Preparation of the HIV Attachment Inhibitor BMS ... - ACS Publications

ACS2GO © 2019. ← → → ←. loading. To add this web app to the home screen open the browser option menu and tap on Add to homescreen...
2 downloads 0 Views 3MB Size
Article pubs.acs.org/OPRD

Preparation of the HIV Attachment Inhibitor BMS-663068. Part 7. Development of a Regioselective Ullmann−Goldberg−Buchwald Reaction

Downloaded via UNIV OF MASSACHUSETTS AMHERST on July 4, 2018 at 12:32:55 (UTC). See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

William P. Gallagher,*,† Maxime Soumeillant,*,† Ke Chen,† Richard J. Fox,† Yi Hsiao,† Brendan Mack,† Vidya Iyer,† Junying Fan,† Jason Zhu,† Gregory Beutner,† Steven M. Silverman,† Dayne D. Fanfair,† Andrew W. Glace,† Adam Freitag,† Jason Sweeney,† Yining Ji,‡ Donna G. Blackmond,‡ Martin D. Eastgate,† and David A. Conlon† †

Chemical & Synthetic Development, Bristol-Myers Squibb Company, One Squibb Drive, New Brunswick, New Jersey 08903-0191, United States ‡ The Scripps Research Institute, 50550 N. Torrey Pines Road, La Jolla, California, United States ABSTRACT: The discovery, development, and optimization of an Ullmann−Goldberg−Buchwald coupling reaction is described. This complex process represents a key transformation in the development of a commercially viable synthesis of the HIV attachment inhibitor prodrug BMS-663068. In this reaction, high regioselectivities were obtained for the coupling of a 1,2,4triazole and a 7-bromoazaindole, preparing BMS-626529, the antepenultimate in good yield and quality. Key challenges associated with developing commercially viable conditions for this copper-mediated coupling include achieving the desired level of regiochemical control, identifying robust isolation conditions and controlling residual copper levels in the isolated product.





INTRODUCTION We recently reported a novel synthesis1 of BMS-663068, a potent and selective HIV-attachment inhibitor.2 As described in the prior articles in this series, this new approach offered significant advantages across all metrics of efficiency and processability for the preparation of this challenging heterocycle.3 One of the key features of this new approach was the invention of a bromination method capable of effecting C2bromination of nitrogen containing heterocycles, a classically difficult process; in reference to the synthesis of BMS-663068, this allowed the functionalization of the parent azaindole core to prepare a C7-bromo-azaindole 1.4 The discovery of this transformation enabled the preparation of this important compound and the application of a subsequent Ullmann− Goldberg−Buchwald coupling to install the triazole, a problematic fragment of the API, BMS-663068. We hoped that installation of the triazole later in the sequence would simplify the synthesis of this compound, as the triazole was a source of several chemical liabilities throughout the prior approaches.5 In addition, the strategy required the regioselective installation of this heterocycle, which we hoped to achieve through a metal mediated process to deliver the desired N-1 regioisomer 3 in good yield and regioselectivity (Scheme 1). The enabling conditions developed during our initial route scouting demonstrated the viability of this strategy, specifically with respect to obtaining a desired level of regiochemical control during the triazole addition and the ability to prepare API meeting our quality criteria. While our initial process supported the selection of this revised synthetic strategy as the commercial synthesis, several challenges still remained.1 To deliver a transformation suitable for validation and commercial production, significant additional optimization, development, and process innovation were required, described herein. © 2017 American Chemical Society

RESULTS AND DISCUSSION As discussed in Part 1 of this series, the 3-methyl-1,2,4-triazole moiety had been installed via a high-temperature SNAr reaction.3 This process required harsh reaction conditions and produced a poor regiochemical outcome with respect to the three triazole nitrogens.1 Despite many attempts, the C7chloride azaindole coupling partner was resistant to the application of metal mediated processes, leaving thermal SNAr as the only viable option. Prior to the invention of the αbromination process,1 the corresponding bromide 1 had been difficult to prepare, inhibiting its exploration as a coupling partner.6 For some time, the formation of the bromide had been seen as a critical objective for the development of an alternate strategy for triazole incorporation; the intrinsically higher reactivity of bromides was anticipated to provide access to a broader array of metal mediated coupling processes.7 Thus, with the bromide 1 in-hand, a broad reactivity screen was conducted surveying a variety of metals, ligands, solvents, and bases. As expected, the initial results indicated that Cumediated Ullmann−Goldberg−Buchwald couplings offered the best opportunity for developing a successful regioselective transformation. Thus, screening focused on the use of known ligand architectures such as phenanthrolines,8 diketones,9 and diamines10 as Cu ligands. Our initial results indicated that diamine scaffolds (such as trans-N,N-1,2-dimethylcyclohexanediamine: DMCHDA) were the preferred ligands for this coupling. Additional rounds of evaluation revealed valuable Special Issue: From Invention to Commercial Process Definition: The Story of the HIV Attachment Inhibitor BMS-663068 Received: May 30, 2017 Published: August 9, 2017 1156

