAR was already detectable at 2 h and became progressively more pronounced up to 24 h (a). The amounts and the processing of rhGAA, analysed by western blot, also improved over time (b).</p><!><p>We next looked at the effects of rhGAA and L-CAR co-dosing on the lysosomal trafficking of the recombinant enzyme. The cells were incubated under the conditions selected in the previous experiments, and co-localization of rhGAA with the lysosomal associated membrane protein 2 (Lamp2), a common lysosomal tag, was analysed by confocal immune-fluorescence microscopy. In all three cells lines the co-localization was improved (Figure 8(a)). This result was confirmed by a quantitative analysis of the total GAA signal (Figure 8(b)) and of the GAA signal co-localized with Lamp2 (Figure 8(c)) performed by ImageJ Software.</p><!><p>Effects of rhGAA and L-CAR co-dosing on lysosomal trafficking of the recombinant enzyme. The cells of three Pompe patients were incubated under the conditions selected in the previous experiments, and co-localization of rhGAA with Lamp2 was analysed by confocal immune-fluorescence microscopy. In all three cells lines the co-localization was improved (a). This result was confirmed by a quantitative analysis of total GAA signal (b) and of GAA signal co-localized with Lamp2 (c).</p><!><p>The problems connected to the inhibitory effect of active-site-directed PCs currently used in clinics for LSD can be addressed by the identification of novel allosteric chaperones, that, not binding to the active site of the enzyme, are non-inhibitory and can be potentially more effective than active-site directed PCs. Several studies demonstrated that this can be a convenient avenue for the treatment of Gaucher53,54 and Pompe diseases41,55. Another limitation of inhibitors acting as PCs is that they are effective in rescuing only some disease-causing missense mutations, mainly located in the catalytic environment of enzyme scaffolds, and are thus potentially effective only in a limited number of patients. For PD, it was proposed that about only 10–15% of patients may be amenable to PCT with the iminosugar DNJ56.</p><p>Based on our previous experience with NAC, and related compounds NAS and NAG41, we embarked on the identification of novel allosteric pharmacological chaperons for PD. In the study presented here, we show that L-CAR and the related compounds D-CAR and A-D-CAR can stabilise GAA without interfering with its activity. In cell-free assays, these PCs prevented the loss of GAA activity at pH 7.0 and increased the enzyme thermal stability in a concentration-dependent manner like previously shown with NAC41,43. In addition, the combination of L-CAR and the active site-directed chaperone DNJ showed clearly an additive effect (Figure 4(b)) confirming that the two molecules bind to different sites of the enzyme.</p><p>The crucial experiment demonstrating the efficacy of L-CAR on PD was the correction of the enzyme defect in patient's fibroblasts to a greater extent than that observed with NB-DNJ. When the recombinant enzyme was administered to the patient's fibroblasts in combination with L-CAR, the lysosomal trafficking, the maturation, and the intracellular activity of the enzymes increased up to 4-fold when compared to the combination ERT/NB-DNJ treatment (Figures 6–8)31.</p><p>The ability of L-CAR and its derivatives to bind to rhGAA is rather surprising and previously unpredictable since L-CAR is structurally different from the DNJ and NAC (Figure 1). The main difference between L-CAR/NAC on one hand and DNJ/NB-DNJ on the other on stabilising rhGAA is the 2-fold higher concentration used for the formers to observe similar ΔTms. The µM affinity of active-site directed iminosugars, as deduced from Ki values, can be explained by the structural similarity with GAA natural substrates. On the other hand, the mM values of the KD of the allosteric pharmacological chaperones indicate lower affinity but a specificity higher than those of chemical chaperones or compatible solutes (e.g. osmolytes, sugars, amino acids, etc.), working at molar concentrations57,58.</p><p>We demonstrated by the inspection of the high-resolution 3 D-structure of the rhGAA/NAC complex, that NAC could bind at the interface between the catalytic and auxiliary domains, thereby explaining its chaperoning activity by enhancing the structural stability of the overall enzyme's scaffold and by preventing deleterious oxidation of Cys93843. The binding mode of L-CAR is currently not known. The sigmoidal saturation curve for rhGAA when incubated with the L-CAR indicated cooperativity in the binding mode of this allosteric chaperone (Figure 2(d)). In addition, the non-additive stabilising effect of L-CAR when used with NAC (Figure 4) and their similar KD, suggesting that these molecules could bind to rhGAA by a similar mechanism. The carboxylate group of NAC makes water-mediated contacts in two different sites of rhGAA43. Possibly, also carnitine derivatives might form weak interactions with the enzyme, through their carboxylates groups like the N-acetylated amino acids (3–8 in Figure 1). Structural data that would be needed to understand the mechanism of action of D- and L-CAR are complicated by the weak binding of these molecules to rhGAA. We endeavoured manifold trials to obtain crystal structures of rhGAA in complex with both L-CAR or D-CAR, employing respectively crystal-soaking and co-crystallisation techniques, but unfortunately, our efforts were not coronated by success. The reasons of these failures might be attributed either to the weak binding of the chaperones to rhGAA (although massive doses had been used) or to the fact that the genuine binding sites were obstructed by molecular packing arrangements within the crystal lattice. In this context it is noteworthy mentioning that crystallisation conditions for rhGAA are extremely stringent, with all the rhGAA structures reported in the Protein Data Bank having been obtained (by us or by others) in exactly the same conditions, explaining why experimental settings are not favourable to unveil binding-sites hidden by crystal-lattice contacts.</p><p>The presence in proteins of weak binding sites for small molecules has been predicted and several experimental and in silico studies showed "hotspots" on protein surfaces that can bind weakly to small molecules, even at low M range, expanding potential druggable sites59–62. Thus, further studies are needed to identify carnitine binding sites on rhGAA, however, our study suggests that other molecules, whose chaperoning activity cannot be simply inferred from their molecular structure, may be effective as PCs for LSDs, thereby opening new and wider opportunities for the identification of novel therapeutic drugs.</p><p>The use of L-carnitine as a drug for the treatment of PD is particularly attractive. L-CAR is involved in fatty acid metabolism and synthesised mainly in the liver and kidneys from the essential amino acids lysine and methionine as ultimate precursors to form trimethyl lysine. L-CAR is not toxic at the concentration normally administrated and its use is approved as nutraceutical. Instead, the use of D-carnitine and acetyl-D-carnitine in clinics is less reliable. D-Carnitine can interfere with the uptake and transport of L-carnitine by inhibiting the carnitine acetyltransferase and its use in patients affected by kidney illnesses is avoided49. In addition, documented clinical use of acetyl-D-carnitine and its pharmacologically acceptable salts is limited to the therapeutic treatment of glaucoma63. Therefore, L-CAR might be promptly included in clinical protocols for the treatment of PD while its D-CAR and A-D-CAR derivatives need more investigations.</p><p>The enzyme/PCs molar ratios used to obtain the stabilisation of rhGAA described here, ranged from 1:102 to 1:104 for DNJ and L-CAR/NAC, respectively. This indicates that even the more specific inhibitor DNJ is used at saturating conditions and that L-CAR, showing a lower affinity for the allosteric binding sites on rhGAA, required relatively high concentrations to promote a stabilising effect. However, the toxicity of L-CAR is reported to be low even at doses higher than those used in our study. In fact, L-CAR, with doses of 3 g daily as an oral supplement, is used to treat patients affected by congestive heart failure, end-stage renal disease, hyperthyroidism, male infertility, myocarditis, polycystic ovary syndrome, and toxic side effects caused by the drug valproic acid. Instead, an intravenous infusion of 60 mg/kg of L-CAR is used for patients suffering from angina pectoris64.</p><p>The synergy between L-Car and ERT demonstrated here may be translated into improved clinical efficacy of ERT, as proposed for other PCs in Gaucher, Pompe, and Fabry diseases20,21,31,32. It is worth noting that, while the activity enhancement of endogenous defective enzymes by chaperones in most cases resulted in minor changes in terms of residual activity, likely leading to a modest impact on patients' outcomes, the synergy between ERT and L-CAR based PCT has the potential to determine remarkable increases of specific activity, independently of mutations affecting individual patients.</p><!><p>Click here for additional data file.</p>

PubMed Open Access

Seroprevalence of SARS-CoV-2 infection among children in Children’s Hospital Zagreb during the initial and second wave of COVID-19 pandemic in Croatia

