|Home | About | Journals | Submit | Contact Us | Français|
An efficient synthetic pathway to the possible stereoisomers of skeletally diverse heterocyclic small molecules is presented. The change in shape brought about by different intramolecular cyclizations of diastereoisomeric amino propargylic alcohols is quantified using principal moment-of-inertia (PMI) shape analysis.
Human genetics and physiology are increasingly revealing the root causes of human disorders, yet many of the relevant targets and processes are considered difficult to modulate with small molecules.1,2 At least part of the difficulty is associated with the candidate small molecules that populate screening collections.(3) We describe an effort to address this shortcoming by applying a build/couple/pair strategy in organic synthesis.4,5 The synthetic pathway we describe focused on the goal of synthesizing stereoisomers of skeletally diverse compounds so that stereochemistry-based structure−activity relationships (SARs) can be derived directly from small-molecule screening. Small-molecule probe-development efforts are significantly enhanced when “hits” from screens can be prioritized and optimized using stereochemistry-based SARs, and we imagine that drug discovery efforts might similarly be facilitated in the future.
We envisioned that densely functionalized chiral amino propargylic alcohols 1a−d could be synthesized as a complete matrix of stereoisomers using simple coupling reactions and readily available building blocks. These coupling products undergo intramolecular cyclization reactions to generate diverse heterocyclic skeletons in 1−2 steps (Figure (Figure11).
We initiated our study with the preparation of densely functionalized amides 2a,b(6) starting from commercial (R)- and (S)-phenylalanine.(7) The 2-nitrobenzensulphonyl (Ns) group was chosen both to modulate the pKa of the attached nitrogen to facilitate subsequent allylation(8) and to introduce an additional chemical handle for the “pair phase” of the pathway.
In the couple phase, an alkynyl moiety was incorporated by triethylsilylacetylide addition to amides 2a,b, generating the corresponding α,β-acetylenic ketones 3a,b in good yield and without racemization, when temperature and reaction time were optimized (Scheme 1). Reactions that enable controlled access to all possible diastereomers are especially valuable in diversity-oriented synthesis.(9) After extensive experimentation, we determined that the carbonyl diastereotopic faces in (R)-3a and (S)-3b can be differentiated efficiently by the chirality of CBS-Me-oxazaborolidine,(10) overriding the influence of the adjacent nitrogen-substituted stereogenic center. The reduction of chiral ynone (R)-3a (98% ee) using the stoichiometric complex (R)-Me-CBS-oxazaborolidine/BH3THF in dry THF at −78 °C delivered the syn-amino propargylic alcohol (2R,3S)-4a in excellent yield and good diastereoselectivity (diasteroisomeric ratio dr = 91:9). When (R)-3a was instead reduced with (S)-Me-CBS-oxazaborolidine/BH3THF, the corresponding anti-amino propargylic alcohol (2R,3R)-4b was obtained in comparable yield and diastereoselectivity. Reduction of (S)-3b (96% ee) under the same reaction conditions provided (2S,3S)-4c and (2S,3R)-4d, respectively (Scheme 1). In contrast, reduction of (R)-3a with NaBH4 in MeOH afforded a 5:1 mixture of anti/syn diastereoisomeric alcohols. Desilylation with a 1:1 molar ratio of TBAF/AcOH efficiently delivered the key intermediates of the pathway 1a−d.(11) The diastereoisomeric ratio for the anti-products (2R,3R)-1b and (2S,3S)-1c was increased to 97:3 after a single crystallization.
