Recent Progress on One-Pot Multisubstrate Screening

Traditionally, new synthetic reactions have been developed using a model substrate to screen reaction conditions before testing the optimized conditions with a range of more complex substrates. In 1998, Gao and Kagan pooled multiple substrates in one pot to study the generality of an enantioselective method. Although such one-pot multisubstrate screenings may be powerful, few applications have appeared in the literature. With the advancement of various chromatography techniques, it may be time to revisit this underutilized platform. This review article discusses the applications of one-pot multisubstrate screenings as a method for developing new synthetic methods.

ABSTRACT: Traditionally, new synthetic reactions have been developed using a model substrate to screen reaction conditions before testing the optimized conditions with a range of more complex substrates. In 1998, Gao and Kagan pooled multiple substrates in one pot to study the generality of an enantioselective method. Although such one-pot multisubstrate screenings may be powerful, few applications have appeared in the literature. With the advancement of various chromatography techniques, it may be time to revisit this underutilized platform. This review article discusses the applications of one-pot multisubstrate screenings as a method for developing new synthetic methods. KEYWORDS: high-throughput screening, enantioselectivity, diastereoselectivity, gas chromatography, liquid chromatography, mass spectrometry S ynthetic methodologies require high efficiency, reproducibility, and generality. Research from the initial discovery of a new reaction to its improvement to meet these criteria demands numerous experiments. Historically, synthetic technologies were optimized through repetitive trial and error in a one-at-atime fashion. To expedite this process, high-throughput experimentation (HTE) has become a viable alternative to the traditional approach (Figure 1). The first parallel HTE was reported in 1996 by the Burgess group in a 96-well format. 1 Today, multiwells (96, 384, or 1536 wells) are used as a platform for HTE. 1−12 Recent work using HTE has provided significant chemical insights and led to more ideal synthetic methodologies. However, HTE requires specialized and expensive infrastructure, and when using 384 or 1536 wells, it is problematic to use volatile solvents (due to solvent evaporation at microliter volumes) or heterogeneous conditions (e.g., Pd/C, alkali metal, etc. due to difficulty in measuring out nanomolar amounts of solvent reproducibly and accurately). Additionally, the analysis of hundreds to thousands of reaction wells remains a bottleneck in HTE even if ultraperformance liquid chromatography (UPLC) is available.
An alternative to HTE was developed in 1998 by Gao and Kagan, which they termed "one-pot multi-substrate screening (OPMSS)." The premise of OPMSS is that multiple parallel reactions in one pot generate a set of data upon analysis of the mixture. This produces many data points with a relatively small number of experiments. Indeed, they stated that "the most usef ul approach in the fast screening of a new asymmetric catalyst is to get a set of information f rom one experiment" and that "the easiness and simplicity of this method make it very convenient for a preliminary evaluation of a chiral reagent or catalyst." 13 Gao and Kagan proposed that OPMSS could be a viable tool to determine the enantioselectivity of reactions. They demonstrated this by applying OPMSS to the asymmetric reduction of a set of ketones using a Corey catalyst. They combined multiple ketones in the same flask, reduced them, and then analyzed the reaction mixture using chiral HPLC to find the enantiomeric excess (ee) of each alcohol product. They compared the ee of each alcohol when the reduction was performed as a mixture of ketones to the ee when each ketone was reduced individually to ensure no interference or induction by the presence of other substrates. The verification that there is no interference between substrates is essential ( Figure 2). As Kagan wrote in a 2005 review article, "the information extracted f rom the one-pot multi-substrate screening will be valid only if there are no interactions between the products and the catalysts or reagents." 14 They reduced each ketone as part of different substrate sets to ensure that different types of substrates were orthogonal with respect to the reduction enantioselectivity. For the first set of ketones, one run of chiral HPLC was able to resolve substrates 1a−d and products 2a−d (Scheme 1). However, for the other, larger sets of ketones, direct HPLC analysis was impossible because of peak overlap, and additional flash column chromatography was necessary to separate the compounds into three fractions, which were then analyzed by chiral HPLC. 13 After Gao and Kagan introduced OPMSS as a feasible method, multiple groups used the platform to develop stereoselective asymmetric catalysis methodologies. Satyanarayana and Kagan reviewed the applications of OPMSS in 2005. 14 This review will briefly discuss the papers already summarized in the 2005 review article and mainly focus on examples of OPMSS published after their review.
