Development of an Effective Scalable Enantioselective Synthesis of the HIV-1 Entry Inhibitor BNM-III-170 as the Bis-Trifluoroacetate Salt

Abstract

We report here the development and optimization of a process synthesis for the HIV-1 entry inhibitor BNM-III-170 bis-TFA salt (1). The synthesis features a dynamic-kinetic resolution (DKR) to establish the initial stereogenicity. By taking advantage of significant sequence modifications of our first generation synthesis, inconjunction with the low solubility of late-stage intermediates, the overall efficiency of the synthesis has been significantly improved, now to proceed in an overall yield of 9.64% for the 16-steps, requiring only a single chromatographic separation.

Graphical Abstract

INTRODUCTION

The AIDS pandemic derives directly from the HIV-1 virus. Currently approximately 36.9 million people in the world are living with AIDS, with an estimate of 2 million new infections reported each year.¹ Although therapeutic combination regimens (i.e., antiretroviral therapy or ART) have proven successful to treat, but not cure HIV-infected patients, the prevention of HIV transmission and/or the development of eradication protocols remain major challenges.

Allosteric activation of the HIV-1 virion surface proteins comprises an obvious tactic to disrupt the HIV-1 entry process.² That is, the binding and subsequent opening of the HIV-1 envelope trimer glycoprotein (gpl20-gp41)³ by a small molecule CD4 mimetic compound prior to virus encounter with the host cell receptor (CD4) holds the promise of epitope exposure leading to immune response and in turn elimination via otherwise rarely-elicited broadly neutralizing antibodies (bNAbs).³

Recently we reported the synthesis of BNM-III-170 bis-trifluoroacetate salt (1),⁴ employing rational structure-based design, in conjunction with X-ray and computational analysis, based on two lead compounds (2 and 3) identified by screening by Debnath el al. (Figure 1).⁵ Compound 1 exhibits low micromolar entry antagonism against a number of HIV-1 strains, and was shown to mimic the host cell CD4 receptor primarily by virtue of binding to the Phe43 cavity on gp120.^2,6 The promising bioactivity of 1, has now resulted in the protection of gp120-immunized non-human primates for up to six months from multiple high-dose intrarectal challenges with a simian-human immunodeficiency virus (SHIV).⁷ Equally important, collaborations have demonstrated that 1 can also sensitize HIV-infected cells toward endogenous antibody neutralization in the sera of the AIDS infected patients.⁸

Figure 1.

BNM-III-170 bis-TFA salt (1), NBD-556 (2), and NBD-557 (3).

Our first-generation synthesis of BNM-III-170 bis-TFA salt (1), disclosed in 2016,⁴ proved capable of producing a few hundred milligrams in 6.2% yield over a 15-step sequence. The assembly of 1 was built on the indane core 4 (Figure 2), employing two building blocks, acid 5 and tri-Boc-guanidine 6.⁹ We began the synthesis with installation of the (methylamino)methyl side chain, onto commercially available 5-bromoindanone 4, via first generation of ethylene glycol ketal 7 in what proved to be a sluggish acid-catalyzed ketalization (Scheme 1). The practicality of this reaction on a large scale was exacerbated by the use of benzene as solvent, in conjunction with the difficulty to achieve full conversion, thereby necessitating a tedious chromatographic separation due to the similar polarities of 4 and 7. Use of toluene instead triggered an undesired aldol condensation of 4, as the corresponding toluene-water azeotrope boiled at a considerably higher temperature. Other nonpolar solvents were inefficient for solubilizing 4. Moving forward, formylation of the aryl bromide in 7, followed by reductive animation of the resulting aldehyde, Boc-protection and ketal removal proceeded smoothly to furnish Boc-carbamate 8. Next, a Claisen condensation led to a racemic mixture of β-ketoester 9 by utilizing 8 with NaHMDS and Mander’s reagent.¹⁰ This was then subjected to our cornerstone construction tactic, a dynamic-kinetic resolution (DKR) under transfer hydrogenation conditions,¹¹ to furnish β-hydroxyester (−)-10 with both excellent yield (91%) and enantioselectivity (er 99:1). With the stereogenicity now defined, we next generated amine (+)-11 via a four-step sequence involving reduction of the ester with LiAlH₄, selective protection of the resulting primary hydroxyl group as a TBS ether, azide formation with complete inversion of stereochemistry,¹² and reduction of the azide via palladium-catalyzed hydrogenolysis. The overall yield for this four-step sequence was 76% yield. The 4-chloro-3-fluoro aromatic moiety was then installed via a TBTU-mediated amide union with known acid 5.¹³ Following silyl group removal with TBAF, Mitsunobu union⁹ of alcohol (+)-13 with urethane 6 gave rise to (+)-14, the fully Boc-masked precursor of BNM-III-170 (1). Upon TFA-mediated deprotection, the small molecule CD4mc (+)-1 was isolated as the bis-TFA salt. The overall sequence proceeded in 15-steps and 6.2% yield, requiring 12 chromatographic separations in addition to an HPLC-purification of (+)-1 after global Boc-deprotection.

