|Home | About | Journals | Submit | Contact Us | Français|
Potential binding modes of (+)-discodermolide at the paclitaxel binding site of tubulin have been identified by computational studies based on earlier structural and SAR data. Examination of the prospective binding modes reveal that the aromatic pocket occupied by the paclitaxel side-chain is unoccupied by (+)-discodermolide. Based on these findings, a small library of (+)-discodermolide-paclitaxel hybrids have been designed and synthesized. Biological evaluation reveals a two- to eight- fold increase in antiproliferative activity compared to the parent molecule using the A549 and MCF-7 cancer cell lines.
(+)-Discodermolide [(+)-1], a potent antitumor polyketide natural product, was first isolated in the early 1990’s from extracts of the Caribbean marine sponge Discodermia dissoluta by Gunasekera and coworkers.1 Confirmation of the assigned structure, as well as the absolute stereochemistry, was subsequently achieved by Schreiber and coworkers via total syntheses of the (−) enantiomer (1993), the natural (+) enantiomer (1994) and a number of novel analogs.2,3 Initially reported to be a potent immunosuppressive agent, (+)-discodermolide was later recognized to possess significant antiproliferative activity both in vitro 4a across the NCI panel of human cancer cell lines,5 including paclitaxel resistant cell lines,6 and in vivo.4b Like paclitaxel (2), (+)-discodermolide stabilizes tubulin thereby modulating microtubule dynamics that leads to cell arrest at mitosis.
As a result of the observed antitumor activity, (+)-discodermolide generated significant interest in the synthetic community: 13 total syntheses have been reported to date.7 Of particular significance, as recently recognized,8 our four generation strategies exploited an early example of the Negishi coupling protocol to access the C14–C15 olefin. 7e,9 Subsequently Novartis employed a similar cross-coupling tactic (i.e. Suzuki) to provide (+)-discodermolide for their Phase I clinical trials.10 Also early on, we both proposed 11 a solution conformation of (+)-discodermolide based on a combination of NMR data and the solid state X-ray structure and carried out a comprehensive structure activity relationship (SAR) study with Kosan Biosciences, Inc. 12,13
More recently we demonstrated that (+)-discodermolide both causes accelerated cell senescence,14 and binds to a region on β-tubulin close to the paclitaxel binding site.15 Of interest here (+)-discodermolide and paclitaxel act synergistically, not only in cell culture16 but also in an ovarian xenograft tumor model in nude mice.17 Subsequent studies, utilizing hydrogen/deuterium exchange and mass spectrometry, indicated that paclitaxel and discodermolide induce stability on opposite sides of the microtubule interface, suggesting a possible explanation for the synergy observed between these two drugs.18 Finally to decrease the pneumotoxicity of (+)-1, the presumptive cause leading to the failure of the Novartis Phase I clinical trial,10 possibly due to metabolites,19 we extended our analog program, wherein the metabolically less stable lactone portion of the molecule 13a–c and C14–C15 olefin,13d were either individually or together, replaced with isosteres. Pleasingly these analogs preserved the cytotoxicity at a level similar to (+)-discodermolide (1).
However despite the availability of extensive SAR data,12 as well as structural,1 computational and NMR studies,11,20 the binding mode of (+)-discodermolide on tubulin remains unclear. Towards this end, we report here two possible binding modes of (+)-discodermolide in conjunction with the design, synthesis and biological evaluation of (+)-discodermolide/paclitaxel hybrid molecules, possessing photoaffinity labels.
In our SAR studies, we demonstrated the following: (1) that changes in the substitution pattern of C(1), C(3) and C(7) centers of (+)-discodermolide are favorable to retention of potent cytotoxicity; (2) that variations at the C(2), C(14) and the carbamate carbons are tolerable; and (3) that changes within the remaining portions of (+)-discodermolide significantly decrease the cytotoxicity.12,13 Based on these observations, in conjunction with our earlier photoaffinity labeling study,12e,f we reasoned that the diene terminus of discodermolide resides in a hydrophobic/aromatic pocket, while the lactone moiety either occupies a binding pocket that can accommodate large groups, or more likely, is partially interfaced with the solvent. Further information gained from our earlier photoaffinity labeling study revealed that attachment of a large aromatic system (cf. benzophenone) and/or alkyl groups to the carbamate nitrogen did not significantly alter the observed activity.12f In fact, in two cases, comprising attachment of a p-dimethylaniline or p-benzophenone substituent at the carbamate nitrogen, a significant increase in tumor cell growth inhibitory activity was observed. We emphasize however that our earlier study was directed specifically at the preparation of photoaffinity labeled (+)-discodermolide analogs, with little or no attention focused on activity changes, except for the possibility of increased hydrophobicity and/or conformational preferences. A closer examination of the earlier carbamate analogs however reveals a strong activity dependency both on the nature of the aromatic rings and linker length. Based on these observations, we have now examined the scope and molecular basis for the observed increase in cytotoxicity of these analogs exploiting detailed computational analysis.