IntroductionThe study aimed to investigate the prevalence and titres of anti-SARS-CoV-2 antibodies in children treated at the Children’s Hospital Zagreb in the first and the second wave of the COVID-19 pandemic. Statistical significance of difference at two time points was done to determine how restrictive epidemiological measures and exposure of children to COVID-19 infection affect this prevalence in different age groups.Materials and methodsAt the first time point (13th to 29th May 2020), 240 samples and in second time point (24th October to 23rd November 2020), 308 serum samples were tested for anti-SARS-CoV-2 antibodies by enzyme-linked immunosorbent assay (ELISA) and electrochemiluminescence immunoassay (ECLIA). Confirmation of results and titre determination was done using virus micro-neutralization test. Subjects were divided according to gender, age and epidemiological history.ResultsSeroprevalence of anti-SARS-CoV-2 antibodies differs significantly in two time points (P = 0.010). In first time point 2.9% of seropositive children were determined and in second time point 8.4%. Statistically significant difference (P = 0.007) of seroprevalence between two time points was found only in a group of children aged 11-19 years. At the first time point, all seropositive children were asymptomatic with titre < 8. At the second time point, 69.2% seropositive children were asymptomatic with titre ≥ 8.ConclusionsThe prevalence of anti-SARS-CoV-2 antibodies was significantly lower at the first time point than at the second time point. Values of virus micro-neutralization test showed that low titre in asymptomatic children was not protective at the first time point but in second time point all seropositive children had protective titre of anti-SARS-CoV-2 antibodies.

seroprevalence_of_sars-cov-2_infection_among_children_in_children’s_hospital_zagreb_during_the_initi

Introduction<!>Study design and subjects<!>Blood sampling<!>Methods<!>ELISA method<!>ECLIA method<!>Virus micro-neutralization test<!>Statistical analysis<!>Results<!>Discussion

