The Evolution of Chemical High-Throughput Experimentation To

Initial forays into the use of HTE in our laboratories for solving chemistry problems ..... of four substrates, six bases, and four temperatures in on...
0 downloads 0 Views 4MB Size
Article pubs.acs.org/accounts

The Evolution of Chemical High-Throughput Experimentation To Address Challenging Problems in Pharmaceutical Synthesis Shane W. Krska,*,‡ Daniel A. DiRocco,† Spencer D. Dreher,‡ and Michael Shevlin† †

Process Research & Development, Merck Sharp & Dohme Corporation, Rahway, New Jersey 07065, United States Chemistry Capabilities and Screening, Merck Sharp & Dohme Corporation, Kenilworth, New Jersey 07033, United States



CONSPECTUS: The structural complexity of pharmaceuticals presents a significant challenge to modern catalysis. Many published methods that work well on simple substrates often fail when attempts are made to apply them to complex drug intermediates. The use of high-throughput experimentation (HTE) techniques offers a means to overcome this fundamental challenge by facilitating the rational exploration of large arrays of catalysts and reaction conditions in a timeand material-efficient manner. Initial forays into the use of HTE in our laboratories for solving chemistry problems centered around screening of chiral precious-metal catalysts for homogeneous asymmetric hydrogenation. The success of these early efforts in developing efficient catalytic steps for late-stage development programs motivated the desire to increase the scope of this approach to encompass other high-value catalytic chemistries. Doing so, however, required significant advances in reactor and workflow design and automation to enable the effective assembly and agitation of arrays of heterogeneous reaction mixtures and retention of volatile solvents under a wide range of temperatures. Associated innovations in high-throughput analytical chemistry techniques greatly increased the efficiency and reliability of these methods. These evolved HTE techniques have been utilized extensively to develop highly innovative catalysis solutions to the most challenging problems in large-scale pharmaceutical synthesis. Starting with Pd- and Cu-catalyzed crosscoupling chemistry, subsequent efforts expanded to other valuable modern synthetic transformations such as chiral phase-transfer catalysis, photoredox catalysis, and C−H functionalization. As our experience and confidence in HTE techniques matured, we envisioned their application beyond problems in process chemistry to address the needs of medicinal chemists. Here the problem of reaction generality is felt most acutely, and HTE approaches should prove broadly enabling. However, the quantities of both time and starting materials available for chemistry troubleshooting in this space generally are severely limited. Adapting to these needs led us to invest in smaller predefined arrays of transformation-specific screening “kits” and push the boundaries of miniaturization in chemistry screening, culminating in the development of “nanoscale” reaction screening carried out in 1536well plates. Grappling with the problem of generality also inspired the exploration of cheminformatics-driven HTE approaches such as the Chemistry Informer Libraries. These next-generation HTE methods promise to empower chemists to run orders of magnitude more experiments and enable “big data” informatics approaches to reaction design and troubleshooting. With these advances, HTE is poised to revolutionize how chemists across both industry and academia discover new synthetic methods, develop them into tools of broad utility, and apply them to problems of practical significance.



INTRODUCTION

commercially available equipment and workflows. Moving from hydrogenation into cross-coupling and other valuable modern chemical methodologies required the creation and validation of higher-complexity workflows. In this Account, we describe the evolution of HTE in our laboratories along a number of dimensions: (1) from simple manual procedures to complex, increasingly automated protocols; (2) from specialized tools used solely by catalysis experts to commoditized kits deployed in every chemistry lab; and (3) from large-scale process chemistry applications to enabling medicinal chemistry. Over the past decade, these collective advances have empowered MSD chemists to bring the diversity of valuable catalytic trans-

Modern catalysis holds tremendous potential for increasing the productivity and cost-effectiveness of pharmaceutical R&D by enabling efficient access to complex molecular architectures. Recognizing this, Merck Sharp & Dohme Corporation (MSD) formed the Catalysis Laboratory in 2002 to develop tools and strategies for applying modern catalysis to drug development projects. We chose to focus initially on implementing asymmetric hydrogenation in the large-scale synthesis of compounds in late clinical development. These early pioneering efforts, summarized in a 2007 Account,1 revealed that high-throughput experimentation (HTE) chemistry tools were ideally suited for solving extremely challenging catalysis problems in fast-moving development projects. With their relatively simple reaction setup, homogeneous hydrogenations proved to be an easy entry point into HTE, requiring relatively little engineering to adapt © 2017 American Chemical Society

