High-Throughput Platform for Novel Reaction Discovery

Abstract

Detection of new chemical bond formation in high-throughput synthesis is limited by the efficiency and scalability of reaction product detection, as conventional methods for product isolation from reaction mixtures are time consuming and labor intensive. Here, we report a miniaturizable purification method that enables rapid, high-throughput isolation of quaternary ammonium-tagged products from reaction mixtures with excellent purity using inexpensive equipment that can be easily set up in a typical organic chemistry laboratory. This novel purification technique enabled us to establish a high-throughput reaction discovery platform. We validated this platform in a screen of 1536 reactions, which identified one previously unreported transformation.

Graphical Abstract

We design a simple technique that can rapidly isolate reaction products with excellent purity by solid phase extraction. This unique technique enables us to develop a high throughput reaction discovery platform. The effectiveness of this strategy was also demonstrated by one newly discovered transformation, which shed new light on the laboratory-scale discovery of new reactions.

Introduction

High-throughput reaction screening (HTRS) enables synthetic chemists to quickly explore a wide range of interrelated variables, accelerating the discovery of new catalysts and novel chemical reactions.^[1] To reliably detect reaction products from complex reaction mixtures in a high-throughput manner, several tag-mediated approaches have been developed, e.g., using substrates functionalized with fluorescent/colorimetric tags,^[2] DNA tags,^[3] mass-spectrometry (MS) tags,^[4] and hapten tags in immunoassays.^[5] However, such methods have generally required specialized and/or expensive instrumentation and have been challenging to scale for high-throughput implementation. Thus, the development of a tag-based strategy for HTRS that can be easily adapted by academic researchers without the need of any costly automation equipment, specialized techniques or expertise is highly desirable.

Quaternary ammonium tag-assisted purification (QATAP)

Based on electrostatic interactions between sorbents and analytes of interest,^[6] ion-exchange resins are widely used in organic synthesis to purify reaction mixtures.^[7] Notably, ion-exchange solid-phase extraction (SPE) methods with cartridges or 96-well microtiter plates have greatly facilitated high-throughput syntheses of new molecules.^[8] Quaternary ammonium (QA) groups have a strong ionic character and carry a permanent charge that is not affected by the pH of the solution. Among the different types of ion-exchange resins, weak cation exchange (WCX) resins offer enhanced selectivity for QA-containing molecules.^[9] A simple SPE using WCX can rapidly separate QA-containing molecules from acidic molecules as well as from neutral or weakly basic molecules (data not shown). These attributes, along with its ease of synthesis and chemically inert nature, make QA an ideal functional group to use in tagging designs for SPE-mediated purification.

In contrast to that of QA moieties, the ionization state of carboxylic acid groups is influenced by both the pH of the solution and the type of solvent. Therefore, the net charge of a QA-tagged molecule that contains a carboxylic group depends on the balance between the carboxylic acid and its carboxylate anion. It was reported that the dissociation of carboxylic acid into carboxylate anion is not favored in aprotic polar solvents.^[10] Therefore, we reasoned that in aprotic polar solvents, the negative charge carried by a carboxylate moiety could be “switched off”. This would confer a positive net charge on a molecule containing a QA tag and a carboxylic acid group, allowing the molecule to bind to a WCX resin. Subsequent exposure to a protic polar solvent, such as MeOH, would “switch on” the negative charge of the carboxylic acid group, making the net charge of the molecule less positive, which would allow the molecule to be eluted out from the WCX resin.

We used two types of QA-tagged molecules in this study. Type I QA molecules contain a QA functional group but no carboxylic acid group, whereas Type II QA molecules contain both a QA functional group and a carboxylic acid group. We envisioned that we could use the inducible switch between carboxylic acids and their carboxylate anions to separate weakly basic (non-QA) molecules, type I QA molecules, and type II QA molecules in SPE cartridges containing WCX resin. To discriminate the WCX binding behaviors of non-QA molecules, type I QA molecules, and type II QA molecules, we examined various elution solvents (see Supplemental Information). We found that aprotic polar solvents such as DMF containing 5% concentrated NH₃ aqueous solution could elute non-QA molecules, but not QA-containing molecules, from the WCX resin, whereas protic polar solvents such as MeOH or EG could elute type II QA molecules, but not type I QA molecule, from the same WCX resin.

Basis on these findings, we developed a quaternary ammonium tag-assisted purification (QATAP) method that uses WCX cartridges to separate type II QA molecules from not only non-QA molecules, but also type I QA molecules. Unlike traditional phase-tag approaches, where separation depends on the solubilities of substrates,^[11] the QATAP method relies on the net charge of the molecules for separation, offering a facile separation strategy for the preparation of molecules containing carboxylic acids.