<p>In December 2019, a novel coronavirus emerged in Wuhan, China (1). The virus was named severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) and disease COVID-19. Due to a rapid spread and possibility of causing severe and life-threatening infections, it has attracted worldwide attention. On 11th March 2020, the World Health Organization (WHO) declared a pandemic of the COVID-19 (2). On the same date, the Croatian Minister of Health declared an epidemic of COVID-19. In Croatia, as in many other countries in the world, restrictive epidemiological measures were introduced (initial lockdown) to prevent the spreading of the disease and to get the time to reorganize the entire healthcare system accordingly (3). With the gradual relaxing of restrictive measures in May and during the summer, the number of new COVID-19 cases gradually increased daily, but in September and October, this number dramatically and exponentially increased in many countries, especially in Europe. It was difficult to talk about the new (or the second) wave of epidemic since the virus had not disappeared at any moment and practically the first wave has not "finished". However, many states introduced new different restrictive measures in September and October 2020 due to the "new" wave of COVID-19. In Croatia, despite a significant daily increase in the number of new cases, new restrictive measures were introduced in late November (4).</p><p>Incidence data of COVID-19 in the paediatric population are available from epidemiological reports of countries with the highest number of cases. Reports show a small proportion of children (0-19 years) in the total number of patients. In China, children under 18 amounted to 2-5%, Italy 1.2%, United States 7.3%, and in Australia 4% of all COVID-19 positive cases (5-9). The first multinational and multicentre study on children with COVID-19 in Europe during the initial peak of the pandemic, which was conducted in 82 tertiary and quaternary paediatric units in 25 European countries, showed that COVID-19 is generally a mild disease in children, including infants, and a proportion of 8% of those COVID-19 positive children developed a severe illness that required intensive care support and prolonged mechanical ventilation. Several predisposing factors for intensive care support have been identified, and it is confirmed that death is rare in children (10).</p><p>All these data suggest that children show clinical symptoms less often than adults, and they also have a milder illness, recover faster and have a better prognosis. The role of asymptomatic or subclinical infection in human-to-human transmission of the virus is not fully understood.</p><p>Due to mild or asymptomatic infections, children are not included in the routine molecular testing (reverse transcriptase polymerase chain reaction; RT-PCR) for COVID-19 and according to existing data, it is impossible to accurately determine the number of infected children (11). Generally, mildly affected or asymptomatic persons are not routinely tested and included in the COVID-19 reports and the number of infections is probably underestimated. In this context, seroprevalence studies are important in the assessment of the extent of infection in the population. As the WHO recommended, monitoring changes in the seroprevalence over time is also crucial to predict dynamics and plan adequate public health measures (12).</p><p>Testing for specific antibodies has far greater potential than molecular testing to detect a past asymptomatic patients or patients with mild symptoms of infection. Antibodies most commonly become detectable 1-3 weeks after symptom onset, at which time evidence suggests that infectiousness likely is greatly decreased and that some degree of immunity has developed (13). Existing commercial assays generally detect SARS-CoV-2 immunoglobulin (Ig)A, IgM and IgG class antibodies separately or total antibodies specific for the nucleocapsid or spike protein of the virus (11).</p><p>Children's Hospital Zagreb (CHZ) is a general hospital for children aged 0 to 19 years whose primary role is not the treatment of COVID-19 and the aim of this study was to investigate the prevalence and titres of anti-SARS-CoV-2 antibodies in children treated at the CHZ in the first and the second wave of the COVID-19 pandemic. A statistical significance of the difference of anti-SARS-CoV-2 prevalence at two time points was done to determine how restrictive epidemiological measures and exposure of children to COVID-19 infection affect this prevalence in children of different age groups. The results of this study represent a contribution to the assessment of the extent of COVID-19 infection in the population, but also help in monitoring seroprevalence changes to predict dynamics and planning appropriate public health measures in Croatian pandemic conditions.</p><!><p>All blood samples of children which arrived at the Department of Laboratory Diagnostics of CHZ that met the residual criterion volume (> 350 µL) after the routine laboratory analysis were included in the study at two time points of COVID-19 pandemic. The first time point was from 13th to 29th May 2020 (just after the lockdown, number of COVID-19 positive patients was 0 to 8 per day at the national level), and in the second time point from 24th October to 23rd November 2020 (the peak of the second wave of the COVID-19 pandemic, until new restrictive measures were introduced, the number of COVID-19 positive patients was 1165 - 3573 per day at the national level).</p><p>Samples collected during the first and second time points came from independent groups of subjects. During the first time point, the number of subjects was defined according to the protocol of multi-center study of anti-SARS-CoV-2 antibodies seroprevalence in the general population where 240 samples were collected in CHZ and tested in Croatian Institute of Public Health (CIPH). At second time point, the CHZ proposed, collected, and determined the samples collected during one month at the peak of the second wave (N = 308).</p><p>All subjects were divided according to gender and age into three groups: (i) < 1 year; (ii) 1-10 years and (iii) 11-19 years. The reason for this age-based classification is the exposure of children to the SARS-CoV-2 virus, which varies according to the age of the child. The assumption is that the newborns and infants are mostly at home with their families, while older children are mostly exposed to the virus in kindergarten and school. In addition to the school exposure, teenagers are exposed due to a more intense social life.</p><p>According to the epidemiological history, the subjects were classified as (i) negative, (ii) potentially positive and (iii) positive for COVID-19. Table 1 shows the criteria for distribution to groups of children according to epidemiological history.</p><p>Samples were taken in the laboratory facilities (outpatients), emergency department (children with acute disease), isolation department (children to be admitted to hospital, prior to the results of the reverse transcription polymerase chain reaction (RT-PCR) test), other hospital department (hospitalized patients), and daily hospital or specialized pediatric clinics (follow-up examinations without active disease). According to the medical history, diagnoses of subjects included: ulcerative colitis, irregular menstruation, obstructive defects of the renal calyx and urethra, conditions following recovery from infectious mononucleosis, bone fractures, cystic fibrosis, appendicitis, fever of unknown origin, Crohn's disease, dermatitis, epilepsy, anaemia, pyelonephritis, asthma as well as children with solid tumors and leukemia. Symptoms at the time of ant-SARS-CoV-2 antibodies determination are grouped and listed in Table 2. Potentially positive children and children suspected for SARS-CoV-2 infection were not considered as an exclusion criteria in order for the level of infection in the general population would not be underestimated.</p><p>Informed consent was obtained from parents of all children included in the study. CIPH and CHZ Ethics Committee approved this study.</p><!><p>Patient's venous blood was collected in a tube with clot activator and gel separator (Vacuette, Greiner Bio-One GmbH, Austria). Centrifugation of the blood samples was applied according to the manufacturer's recommendation (2000xg, 10 minutes), the samples were forwarded for analysis and immediately after analysis (maximum 4 hours after centrifugation) the remaining serum was separated into labelled plastic tubes and stored at - 20 °C until analysis (maximum 3 weeks).</p><p>Blood sampling for outpatients were performed in fasting state in the morning from 8 to 10 am, but for day hospital and hospital departments, patient's blood samples were collected throughout the day in fasting and non-fasting state. Therefore, all lipaemic samples were not included in this study. In addition, haemolysed and icteric samples were also excluded from the study.</p><!><p>Serological testing during the first and second waves of the COVID-19 pandemic was done by different methods: enzyme-linked immunosorbent assay (ELISA) in the first wave at CIPH and electro-chemiluminescence immunoassay (ECLIA) in the second wave at CHZ. Both methods were verified according to the recommended protocol of American Society for Microbiology Clinical and Public Health Microbiology (14). The methods showed good comparability and the results of both methods were confirmed using virus micro-neutralization test (mVNT). The results of mVNT were defined as final.</p><!><p>A commercial ELISA (COVID-19 ELISA IgA+IgM; IgG, Vircell, Granada, Spain) was used for detection of anti-SARS-CoV-2 antibodies in human serum or plasma samples at CIPH. The assay is based on reaction to recombinant spike glycoprotein (S) and nucleocapside protein (N) antigens of SARS-CoV-2. The results were expressed as antibody index [AI = (sample optival density (OD)/ cut off serum mean OD)] x 10 and interpreted as follows: IgG < 4 negative, 4-6 borderline, > 6 positive; IgM/IgA < 6 negative, 6-8 borderline, > 8 positive (15). Positive, negative and controls at borderline level have been run with each test run. Initial test validation has been performed on 30 serum samples collected from patients with RT-PCR confirmed COVID-19 (N = 15) 4-34 days after disease onset and asymptomatic persons (N = 15) with negative RT-PCR and negative mVNT.</p><!><p>Roche Cobas Elecsys Anti-SARS-CoV-2 is an ECLIA test for qualitative detection of IgG and IgM class of antibodies (multiple analytes reported as a single result) developed against anti-SARS-CoV-2 antibodies in human plasma or serum samples on Cobas e-411 immunoassay analysers (Roche Diagnostics GmbH, Mannheim, Germany) at CHZ. The assay is based on a recombinant protein which represents the nucleocapsid (N) antigen of SARS-CoV-2.</p><p>The result of sample testing is given either as reactive or non-reactive as well as in the form of a cut-off index (COI; signal sample/cut-off). Samples results with COI < 1.0 is negative for anti-SARS-CoV-2 antibodies and COI ≥ 1.0 is positive for anti-SARS-CoV-2 antibodies. Verification of the ECLIA method was done according to the American Society for Microbiology recommendations and protocol (14).</p><p>The method verification procedure as well as the testing of the collected samples used the original commercial reagent with the original respective calibrators. Control samples for determining precision and accuracy and as an internal control in each series of sample run used PreciControl Anti SARS-CoV-2 control (Roche Diagnostics GmbH, Mannheim, Germany) in two levels (Level 1: range 0.00 - 0.80 COI, mean: 0.40 COI and Level 2: range 1.73 - 4.64 COI and mean 3.20 COI).</p><p>Verification ECLIA method in CHZ was done at the first time point and parallel serum samples were separated (N = 50) and stored for comparability of results by ECLIA method testing (CHZ) with results by ELISA method testing (CIPH). Verification showed that the method is comparable to ELISA and gives acceptable identical results (35/36 same results, data not published).</p><!><p>All reactive samples (by ELISA and/or ECLIA method) were confirmed using a virus mVNT in cell culture in biosafety level 3 (BSL-3) laboratory at CIPH. The SARS-CoV-2 HR1/8933 strain isolated from the nasopharyngeal swab of COVID-19 patient on Vero E6 cells was used for the mVNT. Maximum cytopathic effect was visible on the 4th day and the virus replication was confirmed by RT-PCR. Heat inactivated serum samples (56 °C/30 min) were tested in duplicate in 96-well plates. An equal volume (25 µL) of two-fold serum dilutions (starting from 1:2) was mixed with the equal volume (25 µL) containing 100 median tissue culture infectious doses (TCID50) of the virus. After 1 h incubation at 37 °C in CO2 incubator, 50 µL of Vero E6 cells in a concentration of 2 x 105 cells/mL were added to each well and incubated for 4 days. The antibody titre was defined as the reciprocal value of the highest serum dilution that showed 100% neutralization in at least half of the infected wells. A titre of ≥ 8 was considered positive and protective and titre of 2-4 was considered positive, but unprotective (16).