DOI: 10.1021/acs.oprd.7b00191 Org. Process Res. Dev. 2017, 21, 1156−1165

Organic Process Research & Development

Article

Scheme 1. Initial Ullmann−Goldberg−Buchwald Conditions

and 1.6 equiv of DMCHDA). Under these conditions we observed significantly improved regioselectivities versus the thermal SNAr (ca. > 10:1:0.5 vs 4:1:1).3 We hypothesized that the improved regioselectivity was due to steric interactions between the triazole and the ligand during complexation of the triazole to the metal center, prior to C−N bond-formation, favoring the N-1 isomer in 3 (Figure 1). The reaction proceeded to high conversion in a variety of solvents (such as acetonitrile, 4-methyl-2-pentanol, and DMAc) and bases. Interestingly, the nature of the cationic component of the base had a significant impact. A successful reaction was only observed utilizing potassium bases (t-BuOK, KOH, K3PO4, or K2CO3); Li or Na derived reagents provided 75 °C to initiate the reaction. Initial lab experiments showed a markedly slower reaction rate than those observed during the initial screening, along with observed reaction stalling. The stalled reactions could not be restarted by additional charges of CuI, ligand, nor additional trizaole 2. However, the addition of water (10 equiv) restored the expected reaction rate. Gratifyingly, we also found that the regioselectivity improved slightly (to ∼15:1) in the presence of water. Although the impact of water on the rate of Ullmann couplings has been documented elsewhere, the reason for this effect is unknown.13 Due to the impact of water on the reaction rate and selectivity, we opted to switch the base for the reaction from t-BuOK to a

40−45 wt % aqueous KOH solution, to ensure a consistent concentration of water was added to the reaction. With preliminary conditions in hand (Figure 1), we performed a “design of experiment” analysis on the reaction conditions, along with initiating detailed mechanistic studies (vide infra) to help us further improve the Ullmann process, resulting in the identification of several influencing factors (Figure 2). It is noteworthy from these results that the main parameter favoring both conversion and regioselectivity was the addition of water (10−30 equiv) across a relatively broad range of CuI concentration (15−30 mol %). Within this Cu range, a high ligand-to-copper ratio had a greater impact on conversion than the observed regioselectivity, as only marginal 3/4 regioselectivity decay was observed when the ligand to copper ratio was less than 6. However, water in excess of 20 equiv, and base in excess of 2.2 equiv, had a detrimental impact on the overall 1158

DOI: 10.1021/acs.oprd.7b00191 Org. Process Res. Dev. 2017, 21, 1156−1165

Organic Process Research & Development

Article

Scheme 3. Preparation of Lithium Salt 3b

metathesis process, converting the solution phase potassium salt to its lithium analogue. In practice, it was found that simply adding lithium salts directly to the K-salt 3a could affect the desired metathesis process inducing the crystallization of Li-3.14 Initially LiBr was utilized as the lithium source; however, this resulted in an uncontrolled precipitation producing sticky solids. While this issue could not be overcome by simple temperature cycles, the addition of alcohols (MeOH, EtOH, and IPA) as cosolvents improved the behavior of the slurry, and compound 3b could be isolated albeit in a modest 35% yield. While the majority of the reaction impurities purged well under these conditions, an oxidative dimer 8 increased by 4-fold vs the end-of-reaction assay. This liability indicated the need to remove copper from the reaction streams prior to isolation. Thus, in parallel to optimizing the crystallization process, we also focused on protocols to efficiently scavenge copper which we hoped would enhance the stability of the reaction stream during the workup and improve the overall robustness of the process. Treatment with aqueous base washes (KOH, K3PO4, and K2CO3) initially showed promise in reducing the copper concentration.15 Unfortunately, significant precipitation of 3a was observed with increasing scale, presumably due to the removal of MeCN into the aqueous phase. The rates of oxidative dimerization and undesired hydrolysis reactions were also found to increase, indicating that a more efficient copper scavenging process was required. We next investigated the application of a mixture of EDTA and KOH. The preliminary results confirmed the high efficiency of the EDTA/KOH treatment, demonstrating the potential to remove 95% of the charged copper during the first wash, and therefore afforded a significantly reduced formation of oxidative dimer 8 (Table 2). Accordingly, after treatment with the EDTA/KOH (45%) mixture, attempted metathesis with LiBr successfully provided a