Received: September 18, 2017 Published: November 27, 2017 2976

DOI: 10.1021/acs.accounts.7b00428 Acc. Chem. Res. 2017, 50, 2976−2985

Article

Accounts of Chemical Research

With these engineering improvements in hand, we validated microscale HTE cross-coupling using literature test reactions to establish excellent reactivity and reproducibility. While our HTE workflow employed automated liquid handling for preparation of ligand libraries and automated solid handling for dosing of inorganic bases, it is important to note that automating the entire cross-coupling process was not practical. We found that manual addition of catalyst precursors and substrate solutions using single- and multichannel pipettes conferred flexibility to the experimental setup; in contrast to automated tools, manual pipettes are inexpensive and easy to use and require minimal training and maintenance. Our initial achiral phosphine libraries, which did not group ligands by structure, proved to be inefficient for screening because of the dramatic reactivity differences observed across structurally diverse ligands. We subsequently redesigned the ligand libraries such that each collection contained functionally similar ligands (e.g., monodentate vs bidentate or electron-rich vs electron-poor ligands) in order to allow a more rational approach to screening of ligand diversity. As the number of ligand libraries grew, we built reaction-specific focused libraries to address common transformations such as Suzuki−Miyaura couplings or Buchwald−Hartwig C−N couplings. Accompanying these libraries were standardized experimental designs consisting of the subsets of catalysts, bases, and solvents found to give optimal performance from key literature reports and internal experience.3 Using these refined approaches, we developed numerous applications of palladium-catalyzed cross-coupling chemistries to drug development projects, including a large-scale Kumada coupling in support of the development of vaniprevir4 and a onepot double borylation/Suzuki coupling for the industrial-scale synthesis of elbasvir.5 Similar approaches were used to develop an efficient, regioselective copper-catalyzed indazole coupling for the synthesis of niraparib.6 Scheme 1 illustrates two examples of the power of HTE for solving complex cross-coupling chemistry problems. With no

formations to bear on the most impactful and challenging problems in pharmaceutical R&D.2



EXPANDING HTE APPLICATIONS TO CROSS-COUPLING Our early research in asymmetric hydrogenation used HTE to rapidly survey the ever-expanding array of commercially available chiral phosphine ligands in combination with various noblemetal precursors.1 Typical screens involved treating predispensed 96-vial ligand libraries in 8 mm × 30 mm glass vial inserts with solutions of Ru, Rh, or Ir precursors to form complexes in situ. After addition of substrate solutions, the reactors were sealed in pressure vessels, pressurized with hydrogen gas, and heated to the desired temperature. Despite the open headspace within the pressure vessels, solvent loss was minimal because of the gastight seal. After many successful applications of this approach in asymmetric hydrogenation, we recognized the potential for these tools to impact many other areas of catalysis. However, doing so would present many new engineering challenges. Our first forays outside of hydrogenation involved Pd- and Cucatalyzed cross-coupling, which required dosing of insoluble inorganic bases into well plates, effective agitation of heterogeneous reaction mixtures, and retention of volatile solvents often heated at or above their boiling points (Figure 1). As with asymmetric hydrogenation, we employed predis-

Scheme 1. Complex Cross-Couplings Enabled by HTE Figure 1. Comparison of microscale HTE for hydrogenation and crosscoupling.

pensed phosphine ligands. Metal precursors were added as stock solutions to form complexes in situ. We initially added inorganic bases to the reaction vials as uniform slurries in a volatile solvent, but this tedious process required large volumes of solvent and introduced a time-consuming evaporation step. We solved this problem with a custom-built Symyx Powdernium MtM soliddispensing robot, which was used to predose arrays of commonly used bases in 2−10 mg quantities. Cross-coupling partners were then added to the reaction vials as solutions to form the complete reaction mixtures. In contrast to our hydrogenation protocols, cross-coupling experiments required direct sealing of individual reaction vials. After evaluation of a large number of different polymeric sealing mats for solvent retention and chemical compatibility, we chose PFA, a perfluoroalkoxyalkane polymer. With this sealing material, 100 μL reactions in tetrahydrofuran could be heated to 100 °C overnight with negligible solvent loss. We found that magnetic tumble stirring provided strong and uniform agitation of heterogeneous cross-coupling reaction screens, in contrast to the irreproducible, nonuniform agitation observed with typical radial magnetic stir plates.