</p><!><p>Results were analysed using Microsoft Excel version 2013 (Microsoft, Redmond, USA) and presented as a total number and a percentage. The age of the children for each age group was expressed as the median and range. Gender was expressed as ratio of the number of girls to the number of boys. Significance of differences of results at two time points was done by chi-square test (for N > 100) and comparison of proportions (with 95% confidence interval) (for N < 100) using MedCalc Statistical Software version 19.5.2 (MedCalc Software Ltd, Ostend, Belgium) (17). A statistically significant difference was defined at P < 0.05.</p><!><p>The first, initial wave of the pandemic was marked by a total lockdown, and samples were collected when the first strict epidemiological measures gradually mitigated. At this first study point, after mVNT, the final number of COVID-19 seropositive children was 2.9% (Table 3). The second point of the study was marked as the peak of the second wave of the pandemic, and out of 308 children samples examined, 8.4% COVID-19 seropositive children were found. Statistically significant difference was found in the number of seropositive children during the first and the second wave of the COVID-19 pandemic (P = 0.010) (Table 3). About 90% of the examined children were defined as negative according to epidemiological history at both observed time points and a significant difference during the first and second wave of the pandemic was observed only in the group of children with a positive epidemiological history (P = 0.009) (Table 3).</p><p>The distribution of results by age/gender groups showed that a statistically significant difference between seropositive children during the first and second wave of the COVID-19 pandemic is found only for the 11-19 age group (P = 0.007) (Table 4). In this age group, at the two time points, proportions of boys and girls are significantly different. Boys during the first wave is significantly higher than in the second wave (P = 0.048) in contrast to girls where proportion of girls in the first wave is significantly less than in the second wave (P = 0.044) (Table 4). Results of the distribution by gender, age, and epidemiological history are presented in Table 4.</p><p>The share of seropositive results in group of children with negative epidemiological history during the first wave was 3.2%, and during the second wave of the pandemic it was 6.6%, with P = 0.107.</p><p>During the first wave of the pandemic, all seropositive children were asymptomatic at the time of anti-SARS-CoV-2 testing. At the second wave, 7/26 children with a positive COVID-19 RT-PCR test were found: 2 children with positive epidemiological history and positive RT-PCR (within 24 hours) and 7 children recovered from COVID-19 from 8 days to 2 months prior to testing and only 5 were seropositive.</p><p>During the second wave of the pandemic, out of a total of 26 seropositive children, 6 children were hospitalized for treatment of acute or chronic diseases, and 4 children were admitted to emergency hospital admission due to acute disease (2 children had a fever > 38 °C with other symptoms: nausea, vomiting/diarrhea, abdominal pain), and 2 children were afebrile (bone fracture, renal colic). Asymptomatic children who underwent triage before entering the day hospital for specialist examination/treatment (7 children) and before arriving at the laboratory for blood sampling (9 children) were seropositive for COVID-19. Only two seropositive children had fever > 38 °C with other symptoms (urological, gastrointestinal and/or other) and 6 children had combination of symptoms (urological, gastrointestinal and/or other) without fever and respiratory symptoms. In total, at the second wave of the pandemic, we found 18/26 seropositive asymptomatic children without the symptoms that could be associated with COVID-19 disease (Table 2).</p><!><p>The results of serological testing for COVID-19 at Children's Hospital Zagreb show a significantly lower seroprevalence in the paediatric population during the first, initial wave, compared to the second wave of the COVID-19 pandemic (2.9% vs. 8.4%, P = 0.010). This significant difference was expected due to the increase of incidence in the country and the second round of seroprevalence study was conducted in order to provide evidence that the increased transmission of infection affected children as well as adults.</p><p>The first published results of COVID-19 seroprevalence from the end of March to up April 8, 2020 in Wuhan City was 9.6% in the general population and first systematic review of COVID-19 seroprevalence in May 2020 reported a very wide range of seroprevalence: from 0.4 to 59.3% in the general population (18, 19).</p><p>There are few published data on the COVID-19 seroprevalence in children and mainly relate to the first wave (from April to May 2020) of the pandemic. Different results in these studies may be due to the type of the studies (the most of published data are case reports or case series), dynamics in new cases per day, the strength of the epidemiological measures applied in countries, and in the definition of age groups (some results are part of the general population report). However, all these data have the same conclusion that COVID-19 seroprevalence in children is low. The results of a study in Switzerland during April and May 2020 show a very low prevalence in children aged 5-9 years (0.8%) compared to children between 10 and 19 years (9.4%) (20). An extensive Spanish study of COVID-19 seroprevalence in the general population during April and May 2020 reported that children aged 0-19 were represented by 3.9% (21). US prevalence study of anti-SARS-CoV-2 antibodies in children (age range 0-18 years) without symptoms of COVID-19 disease who were tested at 28 hospitals showed the prevalence varied from 0% to 2.2%, with a pooled prevalence of 0.65% (95%CI: 0.47% to 0.83%) with significant heterogeneity and significantly associated with weekly incidence of COVID-19 in the general population (22). Also, the study performed at Seattle Children's Hospital during the lockdown in March and April 2020 found only 1% of COVID-19 seropositive children aged 0 to 15 years (23). Multicentre observational cohort study, conducted between April to July 2020 at 5 UK sites, recruited children of healthcare workers, aged 2-16 years publish that total COVID-19 seroprevalence is 6.9% (95% CI 5.4% to 8.6%, N = 992) and varied between sites. Belfast had significantly lower seroprevalence than all other sites at 0.9% (95% CI 0.2% to 3.3%, N = 215 and P < 0.001), and in London seroprevalence was significantly higher than all other sites at 11.6% (95% CI 7.8% to 16.8%, N = 199 and P = 0.007) (9). In the Czech Republic, in April 2020, the overall SARS-CoV-2 seroprevalence was estimated not to exceed 1.3%. In July and August, 2020, 200 children (0 to 18 years of age) from paediatric department of a large hospital in Prague were screened for the presence of anti-SARS-CoV-2 antibodies and zero seropositive subjects were found. Therefore, this study reported a low (< 0.5%) cumulative seroprevalence amongst children in Prague during August, 2020 (24).</p><p>According to the initial reports (up to 29th May 2020) of the CIPH, there were 2.8% children under 10 and 4.0% children and young adults from 11 to 20 years among all COVID-19 patients in Croatia. Until November 2020, the cumulative number of children up to age 10 was 2.4%, and children and young adults ages 11 to 20 years 9.3% (25, 26).</p><p>In our study, the initial, first wave, was marked by strict epidemiological measures (home-working, online schools, closed kindergartens, hospital admissions restricted). With the gradual mitigation of these strict measures in late April and early May 2020, serological testing began. The number of new RT-PCR positive cases at the national level in that period was 0 to 8. According to the literature, children are significantly less likely to suffer from COVID-19 infection, affected children show mild symptoms or are asymptomatic and children's exposure to COVID-19 virus in the initial period was very low (10-12). Our results also show a low seroprevalence (2.9%) for children of age 0 to 19 in the first time point and all of them were asymptomatic.</p><p>The beginning of autumn was marked by a "new normal": the opening of schools, kindergartens, health institutions, sports, social and economic institutions with certain epidemiological measures (social distance, wearing masks indoors, hand disinfection), and by the end of October the steeply upward curve of new infection reached > 3500 new cases per day and children's exposure to COVID-19 infection was significantly higher. At this time point, except negative and potentially positive children according to epidemiological history, we found RT-PCR COVID-19 positive cases, as well as children recovered from COVID-19 infection. This time point being a more realistic state for determining seroprevalence in children than the initial first wave. Seroprevalence in children at the second time point was 8.4%, and 70% of them were asymptomatic.</p><p>When we distributed our subjects according to age groups, taking into account variation of epidemiological measures applied (e.g. infants up to 1 year exposed to COVID-19 infection almost exclusively in contact with mother and household, children staying in kindergarten or attending lower grades at primary schools with a minimum distance and mask-wearing, teenage groups attending school mostly online, and outside of-home-contacts with mandatory mask-wearing indoors by public health and social institutions and ensuring social distance), we noted that a significant difference in seroprevalence is shown exclusively in the age group of teenagers. These results refer that regardless of epidemiological measures and exposure to the COVID-19 virus, the seroprevalence of young children is low. The children are mostly asymptomatic or present a mild form of the disease which is consistent with results of other studies (11, 27).</p><p>The literature cited several possible reasons related to the protective effect from severe clinical forms of COVID-19 and lethal outcomes in children: virological and epidemiological characteristics in children, the immune system and the maturation and low exposure to ACE2 receptors, characteristics of the renin angiotensin system in childhood, as well as the shorter effect of air pollutants on the respiratory system compared to the adult population (11, 27, 28). The contribution to the assumption of protective mechanisms in children is the case of one 2 months old child, the subject of this study who was considered positive according to the epidemiological history due to a COVID-19 RT-PCR positive mother at birth, but found negative for both COVID - 19 (RT-PCR) as well as anti-SARS-CoV-2 antibodies (IgM, IgG).</p><p>This study has some limitations. Although Children's Hospital Zagreb is a specific children's health institution in the Republic of Croatia where children from all parts of the country are treated, the obtained results of COVID-19 seroprevalence in children cannot be considered representataive at the national level.</p><p>Our results do not provide an answer on the transmission of COVID-19 disease from children to adults which should be investigated separately when planning epidemiological measures during a pandemic.</p><p>In conclusion, the prevalence of anti-SARS-CoV-2 antibodies was significantly lower at the first time point than at the second time point. Virus micro-neutralization test values showed that low titre in asymptomatic children was not protective at the first time point, but all seropositive children in the second wave had a protective titre of anti-SARS-CoV-2 antibodies. According to the results, the determination of anti-SARS-CoV-2 antibodies could be useful for children older than 10 years both in terms of their own protection and in terms of COVID-19 transmission.</p><p>Determination of mVNT in epidemiological circumstances of COVID-19 pandemic plays an important role in the verification of qualitative methods as well as their quantification by providing important epidemiological data on the protection of individuals and population. Also, for commercial quantitative tests, it will be helpful in order to define activity of anti-SARS-CoV-2 antibodies antibody as the limit of protection.</p>