In 2001, we reported that the preferred major solution conformation of (+)-discodermolide (1) was similar to the solid state structure,1 notwithstanding the 15 rotatable σ bonds.11 In particular, variable temperature and co-solvent titration NMR studies provided evidence of intramolecular hydrogen bonds between the C(7)-OH and the C(1)-carbonyl, in conjunction with interactions between the carbamate nitrogen and the C(3) and C(7) hydroxyl groups. We postulated that syn-pentane interactions21 were responsible for the proposed turn geometry of the backbone. Importantly, biological studies of epimers bearing inverted stereogenicities at the C(7), C(11), C(16) and C(17) revealed significantly lower activities, which led us to suggest that the conformation of the molecular segment between C(7)–C(17) is conserved upon binding. These conclusions were later confirmed by studies from other research laboratories.20
To understand more fully the solution behavior of (+)-discodermolide, we recently initiated computational studies exploiting molecular dynamics (MD) simulations. Monitoring backbone torsional angles over a 1 microsecond simulation period in water revealed the flexible torsional angles. Of considerable interest, the lactone ring and the diene side chain moieties, were found to display more flexibility compared to the rest of the molecule, when the turn geometry [i.e., C(7)–C(16)] was maintained throughout the simulation.
We next turned to NMR experiments to refine the solution conformation. The DISCON (Distribution of Solution Conformations) program developed in our laboratory22 was employed to obtain a distribution of conformers based on NMR data. Earlier studies had revealed that (+)-discodermolide is more flexible in organic solvents.20a, 20b Based on this observation, acetonitrile was selected as the solvent for the NMR experiments, given that the aim of this study was to identify the “flexible” torsional angles and possible diverse conformational families for our pending docking studies. Proton-proton coupling constants and NOE derived distances reported earlier were utilized.11 Deconvolution of the NMR observables over an ensemble of structures obtained by Monte Carlo conformational searches, followed by the clustering analysis that is embedded in DISCON software, led to an ensemble of four principal conformers, that in combination fit the NMR derived distance and torsional angle data better than any single conformation. Two conformers (Figure 2, A and B), similar to the solid-state structure with differences in the lactone ring orientation, were identified as the major solution conformations (A: 35% and B: 32%), an observation fully consistent with our earlier solution structure.11 Further examination of members of the major conformational families revealed (+)-discodermolide can adopt different conformations by adjusting the torsional angles around C(11)–C(12) and C(7) and C(8) bonds. In addition, the lactone ring can exist in equilibrium between chair, half chair and skew boat conformations. The next most populated conformer (C: 26%) comprised an inverted orientation at the diene region, wherein a different gauche orientation is preferred around the C(16)–C(17) and C(18)–C(19) bonds, with the remaining regions of (+)-discodermolide identical to the solid state structure. The remaining major family (7%) comprises both of these conformational changes, and thereby forces the molecule to occupy an extended conformation. The distribution of the conformations can be rationalized by the observed 3JH16–H17 and 3JH11–H12 values of 6.2 Hz and 6.6 Hz respectively, corresponding to torsional angle averaging and by the observed NOE correlation variations from the solid state structure, in particular those between H(7)–H(9), H(8)–H(12), H(9)–H(11), and H(12)–Me(18). Similar analyses reported earlier by the Jiménez-Barbero20b and the Snyder Laboratories20c are in accord with these observations, wherein differences in distributions were observed, presumably due to the differences in the solvents employed in the NMR experiments, as well as to the force fields and clustering methods employed.