literature precedent for a regioselective Suzuki−Miyaura coupling reaction of dichlorobenzene derivative 1, iterative application of multiple rounds of HTE enabled the identification of ligand 2 in conjunction with a very specific combination of Pd precursor, solvent, and base as being uniquely effective in giving a high yield of the desired product 3 while avoiding the formation 2977

DOI: 10.1021/acs.accounts.7b00428 Acc. Chem. Res. 2017, 50, 2976−2985

Article

Accounts of Chemical Research of regioisomeric, bis-coupled, or dehalogenated byproducts.7 In another example, late-stage introduction of two alkyl side chains to dihalide 4 via sequential sp2−sp3 Suzuki−Miyaura couplings to develop an efficient multikilogram synthesis of NPCL1 inhibitor 7 also required multiple rounds of HTE, examining protecting groups, ligands, Pd sources, solvents, and bases.8 While phosphine ligand libraries provided solutions for many projects, they required a discrete ligand−metal complexation step every time they were used and were limited by the inherent reactivity of the Pd(OAc)2 or Pd2(dba)3 complexes of the ligands. In 2008, palladium precatalysts based upon cyclopalladated 2-phenylethylamine were published by the Buchwald laboratory.9 These compounds provided rapid and quantitative formation of reactive monoligated palladium catalysts upon treatment with base. While these catalysts provided advantageous reactivity and convenience compared with traditional in situ catalyst formation, they were available commercially with only a limited set of ligands and were impractical for large-scale processing because of their synthesis from unstable (TMEDA)PdMe2. These limitations were addressed by the discovery of the analogous cyclopalladated 2-amino-1,1′-biphenyl precatalysts, which could be formed by direct cyclometalation of 2-amino1,1′-biphenyl with Pd(OAc)2 and required weaker bases and lower temperatures for catalyst activation.10 We collaborated with the Buchwald laboratory to further develop these promising precatalysts, resulting in an improved process for large-scale synthesis and broader commercial availability with a wide variety of phosphine ligands.11 These precatalysts were critical for reducing the cost of a key coupling for the industrial-scale synthesis of anacetrapib (Scheme 2).12 Early syntheses of intermediate 11 using 2 mol

Chiral phase-transfer catalysis (PTC) has been extensively investigated as a method for large-scale asymmetric synthesis since its first reported industrial application in 1984.14 Given the promise of this method, a custom library of cinchonine- and cinchonidine-derived phase-transfer catalysts was synthesized, as no commercial sources existed. Because of the biphasic nature of PTC, the results of catalyst screening under these conditions were found to be extremely sensitive to both agitation rate and temperature. This platform was validated for use in standard 96well plates agitated with tumble stirring and subsequently used to develop numerous asymmetric PTC reactions in support of latestage drug development. Examples include an asymmetric cyclopropanation of (E)-N-phenylmethyleneglycine ethyl ester15 en route to a class of macrocyclic HCV protease inhibitors and the development of the chiral-PTC-catalyzed formation of 3,3-spiro-7-azaoxindoles for the synthesis of MK8825 and MK-3207.16 Of particular importance was the serendipitous discovery during the latter study that bisquaternized (bis-quat) PTC impurities led to more active and selective catalysts. These results provided an impetus for the development of subsequent bis-quat PTC libraries that found unique utility in an asymmetric aza-Michael reaction for the construction of the chiral cyclic urea moiety in letermovir.13 While our chiral phosphine libraries were established for asymmetric hydrogenation and continue to be a workhorse in this respect, they have proven to be invaluable in the development of several additional breakthrough catalytic asymmetric methods. The HCV NS5a inhibitors elbasvir and ruzasvir contain a chiral hemiaminal core that was previously inaccessible in enantioenriched form by direct catalytic methods. An unprecedented dynamic kinetic C−N bond formation was evaluated as a possible solution. After extensive experimentation, it was discovered that the combination of QuinoxP*/Pd(OAc)2 was uniquely effective at catalyzing the desired C−N bond formation with high enantioselectivity (Scheme 3).17,18 Further