reaction profile (formation of compounds 7 and 9, see Scheme 2). During the development of these improved reaction conditions, we also focused on understanding the formation of key process impurities, along with defining optimized workup and isolation protocols. The observed impurities could easily be separated into four main categories (Scheme 2): • Triazole regioisomers: N-2 regioisomer 4 and the N-4 regioisomer 5 were observed along with the desired N-1isomer 3. • Competitive C-7 displacement: DMCHDA demonstrated some reactivity and reacted with bromide 1, to produce the ligand adduct 6. Compound 6 was observed in relatively high levels (up to 3.7 RAP). Hydroxide coupled with the activated species, forming the 7-OH azaindole 7. • Oxidative dimerization: Copper promoted dimerization of compound 3 under aereal oxidation, to form dimer 8. This was observed when reaction streams containing both Cu/ligand and base were exposed to air and is a result of the acidity of the triazole methine proton. • Product hydrolysis: Hydrolysis of the benzoyl amide was observed under the basic reaction conditions; thus, the desbenzoyl impurity 9 was also formed. Under the conditions described above, the coupling consistently reached >99% conversion in less than 18 h. In initial experiments, compound 3a was observed to gradually precipitate upon cooling the reaction below 50 °C. The filtration of the precipitated solids led to a modest recovery of the potassium salt 3a (ca. 35−40%), albeit with high organic purity (99 LCAP, 81 wt % vs free acid). XRD analysis indicated that multiple forms of 3a were obtained by this procedure, potentially complicating the workup and isolation. As described in Part 1 of this series, isolation of the free acid 3 was not feasible owing to the long needle morphology that precluded the isolation by filtration.3 Indeed, identification of a crystalline lithium salt form (Li-3b; Scheme 3) had been a key component of the route interrogation strategy, driving the team to reinvestigate the formation of indole 3 as a viable intermediate.1 The initial discovery of the Li-salt 3b showed it had excellent physical properties that resulted in acceptable filtration rates during isolation. In addition, its low solubility in MeCN (99% conversion. MeCN (81 °C) fulfilled that criteria. Reaction temperatures >90 °C led to higher impurity levels. (13) Xie, X.; Chen, Y.; Ma, D. J. Am. Chem. Soc. 2006, 128, 16050. (14) The use of Li bases to facilitate the reaction failed due to the low solubility of the lithium salts of 2 and 3b. (15) The basic aqueous layers were easily separated from the MeCN layer and displayed intense green or blue colorations suggesting significant copper removal. (16) The solubility at the end point solvent composition was used for isolation: KI (>100 mg/mL), KBr (4.5 mg/mL), KCl (3.4 mg/mL). The high solubility of KI allowed the KI formed to remain in solution. With KBr and KCl, both cocrystallized with 3b. (17) Utilizing the APDTC, we developed a modified recrystallization process to remove the Cu to NMT (20 ppm) and provided the desired impurity profile. (18) Maketon, W.; Zenner, C. Z.; Ogden, K. L. Environ. Sci. Technol. 2008, 42, 2124−2129. (19) Wing, R. E.; Rayford, W. E. From Plating and Surface Finishing 1982, 69, 67−71. (20) Gallagher, W. P.; Vo, A. Org. Process Res. Dev. 2015, 19, 1369− 1373. (21) Literature reports have shown that dithiocarbamates form dimeric structures with Cu by X-ray analysis, see: Giovagnini, L.; Sitran, S.; Montopoli, M.; Caparrotta, L.; Corsini, M.; Rosani, C.; Zanello, P.; Dou, Q. P.; Fregona, D. Inorg. Chem. 2008, 47, 6336− 6343. (22) Ammonium pyrrolidinedithiocarbamate (APDTC) was selected for further development due to the quantity needed and the availability at the time. HPLC conditions were developed to analyze for residual APDTC in isolated materials. Although dithiocarbamate does alert for mutagenicity and other EHS relevant end points, the mutagenic potential (Ames) is compound specific. For example, dimethyldithiocarbamate is mutagenic in the Ames assay, but diethyldithiocarbamate is reported as nonmutagenic. We recently tested APDTC and found it to be Ames-negative; thus, it is not considered a genotoxic impurity (GTI). (23) One needs to use at least a 1 μm filter to remove all solids. In some instances, 0.45 μm was utilized due to convenience. (24) Impurity 9 had