Results and Discussion

Use of QATAP to detect new chemical bond formation

We tested the feasibility of using QATAP to screen for reactions in which a new chemical bond is formed between QA-tagged substrates and carboxylic acid-tagged substrates. First, we showed that QATAP could be used to isolate the Buchwald–Hartwig amination product of the reaction between a QA-tagged aryl iodide and 3-amino-4-methoxybenzoic acid with 98% purity (Figure 1a). To further test the ability of QATAP to detect new bond formation by selective separation of type II QA products from reaction mixtures, we performed a series of reactions between QA-tagged substrates and carboxylic acid-tagged substrates using a variety of reagents (bases and metals) and solvents (polar and non-polar). As shown in Figure 1b-g, none of the reagents or solvents had an impact on the purification result; in all cases, the reaction products, all of which were type II QA molecules, were isolated with over 90% purity. Notably, when a buffer (Figure 1c) was used as the reaction solvent, QATAP isolated the desired product with 91% purity, suggesting that QATAP has the potential to detect biocompatible transformations. When the carboxylic acid group was not tolerated in a reaction condition, a methyl ester was used instead, as shown in Figure 1f. After the reaction, the methyl ester was hydrolyzed in situ using 10 equiv. of LiOH, and the reaction mixture was subsequently subjected to the same QATAP procedure used for the other reactions. These examples demonstrated that the QA tags could tolerate a broad range of reaction conditions, and that QATAP could be used to separate the products from a variety of reaction mixtures.

Figure 1.

(a) An example of using QATAP to isolate desired products from reaction mixtures. * Prepared on 1.5 mmol-scale and obtained in 94% isolated yield after purification by SPE using a WCX cartridge. (b-g) QATAP was used to isolate various reaction products under the following conditions: i) Pd(PPh₃)₄ (0.05 equiv.), K₃PO₄ (2.2 equiv.), THF:H₂O = 4:1, 80 °C, MW, 15 min; ii) Cu(OAc)₂ (0.2 equiv.), sodium ascorbate (0.4 equiv.), PBS (1x, pH7.4):DMF:H₂O = 1:1:1, RT, 2 h; iii) Pd₂(dba)₃ (0.05 equiv.), LiCl (3 equiv.), Ac₂O (2 equiv.), DIPEA (2 equiv.), DMF, 80 °C, 18 h; iv) Togni reagent (2 equiv.), 3-ethyl-5-(2-hydroxyethyl)-4-methylthiazol-3-ium (0.1 equiv.), Cs₂CO₃ (1.7 equiv.), MeCN, 60 °C, 12 h; v) Pd(OAc)₂ (0.05 equiv.),Ru(bpy)₃Cl₂·6H₂O (0.01 equiv.), TEA (2 equiv.), DMF, blue LEDs, RT, 24 h; vi) LiOH (10 equiv.), H₂O, 60 °C, 30 min; vii) CuI (1 equiv.), ethylene glycol (4 equiv.), K₃PO₄ (3 equiv.), 2-propanol, 90 °C, 24 h; viii) RT, 16 h.

Implementation of QATAP in a high-throughput platform

Synthetic chemists use high-density microtiter plates to screen hundreds or even thousands of potential reactions in parallel.^{[1b, 12]} This type of HTRS can be set up on a bench top using 384-well plates and manual multichannel pipettes. To enable QATAP to be used to isolate reaction products in HTRS, we designed a simple solid dispensing system to load ion-exchange resins into commercially available 384-well filtration plates (see Supplemental Information). An overview of this HTRS approach using QATAP is given in Figure 2. First, QA-tagged substrates and carboxylic acid-tagged substrates are combined with reagents on 384-well reaction plates. After reaction, the products of successful reactions are isolated by QATAP using 384-well, resin-filled filtration plates. The isolated type II QA-tagged products, defined as screen hits, are detected by a UV-visible absorbance microplate reader. Initial hits are then confirmed by conventional LC/MS. Finally, successful reactions are verified using both tagged and non-tagged substrates.

Figure 2.

Implementation of QATAP in HTRS. Reagent solutions and reaction mixtures are depicted in blue, and the filtrate containing the hits are depicted in green. Experiments are set up on a 384-well reaction plate. After initial incubation, the reaction mixtures are transferred to a pre-conditioned 384-well filtration plate containing WCX resin. Non-QA substances are eluted first from the filtration plate with elution solvent #1. Then, elution solvent #2 is used to elute type II QAs onto a detection plate. The initial hits on the detection plate are determined by UV-Vis spectrometry and subsequently confirmed by LC-MS analysis. Successful reactions are then verified individually using both tagged and non-tagged substrates.