Scheme 2. Synthesis of Anacetrapib Intermediate 11 Using a Biaryl Palladium Precatalyst

Scheme 3. Dynamic Kinetic Asymmetric C−N Coupling To Give a Chiral Hemiaminal Core

% Pd(OAc)2 and 4 mol % Cy3P had reproducibility problems in the catalyst formation step, resulting in a capricious 45−80% product yield. Subsequent catalyst development identified (dtbpf)PdCl2, which provided reliably high yields at 0.5 mol % catalyst loading but contributed significantly to the cost of intermediate 11. In contrast, biaryl precatalyst 10 provided the desired product in 98% yield and contributed an order of magnitude less to the cost of intermediate 11 as a result of the greatly reduced cost of Cy3P compared with dtbpf. The high activity of these biaryl precatalysts was also key in the development of an efficient Heck coupling for the manufacturing route to letermovir.13

analysis of this system identified the bis(phosphine) mono(oxide) complex as the active species, formed in situ from the bis(phosphine) and Pd(OAc)2. These results underline the importance of parallel experimentation, as examination of large arrays of ligands and metal sources often results in key reaction insights that are unlikely to be discovered by smaller numbers of experiments. Over the past decade, photoredox catalysis has revolutionized the way synthetic chemists design new bond-forming reactions.19,20 Modification of our existing reactor design for uniform distribution of light was accomplished by irradiation of individual vials with separate light-emitting diodes (LEDs) arrayed on a circuit board21−23 Using this reactor in conjunction with a comprehensive photocatalyst library has enabled the development of numerous photocatalyzed methods. In one of the first applications of this technique, we developed a late-stage direct alkylation of heterocycles that took advantage of iridium-based excited-state reductants to decompose peracetates to alkyl



DEVELOPING A BROAD PALETTE OF HTE-ENABLED CATALYTIC CHEMISTRIES The success of evolving our HTE protocols to encompass crosscoupling methods emboldened us to further expand this approach to accommodate the rapidly changing landscape of catalysis. Doing so required investments in new technologies and expanded diversity of our catalyst/ligand collections. 2978

DOI: 10.1021/acs.accounts.7b00428 Acc. Chem. Res. 2017, 50, 2976−2985

Article

Accounts of Chemical Research radicals under mild conditions (Figure 2).23 For the first time, this method allowed the introduction of simple methyl groups

Figure 4. Photoredox-catalyzed indoline oxidation.

reaction temperatures and/or times. Figure 5 illustrates one such case, the direct metalation of (hetero)arenes with Knochel− Hauser bases. We developed a fully automated screening protocol using the Chemspeed robotics platform that allows the evaluation of four substrates, six bases, and four temperatures in one experimental run.27 Investigating a complete base/ temperature matrix for each substrate allowed for rapid identification of the optimal conditions for maximum yield and regioselectivity.

Figure 2. Direct methylation of heterocycles via photoredox catalysis.

into complex heterocycles at room temperature (vide infra). This concept was further extended to introduce hydroxymethyl groups into heterocycles using benzoylperoxide as the oxidant and methanol as the hydroxymethyl radical source.24 HTE has also been used to develop photoredox-catalyzed processes for late-stage pharmaceutical development. In collaboration with the Britton group at Simon Fraser University, we developed a one-step direct fluorination of leucine catalyzed by sodium decatungstate to give λ-fluoroleucine, a valuable building block in the synthesis of odanacatib (Figure 3).25