Figure 2.

Strategy for the synthesis of BNM-III-170 (1).

Scheme 1.

The First-Generation Synthesis of (+)-1.

Given the growing interest in the bioactivity profile of (+)-1 as a small molecule CD4mc, generated by a large number of collaborations,⁷ a multigram supply of (+)-1 was clearly required. Equally important, an efficient modular synthesis of (+)-1, in conjunction with the ready availability of late-stage intermediates, would expedite the design and synthesis of potentially more potent analogs. Toward this end, we describe here an enantioselective process scale synthesis, based on our first generation synthesis of BNM-III-170 (Scheme 1) that now is capable of delivering (+)-1 on decagram scale.

RESULTS AND DISCUSSION

Development of a process synthesis of BNM-III-170 as the TFA salt began with elimination of the initial undesirable ketalization protocol (Scheme 1). In addition, we decided to retain the aryl bromide motif in 4 throughout the initial series of transformations (Scheme 2), envisioning a mild carbon-carbon bond forming event on a late-stage intermediate (vide infra). Toward this end, 4 was first subjected to a much less expensive and environmentally benign Claisen condensation, employing NaH as the base (2.0 equiv.) and diethyl carbonate (3 equiv.) as the acylating agentTo attenuate the potential rapid evolution of hydrogen gas over the course of this reaction, 4 was first introduced in a controlled manner into an ice-cold suspension of NaH in THF, and excess diethyl carbonate, prior to warming the reaction mixture to 70 °C, and then aging the reaction mixture for 2 hours to achieve complete conversion. The desired sodium salt of 15, as well as the excess NaH, were then carefully neutralized with diluted hydrochloric acid (3 N) at 0 °C. Upon liquid-liquid extraction with ethyl acetate and subsequent removal of the volatiles, the resulting brown solid 15 as the keto and enol tautomer (6:1) was employed without further purification for our previously reported (Scheme 1) dynamic-kinetic resolution (DKR) to establish the cis-rclationship between the hydroxyl group and the ester group of (−)-16. We were pleased to discover that reducing both the catalyst loading to 0.5 mol% and the amount of formic acid-triethylamine azeotrope did not affect the yield (92% yield) or enantiopurity of the resulting β-hydroxyester (−)-16 (er > 99:1). Reduction of (−)-16 with LiAlH₄ was then followed by a facile Fieser&Fieser workup¹⁴ to furnish diol (−)-17, which could be purified by precipitation from a mixture of ethyl acetate and hexanes (v/v 1:20). We noted that it was crucial to use diethyl ether as opposed to THF as solvent for this reduction in order for the Fieser&Fieser work-up to produce a granular precipitate.

Scheme 2.

Synthesis of the Oxalamide 21

With (−)-17 in hand, we employed the protocols used in our first-generation synthesis, namely a chemoselective silylation followed by a stereospecific azidation, to generate azide (+)-18. Given that the previously employed palladium-catalyzed hydrogenolysis of the azide functional group in (+)-18 would not be viable in the presence of the aryl bromide, we opted for a Staudinger reduction,¹⁵ which was achieved by heating a solution of (+)-18 with PPh₃ (1.5 equiv.) and water (10 equiv.). The silyl protecting group was next conveniently cleaved with HCl in 1:1 (v/v) MeOH-water (pH 0). Following evaporation of the majority of MeOH, the resulting aqueous solution of the hydrochloride salt (19) was then free-based with a 20 wt% NaOH solution to precipitate aminoalcohol (−)-20, as an off-white powder (65% yield over 4 steps). Utilizing this reaction sequence, we were able to eliminate chromatography over the 7-step sequence from 4 to (−)-20 (Scheme 2), as all of the hydrophobic impurities could be removed via acidic workup following the silyl group removal. With (−)-20 now readily available, we proceeded with amide union employing acid 5.¹³ Pleasingly the EDC-mediated coupling reaction in DMF smoothly produced oxalamide (+)-21. Upon completion of this reaction, the reaction mixture was diluted with water, and the resulting precipitate, mostly (+)-21 was collected as a wet solid. This material however proved extremely difficult to dry. In order to obtain pure (+)-21, the wet solid was first dissolved in THF and diethyl ether (60 mL and 30 mL respectively per gram of (−)-20 at 40 °C). The aqueous layer was then removed, and the solvent was concentrated. The residue was further purified by trituration from a mixture of CH₂Cl₂ and hexanes (v/v 1:1) to furnish (+)-21 in 91% yield as an off-white powder.