In summary, (+)-discodermolide in solution displays considerable flexibility in the orientation of the lactone and the diene/carbamate termini, with conservation of the turn region of the conformation comprising the C(7)–C(16) centers. For docking studies, we reasoned that the solution structures hold considerable importance, since the bioactive conformation is more likely to exist in solution given that a lower energy penalty is paid for ligand reorganization upon binding (i.e., pre-organization).23 Stated differently, the 3D binding pharmacophore of the ligand is likely to be found in the conformation(s) that exist in solution, thus permitting easy recognition.24
Having an understanding of the available major ensemble of solution conformations, we turned to define the binding mode of (+)-discodermolide on microtubules. In general, docking studies of microtubule stabilizing agents utilize the heterodimer α-β tubulin structure (PDB code: 1JFF) obtained by high resolution electron microscopy from the Zn induced crystalline sheets of tubulin in the presence of paclitaxel.25 Although the tubulin structure in the crystal was shown to match the microtubule structure, it remains only an approximate model for docking studies for the following reasons: (1) the protein structure is distorted by the bound paclitaxel, and thus does not necessarily represent the unbound microtubule;26 (2) in vivo orientations and conformations of tubulin in microtubules and Zn-induced tubulin crystal sheets are different;27 (3) the flexible M-loop region, at the taxane binding site is not well resolved;25 (4) the lateral contacts between protofilaments, which are important for microtubule formation and tubulin dynamics28 are missing; and (5) in humans the tubulin isotypes have different sequences than found in the reported protein structure.29 Equally significant, the majority of reported (+)-discodermolide docking studies employ either the solid state conformation of (+)-discodermolide as a rigid scaffold, or (+)-discodermolide is treated as a fully flexible molecule without regard to the accessible solution conformations.15,18,30 Finally, earlier docking studies did not take into account the reported SAR data,19 in particular the potent carbamate analogs.13f
More recently Carlomagno20a and then Jiménez-Barbero et al. 20b concluded, based on NMR transfer NOE studies of (+)-discodermolide and tubulin, that the binding conformation of (+)-discodermolide is identical to the solid state conformation with differences in the lactone ring portion, and that (+)-discodermolide binds at the paclitaxel binding site. Paterson and coworkers in turn, based on these proposals, prepared a series of discodermolide-dictyostatin-paclitaxel hybrids which upon biological evaluation revealed a 20–35 fold decrease in tumor cell growth inhibition activity, employing the PANC-1 human cancer cell lines.31 Notwithstanding these disappointing observations, we undertook docking studies employing the solution conformations identified by DISCON (vide infra).
Our 2009 hydrogen-deuterium exchange experiments also revealed that (+)-discodermolide binds to a region quite close to the paclitaxel site on β-tubulin, however the stabilized microtubule geometries were different as evidenced by the observed solvent exposure of the lateral sites,18 with the most important interactions observed in the M-loop region of β-tubulin. Based on this information, we also chose to employ the paclitaxel site for docking studies. Beginning with the paclitaxel bound protein structure (PDB Code: 1JFF), paclitaxel was removed and the remaining protein structure was minimized with the OPLS-2005 force field to remove possible steric clashes. The binding site was not altered excessively, as we did not want to dock to a computationally generated binding pocket. Sets of (+)-discodermolide conformations, belonging to the families identified in solution (A–D, Figure 2), were then docked, employing the Glide32 and Molegro33 software packages, treating the (+)-discodermolide conformations as rigid structures and the protein structure as flexible. Docking modes inconsistent with the SAR data were eliminated by visual inspection. We reason that this approach provides more valid criteria for finding a working binding hypothesis than calculated docking scores and energies, due to the uncertainties (vide infra) in the protein structure (cf. Supporting Information for elimination criteria). Two binding poses for (+)-discodermolide were identified to be consistent with the SAR data.
The first pose (Figure 3B) orients the lactone end of (+)-discodermolide in a region occupied by the baccatin core of paclitaxel. The turn structure of (+)-discodermolide fits tightly in a region under the M-loop (β212–230) with the C(3) and C(7) hydroxyl groups forming hydrogen bonds to Pro272 and Thr274. In this pose, the carbamate moiety is directed towards the aromatic pocket that accommodates the paclitaxel tail (Figure 3A). The diene side chain and the lactone carbonyl group are directed towards the solvent, where additional substituents would appear to be tolerated. Rescoring of this pose via the MM/GBSA protocol34 provided a favorable ΔG value of −12 kcal/mol. The second pose (Figure 3C), similar to pose 1, utilizes the (+)-discodermolide turn scaffold to fit in the hydrophobic pocket under the M-loop. This pose however orients the lactone moiety to the region where the paclitaxel C(2) benzoate group binds. This region would also appear to be able to accommodate larger groups, while maintaining the activity of (+)-discodermolide analogs. That is, replacement of the lactone ring with other moieties might prove feasible. In this pose the known decrease in activity of (+)-discodermolide analogs upon acetylation at the C(11) hydroxyl group can be rationalized by the loss of a hydrogen bond to Thr274. This pose also orients the diene side-chain in line with the paclitaxel benzamide group, with the (+)-discodermolide carbamate group directed to the C(3′)-phenyl pocket. Rescoring employing the MM/GBSA protocol again provided a favorable ΔG value of −18 kcal/mol. Noteworthy, the Jiménez-binding mode differed significantly from the poses presented here, and was found to be inconsistent with our SAR criteria.