In 1998, the Gennari group also used the same approach and combined four substrates to screen chiral catalysts for the nucleophilic addition of diethylzinc to aldehydes 3a−d to form alcohols 4a−d (Scheme 2). 15 However, they did not show that the one-pot multisubstrate reaction did not suffer from interference between substrates. The enantiomeric ratios were determined by chiral gas chromatography (GC).
In 2000, the Liskamp group prepared a library of chiral ligands on a solid support and mixed four substrates in a reaction mixture. They performed a similar addition of diethylzinc to aldehydes 3a, 3b, 3d, and 3e to form alcohols 4a, 4b, 4d, and 4e (Scheme 3). Chiral GC was used to analyze the mixtures of products. A gas chromatogram ( Figure 3) indicates the separation might not have been proactively considered, although they mentioned that GC retention times were important  to verify that data obtained from one-pot multisubstrate reactions are the same as in single-substrate reactions. This is done by performing single-substrate reactions and comparing the results to that of multisubstrate reactions. If the results are consistent, then OPMSS is a valid method to apply to the system being investigated. (d) Two forms of interference that can cause one-pot multisubstrate data to differ from single-substrate reactions.     16 Testing substrates individually to validate their experiments was not reported. In 2001, the Piarulli and Gennari group used the same screening platform to screen different chiral ligands for the conjugate addition of diethylzinc to nitroalkenes 5a and 5b to form 6a and 6b (Scheme 4). 17 The potential interference was not studied.
In 2002, the Wolf group screened enantioselective catalysts for the nucleophilic addition of an ethyl group to aldehydes. 18 This is the same transformation as the Liskamp system. The authors mixed three substrates, 3a, 3b, and 3f, to form products (R)-4a, (R)-4b, and (R)-4f (Scheme 5) and used chiral GC for analysis ( Figure 4).
In 2002, the Feringa group developed an enantioselective conjugate addition of diethylzinc to nitroalkenes. Nine aromatic   substrates 5a−i were pooled into one pot to screen for optimal catalysts to form 6a−i (Scheme 6). 19 Two criteria were articulated: (1) the product peaks should not overlap in the chromatogram and (2) the different substrates and products should not interfere with each other during the reaction. With two optimal catalysts, four aliphatic substrates were tested together to study the generality of the two catalysts. In 2004, Goddard and Reymond pooled 20 substrates to determine the substrate scope for enzyme-catalyzed hydrolysis of esters 7 to carboxylic acid 8 and 1,2-diols 9 (Scheme 7). 20 They used reverse-phase HPLC with different wavelengths to analyze the data and determine reactivity ( Figure 5).
In 2005, the Pfaltz group applied OPMSS to the iridiumcatalyzed enantioselective hydrogenation of alkenes. They hydrogenated four terminal alkenes 10a−d in one pot using an iridium catalyst and H 2 bubbled into the flask (Scheme 8).
They used chiral GC to analyze the reaction mixture, thereby successfully resolving all enantiomers of products 11a−d ( Figure 6). They also demonstrated no interference by other substrates, as they measured the ee values of the products in single-substrate reactions, which corroborated the values in the multisubstrate experiment. 21 In 2005, the Feringa group employed OPMSS for the enantioselective hydrogenation of acyclic enamides 12 to          Organic Process Research & Development pubs.acs.org/OPRD Review systems were previously unknown to display this reactivity. It also shows that when optimizing a reaction using a "model substrate," it is possible to miss specific substrates that require different conditions. 23 In 2008, the Laschat group reported the enantioselective reduction of aryl ketones using α-pinene-derived aminoalcohols and BH 3 (Scheme 11). Chiral GC was used to analyze a mixture of four aryl ketones 14a−c and 14h and the reduced products 15a−c and 15h (Figure 9). They found that only one of their aminoalcohols, 16, provided good enantioselectivity and that adding trimethyl borate increased the ee. Individual experiments were not reported to confirm the lack of interference. 