PubMed Open Access

LRRK2 in peripheral and central nervous system innate immunity: its link to Parkinson's disease

Mutations in the leucine-rich repeat kinase 2 (LRRK2) gene are found in familial and idiopathic cases of Parkinson's disease (PD), but are also associated with immune-related disorders, notably Crohn's disease and leprosy. Although the physiological function of LRRK2 protein remains largely elusive, increasing evidence suggests that it plays a role in innate immunity, a process that also has been implicated in neurodegenerative diseases, including PD. Innate immunity involves macrophages and microglia, in which endogenous LRRK2 expression is precisely regulated and expression is strongly up-regulated upon cell activation. This brief report discusses the current understanding of the involvement of LRRK2 in innate immunity particularly in relation to PD, critically examining its role in myeloid cells, particularly macrophages and microglia.

lrrk2_in_peripheral_and_central_nervous_system_innate_immunity:_its_link_to_parkinson's_disease

Parkinson's disease<!>Systemic and CNS inflammation in PD<!>LRRK2 and microglia-driven neuroinflammation in PD.<!>Schematic of macrophage and microglia differentiation from HiPSCs.<!>Leucine-rich repeat kinase 2<!>LRRK2 expression is precisely regulated in myeloid cells<!>Roles of LRRK2 in macrophages and microglia<!>Cytokine release<!>Migration<!>Phagocytosis<!>Human-induced pluripotent stem cell macrophages and microglia as tools to study LRRK2 biology<!>Concluding remarks<!><!>Funding<!>Competing Interests