Based on the (+)-discodermolide docking modes, we designed a series of analogs that we reasoned would hold the promise of not only confirming our binding hypothesis, based initially on enhanced binding to tubulin and in turn enhanced cytotoxicity, but also defining the true binding pose. Importantly, the design criteria were orchestrated to accommodate incorporation of photoaffinity labels for future structural studies.
Utilizing docking poses 1 and 2, we envisioned filling the aromatic pocket in the vicinity of the carbamate group, which in the case of paclitaxel is occupied by C(3′)-Ph group. In silico docking of our earlier carbamate analogs (vide infra), employing the identified (+)-discodermolide poses, revealed that the earlier analogs did not fully occupy the proposed pocket due to the shorter tether lengths. To test the validity of this scenario, analogs having aromatic groups attached to the carbamate group with differing tether lengths were generated. We hypothesized that if the binding pocket is valid, the cytotoxic activities should vary, with the analog possessing the most suitable linker length displaying the highest tumor cell growth inhibition activity. A second amide linkage was incorporated onto the tether to participate in a possible hydrogen bond, with Asp26 and His229 of β-tubulin, similar to the paclitaxel amide linkage. Finally, we envisioned that photoaffinity tags could be appropriately introduced without deleterious effects on the cytotoxicity. To this end (+)-discodermolide/paclitaxel hybrid structures with a p-azidophenyl substituent (3–5, Figure 4) were generated. Docking studies were then conducted similar to the procedure applied for (+)-discodermolide. In these studies the newly introduced paclitaxel C(3′)-phenyl group surrogates were permitted to be flexible, whereas the (+)-discodermolide core was kept rigid. Examination of the docking scores identified hybrid 4 to fill the aromatic pocket. Two additional hybrids 6 and 7 possessing different photoaffinity tags and optimal linker lengths were also designed, to explore the potential effects of the hybrid structure motif, as well as the photoaffinity groups on the cytotoxicity.
The requisite hybrids, carrying the photoaffinity tags, were envisioned to arise via union of known (+)-discodermolide intermediate 19,7e prepared in connection with our gram-scale synthesis,11 with the photoaffinity units bearing the suitable amine linkers (Scheme 1). One of the attributes of developing preparative scale synthesis of architecturally complex natural products possessing significant bio-regulatory properties is the subsequent ready availability of advanced intermediates for SAR and photoaffinity labeling studies.
To permit facile formation of carbamate linkage, the secondary alcohol on (+)-discodermolide would be activated as a carbonate ester (cf. 20). The amine coupling partners were envisioned to arise via condensation reactions of the photoaffinity tags possessing carboxylic acid functionality with the diamine tethers of suitable length. The photoaffinity tags for this study were prepared following known synthetic routes,35,36,37 wherein acids 8 and 13 were activated as hydroxysuccinamide esters,38 followed by condensation with the alkyl diamine linkers (Scheme 2). For the aryl-(trifluoromethyl)diazirine component 18, the mono-Boc protected amine linker was employed with N,N′-Carbonyl-diimidazole (CDI)39 to reduce possible side reactions; the Boc group was subsequently removed under acidic conditions.
In the event, construction of (+)-discodermolide/paclitaxel hybrids 3–7 possessing the phototags entailed a two-step protocol, wherein the secondary alcohol in (+)-19 was converted to a p-nitrophenylcarbonate ester, followed by union with amines 10–12 and 15 and 18.40 Global deprotection under acidic conditions provided (+)-3–(+)-7 in good to excellent yields. Importantly, the synthetic route not only provides facile access to the designed (+)-discodermolide/paclitaxel hybrids possessing the photoaffinity tags, but was also designed to permit ready access to either 3H or 14C analogs for future labeling experiments.