To determine the consistency and reproducibility of QATAP in 384-well plates, we prepared three 384-well filtration plates with WCX resin and used each well to purify the same reaction mixture shown in Figure 1a. All liquid transfers were handled by hand pipetting. Raw plate reads were rescaled to percent activities as follows: % Activity = (Vreaction - Vneg) / (Vpos - Vneg) × 100%, where Vreaction indicates the absorbance value for a given reaction well, Vneg indicates the median absorbance value of the negative control (DMSO) wells, and Vpos is a high signal that serves as the positive control, which was set to 1.5 for the reproducibility study and 1 for the final HTRS study. In addition, the coefficient of variation (CV) among the Vreaction values was calculated for each plate and across plates at each wavelength.

The CV values varied between plates and across different wavelengths (Figure 3). Intra-plate variation was generally smaller than inter-plate variation. Two of the three plates displayed intra-plate CV values < 10% across all seven wavelengths, whereas one plate had larger CV values ranging from 6% to 14%. The inter-plate signal variation ranged from 7% to 19%. The signal variation was lowest at 300 nm, with all CV values < 10%. In comparison with hand-pipetting, the use of an automated Tecan freedom EVO pipetting system provided more consistent results, with CV values < 10% for the seven wavelengths tested (data not shown). These results indicated that the results of HTRS using QATAP were reproducible and consistent across replicate experiments.

Figure 3.

The CV values varied between plates and across different wavelengths. Three replicate 384-well filtration plates were prepared with WCX resin and used to purify the same reaction mixture shown in Figure 1a. The average CV values among the reaction wells on each replicate plate at each wavelength are shown in blue, orange, and grey. The average CV values among the reaction wells across all three plates at each wavelength are shown in yellow.

Application of QATAP-based HTRS for reaction discovery

As a proof of concept for using QATAP-based HTRS for reaction discovery, we applied 384-well QATAP in a matrix screen to identify novel C–N bond formation using 15 QA-tagged substrates (Figure 4a) and 23 commercially available carboxylic acids (Figure 4b) under 4 reaction conditions at 300 nmol reaction scale using DMSO as a solvent. Each 384-well plate contained all combinations of substrates under one of the four reaction conditions. To ensure that reaction products would be detectable by UV absorption at wavelengths of 250 nm or higher, all QA-tagged substrates contained an aromatic group and show an absorbance above 250 nm in the ultraviolet-visible spectrum (see Supplemental Information). Reaction mixtures were purified using QATAP, and the MeOH/EG eluate, which would contain any type II QA molecules produced in the reactions, was detected directly on a 384-well UV plate using a microplate reader. For each 384-well plate, detection at a single wavelength only took 80 s.

Figure 4.

Photoreaction using Eosin-Y as a catalyst. (A) Structures of the QA-tagged substrates. (B) Structures of the carboxylic acid-tagged substrates. (C) Detection results (255 nm UV wavelength) using a microplate spectrophotometer. (D) Detection results using a hand-held UV light on a TLC plate with UV254 Fluorescent Indicator.

The reaction conditions and the screening results are summarized in Figure 5a. Successful reactions were initially identified based on UV (255 nm) absorption values above arbitrary thresholds of 0.3, 0.4, or 0.5. If the reaction products were confirmed by subsequent LC-MS analyses, the hits were considered confirmed. Using 0.3 as a UV absorption value threshold, a total of 28 reactions across all four reaction conditions were identified as initial hits and subjected to follow up confirmation, which yielded 26 confirmed hits. In total, 23 reactions with the Eosin-Y catalyst and the QA-9 substrate (Figure 4c, row I) were identified as hits. Because one of these reactions was in a control well (well I01) that did not contain a carboxylic acid substrate, we postulated that the other reactions with the same QA substrate and catalyst yielded the same product. This was confirmed by LC-MS analyses (see Supplemental Information). Further analysis indicated that the isolated product, 4-((3-(diethyl(methyl)ammonio)propyl)carbamoyl)benzoate, was generated by an Eosin-Y catalyzed photo-induced decomposition of the phenyloxazole group in substrate QA-9. The transformation of an aryl oxazole to an aryl carboxylic acid by dye-sensitized photooxidation has been reported in literature.^[13] In a follow-up study, we observed the generation of carboxylic acid from the QA-9 as well as from a similar substrate lacking the QA tag (Figure 5b). Another reaction between the QA-2 substrate and the A-14 substrate with the Eosin-Y catalyst was identified as a hit and confirmed by LCMS and NMR analyses. The product of that reaction was generated by formation of a new C–N bond between the two substrates. UV absorption analysis of the purified product indicated that the yield of the reaction in the original screening well was around 15% (see Supplemental Information). As this transformation was also not previously reported, we explored the reaction conditions and substrate scope further (see Supplemental Information). As illustrated in Figure 5c, this new type of C-N bond formation can tolerate a wide range of functional groups presenting on either reaction partner.

Figure 5.