24 In 2008, the Collin and Zouioueche groups described the enantioselective reduction of aryl ketones in water using a ruthenium catalyst. Chiral GC was used to analyze a mixture of the reduction of six aryl ketones, 14a and 14h−l, and their products, 15a and 15h−l (Figure 10a). As a control experiment, they reduced 14l as a single substrate and in the presence of five other ketones. The control experiment showed that the presence of other substrates increased the ee from 55% to 68%, thereby indicating that there was induction by the other substrates. Nevertheless, they performed a multisubstrate screening of the six ketones with eight different ligands and then a different set of six ketones, 14a, 14e, 14h, 14j, 14l, and 14m, with 12 ligands to produce products 15a, 15e, 15h, 15j, 15l, and 15m ( Figure  10b). Ligand 17 was identified to produce the highest ee values for ortho-substituted phenyl ketones, while all other ketones garnered better enantioselectivity with ligand 18 (Scheme 12). They then tested eight different ketones in single-substrate reactions to compare the ee using 17 and 18 and showed the trend that provided superior ee values. 25 In 2010, the Fiaud group reported an enzyme-catalyzed kinetic resolution of racemic secondary alcohols through acylation. They examined the resolution of eight racemic alcohols rac-19a−h catalyzed by the lipase enzyme CAL-B ( Figure 11). They reacted a racemic mixture of the alcohol with    1). This is the second example of applying OPMSS to enzyme catalysis. 26 In 2011, the Collin and Zouioueche groups reported an OPMSS for reducing aliphatic ketones using chiral GC ( Figure  12). They mixed seven ketones 20a−g and reduced them to the corresponding alcohols 21a−g using a ruthenium catalyst in water (Scheme 13). They demonstrated that with ligand 17, the chiral alcohol products of the seven ketones had very similar ee values in single-and multisubstrate screenings. They then reduced those seven ketones with 11 different ligands and found that "enantioselectivity depends both on the nature of ligand and substrate." The results of the multisubstrate screen were applicable to other ketones with similar properties to those in the OPMSS. 27 In 2015, the McIndoe group published their study of the relative reactivity of aryl iodides under copper-free Sonogashira coupling conditions. They pooled six aryl iodides 22a−f and acetylene 24 to form coupled products 23a−f (Scheme 14) and analyzed the kinetics by pressurized sample infusion electrospray ionization mass spectrometry (PSI-ESI-MS). This method allowed for real-time monitoring of the intensities of different mass peaks to determine the reaction kinetics. The reaction was monitored until 90% consumption of the phenylacetylene, and the linear kinetic regime of each product formation was used to determine the rate. This was then used to create a Hammett plot, and ρ was 1.4, which indicated the favorability of the parasubstituted electron-withdrawing groups. Because the purpose of this application of OPMSS was to investigate kinetics, only one catalyst system was used. 28 In 2016, the Flitsch and Barran group reported the biocatalytic PSL lipase enzyme amidation of esters reactions using five substrates in one pot. They used direct infusion ion mobility mass spectrometry (IM-MS) to analyze the products. They used a "twin peak" method involving heavy-isotope labeling of exactly half of each substrate. This resulted in each substrate presenting as one IM peak with two easily recognizable mass peaks, thereby making the identification of a substrate in a complex mixture more facile (Figure 13). Two OPMSS experiments were performed: one screened five esters 25a−e with PSL and piperidine 26a (1:1 unlabeled/deuteriumlabeled), which formed amides 27aa−ea, and the other screened five amines 26a−e with ester 25a (1:1 unlabeled/deuteriumlabeled), which formed amides 27aa−ae (Scheme 15). They successfully resolved and identified the substrate peaks. Using 27aa as a positive control, 27ca and 27ea were primarily formed in the ester screen, and in the amine screen, 27ab and 27ae were the major products. 29 In 2016, the O'Neil group disclosed the use of GC × GC to separate 27 compounds (9 substrates and a pair of enantiomers from the 9 substrates after enantioselective reduction). They used this method to test the reduction of nine ketone substrates 14e, 14i, 14j, and 14m−r, investigated by the Li and Gao group, 30 to their corresponding alcohols 15e, 15i, 15j, and 15m−r in a one-pot fashion (Scheme 16). The ee values were quite similar, and the % conversions followed the same trend between substrates even though they were not always so close to the previously reported single-substrate trials. 31 This is probably the first example of using GC × GC in OPMSS ( Figure 14). The GC × GC technology has excellent potential in OPMSS; however, it has not been used in organic synthesis, possibly because of the scarcity of equipment.