Although we were pleased to have devised a robust and scalable route to (+)-21, the Mitsunobu reaction to attach the guanidine building block (6) proved quite problematic (Scheme 3). Under our first generation conditions,⁶ condensation on a decagram scale of alcohol (+)-21 with the highly insoluble urethane 6, although yielding the desired adduct (+)-22, led to a significant amount of the 2:1 adduct (+)-23 (ca. 15% by weight). Unfortunately, employing a larger excess of 6 was not effective in diminishing the formation of (+)-23. This result did not come as a complete surprise, as the more soluble (+)-22 could compete as an effective Mitsunobu nucleophile to give (+)-23.

Scheme 3.

Formation of (+)-22 and byproduct (+)-23

Despite the difficulty in obtaining (+)-22 with sufficient purity on multigram scale without employing a chromatographic separation, we moved forward with the synthesis to evaluate several transformations we had envisioned in the design stage of the scale-up campaign. Gratifyingly, the proposed Suzuki-Miyaura reaction with vinyltrifluoroborate (24) successfully forged the desired C-C bond to produce (+)-25 in 85% yield without further purification (Scheme 4). Lemieux-Johnson oxidation of (+)-25 then led to the aldehyde (+)-26 in 88% yield, which was directly converted to the amine (+)-27 with methylammonium chloride and NaBH(OAc)₃ under reductive animation conditions. We were however unsuccessful in obtaining (+)-27 with sufficient purity either by precipitation or chromatography. Instead, (+)-27 was directly subjected to 50 equivalents of TFA for global Boc-group removal to furnish (+)-1 in 68% yield over 2-steps, albeit requiring tedious reverse-phase HPLC purification, followed by lyophilization to obtain pure material. Unfortunately, we encountered numerous synthetic challenges during the implementation of these last four steps. In addition, it was undesirable to incorporate heavy metal-mediated transformations within the late portion of the sequence (22 to 26). We therefore deemed this generation of synthesis unsuitable for our purpose.

Scheme 4.

Endgame for the Synthesis of (+)-1

At this juncture, a careful review of our first-generation synthesis, in conjunction with our scale-up modifications, led to a number of issues that would require addressing to achieve a successful high-throughput process synthesis. These were: (A) the need for an efficient approach to attach the requisite (methylamino)methyl group; (B) an alternative strategy to install the guanidine motif without formation of the bis-adduct, and (C) given the poor solubility of (+)-1 in many common organic solvents, a protocol to achieve purification by precipitation. We thus embarked on a revised second generation process synthesis that would not only prove scalable, but also that would not require labor-intensive chromatographic separations.

Further analysis suggested that the 3 early steps (4 to (−)-17, Scheme 2) in our second-generation route would prove suitable for large-scale production. We also recognized, over the course of our optimization studies, that incorporation of the oxalamide moiety invariably significantly decreased the solubility of most intermediates (Scheme 4), and as such their solubility might be capitalized on for isolation/purification. Finally, we wished to exploit a more widely-employed/robust protocol to introduce the guanidine unit at a late stage, although such a tactic might increase the length of the process by one step compared to the first-generation synthesis.

Toward this end, diol (−)-17 (Scheme 5) was selectively protected as the mono-silyl ether with one equivalent of TBSC1 as outlined previously in Scheme 2. Following aqueous washes and trituration, (−)-28 was directly converted to the aldehyde (−)-29 via lithiation followed by capturing the resulting intermediate with DMF. The requisite (methylamino)methyl side chain was then installed via a reductive animation protocol comprising condensation of (−)-29 with aqueous methylamine followed by reduction with NaBH₄. The reaction mixture was then quenched with sat. aqueous NaHCO₃ and extracted with CH₂Cl₂. The resulting secondary amine (−)-30 could then be isolated upon rotary evaporation of the CH₂Cl₂ and immediately protected to furnish the Boc-carbamate (−)-31. The secondary hydroxyl group in (−)-31 was next converted to the corresponding azide (+)-32. Importantly, this reaction proceeded equally well both employing a lower loading of DPPA and DBU (i.e., 1.2 equiv. each, versus 3 equiv. in the first-generation synthesis) and in a shorter reaction time (4 hours instead of overnight). The resultant reaction mixture was then treated with water and extracted with diethyl ether, and the solvent removed by rotary evaporation to give (+)-32 as a red oil. Without further purification, we were then able to carry out a highly efficient homogeneous nickel chloride-catalyzed reduction of the azide (+)-32. This reaction was complete within 4 hours in the presence of 10 mol% NiCl₂•6H₂O, as opposed to 24 hours with 10 mol% Pd/C as employed in the first generation approach.