IMPORTANCE OF HIGH-THROUGHPUT ANALYSIS The ability to run increasingly sophisticated catalytic chemistries in a high-throughput format required a commensurate level of innovation in developing fast analytical methods that provide a meaningful, reliable assay of reaction outcomes with rapid data turnaround to inform the next round of experiments. In this regard, the presence of a dedicated group of analytical chemists at MSD who were simultaneously pushing the boundaries of highthroughput analysis was foundational to the success of the entire HTE initiative. Early efforts focused on automated, multiplexed approaches for fast HPLC method development for reliable quantitation of starting material, desired product, and reaction byproducts.28 In addition to HPLC, supercritical fluid chromatography (SFC) emerged as a powerful method for fast, efficient determination of enantiomeric enrichment in asymmetric catalytic reactions. With optimized semiautomated workflows, a chiral SFC method could usually be developed and deployed within 1 day of receipt of a racemic product standard.29 As the size of HTE experimental arrays continued to increase, the demand for even higher throughput analytical methods drove further innovations. One such example was the multiple injections in a single experimental run (MISER) protocol, which utilized very short columns and closely spaced stacked injections along with mass spectroscopic separation of starting material from product ions to give a rapid first-pass analysis of HTE experiments.30 This method proved to be particularly valuable for triaging of large HTE arrays, filtering out wells that did not produce significant conversion to product and flagging others that were of interest for in-depth chromatographic analysis. As powerful as these new high-throughput analytical methods were, the vast volume of information-rich chromatographic data they generated produced a severe data analysis bottleneck. After

Figure 3. Direct fluorination of leucine via photoredox catalysis.

Because of the requirement of 365 nm light to activate the decatungstate photocatalyst, UV LEDs were incorporated into our parallel reactor. More recently, a unique photocatalyst/ oxidant combination was discovered that allowed selective and highly stereoretentive oxidation of an indoline hemiaminal to the corresponding indole in the context of the synthesis of elbasvir (Figure 4).26 While many of the platforms discussed thus far rely primarily on semiautomated optimization in 96-well arrays, in certain cases we have found great utility in fully automated reaction optimization, especially in those cases that require screening of 2979

DOI: 10.1021/acs.accounts.7b00428 Acc. Chem. Res. 2017, 50, 2976−2985

Article

Accounts of Chemical Research

Figure 5. Automated regioselective deprotonations with Knochel−Hauser bases. Reproduced with permission from ref 27. Copyright 2017 Royal Society of Chemistry.

interrogate the biological system under study. Here synthetic target molecules are always changing in response to data gleaned in the previous design cycle, so synthetic routes and individual steps must be general and flexible to accommodate the many variations in substrate and reaction partners. Many synthetic methods, especially modern catalytic transformations, were developed using very simple model substrates and do not perform well on complex pharmaceutical intermediates.32 If not addressed, this disconnect between the types of substrates amenable to many catalytic methods (generally low-molecular-weight, lipophilic hydrocarbons with few functional groups) and the ideal substrates for drug synthesis (complex N-heterocycles with multiple hydrogen-bond-donor and -acceptor functional groups) can create a synthetic bias in medicinal chemistry programs toward making molecules with less drug-like properties, ultimately leading to lower-quality clinical candidates.33,34 HTE holds great promise to remove this synthetic bias by enabling rapid identification of conditions to give successful synthesis of all designed target molecules regardless of complexity. In order for HTE to prove valuable in this context, two basic criteria must be met: (1) the setup of the experiment must be fast and resource-efficient to match the rapid time cycles in the drug discovery space, and (2) the experiments must be material-sparing since only limited quantities of medicinal chemistry intermediates are typically available and any material consumed by a reaction screen cannot be used to generate new analogues. On the basis of this understanding, our implementation of HTE to solve problems in medicinal chemistry evolved toward the concept of reaction-specific kits: 24-vial arrays with predispensed reagents and simple, recipe-like protocols that captured the current best catalysts and conditions from published literature and our internal experience (Figure 7). Kits for common catalytic transformations, including Suzuki−Miyaura, Buchwald−Hartwig, Heck, and Sonogashira couplings, were developed and validated in-house and mass-produced in our centralized HTE facility.35 A network of HTE-trained chemists across the company assisted medicinal chemists in setting up these kits to solve their individual chemistry problems. While early generations of these kits employed 8 mm × 30 mm vials and

experimenting with a number of commercially available software packages, we elected to employ Virscidian Analytical Studio, which allowed HTE practitioners the ability to interact directly with large arrays of LC−MS chromatograms and mass spectral data, applying corrections to peak integrations, peak assignments, and other parameters in real time with facile export of processed data into spreadsheets and other data analysis tools for more complex manipulations.31