Summary and results of the high-throughput QATAP screen. (a) Summary of the reaction conditions and results. (b) In situ generation of COOH by photo-induced decomposition of the phenyloxazole group in substrate QA-9. (c) The formation of a new C–N bond between QA-2 and A-14. (d) The formation of a new C–C bond between QA-2 and A-1.

A third reaction between the QA-2 substrate and the A-1 substrate with the palladium catalyst generated a confirmed hit; however, the product of that reaction was generated by the formation of a C–C bond between the two substrates (Figure 5d). Although, such C–C bond forming reaction catalyzed by palladium is unknown, the oxidative coupling of quinoxaline-2(1H)-ones with pyrroles via electrochemical dehydrogenation has been reported in literature recently.^[14] Among the other two experimental conditions, no hit was identified in the photoreactions catalyzed by iodine while two hits were identified in the photo reactions catalyzed by an iridium-based photocatalyst (Figure 5a, entries 2 and 3). One hit (well I06) was the same photo-induced decomposition of the phenyloxazole group in substrate QA-9. The other hit (well G07) was considered an unconfirmed hit because no meaningful product was identified by LC-MS analysis. The second unconfirmed hit was found in the reaction between the QA-8 substrate and the A-16 substrate with the Eosin-Y catalyst (well H17).

We next converted the raw absorbance scores into % Activity values and tested various hit-calling conditions. Because the absorption spectrum might be different for different reaction products, we screened for hits at 7 different wavelengths on the plate reader. In addition, because the limit of detection might vary among different products, we tested various hit-calling criteria to reduce the false-negative rate and maximize the true-positive rate. Two types of hit-calling criteria were tested: one based on the standard deviation (SD) of the absorbance values from the negative control wells (i.e., 3SD, 4SD, 5SD, 6SD), and the other based on an arbitrary % Activity cutoff (i.e., 20%, 30%, 40%, 50%).

For each wavelength and hit-calling criterion, we first determined the numbers of true-positives (TP; confirmed reactions that met the detection threshold), false-positives (FP; reactions that met the detection threshold but not subsequently confirmed), true-negatives (reactions that did not meet the detection threshold and did not yield any product), and false-negatives (reactions that did not meet the detection threshold but were subsequently shown to have yielded product). We then evaluated the screen performance in terms of Sensitivity [TP / (TP + FN)], Specificity [TN / (TN + FP)], Positive Predictive Value [TP / (TP + FP)], and Balanced Accuracy [(Sensitivity + Specificity) / 2]. The screen performance under the different hit-calling conditions is summarized in Figure S-17 (see Supplemental Information). The performance varied across different wavelengths, but more stringent cutoff values generally resulted in lower sensitivity and higher positive predictive value. The conditions that produced the best overall performance for the screen were 4SD or 30–40% and 260–270 nm.

In this proof-of-concept screen, we selected model substrates containing chromophoric groups to boost detection. We envisioned that simple, hand-held UV light devices could be used to visualize screen hits by spotting the QA-tagged products with a multichannel pipette on TLC plates containing UV254 Fluorescent Indicator (see Supplemental Information). We attempted this simple detection method for the Eosin-Y–mediated reactions and compared the visual outcome with values generated using the microplate spectrophotometer (Figure 4c). As shown in Figure 4d, all the confirmed hits were identified by this detection method. This suggests that QATAP-based HTRS can be performed in laboratories that do not have access to a microplate spectrophotometer.

Conclusion

We developed a low-cost, high-throughput method to isolate QA-tagged products from chemical reaction mixtures based on the unique WCX SPE separation selectivity among type I QA molecules, type II QA molecules, and non-QA tagged molecules. The QATAP protocol is simple and can be used to explore new bond formation under a broad range of reaction conditions. The reaction purification, screen detection, and data analysis of 1536 reactions in 384-well plate format described here were completed in less than two hours. The effectiveness of the QATAP strategy was demonstrated by the discovery of two new transformations. We suggest that the simplicity and effectiveness the QATAP approach can open new avenues for reaction discovery research in both academic and industrial settings.

Experimental Section

QA-tagged substrates were prepared in house. Stock solutions (0.1 M) were prepared in appropriate solvents as indicated. Master reaction 384-well microtiter plates were prepared by hand pipetting. Once the master reaction plate was prepared, the mixture was diluted with DMSO and then transferred to a DMF pre-treated 384-well filtration plate. The elution solvents were then added to the filtration plate, and the eluates were collected on a collection plate. The filtration/collection plates were centrifuged up to 1000 RPM with first elution solvent (DMF/DMSO/NH4OH = 95/95/10), followed by MeCN wash, and then with the second elution solvent (MeOH/EG). The desired type II QA compounds were collected during the MeOH/EG wash. The eluted samples were then analyzed using a microplate spectrophotometer.