We next explored a series of intramolecular cyclization reactions using the syn- and anti-diastereomers 1a,b(12) (based on symmetry, we expect that the same reactions would work analogously on enantiomers 1c,d, providing products in all stereochemical combinations). Skeletal diversification was achieved by reagent- and substrate-controlled site-selective activation of different pairs of functional groups, strategically placed on the linear template 1a (Scheme 2). We first investigated enyne metathesis to access a substituted vinyl tetrahydropyridine-type skeleton, which is poised for further diversification. We determined that syn-amino propargylic alcohol (2R,3S)-1a in the presence of second generation Hoveyda−Grubbs catalyst (5 mol %) under ethylene atmosphere undergoes endo-mode selective cyclization forming the corresponding cycloheptadiene 5a as the major regioisomer (endo:exo selectivity = 10:1.5).(13) Noting the size of the newly formed ring and the presence of nonsubstituted alkenyl and alkynyl groups in the substrate,(14) we determined that this unusual result was influenced by the free hydroxyl group at the propargylic position. Protection of the alcohol as acetate 6a resulted in exo-mode ring closure to form six-membered ring diene 7a as the only product. We speculate that in the latter case the endo-mode RCM is disfavored as a result of the formation of a sterically hindered alkylidene, compared to the exo-mode RCM intermediate. Under the same reaction conditions, the anti-diastereomer (2R,3R)-1b yielded a mixture of decomposition products, whereas the corresponding propargylic acetate 6b delivered 7b in moderate yield as a single stereoisomer (Figure (Figure22).(15) Gratifyingly, we found that catalytic InCl3 (15 mol %) promoted skeletal reorganization(16) of both syn- and anti-diastereomers 1a,b under microwave conditions, providing the corresponding dienes 8a,b as single products in good yield. We investigated the Pauson−Khand reaction (PKR) as an additional alkyne-alkene functional group-pairing reaction. After conversion of amino propargylic alcohol 1a to acetate 6a, the Co2(CO)6 complex derived from 6a was treated with N-methylmorpholine N-oxide(17) at room temperature to provide cyclopenten-2-enone derivative 9a with excellent stereoselectivity. Similarly, diastereomer 6b formed PKR product 9b stereoselectively. The stereochemistry of 9a and 9b was determined by NOE analyses.(12) The benzyl substituent influenced the diastereofacial control of cyclization, whereas the acetoxy-substituted stereogenic center provided virtually no directing effect.
We next examined polar/polar functional group-pairing reactions for further skeletal diversification (Scheme 2). Treatment of 1a with tetrabutylammonium fluoride (TBAF) in THF at 0 °C provided the Smiles rearrangement(18) product 10a in 50% yield. In contrast, anti-diastereomer 1b rearranged to 10b with high site- selectivity in presence of diaza-(1,3)-bicyclo[5.4.0]undecane (DBU) in refluxing THF. When 1a was treated with NaH in THF, the resultant alkoxide participated in endocyclic nucleophilic aromatic substitution(19) to yield the seven-membered benzothiazepine derivative 11a. Control of the temperature between −10 and 0 °C was needed to obtain 11a selectively in 75% yield. In contrast, anti-diastereomer (2R,3R)-1b only reacted at room temperature under these conditions, generating (3R,4R)-11b in moderate yield due to lower selectivity of the ipso-substitution relative to the Smiles rearrangement.(20)
To access wider skeletal diversity, we sought to take advantage of the nonpolar chemical handles still available on the benzothiazepine core 11a. An enyne metathesis reaction yielded cross-metathesis product 12a in the presence of ethylene gas. Intramolecular enyne metathesis product 13a was obtained selectively in excellent yield by simply treating the substrate with first-generation Grubbs catalyst (3 mol %) in dichloromethane at room temperature followed by addition of lead tetraacetate (15 mol %). The latter was added to scavenge ruthenium and phosphine impurities.(21) The structure of the bridged ten-membered ring diene was confirmed by X-ray analysis.(12) Diastereoemer 11b only reacted in the presence of ethylene gas providing the cross-metathesis product 12b as the major adduct and cyclic diene 13b in combined 58% yield. Skeletal diversification via PKR on 11a afforded bridged ten-membered ring cyclopenten-2-enone derivative 14a in moderate yield as a single stereoisomer. X-ray structure analysis confirmed the stereochemistry of this product.(12) Diastereoemer 11b did not undergo the PKR.
We sought to quantify the extent to which the aforementioned intramolecular cyclizations access different overall molecular shapes. To obtain a simple assessment of diversity in molecular shape among the compounds studied, we calculated normalized principal moment-of-inertia (PMI) ratios,(22) plotted as two-dimensional characteristic coordinates (Ismall/Ilarge, Imedium/Ilarge) for compounds 1a and 5a−14a (Figure (Figure3A).3A). Points in the resulting graph occupy an isosceles triangle defined by the vertices (0,1), (0.5,0.5), and (1,1), corresponding to the shapes of rod, disk, and sphere, respectively.