The designed (+)-discodermolide/paclitaxel hybrids possessing the photoaffinity labels were tested for antiproliferative activity against human lung (A549) and breast (MCF-7) cancer cell lines. Discodermolide (−1) and Taxol (−2) served as controls. In accord with our design strategy, the analogs revealed improved activity in a trend changing with the tether length (Table 1).
Given the similarity of the analogs, the up to an eight-fold increase in potency is believed to be associated with an interaction with the proposed aromatic pocket, and not simply to an increase in hydrophobicity or other factors. Also of importance, (+)-discodermolide and each of the six analogs increased the amount of microtubule polymerization compared to the control that had no drug added (Fig. 6). Taken together these observations strongly suggest that all of the discodermolide molecules have the same mechanism of action. In summary, we have defined two possible binding modes for (+)-discodermolide in the paclitaxel binding pocket. To test the validity of the proposed modes, we designed and synthesized a series of hybrid molecules, wherein the aromatic ring of paclitaxel was attached to (+)-discodermolide, exploiting a tether of suitable length. Biological testing demonstrated that hybrid congeners possess increased potency compared to (+)-discodermolide. Given the availability of the already incorporated phototags, the discodermolide/paclitaxel hybrids hold the promise of defining the (+)-discodermolide binding site and mode of action, as well as determining whether the drug demonstrates selectivity for any of the 7 β-tubulin isotypes. Since certain isotypes, particularly βIII-tubulin, have been associated with paclitaxel resistance,41 an investigation of the interaction between drugs and specific tubulin isotypes becomes an important goal, and in particular definition of human tubulin isotypes that bind (+)-discodermolide. Such studies in conjunction with radiolabeling experiments are ongoing in our laboratories.
Cells lines were obtained from the American Type Culture Collection (ATCC) and maintained in RPMI medium supplemental with 10% fetal bovine serum. Cells (800 A549 or 2000 MCF-7) were added to each well of a 96 well plate. Increasing concentrations of the indicated drugs were added 18 h after plating. IC50 values, the concentration of drug that inhibits cell growth by 50%, were determined after 72 h or 96 h incubation at 37 °C for A549 or MCF-7 cells, respectively, using the SRB method (40). The two cell lines were fixed and stained at different times to take into account their different growth rates.
Bovine brain tubulin (2.5 μM) from Cytoskeleton Inc. in a buffer containing 0.1 M MES, 1 mM EGTA, 1 mM MgCl2 was incubated with 3 μM of the indicated drugs, in the presence of 2 mM GTP, at 37 °C for 30 min. Samples were centrifuged at 120,000 × g at 37 °C for 1 h and the pellet that contains the polymerized tubulin was dissolved in SDS sample buffer and analyzed by SDS-polyacrylamide gel electrophoresis. Proteins were then transferred to nitrocellulose and stained with Ponceau S.
Except as otherwise indicated, all reactions were run under an argon atmosphere in flame- or oven-dried glassware, and solvents were freshly distilled. The argon was deoxygenated and dried by passage through an OXICLEAR™ filter from Aldrich and Drierite tube, respectively. Diethyl ether (Et2O), tetrahydrofuran (THF) and dichloromethane (CH2Cl2) were purchased from Aldrich (HPLC purity) and further purified by Pure Solve™ PS-400. All other reagents were purchased from Aldrich or Acros and used as received. Reactions were monitored by thin layer chromatography (TLC) either with 0.25 mm Silicycle or 0.25 mm E. Merck (Kieselgel 60F254, Merck) pre-coated silica gel plates. Silica gel for flash chromatography (particle size 0.040–0.063 mm) was supplied by Silicycle or Sorbent. Yields refer to chromatographically and spectroscopically pure compounds unless otherwise noted. 1H and 13C spectra were recorded on a Bruker AMX-500 spectrometer. Chemical shifts are reported as δ values relative to internal chloroform (δ 7.26) or benzene (δ 7.15) for 1H and either chloroform (δ 77.0) or benzene (δ 128.0) for 13C. Infrared spectra were recorded on a Jasco FTIR-480plus spectrometer. Optical rotations were measured on a Jasco P-2000 polarimeter in the solvent indicated. High resolution mass spectra were measured at the University of Pennsylvania Mass Spectrometry Center on either a VG Micromass 70/70H or VG ZAB-E spectrometer. Analytical reverse-phased (Sunfire C18; 4.6mm 50mm, 5mL) high-performance liquid chromatography (HPLC) was performed with a Waters binary gradient module 2525 equipped with Waters 2996 PDA and Waters micromass ZQ. All final compounds were analyzed employing a linear gradient from 10% to 90% of acetonitrile in water over 8 min and a flow rate of 1 mL/min, and unless otherwise stated, the purity level was >95%.