Scheme 5.

A New Synthesis of Mesylate (+)-34

With amine (+)-11 in hand, we set out to forge the amide linkage via a more atom-economical tactic employing aminolysis of the known ester 33,¹³ compared to the use of the corresponding acid 5 employed in our first-generation synthesis. We rationalized that such a process would be possible due to the greater electrophilicity of ester 33. Indeed, use of K₂CO₃ in conjunction with 3-nitrophenol cleanly promoted the desired aminolysis at 65 °C in 48 h. Presumably, this union proceeded via the intermediacy of the more reactive aryl ester generated in situ by transesterification of 33 with the phenoxide.¹⁶ Importantly, ¹H NMR analysis indicated complete consumption of (+)-11. Excess 33 was then hydrolyzed employing aqueous K₂CO₃. The resulting potassium salt of 5 was removed by filtration, with the 3-nitrophenol concomitantly removed by the alkaline aqueous wash. Next, the silyl group on amide (+)-12 was removed via addition of a solution of HCl in 3:1:1 (v/v/v) THF-MeOH-water (pH 1) via stirring overnight. The resulting mixture was then diluted with aqueous NaHCO₃, and the precipitated product collected as a damp solid. At this juncture, by taking advantage of the low solubility of (+)-13, purification proved straight forward via an ethereal wash to furnish (+)-13 as an off-white powder. The overall yield for these last 8-steps was 35%. The primary hydroxyl in (+)-13 was then readily derivatized with methanesulfonyl chloride in a 1:4 (v/v) DMF-THF mixture to furnish mesylate (+)-34, which could be easily purified by addition of water to the reaction product, followed by vacuum filtration of the resultant (+)-34 as an off-white solid in 92% yield, thereby setting up the endgame for the process synthesis.

Toward that end, mesylate (+)-34 (Scheme 6) was first converted to the phthalimide (+)-35 via a Gabriel amine synthesis¹⁷ employing potassium phthalimide to introduce the nitrogen atom. The primary amine in (+)-36 was then readily revealed by treatment with hydrazine hydrate (5 equiv.) in THF-ethanol (v/v 1:1) at 50 °C. Pleasingly, the phthalhydrazide byproduct of this hydrolysis protocol dissolves in 5 wt% aqueous Na₂CO₃, whereas the product amine (+)-36 nicely precipitates and could be collected as a pale yellow solid. The requisite guanidine functionality was then introduced by employing commercially available N,N’-Di-Boc-1H-pyrazole-1-carboxamidinc (37) by stirring a mixture of (+)-36 and 37 in DMF overnight. Although 37 was used in a slight excess, the excess amount was readily removed upon addition of a sacrificial polyamine (i.e., ethylenediamine), followed by extraction with 5 wt% aqueous NaHSO₄ and ethyl acetate used as the organic medium. The penultimate product (+)-38 was then obtained following passage through a short plug of silica gel, and solvent removal, to afford an off-white solid in 55% yield. Dropwise addition of a solution of TFA (40 equiv.) in CH₂Cl₂ to an ice-cold solution of (+)-38 in CH₂Cl₂ then led to clean global deprotection, with the TFA salt of (+)-1 readily precipitating in 89% yield from MeOH-ethyl acetate-hexanes (1:4:4) as a white powder.

Scheme 6.

End-Game of the Synthesis of (+)-1

CONCLUSION

In summary, we have achieved an effective enantioselective process scale synthesis of BNM-III-170 as the bis-TFA salt (1), employing 16-steps that proceeded in 9.64% overall yield (Scheme 5 & 6). The scalable process synthesis versus the first generation synthesis pleasingly obviated a tedious end-game HPLC purification, with flash chromatographic separations avoided in all but one step: a final filtration through a plug of silica. Stereogenicity of the BNM-III-170 bis-TFA salt (+)-1 was established via a ruthenium-catalyzed transfer hydrogenation by virtue of a DKR, and the aromatic side chain was installed via alkyllithium-mediated formylation, followed by a reductive amination process. The oxalamide moiety was next installed by an aminolysis protocol, and the primary amine necessary for guanidine introduction was attached via a Gabriel amine synthesis. Pleasingly, the overall process has now been demonstrated on a 20-gram scale to prepare the 6.17 grams of BNM-III-170 bis-trifluoroacetate salt (+)-1.

Supplementary Material

SI

¹H NMR and ¹³C NMR spectra (PDF).