<p>Parkinson's disease (PD) is a complex, multifactorial neurodegenerative disease. In North America, it affects 1.5% of the population over the age of 65. Patients gradually develop motor impairments, caused by a slow and progressive degeneration of dopaminergic neurones in the substantia nigra pars compacta (SNpc). The aetiology of PD is largely unknown, involving a complex interaction between various genetic and environmental factors. To date, 17 distinctive chromosomal locations, named parkin (PARK) 1–18, have been identified in association with inherited PD. Although only ∼10% of PD cases are identified as familial PD, genome-wide association studies (GWAS) have also detected a role for genetic variants in idiopathic PD (reviewed in ref. [1]). Understanding the roles of PD-associated genes, therefore, has increasing significance as this would provide valuable insights into shared pathological mechanisms underlying both inherited and idiopathic PD pathogenesis.</p><p>Interestingly, key PD-associated genes, α-synuclein (SNCA), PARK2, deglycase (DJ-1), leucine-rich repeat kinase 2 (LRRK2), and glucocerebrosidase (GBA), are all expressed in immune cells, implying their potential role in immunity (reviewed in ref. [2]). Neuronal injuries commonly elicit activation of innate immune responses in the central nervous system (CNS), and inflammation-driven neurotoxicity has been suggested to play a central role in progression of various neurodegenerative diseases, including PD (reviewed in ref. [3]). PD-associated genes may have distinct cellular functions in immune cells, and it can also be hypothesized that mutations in these genes commonly contribute to abnormal immune responses, which in turn may act as a driving force to exacerbate the progression of inflammation-mediated neurodegeneration.</p><!><p>The PD brain displays numerous signs of ongoing inflammatory processes. Both PD patients and animal models of PD display higher levels of activated microglia, which remain phagocytic for a prolonged period of time [4,5]. Elevated levels of inflammatory cytokines, especially tumour necrosis factor-α (TNF-α), interleukin-1β (IL-1β), IL-2, IL-6, IL-8, and interferon-γ (IFN-γ), are detected in the brain, cerebrospinal fluid, and blood of PD patients (reviewed in ref. [2]). Infiltration of peripheral immune cells, notably CD4+ and CD8+ T lymphocytes, is found as a consequence of abnormal permeability of the blood–brain barrier, which normally keeps the CNS in an immune-privileged state [6]. The involvement of humoral immunity has also been implicated: Lewy bodies and dopaminergic neurones in the SNpc show strong immunolabelling for immunoglobulin G [7]. Collectively, all of these ongoing inflammatory processes involve activation of microglia, suggesting its relevance to the pathophysiology of PD.</p><p>Microglia, the resident macrophages in the CNS, ensure a healthy environment for neurones by conducting a suite of homoeostatic functions. These include clearing cell debris, extracellular protein aggregates, and excess neurotransmitters. In response to pathological stimuli, microglia become activated, proliferate, and accumulate at the site of injury, where they phagocytose dead cells and secrete inflammatory mediators and a myriad of cytotoxic factors, especially reactive oxygen species (ROS) and nitric oxide [8]. Activation and subsequent down-regulation of microglial activity are strictly controlled, as exaggerated inflammatory responses can be harmful to neurones [9].</p><p>Chronic pathological factors (including repeated exposure to environmental toxins, genetic predispositions, and abnormal immune responses) may prolong the activated state of microglia, potentially instigating a feedforward cycle of chronic degeneration of neurones and inflammation. This self-perpetuating cycle of microglia-mediated neurotoxicity is particularly relevant in PD. First, dopaminergic neurones in the SNpc are intrinsically vulnerable to metabolic stress, particularly that caused by dopamine oxidation or mitochondrial dysfunction. High cytosolic dopamine levels can be dangerous, as dopamine metabolites (e.g. 6-hydroxydopamine) are toxic to neurones. Furthermore, compared with mesolimbic dopaminergic neurones, dopaminergic neurones in the SNpc exhibit a much larger Ca2+ influx, which requires the endoplasmic reticulum (ER)–mitochondrial system to clear excess Ca2+ (reviewed in ref. [10]). Secondly, the SNpc is one of the brain regions with the highest density of microglia and a relatively low density of astrocytes [11,12]. These factors suggest that the feedforward cycle of chronic activation of microglia and chronic damage of dopaminergic neurones would be particularly detrimental in the SNpc.</p><!><p>(Top panel) In this scenario, mutations in PD genes exert their effects directly in neurones, leading to chronic neuronal damage. This triggers microglial activation, leading to a vicious cycle of neuronal death and chronically activated microglia, with inflammatory cytokines causing collateral damage. The threshold for microglial activation may be lowered by peripheral inflammation, which 'primes' microglia. (Bottom panel) In this scenario, mutations in PD genes (in this case LRRK2) also exert effects by expression in macrophages and microglia, leading to dysfunctional immune responses. LRRK2-mutant microglia may exhibit exaggerated responses to neuronal damage, causing an amplified vicious cycle. The threshold for microglia activation could be further lowered by LRRK2 mutations within peripheral immune cells contributing to chronic microglial priming.</p><!><p>For the detailed protocol for HiPSC-derived macrophage differentiation, see ref. [52]. BMP4, bone morphogenetic protein 4; VEGF, vascular endothelial growth factor; SCF, stem cell factor; IL-3, interleukin 3; MCSF, macrophage colony-stimulating factor; d, days.</p><!><p>LRRK2 is a large, multidomain protein, displaying both GTPase and kinase activities. Most PD-causing mutations, notably R1441C/G and G2019S, cluster within these two enzymatic sites, which are surrounded by large protein–protein interacting domains (reviewed in ref. [15]). LRRK2 mutations are one of the most common genetic causes of PD: mutations can account for as much as 40% of familial PD [16] and its variants are also found within idiopathic cases [17]. Unlike other PD-associated genes, LRRK2 Parkinsonism manifests similar clinical phenotypes to idiopathic PD, displaying strong age-dependent development of PD symptoms [18]. Deciphering the role of LRRK2 in PD pathogenesis may reveal common pathological mechanisms underlying idiopathic PD and is consequently of great research interest.</p><p>Despite intense research effort over the past decade, the physiological function of LRRK2 and the contribution of mutations to PD remain largely elusive. This is at least in part because earlier research has mainly focussed on the role of LRRK2 in neurones, in which endogenous expression is low [19]. Many studies have, therefore, relied on overexpression of LRRK2 in non-physiologically relevant cell lines or animal models, but these approaches generate results that do not necessarily reflect the normal physiological interactome of LRRK2. This, together with the complex, multidomain structure of LRRK2, with protein–protein interaction domains at the N- and C-terminal segments, has led to LRRK2 being reported to interact with numerous molecules in a wide variety of cellular pathways, including endosome vesicle trafficking, cytoskeleton reorganization, mitochondrial function, regulation of ER/Golgi retromer complex, autophagy, and various signalling pathways, including wingless/int, TNF-α/Fas ligand (FasL)/Fas-associated protein with death domain, mitogen-activated protein kinase, and nuclear factor κ-light-chain-enhancer of activated B cells pathways [20–23]. Future investigations should examine LRRK2 expressed at physiological levels from its endogenous promoter at its normal human chromosomal location in authentic, relevant human cells, to discern which of these cellular pathways represent the bona fide function(s) of LRRK2.</p><!><p>Although the existence of LRRK2 protein in microglia and astrocytes has been reported in the past [24], and LRRK2 variants have been linked through GWAS to Crohn's disease [25,26] and leprosy [27], it was not until 2010 that researchers found that LRRK2 expression is precisely up-regulated by inflammatory signals in myeloid cells, strongly implicating its potential role as a regulator of immune responses [21].</p><p>Although the level of endogenous LRRK2 is low in resting leukocytes, upon stimulation with IFN-γ, robust up-regulation of endogenous LRRK2 has been consistently detected across various subsets of myeloid cells and lymphocytes, human peripheral blood mononuclear cell-derived CD11b+ monocytes, CD3+ T lymphocytes, CD19+ B lymphocytes [21], human primary monocyte-derived macrophages, mouse primary microglia [28], and transformed cell lines, including human THP-1 monocytic leukaemia cells [29,30], and murine RAW264.7 macrophage-like cells [21]. IFN-γ activation has a direct effect on the LRRK2 promoter region, which contains binding sites for IFN-response factors [21]. Janus kinase/signal transducers and activators of transcription and the extracellular signal-regulated kinase 5 mediate IFN-γ-induced LRRK2 up-regulation in macrophages, although the exact signalling cascades are yet to be elucidated [29].</p><p>LRRK2 is also moderately inducible by other inflammatory mediators, namely IFN-β, TNF-α, and IL-6 [31], whereas the Toll-like receptor 4 (TLR4) agonist, LPS, is found to have an inconsistent effect. Some groups have reported significant up-regulation of LRRK2 protein expression by LPS in primary mouse microglia or in THP-1 cells [28,32], whereas others did not detect any changes in murine immortalised microglia (BV-2), primary mouse microglia [33,34], or mouse bone marrow-derived macrophages (BMDMs) [35]. This discrepancy could be attributed to many factors, such as cell types, experimental conditions, or technical variations. Regardless, activation of TLR4 reproducibly leads to phosphorylation at Ser910/935 residues of LRRK2 in all myeloid cell lineages [19,35,36]. Phosphorylation at Ser910/935 determines its cellular localization, interaction with 14-3-3 protein, dimerization, and translocation from cytosol to the membrane [19,37–39]. However, further experimental investigation is needed to understand the direct physiological consequences of phosphorylation at Ser910/935 residues in myeloid cells.</p><!><p>Microglia and macrophages are both classified as mononuclear phagocytes, sharing common functions of various maintenance and protective roles, but can be distinguished by their ontogeny and transcription profiles (reviewed in ref. [40]). Since biochemical changes in LRRK2 upon inflammatory cues are identical in both systems [19], efforts have been made to inspect the role of LRRK2 in various aspects of innate immunity, summarized in Table 1. These studies imply opposing roles of LRRK2 in peripheral and CNS innate immunity. However, it should be noted that none of these studies has directly compared microglia and macrophages under the same experimental conditions, so direct evidence for opposing roles is still lacking. Moreover, most data are from mouse models that do not faithfully recapitulate all aspects of the human immune system [41]. Table 1Summary of reports on the role of LRRK2 in innate immunitySpeciesCell typesMethodsResultsReferencesLPS-mediated cytokine and chemokine releaseMouseBV-2LRRK2 knockdown (KD) shRNA↓TNF-α, IL-6, Nitrite[42]MousePrimary microgliaLRRK2 KD RNAi Kinase inhibition: sunitinib LRRK2-in-1↓TNF-α[32]MousePrimary microgliaLRRK2 KO Kinase inhibition: LRRK2-in-1 GSK2578215A↓IL-1β, cyclooxygenase-2 mRNA[33]MousePrimary microgliaR1441G↑TNF-α; ↓IL-10[28]MouseBMDMsR1441C LRRK2 KONo difference in IL-6 or keratinocyte chemokine (KC)[43]MouseBMDMsLRRK2 KONo difference in TNF-α, IL-6, KC, IL-1β, IL-10, IL-12[35]MouseThioglycollate-elicited peritoneal macrophages (TEPMs)LRRK2 KONo difference in IL-1β, IL-10, IL-1α, TNF-α, IL-6, KC, granulocyte colony-stimulating factor, monocyte chemoattractant protein-1[44]MouseTEPMsG2019SNo difference in TNF-α[36]MigrationMousePrimary microglia BV-2G2019S↓ADP-induced migration[48]LRRK2 KD shRNA↑ADP-induced migrationMousePrimary microgliaLRRK2 KO↑Fractalkine-induced migration[49]MousePrimary TEPMsG2019S↑ADP-induced migration[36]Kinase inhibition SRI 29451 HG-10-102-01↓G2019S-enhanced ADP-induced migrationPhagocytosisMouseBV-2 RAW264.7LRRK2 KD shRNA Kinase inhibition LRRK2-in-1 GSK2578215A HG-10-102-01No difference in uptake of FITC-conjugated beads[19]MousePrimary TEPMsG2019SNo difference in uptake of fluorescent zymosan bioparticles[36]MouseRAW264.7Salmonella typhimurium infectionLRRK2 localization to phagosomes[21]LRRK2 KD SiRNA↓ROS with zymosan ↑Survival of intracellular S. typhimurium</p><!><p>In mouse primary microglia, R1441G mutation leads to an increase in LPS-driven inflammatory cytokine release [28], and abolishing LRRK2 protein expression has the opposite effect [32,33,42]. However, in mouse primary macrophages, neither R1441G [43] nor G2019S mutations [36], nor LRRK2 knockout (KO) [35,43,44], cause any change in LPS-driven cytokine release. Only Dectin-1 activation by zymosan (yeast) has been shown to produce higher levels of cytokine release in LRRK2 KO mouse macrophages [45]. Therefore, in macrophages, LRRK2 may be dispensable in TLR4-mediated cytokine release, but may serve an important role in responding to other inflammatory stimuli. A recent report has shown that higher levels of peripheral inflammatory cytokines were found in the sera of both asymptomatic LRRK2 G2019S carriers and PD patients carrying LRRK2 G2019S [46], suggesting pathological contributions from LRRK2 mutations within peripheral immune cells. Further studies are merited to establish which specific inflammatory pathways are mediated by LRRK2 in macrophages or microglia. Additionally, the role of LRRK2 in peripheral cytokine release is still equivocal, and it will help the field to have studies that directly compare cytokine release in microglia and macrophages from the same animals under the same experimental conditions, comparing all the key LRRK2 mutations and KO together. Finally, LRRK2-mediated immune responses by more disease-relevant immunogenic agonists, such as aggregates of SNCA, should be explored further, as there is evidence for close interaction between α-synuclein and LRRK2 (reviewed in ref. [47]).</p><!><p>The ability to respond and migrate to the site of injury or infection is a key feature of innate immunity, and recent evidence suggests the involvement of LRRK2 in this process. Abolishing LRRK2 expression in BV-2 microglial-like cells [48] or mouse primary microglia [49] leads to significantly higher migration versus wild type, and the G2019S mutation in mouse primary microglia diminishes ADP-induced migration [48]. However, in mouse primary macrophages, the same mutation has been shown to have the opposite effect, enhancing motility towards ADP [36].</p><!><p>The absence of LRRK2 activity in BV-2 microglial-like cells or RAW 264.7 macrophage-like cells has been reported to have no effect on phagocytosis [19]; likewise, G2019S mutation was not observed in mouse primary macrophages [36]. However, it has been reported that LRRK2 localizes to phagosomes upon bacterial infection, and that lack of LRRK2 expression reduces ROS production and enhances bacterial survival in RAW 264.7 cells [21]. Therefore, LRRK2 may have a specific role during phagocytosis, which may not be detected by simple phagocytosis assays measuring initial uptake of bioparticles. Indeed, LRRK2 is implicated in autophagy in myeloid cells ([19]; reviewed in ref. [50]), a pathway that shares common features with phagocytosis and is also involved in innate immunity (reviewed in ref. [51]). Further exploration of the role of LRRK2 during specific stages of phagocytosis pathways is merited.</p><!><p>It is clear from above that the data on the role of LRRK2 in myeloid cells has so far been collected primarily from murine ex vivo models or from transformed murine cell lines, under a wide variety of experimental conditions. While murine systems are extremely useful, they do not precisely replicate all human cellular and biochemical pathways [41]. Moreover, transformed cell lines poorly reproduce the cellular physiology of authentic primary human macrophages and microglia, which are terminally differentiated cells. Human-induced pluripotent stem cell (HiPSC)-derived macrophages provide an attractive and highly authentic model to study LRRK2 biology (Figure 2). HiPSC-derived macrophages are genetically tractable, can be generated efficiently, and, at scale, become terminally differentiated and accurately recapitulate macrophage functionality [52]. Together with clustered regularly interspaced short palindromic repeats (CRISPR)/Cas-9 gene editing of iPSCs, one can investigate LRRK2 protein at the endogenous level by generating KO lines, correcting and introducing mutations, creating reporter lines, or tagging endogenous proteins. The applicability of this system has been shown already in other diseases, including HIV and chronic granulomatous disease [53,54]. Methods for skewing HiPSC-derived macrophages to microglia are currently under development and will enable direct comparisons of LRRK2 function in human macrophages and microglia.</p><!><p>Recent evidence supports the idea that pathological interplay between peripheral and CNS innate immunity probably contributes to the progression of PD. LRRK2 may be involved in this interplay, as expression of LRRK2 is tightly regulated in both systems and evidence reviewed here implicates LRRK2 in both peripheral and CNS innate immunity. Although the current literature appears to suggest that LRRK2 plays distinct roles in microglia and macrophages, more work needs to be done to unequivocally establish the bona fide function(s) of LRRK2 in human macrophages and microglia, and the role of LRRK2 mutations in these cells in PD. To achieve this, macrophages/microglia differentiated from HiPSCs provide a powerful tool to better understand LRRK2-mediated pathology in PD and also other LRRK2-mediated immune disorders.</p><!><p>bone marrow-derived macrophages</p><p>central nervous system</p><p>clustered regularly interspaced short palindromic repeats</p><p>endoplasmic reticulum</p><p>Fas ligand</p><p>genome-wide association studies</p><p>human-induced pluripotent stem cell</p><p>interferon-γ</p><p>interleukin-1β</p><p>knockout</p><p>lipopolysaccharide</p><p>leucine-rich repeat kinase 2</p><p>parkin</p><p>peripheral blood mononuclear cell</p><p>Parkinson's disease</p><p>reactive oxygen species</p><p>α-synuclein</p><p>substantia nigra pars compacta</p><p>Toll-like receptor</p><p>tumour necrosis factor-α.</p><!><p>We acknowledge financial support from the Wellcome Trust [WTISSF121302], the Oxford Martin School [LC0910-004], the Oxford Parkinson's Disease Centre (OPDC) Monument Trust Discovery Award from Parkinson's UK, a charity registered in England and Wales [2581970] and in Scotland [SC037554], with the support of the National Institute for Health Research (NIHR) Oxford Biomedical Research Centre based at Oxford University Hospitals NHS Trust and University of Oxford, and the NIHR Comprehensive Local Research Network, and European Federation of Pharmaceutical Industries and Associations (EFPIA), European Union Innovative Medicines Initiatives (EU IMI) "StemBANCC", who provide the following statement: The research leading to these results has received support from the Innovative Medicines Initiative Joint Undertaking under grant agreement no. 115439, resources of which are composed of financial contribution from the European Union's Seventh Framework Programme (FP7/2007–2013), and EFPIA companies in kind contribution. This publication reflects that only the author's views and neither the IMI JU (www.imi.europa.eu) nor EFPIA, nor the European Commission, are liable for any use that may be made of the information contained therein.</p><!><p>The Authors declare that there are no competing interests associated with the manuscript.</p>