For 1a and 5a−14a, we observe a band covering the space between the rod- and disk-like regions of the PMI space; 1a is centrally located in this band, and its direct reaction products, 5−6a, 8a, and 10−11a, span nearly the range of rod- and disk-like shapes among all 11 compounds. However, further cyclization of 6a to produce 7a and 9a results in the two most disk-like of all 11 compounds, though each is similar in shape to its precursor 6a. Further reaction of 11a to produce 12−14a similarly results in the three most sphere-like among the 11 compounds. In contrast to cyclization of 6a, these modifications of 11a result in relatively large overall changes in shape.
We also used PMI ratio analysis to quantify the extent to which diastereomeric products 1b and 6b−13b differ from their counterparts 1a and 6a−13a (Figure (Figure3B).3B). We quantified the PMI distance between matched pairs and observed the largest such differences for compounds 12a,b (da−b = 0.364), 11a,b (da−b = 0.211), and 10a,b (da−b = 0.116). Interestingly, for 11 and 12, the a diastereomer is more disk-like, which may be due to extended conformations of acyl and allyl groups perpendicular to the rod-like structure of the molecule, whereas b is more rod-like; the converse is true for 10, where an opposite relationship is observed. The minimum difference we observed was for the pair of amino propargylic alcohols 1a,b (da−b = 0.008), which makes sense due to the short, linear hydroxyl substituent at one of the stereocenters. A similar trend was observed for other compounds with small substituents, e.g., in the case of 13a,b, where only the hydrogen atom at the bridge position is inverted.
To compare additional compounds prospectively available from this chemistry to compounds from other sources (Figure (Figure3C),3C), we enumerated a virtual library of compounds that would be obtained by substituting starting phenylalanines with alternative amino acids to give 6 additional R-groups (bromobenzyl, hydroxymethyl, o-TBS-ethyl, p-hydroxybenzyl, p-azidobenzyl, hydroxyethyl) at carbon 2. We compared the resulting 140-compound library to 140 diverse compounds selected(23) from each of two sources: commercial compounds (CC) typical of those found in many screening collections and natural products (NP).(24) We note that CC compounds cover the left edge of the PMI space, taking shapes intermediate between rods and discs, whereas NP compounds distribute more evenly over the space and populate the intermediate region between disk- and sphere-like shapes. By comparison, the virtual products from this chemistry also provide more broad coverage of the space and include several products among the most sphere-like of all compounds compared.
We initiated this research with the hypothesis that accessing stereoisomers of skeletally diverse compounds will facilitate small-molecule probe and drug discovery efforts. This new DOS pathway will enable further testing of this hypothesis. This work also provides a framework for understanding the consequences of different synthetic decisions in the “pair phase” of build/couple/pair synthesis strategies using PMI shape analysis to understand changes in shape brought about by different intramolecular cyclizations. The analysis of small-molecule screening results using these compounds should enable us to link shape analyses and synthetic decisions to biological assay outcomes in the future.
The NIGMS-sponsored Center of Excellence in Chemical Methodology and Library Development (P50-GM069721) enabled this research. We are grateful to Dr. Peter Muller at MIT for X-ray crystallographic analysis, and Emeline Tissot and Christopher Johnson at the Broad Institute for help with SFC/MS. D.P. thanks Yikai Wang and Dr. Giovanni Muncipinto at the Department of Chemistry and Chemical Biology (Harvard University), Dr. Manuela Rodriquez currently at the Department of Pharmacological Sciences (University of Salerno), and Dr. Masaaki Hirano currently at Astellas Pharma Inc. for helpful discussions. S.L.S. is an investigator with the Howard Hughes Medical Institute.
National Institutes of Health, United States
Experimental procedures and full spectroscopic data for all new compounds. This material is available free of charge via the Internet at http://pubs.acs.org.