Compound (+)-20 (80 mg, 0.0725 mmol) was dissolved in CH2Cl2 (1 mL) and MeOH (1 mL). N-(4-Azido-benzoyl)-1,2-ethanediamine (10) (45 mg, 0.218 mmol) in MeOH (1 mL) and Et3N (40 μL, 0.290 mmol) were added to the above solution. The resultant solution was stirred at room temperature for 6 days. Water (50 mL) was added to the reaction mixture and extracted with CH2Cl2 (3 × 50mL). The combined organic layers were washed with brine (50 mL), dried over MgSO4, filtered and concentrated. Flash chromatography (11 to 25% EtOAc in hexane) provided compound (+)-3a (75.1 mg, 89% yield) as a colorless amorphous solid. [α]22D = +36.0° (c = 0.65, CHCl3). IR (film, NaCl): 3336, 2957, 2929, 2857, 2123, 1718, 1653, 1604, 1539, 1500, 1254, 1044, 836, 775 cm−1. 1H NMR (500 MHz, CDCl3) δ: 7.81 (d, J = 8.3 Hz, 2H), 7.33 (brs, 1H), 7.03 (d, J = 8.4 Hz, 2H), 6.54 (ddd, J = 16.7, 10.6, 10.6 Hz, 1H), 5.88 (apparent t, J = 10.9 Hz, 1H), 5.35 - 5.13 (m, 5H), 5.11 (d, J = 10.2 Hz, 1H), 5.02 (d, J = 9.9 Hz, 1H), 4.81 (apparent t, J = 9.1 Hz, 1H), 4.74 (apparent t, J = 5.8 Hz, 1H), 4.59 (ABq, JAB = 6.9 Hz, ΔAB = 28.4 Hz, 2H), 4.50 (apparent t, J = 10.3 Hz, 1H), 3.63 (brs, 1H), 3.56 - 3.44 (m, 3H), 3.41 (apparent t, J = 4.0Hz, 1H), 3.39 - 3.30 (m, 1H), 3.34 (s, 3H), 3.06 (apparent t, J = 5.6 Hz, 1H), 2.99 - 2.91 (m, 1H), 2.75 - 2.66 (m, 1H), 2.60(qd, J = 7.5, 2.8 Hz, 1H), 2.55 - 2.47 (m, 1H), 2.06 (apparent t, J = 12.7 Hz, 1H), 1.91 - 1.76 (m, 3H), 1.75 - 1.66 (m, 2H), 1.63 - 1.54 (m, 1H), 1.57 (s, 3H), 1.23 (d, J = 7.6 Hz, 3H), 0.96 (d, J = 6.7 Hz, 3H), 0.95 - 0.84 (m, 12H), 0.91 (s, 9H), 0.87 (s, 9H), 0.85 (s, 9H), 0.72 (d, J = 6.7 Hz, 3H), 0.08 (s, 3H), 0.06 (s, 3H), 0.054 (s, 3H), 0.048 (s, 3H), 0.04 (s, 3H), 0.03 (s, 3H). 13C NMR (125 MHz, CDCl3) δ: 173.6, 166.7, 158.8, 143.4, 133.9, 133.4, 132.3, 132.1, 132.0, 131.1, 130.8, 130.0, 129.0, 119.1, 118.5, 97.5, 86.4, 79.5, 77.3, 77.0, 74.9, 64.9, 56.2, 44.2, 42.7, 42.5, 40.5, 37.9, 36.1, 35.8, 35.7, 34.6, 34.4, 34.3, 26.4, 26.1, 25.9, 23.3, 18.7, 18.3, 18.1, 17.7, 16.8, 16.7, 16.5, 14.4, 14.2, 10.2, −3.2, −3.5, −4.2, −4.3, −4.66, −4.68. High resolution mass spectrum (ESI+) m/z 1190.7408 [(M+Na)+; calcd for C62H109N5O10Si3Na: 1190.7380].