EVOLVING HTE FOR DRUG DISCOVERY While HTE efforts at MSD had their genesis in the complex problem-solving environment of process chemistry, it soon became evident that there was an even larger body of opportunities to apply HTE in the realm of medicinal chemistry. Figure 6 compares and contrasts the needs of these two vital

Figure 6. Comparison of complementary catalysis needs for process and medicinal chemistry.

areas of pharmaceutical chemistry research. For process chemistry, the goal is to make one target molecule, the active pharmaceutical ingredient, in the most efficient, robust, economical, and environmentally responsible manner possible. Here HTE plays a critical role in enabling key catalytic transformations that lead to the best overall synthesis. In contrast, in medicinal chemistry the goal is to provide rapid and reliable access to diverse chemical matter that serves to 2980

DOI: 10.1021/acs.accounts.7b00428 Acc. Chem. Res. 2017, 50, 2976−2985

Article

Accounts of Chemical Research

in the CSG has grown considerably. Nowhere has this trend been more evident than in the fast-growing field of C−H functionalization chemistry.36 In medicinal chemistry, application of these new methods for diversification of advanced lead structures has become known as late-stage functionalization (LSF).37 Over the past 5 years, we have made significant investments in LSF, leveraging our expertise in catalysis, HTE, analytical chemistry, and separations science to bring these emerging technologies to bear on active medicinal chemistry programs. As the utilization of HTE in medical chemistry has taken root and chemists have become comfortable with the concept of examining arrays of reaction conditions with catalysis kits, this “HTE mindset” has led to creative applications such as the example shown in Figure 8.38 Here the medicinal chemistry team faced a synthetic bottleneck due to a poorly performing nucleophilic aromatic substitution reaction. In order to find robust reaction conditions for this key disconnection that would be broadly applicable across the desired chemical space to be explored, they adopted a “parallel in parallel” HTE screening approach in which a series of reaction conditions were arrayed against different electrophile−nucleophile pairs. This approach proved to be highly effective, with a dramatic increase in reaction success rate and a corresponding acceleration in the program leading to the identification of molecules with vastly improved properties.

Figure 7. Reaction-specific HTE kits for medicinal chemistry applications.

consumed around 100 mg of substrate, later generations of kits were migrated to the 4 mm × 21 mm vial format, which consumed 4−5 times less material (vide infra). To build on the success of the kits effort, we developed complementary tools to aid medicinal chemists in reaction development. The most widely utilized was the Catalyst Selection Guide (CSG), an extensive document whose purpose was twofold: (1) to suggest the best conditions that could be run as singleton experiments for common types of catalytic transformations, along with tips and tricks for reaction troubleshooting, and (2) to pair this knowledge with the awareness of what kits were available for these reaction classes when the suggested singleton conditions failed. Dog-eared copies of the CSG, now in its eighth edition, can still be found on desks and benchtops across the company. Over the past decade, as established catalytic methods have continued to evolve and new methods have been introduced into practice, the number of screening kits and the associated content



MINIATURIZATION From our very first experiences with HTE for medicinal chemistry applications, it became obvious that our experiments were too material-intensive. To conserve material, the kits approach reduced the number of reactions per screen to 24;

Figure 8. Parallel HTE screen of coupling partners and conditions for an SNAr reaction for library synthesis and impact on medicinal chemistry productivity. 2981

DOI: 10.1021/acs.accounts.7b00428 Acc. Chem. Res. 2017, 50, 2976−2985

Article

Accounts of Chemical Research

Figure 9. 1536-well-plate nanomolar-scale screening of cross-coupling chemistry.

Figure 10. Comparison of photoredox-catalyzed C−N coupling conditions with a Chemistry Informer Library.

however, 8 mm × 30 mm vials still required ∼3−4 mg of substrate per well. To help with this problem, we built a robust workflow using 4 mm × 21 mm “microvials” that could generate reproducible chemistry results with approximately one-fourth the material requirement and translated all of our kits to this format. Because of their smaller size, microvial experiments required significantly more practice and dexterity than the larger

vials, and we recognized that we had reached the practical size limit for manual reaction setup. Given that biology and biochemistry experiments are routinely run at much smaller (