PubMed Open Access

Effect of Adding Cerium on Microstructure and Morphology of Ce-Based Inclusions Formed in Low-Carbon Steel

Intra-granular Acicular Ferrite (IAF), as one of the most well-known desirable microstructure of ferrite with a chaotic crystallographic orientation, can not only refine the microstructure and retard the propagation of cleavage crack but also provide excellent combination of strength and toughness in steel. The effect of adding cerium on microstructure and controlling proper cerium-based inclusions in order to improve properties in low-carbon commercial steel (SS400) were investigated. The type of inclusions can be controlled by changing S/O ratio and Ce content. Without Ce modification, MnS is a dominate inclusion. After adding Ce, the stable inclusion phases change from AlCeO 3 to Ce 2 O 2 S. The optimum amount of cerium, 0.0235 wt.%, lead in proper grain refinement and formation of cerium oxide, oxy-sulfide and sulfide inclusions. Having a high amount of cerium results in increasing the number of inclusions significantly as a result it cannot be effective enough and the inclusions will act like barriers for others. It is found that the inclusions with a size of about 4∼7 μm can serve as heterogeneous nucleation sites for AF formation. Thermodynamic calculations have been applied to predict the inclusion formation in this molten steel as well, which show a good agreement with experimental one.Recently, the issues concerning non-metallic inclusions in steels have become one of the leading subjects of research in the field of metallurgy due to its important effect on the quality of steel. Before, non-metallic inclusions were treated as a detriment but indispensable product of deoxidation and desulphurization of steel, which should be removed or modified as much as possible so it did not decrease mechanical properties. Many researchers have investigated on improving the behavior of non-metallic inclusions during solidification of steel. It was found that depending on the chemical composition and their size, they can have different impact on properties of steel, starting from strongly negative to almost conditioning obtaining desired mechanical properties [1][2][3][4] .Lots of studies have indicated that the inclusions, such as Ti, Al, and Zr oxides, Ti, Nb, and V carbonitrides, would contribute to the IAF nucleation, and the optimal heterogeneous nucleus was Ti 2 O 3 5-9 . Many researchers reported that anisotropic microstructure and elongation of deformable MnS inclusions often happen as a result of the metal forming processes such as rolling and forging. Although the adding sulfur improves the machinability of steel, anisotropy in mechanical and fatigue properties would occur due to the presence of deformable MnS inclusions. In order to modify MnS inclusions, different methods are used by adding of Ca, REM (Rare-Earth-Metals) or Zr in the melt. The modification of non-metallic inclusions by means of Ca-treatment of liquid steels are often limited by the low and unstable yield of the added Ca, due to the high vaporization and low solubility of Ca in the liquid steel. Therefore, steelmaking companies prefer to use some other elements with higher vaporization temperature in the melt such as REM and Zr to modify sulfide inclusions 10 .The studies about the effect of rare earth (RE) elements on the welding microstructures and properties indicated that the RE could react with O and S with the result of forming the high-melting point RE x O y , RE x S y and RE x O y S z 5 . The key factors for the nucleation of intergranular bainite or acicular ferrite are the control of austenite grain size as well as the adjustment of the nature and size of non-metallic inclusions, which are both considered as favorable phases for mechanical properties at room temperature. The Gibbs' free energies of these compounds

effect_of_adding_cerium_on_microstructure_and_morphology_of_ce-based_inclusions_formed_in_low-carbon