In 2019, the List group wrote, "the multi-substrate screening approach has not previously led to a general and broadly applicable catalyst." In light of this, they set out to develop a general catalyst for the enantioselective Diels−Alder reaction between α,βunsaturated aldehydes with cyclopentadiene. Six enals 28a−f were mixed with cyclopentadiene in one pot to form cyclohexenes 29a−f (Scheme 17), and chiral GC was used to analyze the reaction mixture ( Figure 15). They found that imidodiphosphorimidate (IDPi) catalysts provided good reactivity and enantioselectivity, so they used OPMSS to examine multiple IDPi catalysts. Catalysts 30a and 30b worked best and provided excellent conversion and enantioselectivity for all enals. They also found that when single-substrate reactions were performed with the pooled enals, the reactivities and enantioselectivities were similar to those of the multisubstrate screens. They then applied this optimized method to

Organic Process Research & Development
pubs.acs.org/OPRD Review many other substrates and observed excellent reactivity and stereoselectivity. 32 In 2021, the List group reported a new type of reaction: adding silyl enol ethers to silyl nitronates enantioselectively. Using a model-substrate approach, they quickly identified IDPi catalyst 30d as suitable for coupling silyl nitronate 31b with silyl ketene acetal 33. However, when 30d was used with different substrates, the enantioselectivity was much worse. They, therefore, turned to OPMSS to find alternative IDPi catalysts that were effective with more diverse substrates. They pooled silyl nitronates 31a−d and reacted them with 33 to form coupled products 32a−d (Scheme 18). These four substrates were chosen because they had different chemical properties and were separable by chiral HPLC (Figure 16). They found that most of the four substrates had different optimal catalysts, with 30c, 30d, and 30e working best for these substrates. They then expanded the substrate scope by screening new substrates with these three catalysts and could observe good enantioselectivities. 33 Separate and pooled reactions provided essentially the same yields and ee values. 33 In 2021, the Derda group showed the use of a library of fiftythousand peptide aldehydes with phage display and a biotinylated ylide to perform Wittig reactions (Scheme 19). They carried out a pull-down assay in which the aldehydes that reacted in the Wittig reaction would be biotinylated, the products of which could then be isolated by streptavidin beads. They stopped the reaction after approximately 5% conversion. They then analyzed the results using a deep conversion calculation on the basis of deep sequencing, which calculates the enrichment or decrease of a peptide species to determine the relative initial rates. They selected a series of peptide aldehydes to test individually by HPLC to confirm that the results were consistent with the deep conversion method. Using this generated data set, they discovered the cooperative effect of the first and second N-terminal amino acids and the effect of backbone hydrogen bonding on the reaction rate. 34 In 2022, the Jacobsen group reported an enantioselective Pictet−Spengler reaction. They screened 14 aldehyde substrates with 14 different catalysts individually and combinatorially. They then pooled the samples postreaction and analyzed the pooled samples by SFC-MS. Although no catalyst was generally successful for all substrates, they identified catalyst 34 ( Figure  17) as the most generally successful using a constructed generality metric (g). They then screened solvents and found that 2-methyl-THF yielded the highest ee values. 35 Although, by definition, their approach is not OPMSS, their analytical method is closely related. Notably, List et al. stated in the accompanying commentary article that "the next step should be to refine the method to conduct multiple experiments in a single reaction vessel." 36

■ FUTURE PERSPECTIVES
With the advent of machine learning, the generation of thousands of accurate data points from HTE should accelerate discovery and optimization processes in synthetic organic chemistry. 37−40 However, the required infrastructure is costprohibitive in most laboratories. The OPMSS approach is more affordable and may need less time to implement and execute. OPMSS can also rapidly address functional group tolerance in reaction development. 41 Despite the smaller volume of data obtained through OPMSS, if the data quality is as high as HTE, OPMSS may have potential to be integrated with machine learning.
A caveat for OPMSS is that, as some of the research groups experienced in the above examples, there may be orthogonality issues when combining multiple substrates. Therefore, it is imperative to ensure early in the screening process that the substrates in the system of interest are orthogonal by performing single-substrate reactions and confirming that the results are the same as multisubstrate reactions. Additionally, the substrates and products need to be separable, so it is crucial to ensure that the retention times do not overlap.  Like different ingredients are best for different meals, different reagents or catalysts are likely optimal for different substrates for the same type of reaction. The past examples of OPMSS were reported as part of their efforts to discover the most generally successful reaction conditions. However, because, in some cases, steric factors are dominant, and in others, electronics are more important, it may be more advantageous to use OPMSS to find the optimal conditions for particular substrates more rapidly than the traditional approach.