To a solution of compound (+)-3a (67 mg, 0.0573 mmol) in MeOH (5 mL) was added aqueous hydrochloric acid (4M, 4 mL) in 100 – 250 μL portions over 7 h at a rate which minimized precipitation (ca. 8 to 30 min intervals) and the sides of flask were rinsed with MeOH (1 mL). The reaction mixture was stirred at room temperature for 18 h, diluted with EtOAc (70 mL). The resulting solution was neutralized with NaHCO3 (1.35 g in H2O 18 mL) at 0 °C. Phosphate buffer pH7 (1M, 30 mL) and NaCl (17 g) were added and the aqueous layer was extracted with EtOAc (3 × 70 mL). The combined organic layers were washed with brine (100 mL), dried over MgSO4, filtered and concentrated. The residue was purified three times by flash chromatography (80% EtOAc in hexane or 3 – 6% MeOH/CH2Cl2) to afford compound (+)-3 (31.1 mg, 69% yield) as a colorless amorphous solid. [α]22D = +17.4° (c = 0.10, CHCl3). IR (film, NaCl): 3359, 2968, 2123, 1699, 1639, 1604, 1543, 1501, 1457, 1284, 1120, 1032, 731 cm−1. 1H NMR (500 MHz, CDCl3) δ: 7.81 (d, J = 8.5 Hz, 2H), 7.26 (br, 1H), 7.06 (d, J = 8.4 Hz, 2H), 6.55 (ddd, J = 16.6, 10.6, 10.6 Hz, 1H), 5.84 (apparent t, J = 11.0 Hz, 1H), 5.54 - 5.41 (m, 3H), 5.26 (t, J = 10.4 Hz, 1H), 5.17 - 5.05 (m, 3H), 4.76 - 4.66 (m, 2H), 4.61 (apparent t, J = 9.3 Hz, 1H), 3.70 (br m, 1H), 3.57 - 3.44 (m, 2H), 3.43 - 3.31 (m, 2H), 3.29 - 3.16 (m, 3H), 3.00 - 2.91 (m, 1H), 2.83 - 2.74 (m, 1H), 2.68 (qd, J = 7.3, 4.6 Hz, 1H), 2.63 - 2.51 (m, 2H), 2.40 - 2.20 (br, 2H), 2.00 - 1.74 (m, 6H), 1.69 - 1.62 (m, 1H), 1.58 (s, 3H), 1.28 (d, J = 7.3 Hz, 3H), 1.06 (d, J = 6.9 Hz, 3H), 1.01 (d, J = 6.8 Hz, 3H), 0.93 (apparent d, J = 6.8 Hz, 6H), 0.90 (d, J = 6.6 Hz, 3H), 0.82 (d, J = 6.5 Hz, 3H). 13C NMR (125 MHz, CDCl3) δ: 174.6, 167.2, 158.4, 143.6, 134.5, 133.9, 133.6, 133.0, 132.3, 130.6, 130.0, 129.6, 129.1, 119.2, 118.3, 79.4, 78.9, 77.4, 75.2, 73.1, 64.4, 43.3, 42.0, 41.2, 40.5, 37.4, 36.2, 36.0, 35.94, 35.91, 34.9, 33.0, 23.4, 18.8, 17.5, 16.7, 15.7, 14.2, 12.8, 9.3. High resolution mass spectrum (ESI+) m/z 804.4495 [(M+Na)+; calcd. for C42H63N5O9Na: 804.4523].
Support was provided by the National Institutes of Health through the National Institutes of General Medical Sciences and the National Cancer Institute through grants GM-29028 (ABS) and CA-077263 (SBH), and the Breast Cancer Research Foundation (SBH). We also thank Drs. G. Furst and R. Kohli of the Department of Chemistry NMR and MS Facilities for obtaining the NMR and high resolution mass spectra.
1ABBREVIATIONS USED: ESI-HMRS, electrospray ionization high-resolution mass spectrometry; TMS, tetramethylsilane; TLC, thin layer chromatography; PDB, Protein Data Bank; DISCON: Distribution of Solution Conformations; NCI, National Cancer Institute; SAR, Structure activity relationship; MD, Molecular dynamics; CDI, N,N′-Carbonyl-diimidazole.
1H and 13C NMR spectra, experimental methods and synthetic procedures of (+)-3-(+)-20; HPLC data of (+)-3-(+)-7 used in biological assays; NMR data utilized in solution conformation calculations; SAR data employed in computational studies. This material is available free of charge via the Internet at http://pubs.acs.org.