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

<p>at high temperatures are so low that REM elements can combine readily with oxygen and sulfur when added to liquid steel 11,12 . RE elements, with a strong affinity to oxygen and sulfur, were widely applied to spherodizing inclusions (such as MnS) to avoid the anisotropy of mechanical properties in final rolling products.</p><p>Recently it is reported that Ce 2 O 3 with a low misfit value with ferrite can act as the nucleation sites for IAF under fast cooling rate. However, the inclusion characteristic and microstructure of furnace-cooled RE containing sample and inclusion formation evolution have not been discussed yet. Zhang et al. 13 investigated the ability of Mg-based inclusions to induce AF nucleation in SS400 steel, which was Mg-treated using a commercial process. Their results showed that the magnesium-based complex inclusions could act as nucleation sites of AF. Inclusions with a size of about 5 μm can serve as heterogeneous nucleation sites for AF.</p><p>Anmark et al. 10 reviewed and summarized the effect of different non-metallic inclusions on the machinability of various steels. He mentioned that the magnitude of the effect of non-metallic inclusions on the improvement of the machinability of steel matrix, depends on the difference in the thermal expansion coefficients, α , between the steel matrix and the non-metallic inclusions. In the case that the non-metallic inclusions have different compositions and α coefficient than that of the steel matrix, the steel machinability can be affected. However, the value of the α coefficients can be different based on the contents of carbon and alloying elements. He concluded that the effect of the oxides and sulfides of REM and Zr on improving of machinability will be higher than the effect of MnS inclusions, which can be explained by the significantly lower magnitude of the difference between the values of α MnS and α steel for the stainless and high alloyed steels.</p><p>Different researchers worked on different aspects of adding REM in steels such as mechanisms of inclusion evolution, REM effects on the mechanical properties and impact toughness, and in situ observation of the evolution of IAF 4,5,11,14 . Deng et al. 4 proposed a possible inclusion evolution mechanism based on calculated results using both calculations and experiments. Bin et al. 5 worked on the in situ observation of the evolution of intragranular acicular ferrite at Ce-containing inclusions in 16Mn Steel. He also reported that the optimum content of Ce in 16Mn steel is around 0.02 wt%. Although different researchers have been investigated the effects of REM in steels, a comprehensive research in this field for low carbon steel, is still needed. The present study is designed to investigate the effect of cerium addition for grain refinement of SS400 steel comprehensively. The proper amount of Ce and controlling the type of inclusions have been discussed in details in order to find a reasonable relationship for industrial applications.</p><!><p>Commercial SS400 steel was melted at 1873 K in a vacuum induction furnace (100 kHz). Once the alloy was melted in a furnace under argon gas atmosphere, the melt was deoxidized with adding different amount of cerium powder wrapped in pure aluminum foil (99.99%). In order to control the type of Inclusions, S/O ratio has been changed. A wide range of samples are prepared, then the furnace power was turned off and crucible with the melt for the sample was slowly cooled down in the furnace, finally quenched with water. The chemical composition of as-cast samples is analyzed and presented in Table 1.</p><p>The amount of cerium is analyzed by ICP-AES method. In order to clarify the inclusions in steel samples, 1 × 1 × 1 cm 3 cubic samples were cut from sample steel, then ground and polished using 3 and 1 μ m diamond compound. A wide range of characterization methods are used in this research including Laser Scanning Confocal Microscopy (LSCM), ASPEX Explorer SEM/EDS, Scanning Electron Microscopy (SEM-EDS) and Optical Microscope (OM). The in situ observation of microstructure transformation was carried out using LSCM on cylindrical specimens, 8 mm in diameter and 10 mm in length. The sample for inclusion analysis was machined and then ground and polished using diamond compound. Metallographic observations were carried out on the specimens subjected to casted state. The statistic of grain size and inclusions were examined by using image statistical analysis. The types of inclusions and their morphology were extensively analyzed by SEM equipped with energy dispersive X-ray spectroscopy (EDX). In the present study, using ASPEX, the size distribution, composition, number and morphology of inclusions are automatically obtained for each sample. The total area examined for this test was 89.653 mm 2 . For observation of the microstructures, OM is employed for the samples etched for 1-2 mins using 3% Nital.</p><!><p>Thermodynamic analysis of rare earth inclusions formed in SS400. Interaction parameters according to Wagner 15 and Lupis 16 and Elliott have been very successfully used in the study of deoxidation reactions of steel for many years 17,18 . The interaction parameters bear significant correlation with properties that have physical meaning such as heat of formation of the corresponding oxides and atomic number of the deoxidants. These correlations not only help support the soundness of the formalism but also provide an interesting and useful way of checking the consistency of data presented in this formalism, as shown in A. Costa e Silva's work 17 . He investigated the interaction parameters of oxygen and deoxidants in liquid iron. Many researchers were in search of a mathematical way of handling the behavior of solutes in dilute solutions and were aware of the limitations this approach would have for less diluted solutions. This lead to the formalism of interaction coefficients for dilute solutions, widely used today 17 . In the addition of rare earth elements to the molten steel, there is a strong affinity among Ce and O and S. As a result, the thermodynamic calculation can be applied to derive the thermodynamic equations of inclusion formation in this steel. It is reported in the literature 14 that the effect of rare earth is optimal when w (RE)/(w [O] + w [S]) = 3.9. Henrian activity coefficients and Henrian activities (1 wt% standard state) of O, S, Ce and Al in liquid steel can be predicted by interaction coefficients with Wagner's model. In order to control the inclusions in practice, the thermodynamic analysis for the formation of the inclusions is performed. The activity coefficient of each element and activity in liquid steel is calculated by Eqs. ( 1) and ( 2):</p><p>Where ∆ θ G denotes the Gibbs free energy of reaction with the unit of J.mol; −1 R is the gas constant with the unit of J.mol −1 .K; −1 T is temperature with unit of Kelvin; K is the equilibrium constant (without unit); a i is the activity of element i; f i is the Henrian activity coefficient of component i in dilute solution; e i j is the first-order interaction parameters i and j; w[i], w[j] are the mass percentage of elements i and j, respectively.</p><p>Standard Gibbs free energy change for formation of Ce 2 O 3 and Ce 2 O 2 S in liquid steel is expressed respectively by Eq. ( 3) and (4) 14 .</p><p>In addition to cerium oxides and oxy-sulfides, cerium sulfide also can be found in this type of steel. The formation of different type of cerium sulfides may be controlled by following equations 19 . According to Eq. ( 8) and ( 9), the adding rare earth elements can react with the existing Al 2 O 3 to form CeAlO 3 or the rare earth elements directly react with oxygen and aluminum to form CeAlO 3 .</p><p>423900 247 30T (J mol ) (8)</p><p>The activity coefficient of each element and activity in liquid steel is calculated by Eqs. ( 1) and ( 2). The basic chemical compositions are shown as No. 4 in Table 1. Table 2 shows the interaction coefficient e i j of various elements in liquid steel at 1873 K. The corresponding interaction coefficients of O, S, Ce and Al, at 1873K are used from previous researchers 19,20 . On the basis of the data, Henrian activity coefficients of O, S, Ce and Al in liquid steel containing 0.0235 wt% rare earth elements at 1873 K are calculated as 0.111, 0.926, 0.020 and 1.006, respectively. According to them Henrian activities are performed as well. The value of Raoultian activities of Ce 2 O 3 , Ce 2 O 2 S, and CeAlO 3 is assumed to be unity 19,21 .</p><p>The effects of Ce, S and O contents on the stability of inclusions in SS400 steel were studied, as shown in Fig. 1. According to the results, it can be seen that for a certain amount of cerium, the formation of CeO 2 , Ce 2 O 3 , and CeAlO 3 needs a higher amount of oxygen, respectively. Besides, for a low amount of cerium, formation of all these inclusions CeO 2 , Ce 2 O 3 , and CeAlO 3 needs a higher amount of oxygen. For high amount of cerium and S, Ce 2 O 2 S can easily form even in low amount of oxygen. The dominant inclusions are CeO 2 and Ce 2 O 2 S in this study. Generally speaking, as the binding capacity of RE oxide and oxygen is greater than that of aluminum and oxygen, the liquid steel first produces REAlO 3 , and the reaction equation is</p><p>occur when the addition of RE is increased, which can modify Al 2 O 3 inclusions, and play a role of desulfurization 22 . RE aluminates, RE oxides, RE sulfur oxide, and RE sulfides will appear in turn by the free energy calculations with RE addition. According to Fig. 1b, it can be seen that for a certain amount of cerium, the formation of Ce 2 S 3 , Ce 3 S 4 and CeS needs a higher . Different type of inclusions can be found in these samples. Without adding cerium, MnS is a dominate inclusion in samples, which is presented in Fig. 3b. By adding a small amount of Ce (No. 2, 60 ppm), we will have 5 types of inclusions including complex inclusions without Ce (Fig. 3c), complex inclusions with Ce (Fig. 3d), cerium oxy-sulfide (Fig. 3e), MnS (Fig. 3f), CeAlO 3 (Fig. 3g). In sample No. 3 (169 ppm Ce), only cerium oxide (Fig. 3k) and cerium oxy-sulfide (Fig. 3h, 3l) inclusions can be found. There is a direct relationship among amount of Ce, S/O ratio and the type of formed inclusions. By increasing the amount of cerium and decreasing S/O, we have a large amount of inclusions most of them cerium oxy-sulfide and cerium oxides, no cerium sulfide has been detected. In contrast, increasing S/O ratio can result in formation of cerium sulfide, which can be found in sample No. 4 (with 235 ppm Ce). In this sample, different inclusions are cerium oxide (Fig. 3m), cerium oxy-sulfide (Fig. 3n) and cerium sulfide (Fig. 3o). Most of inclusions have been cerium oxides or cerium oxy-sulfide, only a small amount of cerium sulfide has been detected, as the S/O ratio is not that much high to form more amount of cerium sulfide. Sample No. 5 with 1527 ppm cerium, has different types of inclusions including cerium oxy-sulfide with a high amount of S (Fig. 3p), cerium oxy-sulfide (Fig. 3q), and small amount of complex inclusion (Fig. 3r).</p><p>In order to have a better understanding, the relationship between cerium amount, S/O ratio, and type of formed inclusions as well as the thermodynamic explanations have been summarized in Table 3.</p><p>Figure 4. show the microstructure of different samples with changing Ce amount after etching by Nital 3% under Optical Microscopy. Figure 4a (No. 0, without Ce) which consist of white plates of polygonal ferrite (PF) and Pearlite (P). Figure 4b (No. 1, 20 ppm Ce) includes aligned side plate ferrite FS(A), Pearlite (P) and small white plates of polygonal ferrite (PF), which are mainly located in grain boundaries. By increasing amount of cerium in sample No. 2 (60 ppm Ce), Fig. 4c, big white plates of polygonal ferrite (PF) and Pearlite (P), PF is becoming elongated, they also located in grain boundaries. Besides, this sample is full of elongated polygonal ferrite and perlite. By increasing amount of Ce to 235 ppm, shown in Fig. 4e, grain refinement and finer microstructure are clearly observed in this result compared with the non-modified sample or even the previous ones. A large amount of acicular ferrite (AF), the white plate shape regions as polygonal ferrite, and also Pearlite can be found in sample No. 4. This sample has 235 ppm which is result in having a reasonable amount of inclusions. Different types of inclusions such as cerium oxides, sulfides and oxy-sulfides are found in this sample as discussed in last section. The amount of cerium is not that much high and is distributed properly. The highest amount of AF is found in this sample compare with other samples. Having cerium sulfide has been effective for the formation of AF. Our results are in agreement with other researchers' [1][2][3][4][5][6][7][8][9][10][11][12] . Figure 4f is related to sample No. 5 (1527 ppm Ce). Having a high amount of cerium results in increasing the number of inclusions significantly. As a result, high amount of inclusions cannot be effective enough and the inclusions will act like barriers for other. It can be clearly seen that amount of AF is lower than sample No. 4 which can be contributed to this high amount of Ce. A small amount of acicular ferrite (AF), high amount of the white plate shape regions of polygonal ferrite, and also Pearlite can be found. OM results are in agreement with SEM-EDS results. A high amount of inclusions such as cerium oxides, and oxy-sulfides are found in this sample. This sample shows clearly that a proper amount of cerium is necessary; otherwise, the inclusions not only help to grain refinement but also act like barriers to each other.</p><p>The formation of IAF in steel was influenced by various factors, such as the composition of inclusions, the amount and size of inclusions, the cooling rate, and the parent austenite grain size (PAGS) and so on [23][24][25] pointed out that when the PAGS reached the optimum value, the volume fraction of IAF reached the maximum. In situ observation of austenite grain refinement by means of confocal microscopy after REM modification. The austenite grain refinement for different samples with different amount of cerium is presented by CM images in Fig. 6. Different samples with different amount of cerium are prepared in order to see which amount of cerium is effective for austenite grain refinement. Besides, the distribution of inclusions will be considered as well. The parent austenite grain size (PAGS) is considered as one of important factors affect the formation of IAF. Bin et al. 5 reported when the PAGS reached the optimum value, the volume fraction of IAF reached the maximum. The amount of cerium is increased from sample No. 0 to No. 5, respectively. According to the results, it can be seen that the austenite grain size is decreased after adding cerium. Austenite grain size for non-modified sample with Ce is about 130 μ m. For sample No. 4, a medium amount of inclusions has been distributed and they can probably act as the heterogeneous nucleation site for grains, the average austenite grain size for this sample is around 100 μ m. As a result, we will have a finer austenite grain for this sample compare with the others. For sample No. 1 with a small amount of Ce, 20 ppm, only a small refinement can be achieved, but for No. 4 with 235 ppm Ce and No. 5 with higher amount of Ce resulted in austenite grain refinement. Moreover, the amount of inclusions in sample No. 5, is really high. As a result, high amount of cerium cannot be effective enough.</p><!><p>This study examined the effects of adding cerium on microstructure and morphology of Ce-based inclusions formed in commercial SS400. The results are summarized as follows:</p><p>1. Different samples with different amount of cerium have been prepared. For sample with 0.0235 wt.% cerium and S/O ratio around 7, the type of inclusions will be large amount of cerium oxide, cerium oxy-sulfide, and small amount of cerium sulfide. 2. As the binding capacity of Ce oxide and oxygen is greater than that of aluminum and oxygen, the liquid steel first produces CeAlO and Ce sulfides will appear in turn by the free energy calculations with Ce addition. 3. By increasing the amount of cerium and decreasing S/O, a large amount of inclusions can be detected which most of them are cerium oxy-sulfide. It's worth to mention no cerium μ m sulfide is detected when S/O ratio is low. There is a direct relationship among amount of Ce, S/O ratio and the type of formed inclusions. Besides, having a high amount of cerium results in increasing the number of inclusions significantly. As a result, high amount of inclusions cannot be effective enough and the inclusions will act like barriers for others. 4. Inclusions with a size of about 4∼ 7 μ m can serve as heterogeneous nucleation sites for IAF formation. The amount of cerium is not that much high and is distributed properly. The highest amount of AF is found in the sample with 0.0235w% Ce compare with other samples. Having cerium sulfide has been effective for the formation of IAF.</p>

Scientific Reports - Nature